Theimpactofhealthinsurance Fulltext

The Impact of
Health Insurance in
Low- and Middle-Income
Countries
Maria-Luisa Escobar
Charles C. Grifn
R. Paul Shaw
EDITORS
O
ver the past twenty years, many low- and middle-income countries have
experimented with health insurance options. While their plans have
varied widely in scale and ambition, their goals are the same: to make
health services more affordable through the use of public subsidies while also
moving care providers partially or fully into competitive markets.
Colombia embarked in 1993 on a fteen-year effort to cover its entire population
with insurance, in combination with greater freedom to choose among providers.
A decade later Mexico followed suit with a program tailored to its federal system.
Several African nations have introduced new programs in the past decade, and
many are testing options for reform. For the past twenty years, Eastern Europe
has been shifting from government-run care to insurance-based competitive
systems, and both China and India have experimental programs to expand
coverage. These nations are betting that insurance-based health care nancing
can increase the accessibility of services, increase providers productivity, and
change the populations health care use patterns, mirroring the development of
health systems in most OECD countries.
Until now, however, we have known little about the actual effects of these dramatic
policy changes. Understanding the impact of health insurancebased care is
key to the public policy debate of whether to extend insurance to low-income
populationsand if so, how to do itor to serve them through other means.
Using recent household data, this book presents evidence of the impact of
insurance programs in China, Colombia, Costa Rica, Ghana, Indonesia, Namibia,
and Peru. The contributors also discuss potential design improvements that
could increase impact. They provide innovative insights on improving the
evaluation of health insurance reforms and on building a robust knowledge
base to guide policy as other countries tackle the health insurance challenge.
Maria-Luisa Escobar is lead health economist and health systems program
leader at the World Bank Institute. Charles C. Grifn is senior adviser in the
European and Central Asia regional ofce of the World Bank. R. Paul Shaw,
a former World Bank lead economist, advises the Bill and Melinda Gates
Foundation on health economics.
BROOKINGS INSTITUTION PRESS
Washington, D.C.
www.brookings.edu
Cover by Rich Pottern Design
E
s
c
o
b
a
r
/

G
r
i
f
n
/
S
h
a
w
T
H
E

I
M
P
A
C
T

O
F

H
E
A
L
T
H

I
N
S
U
R
A
N
C
E

I
N

L
O
W
-

A
N
D

M
I
D
D
L
E
-
I
N
C
O
M
E

C
O
U
N
T
R
I
E
S
EDITORS
WASHINGTON, DC
Copyright 2010
Te Brookings Institution
1775 Massachusetts Avenue NW
Washington, DC 20036
www.brookings.edu
All rights reserved. No part of this publication may be reproduced or transmitted in any form or by
any means without permission in writing from the Brookings Institution Press.
Library of Congress Cataloging-in-Publication data
Impact of health insurance in low- and middle-income countries / edited by Maria-Luisa Escobar,
Charles C. Grin, and R. Paul Shaw.
p. ; cm.
Includes bibliographical references and index.
ISBN 978-0-8157-0546-8 (pbk. : alk. paper)
1. Health insuranceDeveloping countries. I. Escobar, Maria-Luisa. II. Grin, Charles C.,
1951. III. Shaw, R. Paul. IV. Brookings Institution.
[DNLM: 1. Insurance, Health. 2. Developing Countries. 3. Health Policyeconomics. W 225.1]
HG9399.D442I47 2010
368.3820091724dc22 2010038233
Editing and typesetting by Communications Development Incorporated, Washington, DC.
v
Preface xi
Acknowledgments xiii
Chapter 1 Why and How Are WeStudying Health
Insurance in the Developing World? 1
Objectives of this study and how it was conducted 4
Country case selection 6
. . . and to the book 11
References 12
Chapter 2 A Review of the Evidence 13
Impact of health insurance on access and use 17
Impact of health insurance on nancial protection 19
Impact of health insurance on health status 22
Heterogeneity of health insurance schemes 25
Moving forward 25
Notes 28
References 29
Studies reviewed but not cited 31
Chapter 3 Low-Cost Health Insurance Schemes to
Protect the Poor in Namibia 33
Health and the health care sector in Namibia 34
Coping with health shocks when uninsured 39
Analysis 41
Scope for targeting 48
Contents
vi Contents
Conclusions 51
Notes 56
References 56
Chapter 4 Ghanas National Health Insurance Scheme 58
Ghanas health system 59
Health care nancing 59
Early successes and challenges 62
Evaluation of the National Health Insurance Scheme 64
Determinants of enrollment in the National Health Insurance Scheme 65
Eects of National Health Insurance Scheme implementation on health care use and
spending 69
Dierences in health care use and payment comparing National Health Insurance
Schemeinsured and uninsured in 2007 76
Policy implications 78
Challenges for the future 79
Appendix: Details and methods of the evaluation 81
Notes 86
References 87
Chapter 5 Impact of Health Insurance on Access, Use,
and Health Status in Costa Rica 89
Main characteristics of the Costa Rican health system 90
Who are the uninsured? 93
What are the determinants of insurance status? 99
Impacts of health insurance 99
Conclusions 102
Appendix: Details of the Estimations 103
Notes 105
Chapter 6 Health Insurance and Access to Health Services, Health
Services Use, and Health Status in Peru 106
Te Peruvian health sector 106
Health insurance population coverage in the household surveys 109
Descriptive information 109
Te eects of Seguro Escolar Gratuito/Seguro Materno Infantil and Seguro Integral
de Salud health insurance coverage on health care results using Demographic and
Health Survey data 112
Contents vii
Impact of Seguro Integral de Salud health insurance coverage on health care results
using ENAHO data 115
Summary and conclusions 117
Appendix: Model Specications 119
Notes 121
References 121
Chapter 7 The Impact of Health Insurance on Use,
Spending, and Health in Indonesia 122
Te Indonesian health system 123
Descriptive statistics: the insured and the uninsured 124
Impacts of insurance on adults 125
Impacts of insurance on children 130
Conclusion 134
Notes 136
References 136
Chapter 8 The Impact of Rural Mutual Health Care on
Health Care Use, Financial Risk Protection,
and Health Status in Rural China 137
Rural health care in China 138
Rural mutual health care 139
Study design and data 143
Te enrolled and the nonenrolled 144
Impacts of the Rural Mutual Health Care Scheme 148
Conclusions 150
References 153
Chapter 9 Colombias Big Bang Health Insurance Reform 155
Main features of the Colombian health sector reform 157
Conceptual framework 159
Health insurance coverage 161
Access and use of services 162
Financial protection 167
Health status 171
Conclusion 173
Notes 175
References 176
viii Contents
Chapter 10 Main Findings, Research Issues, and
Policy Implications 178
Six general ndings 180
Research issues 185
Policy implications 189
Conclusion 197
References 198
Editors and Authors 199
Index 205
Boxes
1.1 Impact of health insurance on health-related outcomes in rich countries 5
2.1 Te key analytical question 14
2.2 Robustness of the evidence base 15
3.1 Data collection 35
5.1 Data and methodology 95
7.1 Methodology 124
9.1 Data and methodology 154
Figures
5.1 Age structure of the insured and uninsured, 2006 95
9.1 Health insurance in Colombia, premiums and benets plans, 2010 156
9.2 Colombias health system, nancial ows, and aliation 157
9.3 Te distribution of out-of-pocket spending by households using health services,
thousand pesos, 2003 167
10.1 Factors diminishing the measured impact of health insurance 186
Tables
1.1 Income and health nance indicators for select country groupings, 2007 3
1.2 Indicators for the country cases, 2007 7
3.1 Individuals enrolled in a medical aid fund 36
3.2 Reported prevalence of chronic disease, acute illness, or injuryand incidence of
hospitalization among the insured and uninsured (%) 37
3.3 Use of health services for acute illness or injury and hospitalization 38
3.4 Average per capita annual out-of-pocket health expenditures 39
3.5 Average per capita annual out-of-pocket health expenditures for the uninsured,
by quintile 40
Contents ix
3.6 Health shocks for uninsured and insured households 43
3.7 Te economic consequences of health shocks 45
3.8 HIV infection rates among working-age adults in biomedical Republic of
Namibia Okambilimbili Survey sample 2006 49
A3.1 Economic outcome and coping variables (uninsured households) 53
A3.2 Consequences of health shocks for uninsured and insured households including
income quintiles in regression 55
4.1 National Health Insurance Scheme membership, 2008 63
4.2 Individuals with National Health Insurance Scheme insurance coverage, by
wealth quintile and age category, 2007 endline survey (%) 65
4.3 Predictors of individual enrollment in National Health Insurance Scheme 2007
sample only (multivariate probit model results) 66
4.4 Receipt of exemptions from National Health Insurance Scheme premiums, by
age category and quintile (%) 69
4.5 Change over 200407 in health care utilization and expenditures for illness/
injury in the two weeks prior to the survey 70
4.6 Change over 200407 in probability of hospitalization and in expenditures for
hospitalization episode in the 12 months prior to the survey 72
4.7 Change over 200420 in utilization of and expenditures for prenatal and delivery
care, among women who had delivery in the 12 months prior to the survey 74
A4.1 Health services provision in study districts, 2007 81
A4.2 Coverage of healthcare services by health insurance in study sites 82
A4.3 Sample sizes from household surveys in Nkoranza and Onso 83
A4.4 Sample characteristics (%) 84
5.1 Payroll fees by insurance scheme, 2006 93
5.2 Sociodemographic statistics for the insured and uninsured, 2006 94
5.3 Health status and use of insured and uninsured people, 2006 97
5.4 Comparison of insured and uninsured people who have been discharged from a
hospital, 2006 98
5.5 Probit analysis of the determinants of participationdependent variable:
insurance status, 2006 100
6.1 Health insurance population coverage by region, 2006 (%) 109
6.2 Use of health services, by type of health insurance, 2005 and 2006 (%) 110
6.3 Seguro Escolar Gratuito/Seguro Materno Infantil and Seguro Integral de Salud and
the probability of being fully immunized (children ages 1859 months, %) 113
6.4 SEG/SMI and SIS and the mean percentage of growth monitoring visits attended
(children under age 5) 115
x Contents
6.5 Impact of Seguro Integral de Salud on probability of seeking curative health care
for symptoms, illnesses, or relapses in last four weeks (all ages) 116
6.6 Summary eects of SIS health insurance (%) 117
A6.1 Dependent variables and control variables 120
7.1 Eects of health insurance on use, by gender and location 127
7.2 Eects of health insurance on use, by expenditure distribution 128
7.3 Eects of health status, by gender and location 129
7.4 Eects of health status, by expenditure distribution 131
7.5 Eects of health insurance on youth and health status for children, by gender and
location 132
7.6 Eects of health insurance on use and health status for children, by expenditure
distribution 133
8.1 Contingency table on the distribution of households total annual health
expenditure in Zhangjai Town, Zhenan County, 1999 141
8.2 Rural Mutual Health Care and New Cooperative Medical Scheme benet
packages 143
8.3 Outcome variables at baseline 145
8.4 Matching balancing properties between the Rural Mutual Health Care and
controls 147
8.5 Impacts of the Rural Mutual Health Care Scheme 149
9.1 Health insurance coverage, 19932003 (%) 161
9.2 Propensity score matching results-estimated treatment eect for the subsidized
regime on access and use, national level, radius (bandwidth 0.0001), 2005 163
9.3 Propensity score matching results-estimated treatment eect for the subsidized
regime for access, use, and health status, by rural-urban status, kernel
Epanechnikov, 2005 (bandwidth 0.001) 165
9.4 Instrumental variable results-estimated treatment eect for access and use in the
contributory regime, 2003 166
9.5 Incidence of catastrophic and impoverishing expenditure by income quintile
and insurance status (simple means), population that has used health services,
2003 (%) 168
9.6 Propensity score matching estimated eect of catastrophic and impoverishing
health spending in the subsidized regime 170
9.7 Estimates of the impact of the contributory regime on the incidence of
catastrophic expenditure by type of employment, 2003 172
9.8 Propensity score matching results-estimated treatment eect on subsidized
regime participants for health status, national level, radius (bandwidth 0.0001),
2005 173
10.1 Observed impacts of health insurance on selected indicators 182
xi
While the underinvestment in health technology for poorer countries
has become glaringly obvious and is starting to be rectied, health care
nancing has been an area of neglect. Yet in richer countries, public
policy in health focuses almost exclusively on nancing and incen-
tive issuesand much progress has been made by those countries in
improving access and care. Low- and middle-income countries continue
to limp along with poorly performing public health care delivery sys-
tems, which almost all rich countries have abandoned for mixed systems
nanced through public purchasers or insurers nanced predominately
from tax revenues.
What could be done at the global level to support countries
interested in undertaking fundamental reforms in health nance?
One obvious candidate is health insurance, which for most low- and
middle- income countries would be a paradigm-shifting change in the
technology of health nancing. Te rst question asked, however, is
whether such a change could have demonstrable impacts on the take-up
and impact of health services. We know little about this issue because
economists principal empirical interest in insurance has been its impact
on nancial risk protection, not on health benets.
Tis question of insurance as a tool of health policy is the chal-
lenge addressed by this book. It is a small rst step to explore whether
changing the health nancing method fully or partly into an insurance-
based approachthat is, moving away from the supply side or direct
Preface
xii Preface
service delivery model that dominates low- and middle-income country health
nancing can have benecial eects on health-seeking behavior and, by implica-
tion, health status.
We did not know what to expect when we set out to nd countries, datasets,
and analysts to shed light on this question. Our conclusion is that shifting partly or
fully to insurance-like nancing methods (in which payments are made to provid-
ers contingent on providing services to patients) has positive eects on the health-
seeking behavior of consumers, at least in the countries covered in this book. Even
in a country like Costa Rica, where 80% of the population is covered by insurance
and everyone has access to hospital care when they need it (at no cost if they can-
not aord it), the uninsured behave dierently from the insured. Te association of
insurance with better health-seeking behaviorand in some cases a clear impact of
insurance on better management of a familys healthis strong enough to encour-
age more experimentation and policy innovations.
In a few cases, particularly China and Peru, it is apparent that insurance also
aects provider behavior, although that is not the focus of the book. Moreover,
although we also did not expect it, there are lessons in every chapter about the nuts
and bolts of design and implementation that illuminate some of the tasks reformers
need to do well for a reform to work. Because the hoped-for benets of insurance
depend on how it is designed and who benets from it, no eort should be spared
to get the details right before policy reforms are put in place.
Te authors of the chapters in this book retrotted evaluations as best they
could. It is surprising to us that evaluation had not been built into all of these
insurance reforms from the start. How else can anyone know what is working and
what needs to be changed? How else can the progress of the reform against its
goals be measured? We do not end this eort with a simple call for more research
but with a call for more innovations in health nancing policy like those covered
in this book. But pairing them with a research agenda by building in evaluation
at the start is the only way to improve reforms as learning takes place and impacts
become clear. We hope this book encourages health nance policy innovation,
more international support for it, more learning, and feedback of that learning into
constantly improving policies for better health.
Maria-Luisa Escobar
Charles C. Grin
R. Paul Shaw
xiii
Each chapter was written by a team whom we challenged to do their
best to understand the impact of health insurance reforms using avail-
able household data and their knowledge of health policy in the country
they were working on. We have been impressed by their resourceful-
ness in securing data and nding creative ways to address economet-
ric problems and by their willingness to persist in reanalyzing the data
and rewriting their chapters as the editors and peer reviewers gave them
feedback. Most of all, therefore, we acknowledge the eorts of the chap-
ter authors to produce the ne work that is collected in this volume.
Whatever success we have with this volume is due to the contributors
professionalism and commitment.
We would also like to recognize the contributions of the peer
reviewers, who worked with us and the chapter authors to provide feed-
back not just at the end of the process but during each stage of the
analytical work. We paired two peer reviewers with each set of authors,
with some reviewers working with more than one set of authors. As with
the chapter authors, we appreciate their persistence in reading the same
manuscript several times; in delving into the data, econometric, and
policy issues with the authors; and in participating in long conference
calls to help move the project forward. Tey were an essential resource
and gave their time freely, with little more compensation than the hope
that they were contributing to our understanding of the potential of
insurance as a health policy option for low- and middle-income coun-
tries. Te peer reviewers were Anil Deolalikar, Philip Musgrove, Menno
Pradhan, Jacques van der Gaag, Bill Savedo, and Adam Wagsta.
Acknowledgments
xiv Acknowledgments
We have a great debt to Dan Kress of the Bill & Melinda Gates Foundation
for his willingness to take risks that we could deliver a collection of empirical stud-
ies exploring the impacts of health insurance for countries where it is often not
considered to be a feasible health policy option. We also thank Dan for challeng-
ing us to go beyond the typical economists focus of nancial protection to begin
addressing the question of whether health insurance could be considered a viable
intervention to improve health status. As a result of his prodding, each chapter
attempts to discover proximate measures of outcomes, or outcomes themselves,
that are aected by health insurance. While for many readers this eort may raise
more questions than it answers, that is a good result, as we hope it will prompt
other work. After all, health policy is dominated by physicians and public health
specialists, not economists. If they are convinced that insurance is a side issue that
has no direct eect on health, then it will remain a side issue. We thank Dan for
insisting that we address this question of health impact. We have some tantalizing
results, but much more can and should be done in this area. And we thank Marga-
ret Cornelius, also of the Bill & Melinda Gates Foundation, both for pushing this
issue along with Dan and for working with us to get everything done.
We would also like to thank the Global Economy and Development Program
at the Brookings Institution, including Lael Brainard and her team, for supporting
us while we were in residence at Brookings to undertake this work. On the business
side, Amanda Armah has been tireless in managing the many contracts required
for the work, managing the grant, and keeping us in line. Sun Kordel has back-
stopped Amanda and always helped us nd a way past hurdles. Kyle Peppin worked
with us throughout the long period of gestation of the book and helped enormously
with administrative arrangements, communicating with the authors, and research.
Te book would not have been possible without him.
As the original editors, Maria-Luisa Escobar and Charles Grin would like to
thank Paul Shaw for joining our team midway to oer a third perspective and extra
help to get the volume out after we returned to other duties at the World Bank.
Paul discussed the big picture with us many times, wrote the original versions of
the rst and last chapters, and provided timely feedback to the authors that helped
keep everyone attuned to what we were trying to achieve.
With great respect and thanks, we recognize the contributions of Bruce Ross-
Larson and his team at Communications Development IncorporatedMeta
de Coquereaumont, Rob Elson, Jack Harlow, Christopher Trott, and Elaine
Wilson who whittled more than 700 pages of manuscript down to something con-
cise enough to be readable and consistently on theme despite all the dierent con-
tributors and countries. Tey have made a substantial contribution to this endeavor.
1
More than 2 billion people live in developing countries with health
systems aicted by ineciency, inequitable access, inadequate fund-
ing, and poor quality services. Tese people account for 92% of global
annual deaths from communicable diseases, 68% of deaths from non-
communicable conditions, and 80% of deaths from injuries. Te World
Health Organization (WHO) estimates that more than 150 million
of these people suer nancial catastrophe every year, having to make
unexpected out-of-pocket expenditures for expensive emergency care
(WHO various years).
Within countries, the burden of dysfunctional health systems is dis-
proportionately felt by the poorest households. Teir access and use of
services, such as immunizations and attended deliveries, tend to be half
those of richer households. Tey have limited recourse to purchase qual-
ity services from private providers. Teir enrollment in health insurance
tends to be marginal. And they are unable to shield themselves from
catastrophic health expenditures by drawing on accumulated wealth.
In view of these shortcomings, policymakers in many low- and
middle- income countries are debating the virtues of scaling up health
insurance to improve health outcomes. Major international confer-
ences have been convened in Berlin (2005) on social health insurance in
developing countries and in Paris (2007) on social health protection in
developing countries. Regional conferences have followed, as in Africa
in 2009. Related to these initiatives, the World Health Assembly passed
Why and How Are WeStudying
Health Insurance in the
Developing World?
Maria-Luisa Escobar, Charles C. Grin, and R. Paul Shaw
C
h
a
p
t
e
r

1
2 Chapter 1
a policy resolution whereby the WHO would advocate formally mandated social
health insurance to mobilize more resources for health in low-income countries,
pool risk, provide more equitable access to health care for the poor, and deliver bet-
ter quality care (WHO 2005a).
All rich countries have adjusted their health nance systems to reduce out-
of-pocket expenditures for health, which plunge as per capita income rises across
countries (table 1.1). In terms of purchasing power parity (PPP), our preferred mea-
sure, per capita gross national income (GNI) is 29 times higher in the richest group
than in low-income countries, but health spending per capita is 63 times higher.
Te share of gross domestic product (GDP) devoted to health more than doubles,
the governments share in the total rises, and the burden on individuals plummets
as out-of-pocket spending falls as a proportion of the total. Te bottom of table
1.1 shows how much this result reects the situation in South Asia because of its
large share of the total low-income population. Te situation is slightly less dire in
Africa, but only a bit.
Rich countries achieve these results through general revenue tax nancing in
support of national health insurance or subsidies for specic groups (such as the
poor or the elderly), payroll taxes to support social health insurance, or, most com-
monly, some combination of both. Rich countries provide prepaid entitlement to
health care benets, reduce vulnerability to the expenses of care at times of illness
or injury (nancial risk protection), and use copayments and deductibles chiey to
manage demand rather than to raise revenue. Tey seek to reduce the discontinu-
ity of care so common when people are navigating the system on their own and
paying out of pocket at each point of contact. For the most part, richer countries
have also separated nancing from the provision of care, depend on a mix of public
and private providers that are reimbursed through the insurance system, and rely
increasingly on primary care providers as gatekeepers to more expensive higher level
services. In a nutshell, poor countries want to mimic these successful and desir-
able behaviors of rich countries sooner rather than later. Mysteriously, donors have
historically nanced the direct delivery of health services in poorer countries with
almost no attention paid to helping them build sustainable nancial and purchasing
institutions that could emulate some of the core successes of richer countries.
Whatever policymakers and donors want to do or think they should do to
emulate successful health nancing reforms, there are knowledge gaps that create
enormous risks of failure for any reformer. Tis book attempts to begin lling some
of them, but much more work remains.
Te widest knowledge gap concerns the impact of health insurance on health
status. Do people with health insurance in low- and middle-income countries, or
Why and How Are We Studying Health Insurance in the Developing World? 3
even rich countries, have better health status indicators than those without? Evi-
dence from rich countries suggests yes (box 1.1). But what about low- and middle-
income countries? An armative on this issue would surely seem essential to
consider health insurance as a health policy intervention rather than simply as a
nancial protection intervention. Te vast array of people involved in health care
because they want to improve healthnutrition advocates, family planning advo-
cates, tuberculosis and AIDS activists, vaccine supporters, Millennium Develop-
ment Goal supporters, health systems improverswould have to see health insur-
ance as an intervention that would be more eective in improving health outcomes
than other directly focused options. Obviously, carrying a health insurance card
by itself does not make one healthier, but if that card increases the use of appropri-
ate services, makes a person more likely to access new proven technologies, creates
incentives for providers to deliver the right services, and equalizes use among the
rich and the poor, most analysts would be satised that it can have a powerful
TABLE 1.1
Income and health nance indicators for select country groupings, 2007
Country
group
Gross
national
income
per capita
(US$)
Per capita
health
expenditure
(US$)
Gross
national
income
per capita
(PPP)
Per capita
health
expenditure
(PPP)
Total health
expenditure
in GDP
(%)
Public
share of
total health
expenditures
(%)
Out-of-pocket
share of
total health
expenditures
(%)
Low income 461 27 1,284 69 5 42 48
Lower middle
income 1,752 81 4,234 182 4 42 53
Upper middle
income 6,705 488 11,534 753 6 55 31
High-income
OECD 39,540 4,618 37,328 4,327 11 61 14
East Asia
and Pacifc 2,190 96 4,946 208 4 46 48
Europe and
Central Asia 6,013 396 11,123 647 6 66 29
Latin
America 5,888 475 9,802 715 7 49 35
Middle East
and North
Africa 2,795 151 7,350 364 6 51 46
South Asia 879 26 2,535 98 4 27 66
Sub-Saharan
Africa 966 69 1,858 124 6 41 35
Source: World Bank 2010.
4 Chapter 1
impact on improving health. Tey then can devote themselves to making sure the
services work.
A second knowledge gap concerns the impact of health insurance on out-
of-pocket expenditures for health. Do people with health insurance have lower
out-of-pocket spending than those who do not, especially when they are struck
by health emergencies? Do the uninsured poor pay a higher proportion of their
income for health care than the rich? When out-of-pocket spending is the princi-
pal means of securing health care, emergencies result in people borrowing, selling
assets, not getting needed care, and engaging in other coping mechanisms. A high
proportion of out-of-pocket spending also leads to poorer households spending
more of their income on health care than richer households do, just as they spend
a higher proportion on other necessities, like food and shelter. Health insurance
should address this problem, yet the empirical evidence is slight in our focus coun-
tries. Te more one explores this issue, the more it becomes apparent that success
depends on the design of the program and who is covered; health insurance is
not a homogeneous product. A yes on reducing out-of-pocket spending would be
essential to argue that health insurance can help prevent people from sliding into
health-related poverty.
We can stop with those two questions. Both must be positive to even consider
health insurance as a sensible health policy tool in low- and middle-income coun-
tries. Tere are many other practical questions of implementation, but they reside
in the realm of suciency, not necessity, for considering insurance as a health pol-
icy option rather than just a nancial protection option.
Objectives of this study and how it was conducted
Tis study aims to contribute to current policy debates on scaling up health insur-
ance in low- and middle-income countries by shedding light on these two issues:
its impact on measures of health status and reducing out-of-pocket spending. Four
objectives guide the research and analysis.
Objective 1. Rigorously review and synthesize published and unpublished studies
to determine what we know about the impact of health insurance on access and use
of health services, the impacts on nancial risk protection, and the methodological
and data issues in ascertaining causality.
Objective 2. Undertake new country case studies to assess the impact of health
insurance on access and use of health services as well as nancial risk protection
using the latest data sources and statistical methodologies.
Objective 3. Cast more light on the inclusion of the poorest quintile of the popula-
tion in health insurance in low- and middle-income countries, as well as the ben-
ets they experience compared with the uninsured poor.
Objective 4. Identify the challenges, risks, and opportunities of undertaking ret-
rospective evaluation of health insurance in developing countries using household
data.
Shedding light on these objectives requires more than applying good econometrics.
Researchers require a fundamental understanding of how health systems work to
know what questions to ask and what models to use to nd answers. Tis requires
BOX 1.1
Impact of health insurance on health-related outcomes in rich countries
A committee sponsored by the Institute
of Medicine of the National Academies in
Washington, DC, reviewed 130 research
studies that consider the impact of health
insurance on health-related outcomes for
adults ages 1864 (IOM 2004, updated
in IOM 2009). Findings suggest that
uninsured adults are less than half as likely
as those insured to receive needed care for
a serious medical condition. Uninsured
women and their newborns receive less pre-
natal care and are more likely to have poor
outcomes during pregnancy and delivery,
including maternal complications, infant
death, and low birthweight. In addition,
the uninsured more often:
Lack regular access to medications to
manage conditions, such as hyperten-
sion and HIV.
Do not receive care recommended for
chronic diseases, such as timely eye and
food exams to prevent blindness and
amputations in people with diabetes.
Go without cancer screening tests,
which delays diagnosis and leads to
premature death.
Receive fewer diagnostic and treat-
ment services after a traumatic injury
or a heart attack, causing an in-
creased risk of death even when in the
hospital.
Findings specic to children reveal the
uninsured are:
Less likely to get routine well-child
care.
More likely to receive no care or de-
layed care, thus placing them at greater
risk of being hospitalized for conditions
such as asthma that could have been
treated on an outpatient basis.
Using medical and dental services less
frequently than insured children.
However, as Gruber (2009) observes,
most of these studies simply document a
correlation between no health insurance
and poor health. Almost none attempted to
control for the endogeneity of health insur-
ance coverage with respect to health status.
He cites only a handful of U.S. studies that
have adequately controlled for endogeneity,
but they too show strong impacts of health
insurance coverage on health.
6 Chapter 1
familiarity with design elements that might aect the measured impacts of health
insurance on health outcomes (such as enrollment criteria, benets entitlements,
and copayments). Beyond this, however, the study does not assess whether the
organizational design of health insurance in dierent countries is the most ecient
or most cost-eective arrangement in satisfying clients, providing quality care, pay-
ing providers, or being nancially sustainable over the long term. Tese issues,
while important, are complex and demanding enough to require another volume.
In short, this study focuses on impacts of health insurance schemes as presently
designed and implemented, not what such schemes might accomplish if imple-
mented dierently.
Reading this book may raise more questions than it answers, which is good, as
we want to present the evidence available today on the topic. We began by identify-
ing low- and middle-income countries that had experienced insurance reforms of
interest. We narrowed the list according to whether data existed that could be used
to measure what happened at the household level in response to these insurance
reforms. We sought researchers who knew the country well and were qualied to
do the work. We paired them with advisers and peer reviewers who would commit
to read and advise as drafts of the chapters took shape. We tried to keep all of the
individual projects advancing along the same timeline and hoped that the ensuing
chaos would result in a good collection of work. We did not have the luxury of
perfection in any part of this process.
Tere are many technical impediments to showing an impact of health insur-
ance on anything. Tese are discussed in the literature review in chapter 2. For
some of the chapters readers may conclude that the evidence provided does not get
far past associations; in other chapters the evidence may look conclusive that causa-
tion has been established. Te consistent theme that there is an impactdespite
the variety of situations, data, methods, and policies examinedbecomes inescap-
able after reading all the chapters.
Country case selection
Four considerations guided our selection of country case studies. First, we sought
countries with suciently diverse backgrounds to shed light on the extent nd-
ings could be generalized across dierent contexts. Second, to gain insight into
impacts of scaling up health insurance for relatively disadvantaged or poor
households, we sought countries with a pro-poor orientation in the design and
implementation of health insurance. Tird, we sought countries with suciently
well developed surveys or data systems that would facilitate rigorous statistical
analysis of impacts of health insurance on measures of health status and nancial
risk. Fourth, we sought researchers with a solid knowledge of health insurance
who were capable of performing complex statistical modeling to tease out causal
impacts. Table 1.2 provides summary data on the seven countries in this volume.
We have two giant countries in the mix, China and Indonesia; however, except
for Namibia and Costa Rica, all are sizeable. Te data used in the chapters are
nationally representative except in Namibia, Ghana, and China. Tere are some
important dierences across countries in the state of health, income, and health
spending, but perhaps the widest range lies in out-of-pocket spending, ranging
from 3%8% of total health spending in Namibia and Colombia to over 50% in
China. It is low in Namibia because of good penetration of private insurance; it
is low in Colombia because of high government spending, primarily through its
insurance system.
Namibia. Te Namibian health insurance industry is better developed than that
of most Sub-Saharan African countries. It is organized primarily into nonprot
TABLE 1.2
Indicators for the country cases, 2007
Indicators Namibia Ghana
Costa
Rica Peru Indonesia China Colombia
Population (millions) 2.1 22.9 4.4 28.5 224.7 1,317.9 44.4
Life expectancy (years) 60 56 79 73 70 73 73
Infant mortality rate
(per 1,000 live births) 32 53 10 23 32 19 17
GNI per capita (US$) 4,110 590 5,530 3,340 1,520 2,410 4,070
GNI per capita (PPP) 6,080 1,330 10,530 7,060 3,280 5,430 8,200
Health expenditures
per capita (US$) 319 54 488 160 42 108 284
Health expenditures
per capita (PPP US$) 467 113 878 327 81 233 516
Health expenditures in
GDP (%) 8 8 8 4 2 4 6
Public share of health
spending (%) 42 52 73 58 55 45 84
Out-of-pocket in total
health expenditures (%) 3 38 23 32 30 51 8
Population enrolled in
health insurance (%) <30 61 88 42 36 8090 90
Source: World Bank 2010. Data on population enrolled in health insurance are based on infor-
mation from the chapters in this book.
8 Chapter 1
medical aid fundsabout one-third mandatory social health insurance funds and
about two-thirds voluntary, private plans. Many of the funds are closed, with mem-
bership limited to employees in a particular rm or industry or to government civil
servants. Tis has resulted in large disparities in enrollment across socioeconomic
categories; only 5% of individuals in the poorest quintile are enrolled, compared
with 70% of individuals in the richest quintile. While some private insurance plans
aim to broaden the insured population through low cost plans, the challenge is
huge because of the countrys high prevalence of HIV/AIDS, estimated at 20%
for people ages 1549, concentrated largely among the poor (Feeley, Preker, and
Ly 2007).
Te case study assesses dierences in the consequences of health shocks between
the insured and uninsuredstemming from death, hospitalization, weight loss,
and HIV/AIDSspecic to households in dierent income quintiles. Te impact
of health insurance has been assessed using multiple regression analysis, using 2006
survey data that include both socioeconomic and biomedical information.
Ghana. In 2003 the government passed the National Health Insurance Act, with
a vision of insuring 40% of the population by 2010 and 60% by 2015. About 60%
of the population was enrolled by 2008, exceeding expectations, with the success
attributable to the generous benet package and prior familiarity with enrolling
households in district level mutual health organizations. Other African countries
are closely watching Ghanas attempt to scale up health insurance, given the far
reaching implications for raising funds, purchasing, and providing care to a largely
poor population.
Te case study applies a pre-post evaluation design in two districts, one clas-
sied as deprived, the other as less deprived. Te impact of the health insurance
reform is assessed using pre-post bivariate comparisons of key indicators, multi-
variate regression analysis, and a tentative application of propensity score matching
analysis (tentative because of the small sample sizes), using data from a baseline
household survey in 2004 and an endline survey in 2007.
Costa Rica. Tis country has become a benchmark of health insurance attaining
wide coverage with no copayments, based on a direct delivery model. Social health
insurance was introduced in 1950, and the Universal Coverage Act passed in 1961.
Since then, health insurance coverage grew from about 18% in 1961 to 45% in
1971, 60% in 1975, and a high of 92% in 1990. In 2009 about 88% were covered,
although the surveys used by the authors in this book put coverage closer to 81%
in 2006.
Te case study looks at the 19% of the population without health insurance
in the 2006 surveys to establish dierences in their health status and other char-
acteristics and to investigate whether their health-seeking behavior and results are
dierent. Te impact of health insurance on health and related behaviors has been
assessed using instrumental variables and data from the 2006 National Health Sur-
vey; expenditure results are based on the 2004 Income and Expenditure Survey;
and a database of hospital discharges from 2006 provides a unique perspective on
how the insured and uninsured use the system dierently when they are sick.
Why study a country where everyone is either insured or, if they are not, have
equal access to hospital care if they need it? One would not expect to see dierences
in nancial protection in such a system for sure, but because we are interested also
in health outcomes, it might be a unique opportunity to see whether not being
covered by the formal insurance program has any impact even with Costa Ricas
equal access provision.
Peru. With about 35% of the population covered by employer-mandated social
security and other forms of health insurance, the government consolidated and
began scaling up two pro-poor schemes initiated in 2001: one targeting children in
public schools, the other targeting maternal and child health. Enrollees in the new
consolidated program doubled from 3.6 million in 2001 to 7.3 million in 2007.
Te case study assesses the impact of this publicly subsidized health insurance
program that explicitly targets the poor. Te impact of health insurance has been
assessed with several models, using data from two household surveys: the Demo-
graphic and Health Surveys (DHS) for 2000 and 2004 (heavy on health informa-
tion but light on economic data) and a nationally representative panel survey from
2004 to 2006 (with substantial economic data but limited health data).
Indonesia. With about 36% of the countrys population covered by social secu-
rity schemes as well as a public health insurance scheme, the government greatly
increased public spending on health from about $1 billion in 2001 to $4 billion by
2007. Much of this additional spending was due to the expansion of the Askeskin
health insurance program, which targets the poor.
Our case study examines changes in health status associated with movements
in and out of health insurance, to shed light on how health insurance might aect
health status and nancial risk protection where only formal sector insurance cov-
erage exists. Te impact of health insurance has been assessed with individual xed
eect models, using panel data from longitudinal surveys in 1991, 1997, and 2000.
Te panel data used in this analysis provide a unique contribution even though the
10 Chapter 1
most recent installment of the survey was not yet available to the researchers, which
would have allowed them to include the Askeskin reform in the analysis.
China. In 2002 the government announced a new national policy for rural health
care, the New Cooperative Medical Scheme (NCMS), which aimed to recapture
successes of Chinas past health policies. In the late 1970s Chinas Cooperative
Medical System, a communal-based approach, covered 90% of Chinas rural popu-
lation. But it collapsed after the government introduced the Household Responsi-
bility System in 1979, and communes disappeared as a result. Te revised NCMS is
a voluntary scheme that gives priority to covering catastrophic health expenditures
and subsidizes premiums. By the end of 2008 it was credited with reaching more
than 90% of the rural population.
Te case study reports on a social experiment of a community-based prepay-
ment schemeRural Mutual Health Careundertaken as an implementation
of the NCMS in several counties. Operating from 2002 to 2007, the experiment
aimed to contribute to knowledge on the impacts of insurance, tailored to condi-
tions in the poorest regions of China. Te impact of health insurance has been
assessed using dierences-in-dierences statistical methods and propensity score
matching, using a pre-post treatment-control study design in two of Chinas rural
provinces. A baseline longitudinal survey was conducted, along with two more
panels following the same individuals during implementation of the experiment.
Colombia. A commitment by the government in 1993 to reorganize its dual health
care system (a Ministry of Health direct delivery system alongside a social security
direct delivery system), to expand coverage of the population by insurance, and to
oer more choice to citizens on both insurer and provider oerings, has increased
coverage from 24% in 1993 to 90% in 2007. Health insurance is nanced through
a contributory regime by employees in the formal sector and a subsidized regime
in the informal sector. A major accomplishment of government eorts to scale up
health insurance is an eightfold growth of enrollment among the poorest quintiles.
Te case study uses the gradual implementation and still incomplete coverage
of the subsidized regime to identify dierences in health outcomes between those
with health insurance and those without. Te impact of health insurance has been
assessed with a variety of semiparametric methodsincluding propensity score
matching, double dierence, and matched double dierenceand instrumental
variable analysis, using data from various Colombian DHS (1995, 2000, and 2005)
as well as Living Standard Measurement Survey data for 2003 and administrative
data.
...and to the book
While there is considerably more interest in insurance as a nancing option for
health care even in poor countries, progress has been greatly hampered by a lack of
knowledge of what the future would look like after such a reform. Te rst ques-
tions that arise from policymakers and reformers are:
What country has done something like this that faces our constraints?
What has been the impact?
How did they implement it?
What would they do dierently in hindsight?
Tis book cannot explain much about how the reforms covered were imple-
mented (the third question); that requires a dierent type of case study. But it does
provide considerable information on the rst two questions. On the fourth, each
of the chapters has suggestions for what the authors think the authorities should
have done in hindsight. Whether the suggestions will be taken up is another step
entirely.
One thing that is essential to keep in mind in reading this book is that in
no case is a perfect laboratory experiment being described. In fact, there is no
chance of one being developed to assess the impact of insurance. Why? Because
you can never take away from people all the other options they face. Te most
important other option in this book is the availability, in all cases, of free or
low cost government-provided care in its own facilities. In Namibia the govern-
ment system is reputed to function well and to be well nanced. Yet even with
this option we see substantial dierences between the insured and uninsured. In
Colombia, in comparisons of the insured poor against the uninsured poor, it is
not that the uninsured poor have no services available because they can use the
public system still in place. In Peru the insurance analyzed not only sits next to
the subsidized public system, the insured are required to use the public system.
So, as with all such analysis in low- and middle-income countries, there is always
the unobserved impact of a free or low cost public system option (however well or
poorly it functions) that confounds the results, more than likely by attenuating
the impacts of insurance. For countries considering a complete switch from the
supply side funding of free public services to demand side funding under insur-
ance, we can say only that the evidence in this book is just a starting point.
Te good news for reformers is that this book demonstrateswe thinkthat
to know something about the impact of insurance, clever use of available data
can obtain reasonably robust results. Moreover, to introduce health insurance
as a health policy reform, it is not necessary to wait for results of randomized or
social experiments. By now, we know that insurance can improve access and use of
12 Chapter 1
services and can protect from the risk of nancial loss. We see this in the literature
review and in all the cases in this book. How much and for whom depend on the
specics of the design of the insurance scheme. Despite the statistical challenges
researchers face, countries can reasonably expect that by introducing a pro-poor
insurance scheme they can obtain improved results for access and use of services
and for nancial protection. Te obvious alternative is to invest in providing free
services directly, but we see in Namibia, Costa Rica, Peru, and Colombia that
insurance or an insurance-like alternative may have a greater impact.
Does use of more health services and improved nancial protection lead to
better health? Te cases in this book demonstrate the diculty in establishing that
link with the available data and measures of health outcomes; even so, there are
many tantalizing clues that should encourage more eort in this area. What is
needed are explicit goals for health outcomes embodied in an insurance system,
disaggregated measures of health outcomes that insurance (and alternatives to it)
can aect, and data suitable for measuring impacts without bias. Tere is much
more to be done on this topic and, as well, on the impact of insurance on provid-
ers. In this book for the most part we focus on the demand, or patient, side of the
equation.
Queries about each chapter should be directed to the corresponding authors,
whose email addresses are listed in the Editors and Authors section after chapter 10.
References
Feeley, F., A. S. Preker, and C. Ly. 2007. On a Path to Social Health Insurance? A Look at
Selected Anglophone African Countries. World Bank, Washington, DC.
Gruber, J. 2009. Covering the Uninsured in the U.S. Journal of Economic Literature 46 (3):
571606.
IOM (Institute of Medicine). 2004. Insuring Americas Health: Principles and Recommendations.
Washington, DC: National Academies Press.
. 2009. Americas Uninsured Crisis: Consequences for Health and Health Care. Washington,
DC: National Academies Press.
WHO (World Health Organization). Various years. World Health Report. Geneva: World Health
Organization.
World Bank. 2010. World Development Indicators 2010. Washington, DC: World Bank.
13
C
h
a
p
t
e
r

2
A Review of the Evidence
Ursula Giedion and Beatriz Yadira Daz
We used a detailed protocol to evaluate the robustness of the available
evidence on the impact of health insurance in low- and middle-income
countrieson access, use, nancial protection, and health status (box
2.1). Of 49 quantitative studies, about half provide reasonably robust
evidence. Tey indicate that health insurance improves access and use,
seems to improve nancial protection in most cases, but has no con-
clusive impact on health status. Te third result may be related to the
diculties of establishing a causal link between health insurance and
currently available information on health status.
Te positive eect of health insurance on medical care use has been
widely demonstrated and generally accepted. Hadley (2003), in his
review of research published in the past 25 years on health insurance in
the developed world, concludes that there is a compelling case for the
positive correlation between having health insurance and using more
medical care. Little evidence exists, however, on the impact of health
insurance in the developing world, and only a few studies have tried to
summarize what is available either in some regions or for specic types
of health insurance.
Whether health insurance is a recommendable strategy to improve
access to health care in low- and middle-income countries is hotly
debated but insuciently documented. For example, a resolution
adopted at the 2005 World Health Assembly invited member states
to ensure that their health nancing systems include a method for
14 Chapter 2
prepaying nancial contributions for health care. But a recent joint nongov-
ernmental organization brieng paper laments the lack of evidence on whether
health insurance can really work in low-income countries and concludes that
health insurance so far has been unable to suciently ll nancing gaps in
health systems and improve access to quality health care for the poor (Oxfam
and others 2008).
What do we really know about the impact of health insurance in low- and
middle-income countries? Tis chapter synthesizes the best available evidence
regarding the impact of health insurance in low- and middle-income countries on
access, use, nancial protection, and health status. It emphasizes the results of the
10 studies that provide the most robust evidence and belong to the top quartile
score after applying our quality assessment tool.
1
We extend this analysis to the
second quartile whenever the evidence is especially scarce (box 2.2).
BOX 2.1
The key analytical question
Te purpose of health insurance is three-
fold: increase access and use by making
health services more aordable, improve
health status through increased access and
use, and mitigate the nancial conse-
quences of ill health by distributing the
costs of health care across all members of
a risk pool. Te key analytical question in
this chapter is: What does the literature
say about the impact of health insurance
on access and use of health care, on health
status, and on nancial protection?
Evaluating the impact of health insur-
ance is, methodologically, a challeng-
ing endeavor. It requires econometric
methods to tackle issues such as poten-
tial selection bias and the bidirectional
relationship between health insurance and
health status. It also requires quality data
on households and providers to measure
outcomes of interest, to correct for dier-
ences among the insured and uninsured,
and to account for supply constraints
and, above all, profound knowledge of the
specic health insurance scheme being
evaluated. Often one or more of these
items are missing, and analysts must cope
as best they can.
Some policy reforms aim to use health
insurance to change supplier and pro-
vider behavior as well as to create a more
elastic form of nancing than govern-
ment tax revenue can provide. However,
this review focuses on a circumscribed
number of performance dimensions and
does not include the literature evaluating
other consequences of health insurance,
such as changes in the organization of
health systems or the overall eciency of
health insurance as compared with other
nancing mechanisms. It is limited to
studies that attempt to establish a causal
relationship between health insurance and
health-related outcome indicators. It thus
excludes studies that present descriptive
statistics only or that resort to qualitative
analysis when evaluating health insurance
in low-income countries.
A Review of the Evidence 15
BOX 2.2
Robustness of the evidence base
Te robustness of the evidence was deter-
mined on the basis of ve general criteria:
quality of the study design (selection of the
treatment and control groups), strength
of the impact evaluation methodology
(mostly related to the way the potential
selection bias problem was dealt with), the
rigor with which each method was applied,
and the quality of the discussion related to
the ndings of each study.
Five key issues emerge from the
analysis of the robustness of the litera-
ture. First is dealing with the nonrandom
variation in health insurance status
(endogeneity). Second is considering the
heterogeneity in impact across dier-
ent population groups and insurance
schemes. Tird is exploring the possibil-
ity of spillover eects of health insurance.
Fourth is undertaking the relevance of
timing when evaluating the impact of
health insurance. Fifth is being clear in
the statement of research goals, methods,
and potential limitations.
Te quality evaluation protocol assigns a
maximum score of 100 points according to
the ve criteria. Te scores obtained by our
evidence base varied between a minimum
of 11 points (least robust study) to a maxi-
mum of 83 points (most robust study).
Two-thirds of the studies reviewed scored
45 points or lower, a result that argues for
continuing to support the production of
quality research on the impact of health
insurance in developing countries (box table
1). Note, however, that a small (but growing)
number of studies provide higher quality
evidence of impact. Te 10 studies in the
highest quartile explicitly address endog-
eneity and clearly describe research goals,
methods, results, and the limitations of their
evaluations. Several take into account the
potential heterogeneity of impact across dif-
ferent groups and insurance schemes.
BOX TABLE 1
Distribution of the literature by score quartile
Score quartile Score Total
Access
and use
Financial
protection Health status
Lowest 1144 21 16 11 3
Lower middle 4563 10 9 4 0
Upper middle 6468 10 7 4 2
Upper 69+ 10 7 3 7
Total 51 39 22 12
Source: Authors.
Te top 10 studies provide the best
available evidence on the impact of health
insurance in low- and middle-income
countries (box table 2). Unfortunately, the
number of countries covered by the best
studies (China, Colombia, Costa Rica,
Taiwan, and a cross-country study) is
limited; all the more reason to widen and
deepen the evidence base.
(continued)
16 Chapter 2
BOX TABLE 2
The most robust evidence of the impact of health insurance
Country Author Year Title
Access
and use
Financial
protection
Health
status
China
Wagstaff
and Yu
2007 Do Health Sector Reforms Have
Their Intended Impacts?
The World Banks Health VIII Project
in Gansu Province, China.

Wagstaff and
others
2007 Extending Health Insurance to the
Rural Population: An Impact Evalu-
ation of Chinas New Cooperative
Medical Scheme (NCMS)

Yip, Wang,
and Hsiao
2008 The Impact of Rural Mutual Health
Care on Access to Care: Evaluation of
a Social Experiment in Rural China
Wang and
others
2008 The Impact of Rural Mutual Health
Care on Health Status: Evaluation of
a Social Experiment in Rural China
Colombia
Trujillo,
Portillo, and
Vernon
2005 The Impact of Subsidized Health
Insurance for the Poor: Evaluating
the Colombian Experience Using
Propensity Score Matching
Giedion, Diaz,
and Alfonso
2007 The Impact of Subsidized Health
Insurance on Access, Utilization
and Health Status: The Case of
Colombia

Costa Rica
Dow and
Schmeer
2003 Health Insurance and Child Mortal-
ity in Costa Rica

Dow,
Gonzlez, and
Rosero-Bixby
2003 Aggregation and Insurance-
Mortality Estimation
Cross-
country
Wagstaff and
Moreno-Serna
2007 Europe and Central Asias Great
Post-Communist Social Health Insur-
ance Experiment: Impacts on Health
Sector and Labor Market Outcomes

Taiwan
Chen, Yip,
Chang, Lin,
Lee, Chiu,
and Lin
2007 The Effects of Taiwans National
Health Insurance on Access and
Health Status of the Elderly

Total 7 3 7
Note: Studies are frst ordered alphabetically by country, and then by year of publication.
Source: Authors.
BOX 2.2 (continued)
Robustness of the evidence base
Impact of health insurance on access and use
Overall impact
Besides providing nancial protection from the economic consequences of ill-
ness, health insurance is meant to improve access (Nyman 1999). Seven of the
ten studies in the top quartile evaluate the link between health insurance and
access and use; nine nd a positive and signicant impact of health insurance
on access and use. Similar results are also found when we extend our analysis to
the full evidence base. A majority of the studies (39 of 51) analyze the impact
of health insurance on access and use, and 28 nd evidence indicating that
health insurance increases access to and use of health services. Tis nding
seems consistent with the results of previous reviews in the developed world
(see, for example, Buchmueller and Kronick 2005 and Hadley 2003 for a sum-
mary of this evidence). And it seems to conrm what insurance theory predicts:
health insurance reduces the price of health care and thereby promotes access
and use.
Te one study in the top quartile that does not present conclusive evidence on
the impact of health insurance on access and use compares 28 post-communist
countries in Eastern Europe and Central Asia, some of which have maintained
tax-nanced systems and some of which have switched to a social health insurance
scheme (Wagsta and Moreno-Serra 2007). It nds that social health insurance
has had a small positive impact on some use variables but not on others. One might
wonder, however, whether the heterogeneity in the social health insurance schemes
(in benets packages, institutional implementation, and so on) evaluated allows
for a meaningful cross-country comparison (even after controlling for observable
dierences such as variations in provider payment mechanisms). In this context
several studies nd that an aggregate measure of health insurance may cloud the
impact of health insurance by not taking into account the heterogeneity in impact
across dierent health insurance schemes.
2
Distributional impact of insurance on access and use
Te 10 studies in the top quartile suggest that the impact of health insurance var-
ies across populations but that these dierences vary substantially across countries
and settings. Some studies nd that it is precisely the most vulnerable (low income
and rural) population groups who benet most (Chen 2007; Trujillo, Portillo, and
Vernon 2005; Giedion, Daz, and Alfonso 2007). Others nd that only the better
o are increasing access and use as a result of health insurance (Wagsta and others
2007 on China). Still others nd that the middle-income population is beneting
18 Chapter 2
least (Yip, Wang, and Hsiao 2008). In some instances the impact on use across
population groups varies over time and even across research and health insurance
settings in the same country. Tis variability in results almost certainly stems from
unaccounted for design elements of the programs.
For example, Wagsta and others (2007) nd little impact from Chinas New
Cooperative Medical Scheme (NCMS) on access and use among the poor. Tey
explain this situation by looking at the specicities of NCMS: Given high coin-
surance rates, it is perhaps not surprising that there has been no signicant increase
in utilization among the poorest quintile. Yip, Wang, and Hsiao (2008) present
a more nuanced picture as they look at considerable heterogeneity in benet pack-
ages, coinsurance rates, deductibles, and ceilings across counties and coverage
modes. Tey nd that one modality of NCMS combining an individual savings
account for outpatient care with coverage for catastrophic care and high deduct-
ibles and ceilings has little impact on access and use. But another modality provid-
ing rst dollar coverage with no deductibles but with ceilings does have an impor-
tant impact on access and use, especially among the poorest and highest income
groups.
Te discussion and examples highlight the importance of incorporating the
possible heterogeneity of health insurance schemes and the impact across dierent
population groups into the study design and data collection. Typically household
data used to study insurance programs are collected for other purposes and are dif-
cult to use to understand the impact of insurance program design elements.
Other issues emerging from the literature
One interesting question put forward by several authors is related to the limits of
the concept of use when evaluating the impact of health insurance: if health insur-
ance is found to increase use is this necessarily good? As Nyman (1999) indicates:
Te value of insurance for coverage of unaordable care is derived from the value
of the medical care that insurance makes accessible. In this perspective and given
the substantial access problems in most low- and middle-income countries, observ-
ing improved access and use through health insurance will therefore generally be
considered a welfare gain.
What if health insurance encourages the overuse of health services? Wagsta
and others (2007) indicate, Te aim of health insurance is to reduce risk exposure
and to make necessary health care aordable . . . Teory suggests that the welfare
gains in terms of access must be weighed against the potential welfare loss from
demand-side and supply-side moral hazard. Further research is required to investi-
gate the issue of whether the extra utilization [obtained through health insurance]
is medically necessary or not. In our view, it can be safely assumed that in most
low-income countries and many middle-income countries. Te population and
especially the most vulnerable ones tend to experience severe access problems and
thereby underuse rather than waste and overuse health services, the literature does
not suciently discuss this tradeo between improved access of necessary services
and the potential moral hazard issues.
Much research goes beyond simply stating whether health insurance has a
positive impact on use of health services to ask more interesting questions that
should be further explored. For which services is an increase found and why?
Does use of preventive and curative health services increase (Waters 1999)?
Does insurance induce primarily increases in low cost-eective services or, to
the contrary, does it increase high cost-eective services (Dow, Gonzlez, and
Rosero-Bix 2003)? What do results nding a dierential impact across dierent
services say about the limits of the health benets provided under the insurance
scheme (Smith and Sulzbach 2008)? Is an increase in use accompanied by a sub-
stitution for inexpensive services by more expensive insurance-covered services
(or vice versa) or by a shift from informal and self-medication to formal care
(Hidayat and others 2004)? As the evidence on each of these questions is still
scarce, no generalizations are possible. Further exploration of these issues would
be extremely useful.
Impact of health insurance on nancial protection
General impact
Providing nancial protection against the economic consequences of illness lies at
the heart of the adoption of health insurance. As Nyman (1999) states: Why do
people purchase health insurance? Many economists would answer that it permits
purchasers to avoid risk of nancial loss. It comes then as no surprise that almost
half the studies included in our evidence base (22 of 51) evaluate the impact of
health insurance on nancial protection. However, only two in the top quartile
provide evidence on the impact of health insurance on nancial protection. Both
evaluate Chinas cooperative medical schemes but at dierent times and in dier-
ent parts of the country. Wagsta and Yu (2007) evaluate the impact of the World
Bank Health VIII project (containing an insurance component) in the Gansu
province using data from 2000 (pre-program) and 2004 (post-program). Tey nd
that the project had reduced both out-of-pocket payments and the incidence of cat-
astrophic payments, especially among the poorest. By contrast, Wagsta and others
(2007)evaluating the impact of the NCMS in 12 of Chinas 30 provinces and
20 Chapter 2
using data from 2003 (pre-intervention) and 2005 (post-intervention)nd that
it has had no statistically signicant eect on average out-of-pocket spending by
households, overall or on any specic type of care per contact, for either outpatient
or inpatient care. Indeed, they nd a hint that it may have increased the cost per
inpatient episode. However, across Chinas provinces and counties a lot of variation
exists in how the NCMS is being implemented, and the authors recognize that
this heterogeneity may constitute one important limitation of their study. In con-
trast, Yip, Wang, and Hsiao (2008) reach contrary conclusionsindicating that a
NCMS modality providing rst dollar coverage has indeed reduced out-of-pocket
payments and the incidence of catastrophic payments (Hsiao and Yip 2008).
As the examples illustrate once again, health insurance is not a homogeneous
concept, and in-depth familiarity with the specics of the health insurance scheme
being evaluated is key to interpreting results. Avoiding generalizations across coun-
tries and even across settings in the same country seems advisable unless details of
the plans can be controlled for, which is dicult if not impossible using existing
data and research techniques.
Because few studies in the top quartile evaluate the impact of health insurance
on nancial protection, we extend our analysis to the second quartile. All studies
in the second quartile nd that health insurance has reduced out-of-pocket spend-
ing and the incidence of catastrophic payments. Tis positive evidence on the
impact of health insurance on nancial protection still holds when considering all
studies in the evidence base and despite the fact that studies use many dierent
indicators to measure nancial protection and dierent model specications to
evaluate them.
3
Distributional impact
Two studies in the top quartile (Wagsta and Yu 2007; Wagsta and others
2007) conrm that the impact of health insurance varies across income groups.
Te study evaluating the World Bank Health VIII project in Gansu province
nds that health insurance seems to have had a greater impact in reducing out-
of-pocket payments among the poorest. Te results for Chinas NCMS are more
complicated. It seems to have increased average out-of-pocket spending among the
poorest decile but to have reduced the incidence of catastrophic spending among
this group. By contrast, the NCMS appears to have increased the incidence of
catastrophic spending among deciles 310, leaving average spending unaected
overall.
In the poorest decile no impacts on outpatient use are evident; impacts are
evident only in deciles 210. Te study also nds no impacts on inpatient use for
the poorest decile; statistically signicant positive impacts are found only in deciles
310. NCMS appears, in other words, to have increased average spending per epi-
sode among the poorest (as use has not changed) but reduced the incidence of cata-
strophic payments. Tis result is ascribed mainly to the limited extent of benets,
high copayments, and supply side incentives. Among the better o (deciles 210),
the increase in use and the cost of care per episode seems to have oset the miti-
gating impact of insurance on the price of each service. As this example indicates,
evaluating the impact of health insurance on out-of-pocket payments and the inci-
dence of catastrophic payments is challenging because it is the result of sequential
decisions (whether to use care, what type of care to consume, how much care, and
nally the price to pay for care based on the former sequence). Te positive impact
of health insurance on use may, for example, oset the reduction in price per health
service obtainedor health insurance may involve a substitution from informal
health services to costlier formal health services.
Does health insurance necessarily reduce out-of-pocket and catastrophic
payments?
Clearly not, as the distributional impacts illustrate. Interestingly, several addi-
tional studies in our evidence base (though not in the top quartile) seem to reach
similar conclusions. Ekman (2007b) evaluates the impact of dierent health
insurance schemes in Zambia
4
not only on out-of-pocket expenditure but also
on the broader concept of health carerelated out-of-pocket expenditure (out-of-
pocket spending on transportation, food, and other costs). Being exempted from
paying for care and having access to private or employment-based health insur-
ance signicantly reduces the risk of incurring catastrophic out-of-pocket expen-
diture. When other costs related to health care seeking are included (food and
transportation, for example), the probability of suering from the broader con-
cept of catastrophic health carerelated expenditure actually increases. Te author
puts forward two main reasons for this counterintuitive result. First, the sickest
people self-select into the prepayment scheme and their out-of-pocket payments
may have been even higher had they not been insured. (Because of data limita-
tions the author cannot control for this unobserved heterogeneity.) Second, and
more important, the prepayment scheme facilitates access, but once inside the
health system prepayments may induce the consumption of more costly health
services. Likewise, Trivedi (2002) nds some evidence that Vietnams voluntary
health insurance scheme increased out-of-pocket expenditures, even though the
eect was no longer signicant when commune-xed eects were included in the
regression.
22 Chapter 2
Te ndings are a clear invitation to further explore how health insurance
changes health careseeking behavior in quality, quantity, type, and composition
when evaluating the impact of a health insurance scheme on out-of-pocket and
catastrophic payments.
Other issues
Only one study from the evidence base (Wagsta and Pradhan 2005) goes beyond
evaluating whether health insurance reduces out-of-pocket or catastrophic pay-
ments and tries to understand whether health insurance helps reduce the impact
of illness on household consumption. Much more research of this type should be
undertaken because it helps us understand whether health insurance can really
mitigate the economic consequences of illness at the household level rather than
just indicate whether out-of-pocket spending rises or fallsor whether the level
might be catastrophic.
Impact of health insurance on health status
Health insurance improves health to the extent that it improves access to health
services that have a positive impact on health status. Even though the evidence is
still scarce, the interest from top researchers in documenting this causal link in
developing countries has grown. Only about a fth of the studies in the evidence
base (12) evaluate the impact of health insurance on health status, but most of
them (9) are in the top two quartiles (7 in the top, 2 in the second).
Te analysis here is based on results from the nine studies in the rst two quar-
tiles. Several nd no convincing evidence of an impact of health insurance on the
health status measures available. Te two studies of Costa Ricas social insurance
scheme nd only a small eect of social health insurance on child and infant mor-
tality (Dow, Gonzalez, and Rosero-Bixby 2003; Dow and Schmeer 2003). Simi-
larly, Giedion, Daz, and Alfonso (2007), using data from standard Demographic
and Health Surveys, nd that although the Colombian subsidized health insurance
scheme has greatly improved use of curative and preventive services by the poor, no
convincing evidence emerges of an impact on child mortality, low birthweight, or
self-perceived health status.
Chen and others (2007) use longitudinal data and a dierence-in-dier-
ence methodology to show that, although Taiwans National Health Insurance
greatly increased the use of both outpatient and inpatient services, the increase
did not reduce mortality or lead to better self-perceived general health status for
the Taiwanese elderly. Tey conclude that measures more sensitive than mor-
tality and self-perceived general health may be necessary to discern the impact
of health insurance on health status. Tis could indeed be the case especially
given that they nd that health insurance increases the use of health services,
which may increase awareness of health problems and thereby negatively impact
self-perceived health status. Likewise, and as indicated by Giedion, Daz, and
Alfonso (2007), the mortality rate may be too blunt a measure of health to cap-
ture improvements in health status brought about by health services under the
Taiwanese insurance scheme. Wagsta and Yus (2007) evaluation of the World
Bank Health VIII project in China nds mixed evidence of an impact. While the
evidence points to the projects reducing sick days, the evidence on chronic illness
and self-perceived health status is not conclusive. And it is dicult to attribute
any of these changes to health insurance alone since this project had several sup-
ply side interventions combined with the expansion of health insurance on the
demand side.
Wagsta and Moreno-Serra (2007) evaluate the impact of social health
insurance versus a general tax-nanced system in formerly communist coun-
tries in Eastern Europe and Central Asia on an extensive list of health out-
comes. Tey nd that, once they control for any concurrent dierences in pro-
vider payment systems, social health insurance does not lead to better health
outcomes.
Tree of the nine studies nd that health insurance has improved several health
status measures. Wagsta and Pradhan (2005), using panel data and matched-
double-dierence to evaluate the impact of health insurance on health status, nd
that Vietnams health insurance program favorably aected height-for-age and
weight-for-age among young school childrenand body mass index among adults.
Tis result is only suggestive because the aggregate health measures used by these
authors depend marginally on better access to health care and are strongly inu-
enced by other variables.
Wang and others (2008) nd that the community-based health insurance
scheme implemented in Guizhou province has had a positive eect on health
status among participants. Besides using self-perceived health status, they use
EQ-5D (a proprietary, standardized instrument to measure health outcomes).
Teir results indicate that among the ve EQ-5D dimensions, health insurance
signicantly reduced pain/discomfort and anxiety/depression for the general
populationand had positive impacts on mobility and usual activity for people
above age 55. Tey also nd that the positive impact has been greater among the
poorest.
Nyman and Barleen (2007) evaluate the impact of supplemental private
health insurance on self-perceived health status in Brazil. Tis study is an example
24 Chapter 2
of a way to tackle endogeneity. Te authors try to establish the causality (from
health insurance to an increase in health care and health) by analyzing specic
subpopulations.
Only respondents who indicated a specic acute illness, which would presumably
eliminate the inuence of illness on the decision to become insured. Tus, the
authors estimate the eect of health insurance on the use of health care and
on health, contingent on the respondents reporting an acute medical problem
within the last 30 days. Tis addresses, at least in part, the endogeneity from
the self-selection of sicker individuals into health insurance.
Only respondents who reported a chronic health problem, which eliminates the
inuence of a chronic condition on the decision to become insured. Tus, the
authors determine whether supplemental health insurance generates an
improvement in health status, conditional on the respondent having an acute
or chronic health problem.
Results from both models indicate that supplemental private health insurance
has improved self-perceived health status in Brazil. According to the authors, the
nding that better self-reported health status is associated with health insurance,
given the presence of acute or chronic conditions or other health problems, might
reect better control of symptoms or quicker recovery associated with the increased
access to health care available with health insurance.
Te results provide mixed evidence on the impact of health insurance status.
Te studies reviewed here use dierent measures of health status, so it may not be
surprising that results are inconsistent and hard to compare. It is not clear which
are the most suitable health status measures when evaluating the impact of health
insurance. Whatever the health status indicators nally chosen, they should be
directly and substantially related to the benets provided under the health insur-
ance scheme being evaluated. From this perspective, the current literature on the
impact of health insurance on health status in low- and middle-income countries
seems to still be in its infancy, perhaps related to the limited health status informa-
tion available in standard household surveys.
Te scarcity of the evidence on this issue is likely related, at least in part,
to the methodological challenge of evaluating the impact of health insurance on
health status. Besides the usual problem of unobserved confounding variables,
evaluating the impact of health insurance on health status is further compli-
cated by bidirectional causality: those insured may be healthier because they have
health insurance, but they may buy health insurance in part because they are
healthier, especially if access to insurance is positively related to income and type
of employment.
Heterogeneity of health insurance schemes
Health insurance varies considerably in design, target groups, benets coverage,
nancing mechanisms, and experience. Note that variations are observed both
across and within countries. Dierences in design aect not only what types of
benets are made more aordable and, therefore, what type of results we might
expect but also who tends to aliate. Te latter is important since one of the most
important methodological challenges in evaluating impacts of health insurance
is related to the nonrandom variation in health insurance status and the need to
correct for this possible endogeneity. So, to properly model the impact of health
insurance, it is crucial to understand what determines aliation with one health
insurance scheme or another.
In comparing the impact of health insurance across dierent health insurance
schemes, several studies show that health insurance is by no means a homogeneous
concept and that its impact depends on the specics of the insurance scheme.
A study by Ekman (2007b), evaluating the impact of multiple health insurance
schemes in Jordan, illustrates this point. Ekman rst nds no impact of insurance
coverage on outpatient care use, but when the type of insurance is disaggregated,
it turns out that people with access to the Ministry of Health insurance program
have a signicantly higher probability of seeking outpatient care than do people
covered under other insurance schemes. Similarly, Yip, Wang, and Hsiao (2008)
nd a signicant positive impact of health insurance for one new rural coopera-
tive medical scheme in China on use but only a limited impact for another type of
health insurance.
5
Te implications of this heterogeneity in health insurance schemes are clear.
First, the specication of the models should take this heterogeneity into account.
For example, in some cases it may be necessary to run dierent models for dierent
health insurance schemes,
6
to run dierent models for dierent population groups,
or to include some interaction terms between the health insurance dummy variable
and the groups of interest. Second, the possible heterogeneity of dierent health
insurance schemes indicates the need for care when trying to generalize results
across and even within countries.
Moving forward
Does health insurance matter in low- and middle-income countries? Tis review
indicates that studies show consistently that health insurance improves access and
use. Tis result is found among the 10 studies in the top quartile and among the
general evidence base, a nding consistent with what has been found in the devel-
oped world. Most studies in the top quartile and those in the general evidence base
26 Chapter 2
indicate that health insurance mitigates out-of-pocket expenditures and reduces
the incidence of catastrophic payments. We also found that studies constituting
outliers in this context indicate that the specic design of health insurance schemes
(high copayments and deductibles, little rst dollar coverage), together with the
fact that health insurance may increase use and cost per episode of care more than
it reduces the price to the insured of each service (for example, by providing incen-
tives to switch to costlier care when protected by health insurance), explain why
health insurance may not always increase nancial protection and reduce cata-
strophic costs. Tese results are important, as nancial protection lies at the heart
of any health insurance scheme.
We nd no conclusive evidence of an impact of health insurance on health
status. In this context some crucial issues must be answered, such as the type of
health status variables able to capture changes in health status that may result from
better access and use of health services resulting from health insurance. We ques-
tion whether self-perceived health status measures, indicators of nutritional status,
or blunt mortality information are good ways to go forward, and we suggest that
researchers should perhaps concentrate more on analyzing the impact of health
insurance on health services that are likely to have an important impact on health
status (for example, immunizations) rather than look at health status measures
directly.
What do we know about the robustness of the evidence base, and how could we
improve it? Our search strategy
7
produced more than four dozen studies evaluating
the impact of health insurance in low- and middle-income countries, despite the
restrictive inclusion criteria. Many of these studies have been published recently,
mirroring a growing interest among researchers and policymakers in health insur-
ance as a nancing option. We identied 10 studies providing the best available
evidence. Almost half do not use impact evaluation methods to test the eect of
health insurance and so provide only weak evidence on the impact of health insur-
ance. Tere is considerable room for stronger evidence.
Several methodological recommendations emerge from this review. First,
future studies should shift from purely correlational analysis to causal research that
isolates the impact of health insurance from other confounding variables. Second,
most studies reviewed here use retrospective standard cross-sectional household
data to evaluate impact. Such data, typically available in many countries, can go a
long way toward evaluating the impact of health insurance. But eorts to produce
prospective data should be supported. Care should be taken, however, when ran-
domized controlled trials are promoted as the only valid alternative to evaluate the
impact of health insurance. Quasi-experimental methods can provide reasonably
solid evidence, and social experiments may also suer from limitations, often and
most importantly from limited external validity.
8
Tird, the quality of a study does not depend solely on the sophistication of the
method. It depends on how well researchers understand a health insurance scheme
and how they use this knowledge to nd the most appropriate econometric tool
and the best available information to measure its impact. It is surprising that many
studies reviewed here spend little time establishing a clear link between the specif-
ics of a health insurance scheme and the method used to evaluate it. For example,
endogeneity is not omnipresent and can be context specic. Similarly, local context
may be an important confounding variable in some settings (for example, health
insurance coverage may be highest in places that also have the most complete pro-
vider network). Looking for complementary information may be crucial in such
a setting but will depend on the researchers understanding of the circumstances.
Familiarity with local circumstances thus becomes a key ingredient of quality lit-
erature in this eld.
What are some of the biggest knowledge gaps? Some aspects of the potential
impact of health insurance have yet to receive the full attention of researchers.
Tese include the distributional impact of health insurance; the impact of health
insurance in reducing inequality in use of services, expenditures, and nancial pro-
tection; the dynamics of the health insurance over time; the variation of impact
as the duration of exposure to health insurance varies; the impact of health insur-
ance on household consumption smoothing patterns; the spillover eects of health
insurance; the variation in impact across dierent health insurance schemes; the
variation in impact across health insurance and other supply side (for example, pro-
vider payment reforms) and demand side interventions (for example, equity funds)
targeted toward increasing nancial protection and improving access and use; the
cost-eectiveness of health insurance compared with other interventions; and the
general equilibrium eect of health insurance in recognition of the interrelation-
ship of market price and production.
As the geographic distribution of our evidence base indicates, a few coun-
tries seem to receive substantial attention from the research community (such as
China, Colombia, and Vietnam). But most other countries implementing health
insurance have received little attention, and some regions seem to be receiving
almost none.
Even though much work remains, based on the literature reviewed here, a compel-
ling case can be made for a positive correlation between having health insurance and
two important results: using more medical care and being less exposed to the nan-
cial risks associated with illness (although the latter is sensitive to how the insurance
28 Chapter 2
is designed). Although it is dicult to move from the conclusion that insurance
increases (or is associated with an increase in) the use of medical care to a conclu-
sion that insurance improves health outcomes or health status, almost all the eorts
to improve health in the world involve greater use of medical care: to benet from
specic procedures, to obtain immunizations, to improve knowledge and behavior, to
make births safer, to improve diagnosis, to screen patients, and more. Tat we do not
have a clear link from insurance to outcomesor that we do not better understand
why there is not a strong linkis an important challenge for the research community.
Notes
1. Box 2.2 outlines the protocol followed to evaluate the robustness of each of the arti-
cles reviewed. See the full study from which this chapter is drawn for details of the
conceptual framework, methods, and all 51 studies assessed (Giedion and Diaz 2008).
2. See Ekman 2007 for an excellent example of this issue.
3. Of the 22 studies evaluating nancial protection, 16 nd that health insurance has
improved nancial protection, 4 nd that health insurance has improved nancial
protection for some groups or for some indicators but not for all, 1 nds that health
insurance has actually worsened nancial protection, and 5 measure nancial protec-
tion but do not provide any information on the statistical signicance of their results.
4. A voluntary prepayment scheme, private or employment-based insurance, and a user
fee exemption scheme.
5. Tey nd a positive impact for a scheme operating in two western provinces that
provides rst dollar coverage for both inpatient and outpatient services and uses sup-
ply side interventions to improve quality and reduce ineciencies in health service
delivery, while they nd no signicant positive impact for another scheme common in
the western and central regions of China that provides a medical savings account that
combines an individual medical savings account with high-deductible catastrophic
insurance and that provides mainly catastrophic insurance for expensive hospital ser-
vices. Chapter 8 of this book explains the ndings in detail.
6. Data from a voluntary health insurance scheme for the informal sector may have, for
example, a dierent endogeneity problem than a mandatory social insurance scheme
for formal sector workers.
7. Tis literature review searched studies in online databases, performed manual
searches, reviewed reference lists of related papers, and inspected webpages of major
international organizations and donors. To be included in the list of studies reviewed,
the study must have been about a health-related insurance mechanism; must have
addressed out-of-pocket spending, catastrophic health expenditure, access to care,
use of health services, or health status; must have been quantitative; and must have
appeared in an academic journal or book. Of course, the empirical focus needed to be
a low- or middle-income country. We reviewed papers published between 2000 and
2008 and written in English only.
8. Tere is a vigorous debate about the structural versus program evaluation approach in
econometrics. See Deaton (2010), Heckman (2010), and Imbens (2010).
References
Buchmueller, T., and R. Kronick. 2005. Te Eect of Health Insurance on Medical Care Utili-
zation and Implications for Insurance Expansion: A Review of the Literature. Medical Care
Research and Review 62 (1): 330.
Chen, L., W. Yip, M. Chang, H. Lin, S. Lee, Y. Chiu, and Y. Lin. 2007. Te Eects of Taiwans
National Health Insurance on Access and Health Status of the Elderly. Health Economics 16
(2) 22342.
Deaton, A. 2010. Instruments, Randomization, and Learning about Development. Journal of
Economic Literature 48 (2): 42455.
Dow, W. H., K. Gonzlez, and L. Rosero-Bixby. 2003. Aggregation and Insurance Mortality
Estimation (Costa Rica). NBER Working Paper 9827. Cambridge, MA: National Bureau of
Economic Research.
Dow, W. H., and K. K. Schmeer. 2003. Health Insurance and Child Mortality in Costa Rica.
Social Science & Medicine 57 (6): 97586.
Ekman, B. 2007. Te Impact of Health Insurance on Outpatient Utilization and Expendi-
ture: Evidence from One Middle-Income Country Using National Household Survey Data.
Health Research Policy and Systems 5 (6).
Giedion, U., E. Alfonso, and Y. Diaz. 2007. Measuring the Impact of Mandatory Health Insur-
ance on Access and Use: Te Case of the Colombian Contributory Regime. Working Paper.
Brookings Institution, Washington, DC.
Giedion, U., Y. Daz, and E. A. Alfonso. 2007. Te Impact of Subsidized Health Insurance on
Access, Utilization and Health Status: Te Case of Colombia. World Bank, Washington, DC.
Hadley, J. 2003. Sicker and PoorerTe Consequences of Being Uninsured: A Review of the
Research on the Relationship between Health Insurance, Medical Care Use, Health, Work,
and Income. Medical Care Research and Review 60 (2): 375.
Heckman, J. J. 2010. Building Bridges between Structural and Program Evaluation Approaches
to Evaluating Policy. Journal of Economic Literature 48 (2): 35698.
Hidayat, B., H. Tabrany, H. Dong, and R. Sauerborn. 2004. Te Eects of Mandatory Health
Insurance on Equity in Access to Outpatient Care in Indonesia. Health Policy and Planning
19 (5): 32235.
Hsiao, W., and W. Yip. 2008. Discover New Models of Health Insurance through Social Experi-
mentation. Presentation at the Brookings Institution, April 15, Washington, DC.
30 Chapter 2
Imbens, G. W. 2010. Better LATE Tan Nothing: Some Comments on Deaton (2009) and
Heckman and Urzua (2009). Journal of Economic Literature 48 (2): 399423.
Oxfam, Action for Global Health, Medicins du Monde, Global Health Advocates, Save the Chil-
dren, and Act Up Paris. 2008. Health Insurance in Low-Income Countries: Where Is the Evi-
dence Tat It Works? Joint NGO Brieng Paper 112. [www.oxfam.org.uk/resources/policy/
health/downloads/bp112_health_insurance.pdf ].
Smith, K., and S. Sulzbach. 2008. Community-Based Health Insurance and Access to Maternal
Health Services: Evidence from Tree West African Countries. Social Science & Medicine 66
(12): 246073.
Trujillo, A., J. Portillo, and J. Vernon. 2005. Te Impact of Subsidized Health Insurance for the
Poor: Evaluating the Colombian Experience Using Propensity Score Matching. International
Journal of Health Care Finance and Economics 5 (3): 21139.
Wagsta, A. 2007a. Te Economic Consequences of Health Shocks: Evidence from Vietnam.
Journal of Health Economics 26 (1): 82100.
. 2007b. Health Insurance for the Poor: Initial Impacts of Vietnams Health Care Fund
for the Poor. Policy Research Working Paper 4134. World Bank, Washington, DC.
Wagsta, A., M. Lindelow, G. Junc, X. Ling, and Q. Juncheng. 2007. Extending Health Insur-
ance to the Rural Population: An Impact Evaluation of Chinas New Cooperative Medical
Scheme. Impact Evaluation Series 12. World Bank, Washington, DC.
Wagsta, A., and M. Pradhan. 2005. Health Insurance Impact on Health and Nonmedical Con-
sumption in a Developing Country. Policy Research Working Paper 3563. Washington, DC:
World Bank.
Wagsta, A., and S. Yu. 2007. Do Health Sector Reforms Have Teir Intended Impacts? Te
World Banks Health VIII Project in Gansu Province, China. Journal of Health Economics
26 (3): 50535.
Wang, H., W. Yip, L. Zhang, and W. Hsiao. 2008. Te Impact of Rural Mutual Health Care on Health
Status: Evaluation of a Social Experiment in Rural China. Health Economics 18 (2): 6582.
Waters, H. R. 1999. Measuring the Impact of Health Insurance with a Correction for Selection
BiasA Case Study of Ecuador. Health Economics 8 (5): 47383.
Yip, W., H. Wang, and W. Hsiao. 2008. Te Impact of Rural Mutual Health Care on Access
to Care: Evaluation of a Social Experiment in Rural China. Harvard School of Public
Health, Boston, MA. [www.hsph.harvard.edu/phcf/publications/Yip.Wang.Hsiao.2008
.RMHC.access.pdf ].
Studies reviewed but not cited
Bales, S., J. Knowles, H. Axelson, P. D. Minha, H. D. Long, and T. M. Oanhd. 2007. Te Early
Impact of Decision 139 in Vietnam: An Application of Propensity Score Matching. Vietnam,
Ministry of Health, Hanoi.
Bleich, S. N., D. M. Cutler, A. S. Adams, R. Lozano, and C. J. Murray. 2007. Impact of Insur-
ance and Supply of Health Professionals on Coverage of Treatment for Hypertension in Mex-
ico: Population Based Study. BMJ 335 (7625): 87578.
Dror, D., E. Soriano, M. Lorenzo, J. Sarol Jr., R. Azcuna, and R. Koren. 2005. Field Based Evi-
dence of Enhanced Healthcare Utilization among Persons Insured by Micro Health Insurance
Units in Philippines. Health Policy 73 (3) 263271.
Ekman, B. 2004. Community-based health insurance in low-income countries: a systematic
review of the evidence. Health Policy and Planning 19 (5): 24970.
. 2007. Catastrophic Health Payments and Health Insurance: Some Counterintuitive
Evidence from One Low-Income Country. Health Policy 83 (2): 30413.
Flrez, C. E., U. Giedion, and R. Pardo. Forthcoming. Te Impact of Health Insurance in
Colombia on Financial Protection. Washington, DC: Inter-American Development Bank.
Gertler, P., and J. Gruber. 2002. Insuring Consumption against Illness. American Economic
Review 92 (1): 5170.
Gertler, P., L. Locay, and W. Sanderson. 1987. Are User Fees Regressive? Te Welfare Implica-
tions of Health Care Financing Proposals. Journal of Econometrics 36 (12): 6788.
Gotsadze, G., A. Zoidze, and O. Vasadze. 2005. Reform Strategies in Georgia and Teir Impact
on Health Care Provision in Rural Areas: Evidence from a Household Survey. Social Science
& Medicine 60 (4): 809821.
Heckman, J. J., R. J. LaLonde, and J. A. Smith. 1999. Te Economics and Econometrics of
Active Labor Market Programs. In O. Ashenfelter and D. Card, eds., Handbook of Labor
Economics. Amsterdam: Elsevier.
Jakab, M., A. Preker, C. Krishnan, P. Schneider, F. Diop, J. Jtting, A. Gumber, M. Ranson, and
S. Supakankunti. 2004. Analysis of Community Financing Using Household Surveys. In A.
Preker and G. Carrin, eds, Health Financing for Poor People: Resource Mobilization and Risk
Sharing. Washington, DC: World Bank.
Johar, M. 2007. Te Impact of the Indonesian Health Card Program: A Matching Estimator Approach.
Discussion Paper 2007-30. University of New South Wales, School of Economics, Sydney, Australia.
Jowett, M., P. Contoyannis, and N. Vinh. 2003. Te Impact of Public Voluntary Health Insur-
ance on Rivate Health Expenditures in Vietnam. Social Science & Medicine 56 (2): 33342.
Jtting, J. 2004. Financial Protection and Access to Health Care in Rural Areas of Senegal. In A.
Preker and G. Carrin, eds., Health Financing for Poor People: Resource Mobilization and Risk
Sharing. Washington, DC: World Bank.
Khan, A. A., and S. M. Bhardwaj. 1994 Access to Health Care: A Conceptual Framework and
its Relevance to Health Care Planning. Evaluation & Te Health Professions 17 (1): 6076.
Levy, H., and D. Meltzer. 2001. What Do We Really Know about Whether Health Insurance
Aects Health? ERIU Working Paper 6. University of Michigan, Economic Research Initia-
tive on the Uninsured, Ann Arbor, MI.
32 Chapter 2
Limwattananon, S., V. Tangcharoensathien, and P. Prakongsai. 2007. Catastrophic and Poverty
Impacts of Health Payments: Results from National Household Surveys in Tailand. Bulletin
of the World Health Organization 85 (8): 60006.
Liu, T. C., C.S. Chen, and L. M. Chen. 2002. Te Impact of National Health Insurance on Neo natal
Care Use and Childhood Vaccination in Taiwan. Health Policy and Planning 17 (4): 38492.
Liu, Y., K. Rao, and W. Hsiao. 2003. Medical Expenditure and Rural Impoverishment in China.
Journal of Health, Population, and Nutrition 21 (3): 21622.
Lu, J. F., and W. Hsiao. 2003. Does Universal Health Insurance Make Health Care Unaordable?
Lessons from Taiwan. Health Aairs 22 (3): 7788.
Pagan, J. A., A. Puig, and B. J. Soldo. 2007. Health Insurance Coverage and the Use of Preventive
Services by Mexican Adults. Health Economics 16 (12): 135969.
Pagan, J., S. Ross, J. Yau, and D. Polsky. 2006. Self-Medication and Health Insurance Coverage
in Mexico. Health Policy 75 (2): 17077.
Palmer, N., D. Mueller, L. Gilson, A. Mills, and A. Haines. 2004. Health Financing to Promote
Access in Low Income SettingsHow Much Do We Know? Lancet 364 (9442): 136570.
Penchansky R., and J. W. Tomas. 1981. Te Concept of Access: Denition and Relationship to
Consumer Satisfaction. Medical Care 19 (2): 12740.
Pradhan, M., F. Saadah, and R. Sparrow. 2007. Did the Health Card Program Ensure Access
to Medical Care for the Poor during Indonesias Economic Crisis? World Bank Economic
Review 21 (1): 125151.
Ranson, M. 2004. Te SEWA Medical Insurance Fund in India. In A. Preker and G. Carrin,
eds., Health Financing for Poor People: Resource Mobilization and Risk Sharing. Washington,
DC: World Bank.
Saadah, F., M. Pradhan, and R. Sparrow. 2001. Te Eectiveness of the Health Card as an Instru-
ment to Ensure Access to Medical Care for the Poor during the Crisis. World Bank Economic
Review 21 (1): 125150.
Schneider, P., and F. Diop. 2004. Community-Based Health Insurance in Rwanda. In A. Preker
and G. Carrin, eds., Health Financing for Poor People: Resource Mobilization and Risk Shar-
ing. Washington, DC: World Bank.
Sepehria, A., W. Simpsona, and S. Sarma. 2006. Te Inuence of Health Insurance on Hospital Admis-
sion and Length of StayTe Case of Vietnam. Social Science & Medicine 63 (7): 175770.
Sinha, T., M. Ranson, and A. Mills. 2007. Protecting the Poor? Te Distributional Impact of a
Bundled Insurance Scheme. World Development 35 (8) 140421.
Supakankunti, S. 2004. Impact of the Tailand Health Card. In A. Preker and G. Carrin, eds.,
Health Financing for Poor People: Resource Mobilization and Risk Sharing. Washington, DC:
World Bank.
Yip, W., and P. Berman. 2001. Targeted Health Insurance in a Low Income Country and Its
Impact on Access: Egypts School Health Insurance. Health Economics 10 (3): 20720.
33
C
h
a
p
t
e
r

3
Low-Cost Health Insurance
Schemes to Protect the
Poor in Namibia
Emily Gustafsson-Wright, Wendy Janssens, and Jacques van der Gaag
Investigating alternative mechanisms of health care provision is impor-
tant for African countries, where the epidemics of HIV/AIDS, tubercu-
losis, and malaria increase the demands on the health care sector.
Tis chapter, using a unique combination of household survey data
and a biomedical survey with HIV test data from Greater Windhoek in
Namibia, analyzes the extent to which the fairly well developed public
health care sector in Namibia oers protection from health shocks to
uninsured households. Namibia is in the top tier of African countries
in health expenditures. Not only is government health spending high in
relative terms at almost 8% of gross domestic product (GDP), but out-
of-pocket expenditures are the second lowest among African countries,
after South Africa. So one would expect that the benecial role of public
health care would be particularly visible in this Southern African coun-
try. Namibia is also severely aected by the HIV/AIDS epidemic. Te
latest estimates suggest a prevalence rate of 15% among working-age
adults (UNAIDS 2008).
Te goal of this study is to investigate the potential for health insur-
ance schemes in this setting, given the recent introduction of subsidized
low cost insurance with HIV treatment components in Greater Wind-
hoek. Te chapter begins with a description of health and the health
sector in Namibia. Next, it examines self-reported health status, health
care use, and out-of-pocket health expenditures across insurance status
and consumption quintiles. It then investigates the coping strategies of
34 Chapter 3
uninsured and insured households that face particular health shocks, looking at
a death in the family, extended hospitalization, substantial weight loss, and HIV
infection. Te last section discusses the scope for targeting and subsidizing private
voluntary insurance schemes.
Health and the health care sector in Namibia
Namibia is a lower middle-income country with a gross national income per capita
of US$6,240 in purchasing power parity (PPP) terms in 2008. Te Sub-Saharan
Africa average was US$1,949 (World Bank 2010). However, this number conceals
enormous dierences in wealth within the population. In fact, Namibia has one of
the highest levels of inequality in the world, with a Gini coecient of 0.7. Te rich-
est 10% of the population receives 65% of the countrys income, and about 35% of
the population lives below the poverty line of $1 a day (WHO 2004).
Te Namibian population suers from three major communicable diseases:
HIV/AIDS, tuberculosis, and malaria. Adult (ages 1549) HIV prevalence rates
increased from 2.5% in 1992 to 15.3% in 2007, when there were an estimated
200,000 HIV-infected people, 14,000 of them children under age 15. Te esti-
mated number of people in need of antiretroviral therapy in that year was 59,000,
which could well rise above 200,000 when infected people develop AIDS. Tuber-
culosis, the second major cause of deaths in hospitals, is estimated at 767 cases per
100,000 people per year, with a mortality rate of 96 per 100,000 per year (WHO
2008b). Malaria infects on average 400,000 people per year and causes 877 deaths,
mainly in the north (WHO 2005). Noncommunicable diseases are increasingly
responsible for morbidity and mortality among adults, especially diabetes and car-
diovascular diseases (WHO 2004).
Just after gaining independence from South Africa in 1990, the Namibian
health system was divided. Most health facilities were in urban areas, segregated by
race. Equality gaps in access to health care existed not only between rural and urban
dwellers but also between the rich and the poor. But in the last two decades, a strong
political commitment to upgrade the primary health care system has made health
services more responsive to the needs of the population, albeit slowly (WHO 2004).
Namibia is among the best o African countries in health spending. From
1993 to 2000, 11% of government spending was earmarked for health (WHO
2004). In 2007 it spent 7.6% of GDP on health (PPP US$467 per capita), 42%
nanced by the government and 58% from private payments (of which only 5.8%
was out of pocket). In Sub-Saharan Africa the average is 6.4% of GDP for health
(PPP US$124 per capita), 41% from government and 59% private, of which 60%
was out of pocket (World Bank 2010).
Low-Cost Health Insurance Schemes to Protect the Poor in Namibia 35
Te private health sector in Namibia has a well established for-prot compo-
nent, providing hospital services mainly in the urban centers. Not-for-prot mis-
sionary health facilities operate in the communal areas in the north (WHO 2004).
Private sources of nance include private insurance premiums and user fees in pub-
lic health care facilities. Public facilities charge at user fees, depending on the
facility. Te cost recovery ratio for 2001 was 2% (WHO 2004). Medicines are gen-
erally aordable due to the highly subsidized user fees. But the public sector suers
from long waiting times. And there is a critical shortage of health professionals,
particularly outside urban areas.
Te Namibian health insurance industry, organized primarily into either open
or closed medical aid funds, is better developed than it is in most other African
countries. A fund is a nonprot entity that pays benets directly to medical pro-
viders in proportion to the services rendered to the beneciary. Closed funds limit
membership to employees in a rm or industry. Te government health fund,
PSEMAS, is considered a closed fund since it is limited to those working in govern-
ment. Te open funds are Namibia Medical Care, Namibia Health Plan, Renais-
sance Medical Aid Fund, and Nammed Medical Aid Fund.
3
Insurance enrollment in Greater Windhoek
Te household survey data used in this chapter covers only Greater Windhoek
(box 3.1). Insurance coverage is high for a Sub-Saharan country, with more than
30% of individuals enrolled in a medical aid fund (table 3.1). Te government health
BOX 3.1
Data collection
Te data source for this study is the Re-
public of Namibia Okambilimbili Survey
2006 which includes socioeconomic and
biomedical parts.
1
Te socioeconomic
part was conducted among a representa-
tive, self-weighted sample of the Greater
Windhoek population, including 1,796
households and 7,343 individuals. It
includes data on basic household structure
as well as extensive data on health status,
health expenditures, health insurance,
and willingness to pay for health insur-
ance. Te biomedical part includes a
saliva-based HIV test for those ages
12 and older consenting to participate.
Having such a rich household survey
connected with an HIV test provides a
unique dataset and opportunity to ana-
lyze these data together.
Note
1. Te survey was conducted by sta of the Multi-
disciplinary Research and Consultancy Centre
at the University of Namibia, and the National
Institute of Pathology, in cooperation with the
Amsterdam Institute for International Develop-
ment and the PharmAccess Foundation.
36 Chapter 3
fund, PSEMAS, insured 43% of all insured individuals. Namibia Medical Care and
Namibia Health Plan each insure about a third of the number that PSEMAS does.
Tere are large discrepancies in enrollment across socioeconomic categories.
Only 5% of individuals in the poorest consumption quintile are enrolled in medical
aid, while 70% of individuals in the richest quintile have medical aid benets. Not
surprisingly, the employed are more likely to be insured than the unemployed, with
considerable variation by industry. Tose most likely to be insured are individuals
whose head of household works in government or defense. Household members
with household heads who work in education and health follow close behind. Te
least insured industries are manufacturing, retail/accommodation, and construc-
tion. Te service industry employs the most individuals in Namibia, and 65% of
individuals with a head of household employed in services are uninsured.
TABLE 3.1
Individuals enrolled in a medical aid fund
Household characteristic
Total
population
Individuals
insured (%)
Total number
of households
Households
insured (%)
Consumption quintile
1 (poorest) 1,404 5.3 238 14.3
2 1,381 13.6 285 28.4
3 1,396 25.9 329 40.4
4 1,390 44.0 358 55.3
5 (richest) 1,390 69.1 444 81.3
Employment status (household head)
Employed 6,002 34.9 1,430 52.4
Unemployed 1,249 14.1 320 25.6
Employment sector (household head)
Government and defense 951 58.2 222 93.7
Education 337 51.6 69 82.6
Health 178 49.4 38 76.3
Services 1,853 35.2 441 50.3
Transport and storage 346 33.0 70 50.0
Manufacturing 226 24.3 58 36.2
Retail and accommodation 450 17.1 107 29.9
Construction 346 15.6 103 23.3
Others (such as agriculture and mining) 1,312 24.9 321 38.3
Total 7,343 30.9 1,769 47.2
Note: Total averages by category may differ due to missing observations in some categories.
Source: Authors calculations based on 2006 Republic of Namibia Okambilimbili Survey.
A household is considered insured if at least one of its members is enrolled in a
medical aid fund. In 47% of Greater Windhoek households at least one individual
has medical insurance, a percentage substantially larger than individual cover-
age rates, indicating that insurance is targeted to individuals rather than to whole
families. For example, 94% of households whose head is employed in government
or defense have at least one insured household member. But only 58% of their
household members are insured. Relative enrollment rates across socioeconomic
categories are similar to individual coverage rates: the rich and the employed are
much more likely to participate in a medical aid fund than poorer households.
Inequality in health status, health care use, and health expenditures
Do the insured dier from the uninsured in their health status, and among the
uninsured, do the rich dier from the poor? Given their health status, do the unin-
sured and insured dier in their use of health care? And do the insured and the
uninsured dier in their level and proportion of out-of-pocket payments as part of
their overall spending?
Overall, the insured are more likely to report chronic illness, acute illness, and
hospitalization (table 3.2). One interpretation is that those who are insured have
insurance because they have poorer health. A negative correlation between insured
status and health status could suggest cream-skimming, where insurance compa-
nies insure those in better health, which seems not to be the case here.
In examining dierences across quintiles, chronic disease increases systemati-
cally with income for both the uninsured and the insured. Tis may be because
TABLE 3.2
Reported prevalence of chronic disease, acute illness, or injuryand
incidence of hospitalization among the insured and uninsured (%)
Insured Uninsured
Quintile Chronic Acute Hospitalization Chronic Acute Hospitalization
1 (lowest) 13.51 12.16 9.46 10.20 13.31 7.59
2 14.36 16.49 3.72 11.58 17.67 6.45
3 16.30 22.38 8.56 12.27 17.60 5.22
4 16.53 17.81 6.70 11.44 16.71 4.24
5 (highest) 20.02 19.04 6.66 14.29 13.29 1.86
Total 17.73 18.80 6.83 11.57 15.89 5.73
Note: For acute illness and hospitalization the reported prevalence refects that the individual
experienced at least one episode in the last year.
38 Chapter 3
the poor have less information or awareness about their chronic conditions.
4
For
acute illness, the poorest and the wealthiest are generally less likely to report
an episode than those in the middle quintiles. Although little systematic infor-
mation is conveyed by these data, there appear to be signicant dierences in
reported health status: the insured and the wealthier are more likely to report both
chronic and acute illness, though it is possible that the poor and uninsured may
be underreporting.
Given these dierences in reported health status, is there a signicant dier-
ence in health care use between the insured and uninsured for those who report
having had an acute illness or hospitalization?
5
Te rst notable dierence between
the uninsured and the insured is that the uninsured seek no care for acute illness
more than 20% of the time, compared with 14% for the insured (table 3.3). Tis
could mean that uninsured individuals are forgoing care because they cannot pay
for the health services and the travel costs to get to a health center. Or they choose
to opt out of care because they deem the care to be low quality or because of long
waiting lines in public health service locations.
When services are used for illness, there is a marked dierence in public or pri-
vate facilities between the uninsured and the insured for both acute illness and hos-
pitalization. Among the uninsured, government health facilities are used in 66% of
the cases, while only 10% of insured individuals used government facilities for an
acute illness. A mere 7% of the uninsured used private hospitals for inpatient care,
compared with 63% of insured individuals. Tat uninsured individuals are forgo-
ing care more often than the insured and that the uninsured are possibly under-
reporting illness ag the inequitable and potentially harmful health consequences
for individuals lacking health insurance.
Te uninsured are less likely to report an illness, but when they do report, they
are also less likely to seek care. When individuals without health insurance seek
care, they are more likely to do so in public health facilities.
TABLE 3.3
Use of health services for acute illness or injury and hospitalization
Type of
health facility
Acute illness or injury Hospitalization
Uninsured Insured Total Uninsured Insured Total
None 20.36 14.35 18.23
Government 66.41 10.19 46.47 93.01 36.69 71.24
Private 10.31 71.53 32.02 6.99 63.31 28.76
Traditional 2.93 3.94 3.28
Our third question is whether there are inequities in out-of-pocket payments
for health between the insured and uninsured for those who sought care and
incurred expenses. In absolute terms the insured pay more out of pocket than the
uninsuredup to twice as much for chronic illness and ve times as much for hos-
pitalization (table 3.4). But in relative terms, the uninsured pay more (as a percent-
age of per capita consumption) for both chronic and acute illness.
Tis nding raises another ag. Even in a public health care system that is fairly
well developed and used by the uninsured, the uninsured are disproportionately
aected nancially by out-of-pocket health expenditures relative to the insured. Peo-
ple who are ill, choose to seek care, and must pay for those costs out of pocket pay
twice as much as the insured for acute illness as a proportion of their total income.
Tese inequalities become quite stark for the bottom quintiles among the unin-
sured. In absolute terms people in the higher quintiles spend much more than the
poor on health care, but those in the lower quintiles spend on average a higher
proportion of their consumption per capita on chronic and acute illness, albeit in
a less systematic pattern (table 3.5). Uninsured individuals in the bottom quintiles
spend on average up to 14% of their per capita income on acute illness (quintile 2).
Tis nding again ags the inequities surrounding insurance coverage, even in a
country with a health sector in relatively good shape.
Coping with health shocks when uninsured
Despite the fairly well developed public health care system in Greater Windhoek, the
uninsured and the poor seek treatment less often than the insured and the wealthy.
And if the uninsured and the poor do seek treatment, they fare relatively worse
nancially. Tis section expands on the previous section by investigating in more
detail the economic consequences of health shocks on out-of-pocket expenditures
TABLE 3.4
Average per capita annual out-of-pocket health expenditures
Type of illness Insured Uninsured
Chronic
(US$) 1,078 491
(%) 2.7 4.0
Acute
(US$) 1,377 967.
(%) 3.5 7.8
Hospitalization
(US$) 1,210 226
(%) 3.1 1.8
Note: For all individuals with positive health expenditures.
40 Chapter 3
and income-earning capacity for households without health insurance compared
with those with health insurance. It also analyzes which coping strategies households
adopt in the face of health shocks to deal with medical expenses and reduced income.
Our ability to analyze the mitigating eects of insurance is limited given the
cross-sectional data, so we cannot fully deal with selection and simultaneity bias.
Despite the data constraints, however, the ndings here contribute to the policy
debate on health insurance by highlighting the consequences of health shocks on
uninsured households. Taking advantage of our unique dataset, which combines
socioeconomic data and HIV tests, the results provide insights into the potential
benets of private or community-based health insurance.
Poor households use various coping strategies to buer shocks. A decline in
earned income may be partially oset by an increase in unearned income. Tis
would be the case if the sick individual benets from a social scheme such as a
disability grant or illness compensation. Alternatively, relatives may increase remit-
tances to the aected household. Friends, neighbors, or other community members
may provide gifts and other forms of informal assistance. To deal with the remainder
of the income loss, the medical expenses, and the need for care at home, households
can reallocate the labor of healthy household members. Individuals may increase
their labor supply by working more hours on their job or taking up a second job, or
they may enter the labor market if not working for income yet. Or households can
deplete their savings, sell assets to generate additional monetary resources, borrow
money, or buy goods on credit. A health shock could also induce a household to
reduce food and nonfood consumption or to postpone large nonmedical household
expenditures. Finally, the household may decide to forgo health care altogether.
TABLE 3.5
Average per capita annual out-of-pocket health expenditures for the
uninsured, by quintile
Chronic Acute Hospitalization
Quintile (US$) (%) (US$) (%) (US$) (%)
1 (lowest) 140 5.4 291 11.1 50 1.9
2 250 4.3 806 13.9 54 0.9
3 408 3.9 849 8.2 366 3.5
4 1,123 5.6 978 4.9 182 0.9
5 (highest) 1,015 1.9 3,787 7.2 3,548 6.7
Total 491 4.0 967 7.8 226 1.8
Note: For all individuals with positive health expenditures.
Analysis
We estimate the relationship between health shocks and the following economic
variables: income, medical expenditures, labor supply, consumption, assets, and
credit.
6
In particular, we estimate the following ordinary least squares regression
for uninsured and insured households separately (following Wagsta 2007):
y
h
t
= + s
h
t 1
+ X
h
t
+
n
+
h
t
for i
h
t
= 0, 1
where y
h
t
is the outcome variable for household h at time t, s
h
t 1
is a dummy variable
for each of four health shocks equal to 1 if the health shock occurred to household
h in the 12 months before the time of the survey t, X
h
t
is a vector with household
characteristics (age, age squared, gender, education of the household head, house-
hold size, and number of children) measured at time t,
7
n
captures neighborhood
xed eects such as the presence of health facilities or employment opportunities,
and
h
t
is an unobserved error term. A household is considered to be insured (i
h
t
= 1)
if at least one of its members is enrolled in a health insurance scheme at time t and
is considered uninsured (i
h
t
= 0) otherwise.
Te analysis is at the household level. Tat is, we assume that household mem-
bers pool their income and share the burden of medical expenditures. We are par-
ticularly interested in the coecients , which reect how, given a households
insurance status, a particular health shock is associated with dierences in house-
hold income, medical spending, and coping strategies.
Te cross-sectional data put two important restrictions on the analysis. First,
we cannot perform an impact analysis of the mitigating eects of insurance. Te
decision to enroll in a medical aid fund depends in part on unobserved characteris-
tics. For example, people who are more concerned with their health could be more
likely to take insurance while being less prone to health shocks because of a healthy
lifestyle. Or individuals with private information on particular health care needs
could be more inclined to enroll.
Indeed, we have some indications of adverse selection into private insurance
schemes. Individuals with a parent who suers from a chronic disease, which in turn
increases ones own risk of developing a chronic disease, are more likely to be insured.
But we do not nd a signicant overrepresentation among the insured of individuals
who suer from HIV/AIDS and high blood pressure, the main communicable and
noncommunicable causes of morbidity and mortality in adults. Without panel data or
an experimental setup of the insurance scheme, it is not possible to control for (time-
invariant) unobserved characteristics that aect insurance status. Instead, we stratify
the analysis by insurance status to yield insights in the relationship between health
shocks and economic outcomes for the uninsured and the insured sample separately.
42 Chapter 3
Second, simultaneity eects inuence interpretations of our ndings. Although
the health shocks in our dataset occurred prior to the survey, it is possible that
the direction of causality between a health shock and an outcome variable goes
both ways. For example, a low earned income over the past 12 months may aect
a households vulnerability to health shocks in that same period. In that case, a
signicant negative coecient in the regression for earned income might capture
either the shocks eect on income-generating capacity or the households poverty-
related vulnerability for health shocks, or both. Another source of bias may stem
from omitted variables, such as latent health status. So the ndings should be inter-
preted as correlations, not as causal eects.
Description of the variables
Te analysis looks at the consequences of four types of health shocks in the 12
months prior to the survey. All health shocks relate to working-age household
members only, that is, individuals ages 1565.
Te rst shock is a dummy variable equal to 1 if any working-age household
member reported a loss of weight in the previous 12 months. Tis variable is a
self-reported measure of general health status. Tere is much evidence that losing
weight (or a drop in body mass index) is signicantly related to an overall deterio-
ration of an individuals health status.
8
In 28.8% of the households at least one
working-age individual reported losing weight in the past year.
Te second health shock is whether a working-age member of the household is
infected with HIV. Tis variable is based on a direct medical saliva-based test for
HIV infection among individuals ages 12 and older. Due to a lengthy validation
process of the saliva-based HIV test in Namibia, only 53% of the initial respon-
dents in the relevant age range could be tracked and interviewed at the time of
revisit by nurses ve months after the socioeconomic survey (Janssens, Rinke de
Wit, and van der Gaag 2007). Of the respondents, 20% had relocated to a newly
constructed neighborhood in Greater Windhoek with improved social services and
infrastructure. Tis concerns especially households previously living in areas with
low-quality access to water, sanitation, and other facilities. Sixteen percent were not
present during the revisit for reasons of holidays, leave from work, or working out
of town. Te remainder refused to participate in the biomedical survey. At 86%,
the participation rate in the HIV test among the respondents who could be tracked
is high. A correction for nonresponse based on observed characteristics suggests
that the bias due to refusal would be small. However, further adjustments for
unobserved characteristics indicate that HIV-positive individuals are more likely to
have refused to participate in the HIV test (Janssens, van der Gaag, and Rinke de
Wit 2008).
9
In 20.1% of the fully participating households without health insur-
ance, at least one working-age household member is infected with HIV.
Te third health shock is a dummy variable equal to 1 if a working-age house-
hold member died in the 12 months prior to the survey, and 0 otherwise. On aver-
age, 5.1% of the uninsured households experienced such a death in the past year
(table 3.6).
Te fourth health shock is a dummy variable equal to 1 if a working-age house-
hold member was hospitalized for at least three nights in the 12 months prior to
the survey. In more than one-fth of the uninsured households, 21.3%, at least one
working-age person was hospitalized for three nights or more. Tis cuto excludes
the less serious episodes of hospitalization, 62% of which are to give birth. Te
majority of women who give birth in the hospital are discharged within three days,
indicating a birth without complications. Te main reasons to stay in the hospi-
tal for three nights or more are for treatment (44%), surgery (21%), giving birth
(20%), specialist examination (6%), or acute illness or injury (5%).
TABLE 3.6
Health shocks for uninsured and insured households
Uninsured
households
Total
(n=948)
Mean (#)
Quintile
1
(n=207)
Mean
Quintile
2
(n=207)
Mean
Quintile
3
(n=199)
Mean
Quintile
4
(n=161)
Mean
Quintile
5
(n=84)
Mean
Difference across
quintiles (x
2
)
p-value
Weight loss .288 (272) .348 .232 .297 .288 .313 .140
HIV/AIDS .199 (104) .271 .238 .135 .125 .129 .017**
Death .051 (48) .068 .073 .030 .044 .048 .313
Hospitalization .213 (201) .367 .242 .181 .125 .108 .000***
Insured
households
Total
(n=821)
Mean (#)
Q1
(n=31)
Mean
Q2
(n=78)
Mean
Q3
(n=130)
Mean
Q4
(n=197)
Mean
Q5
(n=360)
Mean
Difference
across
quintiles
(x
2
)
p-value
Difference
uninsured
vs. insured
(F-statistic)
p-value
Weight loss .239 (197) .290 .359 .231 .222 .222 .111 .021**
HIV/AIDS .137 (55) .143 .186 .178 .148 .095 .370 .014**
Death .038 (31) .000 .038 .046 .035 .039 .824 .177
Hospitalization .204 (168) .194 .256 .223 .212 .183 .608 .649
*** signifcant at p < 0.01; ** signifcant at p < 0.05.
Note: Information on HIV/AIDS infection is available only for 524 uninsured households and
402 insured households. Information on consumption quintile is missing for 90 uninsured
households and 25 insured households.
Source: Authors calculations.
44 Chapter 3
Te four health shocks are signicantly more likely to occur in the poorest
quintiles than in the middle or highest quintiles, as calculated with one-by-one
F-tests, (except for the rst quintile in the insured sample that includes relatively
few observations) (see table 3.6). Te overall chi-square value across quintiles is
statistically signicant for hospitalization and HIV infection among the uninsured
households. Death and hospitalization are equally common for uninsured and
insured households. But the uninsured households are signicantly more likely to
have a working-age household member who suered weight loss in the past year or
who is HIV infected.
Te outcome and coping variables are all measured at the level of the house-
hold. But the results do not substantially change if we measure income, expendi-
tures, and consumption on a per capita basis (see table A3.1 in the appendix).
Te results of the regressions by insurance status are in table 3.7. Te rst col-
umn conrms the descriptive statistics. Uninsured households that experience a
health shock face signicantly higher out-of-pocket expenditures for health than
do uninsured households without such a shock. Health insurance, by contrast,
appears to protect households from high medical expenses.
Te consequences of weight loss and HIV infection
Losing weight is an important indicator of worsening health status, with poten-
tially far reaching consequences for the aected household. Indeed, weight loss
is the only shock variable associated with both high medical expenditures (col-
umn 1) and substantially lower earned income (column 2). Te lower earnings may
be caused by an ill household members reduced capacity to work. Conversely, it is
possible that the lower levels of earned income are to some extent responsible for
the weight loss. Tat is, the poorest may be most likely to become ill in the rst
place. But adult weight loss is also signicantly correlated with lower labor produc-
tivity in the household, both in the number of working members and in the aver-
age months worked (column 4 and 5). Tis suggests that the health shock reduces a
households income-earning capacity. For households with health insurance, we do
not nd a signicant correlation between weight loss and reduced income.
Households without health insurance seem to have two main strategies to
cope with the combination of high expenditures and low income. Teir unearned
income (column 3) and their use of credit (column 8) are both signicantly higher
than those of households without an adult losing weight. A closer look at the com-
ponents of unearned income shows that informal support, such as assistance from
relatives and friends, helps in coping with health shocks. A second signicant
source of unearned income is maintenance grants.
T
A
B
L
E

3
.
7

T
h
e

e
c
o
n
o
m
i
c

c
o
n
s
e
q
u
e
n
c
e
s

o
f

h
e
a
l
t
h

s
h
o
c
k
s
U
n
i
n
s
u
r
e
d

(
n
=
9
4
8
)
L
n

M
e
d
i
c
a
l

e
x
p
e
n
d
i
t
u
r
e
s
L
n

E
a
r
n
e
d

I
n
c
o
m
e
L
n

U
n
e
a
r
n
e
d

i
n
c
o
m
e
L
a
b
o
r

(
n
u
m
b
e
r

o
f

w
o
r
k
e
r
s

a
s

p
e
r
c
e
n
t
a
g
e

o
f

w
o
r
k
i
n
g
-
a
g
e

h
o
u
s
e
h
o
l
d

m
e
m
b
e
r
s
)
L
a
b
o
r

(
a
v
e
r
a
g
e

n
u
m
b
e
r

o
f

m
o
n
t
h
s

w
o
r
k
e
d
)
L
n

A
n
n
u
a
l

f
o
o
d

c
o
n
s
u
m
p
t
i
o
n
L
n

A
n
n
u
a
l

n
o
n
f
o
o
d

c
o
n
s
u
m
p
t
i
o
n
C
r
e
d
i
t

a
n
d

b
o
r
r
o
w
i
n
g
A
s
s
e
t
s
W
e
i
g
h
t

l
o
s
s
.
7
0
7
[
.
1
9
7
]
*
*
*
.
5
7
2
[
.
2
8
3
]
*
*
.
5
3
0
[
.
2
5
1
]
*
*
.
0
4
9
[
.
0
3
0
]
.
7
1
8
[
.
3
3
4
]
*
*
.
0
0
5
[
.
0
5
1
]
.
0
0
9
[
.
0
7
4
]
.
0
6
3
[
.
0
3
3
]
*
*
.
0
0
5
[
.
0
3
4
]
H
I
V
-
i
n
f
e
c
t
e
d

h
o
u
s
e
h
o
l
d

m
e
m
b
e
r
.
0
8
2
[
.
2
5
9
]
.
1
4
8
[
.
3
2
8
]
.
1
8
1
[
.
3
3
1
]
.
0
4
2
[
.
0
3
9
]
.
4
9
1
[
.
4
1
6
]
.
1
1
7
[
.
0
7
8
]
.
2
7
3
[
.
1
1
2
]
*
*
.
0
9
8
[
.
0
4
2
]
*
*
.
1
4
3
[
.
0
4
8
]
*
*
*
D
e
a
t
h

i
n

t
h
e

h
o
u
s
e
h
o
l
d
.
6
4
2
[
.
3
3
8
]
*
.
2
3
1
[
.
5
1
2
]
.
1
3
0
[
.
4
3
6
]
.
0
0
5
[
.
0
4
4
]
.
0
1
2
[
.
5
1
2
]
.
0
5
7
[
.
1
2
5
]
.
0
3
6
[
.
1
3
9
]
.
0
2
2
[
.
0
5
9
]
.
1
0
0
[
.
0
5
1
]
*
H
o
s
p
i
t
a
l
i
z
a
t
i
o
n

f
o
r

a
t

l
e
a
s
t

t
h
r
e
e

n
i
g
h
t
s
.
9
1
6
[
.
2
1
3
]
*
*
*
.
2
0
8
[
.
2
3
7
]
.
6
1
6
[
.
3
2
8
]
*
.
0
0
7
[
.
0
2
4
]
.
0
4
7
[
.
2
6
3
]
.
0
5
8
[
.
0
5
6
]
.
1
4
6
[
.
0
7
8
]
*
.
0
4
4
[
.
0
3
4
]
.
1
0
5
[
.
0
4
1
]
*
*
I
n
s
u
r
e
d

(
n
=
8
2
1
)
L
n

M
e
d
i
c
a
l

e
x
p
e
n
d
i
t
u
r
e
s
L
n

E
a
r
n
e
d

i
n
c
o
m
e
L
n

U
n
e
a
r
n
e
d

i
n
c
o
m
e
L
a
b
o
r

(
n
u
m
b
e
r

o
f

w
o
r
k
e
r
s

a
s

p
e
r
c
e
n
t
a
g
e

o
f

w
o
r
k
i
n
g
-
a
g
e

h
o
u
s
e
h
o
l
d

m
e
m
b
e
r
s
)
L
a
b
o
r

(
a
v
e
r
a
g
e

n
u
m
b
e
r

o
f

m
o
n
t
h
s

w
o
r
k
e
d
)
L
n

A
n
n
u
a
l

f
o
o
d

c
o
n
s
u
m
p
t
i
o
n
L
n

A
n
n
u
a
l

n
o
n
f
o
o
d

c
o
n
s
u
m
p
t
i
o
n
C
r
e
d
i
t

a
n
d

b
o
r
r
o
w
i
n
g
A
s
s
e
t
s
W
e
i
g
h
t

l
o
s
s
.
3
2
2
[
.
2
5
2
]
.
2
2
5
[
.
2
6
0
]
.
3
7
4
[
.
3
3
2
]
.
0
1
5
[
.
0
2
1
]
.
1
4
8
[
.
3
0
5
]
.
0
3
5
[
.
0
5
3
]
.
0
7
6
[
.
0
7
1
]
.
0
5
3
[
.
0
4
1
]
.
0
8
1
[
.
0
5
1
]
H
I
V
-
i
n
f
e
c
t
e
d

h
o
u
s
e
h
o
l
d

m
e
m
b
e
r
.
6
7
0
[
.
3
8
5
]
*
.
0
0
2
[
.
2
5
6
]
.
4
6
2
[
.
5
5
3
]
.
0
4
6
[
.
0
3
1
]
.
6
2
9
[
.
3
8
7
]
.
0
0
8
[
.
0
7
5
]
.
2
3
9
[
.
1
0
2
]
*
*
.
0
1
7
[
.
0
7
9
]
.
1
6
9
[
.
0
8
4
]
*
*
D
e
a
t
h

i
n

t
h
e

h
o
u
s
e
h
o
l
d
.
4
5
5
[
.
3
5
4
]
.
5
6
2
[
.
5
2
8
]
.
5
8
8
[
.
7
6
6
]
.
0
4
1
[
.
0
4
9
]
.
4
3
1
[
.
6
1
1
]
.
1
0
1
[
.
1
0
9
]
.
4
1
7
[
.
1
1
4
]
*
*
*
.
1
7
6
[
.
1
0
5
]
.
0
5
7
[
.
1
0
7
]
H
o
s
p
i
t
a
l
i
z
a
t
i
o
n

f
o
r

a
t

l
e
a
s
t

t
h
r
e
e

n
i
g
h
t
s
.
2
4
4
[
.
2
3
6
]
.
2
1
2
[
.
2
0
9
]
.
6
5
2
[
.
3
6
3
]
*
.
0
0
5
[
.
0
2
3
]
.
0
2
1
[
.
2
8
3
]
.
0
3
7
[
.
0
5
5
]
.
1
6
1
[
.
0
6
9
]
*
*
.
0
4
0
[
.
0
3
8
]
.
0
5
1
[
.
0
6
1
]
*
*
*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
0
1
;

*
*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
0
5
;

*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
1
0
.
N
o
t
e
:

N
u
m
b
e
r
s

i
n

b
r
a
c
k
e
t
s

a
r
e

r
o
b
u
s
t

s
t
a
n
d
a
r
d

e
r
r
o
r
s

a
d
j
u
s
t
e
d

f
o
r

c
l
u
s
t
e
r
i
n
g

a
t

t
h
e

P
S
U

l
e
v
e
l
.

D
e
p
e
n
d
e
n
t

v
a
r
i
a
b
l
e
s

a
r
e

a
t

h
o
u
s
e
h
o
l
d

l
e
v
e
l

t
o
t
a
l
s
.

R
e
s
u
l
t
s

f
o
r

p
e
r

c
a
p
i
t
a

d
e
p
e
n
d
e
n
t

v
a
r
i
a
b
l
e
s

a
r
e

s
i
m
i
l
a
r
.

T
h
e

r
e
g
r
e
s
s
i
o
n
s

a
l
s
o

i
n
c
l
u
d
e

h
e
a
d

o
f

h
o
u
s
e
h
o
l
d

c
h
a
r
a
c
t
e
r
i
s
t
i
c
s

(
a
g
e
,

a
g
e

s
q
u
a
r
e
d
,

s
e
x
,

e
d
u
c
a
t
i
o
n
)
,

h
o
u
s
e
h
o
l
d

s
i
z
e
,

n
u
m
b
e
r

o
f

c
h
i
l
d
r
e
n
,

a
n
d

n
e
i
g
h
b
o
r
h
o
o
d

f
x
e
d

e
f
f
e
c
t
s
.

T
h
e

c
r
e
d
i
t

a
n
d

b
o
r
r
o
w
i
n
g

c
o
l
u
m
n

g
i
v
e
s

m
a
r
g
i
n
a
l

p
r
o
b
a
b
i
l
i
t
i
e
s

f
r
o
m

a

p
r
o
b
i
t

e
s
t
i
m
a
t
i
o
n

i
n
s
t
e
a
d

o
f

a
n

o
r
d
i
n
a
r
y

l
e
a
s
t

s
q
u
a
r
e
s

r
e
g
r
e
s
s
i
o
n
.
S
o
u
r
c
e
:

A
u
t
h
o
r
s

c
a
l
c
u
l
a
t
i
o
n
s
.
46 Chapter 3
Te results for HIV infection merit further explanation. Medical expenditures
and earned income are not substantially dierent when it comes to having an HIV-
infected household member. What could cause this absence of impact for one of
the most devastating diseases that exists?
A rst explanation is that coverage of antiretroviral treatment is high in
Namibia, with about 66% of eligible antiretroviral treatment patients on treatment
in March 2007.
10
Nonetheless, a further look at the coping strategies of aected
households suggests that the disease is signicantly correlated with lower expen-
ditures for nonfood items (column 7) and declining ownership of assets (column
9). HIV-aected households are also less likely to have borrowed money, poten-
tially due to reduced access to credit. A similar pattern is discernable among the
insured households. In fact, medical expenditures of HIV-infected households with
health insurance are signicantly higher than those without HIV. Tis suggests
that health insurance may increase demand for treatment among the insured but is
not fully covered by private insurance packages.
A second explanation is that HIV infection is not a health shock per se. At an
average incubation period of about eight years, most HIV-positive individuals have
not developed AIDS yet. So they are currently not ill and still able to function
normally for a number of years. Once HIV-infected individuals start developing
AIDS, they will either get treatment and be able to workor not get treatment
and die within one or at most two years. In other words, most of the HIV-aected
households will not yet suer any of the negative health consequence of AIDS.
But at some point an individuals immune system will be damaged to such an
extent that the person develops AIDS. Tis is often accompanied by substantial
weight loss. Indeed, at the individual level the .061 correlation coecient between
HIV status and weight loss is not perfect but is statistically highly signicant
(p-value .003). Of HIV-negative individuals, 13% experienced a drop in weight in
the past 12 months compared with 19% of HIV-positive individuals. Tus, losing
weight partly proxies for a more developed state of HIV/AIDS which will lead to
additional health problems, medical costs, and a decreasing capacity to work. Over
time, more Namibians infected with HIV will develop AIDS. Without treatment,
the consequences for households will be large because of the increasing pressure on
ones own coping strategies and on the capacity of social networks to keep provid-
ing informal assistance.
Coping with a death in the family
A death in the household leads to substantial medical expenditures for unin-
sured households (column 1) but does not aect earned income in the past year
(column 2). Te latter nding cannot be due to the perverse eect that the death
of a household member with a below-average contribution to household income
might actually increase per capita income. We are looking not at per capita house-
hold income but at total household income. Columns 4 and 5 suggest that the lat-
ter nding is also not due to an increase in labor supply of other household mem-
bers to oset the drop in earnings. Te coecients on a death shock in the labor
regressions are small and not statistically signicant. A potential explanation could
be that terminally ill individuals return to their parental home to die. In that case,
the household will report a deceased family member. But earned income is not
aected because prior to his or her terminal stage, the individual did not contribute
to household income either.
We do not nd evidence of increased remittances and other sources of unearned
income for households confronted with a death in the family (column 3)or of
increased use of credit (column 8) to pay for the medical costs or the funeral for
example. Annual food consumption of those with a deceased household member
(column 6) is not substantially dierent from the consumption levels of other
households. By construction, the extrapolated weekly food expenditures do not
capture a drop in consumption half a year earlier if it was followed by a subsequent
recovery before the previous week. We nd substantially higher expenditures on
nonfood items for the insured who experience a death, but not for the uninsured.
It is possible that uninsured households compensate for the medical and death-
related costs by subsequently reducing consumption of other nonfood items. Te
asset score is substantially lower for uninsured households with a death in the fam-
ily but not for the insured (column 9). Tis suggests that selling assets is one way
for uninsured households to cover death-related (medical) expenditures.
Coping with the consequences of hospitalization
Hospitalization results in high medical costs for uninsured households. Earned
income does not appear to be aected by hospitalization, perhaps indicating that
recovery after treatment is swift enough to prevent income from dropping substan-
tially. Tis interpretation is supported by the insignicant coecient on the labor
outcome variable. Overall, hospitalizations are more prevalent among the lower
quintiles (see table 3.6). So it is unlikely that any negative eects of hospitalization
on income are oset by a positive correlation between income and seeking hospital
treatment.
Two ndings stand out. First, both uninsured and insured households show
signicantly higher unearned income if a household member has been hospital-
ized for at least three nights. Further analysis of the underlying components of
48 Chapter 3
unearned income shows that this is related to two main sources. Te rst is help for
medical expenses from relatives, friends, or employers. Te second is social security
(such as three-month unemployment, maternity leave, or a maintenance grant).
Individuals who expect to receive informal assistance from others could be more
likely to become hospitalized. If households with a strong social network are better
able to aord hospitalization, they might be more likely to seek inpatient treatment
when needed.
A second nding is that both annual nonfood consumption and the asset score
are signicantly lower for uninsured households with a hospitalization shock than
for those without one. One way for uninsured households to cope with high hos-
pitalization costs may be to postpone large nonfood expenditures and to sell dura-
bles. Households with health insurance on the other hand report higher nonfood
expenses if one of their members has been hospitalized.
Scope for targeting
Despite the relatively accessible public health care system in Greater Windhoek,
households without health insurance suer from large medical expenditures after
the death, hospitalization, or weight loss of an adult household member. Although
gifts and support from others help them overcome part of the nancial burden,
ndings indicate that households without health insurance must resort to addi-
tional coping strategies, such as selling assets, reducing nonfood consumption, or
taking out loans. For households with access to private health insurance, the eco-
nomic consequences of health shocks are far less pronounced.
Te results do not show substantial eects related to HIV infection, but for
advanced cases the consequences can be serious. Losing weight is in part a proxy
for a more advanced state of AIDS if the patient is not receiving antiretroviral
therapy. Weight loss is not only associated with high medical costs but also with
substantially lower labor productivity and earned income. Remittances from others
are signicant but not sucient to compensate for all consequences of the health
shock, as the higher use of credit among aected households suggests.
Tis nding is particularly worrisome in view of the high HIV prevalence rates
in Namibia. As more infected people without insurance develop AIDS over time,
the public sector and social support networks will come under increasing pres-
sure. Table 3.8 shows the HIV infection rates of working-age adults across socio-
economic categories. It should be interpreted with some caution, because the sam-
ple is not representative for Greater Windhoek due to the loss of respondents. But
it clearly shows four patterns. First, prevalence rates are signicantly higher among
the poorest quintiles. Second, they are signicantly higher for individuals who have
TABLE 3.8
HIV infection rates among working-age adults in biomedical Republic of
Namibia Okambilimbili Survey sample 2006
Number of observations HIV infection rate (%)
Consumption quintile
1 (poorest) 379 13.2
2 366 12.6
3 370 8.1
4 385 6.8
5 (richest) 306 6.2
Education level
No education 129 15.5
Primary incomplete 208 13.9
Primary complete 123 13.0
Secondary incomplete 806 10.3
Secondary complete 444 5.2
Higher education 182 6.0
Employment status
Employed 1,053 11.1
Unemployed 838 7.8
Industry
Manufacturing 37 8.1
Construction 63 15.9
Retail and accommodation 106 11.3
Transport and storage 40 2.5
Services 345 9.6
Government and defense 135 18.5
Education 50 8.0
Health 35 5.7
Other 239 11.3
Individual insurance status
Insured 553 8.1
Uninsured 1,342 10.2
Total 1,895 9.6
Note: Includes only working-age adults ages 1565 with reliable HIV results.
50 Chapter 3
not completed secondary education, especially for those without education at all.
Tird, HIV infection is more prevalent among the employed, especially among
workers in government and defense. Fourth, HIV infection is higher among indi-
viduals without health insurance.
A high percentage of households involved in government or defense have at
least one household member covered by health insurance (see table 3.1). But other
sectors severely aected by HIV, such as the construction, retail, and accommoda-
tion sectors, show some of the lowest insurance coverage rates. Moreover, the poor-
est and least educated are most likely to be infected and least likely to be insured.
For the uninsured the economic consequences of arriving at a more developed state
of HIV/AIDS are potentially large and reach beyond the aected household into its
social support network.
Our ndings on the inequitable impacts of health shocks on the uninsured
in Namibia are particularly pertinent because of the recent introduction of low
cost subsidized health insurance products with an emphasis on HIV/AIDS treat-
ment in Greater Windhoek. Tese products are among a set of programs initi-
ated by two Dutch organizations, PharmAccess Foundation (PharmAccess) and
the Health Insurance Fund, which currently provide low cost voluntary health
insurance products for low-income workers in Africa using private sector insurance
companies and health maintenance organizations.
A 2004 pilot program introduced this concept of health nancing in Greater
Windhoek. Te Okambilimbili (buttery) project focused on supporting Diamond
Health Services, a newcomer in the Namibian private health care industry. Tis
network of service providers oered an aordable primary health care product that
included HIV/AIDS counseling and treatment (highly active antiretroviral treat-
ment, HAART) and the treatment of tuberculosis and malaria to the uninsured
employed population. Te program emphasized selling insurance through employ-
ers rather than to individual workers. Initially, the product was meant to be subsi-
dized through PharmAccess, so the costs for employees would be kept low. In the
end, however, premiums were subsidized up to 50% by employers.
When it was recognized that engaging the wider medical aid fund industry was
needed to scale up access to HIV/AIDS treatment, PharmAccess initiated negotia-
tions with the private medical aid fund industry. By the end of 2005 three aord-
able health care packages for low- and middle-income employees were available
on the Namibian market: primary health care, HAART, and basic hospitaliza-
tion. New employer-subsidized insurance products were launched in May 2006. To
share the risk for the insurance industry due to the high HIV/AIDS prevalence in
Namibia, PharmAccess supported a risk equalization fund, HEALTH-IS-VITAL,
which became the key part of the PharmAccess program. In this fund the employer-
based insured and the previously uninsured groups contribute monthly premiums
to a risk pool with a dened set of HIV/AIDS treatment benets. Tus they share
the nancial risk of high medical costs related to HIV/AIDS treatment. In addi-
tion to subsidizing the risk equalization fund, part of the Okambilimbili project
budget has been allocated to a treatment literacy campaign, which enables project
partners to focus on awareness raising, treatment education, advocacy, support,
and information sharing.
Tere is substantial demand for the new low cost insurance schemes. Of the
25 companies approached to participate in the new products, 24 agreed. All their
employees have been tested for HIV, and HIV/AIDS treatment and counseling are
included in all new types of low cost insurance packages. Even though the majority
of individual participants belong to the third and fourth income quintiles, substan-
tial subsidies remain necessary to keep the schemes aordable. Te participating
companies contribute to the costs at a 50% employer subsidy of the premium for
the employees. At present, more than 30,000 people are beneting from the new
insurance products.
Our ndings suggest that those lacking private health insurance are substan-
tially aected by health-related shocks when they are forced to resort to coping
strategies, which may leave them with a weak asset base. Te initial success of this
low-cost health insurance program provides some encouragement for the protec-
tion of more individuals from the impacts of health shocks, such as those related
to HIV/AIDS.
Conclusions
Despite widely available and relatively well nanced public care, the economic
consequences of health shocks can be severe for uninsured households who resort
to a variety of coping strategies to deal with high medical expenses and reduc-
tions in income, such as selling assets or taking up credit. HIV infection is not
directly related to severe negative outcomes, but weight loss, a known correlate
with advanced AIDS, is. According to our ndings, the poor in Greater Windhoek
are signicantly more likely to be HIV infected and less likely to be covered by
health insurance, resulting in signicant exposure to health and nancial risks.
One alternative to address poor health outcomes and catastrophic health
expenditures in developing countries is to invest more in the public health care
system. But we nd here that the strong public health care system in Namibia still
leaves the poor and uninsured unprotected from health shocks. Another alterna-
tive is to invest in private health insurance. To date, there is a proven market for
52 Chapter 3
the new low income insurance products being oered in Namibia for those with a
regular income. In the program, both employers and employees are willing to par-
ticipate in the new insurance schemes. Te employees mostly belong to the third
and fourth income quintiles, and so do not represent the poor. But these house-
holds are not rich either, given the high inequality in Namibia.
It is unlikely that the current insurance packages combined with low levels
of subsidy can reach the two lowest quintiles. So, to penetrate these groups with
insurance, the way forward would be to design tailor-made health insurance prod-
ucts with substantially higher subsidies. Financing for such products would most
likely come in a combination of donor and government funds.
We could not estimate the actual impacts of the newly introduced Namibian
health insurance schemes due to the cross-sectional data, but data from follow-up
surveys will allow for such analysis. For now, however, we can conclude that there
is a strong correlation between being uninsured and the negative consequences of
health shocks. In light of these ndings, particularly in a country such as Namibia
with a fairly well developed public health care system, we should continue a seri-
ous evidence-based debate on private health insurance as a potential mechanism to
provide not only nancing for the increasing health care demands facing Africa,
but also protection against the signicant negative economic consequences result-
ing directly from health shocks.
T
A
B
L
E

A
3
.
1

E
c
o
n
o
m
i
c

o
u
t
c
o
m
e

a
n
d

c
o
p
i
n
g

v
a
r
i
a
b
l
e
s

(
u
n
i
n
s
u
r
e
d

h
o
u
s
e
h
o
l
d
s
)
V
a
r
i
a
b
l
e

(
d
e
f
n
i
t
i
o
n
)
N
u
m
b
e
r

o
f

o
b
s
e
r
v
a
t
i
o
n
s
M
e
a
n
S
t
a
n
d
a
r
d

d
e
v
i
a
t
i
o
n
M
i
n
i
m
u
m
M
a
x
i
m
u
m
M
e
a
n

u
n
i
n
s
u
r
e
d
M
e
a
n

i
n
s
u
r
e
d
D
i
f
f
e
r
e
n
c
e

(
p
-
v
a
l
u
e
)
H
e
a
l
t
h

e
x
p
e
n
d
i
t
u
r
e
s

=
N
a
t
u
r
a
l

l
o
g
a
r
i
t
h
m

o
f

t
o
t
a
l

m
e
d
i
c
a
l

e
x
p
e
n
d
i
t
u
r
e
s

b
y

t
h
e

h
o
u
s
e
h
o
l
d

i
n

t
h
e

1
2

m
o
n
t
h
s

p
r
i
o
r

t
o

t
h
e

s
u
r
v
e
y
,

e
x
c
l
u
d
i
n
g

t
h
e

p
r
e
m
i
a

f
o
r

h
e
a
l
t
h

i
n
s
u
r
a
n
c
e
.
1
,
7
4
9
2
.
9
4
2
.
8
7
0
1
1
.
9
8
2
.
4
4
3
.
5
2
.
0
0
0
*
*
*
E
a
r
n
e
d

I
n
c
o
m
e

=
N
a
t
u
r
a
l

l
o
g
a
r
i
t
h
m

o
f

t
o
t
a
l

i
n
c
o
m
e

e
a
r
n
e
d

i
n

t
h
e

1
2

m
o
n
t
h
s

p
r
i
o
r

t
o

t
h
e

s
u
r
v
e
y

b
y

a
l
l

h
o
u
s
e
h
o
l
d

m
e
m
b
e
r
s
,

i
n
c
l
u
d
i
n
g

i
n
c
o
m
e

f
r
o
m

w
a
g
e
,

f
r
o
m

o
w
n

b
u
s
i
n
e
s
s
,

a
n
d

f
r
o
m

a
g
r
i
c
u
l
t
u
r
e
.
1
,
7
7
0
9
.
0
6
3
.
8
5
0
1
4
.
4
0
7
.
9
4
1
0
.
3
7
.
0
0
0
*
*
*
U
n
e
a
r
n
e
d

i
n
c
o
m
e

=
N
a
t
u
r
a
l

l
o
g
a
r
i
t
h
m

o
f

t
o
t
a
l

u
n
e
a
r
n
e
d

i
n
c
o
m
e

i
n

t
h
e

1
2

m
o
n
t
h
s

p
r
i
o
r

t
o

t
h
e

s
u
r
v
e
y
,

i
n
c
l
u
d
i
n
g

g
i
f
t
s

a
n
d

r
e
m
i
t
t
a
n
c
e
s

f
r
o
m

r
e
l
a
t
i
v
e
s
,

f
r
i
e
n
d
s
,

e
m
p
l
o
y
e
r
s
,

a
n
d

o
t
h
e
r
s
;

o
l
d
-
a
g
e

p
e
n
s
i
o
n
s
;

d
i
s
a
b
i
l
i
t
y

g
r
a
n
t
s
;

m
a
i
n
t
e
n
a
n
c
e

g
r
a
n
t
s
;

w
o
r
k
m
a
n
s

c
o
m
p
e
n
s
a
t
i
o
n
;

s
o
c
i
a
l

s
e
c
u
r
i
t
y

(
t
h
r
e
e
-
m
o
n
t
h

e
m
p
l
o
y
m
e
n
t
,

f
u
n
e
r
a
l
,

m
a
t
e
r
n
i
t
y

l
e
a
v
e
)
;

i
n
c
o
m
e

f
r
o
m

r
e
n
t
;

a
n
d

a
n
y

o
t
h
e
r

s
o
u
r
c
e
s

o
f

u
n
e
a
r
n
e
d

i
n
c
o
m
e
.
1
,
7
5
4
3
.
0
4
3
.
9
4
0
1
2
.
5
2
3
.
1
4
2
.
9
4
.
2
9
3
F
o
o
d

c
o
n
s
u
m
p
t
i
o
n

=
N
a
t
u
r
a
l

l
o
g
a
r
i
t
h
m

o
f

t
o
t
a
l

a
n
n
u
a
l

h
o
u
s
e
h
o
l
d

c
o
n
s
u
m
p
t
i
o
n

(
p
u
r
c
h
a
s
e
d
,

r
e
c
e
i
v
e
d
,

p
r
o
d
u
c
e
d
)

o
f

f
o
o
d

i
t
e
m
s

i
n

t
h
e

1
2

m
o
n
t
h
s

p
r
i
o
r

t
o

t
h
e

s
u
r
v
e
y
,

t
h
r
o
u
g
h

a
n

e
x
t
r
a
p
o
l
a
t
i
o
n

o
f

f
o
o
d

c
o
n
s
u
m
p
t
i
o
n

i
n

t
h
e

p
a
s
t

w
e
e
k

t
o

t
h
e

e
n
t
i
r
e

y
e
a
r
.
1
,
7
4
7
9
.
3
5
0
.
8
2
6
.
4
7
1
2
.
3
4
9
.
0
5
9
.
6
9
.
0
0
0
*
*
*
(
c
o
n
t
i
n
u
e
d
)
54 Chapter 3
V
a
r
i
a
b
l
e

(
d
e
f
n
i
t
i
o
n
)
N
u
m
b
e
r

o
f

o
b
s
e
r
v
a
t
i
o
n
s
M
e
a
n
S
t
a
n
d
a
r
d

d
e
v
i
a
t
i
o
n
M
i
n
i
m
u
m
M
a
x
i
m
u
m
M
e
a
n

u
n
i
n
s
u
r
e
d
M
e
a
n

i
n
s
u
r
e
d
D
i
f
f
e
r
e
n
c
e

(
p
-
v
a
l
u
e
)
L
n

N
o
n
f
o
o
d

c
o
n
s
u
m
p
t
i
o
n

=
N
a
t
u
r
a
l

l
o
g
a
r
i
t
h
m

o
f

t
o
t
a
l

a
n
n
u
a
l

h
o
u
s
e
h
o
l
d

c
o
n
s
u
m
p
t
i
o
n

(
p
u
r
c
h
a
s
e
d
,

r
e
c
e
i
v
e
d
,

p
r
o
d
u
c
e
d
)

o
f

f
o
o
d

a
n
d

o
f

n
o
n
f
o
o
d

(
n
o
n
m
e
d
i
c
a
l
)

i
t
e
m
s

i
n

t
h
e

1
2

m
o
n
t
h
s

p
r
i
o
r

t
o

t
h
e

s
u
r
v
e
y
,

t
h
r
o
u
g
h

a
n

e
x
t
r
a
p
o
l
a
t
i
o
n

o
f

f
o
o
d

c
o
n
s
u
m
p
t
i
o
n

i
n

t
h
e

p
a
s
t

w
e
e
k

a
n
d

m
o
n
t
h
l
y

n
o
n
f
o
o
d

e
x
p
e
n
d
i
t
u
r
e
s

i
n

t
h
e

p
a
s
t

m
o
n
t
h

t
o

t
h
e

e
n
t
i
r
e

y
e
a
r
.
1
,
6
5
7
1
0
.
3
5
1
.
3
2
5
.
6
9
1
5
.
1
1
9
.
6
2
1
1
.
1
4
.
0
0
0
*
*
*
L
a
b
o
r

s
u
p
p
l
y
a

=
T
o
t
a
l

n
u
m
b
e
r

o
f

m
o
n
t
h
s

t
h
a
t

h
o
u
s
e
h
o
l
d

m
e
m
b
e
r
s

w
o
r
k
e
d

f
o
r

i
n
c
o
m
e

i
n

t
h
e

1
2

m
o
n
t
h
s

p
r
i
o
r

t
o

t
h
e

s
u
r
v
e
y

d
i
v
i
d
e
d

b
y

t
h
e

t
o
t
a
l

n
u
m
b
e
r

o
f

w
o
r
k
i
n
g
-
a
g
e

h
o
u
s
e
h
o
l
d

m
e
m
b
e
r
s
.
1
,
7
4
2
6
.
5
9
4
.
0
1
0
3
2
5
.
6
9
7
.
6
4
.
0
0
0
*
*
*
A
s
s
e
t
s

=
T
h
e

f
r
s
t

s
c
o
r
e

f
r
o
m

a

f
a
c
t
o
r

a
n
a
l
y
s
i
s

o
f

2
8

t
y
p
e
s

o
f

a
s
s
e
t
s
.
1
,
7
6
0
0
.
0
0
0
.
9
7
1
.
2
2
3
.
8
1
0
.
5
0
0
.
5
8
.
0
0
0
*
*
*
C
r
e
d
i
t

=
D
u
m
m
y

v
a
r
i
a
b
l
e

e
q
u
a
l

t
o

1

i
f

i
n

t
h
e

1
2

m
o
n
t
h
s

p
r
i
o
r

t
o

t
h
e

s
u
r
v
e
y

a
n
y

h
o
u
s
e
h
o
l
d

m
e
m
b
e
r

b
o
u
g
h
t

g
o
o
d
s

o
n

c
r
e
d
i
t

o
r

g
o
t

a

c
a
s
h

l
o
a
n

f
r
o
m

a
n
y

i
n
s
t
i
t
u
t
i
o
n

o
r

a
n
y

p
e
r
s
o
n
,

a
n
d

0

o
t
h
e
r
w
i
s
e
.
1
,
7
7
0
.
3
5
3
.
4
7
8
0
1
.
2
4
1
.
4
8
4
.
0
0
0
*
*
*
*
*
*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
0
1
.
a
.

L
a
b
o
r

s
u
p
p
l
y

o
f

c
h
i
l
d
r
e
n

u
n
d
e
r

a
g
e

1
5

i
s

n
o
t

i
n
c
l
u
d
e
d

a
s

a

p
o
t
e
n
t
i
a
l

c
o
p
i
n
g

v
a
r
i
a
b
l
e

b
e
c
a
u
s
e

t
h
e

i
n
c
i
d
e
n
c
e

o
f

c
h
i
l
d

l
a
b
o
r

i
s

e
x
t
r
e
m
e
l
y

l
o
w

i
n

G
r
e
a
t
e
r

W
i
n
d
h
o
e
k
.
S
o
u
r
c
e
:

A
u
t
h
o
r
s
.
T
A
B
L
E

A
3
.
1

(
c
o
n
t
i
n
u
e
d
)

E
c
o
n
o
m
i
c

o
u
t
c
o
m
e

a
n
d

c
o
p
i
n
g

v
a
r
i
a
b
l
e
s

(
u
n
i
n
s
u
r
e
d

h
o
u
s
e
h
o
l
d
s
)
T
A
B
L
E

A
3
.
2

C
o
n
s
e
q
u
e
n
c
e
s

o
f

h
e
a
l
t
h

s
h
o
c
k
s

f
o
r

u
n
i
n
s
u
r
e
d

a
n
d

i
n
s
u
r
e
d

h
o
u
s
e
h
o
l
d
s

i
n
c
l
u
d
i
n
g

i
n
c
o
m
e

q
u
i
n
t
i
l
e
s

i
n

r
e
g
r
e
s
s
i
o
n
U
n
i
n
s
u
r
e
d

(
n
=
9
4
8
)
L
n

M
e
d
i
c
a
l

e
x
p
e
n
d
i
t
u
r
e
s
L
n

E
a
r
n
e
d

I
n
c
o
m
e
L
n

U
n
e
a
r
n
e
d

i
n
c
o
m
e
L
a
b
o
r

(
n
u
m
b
e
r

o
f

w
o
r
k
e
r
s

a
s

p
e
r
c
e
n
t
a
g
e

o
f

w
o
r
k
i
n
g
-
a
g
e

h
o
u
s
e
h
o
l
d

m
e
m
b
e
r
s
)
L
a
b
o
r

(
a
v
e
r
a
g
e

n
u
m
b
e
r

o
f

m
o
n
t
h
s

w
o
r
k
e
d
)
L
n

A
n
n
u
a
l

f
o
o
d

c
o
n
s
u
m
p
t
i
o
n
L
n

A
n
n
u
a
l

n
o
n
f
o
o
d

c
o
n
s
u
m
p
t
i
o
n
C
r
e
d
i
t

a
n
d

b
o
r
r
o
w
i
n
g
A
s
s
e
t
s
W
e
i
g
h
t

l
o
s
s
.
7
7
5
[
.
2
0
2
]
*
*
*
.
5
4
4
[
.
2
8
1
]
*
.
6
3
2
[
.
2
6
2
]
*
*
.
0
4
9
[
.
0
3
0
]
.
7
8
1
[
.
3
1
9
]
*
*
.
0
2
1
[
.
0
3
9
]
.
0
3
5
[
.
0
4
7
]
.
0
6
7
[
.
0
3
5
]
*
*
.
0
1
0
[
.
0
3
2
]
H
I
V
-
i
n
f
e
c
t
e
d

h
o
u
s
e
h
o
l
d

m
e
m
b
e
r
.
0
0
5
[
.
2
7
5
]
.
0
9
1
[
.
3
3
0
]
.
3
6
7
[
.
3
3
8
]
.
0
4
0
[
.
0
3
7
]
.
4
6
6
[
.
3
9
4
]
.
0
2
3
[
.
0
5
7
]
.
1
7
5
[
.
0
7
9
]
*
*
.
0
9
5
[
.
0
4
1
]
*
*
.
1
1
3
[
.
0
4
6
]
*
*
D
e
a
t
h

i
n

t
h
e

h
o
u
s
e
h
o
l
d
.
5
8
0
[
.
3
1
5
]
*
.
0
1
7
[
.
5
4
7
]
.
0
9
5
[
.
4
4
0
]
.
0
0
8
[
.
0
4
8
]
.
0
6
0
[
.
5
3
3
]
.
0
2
9
[
.
0
9
5
]
.
0
1
7
[
.
0
8
0
]
.
0
3
0
[
.
0
6
1
]
.
0
8
8
[
.
0
4
6
]
*
H
o
s
p
i
t
a
l
i
z
a
t
i
o
n

f
o
r

a
t

l
e
a
s
t

t
h
r
e
e

n
i
g
h
t
s
1
.
0
3
4
[
.
1
9
8
]
*
*
*
.
1
8
4
[
.
2
4
2
]
.
5
8
6
[
.
3
3
0
]
*
.
0
0
7
[
.
0
2
4
]
.
0
3
1
[
.
2
6
7
]
.
0
2
0
[
.
0
5
0
]
.
0
0
6
[
.
0
5
0
]
.
0
3
5
[
.
0
3
7
]
.
0
7
2
[
.
0
3
9
]
*
I
n
s
u
r
e
d

(
n
=
8
2
1
)
L
n

M
e
d
i
c
a
l

e
x
p
e
n
d
i
t
u
r
e
s
L
n

E
a
r
n
e
d

I
n
c
o
m
e
L
n

U
n
e
a
r
n
e
d

i
n
c
o
m
e
L
a
b
o
r

(
n
u
m
b
e
r

o
f

w
o
r
k
e
r
s

a
s

p
e
r
c
e
n
t
a
g
e

o
f

w
o
r
k
i
n
g
-
a
g
e

h
o
u
s
e
h
o
l
d

m
e
m
b
e
r
s
)
L
a
b
o
r

(
a
v
e
r
a
g
e

n
u
m
b
e
r

o
f

m
o
n
t
h
s

w
o
r
k
e
d
)
L
n

A
n
n
u
a
l

f
o
o
d

c
o
n
s
u
m
p
t
i
o
n
L
n

A
n
n
u
a
l

n
o
n
f
o
o
d

c
o
n
s
u
m
p
t
i
o
n
C
r
e
d
i
t

a
n
d

b
o
r
r
o
w
i
n
g
A
s
s
e
t
s
W
e
i
g
h
t

l
o
s
s
.
3
7
9
[
.
2
3
8
]
.
1
9
6
[
.
2
3
4
]
.
4
2
9
[
.
3
4
1
]
.
0
1
5
[
.
0
2
1
]
.
1
4
9
[
.
3
0
5
]
.
0
3
0
[
.
0
4
3
]
.
0
7
1
[
.
0
4
3
]
.
0
5
3
[
.
0
4
1
]
.
0
8
3
[
.
0
4
7
]
*
H
I
V
-
i
n
f
e
c
t
e
d

h
o
u
s
e
h
o
l
d

m
e
m
b
e
r
.
7
5
6
[
.
4
0
9
]
*
.
0
5
6
[
.
2
8
3
]
.
3
4
4
[
.
5
5
6
]
.
0
4
4
[
.
0
3
1
]
.
5
9
8
[
.
3
9
3
]
.
0
0
9
[
.
0
6
7
]
.
2
0
4
[
.
0
6
8
]
*
*
*
.
0
3
0
[
.
0
8
3
]
.
1
4
0
[
.
0
8
0
]
*
D
e
a
t
h

i
n

t
h
e

h
o
u
s
e
h
o
l
d
.
1
8
0
[
.
3
4
6
]
.
3
8
1
[
.
5
2
2
]
.
2
7
9
[
.
7
1
0
]
.
0
2
3
[
0
5
1
]
.
2
2
0
[
.
6
3
3
]
.
1
9
7
[
.
1
1
1
]
*
.
2
3
4
[
.
1
0
9
]
*
*
.
1
4
7
[
.
1
0
8
]
.
0
4
8
[
.
1
0
1
]
H
o
s
p
i
t
a
l
i
z
a
t
i
o
n

f
o
r

a
t

l
e
a
s
t

t
h
r
e
e

n
i
g
h
t
s
.
1
3
2
[
.
2
2
4
]
.
3
4
1
[
.
2
0
9
]
.
7
1
3
[
.
3
6
7
]
*
.
0
0
3
[
.
0
2
3
]
.
1
0
1
[
.
2
8
0
]
.
0
1
6
[
.
0
4
3
]
.
0
5
1
[
.
0
4
8
]
*
.
0
3
8
[
.
0
3
9
]
.
0
0
9
[
.
0
5
5
]
*
*
*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
0
1
;

*
*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
0
5
;

*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
1
0
.
N
o
t
e
:

N
u
m
b
e
r
s

i
n

b
r
a
c
k
e
t
s

a
r
e

r
o
b
u
s
t

s
t
a
n
d
a
r
d

e
r
r
o
r
s

a
d
j
u
s
t
e
d

f
o
r

c
l
u
s
t
e
r
i
n
g

a
t

t
h
e

P
S
U

l
e
v
e
l
.

D
e
p
e
n
d
e
n
t

v
a
r
i
a
b
l
e
s

a
r
e

a
t

h
o
u
s
e
h
o
l
d

l
e
v
e
l

t
o
t
a
l
s
.

R
e
s
u
l
t
s

f
o
r

p
e
r

c
a
p
i
t
a

d
e
p
e
n
d
e
n
t

v
a
r
i
a
b
l
e
s

a
r
e

s
i
m
i
l
a
r
.

T
h
e

r
e
g
r
e
s
s
i
o
n
s

a
l
s
o

i
n
c
l
u
d
e

h
e
a
d

o
f

h
o
u
s
e
h
o
l
d

c
h
a
r
a
c
t
e
r
i
s
t
i
c
s

(
a
g
e
,

a
g
e

s
q
u
a
r
e
d
,

s
e
x
,

e
d
u
c
a
t
i
o
n
)
,

h
o
u
s
e
h
o
l
d

s
i
z
e
,

n
u
m
b
e
r

o
f

c
h
i
l
d
r
e
n
,

a
n
d

n
e
i
g
h
b
o
r
h
o
o
d

f
x
e
d

e
f
f
e
c
t
s
.

T
h
e

c
r
e
d
i
t

a
n
d

b
o
r
r
o
w
i
n
g

c
o
l
u
m
n

g
i
v
e
s

m
a
r
g
i
n
a
l

p
r
o
b
a
b
i
l
i
t
i
e
s

f
r
o
m

a

p
r
o
b
i
t

e
s
t
i
m
a
t
i
o
n

i
n
s
t
e
a
d

o
f

a
n

o
r
d
i
n
a
r
y

l
e
a
s
t

s
q
u
a
r
e
s

r
e
g
r
e
s
s
i
o
n
.
S
o
u
r
c
e
:

A
u
t
h
o
r
s

c
a
l
c
u
l
a
t
i
o
n
s
.
56 Chapter 3
Notes
1. Te authors thank Adam Wagsta, Robert Sparrow, Aparnaa Somanathan, Ingrid
de Beer, and Michael Grimm for comments and suggestions on earlier drafts of this
chapter.
2. WHO 2004.
3. Feeley and others 2006
4. We also found a signicant dierence between the insured and uninsured for each age
group, indicating that age is not the determining factor in this dierence.
5. We would have liked to examine dierences across consumption quintiles also, but
frequencies at that level of disaggregation are so low that analysis of that data would
not be robust.
6. We do not have data on savings. Nor do we have information on whether a household
decided to forgo care in relation to the specic health shocks that we examine here.
7. Our results do not substantially change if we also include income quintile as a control
variable.
8. See, for example, Wagsta (2007) and references therein.
9. To calculate prevalence rates adjusted for bias due to nonresponse, Janssens, van der
Gaag, and Rinke de Wit (2008) use a Heckman selection model. Te explanatory vari-
ables also include a substantial number of biological markers for HIV infection such as
coughing, tuberculosis, and sexually transmitted diseases, as well as attitudes toward
and knowledge of HIV/AIDS.
10. UNAIDS 2008.
11. Labor supply of children under age 15 is not included as a potential coping variable
because the incidence of child labor is extremely low in Greater Windhoek.
References
Beegle, K., J. de Weerdt, and S. Dercon. 2008. Adult Mortality and Consumption Growth in the
Age of HIV/AIDS. Economic Development and Cultural Change 56 (2): 299326.
Ekman, B. 2004. Community-based Health Insurance in Low-Income Countries: A Systematic
Review of the Evidence. Health Policy and Planning 19 (5): 24970.
Ekman, B., N. T. Liem, H. A. Duc, and H. Axelson. 2008. Health Insurance Reform in Vietnam:
A Review of Recent Developments and Future Challenges. Health Policy and Planning 23 (4):
25263.
Feeley, F., I. de Beer, T. Rinke de Wit, and J. van der Gaag. 2006. Te Health Insurance Industry
in Namibia. Baseline Report. Boston University, Boston, MA, and Amsterdam Institute for
International Development and PharmAccess Foundation, Amsterdam.
Fox, M. P., S. Rosen, W. B. Macleod, M. Wasunna, M. Bii, G. Foglia, and J.L. Simon. 2004. Te
Impact of HIV/AIDS on Labour Productivity in Kenya. Tropical Medicine and International
Health 9 (3): 31824.
Gaag, van der J., and V. Stimac. 2008. Towards a New Paradigm for Health Sector Development.
AIID Discussion Series 08-02/1. Amsterdam Institute for International Development, Amsterdam.
Gertler, P., and J. Gruber. 2002. Insuring Consumption against Illness. American Economic
Review 92 (1): 5170.
Janssens, W., T. Rinke de Wit, and J. van der Gaag. 2007. Bio-Medical Baseline Report 2006
2007: Okambilimbili Health Insurance Project in Windhoek, Namibia. Amsterdam Institute
for International Development and PharmAccess Foundation, Amsterdam.
Janssens, W., J. van der Gaag, and T. Rinke de Wit. 2008. Pitfalls in the Estimation of HIV
Prevalence Rates. AIID Research Series 08-03. Amsterdam Institute for International Devel-
opment, Amsterdam.
Sekhri, N., and W. Savedo. 2005. Private Health Insurance: Implications for Developing Coun-
tries. Bulletin of the World Health Organization 83 (2): 12734.
Tirumurthy, H., J. Gra Zivin, and M. Goldstein. 2005. Te Economic Impact of AIDS Treat-
ment: Labor Supply in Western Kenya. NBER Working Paper 11871. Cambridge, MA: National
Bureau of Economic Research.
UNAIDS (Joint United Nations Programme on HIV/AIDS). 2008. Epidemiological Fact Sheets
on HIV and AIDS: Namibia, 2008 Update. Joint United Nations Programme on HIV/AIDS,
Geneva.
Wagsta, A. 2007. Te Economic Consequences of Health Shocks: Evidence from Vietnam.
Health Economics 26 (1): 82100.
Wagsta, A., and M. Lindelow. 2008. Can Insurance Increase Financial Risk? Te Curious Case
of Health Insurance in China. Health Economics 27 (4): 9901005.
WHO (World Health Organization). 2004. WHO Country Cooperation Strategy, Republic of
Namibia, 20042007. World Health Organization, Geneva.
. 2005. Namibia Country Prole: Overview of Malaria Control Activities and Pro-
gramme Progress. World Health Organization, Geneva.
. 2008a. TB Country Prole: Namibia. World Health Organization, Geneva.
. 2008b. World Health Statistics 2008. Geneva: World Health Organization.
58
Ghanas National Health
Insurance Scheme
Slavea Chankova, Chris Atim, and Laurel Hatt
C
h
a
p
t
e
r

4
In 2003 Ghana introduced a National Health Insurance Scheme
(NHIS) that aimed to cover the entire population with aordable access
to basic health services within ve years. Tis chapter provides an over-
view and analysis of the evolution of the NHIS in the rst years of its
operation, and the results from an impact evaluation on the eect the
NHIS has had on use and out-of-pocket expenditures for health care.
Te evaluation was conducted in two districts in Ghana, using a pre-
post evaluation design.
In its rst three years of operation, the NHIS caused an increase
in the use of curative health care services and improved nancial pro-
tection against out-of-pocket expenditures for health care. It did not
increase the use of maternal health care, which remains an area where
nonnancial barriers to access may overshadow the eects of increased
nancial protection. High population coverage has ensured better access
to health services for the majority of Ghanas people, but the NHIS
has not achieved equitable enrollment. Better targeting for poor people
needs to be developed to achieve 100% coverage of the population.
Tree key factors, taken together, threaten the nancial sustain-
ability of the scheme over time: rapidly rising enrollment, the gener-
ous benet package, and a fairly constant insurance revenue base. Te
challenges to sustainability identied in this study may threaten the
successes achieved in the early years of implementation and need to be
addressed without delay.
Ghanas National Health Insurance Scheme 59
Ghanas health system
Ghanas health prole is characteristic of most low-income Sub-Saharan African
countries. Communicable diseases still constitute the major causes of morbidity
and mortality. Malaria accounts for 40% of outpatient visits and has a high mor-
tality rate (13%), aecting mostly young children. Other diseases in the top 10
most common causes of death include respiratory tract infections, skin disease and
ulcers, diarrheal diseases, anemia, and hypertension. Pregnancy and related com-
plications are also among the top 10 and, with yellow fever and meningococcal
meningitis, are major public health concerns (Ghana Health Service 2007).
Ghanas Ministry of Health leads the health sector and is responsible for policy
development, planning, donor coordination, and resource mobilization. Te health
system was restructured beginning in 1993, with an emphasis on decentralizing
from the regional and national levels to district administrations and district health
management teams. Te Ghana Health Service, an autonomous executive agency
under the ministry responsible for implementation of national policies and service
delivery, was established in 1996 as part of these reforms. It is organized in ve
levels: national, regional, district, subdistrict, and community.
Services are available in the public, private (for-prot, mission, and nonprot),
and informal sectors. Private providersa coalition of nongovernmental organiza-
tions, the Christian Health Association of Ghana, Catholic mission hospitals, and
private for-prot providersaccount for 40% of patient care (WHO 2008).
Health care nancing
After Ghana won independence from colonial rule in 1957, its new government
was committed to a welfare state system that included free health care for all.
User fees for health services were low and not aimed at cost recovery: nominal fees
of 20 pence for hospital visits had existed before independence and continued to be
charged thereafter (Dzakpallah 1988). But the general thrust of government policy
was for equitable social development, manifested in a policy to make health care
easily accessible to all at the point of use.
In 1969 the rst post-independence government attempted to institute partial
cost recovery in health and education (Dzakpallah 1988). Such fees quickly made
the government unpopular, ushering in conditions that led to its removal from
oce after only a few years in power. Tis experience also led subsequent govern-
ments to shy away from any meaningful health nancing reforms, including alter-
native ways of nancing health, such as insurance.
During the 1970s health facilities and services entered a long period of decline.
Inadequate resources were allocated to rehabilitate existing facilities falling into
60 Chapter 4
disrepair or to build new ones for rural populations that lacked reasonable access.
Characterizing the decline were severe shortages of essential medicines and other
supplies, badly paid and demoralized sta, illegal under-the-table payments by
patients, and other similar signs of service deterioration. Some eorts at reform
during this period oundered partly because a succession of military dictatorships
lacked the legitimacy to push through painful changes.
User fees and exemptions
With the era of President Jerry Rawlings, which spanned most of the 1980s and
all of the 1990s, health nancing reform returned to the political agenda in a seri-
ous way. In 1985 the cash and carry or user fee system was established, aiming
to recover up to 20% of operational nonsalary costs from patients. From the out-
set, the system was perceived as burdensome (Singleton 2006), and several stud-
ies showed its deterrent eect on use by the poor (Waddington and Enyimayew
1989, 1990; Nyonator and Kutzin 1999). In 1997 the government introduced fee
exemptions for children under age 5, pregnant women, the elderly (older than age
70), extreme indigents, and those suering from certain communicable diseases.
In theory the patients ability to pay for the services would be assessed by the doc-
tor after examination. But in practice the facilities incentive was to collect fees
whenever possible, and patients were often asked to pay a consultation fee at the
registration desk (Atim and others 2001).
Fee exemption policies for children under age 5, pregnant women, and the elderly
also faced various diculties from the start: unclear or nonexistent guidelines, uneven
application, and inadequate budgetary allocations (Atim and others 2001). Tese
persistent diculties aggravated the problems of access for vulnerable people.
1
It was
reported that many patients were observed to have diculty with paying for their
health care (especially admission) costs. Many did not turn up at the hospital until it
was too late or their illness had advanced to a more complicated phase. Some others
who got admitted and were treated subsequently absconded without paying for their
treatment. Many simply could not aord to pay for their care (Atim and Sock 2000).
Growth of risk pooling
Against this background some stakeholders began to explore alternatives to user
fees, especially community-based health insurance schemes. Te rst was the
Nkoranza District Health Insurance Scheme, started in 1992 by the Catholic
Diocese of Sunyani, which managed the Nkoranza District Hospital. Tis was
basically a facilitiy cost-recovery schemea well informed providers response to
patients observed inability to pay for care.
Other stakeholders, including the Ministry of Health, soon began to explore
the possibilities of setting up similar schemes elsewhere in the country. A new
model, the mutual health organization, was introduced around 1999, partly
inspired by experience in francophone Africa. Te model was based on social soli-
darity, community ownership, and democratic control, as opposed to the provider-
driven model typied by the Nkoranza scheme. Tis model spread rapidly in the
country, expanding from 3 schemes in 1999, to 47 in 2001, 159 in 2002, and 258
in 2003 (Atim and others 2001; Atim and Apoya 2003).
Te factors that led to the rise and rapid growth of such schemes, especially the
many problems with the user fee system, did not escape the notice of politicians.
Te leading opposition political party soon took up the issue and promised to do
away with user fees if they came to power in the 2000 election (Rajkotia 2007;
Singleton 2006), which perhaps played a crucial role in its victory. But only in the
third year of the new government, with the approaching election of 2004, was a
law rushed through establishing the NHIS.
Te new national scheme
Te NHIS was established under the National Health Insurance Act of 2003,
which set out three distinct types of health insurance schemes to be established
and operated in Ghana: district mutual health insurance, private commercial
health insurance, and private mutual health insurance. Schemes must apply to the
National Health Insurance Authority, which has the mandate to register, license,
and supervise all schemes.
2
All public health facilities in the country are automati-
cally accredited, but private health facilities have to apply for accreditation by the
authority in order to participate. By December 2008, 1,551 private providers of dif-
ferent categories had been accredited (Ghana National Health Insurance Authority
2008).
Te authority also manages the National Health Insurance Fund (NHIF),
which is nanced primarily by a sales tax levy (a 2.5% earmarked addition to
the value added tax) and 2.5% of formal sector workers contributions to the
Social Security and National Insurance Trust Fund (SSNIT) (Parliament of the
Republic of Ghana 2003). Te NHIF provides a subsidy to the district mutual
schemes to reinsure them against random uctuations in claims expenditures,
to support programs that improve access to health services, and to cover the cost
of health care for indigents and other exempt groups deemed worthy of being
subsidized.
Te NHIS provides an extremely generous benets package, covering more
than 95% of the disease conditions that aict Ghanians, including outpatient and
62 Chapter 4
inpatient care, deliveries (including complications), diagnostic tests, generic medi-
cines, and emergency care. Te district mutual schemes must adhere to this stan-
dard benet package.
To become a member, an individual needs to register with the nearest dis-
trict mutual scheme or through an agent, then wait up to six months to begin
using services. Payment of appropriate premium and registration fees is required
for those not exempt.
3
In general, premiums are meant to be based on income and
capacity to pay, with a nationally determined oor of 72,000 cedis a year (just
over US$5).
4
Districts are authorized to set premium levels, which range in prac-
tice from 72,000 cedis to 480,000 cedis across the country (Asenso-Boadi 2009).
5

Groups exempt from paying premiums include SSNIT contributors (by virtue of
their 2.5% contribution to the National Health Insurance Fund
6
) and pensioners;
people ages 70 and older; children under age 18; indigents; and pregnant women
(as of 1 July 2008, after the nal set of household data used in this chapter had
been collected).
Health care providers participating in the national scheme periodically send
claims for scheme member service use to the district mutual scheme managers,
who in turn send the claims to the national authority for settlement. Te national
scheme reimburses providers through the same path, from the national fund to
the district mutual scheme, which then pays the providers. In exceptional circum-
stances, the national fund may send repayment directly to a provider.
Early successes and challenges
Te national scheme has produced signicant achievements during its short exis-
tence, notably, the remarkable growth of its membership. Tere were 145 district
mutual schemes in operation at the end of 2008. Total membership was just more
than 12.5 million, or 61% of the population, surpassing the NHIS target of 40%
(Ghana National Health Insurance Authority 2008). About 70% of the members
are in the premium-exempt categories (table 4.1), as only informal sector adults
pay the annual premium.
While ocial country data show that an estimated 40% of the population lives
below the national poverty line, indigents account for only 2.4% of members. Te
NHIS means test for indigents is strict, requiring that the person be unemployed
with no visible source of income, be homeless, and have no identiable support
from another person (Republic of Ghana 2004). Tis narrow denition reduces
the incentives of scheme managers to try to identify such persons.
7
As a result, the
NHIS benets are out of reach for many poor people, although premium exemp-
tions for children and the elderly blunt that problem somewhat.
Tere have also been well documented delays in issuing member identication
cards after people have registered. In principle, the cards should be available by the
end of the waiting period for the scheme, but delays well beyond this period are
frequent. Tere is also some evidence that insured people regard the health care
they receive to be of poorer quality than that of noninsured people (Asenso-Boadi
2009). Tere have been reports of negative provider attitudes and practices, such
as illegal fee collection and possibly deliberate delays in seeing insured patients
(World Bank 2007). Weak performance incentives for NHIS-accredited providers
have been cited as one cause for poor quality care. In addition, supply has not kept
pace with the increased demand for health services resulting from NHIS coverage,
and this may compromise the quality of care (Ghana Ministry of Health 2008).
Some argue that the schemes generous benet package, reimbursement systems
used by the NHIS for claims (initially fee-for-service and now diagnosis-related
groups), and weak capacity for verifying provider claims at the scheme level gave
providers incentives to provide more (or more expensive) drugs and services than
necessary to insured patientsand to submit fraudulent claims (Rajkotia 2007; Gar-
shong 2008). NHIS management reports many cases of misapplication of taris and
spurious reimbursement claims by providers (Asenso-Boadi 2009). Delays in claims
payment and the substantial workload for providers to process claims, also well docu-
mented, have caused problems for providers (Asenso-Boadi 2009; World Bank 2007).
TABLE 4.1
National Health Insurance Scheme membership, 2008
Category of membership Total Percent of population
Informal sector adults 3,727,454 29.8
Ages 70 and older 866,956 6.9
Under age 18
a
6,305,729 50.4
SSNIT contributors 811,567 6.5
SSNIT pensioners 71,147 0.6
Pregnant women
b
432,728 3.5
Indigents 302,979 2.4
Total registered 12,518,560 61.3
a. Children under age 18 were initially exempt only if their parents or guardian were scheme
contributors. Since 1 September 2008, however, children under age 18 have been exempt in
their own right (known as decoupling).
b. The exemption for pregnant women became effective on 1 July 2008 for up to four prenatal
visits, delivery care, and one postnatal visit, as well as all other minimum medical benefts
needed during the 12 months following initial registration.
Source: Ghana National Health Insurance Authority 2008.
64 Chapter 4
Evaluation of the National Health Insurance Scheme
In 2004 the United States Agency for International Developmentfunded Partners
for Health Reformplus project, in collaboration with the Health Research Unit of
the Ghana Health Service, initiated an evaluation of the NHIS. Te study focused
on the following research questions:
Who has enrolled in the NHIS?
Do enrollment rates dier across socioeconomic groups?
Is there evidence of adverse selection in NHIS enrollment?
How well targeted have exemptions been?
What is the impact of the NHIS on the use of health services?
What is the impact of the NHIS on out-of-pocket expenditures for health care?
Te study was designed as a pre-post evaluation. (Te appendix contains a
more detailed discussion of study methods). Two districts were selected as study
sites: Nkoranza (in the Brong Ahafo region) and Onso (in the Ashanti region). A
baseline household survey was conducted in September 2004, prior to NHIS roll-
out. In September 2007, more than two years after launch of the NHIS, an endline
household survey was conducted in the same study sites to measure the eects
of NHIS implementation. Te endline survey did not cover the same households
as the baseline. Te baseline and endline surveys collected information on socio-
demographic characteristics of households; health insurance membership; health
care use; and payments associated with injury or illness in the two weeks prior to
the survey, hospitalization in the 12 months prior to the survey, and delivery in the
12 months prior to the survey.
At baseline 23% of the individuals in the sample were members of a community-
based health insurance scheme (35% in Nkoranza and 0% in Onso), whereas at
endline 35% across both districts were enrolled in the NHIS (45% in Nkoranza and
25% in Onso). Te baseline and endline samples of individuals were similar in dis-
tribution by age group, sex, and urban/rural location (see table A4.4 in the appendix).
A signicantly larger proportion of the endline sample belonged to a female-
headed household, a household headed by an individual with some education (rather
than no education), and a household in the top two pooled wealth quintiles. A
smaller proportion of the endline sample was from households headed by a farmer,
while a higher proportion was from households headed by a skilled worker or a gov-
ernment employee (see table A4.4 in the appendix). Tese dierences between the
survey samples imply some overall improvement in the socioeconomic status of the
two districts population in the three years between the baseline and endline surveys.
Te proportion of individuals reporting illness or injury in the two weeks before
the survey declined from 4.3% to 3.1% (p = 0.03, statistically signicant at the 97%
level). Tere was no signicant change in the distribution of these illness/injury epi-
sodes by type of condition or in the proportion that were due to accidents. Tere was
some decrease in the proportion of individuals reporting hospitalization during the
12 months prior to the survey (2.4% to 1.9%) and in the proportion of women ages
1549 reporting a birth in the 12 months prior to the survey (12% to 11%), though
neither dierence was statistically signicant (see table A4.4 in the appendix).
Determinants of enrollment in the National Health Insurance Scheme
Enrollment in the NHIS in 2007 increased with wealth quintile: 52% of those
in the top wealth quintile were enrolled in the NHIS, compared with 18% in the
poorest quintile. Tis pattern of enrollment across wealth quintiles holds within
age groups, including the age groups exempt from premiums (table 4.2). Increasing
rates of enrollment with higher wealth quintiles are also observed within groups
dened by occupation of the head of household. (Tese data were collected before
the 2008 changes that unconditionally exempted all children and pregnant women
from paying enrollment premiums.)
Probit regression analysis indicates that enrollment in NHIS was more likely if
the individual was female, had a reported chronic illness, or belonged to a house-
hold headed by a female or a household participating in a community solidarity
group (table 4.3). Likelihood of NHIS enrollment increased with education of the
head of household and wealth quintile. Children and the elderly (particularly those
ages 70 and older) were more likely to enroll than adults ages 1849, which reects
the age-based premium exemption policies. Residents of Onso were less likely to
enroll than those of Nkoranza, possibly because of the higher premiums charged
by the scheme in Onso. Also, prior experience with community-based health
TABLE 4.2
Individuals with National Health Insurance Scheme insurance coverage,
by wealth quintile and age category, 2007 endline survey (%)
Wealth quintile
Age category
Total 04 517 1849 5069 70+
Poorest 15 20 15 19 40 18
Middle-poor 29 31 25 32 50 30
Middle 37 43 33 48 62 40
Middle-rich 30 44 32 40 63 39
Richest 51 55 43 62 82 52
Total 31 38 30 41 58 35
66 Chapter 4
T
A
B
L
E

4
.
3

P
r
e
d
i
c
t
o
r
s

o
f

i
n
d
i
v
i
d
u
a
l

e
n
r
o
l
l
m
e
n
t

i
n

N
a
t
i
o
n
a
l

H
e
a
l
t
h

I
n
s
u
r
a
n
c
e

S
c
h
e
m
e

2
0
0
7

s
a
m
p
l
e

o
n
l
y

(
m
u
l
t
i
v
a
r
i
a
t
e

p
r
o
b
i
t

m
o
d
e
l

r
e
s
u
l
t
s
)
V
a
r
i
a
b
l
e
T
o
t
a
l

s
a
m
p
l
e
N
k
o
r
a
n
z
a
O
f
f
n
s
o
E
n
r
o
l
l
e
d

i
n

N
a
t
i
o
n
a
l

H
e
a
l
t
h

I
n
s
u
r
a
n
c
e

S
c
h
e
m
e

(
c
o
e
f
f
c
i
e
n
t
)
M
a
r
g
i
n
a
l

e
f
f
e
c
t
E
n
r
o
l
l
e
d

i
n

N
a
t
i
o
n
a
l

H
e
a
l
t
h

I
n
s
u
r
a
n
c
e

S
c
h
e
m
e

(
c
o
e
f
f
c
i
e
n
t
)
M
a
r
g
i
n
a
l

e
f
f
e
c
t
E
n
r
o
l
l
e
d

i
n

N
a
t
i
o
n
a
l

H
e
a
l
t
h

I
n
s
u
r
a
n
c
e

S
c
h
e
m
e

(
c
o
e
f
f
c
i
e
n
t
)
M
a
r
g
i
n
a
l

e
f
f
e
c
t
M
a
l
e
0
.
1
0
8
*
*
*
[
0
.
0
3
2
]
0
.
0
3
9
1
0
.
1
0
8
*
*
[
0
.
0
4
4
]
0
.
0
4
2
6
0
.
1
0
7
*
*
[
0
.
0
4
6
]
0
.
0
3
1
9
A
g
e

g
r
o
u
p

(
r
e
f
e
r
e
n
c
e
:

a
g
e
s

1
8
4
9
)
0
4
0
.
1
0
7
*
*
[
0
.
0
4
1
]
0
.
0
3
9
3
0
.
0
5
2
[
0
.
0
5
1
]
0
.
0
2
0
4
0
.
1
9
3
*
*
*
[
0
.
0
5
8
]
0
.
0
6
0
8
5
1
7
0
.
3
0
1
*
*
*
[
0
.
0
3
7
]
0
.
1
1
1
0
.
2
1
0
*
*
*
[
0
.
0
3
3
]
0
.
0
8
3
3
0
.
4
3
6
*
*
*
[
0
.
0
5
8
]
0
.
1
3
6
5
0
6
9
0
.
2
9
5
*
*
*
[
0
.
0
5
8
]
0
.
1
1
2
0
.
1
2
7
*
[
0
.
0
7
3
]
0
.
0
5
0
5
0
.
5
2
7
*
*
*
[
0
.
0
8
1
]
0
.
1
8
1
7
0
+
0
.
8
6
9
*
*
*
[
0
.
1
1
1
]
0
.
3
3
6
0
.
5
9
7
*
*
*
[
0
.
1
4
7
]
0
.
2
3
2
1
.
1
8
9
*
*
*
[
0
.
1
6
2
]
0
.
4
3
8
E
d
u
c
a
t
i
o
n

o
f

t
h
e

h
e
a
d

o
f

h
o
u
s
e
h
o
l
d

(
r
e
f
e
r
e
n
c
e
:

n
o

e
d
u
c
a
t
i
o
n
)
P
r
i
m
a
r
y
0
.
2
0
4
*
[
0
.
1
0
7
]
0
.
0
7
3
5
0
.
1
2
0
[
0
.
1
2
5
]
0
.
0
4
7
3
0
.
3
5
1
*
[
0
.
1
8
3
]
0
.
1
0
3
S
e
c
o
n
d
a
r
y

o
r

h
i
g
h
e
r
0
.
4
1
8
*
*
[
0
.
1
7
1
]
0
.
1
6
0
0
.
4
5
0
*
*
*
[
0
.
1
5
1
]
0
.
1
7
8
0
.
4
4
9
[
0
.
2
9
2
]
0
.
1
5
1
H
o
u
s
e
h
o
l
d

w
e
a
l
t
h

q
u
i
n
t
i
l
e

(
r
e
f
e
r
e
n
c
e
:

p
o
o
r
e
s
t

q
u
i
n
t
i
l
e
)
P
o
o
r
m
i
d
d
l
e

q
u
i
n
t
i
l
e
0
.
2
4
7
*
[
0
.
1
2
5
]
0
.
0
9
2
1
0
.
3
8
8
*
*
[
0
.
1
8
1
]
0
.
1
5
4
0
.
1
0
6
[
0
.
1
9
1
]
0
.
0
3
2
4
M
i
d
d
l
e

q
u
i
n
t
i
l
e
0
.
4
8
9
*
*
*
[
0
.
1
2
7
]
0
.
1
8
6
0
.
5
0
4
*
*
*
[
0
.
1
7
0
]
0
.
1
9
9
0
.
5
9
8
*
*
*
[
0
.
1
9
3
]
0
.
2
0
3
V
a
r
i
a
b
l
e
T
o
t
a
l

s
a
m
p
l
e
N
k
o
r
a
n
z
a
O
f
f
n
s
o
E
n
r
o
l
l
e
d

i
n

N
a
t
i
o
n
a
l

H
e
a
l
t
h

I
n
s
u
r
a
n
c
e

S
c
h
e
m
e

(
c
o
e
f
f
c
i
e
n
t
)
M
a
r
g
i
n
a
l

e
f
f
e
c
t
E
n
r
o
l
l
e
d

i
n

N
a
t
i
o
n
a
l

H
e
a
l
t
h

I
n
s
u
r
a
n
c
e

S
c
h
e
m
e

(
c
o
e
f
f
c
i
e
n
t
)
M
a
r
g
i
n
a
l

e
f
f
e
c
t
E
n
r
o
l
l
e
d

i
n

N
a
t
i
o
n
a
l

H
e
a
l
t
h

I
n
s
u
r
a
n
c
e

S
c
h
e
m
e

(
c
o
e
f
f
c
i
e
n
t
)
M
a
r
g
i
n
a
l

e
f
f
e
c
t
M
i
d
d
l
e
-
r
i
c
h

q
u
i
n
t
i
l
e
0
.
4
8
3
*
*
*
[
0
.
1
5
3
]
0
.
1
8
3
0
.
6
4
7
*
*
[
0
.
2
5
4
]
0
.
2
5
3
0
.
3
3
1
*
[
0
.
1
8
5
]
0
.
1
0
6
R
i
c
h
e
s
t

q
u
i
n
t
i
l
e
0
.
8
3
3
*
*
*
[
0
.
1
4
5
]
0
.
3
1
8
0
.
9
7
2
*
*
*
[
0
.
2
2
4
]
0
.
3
6
9
0
.
7
5
1
*
*
*
[
0
.
1
9
1
]
0
.
2
5
6
O
f
f
n
s
o

d
i
s
t
r
i
c
t
0
.
5
8
9
*
*
*
[
0
.
0
6
3
]
0
.
2
1
1
U
r
b
a
n
0
.
1
2
4
[
0
.
0
7
4
]
0
.
0
4
3
9
0
.
0
2
9
[
0
.
1
0
1
]
0
.
0
1
1
6
0
.
2
7
0
*
*
*
[
0
.
0
8
3
]
0
.
0
7
4
2
F
e
m
a
l
e

h
o
u
s
e
h
o
l
d

h
e
a
d
0
.
1
2
2
*
[
0
.
0
6
7
]
0
.
0
4
4
9
0
.
1
9
2
*
*
[
0
.
0
7
4
]
0
.
0
7
6
3
0
.
0
1
9
[
0
.
1
2
0
]
0
.
0
0
5
6
3
A
t

l
e
a
s
t

o
n
e

h
o
u
s
e
h
o
l
d

m
e
m
b
e
r

p
a
r
t

o
f

c
o
m
m
u
n
i
t
y

s
o
l
i
d
a
r
i
t
y

s
c
h
e
m
e
0
.
2
5
1
*
[
0
.
1
2
7
]
0
.
0
9
4
4
0
.
2
1
8
[
0
.
1
8
4
]
0
.
0
8
6
6
0
.
3
0
7
*
[
0
.
1
5
7
]
0
.
0
9
9
5
S
e
l
f
-
r
e
p
o
r
t
e
d

c
h
r
o
n
i
c

c
o
n
d
i
t
i
o
n
0
.
4
0
0
*
*
*
[
0
.
1
1
3
]
0
.
1
5
4
0
.
3
6
7
*
*
*
[
0
.
0
9
6
]
0
.
1
4
5
0
.
3
8
2
[
0
.
2
3
0
]
0
.
1
2
8
C
o
n
s
t
a
n
t
0
.
9
0
5
*
*
*
[
0
.
1
1
6
]
0
.
9
1
7
*
*
*
[
0
.
1
2
0
]
1
.
5
9
0
*
*
*
[
0
.
2
4
0
]
O
b
s
e
r
v
a
t
i
o
n
s
1
1
,
6
2
2
1
1
,
6
2
2
5
,
7
4
0
5
,
7
4
0
5
,
8
8
2
5
,
8
8
2
*
*
*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
0
1
;

*
*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
0
5
;

*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
1
0
.
N
o
t
e
:

N
u
m
b
e
r
s

i
n

b
r
a
c
k
e
t
s

a
r
e

s
t
a
n
d
a
r
d

e
r
r
o
r
s
.
S
o
u
r
c
e
:

A
u
t
h
o
r
s

c
a
l
c
u
l
a
t
i
o
n
s
.
68 Chapter 4
insurance in Nkoranza may have increased knowledge of insurance, its costs and
benets, and administrative procedures connected with participation. Regression
analyses run for each district separately show that determinants of NHIS enroll-
ment were generally the same (see table 4.3).
About half the households that paid premiums to enroll in NHIS stated that
the source of payment was income from harvest, 22% said that they used a gift
from a friend or a relative, and 14% said that they used their savings. Te main
reason for nonenrollment in the NHIS, cited by households where no one was
enrolled, was that the premiums were unaordable (76%); fewer than 2% cited
lack of condence in scheme management as a reason.
Adverse selection
We found some evidence of adverse selection in enrollment: those with self- reported
chronic illness were more likely to enroll. In the study sites about 4% of individuals
reported having a chronic illness. Of those, 55% were enrolled in NHIS, compared
with 34% of those who did not report a chronic illness (p < 0.01). Tis pattern was
observed in each of the wealth quintiles. In contrast, self-assessed general health
status did not appear to be associated with NHIS enrollment. Among households
where some but not all members were enrolled in NHIS, only 4% said that they
chose to insure only the sick/ill members.
We did not nd substantial evidence of adverse selection in enrollment related
to pregnancy: 36% of women who had a delivery in the 12 months prior to the
study were insured at time of delivery, compared with 33% of women who did not
have a delivery (p = 0.45). But this dierence was larger for women from the top
two quintiles, indicating that some wealthier women might have been enrolling
for the delivery coverage. Te 2008 reforms exempting pregnant women from pre-
miums were intended to encourage pregnant women, especially those from poorer
quintiles, to use health services.
Premium exemptions for National Health Insurance Scheme enrollment
How did the rules for premium exemption for government employees, children
under age 18, and people ages 70 and older work in the two study districts? Over-
all, 64% of those enrolled in NHIS reported that they were exempt from premi-
ums. Tis gure matches national gures on the share of exempt members. Nearly
all NHIS members in the two districts had paid a registration fee (97%).
Although government employees are supposed to be automatically enrolled in
the NHIS and exempt from paying NHIS premiums, this was not the case in our
two study districts. Among the heads of household who were government employees,
70% were enrolled in the NHIS. All of them had to pay a registration fee, and 42%
paid a premium that ranged from 55,000 cedis to 170,000 cedis. Age-based exemp-
tions have worked as intended in the two study districts but have not benetted dis-
proportionately those in the lowest wealth quintile. Among NHIS members, 99%
of children under age 18 and 98% of people ages 70 and older had been exempted
from paying the premium (table 4.4). In general, this was the case in each of the
wealth quintiles. But for people ages 1869, NHIS members from the poorest quin-
tiles were not more likely to be exempt than those from wealthier quintiles.
Multivariate analyses conrm that, among NHIS members, premium exemp-
tion was more likely among children and the elderly (than adults ages 1849) and
individuals from households headed by a government employee. But NHIS mem-
bers from the poorest wealth quintile were less likely to be exempt than those from
wealthier quintiles. Tose with chronic illness were less likely to benet from an
exemption than those without such illness.
Effects of National Health Insurance Scheme implementation on
health care use and spending
Tis section presents the impact of NHIS implementation on use and out-of-pocket
spending for health care using multivariate regression in the pooled pre-post data.
We report results for three types of health events: illness/injury in the two weeks
prior to the survey, hospitalization in the 12 months prior to the survey, and deliv-
ery in the 12 months prior to the survey.
We report bivariate pre-post comparisons of care-seeking and expenditure indi-
cators and adjusted regression model coecients and marginal eects for the key
TABLE 4.4
Receipt of exemptions from National Health Insurance Scheme premiums,
by age category and quintile (%)
Age category
Wealth quintile
Total Poorest 2 3 4 Richest
04 100 100 100 100 100 100
517 100 100 98 100 99 99
1834 3 10 7 2 8 6
3549 0 2 2 1 8 4
5069 5 11 9 8 18 11
70+ 94 97 100 98 97 98
Total 64 65 59 62 60 62
70 Chapter 4
variable of interest, implementation of NHIS, a dummy variable with a value of 0
for observations from the baseline 2004 survey (when the NHIS was not available)
and a value of 1 for observations from the endline 2007 survey (when the NHIS
was already implemented). Te marginal eect for an indicator shows the change
(from baseline to endline) in the likelihood of a positive outcome for the indi-
cator, controlling for other potential confounders, including the individuals age,
sex, and presence of chronic illness; education, occupation, and sex of the head of
household; household wealth quintile; and urban location and district. In addition,
the models for health care for illness/injury in the two weeks prior to the survey
include self-reported severity of the condition, and the models on maternal care
include parity.
TABLE 4.5
Change over 200407 in health care utilization and expenditures for
illness/injury in the two weeks prior to the survey
Bivariate comparison
Results from multivariate
regressions
a
Baseline
2004
b
Endline
2007
b
Signfcance Coeffcient
b
Marginal
effect
Probit regression model
Sought care at a modern health
care provider
N=413
b
N=411
b
N=814
b
Total 37% 70% *** 1.006***
[0.152]
0.382
Nkoranza 47% 74% *** 0.917***
[0.158]
0.334
Offnso 22% 64% *** 1.347***
[0.377]
0.487
Self-treated or sought care from
informal/traditional providers
N=413
b
N=411
b
N=814
b
Total 76% 44% *** 0.884***
[0.123]
0.328
Nkoranza 70% 40% *** 0.886***
[0.150]
0.340
Offnso 84% 51% *** 0.958***
[0.247]
0.309
Had positive expenditures on
treatment
N=413
b
N=411
b
N=814
b
Total 87% 57% *** 1.025***
[0.142]
0.310
Nkoranza 87% 44% *** 1.504***
[0.177]
0.478
Health care for illness/injury in the two weeks prior to the survey
Te proportion who sought care from a modern provider nearly doubled, from 37% at
baseline to 70% at endline (p < 0.01; table 4.5). Multivariate regression results conrm
this nding, showing a 38 percentage point increase in the likelihood of seeking care
between baseline and endline (see table 4.5). Te proportion seeking care from an
informal provider (such as a chemical seller, pharmacist, herbalist, or traditional healer)
or self-treating at home fell signicantly, from 76% to 44% (p < 0.01; see table 4.5).
Tese positive changes in care-seeking in the period when the NHIS was imple-
mented were accompanied by a substantial reduction in the likelihood of incurring
out-of-pocket spending for health care: average spending for treatment fell from
regressions
a
Baseline
2004
b
Endline
2007
b
Signfcance Coeffcient
b
Marginal
effect
Offnso 86% 76% * 0.403**
[0.153]
0.092
Average out-of-pocket
expenditures on treatment N=413
b
N=411
b
N=814
b
Total 24,437 14,455 **
Nkoranza 25,260 7,689 ***
Offnso 23,229 25,047
Log-linear regression model
Out-of-pocket expenditures on
treatment (among those who had
positive expenditures) N=361
b
N=226
b
N=580
b
Total 28,131 25,545 0.017
[0.264]
Nkoranza 28,875 17,381 0.414
[0.361]
Offnso 27,020 32,991 0.427*
[0.244]
*** signifcant at p < 0.01; ** signifcant at p < 0.05; * signifcant at p < 0.10.
Note: Numbers in brackets are standard errors.
a. Adjusted regression coeffcients and marginal effects for dummy variable that takes the
value 0 for 2004 and 1 for 2007 observations.
b. Total sample; sample sizes for Nkoranza and Offnso are smaller.
TABLE 4.5 (continued)
Change over 200407 in health care utilization and expenditures for
illness/injury in the two weeks prior to the survey
72 Chapter 4
24,437 cedis to 14,455 (p = 0.02). At baseline, 87% of the ill/injured incurred out-
of-pocket spending on treatment, compared with only 57% at endline (p < 0.01).
When spending on transportation to the health facility is included, the proportion
that incurred spending fell by a smaller amount between baseline and endline, but the
dierence remains signicant (88% to 71%, p < 0.01). Tere was no signicant change
in the average amount paid by those who incurred positive out-of-pocket expenditure.
Hospitalization in the 12 months prior to the survey
Results from multivariate analyses show a statistically signicant decrease, by less
than one percentage point, in the likelihood of hospitalization for illness or injury
from 2004 to 2007 (table 4.6). We do not have a measure of the need for hospi-
talization, so the interpretation of this result is ambiguous: it might be due to ear-
lier care-seeking for illness or increased use of preventive care associated with the
NHIS; but the decrease could also reect supply side factors (such as deterioration
of infrastructure for inpatient care) or random uctuations in illness severity.
TABLE 4.6
Change over 200407 in probability of hospitalization and in expenditures
for hospitalization episode in the 12 months prior to the survey
regressions
a
Baseline
2004
Endline
2007 Signifcance Coeffcient
Marginal
effect
Hospitalized in 12 months prior to
the survey N=9,554
b
N=11,770
b
N=20,660
b
Total 2.44% 1.87% 0.168**
[0.076]
0.007
Nkoranza 1.89% 1.60% 0.104
[0.071]
0.003
Offnso 3.45% 2.15% 0.209
[0.126]
0.012
Paid some amount for hospitalization
(among individuals hospitalized in
the 12 months prior to the survey) N=203 N=194 N=396
Total 87% 43% *** 1.817***
[0.289]
0.548
Nkoranza 71% 23% *** 1.489***
[0.368]
0.543
Offnso 100% 55% ***
c
Average out-of-pocket expenditures
for hospitalization N=203 N=194
Total 357,262 199,488 *
However, there was a signicant positive impact of NHIS implementation on
nancial protection from the potentially catastrophic expenditures associated with
hospitalization. Average out-of-pocket spending for hospitalization decreased from
357,262 cedis to 199,488 (p = 0.08). Te proportion of hospitalized individuals
who incurred any out-of-pocket spending for their inpatient treatment was halved,
from 87% to 43% (p < 0.01) (see table 4.6). Pooled probit regression analysis
shows a decline of 55 percentage points in the likelihood of incurring hospitaliza-
tion expenditures between 2004 and 2007. Tere was no signicant change in the
amount paid by those who incurred positive hospitalization expenditures.
Maternal health care
Tere were no signicant changes in the proportion of women who had at least
four prenatal care visits, delivered in a health facility, or delivered by Caesarean sec-
tion (table 4.7). Te proportion of deliveries that took place in a health facility was
54.4% in 2004 and remained virtually unchanged (54.9%) in 2007.
regressions
a
Baseline
2004
Endline
Marginal
effect
Nkoranza 347,668 171,007
Offnso 365,393 216,201
Log-linear
regression model
Out-of-pocket payment for
hospitalization (among those who paid
a positive amount for hospitalization) N=145
b
N=65
b
N=209
b
Total 411,814 468,007 0.286
[0.292]
Nkoranza 488,832 746,825 0.404
[0.490]
Offnso 365,393 398,889 0.419*
[0.225]
c. Not applicable because all who were hospitalized in Offnso at baseline paid for
hospitalization.
Change over 200407 in probability of hospitalization and in expenditures
for hospitalization episode in the 12 months prior to the survey
74 Chapter 4
TABLE 4.7
Change over 200420 in utilization of and expenditures for prenatal and
delivery care, among women who had delivery in the 12 months prior to
the survey
regressions
a
Baseline
2004
Endline
Marginal
effect
Had four or more prenatal care visits N=298
b
N=312
b
N=606
b
Total 73% 68% 0.250
[0.179]
0.083
Nkoranza 75% 67% 0.290
[0.222]
0.096
Offnso 71% 70% 0.243
[0.302]
0.080
Delivery in modern facility N=298
b
N=312
b
N=606
b
Total 54% 55% 0.121
[0.117]
0.047
Nkoranza 63% 57% 0.145
[0.109]
0.055
Offnso 42% 52% 0.003
[0.281]
0.001
Delivery by Caesarian section N=298
b
N=312
b
N=606
b
Total 6% 6% 0.024
[0.176]
0.003
Nkoranza 8% 7% 0.007
[0.228]
0.001
Offnso 4% 6% 0.392
[0.300]
0.001
Paid for prenatal care N=298
b
N=312
b
N=606
b
Total 88% 55% *** 1.158***
[0.170]
0.352
Nkoranza 84% 43% *** 1.214***
[0.138]
0.422
Offnso 94% 70% ** 1.234**
[0.507]
0.205
for prenatal care N=298
b
N=312
b
N=606
b
Total 49,671 42,782
Nkoranza 42,446 18,298 ***
Offnso 61,212 74,704
Paid for delivery
c
N=298
b
N=312
b
N=606
b
Total 74% 47% *** 0.885***
[0.145]
0.326
regressions
a
Baseline
2004
Endline
Marginal
effect
Nkoranza 77% 43% *** 1.020***
[0.212]
0.374
Offnso 69% 53% 0.768***
[0.165]
0.280
for delivery N=298
b
N=312
b
N=606
b
Total 108,217 75,481
Nkoranza 106,017 53,037 ***
Offnso 111,754 105,009
Log-linear regression model
Out-of-pocket expenditures for
prenatal care (among women who
had positive expenditures) N=248
b
N=155
b
N=400
b
Total 54,399 75,641 ** 0.072
[0.143]
Nkoranza 48,215 38,533 0.227
[0.243]
Offnso 63,408 109,239 *** 0.438**
[0.181]
Out-of-pocket expenditures for
delivery (among women who had
positive expenditures) N=221
b
N=135
b
N=353
b
Total 146,642 158,930 0.091
[0.134]
Nkoranza 138,591 122,559 0.136
[0.128]
Offnso 160,896 197,968 0.597**
[0.262]
c. Total sample N=590.
Change over 2004-20 in utilization of and expenditures for prenatal and
delivery care, among women who had delivery in the 12 months prior to
the survey
76 Chapter 4
But there was a substantial reduction in the proportion of women who incurred
out-of-pocket expenditures for maternal care between baseline and endline.
While 88% of women had some prenatal care expenditures at baseline, only 55%
reported expenditures at endline (p < 0.01). Average spending on prenatal care fell
from 49,671 cedis to 42,782, though the change was not statistically signicant
(p = 0.41). Among those who paid a positive amount, expenditures increased sig-
nicantly, from 54,399 cedis to 75,641. Tis increase was driven by the sample in
Onso district.
Average spending for delivery care fell from 108,217 cedis to 75,481 (p = 0.12).
Te proportion of women who had to pay some amount for their delivery fell from
74% to 47% (p < 0.01), and multivariate probit analysis indicates a signicant
decrease of 33 percentage points. Tere was no signicant change in the amount
paid for the delivery among the women who had to pay a positive amount.
Differences in health care use and payment comparing National
Health Insurance Schemeinsured and uninsured in 2007
Below we present comparisons of insured and uninsured individuals within the
2007 survey only. Simple comparisons of insured with uninsured do not control
for self-selection into the insurance scheme. Dierences between these two groups
should not be interpreted as causal or due solely to insurance. Propensity score
matching can reduce the eect of endogeneity and give a closer approximation of
the individual eects of NHIS enrollment. Unfortunately, small sample sizes in
this study and the lack of an appropriate comparison population put severe limita-
tions on propensity score matching methods. While at best indicative, we provide
an overview of the propensity score matching results because they generally con-
rm the results we show from the pre-post analyses and allow for an individual-
level (rather than population-level) interpretation.
Health care for illness/injury in the two weeks prior to the survey
Bivariate and probit analyses indicated that individuals insured by the NHIS were
about twice as likely to seek formal care for illness/injury and about half as likely to
self-treat or seek informal/traditional care, compared with the uninsured. Among
those ill/injured in the two weeks prior to the survey, the insured paid 72% less
than the uninsured for treatment (7,259 cedis and 25,682 cedis respectively, p <
0.01). About 86% of the uninsured had positive expenditures on treatment, com-
pared with 38% of those insured by the NHIS at time of illness (p < 0.01). Enroll-
ment in NHIS also appears to have reduced mean expenditures in the subgroup
of ill/injured individuals with positive expenditures for treatment. Multivariate
regression analysis indicates an 85% decrease in expenditures among those with
positive expenditures. Propensity score matching corroborates these results, but
shows attenuated dierences between NHIS-insured and uninsured individuals.
Hospitalization in the 12 months prior to the survey
Individuals covered by NHIS were signicantly more likely to be hospitalized than
those not covered by insurance. We were not able to t a propensity score matching
model to our data to investigate how much of this dierence might be attributed to
the greater access to needed health care provided by NHIS insurance coverage, as
opposed to adverse selection or unobservable individual characteristics associated
with enrollment and likelihood of hospitalization.
Coverage by the NHIS at the time of hospitalization was associated with substan-
tially reduced hospitalization expenditures. On average, the insured paid 8,010 cedis
for their hospitalization, compared with 477,418 cedis by the uninsured (p < 0.01).
Nearly all uninsured individuals who had been hospitalized incurred positive expen-
ditures for the hospitalization (99%), compared with only 5% of those covered by
NHIS at the time of hospitalization. Results from propensity score matching also
point toward a substantial insurance eect on hospitalization expenditures.
Maternal health care
Among women with a delivery in the 12 months prior to the survey, those enrolled
in the NHIS were signicantly more likely than the uninsured to have four or more
prenatal care visits, to deliver in a modern health care facility, and to deliver by
Caesarean section. Results from propensity score matching indicate that for each
of these indicators much of the dierence between insured and uninsured women
might be attributed to the eects of insurance coverage, rather than merely selec-
tion bias. A recent study using data collected specically for analysis using propen-
sity score matching nds a strong positive impact as well (Mensah, Oppong, and
Schmidt 2010). Tis nding contrasts with our pre-post analysis, which did not
indicate any increase in use of maternal health services as NHIS was implemented.
NHIS coverage had a substantial eect on maternal care expenditures: 81% of
uninsured women had positive prenatal care expenditures, compared with only 13%
of women covered by the NHIS during their pregnancy (p < 0.01). On average, insured
women paid about one-tenth as much as uninsured women for prenatal care (6,293
cedis and 69,710 cedis, respectively; p < 0.01). Average expenditures for delivery care
among uninsured women were 115,189 cedis, compared with 17,138 cedis for insured
women (p < 0.01), and women covered by the NHIS at time of delivery were 70% less
likely to have positive expenditures for delivery care than uninsured women (p < 0.01).
78 Chapter 4
Results from propensity score matching indicate that enrollment in the NHIS pro-
vided nancial protection to women who obtained prenatal and delivery care.
Policy implications
Te NHIS is not achieving equitable enrollment or ensuring well targeted
exemptions
Results from the two study districts show that enrollment in the NHIS increased
with wealth quintile: 52% of those in the top wealth quintile were enrolled, com-
pared with 18% in the poorest quintile. In addition, the evaluation conrms that
exemptions for the indigent have not been well targeted toward the poorest of
the poor. NHIS members from the poorest wealth quintile were in fact less likely
to be exempted than members from wealthier quintiles. Inadequate information
ows and other barriers may be preventing some intended beneciaries from ben-
eting fully from the law. Many formal sector employees paid NHIS premiums,
even though they should fall in the exempt category. But age-based exemptions
have worked as intended in the two study districts. Children under age 18 and the
elderly were more likely to enroll, and nearly all insurees under age 18 and over
age 70 enrolled without paying the premium.
Adverse selection is a concern
Individuals with a chronic illness were more likely to enroll, implying that the wait-
ing period recommended by NHIS regulations for membership has not eliminated
adverse selection. Tere was no signicant evidence of adverse selection related to
expected delivery, though higher rates of enrollment associated with pregnancy
might be considered socially benecial.
NHIS has had a positive impact on the use of modern health care
Use of curative health care services at modern health facilities increased substan-
tially with the NHIS, while there was a signicant decline in self-treatment and
the use of informal/traditional care. Tese positive eects on care-seeking for ill-
ness are likely largely attributable to NHIS implementation. Hospitalization rates
declined over the period of the study, but the interpretation of this result is ambigu-
ous because we do not have a measure of need for hospitalization in our data.
Little impact on use of maternal health services
NHIS implementation was not associated with increased use of maternal health
care services, which may reect the importance of nonnancial barriers to formal
care-seeking for delivery care. Tese barriers (such as poor quality in facilities, cul-
tural preferences to deliver at home, and lack of transportation) deserve further
research and need to be addressed through nonnancial interventions as well.
Positive impact on nancial protection
NHIS implementation was associated with substantial improvements in nancial
protection for health care, including lower out-of-pocket spending on outpatient
curative care, hospitalization, and delivery care. Te changes were substantial for
potentially catastrophic expenditures on hospitalization, where the proportion of
patients who had positive expenditures for inpatient curative care was halved, and
average expenditures fell 44%. In addition, average spending for delivery fell 30%
and the proportion of women with positive delivery expenditures fell from 74% to
47%.
Challenges for the future
In 2003 the rush to keep the campaign promise of abolishing user fees led to a
number of questionable design decisions during the creation of the NHIS. Tese
design problems, as well as district-level implementation challenges, could derail
the substantial achievements documented in this evaluation. Key concerns include
the following.
Financially unsustainable benet package and subsidies
To give people the feeling that the user fee system had been eectively abolished as
promised, the 2003 law provided for a generous but arguably unsustainable benet
package. Tis was accompanied by subsidies to enroll large segments of the popula-
tion without requiring any premiums or co-payments. In essence, this means the
NHIS has become primarily a tax-funded social health insurance system. In 2006,
76% of National Health Insurance Fund income was from the national health
insurance (value added tax) levy, 24% was from SSNIT contributions of formal
sector workers, and only 0.01% was from premiums paid by informal sector mem-
bers (Ghana Health Service 2007).
Principal income source is not related to number of enrollees
Te NHIS, unlike other typical social insurance systems, has an income base that
is not directly or principally linked to the number of enrollees. Te large major-
ity of NHIS members are not social security or informal sector contributors but
individuals who do not pay any form of premium. All other things being equal,
available revenue for the scheme will remain basically constant over time, despite
80 Chapter 4
increases in enrollment. An International Labour Organization actuarial simula-
tion concluded that an imbalance between NHIS revenues and expenditures was
likely to arise within four to ve years of the schemes initiation as membership
gures roseand would appear even sooner, the faster the uptake of the NHIS by
exempted groups (Yankah and Lger 2004).
Government guarantees, reinsurance, and moral hazard
A further consequence of the politically driven process behind the NHIS is that the
government is seen by schemes and their managers as having such a huge stake in
the schemes that it dare not allow a scheme to fail or be unable to provide services
to its members. Schemes therefore do not have a strong sense of responsibility or
attention to sustainability. Moreover, the reinsurance aspect of the National Health
Insurance Fund has turned into a blanket guarantee against all losses rather than
random uctuations. Schemes do not have an incentive to run any surpluses.
Previously existing user fee exemptions, including those to address equity,
remain in limbo
Te NHIS law failed to detail how the scheme would interface with existing
exemption schemes, leading to various implementation problems at facilities. For
pregnant women, this was resolved by adding them to the NHIS exempt groups in
2008. Providing exemptions to indigents has proved more challenging.
Te supply of care may not be keeping pace with expanding demand resulting
from NHIS coverage
Tis emerging concern was documented in a 2007 Ministry of Health indepen-
dent review of the health sector: Tere is a growing need for capital investment,
to address deterioration of existing health infrastructure, provide sta accommo-
dation and infrastructure in deprived areas, expand and improve the quality of
existing facilities to meet increased demand created by the NHIS, and replace or
upgrade vehicles and equipment (Ghana Ministry of Health 2008). Increased
demand without investment in increased supply may lead to worsening of quality
of care over time, which would compromise the gains in health services use in the
early years of the NHIS. Timely investment in expanding and improving existing
health care infrastructure is essential.
Appendix: Details and methods of the evaluation
Study sites
Ghanas administrative classication lists Nkoranza as a deprived district and
Onso as less-deprived. Both are predominantly rural, with agriculture as the
primary economic activity. Table A4.1 summarizes health services availability in
the two districts.
At the time of the baseline survey in 2004, the community-based health
insurance scheme covered about 34% of the district population in Nkoranza. Te
scheme was managed by the districts mission hospital and covered primarily inpa-
tient services, including Caesarean sections. In 2005 this scheme was transformed
into a districtwide mutual health scheme under the NHIS, serving the entire dis-
trict and covering a broader range of health services and providers. In Onso there
were no community-based health insurance schemes in operation before the dis-
trictwide scheme of the NHIS was established in 2005. Table A4.2 summarizes the
characteristics of the health insurance available in the study sites during the period
of the study.
Sample selection
Te study used cross-sectional sampling, with dierent samples of households
selected at baseline and endline. A two-stage cluster sampling design was used, rst
selecting a sample of municipalities, then selecting a sample of households within
the municipalities. In each municipality the total sample was proportional to the
total number of households in the municipality, and approximately equal samples
of insured and uninsured households were selected.
8
Te baseline sample covered
1,805 households, and the endline sample consisted of 2,520 households. Table
A4.3 summarizes the resulting sample sizes.
TABLE A4.1
Health services provision in study districts, 2007
Nkoranza Offnso
District population 128,960 138,676
Area (square kilometers) 2,300 1,254
Number of public health posts and health centers 12 7
Number of mission clinics 0 2
Number of private clinics 1 4
Number of hospitals 1 2
Source: 2000 Population and Housing Census.
82 Chapter 4
Data collection instruments
Te data collection instruments included a household characteristics questionnaire
and a health care questionnaire. Te household questionnaire collected information
on sociodemographic characteristics and composition of households, and details on
TABLE A4.2
Coverage of healthcare services by health insurance in study sites
District
Baseline 2004 Endline 2007
Nkoranza Nkoranza Offnso
Type of health insurance Health insurance scheme
(community-based)
District mutual health
insurance scheme
(under NHIS)
District mutual health
insurance scheme
(under NHIS)
Percent of
population registered
(administrative data)
34% 45% 36%
Registration fee None Previous Nkoranza
scheme members:
20,000 cedis (US$2.15)
New members: 30,000
cedis (US$3.23)
SSNIT contributors:
50,000 cedis (US$5.38)
Other members: 20,000
cedis (US$2.15)
Premium payment
a
Annual premium: 30,000
cedis (US$3.61) per
individual for frst year,
20,000 cedis (US$3.01)
annual renewal
Annual premium: 80,000
cedis (US$8.60)
Annual premium:
150,000 cedis
(US$16.13)
Enrollment and premium
and fee payment
requirements
Entire household
enrollment encouraged
Payment by installment
allowed but upfront
payment encouraged
Payment by installment
allowed but upfront
payment encouraged
Participating providers Only district hospital
(mission health facility)
Primary, secondary, and
tertiary health facilities
in public, private,
and mission sector
Primary, secondary, and
tertiary health facilities
in public, private,
and mission sector
Benefts covered
Outpatient visit Only treatment of dog
and snake bites

Prenatal care
Hospital admission
Normal delivery
Delivery by
Caesarian section

Postnatal care
Drugs and supplies Only for inpatient care
a. Current old Ghanaian cedis, converted at dollar exchange rate at time of each survey.
Source: Authors.
individual health insurance membership. Te health care questionnaire was admin-
istered to those who had been ill or injured in the two weeks prior to the survey,
those who had been hospitalized in the 12 months prior to the survey, and women
who had had a delivery in the 12 months prior to the survey. It collected information
on self-treatment, health careseeking (formal and informal or traditional care) and
payments associated with care. For individuals ill or injured in the two weeks prior
to the survey, the data on out-of-pocket expenditures included payments for infor-
mal/traditional care and itemized expenditures on formal care, including consulta-
tion, lab tests, drugs, x-rays, hospitalization charges, other facility-care expenditures,
and unocial payments to health providers. For individuals hospitalized in the 12
months prior to the survey, the questionnaire asked about total out-of-pocket expen-
ditures for the hospitalization. Women who had had a delivery in the 12 months
prior to the survey were asked how much they paid for prenatal care during the entire
pregnancy and how much they and their family paid for the delivery.
Analytical methods
Sampling weights reecting the probability of selection in the sample were assigned
to each household and used in all analyses presented here. Data on household assets
and housing quality were used to construct wealth indexes at baseline and endline,
using principal components analysis (Filmer and Pritchett 2001).
9
Te indexes dif-
ferentiated households in ve asset-based wealth groups (wealth quintiles) in each
time period. A separate pooled wealth index was developed for the analyses using
the pooled data (across baseline and endline samples).
Te methods explored include pre-post bivariate comparisons of key indicators
related to our research questions; multivariate regression analyses on pooled pre-
post data to measure the eects of NHIS implementation on these indicators, con-
trolling for other potentially confounding variables; multivariate regression analysis
in the 2007 sample to identify characteristics associated with individual NHIS
TABLE A4.3
Sample sizes from household surveys in Nkoranza and Ofnso
Baseline 2004 Endline 2007
Number of households 1,805 2,520
Number of individuals 9,554 11,770
Individuals reporting illness and injury in the two weeks prior to the survey 413 411
Individuals reporting hospitalization in the 12 months prior to the survey 203 208
Women reporting delivery in the 12 months prior to the survey 298 312
84 Chapter 4
TABLE A4.4
Sample characteristics (%)
Pre-post comparison 2007 sample only
Variable 2004 2007 p-value
a
Not
insured
Insured under
National Health
Insurance Scheme p-value
a
Sample size
(number of individuals)
b
N=9,554 N=11,757 N=6,794 N=4,963
Have health insurance 23 35 <0.01
Age <0.01 <0.01
04 15 15 16 13
517 34 35 34 38
1849 39 36 39 30
5069 8 9 8 10
70+ 3 5 3 8
Missing 1 0 0 0
Male 48 47 0.34 48 44 0.01
Reported chronic illness
c
4 3 6 <0.01
Self-assessed health status
c
0.24
Very good 63 63 62
Good 36 36 35
Average 1 1.1 1.9
Poor 0 0.3 0.4
Household heads level of
education 0.01 0.02
No education 42 34 36 29
Primary/Junior secondary 50 57 56 58
Secondary or higher 8 9 7 13
Household heads current
occupation <0.01 0.19
Not working 9 8 9 9
Farmer/fsher 77 69 71 64
Government worker 3 5 4 7
Artisan/trader 14 15 13 17
Other 0 3 2 3
Head of household is female 25 32 <0.01 30 35 0.08
Urban location 15 12 0.29 12 13 0.47
Asset index quintiles
d
<0.01 <0.01
Poorest quintile 27 16 27 11
Poor-middle quintile 22 20 22 17
enrollment; and propensity score matching analyses in the 2007 data to measure
the eect of individual NHIS membership on key indicators, addressing the poten-
tial endogeneity of insurance enrollment.
Limitations of the study
Our study results have ve important limitations. First, the study included only 2
of Ghanas 138 districts, which limits the generalizability of study results regionally
or nationally. Second, the pre-post design of our study means that the eects of
NHIS implementation measured may be confounded by the eect of other health-
related policy interventions that may have occurred in the study districts in the
three years between baseline and endline data collection. Tird, the lack of panel
data or an instrumental variable constrains our ability to account for endogene-
ity when measuring the impact of individual NHIS membership on use of care
Pre-post comparison 2007 sample only
Variable 2004 2007 p-value
a
Not
insured
Insured under
National Health
Insurance Scheme p-value
a
Sample size
(number of individuals)
b
N=9,554 N=11,757 N=6,794 N=4,963
Middle quintile 21 18 17 21
Middle-rich quintile 16 23 18 21
Richest quintile 14 23 15 29
At least one household member
in a community-solidarity group
c
12 10 15 0.03
Reported illness or injury in
two weeks prior to the survey 4 3 0.03 2 5 <0.01
Hospitalized in 12 months
prior to the survey 2 2 0.15 1 3 <0.01
Gave birth in 12 months prior
to the survey (women ages
1549 only) 12 11 0.22 10 12 0.45
Mean household size 7 6 <0.01 6.32 5.93 0.10
a. Statistical signifcance of difference in means (t-test or chi-square).
b. Sample size is smaller for some variables because of observations with missing data.
c. Question not asked at baseline.
d. Quintiles developed from pooled baseline and endline data shown for pre-post comparison.
Quintiles developed from endline-only data shown for endline comparison.
TABLE A4.4 (continued)
Sample characteristics (%)
86 Chapter 4
(Waters 1999).

Fourth, the study does not have data on changes in the quality of
health care in the study sites, which could be inuenced by NHIS implementa-
tion and may be a potential confounder of the use and nancial protection eects.
Fifth, sample sizes for some of our key indicators of health care use and spending
are small (particularly for rare health-related events, such as hospitalization) and
limit our ability to detect the eects of NHIS implementation.
Notes
1. For instance, the 2004 Health Sector Review (Ghana Ministry of Health 2008) found
that, because of insucient funding, only 67% of facilities claims for free services in
2004 were reimbursed by the Ministry of Health.
2. Te availability of substantial subsidies for the district mutual schemes led to the
demise of virtually all previously existing nondistrict mutual health organizations,
after the law was implemented. Te members of the previous nondistrict schemes
became integrated into the district mutual schemes, providing skills and personnel
for their operation. Te regions where district mutual schemes are most developed
todaythe Brong Ahafo, Eastern, Ashanti, and Northern regionswere also those
where the previous mutual health organizations were most highly developed.
3. District mutual schemes organize regular sensitization exercises in the communities to
increase enrollment. Terms of premium payments are decided by each scheme. Mem-
bers who pay premiums directly to the schemes (informal sector adults) may pay their
dues upfront at the beginning of the scheme year, or by installments in accordance
with arrangements reached with the scheme managers.
4. On 1 July 2007 the Ghanaian cedi was redenominated and is now worth 10,000
old cedis. We report all currency amounts in old cedis. Expenditures in 2007 were
adjusted for ination (40% over three years), so all expenditures are reported in 2004
old cedis.
5. Setting premium rates is largely at the discretion of district mutual scheme manag-
ers, who take into account the premium guidelines given by the National Health
Insurance Authoritynot below 72,000 cedis and graduated according to income
if possible and use their knowledge of incomes in the community and how much
people typically pay for health services to decide on the premium for their scheme
(Ghana Health Service 2009).
6. Before the NHIS was introduced, public servants were entitled to a package of free
health care benets, more limited than the NHIS package. Tey had to pay for services
out of pocket and then seek reimbursement, which typically took months to process
and was frequently below the amount incurred due to annual budget ceilings (Atim
and others 2001).
7. By this denition, it is doubtful that any but the homeless beggars in Accra and other
major urban centers could qualify to receive benets, and it rules out many of the poor
in the rural areas where taboos, family pride, and social solidarity systems prevent even
poor relatives from being cast out to the streets.
8. In Onso the baseline sample was selected using systematic random sampling in each
selected municipality.
9. Te variables used in the index are type of cooking fuel, main source of drinking
water, type of toilet, type of oor, people per room, has electricity, and ownership of
radio, television, fridge, phone, bicycle, motorcycle, and car.
References
Asenso-Boadi, F. 2009. Ghanas National Health Insurance System: Design, Implementation
Perspectives. Presentation on behalf of Ras Boateng, at the African Health Economics and
Policy Association Conference, 10 March, Accra.
Atim, C., and M. Sock. 2000. An External Evaluation of the Nkoranza Community Financing
Health Insurance Scheme, Ghana. Technical Report 50. Partnerships for Health Reform Proj-
ect and Abt Associates Inc., Bethesda, MD.
Atim, C., S. Grey, P. Apoya, S. J. Anie, and M. Aikins. 2001. A Survey of Health Financing Schemes
in Ghana. Bethesda, MD: Partners for Health Reformplus Project and Abt Associates Inc.
Atim, C., and P. Apoya. 2003. A Survey of Mutual Health Organizations in Ghana, November 2003.
Bethesda, MD: Partners for Health Reformplus Project and Abt Associates Inc.
Dakpallah, George. 1988. Financing of Health Services: Eects of Cost Recovery Mechanisms
on Health Services in Developing CountriesCase Study of Ghana MSc. Tesis. University
of Wales, Swansea.
Filmer, D., and L. H. Pritchett. 2001. Estimating Wealth Eect without Expenditure Dataor
Tears: An Application to Educational Enrollments in States of India. Demography 38 (1):
11532.
Garshong, B. 2008. Evaluating the Eects of the National Health Insurance Act in Ghana: Sum-
mary of In-depth Interviews with District Mutual Health Organizations in Six Districts.
Unpublished manuscript.
Ghana Health Service. 2007. Annual Report 2007. Accra: Ghana Health Service.
. 2009. Evaluation of the Ghana National Health Insurance Scheme. Bethesda, MD: Health
Systems 20/20 Project and Abt Associates Inc.
Ghana Ministry of Health. 2008. Independent Review, Health Sector Programme of Work
2007. Accra: Ghana Ministry of Health.
Ghana National Health Insurance Authority. 2008. National Health Insurance Authority Status
Report: Operations, December 2008. Accra: Ghana National Health Insurance Authority.
Ghana, Republic of. 2004. National Health Insurance Regulations, 2004 (L.I. 1809). Accra.
88 Chapter 4
Mensah, J., J.R. Oppong, and C.M. Schmidt. 2010. Ghanas National Health Insurance Scheme
in the Context of the Health MDGs: An Empirical Evaluation Using Propensity Score Match-
ing. Health Economy 19 (Suppl.): 95106.
Nyonator, F., and J. Kutzin. 1999. Health for Some? Te Eects of User Fees in the Volta Region
of Ghana. Health Policy and Planning 14 (4): 32941.
Parliament of the Republic of Ghana. National Health Insurance Act, 2003 (Act 650).
Rajkotia, Y. 2007. Te Political Development of the Ghanaian National Health Insurance System:
Lessons in Health Governance. Bethesda, MD: Health Systems 20/20 Project and Abt Associ-
ates Inc.
Singleton, J. L. 2006. Negotiating Change: An Analysis of the Origins of Ghanas National Health
Insurance Act. Honors Project. Macalester College, Department of Sociology, St. Paul, MN.
Waddington, C., and K. A. Enyimayew. 1989. A Price to Pay: Te Impact of User Charges in
Ashanti-Akim District, Ghana. International Journal of Health Planning and Management 4
(1): 1747.
. 1990. A Price to Pay, Part 2: Te Impact of User Charges in the Volta Region of Ghana.
International Journal of Health Planning and Management 5 (4): 287312.
Waters, H. 1999. Measuring the Impact of Health Insurance with a Correction for Selection
BiasA Case Study of Ecuador. Health Economics 8 (5): 47383.
WHO (World Health Organization). 2008. Country Cooperation Strategy: At a Glance. World
Health Organization, Geneva.
World Bank. 2007. Project Appraisal Document on a Proposed Credit to the Republic of Ghana
for a Health Insurance Project. World Bank, Washington, DC.
Yankah, B., and F. Lger. 2004. Financial Analysis of the National Public Health Budget and
of the National Health Insurance Scheme. ILO Discussion Paper 4. International Labour
Organization, Geneva.
89
Impact of Health Insurance
on Access, Use, and Health
Status in Costa Rica
James Cercone, Etoile Pinder, Jose Pacheco Jimenez, and Rodrigo Briceno
C
h
a
p
t
e
r

5
Costa Rica, as a middle-income country that has largely achieved uni-
versal health coverage, allows for analysis of the dierences in behavior
and care for the small share of citizens who remain uninsured. Tis
chapter sheds light on the impact of being covered by insurance in a
country where access is guaranteed even if uninsured and on the costs
and benets of covering the last 10%20% of the population with
insurance or other approaches.
A country of 4.5 million people, Costa Rica has a per capita gross
domestic product (GDP) of US$5,600 (US$10,700 in purchasing power
parity terms), and in 2007 it spent about 7.1% of GDP on health care.
In 2008 the infant mortality rate was less than 10 deaths per 1,000 live
births, and average life expectancy was 80 years for women and 76 years
for men. Average life expectancy exceeds that of the United States by a
year, even though U.S. GDP per capita is four times that of Costa Rica.
Costa Rica has mandatory health insurance coverage, established
in 1941, and a comprehensive primary health care model that reaches
all citizens. Te Caja Costarricense del Seguro Social (the Caja) is an
autonomous government institution that is both insurer and provider
of care. Nearly 90% of the countrys 4.5 million people are covered.
Te health insurance system is based on traditional Bismarkian social
insurance, with an expanded role of the central government to cover the
uninsured population. It provides equal access to health care services,
irrespective of income or contribution. Te formal sector contributes
90 Chapter 5
14.75% of payroll income to sustain the system. Te poor and indigent are cov-
ered by the noncontributory and insured by state regimes, which have led to
equal access to health services for the poor and wealthy, something not seen in
any of Costa Ricas neighboring countries. In addition, the absence of copayments
removes another possible barrier to equal access.
Main characteristics of the Costa Rican health system
Structure
Te Costa Rican health system includes a wide range of entities; the most relevant
for this study are the Ministry of Health, the National Insurance Institute (INS),
and the Caja. Te ministry oversees the performance of the essential public health
functions and exercises the stewardship role in the health sector, while the INS
oers protection against occupational risks and trac accidents as well as accident
liability and a voluntary insurance plan for health care.
Te Caja is the key institution for this study. It manages and organizes manda-
tory health insurance and is an autonomous institution with technical, administra-
tive, and functional independence. It manages the compulsory health insurance
funds that come from payroll taxes and provides the highest proportion of health
care services in the country, covering roughly 90% of the population with a broad
package of services. Besides health services, it provides social security protection
to insured individuals and poor households through the Disability, Old Age, and
Funeral regime. In Costa Rica there is an administrative purchaser-provider split
between the nancial network and the provider network of the Caja. Nearly all
provision is through the Caja network; however, the Caja also contracts with pri-
vate providers and nongovernmental organizations for some services.
Te network of providers belongs to the Caja, which is organized as a pyramid-
style network with primary care at the bottom and tertiary hospital care at the top.
Primary care consists of 104 health regions and 953 basic care teams (Equipos Bsi-
cos de Atencin Integral en Salud, or EBAIS). Each EBAIS covers 3,5004,000
people and consists of a general practitioner, an auxiliary level nurse, and a pri-
mary care technician. All members must be registered with a primary care provider.
Recently, the Caja has expanded its purchasing options, and some primary care ser-
vices, such as minor surgeries and diagnoses, are purchased from nonpublic agents.
Secondary care consists of 10 major clinics, 13 suburban hospitals, and 7
regional facilities specializing in hospital services. Tertiary care has three general
hospitals and ve specialized hospitals (women, children, geriatrics, psychiatry, and
rehabilitation). General and regional hospitals have a set number of people in their
Impact of Health Insurance on Access, Use, and Health Status in Costa Rica 91
catchment areas, so clients cannot freely select their hospitalthat is, every person
must register in the hospital in the zone where he or she lives.
Te health care sector also has a private subsector, which has expanded sub-
stantially in recent years, exemplied by its increasing number of medical profes-
sionals. In the 1990s the share of medical sta in the private sector increased from
10% to 24%. Household surveys show that about 30% of the population uses
private health services at least once a year.
Eligibility and coverage
According to the constitution and founding laws of the Caja, mandatory health
insurance in Costa Rica covers the risks of illness, maternity, disability, aging, and
involuntary unemployment. Te insurance also partly covers nancial burdens due
to maternity, widowhood, orphanage, and burial. All wage-earners must be cov-
ered by health insurance, which also covers the workers dependents. Coverage
for poor families and the indigent is dened by the Caja board of directors and
is nanced by the government from general tax revenues. All pensioners are auto-
matically covered by health insurance.
Health insurance was originally established to protect workers against the risk
of illness, maternity, and labor injuries, and initially no other group was covered by
the insurance. Te mandatory health insurance scheme has evolved over time. In
1961 congress established universal health insurance for workers and their families.
Te system expanded rapidly in the 1960s, when coverage almost tripled to half the
population. In 1975 health insurance was expanded to cover agricultural workers.
In 1978 the Caja created the voluntary health insurance scheme for independent
workers. In 1984 it created the special regime, funded from general tax revenues,
to cover the indigent. In 1993 all provision of care was moved to the Caja, and
the Ministry of Health became the regulator and coordinator of the sector. Te
fundamental dierence between Costa Rica and the rest of Latin America is that
Costa Rica unied its public delivery system under social security, eliminating the
parallel system operated by the Ministry of Health.
In the 2000s the approval of the Act for the Workers Protection set the manda-
tory enrollment of independent workers. Recent decisions by the Caja board have
aimed at expanding coverage to specic vulnerable groups. For instance, the ben-
ets of health insurance can now cover a brother or sister of a contributory member
if he or she is disabled or taking care of their parents.
Roughly 88% of Costa Ricas residents have health insurance. Coverage
increased from 60% in 1975 to a high of 92% in 1990. Since then, coverage rates
have remained at 86%88%.
92 Chapter 5
Te Caja law establishes universal coverage of health servicesthat is, no per-
son can be denied health care services even if uninsured. Emergency services are
free to both the insured and the uninsured. If uninsured persons go to a public
facility, two possibilities arise. First, they can pay for the services. Second, if they
lack resources and are deemed indigent, they can become insured by the state.
Te process of aliation and access to the health insurance system
Any person living in Costa Rica may become aliated with the social health insur-
ance scheme in any one of at least ve ways.
As a formal or self-employed worker. Health insurance is mandatory in Costa
Rica for all categories of paid workers. Until 1999 this obligation was dened
for formal salaried workers only. Since then, the Workers Protection Act estab-
lished that self-employed workers must also be aliated with the Caja pro-
gram. Migrants are covered by the same legislation as citizens or residents. Te
contribution rate for formal employees is a 14.75% payroll tax and is 10.25% of
reported income for independent workers.
As a pensioner. All pensioners, either members of the contributory or noncon-
tributory schemes, are automatically aliated with the national health insur-
ance program. Noncontributory pensioners receive a pension from the state even
though they never contributed. Contributions are set at 14% for pensioners.
As state-covered members. Under this program, aliation is usually dened at
demandthat is, the uninsured person rst receives health care services at the
public facility, then is requested to pay for the services or to become enrolled
by the state, based on a means test applied by the Caja. Te program is funded
by special taxes on luxury goods, liquor, beer, cola, and other similar imported
goods. It accounts for about 12% of the insured.
As a voluntary member. A special option gives individuals the possibility to
enroll in the insurance program if they do not belong to any of the other
groups. In this option, the person enrolls and pays a regular fee of about $25 a
month. Tere is no dierence in the package of services this category receives.
As an indirect member. Dependents of members are automatically covered.
Spouses and children studying until age 25 are covered by the direct member.
As mentioned above, some additional options exist, such as dependency status
for a contributing members brother or sister who is taking care of his or her
parents and is older than 60.
If a person cannot be included in any of the ve groups, the chances of being
insured are essentially nil because of the absence of private health insurance options
prior to 2009. Te Free Trade Agreement with the United States opened the Costa
Rican insurance market, a public monopoly since 1924, allowing international
insurance companies to establish, operate, and sell insurance plans.
1
Currently, the
INS, the public insurance monopoly, operates one health insurance scheme with
approximately 10,000 aliates nationwide.
2
Some transnational companies have
insured their employees in the past through international health insurance pro-
grams, allowing their workers to obtain health care in private hospitals and clinics.
Table 5.1 summarizes sources of revenue for the Caja. Te contributory portion of
the system accounts for about 76% of revenues. For poor households, the Caja actuar-
ial department estimates the number of poor households based on a household survey.
Ten it estimates an average premium, based on average wages and the application of
a 14% payroll tax rate to that average. Te number of estimated poor people times the
average premium determines the expected global contribution from the state.
Who are the uninsured?
An analysis of the National Health Survey (ENSA 2006) shows signicant dif-
ferences in the sociodemographic proles of the insured and uninsured in Costa
Rica (table 5.2; see box 5.1 for a description of the data and methodology used
in the analysis). According to the survey, 81% are insured and 19% are not. Te
uninsured are less likely to be female (54% of the insured are female, compared
with 46% of the uninsured) and less likely to be married (32% of the insured are
married, compared with 20% of the uninsured). No signicant dierence in aver-
age age exists between the two groups, yet the age structure is substantially dier-
ent (gure 5.1). Adults ages 1954 comprise 50% of the insured but 62% of the
uninsured. Adults ages 55 and older comprise 16% of the insured but 12% of the
TABLE 5.1
Payroll fees by insurance scheme, 2006
Type of health insurance
Contribution by type of contributor (%)
Employee Employer State
Pension
regime Total
Salaried 5.50 9.25 0.25 15.00
Independent 4.75 5.50 10.25
Voluntary 4.65 5.50 10.15
Contributory pensioner 5.00 0.25 8.75 14.00
Noncontributory pensioner 0.25 13.75 14.00
Insured by state 14.00 14.00
not applicable.
Source: Caja Costarricense del Seguro Social.
94 Chapter 5
uninsured. Tese statistics suggest that there can be self-selection of the healthy
into uninsured status even in a system with universal coverage.
Te uninsured are more likely to have completed secondary education and to
work for a small company with fewer than 10 employees. Tey are signicantly
more likely to be independent workers17% of uninsured are self-employed, com-
pared with 9% of the insuredand to have a lower monthly income ($135 for
the un insured, compared with $159 for the insured). Te uninsured are also more
likely to be immigrants (11% of the uninsured, compared with 3% of the insured).
Immigrants tend to work in low paying informal sector jobs, and their employers
TABLE 5.2
Sociodemographic statistics for the insured and uninsured, 2006
Variable Defnition
Insured
N=3,988 (81.30%)
Uninsured
N=917 (18.70%)
Difference
insured
uninsured
Signif-
cance Mean
Standard
deviation Mean
Standard
deviation
Age Years 31.323 20.310 30.778 18.036 0.545
Sex Female 0.536 0.499 0.458 0.498 0.078 ***
Income Per capita
monthly
household
income (US$) 159.11 153.77 134.95 145.90 24.16 ***
1st quintile 27.41 16.78 26.30 18.16 1.11
2nd quintile 71.31 11.31 73.13 11.81 1.81
3rd quintile 115.17 14.77 114.77 13.40 0.41
4th quintile 185.39 26.17 183.68 23.60 1.71
5th quintile 401.86 190.82 378.27 225.25 23.59
Civil status Married 0.322 0.467 0.200 0.400 0.122 ***
Household head Household head 0.214 0.410 0.236 0.425 0.022
Education Primary 0.487 0.500 0.472 0.499 0.015
Secondary 0.377 0.485 0.420 0.494 0.043 **
University 0.113 0.317 0.096 0.295 0.017
Employment Self-employed 0.086 0.280 0.172 0.377 0.086 ***
Size of frm Small frms have
<10 workers
(0=small, 1=big) 0.073 0.260 0.021 0.145 0.052 ***
Nationality Costa Rican 0.967 0.179 0.894 0.308 0.073 ***
Urban status 1 = urban 0.684 0.465 0.662 0.473 0.022
Source: ENSA 2006.
BOX 5.1
Data and methodology
We use data from three sources: the
nationally representative National Health
Survey (ENSA 2006), the administrative
database of hospital discharges for 2006
(Caja Costarricense del Seguro Social
2006), and the nationally representative
Income and Expenditure Survey (IES
2004). For ENSA there are a total of 7,522
households. We removed 54 with missing
insurance status as well as 1,409 public
employees and their dependents and 1,154
pensioners (because neither group has a
choice whether to be insured, and our
population of interest is those who have
a choice). For the hospital data we also
removed public employees and pensioners,
which reduced the number of discharges
from 326,583 to 267,325. We deleted
these observations because the only people
with the ability to avoid enrolling are the
self-employed and private employers who
choose illegally not to enroll themselves or
their employees.
Our strategy for identifying the im-
pact of insurance was to nd instrumental
variables that would explain aliation
with the Caja but not health or nancial
protection behavior. Depending on the
dataset, we used a subset of the follow-
ing for this purpose: size of rm where
the individual is employed (we used the
mean size within a household), sector of
economic activity, occupation type, self-
employment, having multiple jobs, and
canton code. No exclusion restrictions are
perfect, but we believe that in the context
of Costa Rica, these variables would af-
fect the dependent health variables only
through their impact on aliation with
the Caja.
Insured
018
26%
3054
34%
5564
7%
1929
28%
65 and older
5%
Uninsured
018
34%
3054
31%
5564
10%
1929
19%
65 and older
6%
FIGURE 5.1
Age structure of the insured and uninsured, 2006
Source: ENSA 2006.
96 Chapter 5
often do not pay taxes for them. Te authorities have few tools to detect and pre-
vent self-exclusion and to collect premiums for or from independent or itinerant
workers.
Insured people tend to have a higher burden of disease (a metric that summarizes
mortality and morbidity conditions in a determined population) and higher preva-
lence of specic diagnosed diseases, such as diabetes, hypertension, and asthma than
uninsured people (table 5.3). Tis does not imply a negative impact of insurance on
health status (a causal relationship cannot be inferred), but it can be considered that
if a person feels healthy, they are more inclined to believe that the benets of paying
insurance premiums outweigh the nancial cost. Of course, it is also consistent with
underdiagnosis in the uninsured population. But there is no statistically signicant
dierence between the two groups self-perceived health status.
Tere is no statistically signicant dierence between the insured and unin-
sured for general utilization statistics (see table 5.3). For outpatient services, 65%
of the insured and the uninsured report having a visit during the last year. For hos-
pital services, 5% in both groups report a hospital admission for at least one night
in the last year. Tere is only one signicant dierence in use, but it is important:
while 49% of insured women ages 40 and older report having received a mammo-
gram, fewer than 40% of uninsured women report having received one. Given the
need for a patient to be referred to a high technology diagnostic imaging appoint-
ment, it is logical this would be an area where the uninsured are at a disadvantage.
According to the database of hospital discharges in 2006, 16% of people
discharged were uninsured, compared with 19% according to household survey
data (table 5.4). With childbirth as a leading cause of admissions, substantially
more women than men are in the hospital discharge database than in the ENSA
household data. But dierences in insurance status are not large. ENSA shows that
85.3% of men and 79.7% of women are insured, while at discharge, 81.2% of men
and 85.4% of women were insured.
As with the survey data, the discharge data show that the uninsured were more
likely to be unmarried, but there is a signicant dierence between the percentage
of survey respondents and discharges who declare they are cohabitating. Only
6% of insured survey respondents state they are cohabitating, while 19% of insured
discharges say they are. Tere is an incentive to claim cohabitation with a Caja
member because it qualies the patient for dependent benets.
Te discharge data paint a more nuanced picture of dierences in behavior
between the insured and uninsured. Te rst is how they are admitted. Some
87% of the uninsured were admitted to hospitals through the emergency room,
compared with 58% of the insured. While 39% of the insured are referred to the
TABLE 5.3
Health status and use of insured and uninsured people, 2006
Variable Defnition
Insured
N=3,988 (81.30%)
Uninsured
N=917 (18.70%)
Difference
insured
uninsured
Signif-
cance Mean
Standard
deviation Mean
Standard
deviation
Chronic
disease
More than one diag-
nosed disease 0.269 0.444 0.147 0.354 0.122 ***
Hypertension Diagnosed arterial
hypertension 0.095 0.293 0.039 0.195 0.056 ***
Diabetes Diagnosed diabetes 0.032 0.177 0.018 0.132 0.014 **
Asthma Diagnosed asthma or
bronchitis 0.041 0.198 0.009 0.095 0.032 ***
Diagnosed
disease status
Index, 0, lowest
burden of disease, to
100, highest burden
of disease 4.565 10.665 2.671 7.955 1.894 ***
Self-reported
health
Scale, 1, very good,
to 5, very bad 2.085 0.707 2.124 0.813 0.039
Visit 1 = person visited
doctor at least once
during last year 0.647 0.478 0.655 0.476 0.009
Hospitalization 1 = person hospital-
ized at least one night
in last year 0.052 0.221 0.050 0.219 0.001
Emergency 1 = person used
emergency services
at least once during
last year 0.117 0.321 0.102 0.303 0.015
Mammogram Woman ages 40 and
older received mam-
mogram 0.490 0.500 0.397 0.491 0.094 **
Cytology Woman ages 18 and
older received pap
smear 0.934 0.248 0.921 0.271 0.014
Vaccines Person under age
18 completed
vaccinations 0.826 0.379 0.812 0.392 0.015
Diabetes
medicine
Diabetics took
diabetes medicines in
the two weeks prior to
the study 0.733 0.443 0.706 0.462 0.027
Hypertension
medicine
Hypertensives took
hypertension medicines
in the two weeks prior
to the study 0.734 0.442 0.765 0.427 0.031
Source: ENSA 2006.
98 Chapter 5
hospital through an outpatient provider, only 8% of the uninsured follow this
route. What happens to them in the hospital is also dierent. About 17% of the
insured and 21% of the uninsured have minor surgery, but 16% of the insured
undergo major ambulatory surgery, compared with only 2% of the uninsured.
Te uninsured experience considerably longer stays (5.2 days, compared with 3.8
days) and are far more likely to end up in the intensive care unit (3% of uninsured
TABLE 5.4
Comparison of insured and uninsured people who have been discharged
from a hospital, 2006
Variable Defnition
Insured
N=224,800 (84.1%)
Uninsured
N=42,525 (15.9%)
Difference
insured
uninsured
Signif-
cance Mean
Standard
deviation Mean
Standard
deviation
Age Years 28.48 18.567 28.23 19.662 0.253 **
Sex Female 0.700 0.458 0.634 0.482 0.066 ***
Civil status Married 0.358 0.479 0.186 0.389 0.172 ***
Nationality Costa Rican 0.930 0.254 0.771 0.420 0.159
Death Discharged dead 0.0097 0.0978 0.0296 0.170 0.0199 ***
Average length
of stay
Days
3.758 5.026 5.238 6.951 1.480 ***
Number of
previous
admissions 0.047 0.246 0.026 0.177 0.022 ***
Number of
medical visits 1.339 2.892 1.062 2.456 0.277 ***
Admission
source
Outpatient care
0.387 0.487 0.082 0.274 0.305
Emergency room 0.584 0.493 0.874 0.332 0.290 ***
Childbirth 0.029 0.168 0.044 0.206 0.015 ***
Admission
service
Medicine
0.098 0.298 0.136 0.343 0.038 ***
Surgery 0.167 0.373 0.209 0.407 0.043 ***
Gyno-obstetrics 0.407 0.491 0.399 0.490 0.008 ***
Pediatrics 0.162 0.368 0.168 0.373 0.006 ***
Major ambulatory
surgery 0.156 0.363 0.017 0.130 0.139 ***
Psychiatry 0.008 0.087 0.040 0.197 0.033 ***
Intensive care unit 0.002 0.048 0.029 0.167 0.027 ***
Source: Hospital discharges database.
admissions, compared with less than 0.3% of insured). Te ENSA data show that
the insured appear to be in worse health status than the uninsured (only 14% of the
uninsured report having more than one diagnosed chronic condition, compared
with 29% of the insured). Finally, the uninsured have a statistically signicant
higher chance of inpatient death (3%, compared with only 1% for the insured).
In short, despite being healthier, the uninsured enter hospitals directly in more
apparent trauma and with less previous attention and planning than the insured.
Tey are signicantly more likely to die. Tese results have implications for patient
health outcomes and for health care costs, as hospitals are by far the most expensive
places to receive care.
What are the determinants of insurance status?
Probit analysis indicates that insurance status depends primarily on age, employ-
ment, nationality, education, and marital status (table 5.5). Not being married
signicantly decreases the chances of being insured (being married increases an
individuals chance by 12 percentage points). Being an immigrant decreases the
likelihood of being insured by 19 percentage points relative to being a native Costa
Rican, and having less than a secondary education also decreases the probability.
Another large contributor is self-employment. Compared with being unem-
ployed (individuals employed by the government and pensioners are excluded from
the analysis because they have no choice whether to be covered), the likelihood of
being insured if an individual is self-employed is 12 percentage points lower, again
pointing to the independent workers decision to self-select out of the insurance
plan to avoid paying the taxa problem in all payroll taxnanced systems, made
worse as tax rates rise.
Impacts of health insurance
Tis section discusses the impact of insurance coverage on a number of outcome
variables, based on the estimation strategy shown in the appendix.
Does health insurance aect access and use?
Using the ENSA 2006 data, we explored the impact of insurance for the full sam-
ple and for subsamples of the household data. Tere is no evidence of a statistically
signicant dierence between the insured and uninsured in the use of outpatient
care, hospitalizations, or emergency services. Te lack of a dierence extends to
insured and uninsured individuals from subsamples of the poorest 40% and the
wealthiest 40%. Tis result reinforces the descriptive data, which show few dier-
ences between the insured and the uninsured.
3
100 Chapter 5
We also analyzed subsamples with chronic conditionsone group diagnosed
with at least one disease, one group diagnosed with hypertension, and a third diag-
nosed with diabetes. Tere is a statistically signicant impact only for diabetes:
health insurance reduces the probability of both inpatient care and emergency
room care for diabetics. Uninsured diabetics (all other things being equal) are more
likely to end up in an inpatient bed or the emergency room. In addition, insurance
reduces the use of medicines for the diabetic population. Tese ndings suggest
that insurance coverage in Costa Rica results in better and safer management of
diabetes, probably associated with primary care.
TABLE 5.5
Probit analysis of the determinants of participationdependent variable:
insurance status, 2006
Variable Partial effect Standard error Signifcance
Sociodemographic
Age 0.004 0.002 *
Sex (women=1) 0.013 0.013
Log of the household monthly income per capita 0.018 0.008 **
Civil status (reference = single)
Married (married=1) 0.122 0.022 ***
Relationship to household head
Head of household (head of household=1) 0.017 0.015
Education (reference = without formal education, primary incomplete and primary complete)
Secondary 0.027 0.015 *
University 0.033 0.017 *
Employment category (reference = not employed)
Patron 0.077 0.062
Self-employed 0.125 0.060 **
Private 0.017 0.041
Nationality
Costa Rican (Costa Rican=1) 0.192 0.042 ***
Urban zone (urban=1) 0.007 0.013
Quality of health care services (self-reported) 0.022 0.006 ***
Reference population in the EBAIS to attend 0.000 0.000
Number of observations 3,070
Pseudo R
2
0.0668
Lod pseudo log likelihood 1,428.41
Pred. P. 0.871
Source: ENSA 2006.
Health insurance does not appear to have any impact on access to diagnostic tests
like pap smears in women ages 18 and older and mammograms in women ages 40
and older. Te mammography result is dierent from the descriptive analysis, indi-
cating that controlling for other factors (such as education) eliminated the disparity
between the insured and uninsured. In contrast, health insurance improves the likeli-
hood of completing the full series of vaccinations for children age 18 and younger.
From the 2006 hospital discharge data analysis, the insured were signicantly
less likely to access the hospital through the emergency room, consistent with the
descriptive data. Te likelihood of having an avoidable hospitalization for a condi-
tion that could be managed in an outpatient setting was signicantly less for the
insured. Tis result, consistent with the household data result for diabetes, again
points to the importance of quality of health care services over quantity. Because
no one is denied care in Costa Rica, uninsured people who are sick are generally
able to receive treatment, but apparently often later than they should and in a less
than optimal setting. For health outcomes and cost-eectiveness, the uninsured
are in a position inferior to the insured, who are enrolled with a primary care pro-
vider through their EBAIS, and therefore receive appropriate preventive and main-
tenance care. Tat, in turn, reduces chronic disease complications and expensive
hospitalizations for those diseases.
Does health insurance aect health status?
ENSA includes the variable self-reported health status, which reects whether
individuals describe their health as very good, good, okay, bad, or very bad. Over-
all, health insurance signicantly improves an individuals self-perception of health
status. But it reduces the self-perceived health status for diabetics, interesting
because insured diabetics are less likely to need hospital services. Perhaps insured
diabetics are more educated about the serious complications associated with their
disease and thus consider themselves to be in a worse state of health.
From the hospital discharge data analysis, we have more objective information
on health status. Insured mothers are less likely to have babies with low birthweight,
which would be consistent with better access to prenatal care through EBAIS. In
addition, if we measure severity of illness by the number of days of hospitalization,
insured people experience substantially shorter hospital stays. When insured and
uninsured people are hospitalized, the insured are healthier by this measure.
Does health insurance aect out-of-pocket expenditures?
Te impact of health insurance on nancial protection of insured people was
estimated using a third data source, the nationally representative Income and
102 Chapter 5
Expenditure Survey of 2004. We estimated per capita out-of-pocket spending
on health as a proportion of per capita expenditures and as a proportion of pay-
ment capacity, dened as total household expenditures minus household food
expenditures.
Te average monthly out-of-pocket health expenditure by Costa Ricans in
2004 was US$8.50, but with a high degree of variability (coecient of variation
was estimated at 321%). Per capita out-of-pocket health expenditures represent
nearly 3% of per capita expenditure and 3.5% of the payment capacity of indi-
viduals. Out-of-pocket health expenditures represent only 2% of the poorest third
of the populations per capita expenditures but 4% of the wealthiest thirds, the
reverse of the usual tendency in the absence of eective nancial protection.
Our analysis found no signicant impact of health insurance on a Costa
Ricans out-of-pocket health expenditures. It is likely this result is due largely to
the fact that no person can be denied care in Costa Rica. An individual who arrives
at an emergency room needing to be admitted to hospital is admitted, regardless
of the ability to pay. So there is no dierence in out-of-pocket health expenditures
between insured and uninsured individuals.
Conclusions
Te main distinguishing characteristic of health insurance in Costa Rica relative to
other countries of Latin America and many other middle-income countries is that
approximately 81% of the population is aliated with the Caja and thus covered,
but even those who are not covered are guaranteed access to emergency and hospi-
tal care provided by the Caja when they are sick or need care. Tey are not shunted
into a separate lower quality system. Terefore everyone is covered by catastrophic
insurance; the major dierence is that the 19% not aliated with the Caja do not
benet from assignment to a primary care provider and must seek and pay for those
services in the market.
Te uninsured are somewhat less educated and more likely to be immigrants,
have lower income, be self-employed, and come from healthier age groups than
the insured, but in their overall use of health care resources they are similar. If we
had only the household data, we would have concluded that the insured gained no
advantage over the uninsured except for a higher probability of children receiving
all immunizations and better care for diabetics.
However, the hospital discharge data raise concerns. Te uninsured are far
more likely to enter the hospital through the emergency room. Tey are likely to
have surgery but not to have a planned major surgery; even so, they experience a
36% longer stay. We estimate that simply reducing their length of stay to that of
the insured would save about US$8.5 million, or about US$100 per uninsured
probably enough to nance a reasonable level of access to primary care for them.
Te uninsured are more likely to be hospitalized for a condition that can be
managed in an ambulatory care setting, to end up in the intensive care unit, and
to be discharged dead. Te hospital data analysis conrms a higher probability of
emergency room and inpatient care for uninsured diabetics.
We expected to nd few dierences between the insured and uninsured in
Costa Rica because of the equal access rule. For measures of nancial protection
and use of services, our expectations were met. Yet in signicant ways the unin-
sured are disadvantaged from a health standpoint. Tey use medical care resources
more haphazardly than the insured. We hypothesize that this happens principally
because the insured enter the pyramidal Caja system at the bottom, or primary
care level, while the uninsured tend to enter closer to the top. Perhaps, in light of
these ndings, it would be possible for the Caja to reallocate resources to cover the
uninsured in a more health-friendly manner.
104 Chapter 5
Appendix: Details of the Estimations
Te estimated models when we use the ENSA 2006 database are the following:
First stage:
i
(insured
i
) = a
0
+ a
1
sex
i
+ a
2
age
i
+ a
3
linc
i
+ a
4
married
i
+ a
5
hoh
i
+ a
6
secondary
i
+ a
7
univ
i
+
a
8
costarican
i
+ a
9
employed
i
+ a
10
zone
i
+
r=11
15
a
r
region
ri
+ a
16
iv
i
+ a
17
disease_status
i
+
a
18
ebaisxh
i
+ e
i
Second stage:
H
i
= b
0
+ b
1
sex
i
+ b
2
age
i
+ b
3
linc
i
+ b
4
married
i
+ b
5
hoh
i
+ b
6
secondary
i
+
b
7
univ
i
+ b
8
costarican
i
+ b
9
employed
i
+ b
10
zone
i
+
r=11
15
b
r
region
ri
+ b
16
i
(insured
i
) +
b
17
disease_status
i
+ b
18
ebaisxh
i
+ m
i
where H
i
is the result variable, sex is a dummy variable that takes the value 1 if the
individual is a woman and 0 if the individual is a man, age is the age of the indi-
vidual, linc is the logarithm of per capita household income, married is a dummy
variable that takes the value 1 if the individual is married and 0 otherwise, hoh is
a dummy variable that takes the value 1 if the individual is the head of the house-
hold and 0 otherwise, secondary (univ) is a dummy variable that takes the value 1 if
the individual has secondary (university) completed as his highest education level,
costarican is a dummy variable that takes the value 1 if the individual is a citizen
and 0 otherwise, employed is a dummy variable that takes the value 1 if the person
works and 0 otherwise, region
r
are regional dummy variables, insured is a dummy
variable that indicates the coverage of the health insurance system (
i
(insured
i
) is
the estimated value of this variable from the rst stage), disease_status is an index
that shows health status, and ebaisxh is the number of people in each health area.
For self-reported health status models we add other independent variables: consult
(a dummy variable that takes the value 1 if if the individual has visited a clinic
or hospital during the last year and 0 otherwise), hospital (a dummy variable that
takes the value 1 if the individual has been hospitalized at least one night during
the last year and 0 otherwise), and emergency (a dummy variable that takes the
value 1 if the person used emergency services at least once during the last year and
0 otherwise).
Te estimated models when we use the 2006 hospital discharge database are
detailed as follows:
First stage:
i
(insured
i
) = a
0
+ a
1
sex
i
+ a
2
age
i
+ a
3
married
i
+ a
4
costarican
i
+ a
5
los
i
+ a
6
previous
i

+
r=11
13
a
p
prov
pi
+ a
16
iv
i
+ e
i
Second stage:
H
i
= b
0
+ b
1
sex
i
+ b
2
age
i
+ b
3
married
i
+ b
4
costarican
i
+ b
5
los
i
+ b
6
previous
i
+
p=7
13
b
p
prov
pi

+ b
14
i
(insured
i
) + m
i
where H
i
is the result variable, sex is a dummy variable that takes the value 1 if the
individual is a woman and 0 if the individual is a man, age is the age of the individ-
ual, married is a dummy variable that takes the value 1 if the individual is married
and 0 otherwise, costarican is a dummy variable that takes the value 1 if the indi-
vidual is a citizen and 0 otherwise, los is the number of days an individual remained
in a hospital, previous indicates the number of earlier entrances to the clinic or
hospital, prov
p
are province dummy variables, and insured is a dummy variable that
takes the value 1 if the individual is covered by the health insurance system and 0
otherwise (with
i
(insured
i
) being the estimated value from the rst stage).
Notes
1. Te market is to be opened in 2009 and 2012, depending on the type of insurance
plan.
2. Te Free Trade Agreement breaks this monopoly so international insurance companies
can now arrive in Costa Rica, operate their oces there, and oer the public not only
health insurance but life and automobile insurance.
3. Insurance has a puzzlingly signicant negative impact on use of outpatient care for
nonCosta Ricans, but that is for a tiny subsample of only 175.
106
Health Insurance and Access to
Health Services, Health Services
Use, and Health Status in Peru
Ricardo Bitrn, Rodrigo Muoz, and Lorena Prieto
C
h
a
p
t
e
r

6
Tis chapter examines how Perus eight-year-old Integral Health Insur-
ance (Seguro Integral de Salud, or SIS) has aected access to health ser-
vices and out-of-pocket spending by its beneciaries. We use data from
the Demographic and Health Survey (DHS), which contains two cross-
sectional samples, one for 2000 and another for 200407, with a sample
spread over ve years. We also use data from the National Household
Survey (ENAHO), a panel collected over 200206.
Classied by the World Bank as a lower middle-income country in
2008, Peru had per capita gross national income of US$7,950 annually
in purchasing power parity terms. Life expectancy at birth is 73 years,
under-ve mortality is 24 per 1,000 in 2008, and 99% of children
receive the full course of three doses for their DPT vaccinations by their
23rd month. Total health spending is about 4.3% of gross domestic
product (GDP), about 58% of it public and 42% private. Nearly 75% of
private spending is out of pocket (World Bank 2010).
The Peruvian health sector
Perus mixed health system includes a social securitynanced sec-
tor (EsSalud), a tax-nanced public sector (the Ministry of Health, or
MINSA), a private sector, and a national police and armed forces sec-
tor. Te private sector oers health plans with dierent copayments,
deductibles, and ceilings. Te main insurers in the system are EsSalud,
SIS, and private insurance companies. EsSalud covers about 17% of the
Health Insurance, Access to Health Services, and Health Status in Peru 107
population; SIS, 15%; the national police and armed forces, about 1.6%; and the
private sector, 1.7% (Portocarrero, Margarita, and Vallejo 2007). Health service
providers in the system are MINSA, which is the largest provider network; the par-
allel EsSalud system; private providers; and the national police and armed forces.
Each of these systems provides comprehensive services from the primary level up to
complex hospital care. Te Health Providing Institutions (Entidades Prestadoras de
Salud, or EPS) providers, explained below, deliver mainly low complexity services
to supplement EsSalud.
Health social security
EsSalud is the typical social security health insurance system in Latin America. It
is nanced with a 9% income-based contribution paid by employers and directly
provides health services to its beneciaries through its own network of clinics and
hospitals. Beneciaries are mainly formal sector workers and their familiescalled
regular insureeswhich account for 75% of all beneciaries. EsSalud also has
three other types of insurees: retired (17%), subsidized (5%), and elective (3%).
Retirees contribute 4% of their pensions. Subsidized insurees include shers, agrar-
ian workers, and home workers.
1
Voluntary insurees are those who purchase an
EsSalud health plan and pay a premium based on the health plan they choosein
competition with private health insurance plans. Voluntary insurees are typically
self-employed.
EsSalud has an explicit benet package (Plan Mnimo de Atencin) of 752
diagnoses, which makes it comprehensive rather than minimal. Te package
includes preventive, promotional, and curative care; welfare and social promotion
services; subsidies for temporary disability and maternity; and burial services. It is
the same for all but voluntary beneciaries, who are entitled to a reduced benets
package. EsSalud provides services that cover at least this package through its own
provider network or through contracts with other health service providers.
Integral Health Insurance
SIS began in 2001 by merging two health insurance programs: free health insur-
ance for children in public schools (Seguro Escolar Gratuito, or SEG), which was
launched in 1997, and the maternal and child health insurance (Seguro Materno
Infantil, or SMI), which was launched in 1998 and gradually expanded from two
provinces to the whole country. Both programs were targeted to specic populations:
SEG to children ages 317 in public schools, excluding private school children, and
SMI to all pregnant women and children up to age 4 if they were not covered by
another insurance system. Tese programs created for the rst time a demand side
108 Chapter 6
subsidy that was paid to MINSA providers when services were provided in lieu of
simply increasing MINSAs budget to try to reach these groups more eectively.
SIS is an agency under MINSA with its own allocation of resources from the
Ministry of Economics and Finance. It has ve plans: Plan A for children ages 04;
Plan B for children and adolescents ages 517; Plan C for women before, during, and
after childbirth; Plan D for adults in emergencies; and Plan E for poor adults. A new
partially subsidized or contribution plan has been added to extend the program to
households with some capability to pay. Nevertheless, SIS remained fundamentally a
maternal and child health insurance program during the period of our analysis, with
93% of its resources going to Plans A, B, and C in 2004 (Cotlear 2006; Jaramillo
and Parodi 2004). Its beneciaries seek care from MINSAs provider network and
make no copayments, because SIS oers full nancial coverage. In turn, SIS pays
MINSA providers for these interventionsbased on a price list previously agreed
with MINSAthrough a fee-for-service mechanism that covers only the variable
costs of care. SIS administrative data show that the number of members increased
from 3.6 million in January 2007 to 7.3 million in January 2008 to 12.7 million in
the rst half of 2010 (see www.sis.gob.pe). In June 2007 SIS covered approximately
33.1% of its target population (based on ENAHO 2004 incidence of poverty).
Eligibility to become beneciaries is explicitly dened and targeted to the
poor and uninsured. Individuals must go to the nearest MINSA health facility
and request aliation by presenting their national identication card. SIS rep-
resentatives then apply the socioeconomic evaluation form (Ficha de Evaluacin
Socio Econmica) to determine whether they are extremely poor, poor, or non-
poor. Membership is automatic for those who are categorized as poor or extremely
poor and who sign a contract with a SIS representative. With the partially subsi-
dized plan, SIS expanded its target population to include all poor families with the
same benets package (preventive and ambulatory care are emphasized, but it also
includes an array of priority hospital services), but with a monthly premium that
varies between S/.10 ($3.50) and S/.30 ($10.00), depending on income and whether
it is for an individual or a family.
It is important to keep in mind, especially when the results of the analysis
are presented, what a limited program SIS is. Te covered population already has
access to MINSA facilities, which they are required to use to benet from SIS. Lit-
tle has changed in the operation of MINSA facilities. Employees are civil servants
whose salaries are paid as usual. Basically the only dierence from regular opera-
tions is that for the covered services, the facility receives an extra payment from SIS
when it can show it delivered the service to an eligible member of SIS, creating an
incentive to provide the covered services to SIS beneciaries.
Health insurance population coverage in the household surveys
EsSalud and SIS are the health insurers with the highest population coverage,
together covering 34.8% of the population (table 6.1). An additional 4% of the
population is covered by other types of health insurance, including the national
police and armed forces, private insurance plans, and university and private school
insurance. Overlapping coverage between EsSalud and SIS and between SIS and
other health insurances is negligible; however, about 5% of EsSalud beneciaries
overlap with other types of health insurance. SIS beneciaries are concentrated in
areas with higher informal labor and poverty, like the mountains and jungle. Te
presence of SIS in these areas helps close the gap in EsSalud population cover-
age in these regions, yet 62% of the population is still not covered by any type of
insurance.
EsSalud beneciaries are concentrated in the coastal region, with more than
half living in Metropolitan Lima. Te percentage of the population covered by
EsSalud is highest in Metropolitan Lima, reecting the higher formal employment
rate in this region. Regions with less formal labor, like the mountains and the jun-
gle, have lower EsSalud coverage rates. Coverage of EsSalud has remained relatively
constant over 200007. Although not presented here, coverage estimates based on
DHS data are almost identical.
Descriptive information
SIS beneciaries are mainly poor families. Survey data show that most beneciaries
are children, because the SIS benet package is focused on maternal-child inter-
ventions. In contrast, EsSalud beneciaries are mainly formal sector workers and
their families. Tey have a slightly older age structure than the general popula-
tion and the uninsured. Tey are concentrated in the richer quintiles of per capita
expenditure. Te uninsured are distributed more homogeneously (except for the
richest quintile, which has a lower percentage of uninsured than the rest).
TABLE 6.1
Health insurance population coverage by region, 2006 (%)
Region EsSalud SIS Any health insurance
Metropolitan Lima 27.5 6.1 41.1
Coast 22.4 13.4 37.8
Mountains 11.5 22.7 35.2
Jungle 10.0 27.5 38.4
Total 18.4 16.4 37.9
Source: ENAHO.
110 Chapter 6
Table 6.2 presents descriptive data for several health service demand indica-
tors. Since we are not controlling for the variables used in the regression analyses,
the dierences between indicators do not necessarily reect causality by insurance
TABLE 6.2
Use of health services, by type of health insurance, 2005 and 2006 (%)
Indicator Uninsured
EsSalud
only SIS only Other Total
ENAHO 2006
Presence of chronic disease 22.7 32.7 10.1 25.9 22.5
Population with health problem in the four weeks
prior to the survey 52.3 53.0 51.3 43.4 51.9
Hospitalized in the 12 months prior to the survey 3.4 6.7 2.6 6.7 4.0
Sought formal care (with health professional) in
the four weeks prior to the survey 19.5 45.3 46.8 46.9 29.4
Use of child growth monitoring in the three
months prior to the survey 1.5 3.1 18.5 2.8 4.6
Use of family planning in the three months prior
to the survey 4.3 3.4 0.5 1.5 3.4
Use of vaccines in the three months prior to the
survey 17.8 16.4 38.0 18.6 20.8
Use of child iron supplements in the three
months prior to the survey 0.4 0.7 4.1 0.7 1.1
Use of disease prevention in the three months
prior to the survey 1.2 2.3 1.6 2.6 1.5
DHS 2005
Womens health
Pap-smear exam in the fve years prior to the
survey 38.7 62.6 29.3 66.4 43.9
Breast exam in the fve years prior to the survey 16.1 42.3 11.0 43.9 21.4
Number of prenatal visits
None 8.3 2.1 11.5 0.6 7.2
14 36.7 19.6 34.8 9.0 32.9
5 or more 55.0 78.4 53.7 90.5 59.8
Person providing assistance at delivery (among women with births)
No assistance 0.3 0.0 1.3 0.0 0.3
Doctor 35.5 69.5 20.9 92.8 42.1
Midwife 30.1 25.2 29.9 7.2 28.5
Nurse, promoter, or feldworker 2.7 0.5 8.8 0.0 2.6
Comadrona or partera 16.2 2.5 21.8 0.0 13.8
Other 15.2 2.4 17.3 0.0 12.8
Immunization
Children ages 1859 months fully immunized 47.5 64.7 64.7 70.7 58.8
(continued)
Indicator Uninsured
EsSalud
only SIS only Other Total
Children with diarrhea
Children under age 5 with diarrhea in the four
weeks prior to the survey 16.3 9.5 18.1 9.8 15.9
Sought formal treatment for diarrhea 27.5 44.2 46.1 37.4 39.2
Place of treatment
Traditional healer, pharmacy, at home, or with
friend 37.4 2.6 15.8 29.9 21.8
Community health worker 0.7 0.0 0.8 0.0 0.7
MINSA facility 48.9 13.1 78.4 0.0 62.1
EsSalud facility 0.0 44.4 0.0 0.0 3.5
Army or private facility 0.0 0.0 0.0 0.0 0.0
Other type of health facility 13.0 37.0 1.4 70.1 9.5
Treated in more than one type of health facility 0.0 2.9 3.6 0.0 2.4
Children with acute respiratory infection
Children under age 5 with acute respiratory
infection in the four weeks prior to the survey 16.7 18.3 20.4 11.9 18.5
Sought formal treatment for acute respiratory
infection 54.7 70.4 69.0 90.2 65.4
Place of treatment
Traditional healer, pharmacy, at home, or with
friend 21.6 11.8 8.9 10.5 13.1
Community health worker 0.0 0.0 0.7 0.0 0.4
MINSA facility 45.0 23.4 88.0 7.1 62.7
EsSalud facility 0.0 34.2 0.0 0.0 5.3
Army or private facility 3.8 0.0 0.0 0.0 1.1
Other type of health facility 20.5 21.8 2.2 49.7 12.2
Treated in more than one type of health facility 9.1 8.9 0.3 32.6 5.3
Family planning
Current use of modern family planning methods 30.4 40.1 5.3 39.5 31.1
Number of growth monitoring visits (children ages 14)
0 10.0 2.8 5.7 0.0 6.4
1 8.3 2.8 4.8 2.2 5.6
2 5.1 4.4 5.3 2.1 4.9
3 6.7 3.8 8.0 3.0 6.7
4 29.6 33.3 31.5 14.9 30.4
5 40.4 52.9 44.7 77.9 45.9
Source: ENAHO and DHS.
Use of health services, by type of health insurance, 2005 and 2006 (%)
112 Chapter 6
aliation. For example, the higher use of child growth monitoring among SIS ben-
eciaries probably appears because SIS beneciaries are much younger than those
of EsSalud and the general population. Another indicator the probability of seek-
ing formal care in the four weeks prior to the surveyis less aected by demo-
graphic variables and shows practically no dierence between SIS and EsSalud.
It is interesting to note, however, that SIS immunization rates for children are
much higher than those for the uninsured and that nearly half of SIS children
with diarrhea were taken for treatment, almost exclusively at MINSA facilities. Te
same is true for children with acute respiratory infections. SIS mothers match or
exceed all others in using prenatal care and growth monitoring for their children.
As we will see, there are interesting wrinkles in each of these ndings when other
variables are controlled.
The effects of Seguro Escolar Gratuito/Seguro Materno Infantil and
Seguro Integral de Salud health insurance coverage on health care
results using Demographic and Health Survey data
We were able to use the 2000 and 200407 DHS. Because of limitations, only a
cross-sectional regression could be done for 2000 and for 200407. In 2000 the
SEG/SMI programs were in place; in 200407 they had been merged to create SIS.
Moreover, the 200407 surveys did not report child insurance status, so that was
imputed based on the mothers status and rules governing which children would be
covered. Te full specication is shown in table A6.1 in the appendix; only the key
results are discussed in this section.
We know that SIS is targeted, but we do not know the algorithm for the test. To
limit the analysis to women who would be eligible for SIS under a perfect targeting
regime, we removed the upper two quintiles from the analysis. In the DHS wealth
variables are used to create these quintiles. In DHS-based regressions we cannot
control for the endogeneity, if it existed for participation in SIS, so we ran univariate
regressions where participation in SIS is simply included as a dummy variable.
Probability of being fully immunized (children ages 1859 months)
Health insurance has a positive association with being fully immunized for chil-
dren aged 1859 months. Table 6.3 shows both descriptive statistics and results
of a logit regression. Te descriptive statistics show that the insured population
has higher coverage of full immunization among children than the uninsured but
eligible population. Te logit regression, which controls for other factors that aect
the immunization coverage besides health insurance, conrms that the eect of
health insurance is positive: SEG/SMI health insurance increases the probability
of being fully immunized by 4 percentage points in 2000, and SIS does so by 14
percentage points in 2004.
Te eect of SEG/SMI decreases gradually with income and is statistically sig-
nicant only in quintile 1. In contrast, the eect of SIS increases steadily with
income, yet the eect of SIS is larger than the eect of SEG/SMI even in quintile 1.
Probability of receiving pap-smear exam in last ve years (women ages 1549)
Te percentage of women receiving pap smears in the last ve years is lower among
the insured SIS population (29%) than among the uninsured but eligible popula-
tion (39%). When we control for other factors such as age, education, and marital
status in a logit regression, we show that the eect of health insurance is positive:
SIS is associated with a 3 percentage point increase in the probability of receiving
a pap-smear exam.
We found a statistically signicant interaction eect between SIS and age. Te
probability of receiving a pap smear exam always increases with age, but does so
faster with SIS health insurance. Tus, there is a positive eect of SIS, which is lower
among young women and higher among older women; in other words, SIS is associ-
ated with an increase in the probability of a pap smear when it is needed more.
TABLE 6.3
Seguro Escolar Gratuito/Seguro Materno Infantil and Seguro Integral de
Salud and the probability of being fully immunized (children ages 1859
months, %)
Subpopulation
SEG/SMI Health Insurance, 2000 SIS Health Insurance, 200407
All
Quintile
1
Quintile
2
Quintile
3 All
Quintile
1
Quintile
2
Quintile
3
Observed mean of dependent variable
Among population with
health insurance 67 67 68 64 65 62 68 66
Among uninsured but
eligible population 62 60 62 65 50 51 50 49
Among total population 63 61 63 65 60 59 62 57
Logit model (predicted mean of dependent variable)
Simulation with health insurance = 1 64 65 65 63 65 62 67 71
Simulation with health insurance = 0 60 57 61 64 51 52 52 49
Effect of health insurance
on dependent variable
(percentage points) +4*** +7*** +4 1 +14*** +10*** +14*** +22***
*** signifcant at p < 0.01.
Source: Authors calculations based on DHS 2000 and 2004 datasets.
114 Chapter 6
Probability of delivery attended by a skilled health professional (women ages
1549)
Results show no association between SEG/SMI or SIS and the probability of hav-
ing a delivery attended by a skilled health professional. Although the insured
population shows lower rates of assisted deliveries than the uninsured, these are
caused by other factors: income, location, education, gender of the household
head, marital status, and the number of children living in the household. When
these factors are controlled, we nd no statistically signicant eect of health
insurance.
Mean percentage of growth monitoring visits attended (children under age 5)
On average, SIS children attended 54% of their regular child growth monitoring
visits, compared with 44% for similar uninsured children (table 6.4). Te dier-
ence in favor of SIS children is observed across the three bottom quintiles. When
controlling for other variables, SIS is associated with a 9 percentage point increase
in child growth monitoring visits, with a much higher eect on the bottom quin-
tile. We nd a similar result for SEG/SMI in 2000.
We found a statistically signicant interaction eect between SIS and the edu-
cation of the mother. For all levels of mothers education, the predicted number
of growth monitoring visits is higher for SIS children, reecting the eect of the
health insurance variable. SIS also attenuates the impact of education. Te result is
that a child born to a woman with no education, if covered by SIS, is more likely to
have a complete cycle of growth monitoring or well-baby visits than if the mother
has 15 years of education but is not covered by SIS.
Probability of being formally treated for diarrhea (children under age 5)
SIS is associated with a probability of receiving formal treatment for diarrhea that
is 20 percentage points higher relative to the uninsured, with the eect statistically
signicant in the two bottom quintiles only. We sought possible interaction eects
between the health insurance variable and the childs gender and age, the mothers
age and education, and the households urban/rural location, but none of the inter-
action terms was signicant. Te eect is similar for SEG/SMI in 2000.
Probability of being formally treated for acute respiratory infection (children
under age 5)
SIS is associated with a 23 percentage point increase in the probability of receiving
formal treatment for acute respiratory infection, with the eect statistically signi-
cant in all quintiles. We found no signicant interaction eects.
Impact of Seguro Integral de Salud health insurance coverage on
health care results using ENAHO data
We were able to use the ENAHO data for 200206, which contain less informa-
tion about health than the DHS but more information on economic variables.
We use a 30% panel subsample for which we have ve years of data. Because SIS
members tend to join when they present themselves for care, we found that we
could use double-lagged health insurance status (two years prior to the survey) as
an identifying variable to control for endogeneity, leaving us with three years of the
panel for analysis (200406). As in the previous analysis, the sample is restricted
to the bottom three quintiles to reduce the possibility that we are comparing SIS
aliates to individuals who would not qualify for SIS.
Probability of seeking curative health care for symptoms, illnesses, or relapses
in the four weeks prior to the survey
Te probability of seeking curative health care, with a doctor or other qualied
health professional, for symptoms, illnesses, or chronic disease relapses in the four
TABLE 6.4
SEG/SMI and SIS and the mean percentage of growth monitoring visits
attended (children under age 5)
Subpopulation
SEG/SMI Health Insurance, 2000 SIS Health Insurance, 200407
All
Quintile
1
Quintile
2
Quintile
3 All
Quintile
1
Quintile
2
Quintile
3
Among population with health
insurance 41 40 44 41 54 50 55 60
Among uninsured but eligible
population 39 33 41 45 44 38 43 50
Among total population 39 34 42 44 50 47 51 56
Ordinary least squares model
Number of observations 8,060 3,762 2,640 1,658 3,140 1,405 1,066 669
Adjusted R
2
0.11 0.09 0.06 0.11 0.12 0.15 0.10 0.08
Predicted mean of dependent variable
Simulation with health insurance
= 1 44 41 45 50 54 50 54 59
Simulation with health insurance
= 0 37 32 40 43 45 37 46 52
Effect of health insurance on depen-
dent variable (percentage points) +7*** +9*** +6*** +7*** +9*** +13*** +8*** +6***
Source: Authors calculations based on DHS 2000 and 2004 datasets.
116 Chapter 6
weeks prior to the survey, excluding accidents. Even though there is a large discrep-
ancy between the insured and uninsured in likelihood of seeking care, we nd that
much of the dierence is explained by variables other than insurance. While insur-
ance coverage has a signicant, positive eect on the probability of seeking care,
its independent eect ranges from 19 percentage points in 2004 to 6 percentage
points in 2005 and 2006 (table 6.5).
Probability of spending a positive amount among those receiving formal care
in last four weeks, how much is spent, and catastrophic spending (more than
30% of household spending)
Te negative impact of SIS on the probability of spending any amount for those
receiving care is high. With SIS coverage, the predicted probability of spending
anything in 2004 is 13%, for example, compared with 86% for those without
coverage. As a result, SIS coverage reduces the probability of spending anything,
TABLE 6.5
Impact of Seguro Integral de Salud on probability of seeking curative
health care for symptoms, illnesses, or relapses in last four weeks (all
ages)
2004 2005 2006
Among population with health insurance 50 42 44
Among uninsured but eligible population 16 16 16
Among total population 24 22 21
Endogeneity tests
Value of rho in bivariate probit 0.204 0.143 0.312**
Signifcance of predicted health insurance in primary equation 0.196 0.424 0.114
Model results
Type of model probit probit biprobit
Number of observations 2,174 1,912 2,008
Pseudo R2 0.13 0.08 na
Predicted mean of dependent variable
Simulation with health insurance = 1 33 21 7
Simulation with health insurance = 0 14 15 1
Effect of health insurance on dependent variable
(percentage points) 19*** 6*** 6***
na not applicable.
Source: Authors calculations based on ENAHO 20022006 panel dataset.
reducing the sample sizes so much for those with any spending that further analysis
is not possible.
Summary and conclusions
SIS reduces in an important way the likelihood that those insured will have to
spend money out of pocket for health care. At the same time, SEG/SMI and SIS
are associated with increased use for a variety of services, both preventive and cura-
tive (table 6.6). Te biggest eect occurs in the case of formal treatment for diar-
rhea and acute respiratory infections for children under age 5. Yet it also increases
use of ambulatory care for all members when they are sick. Among preventive ser-
vices, the biggest positive eect of SIS is for immunizations, followed by growth
monitoring.
It is clear that SIS has achieved important gains for its beneciaries, in lower
out-of-pocket spending and higher use of services, but we were surprised to nd
TABLE 6.6
Summary effects of SIS health insurance (%)
Indicator
SEG/SMI SIS
Baseline Value
Pro-
poor
effect Baseline Value
Pro-
poor
effect
Probability of being fully immunized
(children ages 1859 months) 62 +4*** Yes 50 +14*** No
Probability of receiving pap-smear exam in
the fve years prior to the survey (women
ages 1549) 44 +7*** No 38 +3* No
Probability of having delivery attended by a
skilled health personnel (women ages 1549) 41 +4 No 56 +3 No
Mean percentage of growth monitoring
schedule completed (children under age 5) 39 +7*** No 44 +9*** Yes
Probability of being formally treated for
diarrhea (children under age 5) 34 +16*** No 29 +20*** Yes
Probability of being formally treated for acute
respiratory infection (children under age 5) 50 +15*** No 52 +23*** No
Probability of seeking curative health care
for symptoms, illnesses or relapses in the
four weeks prior to the survey 16 +6~+19***
Probability of spending a positive amount
among those receiving formal care in the
four weeks prior to the survey 86~92 67~81***
Source: Authors calculations based on DHS and Encuesta Demogrfca y de Salud Familiar.
118 Chapter 6
that SIS coverage does not have an eect on institutional deliveries, which is a goal
of the program. Te reasons will have to be understood better and addressed in
future policy changes. Assessing the causes of this partial failure should be a prior-
ity for SIS strategists.
Finally, we consider the eect of SIS to be pro-poor when its eect is greatest
in quintile 1 and lowest in quintile 3, and not pro-poor otherwise. For example,
the eect of SIS on the probability of being fully immunized is greater in quintile 3
than in quintile 1 and thus not considered pro-poor. Te eect on the probability
of seeking care for acute respiratory infection is constant in every quintile and also
not considered pro-poor. Te eect on growth monitoring visits, however, is high-
est in quintile 1 and lowest in quintile 3 and thus considered pro-poor. Although
all three bottom quintiles are considered poor and eligible for SIS, it is desirable
that the benets of SIS reach the poor progressively, or at least proportionally, with
their income. Cases where most of the eect reaches the less poor instead of the
extreme poor (as for immunizations) reect coverage inequalities that should be
addressed.
Appendix: Model specications
Table A6.1 shows the model specication for each dependent variable, indicating
which control variables were included in the regressions. To explore whether the
eects of health insurance vary between certain population groups, we included
interactions terms between the health insurance dummy variable and the following
variables: gender, age, years of education (of patient or patients mother, depending
on the age of the patient), and location.
120 Chapter 6
TABLE A6.1
Dependent variables and control variables
DHS ENAHO
P
r
o
b
a
b
i
l
i
t
y

o
f

b
e
i
n
g

f
u
l
l
y

i
m
m
u
n
i
z
e
d

(
c
h
i
l
d
r
e
n

a
g
e
d

1
8
5
9

m
o
n
t
h
s
)
P
r
o
b
a
b
i
l
i
t
y

o
f

r
e
c
e
i
v
i
n
g

p
a
p
-
s
m
e
a
r

e
x
a
m

i
n

l
a
s
t

5

y
e
a
r
s

(
w
o
m
e
n

1
5
4
9
)
P
r
o
b
a
b
i
l
i
t
y

o
f

h
a
v
i
n
g

d
e
l
i
v
e
r
y

a
s
s
i
s
t
e
d

b
y

a

d
o
c
t
o
r

(
w
o
m
e
n

1
5
4
9
)
P
e
r
c
e
n
t
a
g
e

o
f

c
h
i
l
d

g
r
o
w
t
h

m
o
n
i
t
o
r
i
n
g

s
c
h
e
d
u
l
e

c
o
m
p
l
e
t
e
d

(
c
h
i
l
d
r
e
n

u
n
d
e
r

5
)
P
r
o
b
a
b
i
l
i
t
y

o
f

b
e
i
n
g

f
o
r
m
a
l
l
y

t
r
e
a
t
e
d

f
o
r

d
i
a
r
r
h
e
a

(
c
h
i
l
d
r
e
n

u
n
d
e
r

5
)
P
r
o
b
a
b
i
l
i
t
y

o
f

b
e
i
n
g

f
o
r
m
a
l
l
y

t
r
e
a
t
e
d

f
o
r

A
R
I

(
c
h
i
l
d
r
e
n

u
n
d
e
r

5
)
P
r
o
b
a
b
i
l
i
t
y

o
f

s
e
e
k
i
n
g

c
u
r
a
t
i
v
e

h
e
a
l
t
h

c
a
r
e

f
o
r

s
y
m
p
t
o
m
s
,

i
l
l
n
e
s
s
e
s

o
r

r
e
l
a
p
s
e
s

i
n

l
a
s
t

4

w
e
e
k
s

(
a
l
l

a
g
e
s
)
P
r
o
b
a
b
i
l
i
t
y

o
f

s
p
e
n
d
i
n
g

a

p
o
s
i
t
i
v
e

a
m
o
u
n
t

a
m
o
n
g

t
h
o
s
e

r
e
c
e
i
v
i
n
g

f
o
r
m
a
l

c
a
r
e

i
n

l
a
s
t

4

w
e
e
k
s

(
a
l
l

a
g
e
s
)
A
m
o
u
n
t

s
p
e
n
t

b
y

t
h
o
s
e

w
i
t
h

p
o
s
i
t
i
v
e

s
p
e
n
d
i
n
g

(
a
l
l

a
g
e
s
)
P
r
o
b
a
b
i
l
i
t
y

o
f

s
p
e
n
d
i
n
g

o
n

h
e
a
l
t
h

m
o
r
e

t
h
a
n

3
0
%

o
f

t
o
t
a
l

h
o
u
s
e
h
o
l
d

e
x
p
e
n
d
i
t
u
r
e
,

e
x
c
l
u
d
i
n
g

s
u
b
s
i
s
t
e
n
c
e

n
e
e
d
s

(
e
x
t
r
e
m
e

p
o
v
e
r
t
y

l
i
n
e
)
Quintile 1 Omitted
Quintile 2
Quintile 3
Quintile 4
Quintile 5
Sex (Female = 1)
Age
Age squared (100)
Years of education
Married/concubinate
Female household head
Age of mother
Years of education of
mother

Mother is married/
concubinate

Number of children under
age 5
Owner of dwelling Omitted
Lives in urban area
Lives in coast region
Lives in mountains region
Lives in Lima region Omitted
Lives in higher jungle
Omitted
Lives in lower jungle
Health post in community Not available
: indicates variable included in the regression.
: indicates variable included in the regression, plus a term to capture its interaction with the
health insurance dummy.
Source: Authors.
Note
1. Te subsidized scheme was implemented in phases: for shers it began in 1997, for
agrarian workers in 2002, and for home workers in 2005.
References
Cotlear, D. (ed.). 2006. A New Social Contract for Peru: An Agenda for Improving Education, Health
Care, and the Social Safety Net. Washington, DC: World Bank.
Jaramillo, M., and S. Parodi. 2004. El Seguro Escolar Gratuito y el Seguro Materno Infantil:
Anlisis de su incidencia e impacto sobre el acceso a los servicios de salud y sobre la equidad en
el acceso. Documento de Trabajo 46Lima: Grupo de Anlisis para el Desarrollo.
Portocarrero, J., P. Margarita, and C. Vallejo. 2007. Cuentas Nacionales de Salud: Per, 1995
2005 Ocina General de Planeamiento y Presupuesto del Ministerio de Salud (MINSA) y
Observatorio de la Salud del Consorcio de Investigacin Econmica y Social, Lima.
World Bank. 2010. World Development Indicators 2010. Washington DC: Te World Bank. Avail-
able at http://data.worldbank.org/sites/default/les/wdi/complete.pdf.
122
The Impact of Health
Insurance on Use, Spending,
and Health in Indonesia
Facundo Cuevas and Susan W. Parker
Classied by the World Bank as a lower middle-income country, Indo-
nesia had per capita gross national income of US$3,600 in purchasing
power parity terms in 2008. With about half the population residing in
rural areas, life expectancy remains fairly lowat 69 years for men and
73 years for womenand under-ve mortality highat 41 per 1,000
in 2008. Child vaccination remains far from universal: in 2008, 77%
of children had received the full course DPT vaccination (three doses)
by their 23rd month, a good metric for the performance of immuniza-
tion programs. Total health spending was about 2.2% of gross domestic
product (GDP) in 2008, about 55% of it public and 45% private (World
Bank 2010).
In this chapter, we study health insurance coverage and its relation-
ship to health, health care use, access to services, and nancial protec-
tion in Indonesia. For the empirical analysis, we use the 1993, 1997, and
2000 waves of the Indonesian Family Life Survey (IFLS), a multipur-
pose longitudinal household survey. (A more recent wave was collected
in 2008 but was not available at the time of this analysis.) Observing
outcomes and decisions of the same individuals over almost a decade
provides a unique opportunity to understand the impact of insurance
coverage and yields estimates that are purged of some of the most con-
cerning sources of bias.
C
h
a
p
t
e
r

7
The Impact of Health Insurance on Use, Spending, and Health in Indonesia 123
The Indonesian health system
Te Indonesian constitution stipulates that every citizen has the right to social
security and emphasizes the role of the state in providing universal social secu-
rity coverage. But coverage remains far from universal. Te social security systems
include Askes, a mandatory social health insurance program for civil servants, and
Asabri, for police, military, and their dependents. Jamsostek insurance for private
sector workers is theoretically mandatory, but rms can opt out if they provide
similar health services.
Askes covers an estimated 13.8 million people (plus 1.4 million commercial
members), and Jamsostek 2.7 million (1.5 million of them workers). So formal
health insurance schemes cover about 18 million people. Adding those covered by
private or employer-funded health insurance, about 30 million individuals, or 15%
of the population, are estimated to have health insurance.
In September 2004 the Indonesian House of Representatives endorsed a law on
the National Social Security System mandating several social security schemes for
citizens: old-age pension, old-age savings, national health insurance, work injury
insurance, and death benets for survivors of deceased workers. Te new scheme
aims to cover all Indonesian citizenswhether formal or informal workers or self-
employed (World Bank 2010). Te main program is Askes, which provides basic
health benets to poor individuals.
1

Te proportion covered by health insurance is remarkably constant over time,
at about 15%, with a majority receiving coverage through the workplace in all three
survey years. In 1993 we cannot disaggregate by type of coverage, but data from
1997 and 2000 show that most insurance is through Askes, then Jamostek. Nearly
all insurance plans cover outpatient care and some hospitalization as well as some
types of surgery.
Usually there is an assumption of fairly static coverage of formal sector and
government workers, that they get and retain insurance coverage upon enter-
ing a covered job. We discovered the opposite; in fact, our empirical strategy
uses changes in insurance status among the same people to analyze the impact
of insurance. Of those insured in 1993 (about 15% of the sample), only about
two-thirds remained insured in 1997. Similarly, of individuals with insurance in
1997, only about two-thirds remained insured in 2000. Of the uninsured popu-
lation in 1993 (86%), about 5% gained insurance by 1997, and of the uninsured
population in 1997, about 5% gained insurance by 2000. In sum, whereas about
15% of the population is covered by health insurance in any round of the IFLS,
only about two-thirds of them retain insurance coverage in the next wave of data
collection.
124 Chapter 7
About 23% of the population is covered by insurance in at least one of the rounds.
Only about 8% of the population are constantly insured. Overall, then, about 15%
of the population have changes in insurance statusthat is, are sometimes insured.
Tis population is the source of variation for the econometric analysis.
Most of those leaving the insured state were previously covered by Askes or
Jamostek. But it is more probable that uninsured individuals lost Askes insurance
than insured individuals gained Askes insurance between 1997 and 2000. Te
same is true for reimbursement insurance.
2
For Jamsostek the proportion of those
losing insurance is about the same as the proportion of those gaining it. Tere is
also movement between the covered population in the type of health insurance.
For instance, of the 63% reporting Askes coverage in 1997, 88% remained with
Askes, and the rest reported dierent coverage in 2000.
Descriptive statistics: the insured and the uninsured
Overall, the insured have higher rates of outpatient care use during the four weeks
prior to the survey and in each of the three IFLS rounds. In 1993, 27% of the
insured adult population received outpatient care, compared with 19% of the
uncovered population.
3
For inpatient care for individuals ages 15 and older, those
with insurance report higher use of inpatient care during the four weeks prior to
the survey than do those without insurance over the three rounds. Also in all three
rounds, the proportion of households incurring positive health expenditures is less
for insured individuals than for uninsured individuals. Insured individuals have a
higher body mass index than the uninsured, weighing on average about ve kilo-
grams more. About 22% of the insured population is overweight or obese, com-
pared with about 13% of the uninsured population. Other health indicators do not
show worse health for the insured than the uninsured.
In 1993, 27% of insured children reported using outpatient care in the four
weeks prior to the survey, compared with 20% of uninsured children. Te over-
all proportion of children using health care decreases over the panel, as expected
because older children are less likely to have illnesses. For health status, insured
children tend to be taller than uninsured children. Body mass indexes are slightly
higher for insured than uninsured children, and these dierences increase over
time. Anemia levels for insured children are 45 percentage points lower than for
uninsured children. For such health symptoms as diarrhea and headache, there are
no clear patterns of dierences between insured and uninsured children.
We divide the population into four groups between each two successive rounds:
those who have insurance in both rounds, those who have insurance in neither
round, those who gain insurance over time, and those who lose insurance over
time. Te use of outpatient care falls between 1993 and 1997 for all of the groups,
except that those without insurance in 1993 who gained insurance by 1997 main-
tained their 1993 level of use across time. In other words, those gaining insurance
over time increased use relative to those losing insurance.
Te probability of spending money on health care between 1993 and 1997 falls
for all four groups, but the group gaining insurance over time shows a greater fall
in the probability of health expenditures. For health status, the four groups tend to
show the same tendencies over time. Te always insured group tends to have higher
obesity rates than the other groups, but other health measures (including the preva-
lence of anemia and hypertension) tend to be similar. In sum, these descriptive sta-
tistics are consistent with an impact of insurance on health use and expenditures,
but not on health status.
Use increases over time for the group that gains insurance and is reduced for
the group that loses insurance. Te group with insurance in both periods and those
without insurance in both periods do not show changes in overall health care use
between 1997 and 2000. Te probability of making an expenditure increases for
all three groups between 1997 and 2000, but the changes for the group that gains
insurance are lower than those for the group that loses insurance. Again, for health
status, there is little to dierentiate the four groups.
We nd similar trends for children. Children who gain insurance increase use
relative to those who lose insurance. For instance, between 1997 and 2000 the pro-
portion of children using outpatient care increased slightly for the group gaining
insurance, from 12% to 13%, but fell for the group losing insurance, from 20% to
15%. Health status indicators for children show no obvious patterns for gaining or
losing insurance.
In sum, individuals switching coverage have similar health status and symp-
toms during the four weeks prior to the survey. However, those gaining insurance
are more likely to use care over time than those losing it, and the probability of
those gaining insurance to incur health expenditures rises more slowly than that of
those losing insurance.
Impacts of insurance on adults
Overall, having health insurance shows important positive and signicant eects
on indicators of health care use, both in inpatient and outpatient care (see box 7.1
for information on the methodology behind this study). For instance, for adults,
having health insurance increases by about 4 percentage points the probability of
having outpatient care in the four weeks prior to the survey and the probability
of having inpatient care by about 1 percentage point (top 3 rows in table 7.1). For
126 Chapter 7
health spending, insurance signicantly reduces the probability of any household
health spending, but the impact on average per capita health spending is insigni-
cant. By gender, the impact of health insurance on men and women is similar for
inpatient and outpatient care as well as for household health care expenditures.
Insurance increases the probability of having any outpatient care by 5 per-
centage points for adults in rural areas, compared with 4 percentage points for
urban adults (middle and bottom 3 rows in table 7.1). For inpatient care, rural areas
show positive and signicant impacts of insurance for the group of all adults, but
in urban areas, there are no statistically signicant impacts. Also in rural areas,
insurance signicantly reduces the probability of household health care spending
(mostly for female insured adults), and there are signicant impacts, albeit smaller,
in urban areas. Overall, then, insurance has slightly larger impacts for the rural
population, especially for women.
Impacts on use of outpatient care are higher for adults in the bottom 50% of
the expenditure distribution (a proxy for relative income), compared with adults
in the top 50% of the distribution (table 7.2). For instance, whereas the impact of
insurance on the probability of having any outpatient care is 4 percentage points
BOX 7.1
Methodology
Tis study uses longitudinal information
in the Indonesian Family Life Survey to
estimate individual xed eect models
that use changes in insurance status for
the same people over time to estimate the
impacts of insurance coverage. Cross-
sectional analyses are likely to suer from
selection problems in who has insurance at
a given time. Individual xed eect estima-
tors rely on variation in health insurance
coverage for the same individual (rather
than comparing dierent individuals) and
isolate the impact estimates from unob-
served time-invariant individual character-
istics that could create spurious correla-
tions among insurance, health status, and
the use of services.
For the impacts of health insurance, we
focus primarily on the use of services and
health spending, as well as some health sta-
tus indicators that might respond quickly
to changes in insurance. Such indicators
potentially include symptoms such as
cough, diarrhea, the u, high blood pres-
sure (which can be lowered by appropri-
ate medicine), and anemia (which can be
remedied by iron tablets)but we do not
include illnesses such as diabetes. We thus
concentrate on variables where changes in
insurance coverage might have fairly quick
eects.
Te descriptive analysis shows sucient
variation in insurance coverage for indi-
vidual xed eects models to be estimated.
Tese estimates are of course based only
on the individuals experiencing changes in
insurance coverage over the seven years of
the panel, about 15% of the sample.
for adults in the top 50% of the distribution, it is 6 percentage points for those in
the bottom 50% of the distribution. For outpatient care during the four weeks
prior to the survey, insurance increases the number of visits by 0.07 for adults
in the top 50% of the distribution, compared with 0.17 for those in the bottom
50%. Surprisingly, however, the impact of health insurance on reducing household
health spending is negative and signicant only for women in the lower income
groups, whereas for upper income groups the impact is negative and signicant
only for men.
We now turn to potential impacts of insurance on indicators more closely asso-
ciated with health status, including weight, body mass index, high blood pressure,
TABLE 7.1
Effects of health insurance on use, by gender and location
Individual fxed effects on adults older than age 15 at baseline
Had any
outpatient
care in the four
weeks prior
to the survey
Number of
times that had
outpatient
care in the
four weeks
prior to the
survey
Had any
inpatient care
in the 12
months prior
to the survey
Number
times that
had inpatient
care in the 12
months prior
to the survey
Household
health
expenditures
(>0)
Monthly
per capita
expenditure
on health
(real rupiahs)
Adults 0.04
[0.0111]***
0.10
[0.0219]***
0.01
[0.0057]*
0.01
[0.0061]**
0.03
[0.0126]**
349.55
[435.5682]
Female 0.04
[0.0161]**
0.11
[0.0327]***
0.01
[0.0086]
0.01
[0.0093]
0.04
[0.0143]**
297.17
[495.9211]
Male 0.05
[0.0142]***
0.09
[0.0278]***
0.01
[0.0071]**
0.01
[0.0075]**
0.03
[0.0148]*
374.14
[466.2524]
Adults living
in rural areas
0.05
[0.0200]**
0.12
[0.0436]***
0.02
[0.0087]*
0.02
[0.0096]*
0.04
[0.0218]*
319.06
[389.1802]
Female 0.05
[0.0294]*
0.14
[0.0669]**
0.02
[0.0134]
0.02
[0.0149]
0.05
[0.0251]**
687.37
[448.7407]
Male 0.0474
[0.0268]*
0.0895
[0.0567]
0.0144
[0.0111]
0.0121
[0.0117]
0.0335
[0.0250]
44.8885
[452.5782]
Adults living
in urban
areas
0.04
[0.0133]***
0.09
[0.0251]***
0.01
[0.0072]
0.01
[0.0077]
0.03
[0.0151]*
354.04
[574.0108]
Female 0.04
[0.0193]*
0.10
[0.0372]**
0.00
[0.0107]
0.01
[0.0116]
0.03
[0.0171]
230.62
[648.6015]
Male 0.04
[0.0166]***
0.09
[0.0311]***
0.01
[0.0090]
0.02
[0.0095]*
0.03
[0.0180]
505.88
[616.9774]
Source: Authors calculations based on 1993, 1997, and 2000 Indonesian Family Life Surveys.
128 Chapter 7
hemoglobin levels and anemia, and activities of daily living, including ability to
carry a heavy load, walk 5 kilometers, and kneel (table 7.3). Among these indica-
tors, we have tried to choose some, among those available, where it is feasible that
health insurance might have an impact fairly quickly. Even so, the health status
variables chosen are likely to take longer to react to having insurance than such
variables as health clinic visits or health care expenditures.
Overall, there are few consistent ndings of signicant and positive impacts of
health insurance. Tere are no signicant coecients on the impact of insurance
on obesity, hemoglobin levels, or high blood pressure, except for insurance reduc-
ing the prevalence of hypertension stage 1 for women in rural areas. Tere are a
couple modest impacts of insurance on some activities of daily living. In particular,
TABLE 7.2
Effects of health insurance on use, by expenditure distribution
Individual fxed effects on adults older than age 15 at baseline
Had any
outpatient
care in the
four weeks
prior to the
survey
Number
of times
that had
outpatient
care in the
four weeks
prior to the
survey
Had any
inpatient
care in the
12 months
prior to the
survey
Number
times that
had inpatient
care in the
12 months
prior to the
survey
Household
health
expenditures
(>0)
Monthly
per capita
expenditure
on health
(real rupiahs)
Adults in
top 50% of
expenditure
distribution
0.04
[0.0136]***
0.07
[0.0249]***
0.01
[0.0073]
0.01
[0.0075]
0.03
[0.0144]**
562.42
[601.7865]
Female 0.04
[0.0196]*
0.09
[0.0375]**
0.01
[0.0108]
0.01
[0.0113]
0.03
[0.0169]
516.71
[675.4978]
Male 0.04
[0.0167]**
0.04
[0.0303]
0.01
[0.0091]
0.01
[0.0093]
0.04
[0.0167]**
599.81
[650.0818]
Adults in
bottom
50% of
expenditure
distribution
0.0547
[0.0190]***
0.1733
[0.0437]***
0.0087
[0.0084]
0.0165
[0.0103]
0.02
[0.0252]
398.5821
[226.5667]*
Female 0.04
[0.0279]
0.15
[0.0655]**
0.00
[0.0131]
0.01
[0.0162]
0.05
[0.0267]*
485.23
[361.6902]
Male 0.0634
[0.0268]**
0.1949
[0.0587]***
0.0212
[0.0103]**
0.0254
[0.0126]**
0.0088
[0.0307]
285.6729
[219.4914]
Source: Authors calculations based on 1993, 1997, and 2000 Indonesian Family Life Surveys.
T
A
B
L
E

7
.
3

E
f
f
e
c
t
s

o
f

h
e
a
l
t
h

s
t
a
t
u
s
,

b
y

g
e
n
d
e
r

a
n
d

l
o
c
a
t
i
o
n
I
n
d
i
v
i
d
u
a
l

f
x
e
d

e
f
f
e
c
t
s

o
n

a
d
u
l
t
s

o
l
d
e
r

t
h
a
n

1
5

y
e
a
r
s

a
t

b
a
s
e
l
i
n
e
B
o
d
y

m
a
s
s

i
n
d
e
x

(
B
M
I
)
O
v
e
r
w
e
i
g
h
t

(
B
M
I

b
e
t
w
e
e
n

2
5
2
9
.
9

k
g
/
m
2
)
O
b
e
s
i
t
y

(
B
M
I

3
0

k
g
/
m
2
)
H
e
m
o
g
l
o
b
i
n

(
g
/
d
l
)
L
o
w

h
e
m
o
g
l
o
b
i
n

(
i
r
o
n

d
e
f
c
i
t
:

h
b
1
2

w
o
m
e
n
,

h
b
1
3
.
5

m
e
n
,

h
b
1
1

c
h
i
l
d
)
H
i
g
h

b
l
o
o
d

p
r
e
s
s
u
r
e
H
y
p
e
r
t
e
n
s
i
o
n

s
t
a
g
e

1

(
s
y
s
t
o
l
i
c

1
4
0
1
5
9

o
r

d
i
a
s
t
o
l
i
c

9
0
9
9
)
H
y
p
e
r
t
e
n
s
i
o
n

s
t
a
g
e

2

(
s
y
s
t
o
l
i
c
1
6
0

o
r

d
i
a
s
t
o
l
i
c
1
0
0
)
D
i
f
f
c
u
l
t
y
/

u
n
a
b
l
e

t
o

c
a
r
r
y

h
e
a
v
y

l
o
a
d
D
i
f
f
c
u
l
t
y
/

u
n
a
b
l
e

t
o

w
a
l
k

5

k
i
l
o
m
e
t
e
r
s
D
i
f
f
c
u
l
t
y
/

u
n
a
b
l
e

t
o

k
n
e
e
l
A
d
u
l
t
s
0
.
0
3
[
0
.
0
4
5
8
]
0
.
0
1
[
0
.
0
0
7
9
]
0
.
0
0
[
0
.
0
0
3
9
]
0
.
0
4
[
0
.
0
5
3
5
]
0
.
0
1
[
0
.
0
1
7
3
]
0
.
0
1
[
0
.
0
1
7
4
]
0
.
0
0
[
0
.
0
1
4
9
]
0
.
0
1
[
0
.
0
0
8
6
]
0
.
0
0
[
0
.
0
0
8
3
]
0
.
0
1
[
0
.
0
1
0
4
]
0
.
0
1
[
0
.
0
0
5
1
]
F
e
m
a
l
e
0
.
0
2
[
0
.
0
6
5
0
]
0
.
0
0
[
0
.
0
1
1
9
]
0
.
0
1
[
0
.
0
0
6
2
]
0
.
0
1
[
0
.
0
6
7
9
]
0
.
0
2
[
0
.
0
2
3
4
]
0
.
0
1
[
0
.
0
2
3
0
]
0
.
0
1
[
0
.
0
1
8
9
]
0
.
0
2
[
0
.
0
1
1
5
]
0
.
0
1
[
0
.
0
1
2
9
]
0
.
0
1
[
0
.
0
1
6
1
]
0
.
0
0
[
0
.
0
0
7
4
]
M
a
l
e
0
.
0
9
[
0
.
0
6
0
8
]
0
.
0
1
[
0
.
0
1
0
4
]
0
.
0
0
[
0
.
0
0
4
2
]
0
.
0
6
[
0
.
0
7
9
7
]
0
.
0
0
[
0
.
0
2
3
9
]
0
.
0
2
[
0
.
0
2
5
6
]
0
.
0
0
[
0
.
0
2
2
6
]
0
.
0
0
[
0
.
0
1
3
0
]
0
.
0
2
[
0
.
0
0
8
8
]
*
0
.
0
0
[
0
.
0
1
1
2
]
0
.
0
1
[
0
.
0
0
6
7
]
*
A
d
u
l
t
s

l
i
v
i
n
g

i
n

r
u
r
a
l

a
r
e
a
s
0
.
0
9
[
0
.
0
7
9
2
]
0
.
0
1
[
0
.
0
1
2
5
]
0
.
0
0
[
0
.
0
0
4
8
]
0
.
0
2
[
0
.
0
9
8
0
]
0
.
0
1
[
0
.
0
3
2
2
]
0
.
0
0
[
0
.
0
3
4
1
]
0
.
0
3
[
0
.
0
2
8
5
]
0
.
0
0
[
0
.
0
1
7
5
]
0
.
0
1
[
0
.
0
1
4
1
]
0
.
0
1
[
0
.
0
1
9
0
]
0
.
0
2
[
0
.
0
0
9
0
]
*
F
e
m
a
l
e
0
.
0
9
[
0
.
1
1
6
7
]
0
.
0
1
[
0
.
0
2
0
3
]
0
.
0
0
[
0
.
0
0
8
6
]
0
.
0
6
[
0
.
1
3
2
6
]
0
.
0
3
[
0
.
0
4
1
9
]
0
.
0
3
[
0
.
0
4
7
4
]
0
.
0
6
[
0
.
0
3
5
0
]
*
0
.
0
2
[
0
.
0
2
3
8
]
0
.
0
2
[
0
.
0
2
1
5
]
0
.
0
3
[
0
.
0
3
1
2
]
0
.
0
3
[
0
.
0
1
3
4
]
*
*
M
a
l
e
0
.
1
1
[
0
.
0
9
4
5
]
0
.
0
2
[
0
.
0
1
2
6
]
0
.
0
0
[
0
.
0
0
3
1
]
0
.
0
1
[
0
.
1
4
3
5
]
0
.
0
2
[
0
.
0
4
5
6
]
0
.
0
2
[
0
.
0
4
5
9
]
0
.
0
0
[
0
.
0
4
2
2
]
0
.
0
1
[
0
.
0
2
5
8
]
0
.
0
0
[
0
.
0
1
6
1
]
0
.
0
0
[
0
.
0
1
9
0
]
0
.
0
0
[
0
.
0
1
2
7
]
A
d
u
l
t
s

l
i
v
i
n
g

i
n

u
r
b
a
n

a
r
e
a
s
0
.
0
2
[
0
.
0
5
6
2
]
0
.
0
1
[
0
.
0
1
0
0
]
0
.
0
1
[
0
.
0
0
5
2
]
0
.
0
6
[
0
.
0
6
3
8
]
0
.
0
2
[
0
.
0
2
0
5
]
0
.
0
1
[
0
.
0
2
0
2
]
0
.
0
1
[
0
.
0
1
7
4
]
0
.
0
1
[
0
.
0
0
9
8
]
0
.
0
1
[
0
.
0
1
0
1
]
0
.
0
1
[
0
.
0
1
2
5
]
0
.
0
0
[
0
.
0
0
6
1
]
F
e
m
a
l
e
0
.
0
6
[
0
.
0
7
9
0
]
0
.
0
0
[
0
.
0
1
4
6
]
0
.
0
1
[
0
.
0
0
7
9
]
0
.
0
3
[
0
.
0
7
9
0
]
0
.
0
2
[
0
.
0
2
8
1
]
0
.
0
1
[
0
.
0
2
6
2
]
0
.
0
3
[
0
.
0
2
2
5
]
0
.
0
2
[
0
.
0
1
3
2
]
0
.
0
0
[
0
.
0
1
5
9
]
0
.
0
1
[
0
.
0
1
8
8
]
0
.
0
1
[
0
.
0
0
8
9
]
M
a
l
e
0
.
1
0
[
0
.
0
7
7
3
]
0
.
0
1
[
0
.
0
1
4
0
]
0
.
0
0
[
0
.
0
0
6
0
]
0
.
0
7
[
0
.
0
9
6
2
]
0
.
0
1
[
0
.
0
2
7
8
]
0
.
0
5
[
0
.
0
3
0
6
]
0
.
0
1
[
0
.
0
2
6
8
]
0
.
0
0
[
0
.
0
1
4
8
]
0
.
0
2
[
0
.
0
1
0
5
]
*
*
0
.
0
0
[
0
.
0
1
3
8
]
0
.
0
2
[
0
.
0
0
7
9
]
*
*
*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
0
5
;

*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
1
0
.
N
o
t
e
:

N
u
m
b
e
r
s

i
n

b
r
a
c
k
e
t
s

a
r
e

s
t
a
n
d
a
r
d

e
r
r
o
r
s
.
S
o
u
r
c
e
:

A
u
t
h
o
r
s

c
a
l
c
u
l
a
t
i
o
n
s

b
a
s
e
d

o
n

1
9
9
3
,

1
9
9
7
,

a
n
d

2
0
0
0

I
n
d
o
n
e
s
i
a
n

F
a
m
i
l
y

L
i
f
e

S
u
r
v
e
y
s
.
130 Chapter 7
health insurance reduces the probability of diculties in kneeling (for women in
rural areas and men in urban areas) and in carrying heavy objects for men.
Disaggregating by spending category, the results show no signicant impacts of
insurance on the health status indicators for adults in wealthier households (table 7.4).
But for adults in poorer households, there is an unexpected signicant increase in the
probability of having anemia for females and a signicant reduction in blood pressure
associated with health insurance for adult males. Overall, then, the impacts on health
status are much less widespread than those on use by adults and do not provide con-
vincing evidence of a strong eect of insurance on health status.
Before turning to estimates of the impact of insurance on children, it is worth
noting that, because our sample includes all adults ages 15 and older, some of the
transitions from having health insurance to not having health insurance are related
to young individuals entering the labor market for the rst time and retirement
decisions of older individuals. To ensure that such transitions are not driving the
results, we re-estimate the previous tables for adults using the population of adults
ages 3550. Overall, results (available upon request) of the impact of insurance on
health care use are quite similar to previous results. For health status indicators, as
with the entire adult population, there are few signicant impacts of insurance in
this population. Tere continues to be an impact of insurance on reducing prob-
lems associated with kneeling for adults in rural areas. We conclude that transitions
associated with entering the labor market for the rst time or exiting due to retire-
ment are not driving the reported results.
Impacts of insurance on children
We now turn to insurance impacts on children. Overall, insurance signicantly
increases the use of outpatient care for children, but curiously only for female chil-
dren in both rural and urban areas, with a 4 percentage point increase in the prob-
ability of using outpatient care in the four weeks prior to the survey (table 7.5).
Insurance signicantly increases the use of inpatient care but notably again only
for girls, both in rural and urban areas. Insurance reduces the probability of mak-
ing positive household health expenditures for male children, although there is no
overall eect on household health expenditures. Table 7.5 also includes the impacts
of insurance on body mass index, obesity, and hemoglobin levels for children, but
there do not appear to be any signicant impacts of health insurance on these vari-
ables, with the exception of a surprising positive impact of insurance on increasing
obesity of girls living in rural areas.
Tere are some signicant impacts of insurance on children from both upper
income and lower income households (table 7.6), but again mainly for girls. In
T
A
B
L
E

7
.
4

E
f
f
e
c
t
s

o
f

h
e
a
l
t
h

s
t
a
t
u
s
,

b
y

e
x
p
e
n
d
i
t
u
r
e

d
i
s
t
r
i
b
u
t
i
o
n
I
n
d
i
v
i
d
u
a
l

f
x
e
d

e
f
f
e
c
t
s

o
n

a
d
u
l
t
s

o
l
d
e
r

t
h
a
n

1
5

y
e
a
r
s

a
t

b
a
s
e
l
i
n
e
B
o
d
y

m
a
s
s

i
n
d
e
x

(
B
M
I
)
O
v
e
r
w
e
i
g
h
t

(
B
M
I

b
e
t
w
e
e
n

2
5
2
9
.
9

k
g
/
m
2
)
O
b
e
s
i
t
y

(
B
M
I

3
0

k
g
/
m
2
)
H
e
m
o
g
l
o
b
i
n

(
g
/
d
l
)
L
o
w

h
e
m
o
g
l
o
b
i
n

(
i
r
o
n

d
e
f
c
i
t
:

h
b
1
2

w
o
m
e
n
,

h
b
1
3
.
5

m
e
n
,

h
b
1
1

c
h
i
l
d
)
H
i
g
h

b
l
o
o
d

p
r
e
s
s
u
r
e
H
y
p
e
r
t
e
n
s
i
o
n

s
t
a
g
e

1

(
s
y
s
t
o
l
i
c

1
4
0
1
5
9

o
r

d
i
a
s
t
o
l
i
c

9
0
9
9
)
H
y
p
e
r
t
e
n
s
i
o
n

s
t
a
g
e

2

(
s
y
s
t
o
l
i
c
1
6
0

o
r

d
i
a
s
t
o
l
i
c
1
0
0
)
D
i
f
f
c
u
l
t
y
/

u
n
a
b
l
e

t
o

c
a
r
r
y

h
e
a
v
y

l
o
a
d
D
i
f
f
c
u
l
t
y
/

u
n
a
b
l
e

t
o

w
a
l
k

5

k
i
l
o
m
e
t
e
r
s
D
i
f
f
c
u
l
t
y
/

u
n
a
b
l
e

t
o

k
n
e
e
l
A
d
u
l
t
s

i
n

t
o
p

5
0
%

o
f

e
x
p
e
n
d
i
t
u
r
e

d
i
s
t
r
i
b
u
t
i
o
n
0
.
0
2
6
5
[
0
.
0
5
6
8
]
0
.
0
0
7
5
[
0
.
0
0
9
7
]
0
.
0
0
4
1
[
0
.
0
0
5
0
]
0
.
0
4
5
2
[
0
.
0
6
3
5
]
0
.
0
1
0
1
[
0
.
0
2
0
6
]
0
.
0
1
0
9
[
0
.
0
2
0
8
]
0
.
0
1
1
2
[
0
.
0
1
7
9
]
0
.
0
0
3
7
[
0
.
0
1
0
2
]
0
.
0
0
4
1
[
0
.
0
1
0
1
]
0
.
0
0
2
8
[
0
.
0
1
2
3
]
0
.
0
0
4
1
[
0
.
0
0
6
2
]
F
e
m
a
l
e
0
.
0
5
0
4
[
0
.
0
7
8
1
]
0
.
0
0
2
2
[
0
.
0
1
4
1
]
0
.
0
0
7
1
[
0
.
0
0
7
7
]
0
.
0
3
2
2
[
0
.
0
7
8
3
]
0
.
0
1
3
2
[
0
.
0
2
7
7
]
0
.
0
1
2
8
[
0
.
0
2
8
4
]
0
.
0
2
3
7
[
0
.
0
2
3
0
]
0
.
0
0
7
6
[
0
.
0
1
4
1
]
0
.
0
0
5
7
[
0
.
0
1
5
6
]
0
.
0
0
0
7
[
0
.
0
1
9
2
]
0
.
0
0
0
5
[
0
.
0
0
9
0
]
M
a
l
e
0
.
1
1
5
5
[
0
.
0
7
7
2
]
0
.
0
1
9
6
[
0
.
0
1
3
9
]
0
.
0
0
1
1
[
0
.
0
0
5
7
]
0
.
0
4
6
8
[
0
.
0
9
4
7
]
0
.
0
0
8
7
[
0
.
0
2
8
1
]
0
.
0
0
3
8
[
0
.
0
3
0
5
]
0
.
0
0
3
2
[
0
.
0
2
7
2
]
0
.
0
0
0
6
[
0
.
0
1
5
2
]
0
.
0
1
6
5
[
0
.
0
1
0
7
]
0
.
0
0
3
4
[
0
.
0
1
3
0
]
0
.
0
0
9
6
[
0
.
0
0
8
4
]
A
d
u
l
t
s

i
n

b
o
t
t
o
m

5
0
%

o
f

e
x
p
e
n
d
i
t
u
r
e

d
i
s
t
r
i
b
u
t
i
o
n
0
.
0
6
5
1
[
0
.
0
7
7
2
]
0
.
0
1
0
6
[
0
.
0
1
3
7
]
0
.
0
0
5
3
[
0
.
0
0
5
8
]
0
.
0
1
1
7
[
0
.
0
9
8
3
]
0
.
0
5
8
[
0
.
0
3
1
7
]
*
0
.
0
4
7
9
[
0
.
0
3
1
8
]
0
.
0
2
4
1
[
0
.
0
2
6
4
]
0
.
0
1
4
7
[
0
.
0
1
6
3
]
0
.
0
0
4
2
[
0
.
0
1
4
6
]
0
.
0
1
9
4
[
0
.
0
1
9
4
]
0
.
0
1
4
6
[
0
.
0
0
8
6
]
*
F
e
m
a
l
e
0
.
0
7
5
5
[
0
.
1
2
0
2
]
0
.
0
1
8
[
0
.
0
2
1
9
]
0
.
0
0
8
9
[
0
.
0
0
9
8
]
0
.
0
5
1
1
[
0
.
1
3
2
5
]
0
.
0
8
7
7
[
0
.
0
4
3
3
]
*
*
0
.
0
1
7
6
[
0
.
0
3
9
0
]
0
.
0
3
5
5
[
0
.
0
3
3
4
]
0
.
0
4
0
5
[
0
.
0
2
0
1
]
*
*
0
.
0
0
9
5
[
0
.
0
2
2
8
]
0
.
0
4
7
3
[
0
.
0
2
9
5
]
0
.
0
1
3
[
0
.
0
1
3
2
]
M
a
l
e
0
.
0
6
6
3
[
0
.
0
9
3
6
]
0
.
0
0
2
8
[
0
.
0
1
3
4
]
0
.
0
0
1
4
[
0
.
0
0
5
3
]
0
.
0
6
7
5
[
0
.
1
4
7
1
]
0
.
0
1
8
7
[
0
.
0
4
4
9
]
0
.
0
8
3
2
[
0
.
0
4
6
9
]
*
0
.
0
0
9
4
[
0
.
0
4
0
2
]
0
.
0
1
2
9
[
0
.
0
2
4
9
]
0
.
0
1
9
9
[
0
.
0
1
5
6
]
0
.
0
1
2
7
[
0
.
0
2
1
1
]
0
.
0
1
7
3
[
0
.
0
1
1
1
]
*
*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
0
5
;

*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
1
0
.
N
o
t
e
:

N
u
m
b
e
r
s

i
n

b
r
a
c
k
e
t
s

a
r
e

s
t
a
n
d
a
r
d

e
r
r
o
r
s
.
S
o
u
r
c
e
:

A
u
t
h
o
r
s

c
a
l
c
u
l
a
t
i
o
n
s

b
a
s
e
d

o
n

1
9
9
3
,

1
9
9
7
,

a
n
d

2
0
0
0

I
n
d
o
n
e
s
i
a
n

F
a
m
i
l
y

L
i
f
e

S
u
r
v
e
y
s
.
132 Chapter 7
T
A
B
L
E

7
.
5

E
f
f
e
c
t
s

o
f

h
e
a
l
t
h

i
n
s
u
r
a
n
c
e

o
n

y
o
u
t
h

a
n
d

h
e
a
l
t
h

s
t
a
t
u
s

f
o
r

c
h
i
l
d
r
e
n
,

b
y

g
e
n
d
e
r

a
n
d

l
o
c
a
t
i
o
n
H
a
d

a
n
y

o
u
t
p
a
t
i
e
n
t

c
a
r
e

i
n

t
h
e

f
o
u
r

w
e
e
k
s

p
r
i
o
r

t
o

t
h
e

s
u
r
v
e
y
N
u
m
b
e
r

o
f

t
i
m
e
s

t
h
a
t

h
a
d

o
u
t
p
a
t
i
e
n
t

c
a
r
e

i
n

t
h
e

f
o
u
r

w
e
e
k
s

p
r
i
o
r

t
o

t
h
e

s
u
r
v
e
y
H
a
d

a
n
y

i
n
p
a
t
i
e
n
t

c
a
r
e

i
n

t
h
e

1
2

m
o
n
t
h
s

p
r
i
o
r

t
o

t
h
e

s
u
r
v
e
y
N
u
m
b
e
r

t
i
m
e
s

t
h
a
t

h
a
d

i
n
p
a
t
i
e
n
t

c
a
r
e

i
n

t
h
e

1
2

m
o
n
t
h
s

p
r
i
o
r

t
o

t
h
e

s
u
r
v
e
y
H
o
u
s
e
h
o
l
d

h
e
a
l
t
h

e
x
p
e
n
d
i
t
u
r
e
s

(
>
0
)
M
o
n
t
h
l
y

p
e
r

c
a
p
i
t
a

e
x
p
e
n
d
i
t
u
r
e

o
n

h
e
a
l
t
h

(
r
e
a
l

r
u
p
i
a
h
s
)
B
o
d
y

m
a
s
s

i
n
d
e
x

(
B
M
I
)
O
v
e
r
w
e
i
g
h
t

(
B
M
I

b
e
t
w
e
e
n

2
5
2
9
.
9

k
g
/
m
2
)
O
b
e
s
i
t
y

(
B
M
I
3
0

k
g
/
m
2
)
H
e
m
o
g
l
o
b
i
n

(
g
/
d
l
)
L
o
w

h
e
m
o
g
l
o
b
i
n

(
i
r
o
n

d
e
f
c
i
t
:

h
b
1
2

w
o
m
e
n
,

h
b
<
1
3
.
5

m
e
n
,

h
b
<
1
1

c
h
i
l
d
)
C
h
i
l
d
r
e
n
0
.
0
3
[
0
.
0
1
1
8
]
*
*
*
0
.
0
4
[
0
.
0
1
8
3
]
*
0
.
0
1
[
0
.
0
0
4
2
]
*
*
0
.
0
1
[
0
.
0
0
4
9
]
*
*
0
.
0
5
[
0
.
0
1
7
0
]
*
*
*
3
0
0
.
4
5
[
4
2
9
.
2
0
8
4
]
0
.
0
3
[
0
.
0
6
7
3
]
0
.
0
1
[
0
.
0
0
6
5
]
0
.
0
0
[
0
.
0
0
3
5
]
0
.
0
1
[
0
.
0
6
0
5
]
0
.
0
1
[
0
.
0
1
8
7
]
F
e
m
a
l
e
0
.
0
5
[
0
.
0
1
6
9
]
*
*
*
0
.
0
5
[
0
.
0
2
6
0
]
*
0
.
0
1
[
0
.
0
0
5
8
]
*
0
.
0
1
[
0
.
0
0
7
5
]
*
0
.
0
2
[
0
.
0
1
9
6
]
6
2
4
.
6
8
[
7
3
8
.
7
3
0
0
]
0
.
0
5
[
0
.
0
9
3
6
]
0
.
0
1
[
0
.
0
0
9
1
]
0
.
0
0
[
0
.
0
0
4
3
]
0
.
0
6
[
0
.
0
7
7
5
]
0
.
0
4
[
0
.
0
2
6
4
]
M
a
l
e
0
.
0
2
[
0
.
0
1
6
5
]
0
.
0
2
[
0
.
0
2
6
2
]
0
.
0
1
[
0
.
0
0
6
2
]
0
.
0
1
[
0
.
0
0
6
4
]
0
.
0
8
[
0
.
0
2
1
5
]
*
*
*
8
0
.
2
4
[
3
5
7
.
9
2
3
5
]
0
.
0
1
[
0
.
0
8
8
2
]
0
.
0
1
[
0
.
0
0
9
0
]
0
.
0
0
[
0
.
0
0
5
4
]
0
.
0
4
[
0
.
0
8
4
6
]
0
.
0
0
[
0
.
0
2
5
0
]
C
h
i
l
d
r
e
n

l
i
v
i
n
g

i
n

r
u
r
a
l

a
r
e
a
s
0
.
0
4
[
0
.
0
2
0
8
]
*
0
.
0
7
[
0
.
0
3
4
1
]
*
*
0
.
0
1
[
0
.
0
0
6
5
]
*
0
.
0
1
[
0
.
0
0
7
3
]
*
*
0
.
0
4
[
0
.
0
2
7
6
]
6
0
6
.
9
8
[
6
0
9
.
8
5
4
9
]
0
.
0
5
[
0
.
1
0
6
6
]
0
.
0
0
[
0
.
0
0
8
9
]
0
.
0
0
[
0
.
0
0
2
9
]
0
.
0
2
[
0
.
0
9
3
8
]
0
.
0
1
[
0
.
0
3
1
1
]
F
e
m
a
l
e
0
.
0
5
[
0
.
0
2
7
9
]
*
0
.
0
7
[
0
.
0
4
7
3
]
0
.
0
2
[
0
.
0
0
8
7
]
*
*
0
.
0
3
[
0
.
0
1
0
7
]
*
*
0
.
0
3
[
0
.
0
3
2
8
]
1
0
7
1
.
5
6
[
1
,
0
3
4
.
9
9
6
0
]
0
.
0
6
[
0
.
1
4
7
7
]
0
.
0
1
[
0
.
0
1
4
5
]
0
.
0
0
[
0
.
0
0
0
8
]
*
*
0
.
1
1
[
0
.
1
3
3
7
]
0
.
0
4
[
0
.
0
3
8
2
]
M
a
l
e
0
.
0
3
[
0
.
0
3
1
0
]
0
.
0
6
[
0
.
0
5
1
9
]
0
.
0
0
[
0
.
0
0
9
5
]
0
.
0
0
[
0
.
0
0
9
5
]
0
.
0
6
[
0
.
0
3
4
0
]
*
4
8
.
2
4
[
4
1
1
.
7
4
1
1
]
0
.
0
3
[
0
.
1
3
6
1
]
0
.
0
0
[
0
.
0
0
8
5
]
0
.
0
0
[
0
.
0
0
6
3
]
0
.
0
7
[
0
.
1
2
9
6
]
0
.
0
1
[
0
.
0
4
9
9
]
C
h
i
l
d
r
e
n

l
i
v
i
n
g

i
n

u
r
b
a
n

a
r
e
a
s
0
.
0
3
[
0
.
0
1
4
3
]
*
*
0
.
0
2
[
0
.
0
2
1
6
]
0
.
0
1
[
0
.
0
0
5
4
]
0
.
0
1
[
0
.
0
0
6
3
]
0
.
0
5
[
0
.
0
2
1
1
]
*
*
1
4
8
.
6
0
[
5
6
0
.
3
7
7
6
]
0
.
0
1
[
0
.
0
8
4
7
]
0
.
0
1
[
0
.
0
0
8
7
]
0
.
0
0
[
0
.
0
0
5
0
]
0
.
0
1
[
0
.
0
7
7
1
]
0
.
0
2
[
0
.
0
2
3
3
]
F
e
m
a
l
e
0
.
0
5
[
0
.
0
2
1
2
]
*
*
0
.
0
4
[
0
.
0
3
1
2
]
0
.
0
0
[
0
.
0
0
7
6
]
0
.
0
1
[
0
.
0
1
0
0
]
0
.
0
1
[
0
.
0
2
4
2
]
3
5
9
.
7
6
[
9
7
9
.
4
1
7
2
]
0
.
0
2
[
0
.
1
1
9
0
]
0
.
0
1
[
0
.
0
1
1
6
]
0
.
0
0
[
0
.
0
0
6
4
]
0
.
0
5
[
0
.
0
9
5
2
]
0
.
0
4
[
0
.
0
3
4
9
]
M
a
l
e
0
.
0
2
[
0
.
0
1
9
6
]
0
.
0
1
[
0
.
0
3
0
3
]
0
.
0
1
[
0
.
0
0
7
9
]
0
.
0
1
[
0
.
0
0
8
2
]
0
.
0
9
[
0
.
0
2
6
9
]
*
*
*
1
2
1
.
8
7
[
4
7
9
.
3
1
7
8
]
0
.
0
4
[
0
.
1
1
1
3
]
0
.
0
1
[
0
.
0
1
2
3
]
0
.
0
0
[
0
.
0
0
7
2
]
0
.
0
2
[
0
.
1
0
7
2
]
0
.
0
0
[
0
.
0
2
8
6
]
*
*
*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
0
1
;

*
*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
0
5
;

*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
1
0
.
N
o
t
e
:

N
u
m
b
e
r
s

i
n

b
r
a
c
k
e
t
s

a
r
e

s
t
a
n
d
a
r
d

e
r
r
o
r
s
.
S
o
u
r
c
e
:

A
u
t
h
o
r
s

c
a
l
c
u
l
a
t
i
o
n
s

b
a
s
e
d

o
n

1
9
9
3
,

1
9
9
7
,

a
n
d

2
0
0
0

I
n
d
o
n
e
s
i
a
n

F
a
m
i
l
y

L
i
f
e

S
u
r
v
e
y
s
.
T
A
B
L
E

7
.
6

E
f
f
e
c
t
s

o
f

h
e
a
l
t
h

i
n
s
u
r
a
n
c
e

o
n

u
s
e

a
n
d

h
e
a
l
t
h

s
t
a
t
u
s

f
o
r

c
h
i
l
d
r
e
n
,

b
y

e
x
p
e
n
d
i
t
u
r
e

d
i
s
t
r
i
b
u
t
i
o
n
I
n
d
i
v
i
d
u
a
l

f
x
e
d

e
f
f
e
c
t
s

o
n

c
h
i
l
d
r
e
n

a
g
e
s

1
5

o
r

y
o
u
n
g
e
r

a
t

b
a
s
e
l
i
n
e
H
a
d

a
n
y

o
u
t
p
a
t
i
e
n
t

c
a
r
e

l
a
s
t

4

w
e
e
k
s
N
u
m
b
e
r

o
f

t
i
m
e
s

t
h
a
t

h
a
d

o
u
t
p
a
t
i
e
n
t

c
a
r
e

l
a
s
t

4

w
e
e
k
s
H
a
d

a
n
y

i
n
p
a
t
i
e
n
t

c
a
r
e

l
a
s
t

1
2

m
o
n
t
h
s
N
u
m
b
e
r

o
f

t
i
m
e
s

t
h
a
t

h
a
d

i
n
p
a
t
i
e
n
t

c
a
r
e

l
a
s
t

1
2

m
o
n
t
h
s
H
o
u
s
e
h
o
l
d

h
e
a
l
t
h

e
x
p
e
n
d
i
t
u
r
e
s

(
>
0
)
M
o
n
t
h
l
y

p
e
r

c
a
p
i
t
a

e
x
p
e
n
d
i
t
u
r
e

o
n

h
e
a
l
t
h
(
r
e
a
l

r
u
p
i
a
h
s
)
B
o
d
y

m
a
s
s

i
n
d
e
x

(
B
M
I
)
O
v
e
r
w
e
i
g
h
t

(
B
M
I

b
e
t
w
e
e
n

2
5
2
9
.
9

k
g
/
m
2
)
O
b
e
s
i
t
y

(
B
M
I
>
=
3
0

k
g
/
m
2
)
H
e
m
o
g
l
o
b
i
n

(
g
/
d
l
)
L
o
w

h
e
m
o
g
l
o
b
i
n

(
i
r
o
n

d
e
f
c
i
t
:

h
b
1
2

w
o
m
e
n
,

h
b
<
1
3
.
5

m
e
n
,

h
b
<
1
1

c
h
i
l
d
)
C
h
i
l
d
r
e
n

i
n

t
o
p

5
0
%

o
f

e
x
p
e
n
d
i
t
u
r
e

d
i
s
t
r
i
b
u
t
i
o
n
0
.
0
3
0
5
[
0
.
0
1
4
6
]
*
*
0
.
0
2
7
3
[
0
.
0
2
3
3
]
0
.
0
0
6
5
[
0
.
0
0
5
5
]
0
.
0
0
8
1
[
0
.
0
0
6
4
]
0
.
0
5
1
7
[
0
.
0
1
9
9
]
*
*
*
7
.
9
5
0
4
[
5
5
2
.
6
4
1
5
]
0
.
0
4
4
8
[
0
.
0
8
1
9
]
0
.
0
1
7
2
[
0
.
0
0
8
0
]
*
*
0
.
0
0
0
7
[
0
.
0
0
4
5
]
0
.
0
6
6
1
[
0
.
0
7
4
7
]
0
.
0
2
9
3
[
0
.
0
2
2
6
]
F
e
m
a
l
e
0
.
0
4
8
9
[
0
.
0
2
0
4
]
*
*
0
.
0
5
0
5
[
0
.
0
3
1
4
]
0
.
0
1
0
1
[
0
.
0
0
7
7
]
0
.
0
1
2
3
[
0
.
0
0
9
9
]
0
.
0
1
5
8
[
0
.
0
2
2
7
]
1
5
.
2
1
6
4
[
9
3
0
.
6
2
1
3
]
0
.
0
8
5
6
[
0
.
1
1
2
3
]
0
.
0
2
4
6
[
0
.
0
1
1
4
]
*
*
0
.
0
0
4
4
[
0
.
0
0
5
7
]
0
.
1
0
5
[
0
.
0
9
1
6
]
0
.
0
5
6
5
[
0
.
0
3
1
6
]
*
M
a
l
e
0
.
0
1
3
9
[
0
.
0
2
0
9
]
0
.
0
0
6
2
[
0
.
0
3
4
4
]
0
.
0
0
3
5
[
0
.
0
0
7
8
]
0
.
0
0
4
6
[
0
.
0
0
8
1
]
0
.
0
8
8
6
[
0
.
0
2
5
3
]
*
*
*
5
7
.
3
1
2
1
[
5
0
2
.
5
4
1
1
]
0
.
0
1
1
5
[
0
.
1
0
9
9
]
0
.
0
0
7
5
[
0
.
0
1
1
0
]
0
.
0
0
4
9
[
0
.
0
0
7
1
]
0
.
0
3
2
[
0
.
1
0
7
0
]
0
.
0
0
5
3
[
0
.
0
2
9
9
]
C
h
i
l
d
r
e
n

i
n

b
o
t
t
o
m

5
0
%

o
f

e
x
p
e
n
d
i
t
u
r
e

d
i
s
t
r
i
b
u
t
i
o
n
0
.
0
4
0
7
[
0
.
0
1
9
7
]
*
*
0
.
0
5
2
3
[
0
.
0
2
7
9
]
*
0
.
0
1
1
3
[
0
.
0
0
5
4
]
*
*
0
.
0
1
3
7
[
0
.
0
0
6
4
]
*
*
0
.
0
3
5
[
0
.
0
3
2
3
]
1
,
1
3
5
.
1
1
[
6
2
7
.
3
4
7
4
]
*
0
.
0
1
0
9
[
0
.
1
1
5
3
]
0
.
0
0
7
8
[
0
.
0
1
1
0
]
0
.
0
0
0
5
[
0
.
0
0
5
3
]
0
.
1
1
8
9
[
0
.
1
0
2
1
]
0
.
0
1
9
9
[
0
.
0
3
3
3
]
F
e
m
a
l
e
0
.
0
5
3
6
[
0
.
0
3
0
3
]
*
0
.
0
5
5
5
[
0
.
0
4
7
5
]
0
.
0
1
0
1
[
0
.
0
0
7
5
]
0
.
0
1
5
7
[
0
.
0
1
0
2
]
0
.
0
1
6
5
[
0
.
0
3
8
4
]
2
,
2
3
5
.
0
8
[
1
,
1
8
7
.
3
9
0
8
]
*
0
.
0
2
7
7
[
0
.
1
6
8
7
]
0
.
0
2
5
2
[
0
.
0
1
4
5
]
*
0
.
0
0
0
9
[
0
.
0
0
5
6
]
0
.
0
4
8
7
[
0
.
1
4
5
1
]
0
.
0
0
9
4
[
0
.
0
4
7
8
]
M
a
l
e
0
.
0
3
1
2
[
0
.
0
2
5
5
]
0
.
0
5
5
2
[
0
.
0
3
3
9
]
0
.
0
1
0
9
[
0
.
0
0
9
7
]
0
.
0
1
0
3
[
0
.
0
0
9
7
]
0
.
0
5
4
6
[
0
.
0
4
0
4
]
2
8
.
3
9
8
3
[
2
6
9
.
1
9
2
3
]
0
.
0
2
0
6
[
0
.
1
3
8
6
]
0
.
0
1
1
1
[
0
.
0
1
5
4
]
0
.
0
0
0
6
[
0
.
0
0
6
8
]
0
.
2
3
0
3
[
0
.
1
3
4
4
]
*
0
.
0
3
1
2
[
0
.
0
4
6
4
]
*
*
*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
0
1
;

*
*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
0
5
;

*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
1
0
.
N
o
t
e
:

N
u
m
b
e
r
s

i
n

b
r
a
c
k
e
t
s

a
r
e

s
t
a
n
d
a
r
d

e
r
r
o
r
s
.
S
o
u
r
c
e
:

A
u
t
h
o
r
s

c
a
l
c
u
l
a
t
i
o
n
s

b
a
s
e
d

o
n

1
9
9
3
,

1
9
9
7
,

a
n
d

2
0
0
0

I
n
d
o
n
e
s
i
a
n

F
a
m
i
l
y

L
i
f
e

S
u
r
v
e
y
s
.
134 Chapter 7
particular, girls in both lower and higher income households show signicant
impacts of health insurance on the use of outpatient care. Insurance also increases
the use of inpatient care for children in lower income households. Yet it decreases
the likelihood of having any health expenditures among the rich by about 5 per-
centage points, with no impact on poorer households. For poorer insured house-
holds there is an increase in per capita spending for girls. For health status variables
there are some signicant impacts, with some inconsistencies. Health insurance
is associated with higher proportions of overweight children in upper income
households and a lower proportion of overweight female children in lower income
groups. Health insurance is also signicantly associated with lower rates of anemia
for girls in upper income groups.
Te results are for all children younger than age 15. But health insurance might
be more important for health outcomes for younger children than older children,
particularly in the rst few years of life. So we repeat our analysis of the impacts
of insurance for three groups of children, those ages 3 and younger in 1993, those
ages 5 and younger, and those ages 8 and younger. We might expect that any
impacts of health insurance would be stronger for the younger age groups, but
concentrating on these age groups signicantly reduces the sample size. We also
include child height as an extra outcome variable.
While the coecient on the impact of insurance on health care use continues
to be positive, the results generally lose signicance, consistent with the smaller
sample size or suggesting that insurance might have less impact on health use for
younger age groups. For health status measures, for younger children, the results
show an overall signicant reduction in obesity for children under age 3 and chil-
dren under age 5 at baseline and a slight increase for young girls in the proportion
of those overweight for those insured. But for female children in rural areas, there
is a positive impact of insurance on increasing both body mass index and the prob-
ability of being overweight (but not obese). Young male children under age 3 show
a signicant reduction in the proportion obese of about 3 percentage points. Tere
are no signicant impacts of insurance on child height.
Conclusion
Our ndings show consistently that health insurance increases the use of both
inpatient and outpatient care over the length of the panel, for both adults and
children. Tere is also evidence that health insurance reduces the probability of
any health care spending; nevertheless, while the eect on the overall amount of
per capita spending is consistently negative, it is not statistically signicant. It is
important to note some dierences by subgroup in the impacts of insurance. In
particular, lower income groups tend to show higher impacts of use with insurance,
and there is some evidence of higher impacts of insurance in rural areas than in
urban areas.
For health status, the evidence is less uniform, with many indicators of health
status for both adults and children showing little relationship with insurance cover-
age. Tere are, however, a few exceptionsparticularly some small positive eects
of insurance on reducing problems associated with activities of daily living for
adults and a potential impact of reducing high blood pressure for adults in lower
income groups. But for children there appear to be few impacts of health insur-
ance on health status indicators. Tere are some suggestions that insurance reduces
child obesity for some age groups but increases it for others, so there are no obvious
conclusions here. Child height is unaected by insurance status in all of the groups
studied.
How can insurance increase overall use of services without increasing, at least
in clearly measurable ways, health status? One possibility might be that those
without insurance nd ways of coping when faced with health issues that might
aect their health status. For example, they see free or low cost public providers,
use savings, and ask friends or pharmacists for assistance. Alternatively, illnesses
that substantially aect health status might be infrequent enough that impacts
on these rare events are dicult to observe. It may be that compared with other
factors aecting health, such as genetic, behavioral, or random components
insurance could at best have very small impacts. Finally, some aspects of health
are likely dicult to alter no matter how much attention is receivedfor
instance, obesity, an important indicator of health but notoriously dicult to
overcome.
In chapter 2 of this volume Giedion and Diaz argue that future studies of the
impact of health insurance should focus on health outcomes more likely to be alter-
able and measurable. We agree that this should form the basis for future research.
It seems unlikely that health insurance might directly aect many standard self-
reported measures such as days ill or getting a fever or headache. It seems equally
unlikely that insurance would aect the probability of certain types of illness, such
as breast cancer. Insurance might, however, promote earlier diagnoses of chronic
illnesses, though for this study we did not have information on such variables.
Finally, there may be contexts where increased use does not necessarily lead to
an improvement in health because the client was either not particularly ill or the
ailment does not respond well to treatment, such as the common cold. In these
circumstances, use might mitigate symptoms but not necessarily measured physi-
cal health.
136 Chapter 7
Notes
1. See Setiana (2005) for a short but comprehensive history of the health nancing policy
in Indonesia as well as the reform program.
2. Instead of Jamsostek, rms may opt for a system where the rm reimburses the
employee for medical expenditures.
3. We also carried out descriptive analysis comparing the groups of those always having
insurance to those never having insurance. But these results were qualitatively similar
to those presented here, so we omit them to save space.
References
Setiana, Adang. 2005. Social Health Insurance Development as an Integral Part of the National
Health Policy: Recent Reform in Indonesian Health Insurance System. Paper presented at
the International Conference on Social Health Insurance in Developing Countries, December
57, Berlin. [www.socialhealthprotection.org/pdf/SHI-ConfReader_Druckkomprimiert.pdf ].
. 2008. Investing in Indonesias Health: Challenges and Opportunities for Future Public
Spending. Health Public Expenditure Review. Washington, DC: World Bank.
137
C
h
a
p
t
e
r

8
In 2002 the Chinese government announced a new national policy for
rural health carethe New Cooperative Medical Scheme (NCMS).
First rolled out in a small number of pilot counties in 2003 and targeted
to cover the entire rural population by 2010, the goals are to improve
access to health care and reduce inequality and medical impoverish-
ment. Te government has allocated new resources to the scheme, tar-
geting the poor western and central regions. Te national policy guide-
lines for the scheme have only two requirements: voluntary enrollment
and priority to cover catastrophic health expenditures. Apart from this,
local governments are free to design their own programs, turning China
into a laboratory for experimentation.
To assist China in developing a rural health care system tailored to
conditions in poorer regions and designed to be sustainable in the long
run, we conducted a social experiment of a community-based prepay-
ment schemeRural Mutual Health Care (RMHC)following the
national guidelines but augmented with other interventions to improve
quality and eciency.
Te primary objective of this chapter is to empirically evaluate the
RMHCs impact on access to care, nancial risk protection, and health
status. Using a pre-post treatment-control study design and longitudi-
nal household/individual surveys one year before the interventions and
annually for three years after the interventions, we estimate the impact
eects of the RMHC, combining dierence-in-dierence estimation
The Impact of a Social
ExperimentRural Mutual Health
Careon Health Care Use,
Financial Risk Protection, and
Health Status in Rural China
Winnie Yip and William Hsiao
138 Chapter 8
with propensity score matching to control for observable and unobservable time-
invariant dierences between the treatment and control groups.
Rural health care in China
Key challenges confronting the rural health care system
From the early 1950s to 1980 Chinas strategy for rural health care emphasized
prevention and basic health care. It developed a three-tiered organization for deliv-
ery of health care. In rural areas this consisted of village health posts, township
health centers, and county hospitals, which together provided a structure for e-
cient patient referrals to treat health problems. Te Cooperative Medical System
provided nearly universal insurance coverage in rural areas. Financed primarily by
the welfare fund of the communes (collective farms), the system organized health
stations, paid village doctors to deliver primary care, and provided drugs. It also
partially reimbursed patients for services received at township and county facili-
ties. At its peak in 1978 it covered 90% of Chinas rural population, making basic
health care accessible and aordable and oering peasants nancial protection
against large medical expenses.
When China reformed its rural economy in 1979 and introduced the House-
hold Responsibility System, the communes disappeared, and without this fund-
ing base, the Cooperative Medical System collapsed, leaving 90% of all peasants
uninsured. Village doctors became private practitioners with little government
oversight, earning their income from patients on a fee-for-service basis. Further,
like all transition economies, China experienced a drastic reduction in the govern-
ments capacity to fund health care as government revenue shrank. Government
subsidies as a share of public health facilities total revenues fell to a mere 10% by
the early 1990s. To keep health care aordable, the government maintained strict
price controls by setting prices for basic health care below cost. At the same time,
the government wanted facilities to survive nancially, so it set prices for new and
high-tech diagnostic services above cost and allowed a 15% prot margin on drugs.
Tese policies created perverse incentives for providers who had to generate
90% of their budget from revenue-producing activities, turning hospitals, town-
ship health centers, and village doctors alike into prot-seeking entities. Provid-
ers overprescribe drugs and tests while hospitals race to introduce high-tech ser-
vices and expensive imported drugs that give them higher prot margins (Liu and
Mills 1999). To increase their prots village doctors often buy cheap counterfeit
or expired drugs and sell them to patients at the higher ocial price (Blumenthal
and Hsiao 2005). Referrals within the three-tiered delivery system also collapsed,
Impact of Rural Mutual Health Care on Use, Financial Risk, and Health Status in Rural China 139
as each level of provider competed with the other for patients, duplicating tests and
services. Health care spending soared, growing 16% a year7 percentage points
faster than gross domestic product (GDP) growthand patient out-of-pocket
health spending also grew at an average of 16% from 1978 to 2003 (Blumenthal
and Hsiao 2005; Smith, Wong, and Zhao 2005).
In less than two decades China had transformed its rural health care system
from one that provided prevention and aordable basic health care for all to one in
which people could not aord basic health care, driving many families into pov-
erty (Hesketh and Zhu 1997a, 1997b; Hsiao 1984; Lindelow and Wagsta 2005;
Watts 2006, 2007). Te 2003 National Health Survey found that 46% of the rural
Chinese who were ill did not seek health care, and among them, 40% cited cost
as the main reason (Center for Health Statistics and Information 2004). Another
22% of those advised by physicians to be hospitalized refused to do so because they
could not aord it. Of those who did become hospitalized, about 35% discharged
themselves against their doctors advice because of cost. Studies have found that
30%40% of those below the poverty line attributed their poverty to medical
expenditures (Center for Health Statistics and Information 2004; Watts 2006).
Chinas national policy for rural health care
In 2002 the Chinese government announced the NCMS, but conspicuously absent
from its stated goals are improving health outcomes and reducing ineciencies in
health care delivery. Targeted to cover the entire rural population by 2010, more
than 90% of the rural population was covered by the end of 2008.
For the schemes initial waves the government subsidized each farmer in west-
ern and central provinces with 20 RMB (1 RMB = US$0.125), shared equally
between the central and local governments, if the farmer pays an annual premium
of at least 10 RMB to enroll (Central Committee of CPC 2002; Watts 2006). Te
subsidy was increased to 40 RMB (US$5) in 2006, then again to 80 RMB in 2007,
100 RMB in 2008, and 120 RMB in 2009, with the individuals contribution
rising to 20 RMB (Anonymous 2009). Exactly how such a scheme will address
the multiple challenges in rural health care has been left open, and the central
government encourages local governments to experiment with dierent workable
schemes.
Rural mutual health care
Beyond the governments limited goals for the NCMS, the RMHC project aimed
to improve the eciency and quality of health care and the health status of people.
Te project simulated the government subsidy of 20 RMB for each villager who
140 Chapter 8
prepaid a premium to enroll. It followed the two government guidelines of volun-
tary enrollment and coverage of hospitalization. But to improve quality and e-
ciency, it adopted several features targeted at the village doctors, who, because of
their convenient location, provide most of the services.
Design of rural mutual health care
Te RMHC has three major design features: nancing and benet packages; orga-
nization of service provision, including provider payment method; and the use of
community governance in management.
Financing and benet packages
Te RMHC integrated both the (simulated) government subsidies and the villag-
ers premiums into one single risk-pooled fund to cover primary care as well as
hospital services. Coverage of primary care is a cost-eective way to improve health
outcomes, providing incentives for the patients to use basic and primary health care
rather than to seek care in hospitals. It also makes villagers more willing to prepay
into the RMHC and reduces adverse selection. In any voluntary scheme, villagers
are more willing to enroll if the expected benets of enrollment exceed the cost of
premiums. Because the distribution of health risks is such that a small proportion
of individuals use a large share of total health expenditure, as shown in the contin-
gency table (table 8.1), a scheme that covers only hospitalization would more likely
attract the old and the sick, making it nancially unsustainable.
We oered three benet packages with annual premiums ranging from
12 RMB to 18 RMB. All covered primary care, hospital services, and drugs at
all levels of facilities with no deductibles. To nance such comprehensive cover-
age with limited funding (3238 RMB per person, when health expenditures per
person were about 150180 RMB), coinsurance rates ranged from 55% to 60%,
with rates for visits to village clinics lower than those for visits to higher facilities,
to encourage use of the village clinic. Ceilings for hospitalizations were also intro-
duced (400 RMB for admissions to township health centers and 8,000 RMB for
admissions to county hospitals).
Organization of provision and provider payment
By covering both primary care and hospitalization the RMHC had the nancial
power to introduce interventions improving the eciency and quality of the deliv-
ery system by changing the organizational and incentive structure on the supply
side. Te RMHC Fund Oce, a single purchaser, selected the best village doctors
(often two to three per village) on a competitive basis. Since no clinical performance
records exist at the village level in China (as in most countries), selection is based
on a combination of qualications and villager voting. Te oce contracted with
these village doctors, compensating them with a salary plus a bonus based on per-
formance measures, including conforming to established protocols of treatment for
common diseases such as upper respiratory infection and diarrhea; maintaining
patient medical records; delivering public health functions such as immunizations;
and receiving high patient satisfaction ratings. Tis de-linked village doctors com-
pensation from their drug-dispensing activities and aimed to reduce overprescrib-
ing and sales of fraudulent drugs.
Te contracts and their annual renewals also allowed the RMHC to screen
and regulate village doctors. Provider contracts would explicitly outline provider
responsibilities and payment, restricting village doctors to tasks within their level
of competency (whereas they previously competed with township and county
providers) and to primary care and prevention. In addition, village doctors were
not allowed to purchase drugs directly. Instead, township health centers (covering
10,00015,000 people) purchased drugs in bulk through competitive bidding and
TABLE 8.1
Contingency table on the distribution of households total annual health
expenditure in Zhangjai Town, Zhenan County, 1999
Concentration of households
(%)
Accumulated amount spent
(RMB)
Accumulated spending as a share
of total health expenditure (%)
1 96,040 20.7
5 242,530 42.1
10 330,378 61.0
20 408,967 77.9
30 442,590 85.2
40 458,602 92.6
50 463,577 97.7
60 464,831 99.9
70 465,124 100.0
80 465,124 100.0
90 465,124 100.0
100 465,124 100.0
Note: Population of 9,784.
Source: Authors compilation of claims data from Zhangjai Towns Cooperative Medical
System.
142 Chapter 8
distributed them directly to village doctors through a central distribution system,
helping to assure drug safety at minimum cost. Doctors not selected saw their
patient loads drastically reduced because villagers who enrolled in the RMHC
could receive reimbursements only if they consulted the contracted village doctors.
In our experiment, in which close to 80% of the villagers enrolled, village doctors
had very high incentives to improve their performance to increase their chances of
being selected.
Community governance
Te RMHC used community governance rather than management by the govern-
ment alone. Since villagers had an interest in ensuring that they would benet from
the scheme, they were in the best position to see that the funds were properly used and
to choose the most attractive benet package for the community. Each village elected
a representative to serve with government ocials, township health center directors,
and town nancial auditors on the Fund Board at the township level, where the risks
were pooled. Te board decided on the benet package that best reected their fellow
villagers preferences; for example, all chose coverage of primary care and drugs at
the village level because villagers desired basic health care and drugs at nearby loca-
tions. Te board also managed and controlled the Fund Oce that nanced and
contracted with service providers. To monitor the daily activities of village doctors, vil-
lagers elected ve volunteers to form a village management committee to check village
clinics hours of opening, their cleanliness, whether essential drugs are available, and
most important the attendance of village doctors at their clinics.
Te prevailing New Cooperative Medical Scheme model in the western and
central regions
Many counties in the western and central regions adopted a model that combines
an individual medical savings account with high-deductible catastrophic insur-
ance. Typically, this scheme collects 10 RMB from the farmer and assigns an aver-
age of 8 RMB (US$1) to an individual savings account that can be used by the
farmer to pay for outpatient visits. Te governments 20 RMB subsidy plus the 2
RMB remaining premium would be used for risk-pooling to cover inpatient hos-
pital expenses that exceeded a deductible (for example, the NCMS site near our
RMHC intervention site has a deductible of 800 RMB). Besides the deductible, the
patients still have to pay 40%60% of covered inpatient hospital costs. Te benet
package also caps the benet payment between 10,00020,000 RMB (US$1,250
US$2,000) (Mao 2005; Ministry of Health 2007). But there is no supply side
intervention to deal with the waste caused by unnecessary treatments and drugs.
Te key features of the RMHC and NCMS benet packages commonly found
in the western and central regions are compared in table 8.2. In both models, the
premium is similar, at about 30 RMB per person.
Study design and data
We chose the study design, data collection, and analytical methods that would allow
us to conduct a prospective impact evaluation and draw evidence-based conclusions.
Study design
Te RMHC adopted a pre-post, treatment-control study design. RMHC was imple-
mented in three towns, one in Guizhou and two in Shaanxi provinces. We chose
these western provinces because our goal is to help China nd a rural health care
model suitable for poorer regions and because the health ocials in these provinces
invited us to conduct our experiments there. In these two provinces we rst iden-
tied towns representative of the socioeconomic conditions in Chinas low-income
regions and randomly selected three as our intervention sites. Together, the three
towns encompassed a population of about 60,000 people (15,000 in each of the two
towns in Shaanxi and 30,000 in Guizhou). We further selected two control towns
(one in each province) that matched the RMHC intervention sites in socioeconomic
conditions, availability of health facilities, and distance to city centers, based on avail-
able ocial statistics. Tese control sites did not experience any intervention in health
TABLE 8.2
Rural Mutual Health Care and New Cooperative Medical Scheme benet
packages
RMHC NCMS
Deductible 0 800 RMB
Individual savings account 0 Deposit 8 RMB each year
Reimbursement rate for outpatient visit
Village health posts 45% 0
Township health centers and above 40% 0
Reimbursement rate and caps for hospitalization
Township health centers 40%, capped at 400 RMB 40%60% of the amount
exceeding the deductible,
capped at 10,00020,000 RMB
County hospitals and above 40%, capped at 8,000 RMB 40%60% above deductible,
capped at 10,00020,000 RMB
Source: Authors design for Rural Mutual Health Care and authors estimates for New Coop-
erative Medical Scheme.
144 Chapter 8
care until 2006, when the government introduced its schemes. In 2003 the average
annual income per person was about 1,4001,800 RMB (US$175US$225) at the
study sites and, on average, villagers spent about 8%10% of their annual income on
health care. Te RMHC began its initial enrollment in October 2003 and went into
full operation immediately thereafter. To reduce adverse selection, enrollment was by
household. Te experiment was planned for three years and concluded in early 2007.
Data collection
Te ndings here are based primarily on analyses using data from longitudinal
household (and individual) surveys conducted one year prior to the intervention
(December 2002) and each of the three years after2004, 2005, and 2006. Te
same survey instruments were used for all waves and for both the intervention and
control sites. For this chapter we included data only from 2002 (baseline) and 2005
because in 2006, new interventions were introduced in the control sites, and 2004,
a year after the interventions, captures only short-term eects.
Within the three intervention and two control towns, we randomly selected 18
villages (the number of villages selected within a town was proportional to its popu-
lation). In the pre-intervention year we randomly selected one out of every three
households in each village, yielding a total sample of 2,329 households (8,582 indi-
viduals) in the intervention sites and 752 households (2,865 individuals) in the con-
trol sites. We successfully re-interviewed 87% (83%) of these households (individu-
als) in 2005. Attrition was primarily due to households migrating from the town.
Households dropped out in each wave were replaced with households with similar
income and household size. Response rates in both rounds were high, close to 98%.
Te household/individual questionnaire was designed to collect data measur-
ing three primary outcome variables: access to care, nancial risk protection, and
health status as well as a set of control variables (table 8.3).
The enrolled and the nonenrolled
Te RMHC achieved average enrollment rates of 78%, increasing from 70% in the
initial year to almost 85% in the nal year as villagers gained trust in the scheme
and experienced its benets.
Table 8.3 shows that our treatment sample, those who enrolled in the RMHC,
are dierent from those who chose to not enroll in the RMHC (since enrollment
is voluntary) and from those in the control site. Te enrolled-in RMHC sites had
worse health, had higher use rates, and were more likely to experience catastrophic
expenditure in the baseline than the nonenrolled. Te data reect adverse selec-
tionthat is, those who chose to enroll in RMHC are those who are more likely to
TABLE 8.3
Outcome variables at baseline
Variable
RMHC
Control Enrolled Nonenrolled
Use
a
Visit an outpatient provider in the two weeks prior to the survey (1/0) 0.173 0.094 0.132
Visit a village clinic in the two weeks prior to the survey (1/0) 0.141 0.056 0.087
Visit a township health center in the two weeks prior to the
survey (1/0)
0.022 0.032 0.028
Visit a county hospital in the two weeks prior to the survey (1/0) 0.010 0.006 0.017
Number of outpatient visits in the two weeks prior to the survey 0.352 0.185 0.220
Self-treat in the two weeks prior to the survey (1/0) 0.056 0.040 0.028
Hospitalized in the 12 months prior to the survey (1/0) 0.033 0.022 0.039
Catastrophic expenditure
b
Out-of-pocket health expenditure more than 10% income net of
food expenditure
0.296 0.211 0.259
More than 15% 0.245 0.190 0.227
More than 20% 0.212 0.164 0.193
More than 30% 0.171 0.130 0.147
Impoverishment
b
Percentage below $1/day: full sample 0.221 0.275 0.183
Percentage below $1/day: lowest 25% income sample 0.626 0.781 0.502
Health status (1=problem, 0=no problem)
c
Any of the fve dimensions with problem 0.49 0.37 0.375
Mobility 0.08 0.048 0.055
Self-care 0.05 0.030 0.036
Usual activity 0.11 0.058 0.103
Pain/discomfort 0.31 0.148 0.226
Anxiety/depression 0.40 0.180 0.307
Socioeconomics
Income per capita 1,885 1,700 2,481
Household wealth 0.65 0.68 0.51
Illiterate education 0.27 0.25 0.26
Primary education 0.46 0.44 0.40
Junior high education 0.22 0.25 0.27
Senior high education 0.03 0.04 0.05
Tertiary education 0.01 0.02 0.02
Sociodemographics
Male 0.50 0.59 0.51
(continued)
146 Chapter 8
have higher use and expenditures, including the less healthy and older individuals.
Tis supports our choice of using the control site, rather than the nonenrolled as
the comparison group. Comparing the sample that enrolled in the RMHC with
those in the control sites, the dierences are smaller than those observed with the
nonenrolled, but dierences still exist.
Table 8.4 shows the balancing properties of the propensity score matching and
resulting reductions in observable dierences between the treatment and control
Variable
RMHC
Control Enrolled Nonenrolled
Age 40 36 42
Single 0.17 0.34 0.09
Married 0.76 0.61 0.86
Divorced/separated 0.01 0.01 0.01
Widowed/other 0.06 0.04 0.04
Migrant 0.07 0.11 0.04
Health status
Ill in last month 0.26 0.16 0.17
1+ chronic conditions 0.17 0.13 0.14
Current smoker 0.34 0.32 0.41
Current drinker 0.27 0.27 0.19
Very good sexual and reproductive health 0.10 0.14 0.11
Good sexual and reproductive health 0.24 0.31 0.19
Average sexual and reproductive health 0.41 0.38 0.51
Bad/very bad sexual and reproductive health 0.24 0.17 0.18
Household characteristics
Household size 4.0 4.1 3.8
Distance from village clinic (miles) 2.1 2.1 2.5
Distance from township health center (miles) 14 13 10
Distance from county hospital (miles) 65 74 63
a. Unit of observation is individuals older than age 15 who self-responded and children under
age 15 proxied by their parents. Sample sizes for the enrolled, nonenrolled, and controls were
2,998, 1,134, and 1,745, respectively.
b. Unit of observation is households. Sample sizes for the enrolled, nonenrolled, and controls
were 1,519, 507, and 692, respectively.
c. Unit of observation is individuals older than age 15 who self-responded (as opposed to
using proxies) to the surveys. Sample sizes for the enrolled, nonenrolled, and controls were
1,665, 610, and 1,219, respectively.
Source: Authors.
Outcome variables at baseline
TABLE 8.4
Matching balancing properties between the Rural Mutual Health Care and
controls
Prematching
Postmatching
Kernel
Standardized
difference
a
t-statistic
Standardized
difference
a
t-statistic
Bias
reduction
b
Socioeconomics
Log (income/capita) 42.2 12.86 1.4 0.59 96.6
Log (income/capita) squared 42.8 13.1 1.5 0.62 96.6
Household wealth 17.3 5.25 5.7 2.34 67
Household wealth squared 1.6 0.49 4.1 1.62 149.9
Primary education 12.7 3.72 6.1 2.32 52
Junior high education 11.1 3.32 2.9 1.12 74.4
Senior high education 9.9 3.05 6.1 2.43 38.6
Tertiary education 7.5 2.37 1.8 0.81 75.6
Sociodemographics
Male 0.5 0.16 1.9 0.71 239.4
Age 10.4 2.95 11.1 4.04 7.1
Age squared 4 1.15 9 3.21 126
Married 25 7.07 11.8 4.3 53
Divorced/separated 4 1.12 2.5 0.9 37.9
Widowed/other 6.5 1.86 3.7 1.36 43.9
Migrant 15.4 4.28 1.5 0.51 90.1
Health status
Ill in last month 24.4 6.93 3.5 1.24 85.6
1+ chronic conditions 10.5 3.02 1.6 0.58 84.9
Current smoker 13.1 3.89 4.9 1.9 62.7
Current drinker 17.6 5.06 4.3 1.56 75.4
Good sexual and reproductive health 11.8 3.4 0.8 0.29 93.4
Average sexual and reproductive health 19.7 5.81 5.4 2.08 72.5
Bad/very bad sexual and reproductive health 14.7 4.23 4.4 1.61 70.4
Household characteristics
Household size 16.3 4.71 1.4 0.52 91.6
Distance from village clinic
c
22.6 6.79 6.8 2.85 69.8
Distance from township health center
c
47.5 14.06 8.2 3.48 82.8
Distance from county hospital
d
6.0 1.64 18.7 7.33 209.8
LR-Chi square (27 df) 1,242.52 180.68
a. The raw differences in intervention/control sample means as a percentage of the square
root of the average of the intervention/control sample variances respectively.
b. The percentage reduction in standardized differences.
c. 010 or more miles, increments of 1 mile.
d. 0100 or more miles, increments of 10 miles.
Source: Authors.
148 Chapter 8
groups. Matching has reduced the baseline dierences signicantly for the major-
ity of observable characteristics between the two groups. In addition, we also con-
trolled for changes in these variables between the baseline and 2005 in the estima-
tion. Standard errors were bootstrapped and clustering at the household level was
accounted for.
Impacts of the Rural Mutual Health Care Scheme
Table 8.5 presents the dierence-in-dierence plus propensity score matching esti-
mates, or the impact estimates, and the baseline values for comparison.
Access to care
Te estimates show that the RMHC increased the probability of an outpatient
visit by 0.12 (p < 0.01), from a baseline of 0.173. Te increase primarily represented
visits to village doctors, followed by visits to township health centers. Te number
of visits increased by 0.155 (p < 0.01) from a baseline of 0.35. Te RMHC also
reduced the probability of self-medication by about two-thirds (by 0.038 from a
baseline of 0.056; p < 0.05). But we did not nd any statistically signicant impact
on hospitalization.
Te RMHC has, in addition, some distribution eects. Te lowest and highest
income individuals experienced the greatest increases in outpatient use of village
doctors, while the middle-income group experienced a substantial increase in the
use of township health center services (results not presented here). A full benet-
incidence analysis is beyond the scope of this chapter because, in particular, with-
out knowing the content of the services, we cannot assess how much of the increase
in township health center use is health improving (a real benet) and how much is
waste (with a neutral or negative eect on health).
Financial risk protection
Dening catastrophic health spending as out-of-pocket health spending greater
than 10%, 15%, 20%, and 30% of household income net of food expenditures, the
RMHC reduced the rates of catastrophic health spending by 0.075 (from 0.296;
p < 0.05), 0.076 (from 0.245; p < 0.05), 0.28 (from 0.212; p = 0.44) and 0.050 (from
0.171; p = 0.14), respectively. It also reduced medical impoverishment by 0.129
(from a baseline rate of 0.626; p < 0.05) for those in the lowest income quartile.
Health status
Te RMHC signicantly reduced the probability of having a problem in any of
the ve self-reported dimensions: mobility, self-care, usual activities (work, study,
housework, family, or leisure), pain/discomfort, and anxiety/depression. Te impact
estimate is 0.244 (p < 0.01)that is, adjusting for dierences between the treat-
ment and comparison groups, the treatment group experienced a greater reduction
(by 0.244 percentage points) in the probability of having a problem in any of the
ve dimensions after the RMHC than did the comparison group. Compared with
TABLE 8.5
Impacts of the Rural Mutual Health Care Scheme
Baseline
Impact
estimates
95% confdence
interval p-value
Use
Visit an outpatient provider in the last 2 weeks (1/0) 0.173 0.122 0.0701, 0.1748 0.000
Visit a village clinic in the last 2 weeks (1/0) 0.141 0.098 0.059, 0.138 0.000
Visit a township health center in the last 2 weeks (1/0) 0.022 0.020 0.002, 0.039 0.030
Visit a county hospital in the last 2 weeks (1/0) 0.010 0.001 0.018 0.019 0.926
Number of outpatient visits in the last 2 weeks 0.352 0.155 0.0516, 0.2589 0.003
Self-treat in the last 2 weeks (1/0) 0.056 0.038 0.0682, 0.0072 0.016
Hospitalized in the last year (1/0) 0.033 0.009 0.0300, 0.0110 0.365
Catastrophic expenditure
Out-of-pocket health expenditure more than10%
income net of food expenditure
0.296 0.075 0.1489, 0.0011 0.047
More than 15% 0.245 0.076 0.1497, 0.0021 0.044
More than 20% 0.212 0.028 0.1003, 0.0439 0.444
More than 30% 0.171 0.050 0.1159, 0.0160 0.137
Impoverishment
% below $1/day: full sample 0.221 0.024 0.0822, 0.0344 0.421
% below $1/day: lowest 25% income sample 0.626 0.129 0.2370, 0.0204 0.020
Health status (1=problem, 0=no problem)
Any of the fve dimensions with problem 0.49 0.244 0.3106, 0.1773 0.000
Mobility 0.08 0.024 0.0502, 0.0019 0.069
Self-care 0.05 0.007 0.0302, 0.0171 0.587
Usual activity 0.11 0.018 0.0540, 0.0178 0.322
Pain/discomfort 0.31 0.095 0.1528, 0.0376 0.001
Anxiety/depression 0.40 0.252 0.3185, 0.1859 0.000
Note: The impact estimates are based on the differences-in-differences, combined with
propensity score matching estimation. The results shown here used the kernel matching
algorithm. We also conducted sensitivity analyses using other matching algorithms, such as
nearest four neighbor and local linear matching, and the results do not change the conclu-
sions. Changes in income, household wealth, and other household and individual characteris-
tics that occurred between baseline and 2005 were also controlled for in the estimation.
Source: Authors.
150 Chapter 8
a baseline value of 0.49, a reduction of about 49%. We also found that the RMHC
signicantly reduced the probability of pain/discomfort and anxiety/depression by
0.095 (p < 0.01) and 0.252 (p < 0.01), from baselines of 0.31 and 0.40, respectively.
We did not nd statistically signicant reduction in usual activity, problems
with mobility, or use of self-care. To investigate why the RMHC did not have any
signicant impact on these dimensions, we examined subgroup dierences by age
and gender. For those older than 55, the RMHC signicantly reduced the prob-
ability of having problems in mobility (impact estimate = 0.060; baseline = 0.193;
p < 0.01) and usual activity (impact estimate = 0.094; baseline = 0.261; p < 0.01).
Tose older than 55 experienced a reduction in probability of 0.096 (baseline =
0.449; p < 0.01) for pain/discomfort, whereas those younger than 35 experienced a
reduction of only 0.054 (baseline = 0.311; p < 0.01). In contrast, for anxiety/depres-
sion, the estimates for those older than 55 were not signicant, while the estimates
were 0.163 (p < 0.01) for those younger than 35, and 0.104 (p < 0.01) for those
between 35 and 55 (Wang and others 2009).

We found no statistically dierent
impact on the health dimensions by gender, conditional on the same age group.
Conclusions
Te RMHC improved the populations access to basic health care and their health
status, while reducing the risk of catastrophic health expenditure and impoverish-
ment. Because the interventions were not phased in, we cannot isolate the inde-
pendent eect of each on the observed impacts. But because the design of the
RMHC was based on generalizable theories and concepts, we can draw some logi-
cal conclusions.
By covering primary care in addition to hospitalization, the RMHC reduced
nancial barriers to accessing primary health care, which in turn improved health
status. Our ndings that the oldest age group experienced signicant improve-
ments in mobility and usual activities and reductions in pain/discomfort provide
suggestive evidence for the access hypothesis. Many older people in our sample (as
in many rural areas of China) suer from chronic pain. Making access aordable
allows them to seek care to relieve their pain and discomfort and improve their
mobility and daily activities.
We did not nd any statistically signicant impact of RMHC on hospital-
ization. Tis could be because RMHC is ineective in improving access to hos-
pitalization. But it could also be due to a substitution of outpatient use for less
serious casessince the RMHC now provided coverage for outpatient careor to
an improvement in health status as a result of improved access to basic health care.
Unfortunately, we were not able to test these hypotheses.
Covering both primary care and hospitalization contributed to the reduction in
medical impoverishment and catastrophic expenditure as well. In fact, in a separate
analysis, we compared the impact of RMHC on nancial risk protection with an
alternative scheme commonly found among NCMS models in western China and
that covers only hospitalization but with higher caps in reimbursement. To our
surprise, we found the RMHC to be more eective in reducing medical impover-
ishment than this catastrophic insurance scheme. Te primary reason is that the
NCMS does not address a major cause of medical impoverishment: expensive out-
patient services for chronic conditions. By covering only hospitalizations, it does
not protect chronic patients with major outpatient health expenditures (Yip and
Hsiao 2009).
Another possible explanation for the achievements in access, health status, and
nancial risk protection was the reduction in ineciencies and inappropriate treat-
ment brought about by changes in the organization and incentive structure of the
delivery system under the RMHC. For example, at the primary care level, spending
per visit to the village doctor in RMHC sites dropped from 16 RMB in the base-
line to about 10 RMB in 2005, whereas spending per visit in the control site grew
to about 18 RMB over the same period. By analyzing prescription records collected
by the RMHC Fund Oce, we ascertained that the reduction in expenditure per
visit was largely attributed to a reduction in drug prices of almost 30%and to
reductions in the number of drugs prescribed, the number of prescriptions for
antibiotics and steroids, and the number of intravenous injections for treatment of
common cold. After implementation of the drug bulk purchasing policy and audits
by the Fund Oce, the use of fake and expired drugs was eliminated, whereas the
use of counterfeit drugs in the control sites remained at about 30%.
A nal important aspect of the RMHC is that it placed some power in the
hands of the community, who used it to ensure that the program was organized
and managed to yield the most benets for their community. While we did not
evaluate this aspect, we conducted focus groups and interviews with villagers at
large and villagers elected to serve on the Board and village management com-
mittee. We found that the villagers were willing to monitor the operations of the
RMHC and were eective in doing so. Teir involvement resulted in a better use
of funds, reductions in waste and eciencies, and a benet package that reected
villagers preferences and for which they were willing to prepay.
At the conclusion of the experiment, the government of Guiyang, where one
of the intervention sites was located, immediately replicated our scheme to cover
around 1.7 million rural Chinese. In 2008 Shaanxi province followed suit and
began to replicate the RMHC throughout the province, which has more than
152 Chapter 8
30 million rural inhabitants. When the ndings of this social experiment were
presented to Chinas top ocials, the government revised its policies for the
NCMS. It decided that NCMS benet packages should cover both primary
care and hospitalizations, that bulk purchasing and central distribution of drugs
should be established, and that community governance should be greatly encour-
aged (Ministry of Health and Ministry of Finance 2007, 2008; Zhu 2008). Many
provincial health leaders have come to recognize that unless ineciencies are
reduced, the NCMS will not be sustainable in the long run. Tey are beginning
to experiment with ways to reform payment incentives and the organization of the
delivery system.
Tere are several plausible explanations for why our ndings inuenced both
central and local government policies. First, our ndings provide evidence for the
decisionmakers. Second, we engaged top health and political leaders as partners
throughout the social experiments by regularly brieng them on the ndings and
by incorporating Chinese policy needs and current situations into the design. Tese
leaders felt a sense of ownership of the experiments and were more motivated to
incorporate the lessons into their policies and scale up the experiments to a larger
population. Te fact that the RMHC consistently received high public satisfac-
tion ratings85% of the enrolled were satised with the scheme, with almost
90% wishing to continue their enrollmentalso appealed to the policymakers
because they knew that there would not be major resistance when they scaled up
the RMHC.
Several limitations to our study should be taken into account for future studies
and experiments. First, caution should be exercised in extrapolating our ndings to
settings that have dierent socioeconomic conditions than our intervention sites.
Te fact that Guizhou and Shaanxi agreed to be our experimental sites may mean
that their local leaders are more progressive and our ndings may not generalize to
localities lacking such leaders. Second, our health outcome measurements are self-
perceived; we do not have objective health measures. Tis limitation is mitigated
somewhat by other studies that have already demonstrated the strong correlation
between self-perceived health status and other more objective measurements such
as mortality and disability. Tird, our study design prevents us from separating
the independent eects of the various aspects of interventions. Another possible
bias stems from attrition. But this is of little concern since the attrition rate for
the treatment group is less than 9%, and although the attrition rate for the control
group is higher (about 20%), we did not nd any statistically signicant dierences
in the baseline characteristics between the followed-up and the lost samples, except
for marital status and whether the person has a chronic condition. To the extent
that these characteristics do not change much over the three-year period of study,
they are controlled for by a dierence-in-dierence model.
Although we should be conservative about the conclusions, this chapter dem-
onstrates that a well designed social experiment can inform policy actions and gen-
erate knowledge about health system strengthening. While experimentation and
prospective evaluation have been used and advocated for evaluating the eective-
ness of development programs, especially for education, their use for health systems
is still limited and in its infancy (Duo and Kremer 2005). No single health care
system model would suit the needs of the world. As countries continue to search for
the models that best match their needs and conditions, well designed social experi-
ments with objective evaluations oer a promising way forward.
References
Anonymous. 2009. Te standing conference of State Council of China adopted Guidelines for
Furthering the Reform of Health-care System in principle. Ministry of Health of China.
http://www.moh.gov.cn/publicles/business/htmlles/mohbgt/s3582/200901/38889.htm.
(in Chinese) (accessed Mar 13, 2009).
Blumenthal, D., and W. Hsiao. 2005. Privatization and its discontentsthe evolving Chinese
health care system. New England Journal of Medicine 353 (11): 116570.
Center for Health Statistics and Information. 2004. An analysis report of the National Health Survey
in 2003. Beijing, China: Ministry of Health.
Central Committee of CPC (the Communist Part of China). 2002. Decisions of the central com-
mittee of the communist party of China and the State Council on further strengthening rural health
work. Beijing, China.
Duo, E., and M. Kremer. 2005. Use of Randomization in the Evaluation of Development
Eectiveness. In G. Pitman, O. Feinstein, and G. Ingram, eds., Evaluating Development Eec-
tiveness. New Brunswick, NJ: Transaction Publishers.
Hesketh, T., and W. Zhu. 1997a. Health in China: From Mao to market reform. BMJ 314
(7093): 1543.
. 1997b. Health in China: Te healthcare market. BMJ 314 (7094): 1616.
Hsiao, W. C. 1984. Transformation of health care in China. New England Journal of Medicine
310 (14): 93236.
Lindelow, M., and A. Wagsta. 2005. Chinas health sectorwhy reform is needed. Washington,
DC: World Bank.
Liu, X., and A. Mills. 1999. Evaluating payment mechanisms: how can we measure unnecessary
care? Health Policy and Planning 14 (4): 40913.
Mao, Z. 2005. Pilot program of NCMS in China: System design and progress. Washington, DC:
World Bank.
154 Chapter 8
Ministry of Health. 2007. Chinas New Cooperative Medical Scheme: Report on Development and
Results. In Chinese. Beijing, China: Peking Union Medical College Press.
Ministry of Health and Ministry of Finance. 2007. Directions for Improving Risk Pooling and
Reimbursement Policies for New Cooperative Medical Scheme. In Chinese. Document no.
253. Beijing, China.
. 2008. Announcement on New Cooperative Medical Scheme. In Chinese. Document
no. 17. Beijing, China.
Smith, P., C. Wong, and Y. Zhao. 2005. Public expenditure and the role of government in the Chinese
health sector. Washington, DC: World Bank.
Wang, H., W. Yip, L. Zhang, and W. Hsiao. 2009. Te impact of rural mutual health care
on health status: evaluation of a social experiment in rural China. Health Economics 18 (2):
6582.
Watts, J. 2006. Chinas rural health reforms tackle entrenched inequalities. Lancet 367 (9522):
156465.
. 2007. Protests in China over suspicions of a pay-or-die policy. Lancet 369 (9556):
934.
Yip, W., and W. C. Hsiao. 2009. Non-evidence-based policy: how eective is Chinas new coop-
erative medical scheme in reducing medical impoverishment? Social Science & Medicine 68
(2): 2019.
Zhu, C. 2008. Directions for Health Reform and Its Companion Policies. Beijing, China: News
Oce of the Ministry of Health.
155
C
h
a
p
t
e
r

9
Colombias Big Bang Health
Insurance Reform
Ursula Giedion, Carmen Elisa Flrez, Beatriz Yadira Daz,
Eduardo Alfonso, Renata Pardo, and Manuela Villar
Colombia is one of the few developing countries that have introduced
government- subsidized universal health insurance by drastically chang-
ing social security schemes and breaking the public sector monopoly.
Te reforms began in 1993, when approximately 28% of the popula-
tion was covered by insurance through the traditional Latin American
approach of a social security system delivering services directly to the
covered population. By 2005 health insurance (with choice of provider)
reached more than 70% of the total population and close to 60% of the
lowest two income quintiles.
Recent estimates suggest that insurance coverage reached 86%
of the population by the end of 2006, with another 2% covered by
military and other programs. Te population is covered through two
regimes: the contributory regime for the employed and self-employed
(covering 40% of the population in 2006) and the subsidized regime for
the poor (covering 46% of the population in 2006; Clavijo 2009). By
2009 coverage had expanded to 89% (Tsai 2010).
For this chapter the gradual implementation of the subsidized health
insurance regime for the poor provides a unique opportunity to apply
semiparametric methods (propensity score matching, double dierence,
and matched double dierence) to identify dierences in health-related
outcomes between those with insurance and those without (see box 9.1
for details on the data and methodology of this study). Te impact
of the contributory regime on similar variables is analyzed using an
156 Chapter 9
instrumental variable approach. Tose without insurance nevertheless remain eli-
gible for services provided directly by the government through public facilities, so
those without insurance in this analysis retain access to traditional public providers.
BOX 9.1
Data and methodology
No single household survey in Colombia
synthesizes data on access, use, health
status, and nancial protection for the
population. Whereas recent Colombian
Demographic and Health Surveys (DHS)
(1995, 2000, and 2005) oer household
data on access, use, and health status for
small children and women of child-bearing
age, the Living Standards Measurement
Survey 2003 (LSMS) provides information
on general use of health care and out-of-
pocket spending as well as a wealth of
socioeconomic data (including information
on employment).
We used DHS data to evaluate the
subsidized regime on access, use, and
health status. In contrast to the contribu-
tory regime, the gradual implementation
and still incomplete coverage of the sub-
sidized regime among the poor allowed
us to apply semiparametric methods to
identify dierences in health outcomes
between those with insurance and those
without.
We used LSMS data to evaluate the im-
pact of the contributory regime on health-
related outcome variables. LSMS data do
not, however, include any health status
variables that can be expected to change as
a consequence of benets provided under
the contributory regime. Te analysis of the
contributory regime therefore did not try
to look at the impact of health insurance
on health status. Further, we had to resort
to either propensity score matching or in-
strumental variables to evaluate the impact
of the contributory regime, as we had only
one cross-sectional data set (LSMS 2003).
Given that almost all of those working as
formal workers participate, matching ali-
ates (through propensity score matching) to
similar nonaliates was impossible, so we
used instrumental variables. Te table below
summarizes the data used in the analysis.
Summary of data sources for evaluation
Subsidized regime Contributory regime
Access,
utilization,
and health
status
Financial
protection
Access
and
utilization
Financial
protection
LSMS 2003 (cross-sectional data)
DHS 1995, 2000, and 2005 (cross-sectional and
repeated cross-sectional data)
Administrative data at the municipal and state level
Census data at the block level
Source: Authors.
Colombias Big Bang Health Insurance Reform 157
Te main goal of this chapter is to analyze existing household data to provide
information on the impact of the Colombian health insurance scheme on key per-
formance indicators of the health system. Specically, it seeks answers to the fol-
lowing questions: Has insurance improved access to and use of health services for
individuals in case of an adverse health event? Has insurance reduced the risk of
having to confront an out-of-pocket health payment that destabilizes the nancial
welfare of the household or causes the household to fall below a poverty line? Has
insurance improved health outcomes?
Main features of the Colombian health sector reform
Colombia is a lower middle-income Andean country of 45 million inhabitants,
more than 70% in urban areas. Te country is divided into 32 departments, 1,099
municipalities, and 3 special districtsBogot, Cartagena, and Santa Marta
(Ministry of Planning 2008). Per capita gross national income was US$8,430 in
2008, in purchasing power parity terms. Adult literacy was 93% in 2008, and basic
service coverage is good, with 99% of urban households having access to improved
water sources (77% rural) and 78% of all households having access to improved
sanitation facilities. About 45% of the population was below the national poverty
line, and about 28% lived on US$2 or less per day in 2006. Te infant mortality
rate is 16 per 1,000 live births, and average life expectancy is above 70 years. About
6.1% of GDP is spent on health, with 84% of that amount nanced through taxes.
Of the 16% coming from private funds, only about half is out of pocket. About
US$284 per capita is spent on health (World Bank 2010).
Before the health reform in the early 1990s, Colombia had a health system
similar to others in Latin America. A vertically integrated social insurance system,
based on payroll taxes for formal workers, basically covered only the employed
worker.
1
A tax-nanced system of public providers served the poor and the not so
poor, the latter especially for hospital and surgical services. A private provider sys-
tem operated for all those with the ability to pay or for those dissatised with the
services provided in the traditional social insurance and the public system.
2
With Law 100 in 1993 Colombia introduced universal health insurance, a pol-
icy implemented in only a few Latin American countries. For those with sucient
income (above one minimum wage, about US$170), a payroll tax of 12.5% is col-
lected and a comprehensive insurance plan (plan obligatorio de salud, or mandatory
health plan) is provided within the rgimen contributivo (contributory regime). For
the poor whose eligibility to subsidies is determined by a proxy means test called
sistema de identicacin de beneciarios, or beneciary identication system , the
government purchases, with a mixture of tax revenue and a solidarity contribution
158 Chapter 9
from payroll taxes,
3
insurance coverage in the rgimen subsidiado (subsidized
regime).
In both cases the aliated family or individual chooses a health insurance
companyentidad promotora de salud, or EPS (health-promoting entity)whose
ownership may be public, private, or mixed, and which may be for prot or non-
prot. Te insurance company, in turn, contracts health services with a network of
public, private, or own service providers.
Te government establishes the benets package and sets the premium to be paid
to each insurance company for each individual, with a risk adjustment by age, sex,
and location. Te premium is about US$252 per person annually in the contribu-
tory regime and US$146 per person in the subsidized regime.
4
Tose insured through
the contributory regime have access to a benets package more comprehensive than
that provided by the subsidized regime (gure 9.1, which shows clearly the cut-out
method of managing the cost of the benets package). Services not included in the
subsidized regime package are the responsibility of the public hospital network, as
are all services for the uninsured. Insurable benets for the subsidized regime were
to expand progressively to converge with those covered by the contributory regime.
According to Law 100/1993, the two plans were supposed to converge by 2000, a
promise still unfullled. However, Constitutional Court ruling T-760 on 31 July
2008, mandated the immediate unication of benets plans for children and concrete
steps to move toward equal benets for all (Tsai 2010; Yamin and Parra-Vera 2009).
Contributory regime, 2010
C
o
m
p
l
e
x
i
t
y

o
f

c
a
r
e
C
o
m
p
l
e
x
i
t
y

o
f

c
a
r
e
Subsidized regime, 2010
Care for catastrophic
illnesses
Care for catastrophic
illnesses
Primary care Primary care
Hospital care Hospital care US$146 US$252
FIGURE 9.1
Health insurance in Colombia, premiums and benets plans, 2010
Source: Ministry of Social Protection.
In each regime benets are unrelated to the nancial contribution of the ali-
ate. Te nancial contribution is established as a percentage of income and is thus
independent of the risk of the insured. Te government has established a mecha-
nism for channeling resources from individuals whose payroll contributions are
greater than the premiums for themselves and their families toward individuals
whose contributions are less. Te solidarity fund ( fondo de solidaridad y garantia)
receives the excess contributions and reassigns funds toward those whose contri-
butions fall short of the capitation rate of the subsidized regime. Tis equalization
fund, through a complex process, makes sure that payroll contributions based on
income are transformed into risk-adjusted premiums for all insured, both for con-
tributor and dependents in the contributory regime and on a solidarity contribu-
tion for the poor. In this integrated risk-pooling scheme individuals at high risk
of disease subsidize those at low risk, those with higher ability to pay subsidize
those without, and those in productive ages subsidize the young and the elderly.
Figure 9.2 presents a simplied version of the ows of funds and aliations within
the current Colombian health system.
Conceptual framework
Health insurance in Colombia serves the dual purpose of promoting health by
making routine health care services more accessible and protecting individuals and
families against large nancial losses in case of an adverse health event.
Government funds
National
Insurance Fund
Population with
ability to pay
Poor population
Identied
by proxy
means test
$
Pays on behalf
of the poor
$
Payroll tax and
solidarity contribution
Chooses
health insurer
Insurer provides
pre-established
benets package
Chooses providers
within insurers network
$
Contracts
health services
Health insurers
(public and private)
Providers
(public and private)
$ Risk-based
premiums
FIGURE 9.2
Colombias health system, nancial ows, and afliation
Source: Authors.
160 Chapter 9
We hypothesize that when services in Colombia become more aordable
through health insurance, patients will use them more often, will seek care with
less delay, and may be more likely to have a regular source of care. Tis reason-
ing follows standard economic theory, which says that health insurance coverage
induces greater medical care use by reducing the cost of care to patients (Phelps
2009), as well as evidence in developed countries (Institute of Medicine 2009).
We speculate that, all other things being equal, those insured by the subsidized
regime and contributory regime experience fewer nancial barriers to access and
use more health care than the uninsured do. We further speculate that health sta-
tus improves as a result of increased access to health care and that health insurance
provides nancial protection to individuals by reducing catastrophic out-of-pocket
health spending.
Several comments are relevant to this framework.
Health insurance is not a homogeneous good. It varies in both extent (benets,
level of copayment, and conditions of access) and duration. In this study the
benets package provided under the contributory regime is much larger than
that oered by the subsidized regime, so we decided to evaluate the impact
of each regime separately. Further, health insurance does not improve access
to all health services. We cannot expect changes in access, use, and outcome
variables that are not related to the benets oered under the insurance plan
we are evaluating.
Te eect of health insurance may vary across population groups. Geographic
variation in the supply of care to insured and uninsured individuals is one
potentially important source of heterogeneity.
Health outcomes depend on many more variables than just health insurance
coverage, and people who have health insurance and those who do not dier in
many ways other than in their health insurance coverage.
Health outcomes partially determine health insurance coverage, and vice versa.
So health status will most probably dier systematically among individuals
grouped by health insurance categories.
Health insurance does not have a direct impact on health. Instead, it changes
individuals and households decisions related to the use of health care services
by reducing nancial barriers to access.
Health status is a complex concept, and the impact of health insurance on
health status depends on the health status variable we choose. If better access
is the means by which health status may improve due to health insurance, we
should concentrate on the measures of health status that can be reasonably well
connected to access to health services.
Health insurance coverage
Te increase of health insurance coverage among Colombians is the one successful
outcome on which most observersfoes or friends of the reformwould prob-
ably agree. Household survey data from 2003 presented by Escobar (2005) indi-
cate that overall health insurance (subsidized regime and contributory regime)
increased from less than a quarter of the population before the reform (1993)
to almost two-thirds a decade later (table 9.1). Te most recent administrative
data from the Ministry of Social Protection indicate that health insurance reached
more than 90% of the population by 2007. Growth has been especially fast
among the poorest 20% of the population, with an almost eightfold increase in
one decade.
5
No major dierences are detected in coverage of the target population (in the
lower quintiles) by gender. Small children (ages 05) in the poorest SISBEN level
are least protected by the subsidized regime. Tis result is worrisome, both because
this group is especially vulnerable to adverse health events and because the cover-
age policy of the subsidized regime ocially gives priority to this vulnerable group.
Substantial dierences can be observed at the municipal level: in 2004 more
than 40% of all municipalities had coverage of more than 80%, but close to 20%
had coverage still below the 20% level.
6
Tis is explained mainly by the inequity
in the public per capita health resources available locally to nance the subsidized
regime.
Tough less dramatic, growth in the coverage of the contributory regime has
also been important. In 1993 the contributory mandatory social insurance scheme
covered around 9.4 million people (about 26% of the population). A decade later,
the number of aliates had grown to 17 million (39% of the population), increas-
ing coverage by close to 80% in one decade. Before the reform, mostly formal
TABLE 9.1
Health insurance coverage, 19932003 (%)
Quintile 1993 1997 2003
1 (poorest) 6.1 43.4 46.5
2 16.5 48.7 52.5
3 27.5 59.0 58.2
4 35.3 65.7 69.3
5 (richest) 43.1 76.7 82.7
Total 23.8 57.1 61.8
Source: Escobar 2005, based on Casen survey (1993) and LSMS (1997, 2003) household
surveys.
162 Chapter 9
sector workers of the private and public sector had access to a full benet plan, and
family coverage was limited. Under the Social Security Institute (Instituto de Segu-
ros Sociales, or ISS), which covered formal private sector workers, only pregnant
and dependent wives and their small children (younger than age 1) had access to a
few birth-related services, leaving the remaining services and all other dependent
family members uncovered.
7,8
With reform the number of beneciaries (dependent, nonpaying family mem-
bers) more than doubled between 1993 and 2003 (+129%), and the number of
contributing aliates increased 36%, rising from 4.9 million contributing aliates
to 6.8 million a decade later. Independent workers, fewer than 10% of aliates
before the reform, represented more than 18% a decade later, and their number had
increased more than threefold. Family coverage was nationwide and unrelated to
type of employment. Under the reform, eligibility started to be based on income:
all workers declaring a monthly income equal to or above one ocial minimum
salary as a basis for payroll taxes were required to aliate with the contributory
regime. From then on, the whole family was to be insured (spouses, partners,
dependents, and parents in some cases).
Because of these modications, most contributory regime aliates were in
urban areas (more than 90%) in 2005, compared with 78% of the general popula-
tion (CEPAL 2006). Coverage of aliates older than age 57 increased more than
did any other group in the last decade. Tese adults, either retired or close to retire-
ment, represented 9% of aliates in 1993 and about 13% a decade later, larger
than their share in the population.
Access and use of services
We hypothesize that both the subsidized and the contributory health insurance
regimes introduced in Colombia in 1993 have improved the health status of the
insured by making access more aordable. As a result, we are speculating that,
all other things being equal, aliates of the subsidized regime and contributory
regime experience fewer nancial barriers to access, use more health care, and get
health care earlier than the uninsured.
Subsidized regime
Propensity score matching estimates conrm that the subsidized health insurance
scheme increases access for the poor (table 9.2).
9
Tose aliated with the subsi-
dized regime are 40% more likely to have used outpatient health services in the
month prior to the survey than the uninsured and less than half as likely to have
experienced barriers to access when needing care (38%).
Barriers to access may lie on the demand side with the household (such as
income, knowledge, and the like) or on the supply side, as a distant health facility,
bad service, delayed appointments, excessive procedures and formalities to get an
appointment, asking for service but not getting it, or consulting without results.
Insurance in Colombia reduces demand side nancial barriers and creates an ali-
ation with service providers, so it should reduce demand side barriers consider-
ably, and this is borne out by the data. For the insured, barriers to access are more
related to supply of health services (+120%) when compared with the uninsured.
Tere is a clear benet for insured pregnant women in accessing prenatal and
post-partum care: insured women have more prenatal visits (+6%), are more likely
to give birth in a health facility (+7%), and have a higher probability of being
assisted either by a doctor (+8%) or by skilled personnel during childbirth (+7%).
TABLE 9.2
Propensity score matching results-estimated treatment effect for the
subsidized regime on access and use, national level, radius (bandwidth
0.0001), 2005
Outcome variable
Enrolled
in the
subsidized
regime
(treatment
group)
Not insured
(control
group)
Difference
(%) Signifcance
Used ambulatory services in the 12 months prior
to the survey 0.686 0.492 40 ***
Birth attended by a doctor 0.819 0.757 8 ***
Birth attended by a skilled professional 0.862 0.808 7 ***
Birth in a health facility 0.869 0.812 7 ***
Immunization child complete 0.395 0.365 8 *
Not receiving medical care when needed 0.259 0.420 38 ***
Not receiving medical care when needed due to
supply reasons 0.289 0.131 120 ***
Number of prenatal visits 5.560 5.261 6 ***
Child taken to a health care facility when coughing 0.451 0.378 20 ***
Child taken to a health care facility when having
diarrhea 0.329 0.235 40 *
*** signifcant at p < 0.01; * signifcant at p < 0.10.
Note: Other matching methods (double difference and matched double difference) were imple-
mented with similar results. Results from these methods can be obtained from the authors
on request.
Source: Authors calculations based on DHS 2005 data.
164 Chapter 9
Te benets extend to their children as well. Parents of insured children in the sub-
sidized regime have a greater probability of taking children to a health care facil-
ity when they are coughing (+20%) or suering from diarrhea (+40%). Aliated
children are more likely to have their immunization series completed for their age
(+8%), even though immunizations are widely available, free for all, and heavily
promoted.
Urban-rural dierences
Insurance seems to have been important in bringing about improvements for the
rural population often greater than for the urban population (table 9.3). In many
cases the dierence between the treatment and the control groups in rural areas is
twice that in urban areas. For a child receiving health services when he or she has a
cough, there is a 15% dierence between the insured and uninsured in urban areas,
compared with 33% in rural areas.
Contributory regime
Self-employed and their families
Health insurance under the contributory regime improves access and use indica-
tors of independent workers for most variables, and the results are statistically
signicant for almost all coecients (table 9.4).
10
Looking at the self-employed
11

separately from the rest of the insured in the contributory regime (the left side
of table 9.4), on average 20% of them face nancial barriers to access when in
need of health services. According to our instrumental variable estimates, the
contributory regime reduces the incidence of nancial barriers by 47 percentage
points for the self-employed, a large improvement in access. In contrast to the
subsidized regime, there is no statistically signicant impact of the contributory
regime on the probability of suering from supply side access barriers for this
group.
In the descriptive statistics we also nd that only 26% of all self-employed
whether aliated with the insurance program or notreceive all medicines pre-
scribed. Health insurance provided by the contributory regime increases by 52 per-
centage points the probability of receiving all prescribed medicines (see table 9.4).
Even so, there is still much to do, as only 53% of the insured in the contributory
regime reported receiving all prescribed medicines at the time of the survey.
Sample means indicate that, on average, the self-employed without insurance
use at least some (39%) dentist or physician preventive care during the year, or
both (16%). In contrast, 42% of those aliated with the contributory regime went
T
A
B
L
E

9
.
3

P
r
o
p
e
n
s
i
t
y

s
c
o
r
e

m
a
t
c
h
i
n
g

r
e
s
u
l
t
s
-
e
s
t
i
m
a
t
e
d

t
r
e
a
t
m
e
n
t

e
f
f
e
c
t

f
o
r

t
h
e

s
u
b
s
i
d
i
z
e
d

r
e
g
i
m
e

f
o
r

a
c
c
e
s
s
,

u
s
e
,

a
n
d

h
e
a
l
t
h

s
t
a
t
u
s
,

b
y

r
u
r
a
l
-
u
r
b
a
n

s
t
a
t
u
s
,

k
e
r
n
e
l

E
p
a
n
e
c
h
n
i
k
o
v
,

2
0
0
5

(
b
a
n
d
w
i
d
t
h

0
.
0
0
1
)
U
r
b
a
n
R
e
l
a
-
t
i
v
e

i
m
p
a
c
t
R
u
r
a
l
E
n
r
o
l
l
e
d

i
n

t
h
e

s
u
b
s
i
d
i
z
e
d

r
e
g
i
m
e

(
t
r
e
a
t
m
e
n
t

g
r
o
u
p
)
N
o
t

i
n
s
u
r
e
d

(
c
o
n
t
r
o
l

g
r
o
u
p
)
D
i
f
f
e
r
e
n
c
e

(
p
e
r
c
e
n
t
)
S
i
g
n
i
f
-
c
a
n
c
e
E
n
r
o
l
l
e
d

i
n

t
h
e

s
u
b
s
i
d
i
z
e
d

r
e
g
i
m
e

(
t
r
e
a
t
m
e
n
t

g
r
o
u
p
)
N
o
t

i
n
s
u
r
e
d

(
c
o
n
t
r
o
l

g
r
o
u
p
)
D
i
f
f
e
r
e
n
c
e

(
p
e
r
c
e
n
t
)
S
i
g
n
i
f
-
c
a
n
c
e
U
s
e
d

a
m
b
u
l
a
t
o
r
y

s
e
r
v
i
c
e
s

i
n

t
h
e

1
2

m
o
n
t
h
s

p
r
i
o
r

t
o

t
h
e

s
u
r
v
e
y
0
.
7
0
2
0
.
5
3
0
3
3
*
*
*
<
0
.
6
5
7
0
.
4
4
0
4
9
*
*
*
B
i
r
t
h

a
t
t
e
n
d
e
d

b
y

a

d
o
c
t
o
r
0
.
9
0
1
0
.
8
7
5
3
*
*
*
<
0
.
6
7
8
0
.
6
3
7
6
*
*
B
i
r
t
h

a
t
t
e
n
d
e
d

b
y

a

s
k
i
l
l
e
d

p
r
o
f
e
s
s
i
o
n
a
l
0
.
9
3
3
0
.
9
1
4
2
*
*
<
0
.
7
4
0
0
.
6
9
8
6
*
*
B
i
r
t
h

i
n

a

h
e
a
l
t
h

f
a
c
i
l
i
t
y
0
.
9
4
2
0
.
9
2
7
2
*
<
0
.
7
4
7
0
.
6
8
4
9
*
*
*
I
m
m
u
n
i
z
a
t
i
o
n

c
h
i
l
d

c
o
m
p
l
e
t
e
0
.
4
5
0
0
.
4
2
1
7
*
0
.
3
2
9
0
.
3
1
1
6
N
o
t

r
e
c
e
i
v
i
n
g

m
e
d
i
c
a
l

c
a
r
e

w
h
e
n

n
e
e
d
e
d
0
.
2
0
8
0
.
3
5
2
4
1
*
*
*
>
0
.
3
2
4
0
.
4
9
8
3
5
*
*
*
N
o
t

r
e
c
e
i
v
i
n
g

m
e
d
i
c
a
l

c
a
r
e

w
h
e
n

n
e
e
d
e
d

d
u
e

t
o

s
u
p
p
l
y

r
e
a
s
o
n
s
0
.
2
6
7
0
.
1
3
3
1
0
1
*
*
*
0
.
3
4
8
0
.
1
7
2
1
0
2
*
*
*
N
u
m
b
e
r

o
f

p
r
e
n
a
t
a
l

v
i
s
i
t
s
5
.
7
6
7
5
.
8
0
5
1
*
*
*
<
5
.
1
6
7
4
.
5
4
9
1
4
*
*
*
C
h
i
l
d

t
a
k
e
n

t
o

a

h
e
a
l
t
h

c
a
r
e

f
a
c
i
l
i
t
y

w
h
e
n

c
o
u
g
h
i
n
g
0
.
4
7
7
0
.
4
1
4
1
5
*
*
*
<
0
.
4
0
4
0
.
3
0
3
3
3
*
*
*
C
h
i
l
d

t
a
k
e
n

t
o

a

h
e
a
l
t
h

c
a
r
e

f
a
c
i
l
i
t
y

w
h
e
n

h
a
v
i
n
g

d
i
a
r
r
h
e
a
0
.
3
6
2
0
.
3
2
6
1
1
0
.
3
2
4
0
.
2
3
5
3
8
*
*
*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
0
1
;

*
*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
0
5
;

*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
1
0
.
N
o
t
e
:

O
t
h
e
r

m
a
t
c
h
i
n
g

m
e
t
h
o
d
s

(
d
o
u
b
l
e

d
i
f
f
e
r
e
n
c
e

a
n
d

m
a
t
c
h
e
d

d
o
u
b
l
e

d
i
f
f
e
r
e
n
c
e
)

w
e
r
e

i
m
p
l
e
m
e
n
t
e
d

w
i
t
h

s
i
m
i
l
a
r

r
e
s
u
l
t
s
.

R
e
s
u
l
t
s

f
r
o
m

t
h
e
s
e

m
e
t
h
-
o
d
s

c
a
n

b
e

o
b
t
a
i
n
e
d

f
r
o
m

t
h
e

a
u
t
h
o
r
s

o
n

r
e
q
u
e
s
t
.
S
o
u
r
c
e
:

A
u
t
h
o
r
s

c
a
l
c
u
l
a
t
i
o
n
s

b
a
s
e
d

o
n

D
H
S

2
0
0
5

d
a
t
a
.
166 Chapter 9
to a general physician or dentist at least once during the year prior to the survey.
Isolating the eect of insurance from other variables using either a dentist or a phy-
sician at least once a year for checkups or preventive care improves by 46 percent-
age points for the self-employed, and the use of both improves by 15 percentage
points (see table 9.4). Furthermore, health insurance under the contributory regime
increases the probability of using formal care (+26 percentage points) rather than
informal care
12
(5 percentage points) and reduces self-medication when having a
health problem (15 percentage points).
TABLE 9.4
Instrumental variable results-estimated treatment effect for access and
use in the contributory regime, 2003
Outcomes
Self-employed, or
independent, workers
and their families
Relative
impact
Formally employed, or
dependent, workers
and their families
Difference p Difference p
Preventive health care use (physician or
dentist visit at least once per year) 0.459 *** > 0.342 ***
Preventive health care use (physician
and dentist at least once a year) 0.152 [0.0138]*** < 0.272 **
Formal health care services use 0.256 [0.0328]*** < 0.567 ***
Informal health care services use 0.052 [0.0136]*** > 0.027 *
Self-medication when having a health
problem 0.148 [0.0266]*** < 0.276 [0.0392]***
No health care use when having a health
problem 0.021 [0.0064]** 0.020 [0.0083]*
Supply side barrier to access 0.090 0.045 *
Demand side barrier to access 0.353 [0.1766]* > 0.210 [0.0506]***
Financial barrier of access 0.473 [0.1524]** > 0.144 [0.0337]***
Access to medications (patients given at
least some of the prescribed medicines) 0.755 [0.0325]*** 0.760 [0.0248]***
Access to medications (patient was
given all of the prescribed medicines) 0.516 [0.0390]*** < 0.568 [0.0268]***
Timeliness of service for general
physician and dentist 1.525 [1.0454] 0.910 [0.4151]*
Timeliness of service for visit to
specialist 1.948 [1.5022] 0.322 [0.6083]
Note: Relative impact indicates for which population group (independent or dependent) the
impact has been more important in case both variables are found to be statistically signif-
cant. Numbers in brackets are standard errors.
Source: Authors.
Te formally employed and their families
Results for the formally employed and their families are similar to those for the
self-employed. Te contributory regime has improved access to and use of pre-
ventive, curative, and formal health care services and reduced access barriers and
self-medication. It reduces the incidence of supply side barriers by 5 percentage
points and the number of days waiting for an appointment to the general physician
or dentist by about one day (0.91), considerable given that waiting times for the
uninsured are on average six days. Note that for supply side barriers and waiting
times the results are statistically signicant only for dependent workers.
Comparing self-employed and formally employed workers
Using the instrumental variable approach we nd that for access barriers, the impact
of the contributory regime seems to be more important for the self-employed than
the formally employed (see table 9.4), especially for the probability of suering
from nancial barriers to access. For example, the probability of encountering a
nancial barrier drops 14 percentage points among the formally employed due to
insurance coverage, whereas it drops almost 50 percentage points among the self-
employed insured. Te self-employed face, on average, worse basic access condi-
tions than the rest of the sample, so insurance might have a more important mar-
ginal impact for them because diminishing marginal returns set in for these basic
access variables for the formally employed.
Te formally employed and their families beneted more than the self-
employed for several other variablesincluding the probability of using formal
health care services (general physician, specialist and dentist visit, services provided
in a health facility by a nurse or medical caretaker), the probability of using preven-
tive dental and general physician visits at least once a year, and reduction in the
use of self-medication to confront medical problems. On the use of preventive care
we might hypothesize that the opportunity cost of time is lower for the formally
employed (they can take o from work for a preventive care visit without losing
income) than for the self-employed (they have to pay themselves for the time lost
for a preventive visit).
Financial protection
According to our calculations, 14% of Colombian households that use health ser-
vices devote on average more than 30% of their monthly nonsubsistence income
to health-related out-of-pocket spending. Tis percentage drops to 5% for the total
population.
13
It is dicult to tell whether this incidence is high or low because
no universal benchmark exists and because comparisons across countries can be
168 Chapter 9
misleading and are therefore not advisable due to the dierences in contexts and
methods. As expected, vulnerability to catastrophic spending increases for the
poorest individuals. Of the richest 20% who use health services, 6.6% have out-of-
pocket spending greater than 30% of their monthly nonsubsistence income, com-
pared with 16.9% for those in the poorest quintile and 31.7% in the next poorest
quintile (table 9.5).
If we raise the threshold from 10% to 40% of nonsubsistence income, the inci-
dence of catastrophic expenditure drops from 32% to 11%, a consequence of the
well known skewed distribution of health spending (gure 9.3) Most households
using health services spend less than 20,000 Colombian pesos (US$10 in 2003)
when using health services, or between 0% and 15% of their available income. Te
frequency of spending more drops rapidly thereafter, and only a fraction of house-
holds spend more than 200,000 pesos (US$100). Tis has important consequences
for the sample size available to estimate the impact of health insurance on nancial
protection: the higher the threshold beyond which an out-of-pocket expenditure
is considered catastrophic, the smaller the sample size and therefore the less the
information available to generate signicant results.
Only 4% of households that used health services in 2003 fell below the endog-
enous poverty line when incurring out-of-pocket expenditures. Comparing results
by insurance status, we nd that, with a threshold of 30%, the incidence of cata-
strophic expenditure is highest among those lacking insurance (34%) followed by
TABLE 9.5
Incidence of catastrophic and impoverishing expenditure by income
quintile and insurance status (simple means), population that has used
health services, 2003 (percent)
Insurance status and
income quintile
Capacity to pay (percentage of nonsubsistence
income defning a catastrophic expense)
10% 20% 30% 40%
Total 32.0 21.0 14.0 11.0
Uninsured 63.9 45.4 34.0 23.9
Subsidized regime 37.9 27.6 20.8 17.5
Contributory regime 16.9 8.5 4.4 3.4
Quintile 1 (poorest) 37.6 25.0 16.4 11.5
Quintile 2 51.2 40.5 31.7 26.2
Quintile 3 29.7 19.0 11.6 6.6
Quintile 4 20.5 10.0 6.2 5.0
Quintile 5 (richest) 20.4 9.3 6.6 5.4
Source: Flrez, Giedion, and Pardo 2010.
those aliated with the subsidized regime (20.8%), and nally those aliated with
the contributory regime (4.4%) (see table 9.5).
Tese dierences by insurance status are important, but to know whether they
are due to insurance or other characteristics that might systematically dier across
groups, we used propensity score matching in the subsidized regime and instru-
mental variables to correct for selection bias.
Subsidized regime
Te average monthly income of households in the subsidized regime is US$180,
and their capacity to pay roughly US$98 (Flrez, Giedion, and Pardo 2010). Com-
pared with uninsured households, households insured with the subsidized regime
have a smaller probability (21 percentage points) of having an out-of-pocket pay-
ment that is greater than 10% of their capacity to pay (table 9.6). By the same
token, insured households in the subsidized regime have a lower probability of
facing an out-of-pocket health expenditure greater than 20% or 30% of their non-
subsistence income (14 and 11 percentage points respectively), compared with
uninsured households. When households face out-of-pocket health expenditures
more than 40% of their ability to pay, the positive impact of the subsidized regime
is lowerbut a dierence in favor of the insured households can still be observed
(4 percentage points).
Density
Out-of-pocket spending for health
Households that experience an illness
0 250 500 750 1,000
0.000
0.005
0.010
0.015
0.020
FIGURE 9.3
The distribution of out-of-pocket spending by households using health
services, thousand pesos, 2003
Source: Flrez, Giedion, and Pardo 2010.
170 Chapter 9
Te results show that as the cost of the catastrophic event increases, the protec-
tive eect of insurance decreases, probably a reection of the level of coverage of
the benets package for the poor, because the plan covers ambulatory and cata-
strophic care (low frequency, high cost) and coverage is limited for standard hospi-
tal care except some frequent surgeries such as appendectomies and hysterectomies.
So, with a threshold of 40% any out-of-pocket spending more than US$40 would
be catastrophic. Given that coverage for hospital services is limited it is not surpris-
ing that dierences between those insured under the subsidized regime and their
comparable counterparts tend to be small.
No solid evidence emerges for the impact of the subsidized regime on impoverish-
ment due to health-related out-of-pocket spending. Results are barely or not statisti-
cally signicant for two of the four poverty lines. It is dicult to identify the mitigat-
ing impact of the subsidized regime on impoverishing health spending due to its low
incidence and the similarity of incidence between insured and uninsured households.
Contributory regime
Te mitigating eect of the contributory regime for insured households facing
catastrophic health spending follows a similar pattern to that of the subsidized
TABLE 9.6
Propensity score matching estimated effect of catastrophic and
impoverishing health spending in the subsidized regime
Propensity score matching
(probit includes proxy for household health status)
Enrolled in the
subsidized regime
(treatment group)
Not insured
(control
group) Difference Signifcance
Catastrophic spending (10% capacity to pay) 0.3942 0.6080 0.21 ***
Catastrophic spending (40% capacity to pay) 0.1594 0.1956 0.04 *
Falls below the endogenous poverty line 0.0609 0.0513 0.01
Falls below the national poverty line 0.0638 0.1003 0.04 *
Falls below the national indigence line 0.0435 0.0414 0.00
Falls below the endogenous or the national
poverty line
a
0.0986 0.1407 0.04 *
*** signifcant at p < 0.01; * signifcant at p < 0.10.
a. Endogenous (to the household) poverty line = basic household basket of goods and
services.
Source: Authors.
regime.
14
As out-of-pocket expenditures for health increase as a percentage of the
households nonsubsistence income,

the mitigating eect of insurance decreases.
15

Te contributory regime has an explicit benets plan that requires copayments at
the time of the use of services. For adverse health events treatable with technolo-
gies that are part of the insurance plan, the contributory regime has the capacity to
mitigate the nancial impact. But for technologies outside of the plan, there is no
protection for individuals.
Health insurance under the contributory regime improves nancial protec-
tion for the formally employed as well as for the self-employed and their families
(table 9.7). Te impact of health insurance on nancial protection is greater, and
results tend to be more signicant, among the self-employed than other insured
workers. For instance, for a catastrophic threshold of 30%, health insurance under
the contributory regime reduces the probability of suering from a catastrophic
out-of-pocket health expenditure by 27% among the self-employed but only 4%
(and without statistical signicance) among other workers. For impoverishing
adverse health events, results were not statistically signicant.
Health status
Subsidized regime
Evaluating the impact of health insurance on health status is extremely complex. Pos-
sibly the most challenging issue is to nd health status variables subtle enough to
capture changes underlying the quality of life that can be related to improved access to
services covered under the benets packages. In addition, without any real panel data,
we will never know whether observed dierences are the result of health insurance or
whether health insurance is partly the result of observed health status. If we observe
that those aliated have, on average, a worse perception of their health status, is that
because health insurance worsens health or because those ill are more prominent
among those seeking aliation? It comes therefore as no surprise that no conclusive
evidence emerges on the impact of health insurance on the health status variables that
are available in the DHS. For the impact of the subsidized regime on health status,
the results of the analysis based on the simple comparison of means provide mixed
evidence. No signicant dierences are observed for the survival of small children.
Aliates have a higher incidence of low birthweight (+43%) and complications after
delivery (+5%) but a less favorable perception of their health status (3%) and a lower
incidence of extremely low birthweight (77%). After controlling for other variables,
insurance does not appear to have a signicant impact in explaining dierences in
health status, except low birthwweight, which is worse for the uninsured (table 9.8).
172 Chapter 9
T
A
B
L
E

9
.
7

E
s
t
i
m
a
t
e
s

o
f

t
h
e

i
m
p
a
c
t

o
f

t
h
e

c
o
n
t
r
i
b
u
t
o
r
y

r
e
g
i
m
e

o
n

t
h
e

i
n
c
i
d
e
n
c
e

o
f

c
a
t
a
s
t
r
o
p
h
i
c

e
x
p
e
n
d
i
t
u
r
e

b
y

t
y
p
e

o
f

e
m
p
l
o
y
m
e
n
t
,

2
0
0
3
F
o
r
m
a
l
l
y

e
m
p
l
o
y
e
d

w
o
r
k
e
r
s
S
e
l
f
-
e
m
p
l
o
y
e
d

w
o
r
k
e
r
s
S
i
m
p
l
e

d
i
f
f
e
r
e
n
c
e
P
r
o
b
i
t

c
o
n
t
r
o
l
I
v
p
r
o
b
i
t
B
i
p
r
o
b
i
t
i
v
S
i
m
p
l
e

d
i
f
f
e
r
e
n
c
e
P
r
o
b
i
t

c
o
n
t
r
o
l
I
v
p
r
o
b
i
t
B
i
p
r
o
b
i
t
i
v
C
a
t
a
s
t
r
o
p
h
i
c

s
p
e
n
d
i
n
g

(
1
0
%

c
a
p
a
c
i
t
y

t
o

p
a
y
)
0
.
3
2
8
1
[
0
.
0
7
4
7
]
*
*
*
0
.
2
7
1
2
[
0
.
0
7
7
5
]
*
*
*
0
.
2
4
0
8
[
0
.
0
8
2
6
]
*
*
0
.
3
1
4
3
[
0
.
1
5
6
7
]
*
0
.
4
6
8
5
[
0
.
0
5
3
9
]
*
*
*
0
.
5
2
1
2
[
0
.
0
6
9
9
]
*
*
*
0
.
6
2
7
8
[
0
.
2
0
6
4
]
*
*
0
.
6
2
2
7
[
0
.
1
5
6
5
]
*
*
*
N
u
m
b
e
r

o
f

o
b
s
e
r
v
a
t
i
o
n
s
1
,
1
1
7
1
,
1
1
7
1
,
1
1
7
1
,
1
1
7
1
,
0
3
1
1
,
0
3
1
1
,
0
3
1
1
,
0
3
1
H
a
u
s
m
a
n
n
0
.
1
2
2
6
0
.
1
7
8
5
r
h
o
-
w
a
l
d

t
e
s
t
0
.
1
1
8
0
.
6
9
5
2
*
C
a
t
a
s
t
r
o
p
h
i
c

s
p
e
n
d
i
n
g

(
2
0
%

c
a
p
a
c
i
t
y

t
o

p
a
y
)
0
.
2
1
6
5
[
0
.
0
5
2
9
]
*
*
*
0
.
1
3
1
[
0
.
0
3
7
6
]
*
*
*
0
.
0
7
1
6
[
0
.
0
2
7
5
]
*
*
0
.
0
7
8
6
[
0
.
0
6
3
3
]
0
.
3
6
8
5
[
0
.
0
4
3
6
]
*
*
*
0
.
3
2
7
4
[
0
.
0
4
4
0
]
*
*
*
0
.
5
6
4
4
[
0
.
1
8
2
7
]
*
*
0
.
6
1
2
7
[
0
.
0
8
7
0
]
*
*
*
N
u
m
b
e
r

o
f

o
b
s
e
r
v
a
t
i
o
n
s
1
,
1
1
7
1
,
1
1
7
1
,
1
1
7
1
,
1
1
7
1
,
0
3
1
1
,
0
3
1
1
,
0
3
1
1
,
0
3
1
H
a
u
s
m
a
n
n
0
.
3
5
7
9
0
.
8
9
1
6
r
h
o
-
w
a
l
d

t
e
s
t
0
.
2
4
7
9
0
.
8
0
3
2
*
C
a
t
a
s
t
r
o
p
h
i
c

s
p
e
n
d
i
n
g

(
3
0
%

c
a
p
a
c
i
t
y

t
o

p
a
y
)
0
.
1
4
1
[
0
.
0
3
7
1
]
*
*
*
0
.
0
4
1
9
[
0
.
0
2
2
8
]
0
.
0
1
8
6
[
0
.
0
1
2
8
]
0
.
0
1
7
4
[
0
.
0
2
2
1
]
0
.
2
9
2
[
0
.
0
3
9
6
]
*
*
*
0
.
2
7
2
7
[
0
.
0
3
5
0
]
*
*
*
0
.
4
2
8
3
[
0
.
1
1
5
3
]
*
*
*
0
.
5
1
7
4
[
0
.
1
0
7
0
]
*
*
*
N
u
m
b
e
r

o
f

o
b
s
e
r
v
a
t
i
o
n
s
1
,
1
1
7
1
,
1
1
7
1
,
1
1
7
1
,
1
1
7
1
,
0
3
1
1
,
0
3
1
1
,
0
3
1
1
,
0
3
1
H
a
u
s
m
a
n
n
0
.
4
0
9
7
0
.
7
7
3
3
r
h
o
-
w
a
l
d

t
e
s
t
0
.
3
5
5
1
0
.
2
3
5
6
C
a
t
a
s
t
r
o
p
h
i
c

s
p
e
n
d
i
n
g

(
4
0
%

c
a
p
a
c
i
t
y

t
o

p
a
y
)
0
.
1
0
9
1
[
0
.
0
3
2
6
]
*
*
0
.
0
3
9
2
[
0
.
0
2
5
4
]
0
.
0
1
8
5
[
0
.
0
1
1
6
]
0
.
0
2
1
3
[
0
.
0
3
5
5
]
0
.
1
8
6
1
[
0
.
0
4
1
0
]
*
*
*
0
.
1
3
9
4
[
0
.
0
3
0
3
]
*
*
*
0
.
1
4
6
7
[
0
.
1
0
4
1
]
0
.
2
0
2
9
[
0
.
1
3
2
6
]
N
u
m
b
e
r

o
f

o
b
s
e
r
v
a
t
i
o
n
s
1
,
1
1
7
1
,
1
1
7
1
,
1
1
7
1
,
1
1
7
1
,
0
3
1
1
,
0
3
1
1
,
0
3
1
1
,
0
3
1
H
a
u
s
m
a
n
n
0
.
1
2
4
0
.
1
0
7
7
r
h
o
-
w
a
l
d

t
e
s
t
0
.
2
5
2
8
0
.
2
0
9
4
*
*
*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
0
1
;

*
*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
0
5
;

*

s
i
g
n
i
f
c
a
n
t

a
t

p

<

0
.
1
0
.
N
o
t
e
:

R
e
s
u
l
t
s

f
o
r

e
a
c
h

v
a
r
i
a
b
l
e

a
r
e

p
r
e
s
e
n
t
e
d

w
i
t
h

t
w
o

r
o
w
s

b
e
l
o
w

t
h
e

c
o
e
f
f
c
i
e
n
t
.

T
h
e

f
r
s
t

r
o
w

c
o
r
r
e
s
p
o
n
d
s

t
o

t
h
e

H
a
u
s
m
a
n
n

t
e
s
t

a
n
d

t
h
e

s
e
c
o
n
d

t
o

r
h
o
-
w
a
l
d
.

T
h
e
s
e

a
r
e

t
w
o

t
e
s
t
s

f
o
r

e
x
o
g
e
n
e
i
t
y

o
f

t
h
e

a
f
f
l
i
a
t
i
o
n

v
a
r
i
a
b
l
e
.

T
h
e

f
r
s
t

t
e
s
t

c
o
r
r
e
s
p
o
n
d
s

t
o

t
h
e

t
w
o

s
t
a
g
e

p
r
o
b
i
t

m
o
d
e
l

(
p
r
o
b
i
t

i
n
s
t
r
u
m
e
n
t
a
l

v
a
r
i
a
b
l
e
)

a
n
d

t
h
e

s
e
c
o
n
d

c
o
r
r
e
s
p
o
n
d
s

t
o

t
h
e

b
i
v
a
r
i
a
t
e

p
r
o
b
i
t

m
o
d
e
l
.

R
e
j
e
c
t
i
o
n

o
f

t
h
i
s

t
e
s
t

m
e
a
n
s

t
h
a
t

a
f
f
l
i
a
t
i
o
n

i
s

n
o
t

e
x
o
g
e
n
o
u
s

i
n

t
h
e

o
r
i
g
i
n
a
l

m
o
d
e
l

a
n
d

t
h
a
t

i
t

i
s

n
e
c
e
s
s
a
r
y

t
o

u
s
e

t
h
e

i
n
s
t
r
u
m
e
n
t
a
l

v
a
r
i
a
b
l
e

a
p
p
r
o
a
c
h
.

O
n

t
h
e

o
t
h
e
r

h
a
n
d
,

w
h
e
n

t
h
e

H
o

i
s

a
c
c
e
p
t
e
d
,

o
n
e

c
o
n
t
i
n
u
e
s

t
o

u
s
e

t
h
e

p
r
o
b
i
t

m
o
d
e
l

w
i
t
h

c
o
n
t
r
o
l
s

b
e
-
c
a
u
s
e

t
h
e

t
e
s
t

i
n
d
i
c
a
t
e
s

t
h
a
t

t
h
e

a
f
f
l
i
a
t
i
o
n

i
s

n
o
t

e
n
d
o
g
e
n
o
u
s
.

I
f

a
f
f
l
i
a
t
i
o
n

i
s

n
o
t

e
n
d
o
g
e
n
o
u
s
,

s
e
l
e
c
t
i
o
n

b
i
a
s

i
n

t
h
i
s

r
e
s
u
l
t

v
a
r
i
a
b
l
e

i
s

d
u
e

o
n
l
y

t
o

o
b
s
e
r
v
a
b
l
e

v
a
r
i
a
b
l
e
s

t
h
a
t

a
r
e

i
n
c
l
u
d
e
d

i
n

t
h
e

m
o
d
e
l
.

F
o
r

t
h
i
s

c
a
s
e

o
n
l
y

t
h
e

p
r
o
b
i
t

m
o
d
e
l

i
s

n
e
c
e
s
s
a
r
y
.

T
h
e

m
o
s
t

r
o
b
u
s
t

c
o
e
f
f
c
i
e
n
t
s

a
r
e

m
a
r
k
e
d

i
n

b
o
l
d
.
S
o
u
r
c
e
:

A
u
t
h
o
r
s
.
Conclusion
Insurance coverage has increased greatly not only for the general population but
also for the poor. Both the subsidized regime and the contributory regime have
increased access to and use of health services among their beneciaries and made
positive changes in nancial protection in the case of an adverse health event.
But, to estimate the impact of insurance on health outcomes, we need data dier-
ent from those available. Our results are inconclusive for those in the subsidized
regime, and the LSMS data do not support such an analysis for those in the con-
tributory regime.
Insured individuals are much less likely to experience barriers to access when
needing care, and when facing such barriers they are less likely to be nancial and
more likely to be supply reasons such as excessive waiting times, low quality, and
unfriendly personnel. Tis result shows that the improvement in access through
insurance coverage could now be further enhanced by emphasizing policy mea-
sures to improve quality of care.
Te insured also use ambulatory health services more often than the unin-
sured. Poor and insured children suering from diarrhea or respiratory infections,
still among the main causes of premature death among small children in Colom-
bia, are more likely to visit a health care facility than their uninsured counterparts.
Despite the fact that immunizations are provided directly by the local health
authorities for free and irrespective of the individuals insurance status, insurance
TABLE 9.8
Propensity score matching results-estimated treatment effect on
subsidized regime participants for health status, national level, radius
(bandwidth 0.0001), 2005
Health status outcome variable
Enrolled in the
subsidized regime
(treatment group)
Not insured
(control group)
Difference
(percent) p
Complication after delivery 0.313 0.322 3
Extremely low birthweight 0.006 0.006 2
Extremely low birthweight from card 0.000 0.010 100
Low birthweight 0.085 0.052 63 ***
Low birthweight from card 0.077 0.038 100
Note: Other matching methods (double difference and matched double difference) were imple-
mented with similar results. Results from these methods can be obtained from the authors
on request.
Source: Authors calculations based on DHS 2005 data.
174 Chapter 9
increases immunization coverage. Tis shows that the benets of health insurance
are not limited to reduced nancial barriers and may provide other more indirect
paybacks. Tis eect might be related to increased knowledge connected to an
aliated mothers greater exposure to preventive health information and prodding
by providers. Similarly, aliated women benet from improved access to delivery-
related care. Tey receive more prenatal visits, and they are more likely to give birth
in a health facility and to be attended by a doctor or a skilled professional.
Te self-employed and their families in the contributory regime, who were
uninsured under the social security system operating until 1993, seem to have ben-
eted most. Nevertheless, descriptive data indicate that much remains to be done
on some key health indicators, as close to a third of the self-employed insured still
do not use any preventive health care services and 45% do not receive all medicine
prescribed for them.
Te nancial protection eects of insurance, for both schemes, are greater
when the household faces low health-related expenses. Tis means that the protec-
tive eects of insurance fall as the cost of the adverse health eect rises, exactly the
opposite of what should happen. Within the contributory regime, the self-employed
benet more from insurance during nancially catastrophic health events than
do other workers. In any case, health insurance in Colombia does provide nan-
cial protection to households, mitigating the nancial eects of an adverse health
event, but surely the design could be improved to provide more protection when
households face greater risks.
Te evidence on the impact of the subsidized regime on health status seems
less convincing. Tis result is related primarily to the quantity, characteristics, and
quality of the health status variables in the DHS. Tose surveys concentrate mainly
on health status variables related to women of child-bearing age and small children
(child survival, complications after delivery, and birthweight, for example). Health
insurance denitely increases use of professional care for all aspects of child deliv-
ery and care; the challenge is the contribution of those services to improving the
outcomes measured in the survey. As argued in chapter 2, connecting outcome
variables in surveys more directly to what insurance can do would help sort this
out.
Despite popular belief in Colombia, our results indicate that insurance matters
for the rural population and has improved access to and use of care, particularly for
the rural poor. Moreover, although social health insurance schemes are criticized
for the diculty of attracting the self-employed, our results show that the benets
of insurance are even more important among this group than for others insured in
the contributory regime.
Colombians need to address several important challenges if they want to fur-
ther improve the benets from the health insurance scheme and make their system
more sustainable. Now that access has substantially improved and nancial barriers
have been reduced, improving service quality becomes a key issue for researchers
and policymakers to consider and a key incentive issue in the design of reimburse-
ment policies. Similarly, the goal of the current administration to aliate more
than 50% of the population with the subsidized regime will require careful consid-
eration of the nancial sustainability of this subsidy expansion. Most important,
eligibility for the subsidized regime should be a transient feature conditioned on
nancial need, not a permanent and rigid right. Incentives must be created to foster
mobility from the subsidized regime to the contributory regime for those escaping
poverty. Te limits of the nancial protection oered by the insurance system are
often amenable to relatively inexpensive xes by the insurer that can provide tre-
mendous benets to those incurring catastrophic expenses.
Notes
1. Before the reform, mostly only formal sector workers of the private and public sector
had access to a full benet plan, and family coverage was limited. Under the Social
Security Institute (Instituto de Seguros Sociales, or ISS), which covered formal private
sector workers, only pregnant and dependent wives and their small children (under
age 1) had access to a limited array of birth-related services, leaving the remaining
services and all other dependent family members uncovered.
2. See Harvard Master Plan of Health Reform Implementation, 1997, for a synthesis of
the situation prior to the reform.
3. 1.5 percentage points of the 12.5% payroll tax contribution is channeled to the sub-
sidized regime.
4. Since 2005 those not poor enough to qualify for the subsidized regime but not wealthy
enough to be aliated with the contributory regime are aliated with a partial sub-
sidy system. Benets covered under the system are limited to coverage for catastrophic
illnesses (such as cancer, AIDS, and diabetes). Given that the data used in this study
stem from 2005 (when the aliation under the system just started), the impact of the
partial subsidy system is not analyzed here.
5. Detailed information on coverage by SISBEN levels and by income is in Giedion,
Daz, and Alfonso (2007).
6. Coverage has since increased but no updated information is available at the municipal
level.
7. Tis section is based on a conversation with Gilberto Barn, director of the Planning
Division of the ISS prior to the reform.
176 Chapter 9
8. In the late 1970s the ISS introduced family coverage in some small towns and villages.
In the early 1990s coverage was extended to some special economic groups, such as
priests, self-employed, and independent workers, and domestic helpers, prior to the
major reform of 1993.
9. Although results presented here are for analysis done using propensity score matching,
analysis was also done using double dierence and matched double dierence produc-
ing similar results to propensity score matching.
10. Information for the contributory regime refers to 2003, and the household dataset
come from LSMS 2003.
11. Excludes the population aliated and qualifying for the subsidized regime.
12. Informal care includes consulting a druggist, apothecary, quack, and the like when
facing a health problem.
13. Detailed statistics for the incidence of catastrophic and impoverishing expenditure
in the total (not just the user) population can be requested from the authors and are
presented in detail in Flrez, Giedion, and Pardo (2010).
14. Note that the use of dierent methods to evaluate the impact of the subsidized regime
(propensity score matching) and contributory regime (instrumental variables) on
nancial protection does not allow a straightforward comparison of the coecients.
Te sign and statistical signicance of results can, however, be compared.
15. As measured by the households nonsubsistence income, as indicated earlier.
References
Barn, G. 2006. Cuentas de Salud de Colombia 1993-2003: El gasto nacional en salud y su nancia-
miento. Bogota: Colombian Ministry of Social Protection.
Clavijo, S. 2009. Social Security Reforms in Colombia: Striking Demographic and Fiscal Bal-
ances. Working Paper WP/09/58. International Monetary Fund, Washington, D.C.
Buchmueller, T., R. Kronick, K. Grumbach, and J. G. Kahn. 2005. Te Eect of Health Insur-
ance on Medical Care Utilization and Implications for Insurance Expansion: A Review of the
Literature. Medical Care Research and Review 62 (1): 330.
CEPAL (Economic Commission for Latin America and the Caribbean). 2006. Anuario Estadstico
2006. Santiago: Economic Commission for Latin America and the Caribbean.
Escobar, M. L. 2005. Health Sector Reform in Colombia. Washington, DC: World Bank,
Development Outreach.
Flrez, C., U. Giedion, and R. Pardo. 2010. Te Impact of Health Insurance in Colombia on
Financial Protection. In A. Glassman, A. Giurida, M. Escobar and U. Giedion, eds., From
Few to Many: Ten Years of Health Insurance Expansion in Colombia Latin America, Washington
DC: Brookings Institution Press.
Gaviria, A., C. Medina, and C. Meja. 2006. Assessing Health Reform in Colombia: From Te-
ory to Practice. Economa (7)1: 2972.
Institute of Medicine. 2009. Americas Uninsured Crisis: Consequences for Health and Health Care.
Washington, DC: Te National Academies Press.
Khan, A. A., and S. M. Bhardwaj. 1994. Access to Health Care: A Conceptual Framework and
its Relevance to Health Care Planning. Evaluation & the Health Professions 17 (1): 6076.
Levy, H., and D. Meltzer. 2001. What Do We Really Know about Whether Health Insurance
Aects Health? ERIU Working Paper 6. University of Michigan, Economic Research Initia-
tive on the Uninsured, Ann Arbor, MI.
McLaughlin, C. G., and L. Wyszewianski. 2002. Access to Care: Remembering Old Lessons.
Health Services Research 37 (6): 144143.
Penchansky, R., and J. W. Tomas. 1981. Te Concept of Access: Denition and Relationship to
Consumer Satisfaction. Medical Care 19 (2): 12740.
Phelps, C. 2009. Health Economics (4th Edition). Boston, MA: Addison Wesley Higher Education
Trujillo, A., J. Portillo, and J. Vernon. Te Impact of Subsidized Health Insurance for the Poor:
Evaluating the Colombian Experience Using Propensity Score Matching. International Jour-
nal of Health Care Finance and Economics 5 (3): 21139.
Tsai, T. 2010. Second Chance for Health Reform in Colombia. Te Lancet 375 (9709): 10910.
World Bank. 2007. Healthy Development: Te World Bank Strategy for Health, Nutrition, and Popu-
lation Results. Washington, DC: International Bank for Reconstruction and Development.
. 2010. World Development Indicators 2010. Washington, DC: World Bank.
Yamin, A. and O. Parra-Vera. 2009. How Do Courts Set Health Policy? Te Case of the Colom-
bian Constitutional Court, PLoS Medicine 6(2):147-150. Available at http://www.ncbi.nlm.
nih.gov/pmc/articles/PMC2642877/pdf/pmed.1000032.pdf.
178
Main Findings, Research Issues,
and Policy Implications
Maria-Luisa Escobar, Charles C. Grin, and R. Paul Shaw
Tis book contains rich and varied analyses of the impact of health
insurance in dierent socioeconomic and organizational settings. It
begins with a comprehensive literature review that distills ndings on
prior studies that examine causal eects between health insurance and
health outcomes. Tis is followed by seven country case studies, most of
which use advanced statistical techniques and new data sources to shed
light on how health insurance improves health outputs and outcomes.
Tis chapter summarizes the main ndings, the methodological issues
that can understate or diminish the estimated impact of health insur-
ance on health, and the country scenarios that illustrate the art of the
possible for policymakers interested in scaling up well designed health
insurance programs.
To a large extent, selecting the countries was opportunistic, as
explained in chapter 1. In view of pressing policy concerns in many
low-income countries, as well as major gaps in our knowledge of the
impacts of health insurance, we chose countries where scaling up
health insurance aims to be more inclusive of the poor and where
available data permit explorations of the impact on health status. No
pretense is made, therefore, that ndings reported here come from a
uniform dataset or research methodology applied to all countries or
that all studies satisfy the gold standard for empirical robustness as
described in chapter 2. Instead, several caveats and qualiers should be
kept in mind.
C
h
a
p
t
e
r

1
0
Main Findings, Research Issues, and Policy Implications 179
Above all, we have learned that health insurance is not a homogeneous product
(like an approved oral medicine or vaccine). It tends to be heterogeneous in the
entitlements to medical goods and services created for health insurance members;
the quantity, quality, and distribution of providers where members can access ser-
vices; the extent that copayments and deductibles aect out-of-pocket spending by
members; and so on. Tis heterogeneity cannot be controlled or made uniform.
It shapes the impact that health insurance has on outputs and outcome measures
in one country versus another, so that the measured eects of health insurance on
access, service uptake, and out-of-pocket spending vary widely across countries. Te
temptation to generalize ndings across countries must be tempered accordingly.
We have also learned that the extent to which health insurance succeeds in
being pro-poor has more to do with a purposive eort to design health insurance
in a way that benets the poor than with any presumption that health insur-
ance is automatically and intrinsically pro-poor or anti-poor. Health insurance
has important design features that can benet low-income households, such as
pooling contributions by rich and poor households, then paying for treatment of
illnesses that disproportionately fall on the poor. But if health insurance fails to
enroll the poor or extend services to them, the distributional impacts of health
insurance on equity will likely be muted. Accordingly, to generalize that health
insurance does, indeed, contribute to greater equity in health care consumption
will be conditional on successful pro-poor design features. Tis caveat also applies
to generalizations we might be inclined to make about distributional impacts
of health insurance on women and children. Pooling risks is equity improving
within the risk pool, but who benets depends on who is in the risk pool and how
it is designed to function.
Finally, we have learned that the robustness of empirical analysis varies across
case studies. Robustness depends on the comprehensiveness and quality of datasets,
availability of appropriate measures of impact, success in controlling for endo-
geneity, and appropriate application of statistical models. Tis caveat forms the
backbone of chapter 2, where checklists of key methodological concerns are used
to score the quality of a wide variety of studies and the robustness of the empirical
estimates they have reported. Managing these problems well is critical if empirical
ndings are to be taken seriously.
With these caveats in mind, the collective ndings in this book do lend them-
selves to some conservative generalizations that advance not only the evidence base
but also contribute information to current policy debates on the desirability of scal-
ing up health insurance. Te next section considers the accumulation of evidence
in six areas.
180 Chapter 10
Six general ndings
In this volume the estimated benets of health insurance among the insured
are measured relative to conditions aecting the uninsured. In this sense, the
uninsured represent a baseline for determining the value added of health insur-
ance. Tis prompts the questions: What does it mean to be uninsured? What is
health insurance trying to improve on? Te rst case study, on Namibia, pro-
vides useful insights into these questions in an African context. First, it reveals
that even though uninsured households (about half ) presumably have access to
a reasonably well functioning public health system, they are less likely to report
an illness than those with public or private insurance. And when they do report
illness, they are less likely to seek care. Moreover, the health shocks experi-
enced by those without insurance lead to higher medical expenses, reduced
food and nonfood consumption, and fewer assets than among insured house-
holds. Consequences of being without insurance are likely to be particularly
dire for Namibian households in the bottom income quintile: they are three
times more likely to have a hospitalization (three or more days) than those in
the top quintile and one and a half times more likely to have HIV/AIDS or die.
Te data point to substantial dierences among population groups even in the
presence of a relatively well nanced and functioning public system of direct
service delivery.
Findings from the other case studies, while not explicitly designed to prole
households without health insurance, deepen the foregoing perspective by convey-
ing that those without insurance are more prone to:
Go without treatment.
Self-treat and self-medicate.
Benet less from preventive services.
Have much higher shares of out-of-pocket spending as a percentage of their
disposable income.
Incur catastrophic nancial loss, borrowing, or indebtedness.
Have poorer self-perceptions of their health status.
Although young single people tend to be more prevalent among the uninsured,
especially if health insurance requires voluntary enrollment, a large share of the
uninsured in the low- and middle-income countries covered in this book tend to
be relatively poor families with low levels of education, self-employed workers,
migrants, and people living in rural and remote areas.
Turning now to the impact of health insurance in the countries examined in
this book, table 10.1 distills the main ndings. Tey are based on empirical esti-
mates that for the most part have attempted to purge the eects of endogeneity
and quantify impacts of health insurance on the insured relative to the uninsured.
We conclude that the evidence reasonably supports six generalizations, all of which
provide further support for and extend the ndings from the global literature
review in chapter 2.
Health insurance can produce signicant positive impacts on access and use
In Colombia low-income health insurance members are 41% more likely to have
had an outpatient visit in the 12 months prior to the survey than low-income non-
members even though the latter have access to public clinics and hospitals. Insured
Ghananians had 72% lower outlays than the uninsured but were twice as likely to
use formal care and half as likely to self-treat or use informal practitioners. Insured
mothers paid 90% less for prenatal and delivery care. Only 13% paid anything,
compared with 81% of the uninsured. While the magnitude of impacts varies by
study and country, these ndings are consistent with the hypothesis that insur-
ance removes barriers to access. All studies found an increase in use of services (see
table 10.1).
Increased access is the rst step toward increased uptake of cost-eective
outpatient services known to aect preventable communicable diseases, such as
immunizations, as reported in Peru. Tis is a major appeal of health insurance in
countries where health services are underused, especially by women and children.
Te complement of increased access and use is reduced prevalence of no treatment,
self-treatment, and informal care. Ghana and China both reported self-treatment
among the insured around 30% lower than among the uninsured; in Colom-
bia bringing the self-employed into the contributory scheme caused substantial
improvements in their use of preventive and dental services as well as medicines,
and it lowered their use of self-medication and informal care.
Health insurance can and does benet poorer households as much as better
o households, if not more
To a large extent, this nding is associated with well designed health insurance pro-
grams that target low-income households, as in Peru, Colombia, and Costa Rica.
Te voluntary scheme in Ghana has few contributors, and the targeting of subsidies
except by age needs improvement. Chinas experiment is targeted to poorer rural
provinces, but because it is voluntary, the poorest households are the last to enroll.
Integral Health Insurance (SIS) in Peru, the subsidized regime in Colombia, and
the noncontributory regime in Costa Rica use means tests to achieve this goal.
Te poor in Peru and Colombia gain substantial benets from insurance coverage.
Te case study on Indonesia demonstrates that health insurance can exert positive
182 Chapter 10
T
A
B
L
E

1
0
.
1

O
b
s
e
r
v
e
d

i
m
p
a
c
t
s

o
f

h
e
a
l
t
h

i
n
s
u
r
a
n
c
e

o
n

s
e
l
e
c
t
e
d

i
n
d
i
c
a
t
o
r
s
C
o
u
n
t
r
y
I
n
c
r
e
a
s
e
s

u
s
e
R
e
d
u
c
e
s

s
e
l
f
-
t
r
e
a
t
m
e
n
t
R
e
d
u
c
e
s

o
u
t
-
o
f
-
p
o
c
k
e
t

s
p
e
n
d
i
n
g
B
e
n
e
f
t
s

r
e
l
a
t
i
v
e
l
y

p
o
o
r

h
o
u
s
e
h
o
l
d
s
I
m
p
r
o
v
e
s

s
e
l
f
-
a
s
s
e
s
s
e
d

h
e
a
l
t
h

s
t
a
t
u
s

a
n
d
/
o
r

u
p
t
a
k
e

o
f

c
o
s
t
-
e
f
f
e
c
t
i
v
e

p
r
e
v
e
n
t
i
v
e

s
e
r
v
i
c
e
s
N
a
m
i
b
i
a
n
a
n
a
n
a
N
o
,

v
o
l
u
n
t
a
r
y

i
n
s
u
r
a
n
c
e

t
a
k
e
n

u
p

f
r
s
t

b
y

t
h
e

r
i
c
h
n
a
G
h
a
n
a
Y
e
s
Y
e
s
Y
e
s
,

a

l
a
r
g
e

r
e
d
u
c
t
i
o
n

i
n

h
a
v
i
n
g

a
n
y

s
p
e
n
d
-
i
n
g

f
o
r

a
c
u
t
e

c
a
r
e

a
n
d

d
e
l
i
v
e
r
i
e
s
N
o
,

v
o
l
u
n
t
a
r
y

i
n
s
u
r
a
n
c
e

t
a
k
e
n

u
p

f
r
s
t

b
y

t
h
e

r
i
c
h
;

e
x
e
m
p
t
i
o
n
s

n
o
t

w
e
l
l

t
a
r
g
e
t
e
d
N
o

e
f
f
e
c
t
s

f
o
u
n
d

i
n

p
r
e
-
p
o
s
t

a
n
a
l
y
s
i
s
,

b
u
t

p
r
o
p
e
n
s
i
t
y

s
c
o
r
e

m
a
t
c
h
i
n
g

s
u
g
g
e
s
t
s

a

l
a
r
g
e

i
n
s
u
r
a
n
c
e

i
m
p
a
c
t

o
n

p
r
e
n
a
t
a
l

c
a
r
e
,

f
a
c
i
l
i
t
y

d
e
l
i
v
e
r
y
,

a
n
d

C
a
e
s
a
r
e
a
n

s
e
c
t
i
o
n
.

I
n
s
u
r
a
n
c
e

s
u
b
s
t
a
n
t
i
a
l
l
y

r
e
d
u
c
e
d

e
x
p
e
n
d
i
t
u
r
e
s

o
n

t
h
e
s
e

s
e
r
v
i
c
e
s

d
e
s
p
i
t
e

h
i
g
h
e
r

u
s
e
.
C
o
s
t
a

R
i
c
a
N
o

e
f
f
e
c
t

b
e
c
a
u
s
e

a
l
l

n
e
e
d
i
n
g

c
a
r
e

m
a
y

r
e
c
e
i
v
e

i
t

b
y

p
a
y
i
n
g

o
r

f
o
r

f
r
e
e

i
f

i
n
d
i
g
e
n
t
,

e
v
e
n

i
f

n
o
t

i
n
s
u
r
e
d
n
a
N
o

i
m
p
a
c
t

o
n

o
u
t
-
o
f
-
p
o
c
k
e
t

h
e
a
l
t
h

s
p
e
n
d
i
n
g

b
e
c
a
u
s
e

e
m
e
r
g
e
n
c
y

c
a
r
e

i
s

f
r
e
e

t
o

a
l
l
;

n
o
n
e
m
e
r
-
g
e
n
c
y

c
a
r
e

f
r
e
e

t
o

i
n
d
i
g
e
n
t
s
N
o

d
i
f
f
e
r
e
n
c
e

a
m
o
n
g

q
u
i
n
t
i
l
e
s

a
s

8
1
%

o
f

t
h
e

s
a
m
p
l
e

h
a
v
e

h
e
a
l
t
h

i
n
s
u
r
a
n
c
e

a
n
d

i
n
d
i
g
e
n
t
s

r
e
c
e
i
v
e

f
r
e
e

c
a
r
e

e
v
e
n

i
f

n
o
t

i
n
s
u
r
e
d
+

i
m
p
a
c
t

o
n

v
a
c
c
i
n
a
t
i
o
n
s

f
o
r

c
h
i
l
d
r
e
n

u
n
d
e
r

a
g
e

1
8
.

R
e
d
u
c
e
d

u
s
e

o
f

i
n
p
a
t
i
e
n
t

a
n
d

e
m
e
r
g
e
n
c
y

c
a
r
e

f
o
r

t
h
o
s
e

w
i
t
h

d
i
a
b
e
t
e
s
,

m
o
r
e

a
p
p
r
o
p
r
i
a
t
e

d
o
s
a
g
e
s

f
o
r

i
n
s
u
r
e
d

d
i
a
b
e
t
i
c
s
.
U
n
i
n
s
u
r
e
d

a
r
e

y
o
u
n
g
e
r
,

h
e
a
l
t
h
i
e
r

b
u
t

m
o
r
e

l
i
k
e
l
y

t
o

e
n
t
e
r

h
o
s
p
i
t
a
l

t
h
r
o
u
g
h

e
m
e
r
g
e
n
c
y

r
o
o
m
,

t
o

h
a
v
e

a

l
o
n
g
e
r

s
t
a
y
,

a
n
d

t
o

d
i
e
.
+

e
f
f
e
c
t

o
n

p
e
r
c
e
i
v
e
d

h
e
a
l
t
h

s
t
a
t
u
s

(
e
x
c
e
p
t

d
i
a
b
e
t
i
c
s
)
.
R
e
d
u
c
e
s

h
o
s
p
i
t
a
l

l
e
n
g
t
h

o
f

s
t
a
y

a
n
d

l
o
w

b
i
r
t
h
w
e
i
g
h
t

n
e
w
b
o
r
n
s
.
P
e
r
u
Y
e
s
n
a
Y
e
s
,

o
u
t
-
o
f
-
p
o
c
k
e
t

a
n
d

c
a
t
a
s
t
r
o
p
h
i
c

s
p
e
n
d
i
n
g
;

l
a
r
g
e

r
e
d
u
c
t
i
o
n

i
n

p
r
o
b
a
b
i
l
i
t
y

o
f

a
n
y

s
p
e
n
d
i
n
g
Y
e
s
,

e
x
p
l
i
c
i
t
l
y

t
a
r
g
e
t
e
d

h
e
a
l
t
h

i
n
s
u
r
a
n
c
e

f
o
r

l
o
w
-
i
n
c
o
m
e

h
o
u
s
e
h
o
l
d
s

b
u
t

s
o
m
e

e
f
f
e
c
t
s

m
o
r
e

p
r
o
-
p
o
o
r

t
h
a
n

o
t
h
e
r
s
S
t
r
o
n
g

+

c
h
a
n
g
e

f
o
r

h
a
v
i
n
g

f
u
l
l

i
m
m
u
n
i
z
a
t
i
o
n

s
e
r
i
e
s

c
o
m
p
l
e
t
e

f
o
r

c
h
i
l
d
r
e
n

+

f
o
r
m
a
l

t
r
e
a
t
m
e
n
t

o
f

c
h
i
l
d

d
i
a
r
r
h
e
a

a
n
d

a
c
u
t
e

r
e
s
p
i
r
a
t
o
r
y

i
n
f
e
c
t
i
o
n
,

e
s
p
e
c
i
a
l
l
y

p
o
o
r
e
r

h
o
u
s
e
h
o
l
d
s
.
+

e
f
f
e
c
t

o
n

g
r
o
w
t
h

m
o
n
i
t
o
r
i
n
g

v
i
s
i
t
s

f
o
r

c
h
i
l
d
r
e
n

u
n
d
e
r

a
g
e

5
.
S
m
a
l
l

+

e
f
f
e
c
t

o
n

a
t
t
e
n
d
e
d

d
e
l
i
v
e
r
i
e
s

a
n
d

o
n

p
r
o
b
a
b
i
l
i
t
y

o
f

r
e
c
e
i
v
i
n
g

a

P
a
p

s
m
e
a
r
,

b
u
t

o
n
l
y

f
o
r

o
l
d
e
r

w
o
m
e
n

(
a
p
p
r
o
p
r
i
a
t
e
l
y
)
.
N
o

e
f
f
e
c
t

o
n

i
n
s
t
i
t
u
t
i
o
n
a
l

d
e
l
i
v
e
r
i
e
s
.
(
c
o
n
t
i
n
u
e
d
)
C
o
u
n
t
r
y
I
n
c
r
e
a
s
e
s

u
s
e
R
e
d
u
c
e
s

s
e
l
f
-
t
r
e
a
t
m
e
n
t
R
e
d
u
c
e
s

o
u
t
-
o
f
-
p
o
c
k
e
t

s
p
e
n
d
i
n
g
B
e
n
e
f
t
s

r
e
l
a
t
i
v
e
l
y

p
o
o
r

h
o
u
s
e
h
o
l
d
s
I
m
p
r
o
v
e
s

s
e
l
f
-
a
s
s
e
s
s
e
d

h
e
a
l
t
h

s
t
a
t
u
s

a
n
d
/
o
r

u
p
t
a
k
e

o
f

c
o
s
t
-
e
f
f
e
c
t
i
v
e

p
r
e
v
e
n
t
i
v
e

s
e
r
v
i
c
e
s
I
n
d
o
n
e
s
i
a
Y
e
s
n
a
Y
e
s
,

r
e
d
u
c
e
s

p
r
o
b
a
b
i
l
i
t
y

o
f

a
n
y

s
p
e
n
d
i
n
g

b
u
t

n
o
t

o
v
e
r
a
l
l

a
m
o
u
n
t

s
p
e
n
t
Y
e
s
,

l
a
r
g
e
r

u
s
e

i
m
p
a
c
t
s

f
o
r

r
u
r
a
l

a
n
d

p
o
o
r

h
o
u
s
e
h
o
l
d
s
,

b
u
t

d
o
e
s

n
o
t

a
p
p
l
y

t
o

o
u
t
-
o
f
-
p
o
c
k
e
t

s
p
e
n
d
i
n
g
+

e
f
f
e
c
t

f
o
r

o
u
t
p
a
t
i
e
n
t

c
a
r
e

a
n
d

s
p
e
n
d
i
n
g

f
o
r

f
e
m
a
l
e

c
h
i
l
d
r
e
n
.
S
m
a
l
l

e
f
f
e
c
t

t
o

r
e
d
u
c
e

p
r
o
b
l
e
m
s

w
i
t
h

a
c
t
i
v
i
t
i
e
s

o
f

d
a
i
l
y

l
i
f
e

a
n
d

r
e
d
u
c
t
i
o
n

i
n

h
i
g
h

b
l
o
o
d

p
r
e
s
s
u
r
e

f
o
r

a
d
u
l
t
s

i
n

l
o
w
e
r

i
n
c
o
m
e

g
r
o
u
p
.
C
h
i
n
a
Y
e
s
,

i
n
c
r
e
a
s
e
d

o
u
t
p
a
t
i
e
n
t

b
u
t

n
o
t

i
n
p
a
t
i
e
n
t
Y
e
s
,

s
e
l
f
-
m
e
d
i
c
a
t
i
o
n
Y
e
s
,

o
u
t
-
o
f
-
p
o
c
k
e
t

a
n
d

c
a
t
a
s
t
r
o
p
h
i
c

s
p
e
n
d
i
n
g
Y
e
s
,

e
x
p
l
i
c
i
t

t
a
r
g
e
t
i
n
g

o
f

t
h
e

p
o
o
r

b
y

d
e
s
i
g
n

(
p
o
o
r

g
e
o
g
r
a
p
h
i
c
a
l

a
r
e
a
)
;

w
i
t
h
i
n

s
a
m
p
l
e

r
i
c
h
e
r

h
o
u
s
e
h
o
l
d
s

e
n
r
o
l
l
e
d

i
n

v
o
l
u
n
t
a
r
y

s
c
h
e
m
e
I
m
p
r
o
v
e
d

s
e
l
f
-
a
s
s
e
s
s
e
d

h
e
a
l
t
h

s
t
a
t
u
s

a
n
d

m
o
b
i
l
i
t
y

a
m
o
n
g

o
l
d
e
s
t

a
g
e

g
r
o
u
p
.

V
a
s
t
l
y

i
m
p
r
o
v
e
d

p
r
e
s
c
r
i
b
i
n
g

b
e
h
a
v
i
o
r

o
f

p
r
o
v
i
d
e
r
s

a
n
d

e
l
i
m
i
n
a
t
e
d

f
a
k
e

a
n
d

e
x
p
i
r
e
d

d
r
u
g
s
.
S
i
g
n
i
f
c
a
n
t
l
y

r
e
d
u
c
e
d

t
h
e

p
r
o
b
a
b
i
l
i
t
y

o
f

p
a
i
n
/
d
i
s
c
o
m
f
o
r
t

a
n
d

a
n
x
i
e
t
y
/
d
e
p
r
e
s
s
i
o
n
.
C
o
l
o
m
b
i
a
Y
e
s
Y
e
s
,

c
o
n
t
r
i
b
u
t
o
r
y

r
e
g
i
m
e

r
e
d
u
c
e
s

i
n
f
o
r
m
a
l

c
a
r
e

u
s
e
,

s
e
l
f
-

t
r
e
a
t
m
e
n
t
,

n
o

h
e
a
l
t
h

c
a
r
e
Y
e
s
,

o
u
t
-
o
f
-
p
o
c
k
e
t

a
n
d

c
a
t
a
s
t
r
o
p
h
i
c

s
p
e
n
d
i
n
g
;

b
u
t

s
t
i
l
l

u
n
n
e
c
e
s
s
a
r
i
l
y

h
i
g
h

d
u
e

t
o

d
e
s
i
g
n
Y
e
s
,

t
h
r
o
u
g
h

t
h
e

s
u
b
s
i
d
i
z
e
d

r
e
g
i
m
e
,

l
a
r
g
e
r

i
m
p
a
c
t
s

f
o
r

l
o
w
-
i
n
c
o
m
e

r
u
r
a
l

h
o
u
s
e
h
o
l
d
s
+

i
m
p
a
c
t
s

o
f

s
u
b
s
i
d
i
z
e
d

r
e
g
i
m
e

o
n

r
e
c
e
i
v
i
n
g

c
a
r
e

w
h
e
n

n
e
e
d
e
d
,

s
k
i
l
l
e
d

b
i
r
t
h

a
t
t
e
n
d
a
n
c
e
,

b
i
r
t
h

i
n

h
e
a
l
t
h

f
a
c
i
l
i
t
y
,

n
u
m
b
e
r

o
f

p
r
e
n
a
t
a
l

v
i
s
i
t
s
,

c
h
i
l
d

g
e
t
t
i
n
g

t
r
e
a
t
m
e
n
t

f
o
r

c
o
u
g
h

o
r

d
i
a
r
r
h
e
a
,

a
n
d

f
u
l
l

i
m
m
u
n
i
z
a
t
i
o
n

o
f

c
h
i
l
d
.
+

i
m
p
a
c
t
s

o
f

c
o
n
t
r
i
b
u
t
o
r
y

r
e
g
i
m
e

o
n

m
a
k
i
n
g

p
r
e
v
e
n
t
i
v
e

d
o
c
t
o
r
/
d
e
n
t
a
l

v
i
s
i
t
s

a
n
d

g
e
t
t
i
n
g

f
u
l
l

c
o
u
r
s
e

o
f

m
e
d
i
c
i
n
e
s
.
N
o
t
e
:

n
a
=
n
o

a
c
c
o
u
n
t

o
r

n
o
t

e
x
p
l
i
c
i
t
l
y

e
v
a
l
u
a
t
e
d
.
T
A
B
L
E

1
0
.
1

(
c
o
n
t
i
n
u
e
d
)

O
b
s
e
r
v
e
d

i
m
p
a
c
t
s

o
f

h
e
a
l
t
h

i
n
s
u
r
a
n
c
e

o
n

s
e
l
e
c
t
e
d

i
n
d
i
c
a
t
o
r
s
184 Chapter 10
distributional eects on rural versus urban women and children, even in a fairly
short timeframe, and in that case it appears that poorer households gained the
most from insurance coverage in gaining access.
Te impact of health insurance on measures of health status is weak and
irregular but strong for self-perceptions
Tis nding is interesting but far less important than the impact on actual behav-
ior. Two studies examine the impacts on self-reported health status. Using a stan-
dardized measure of self-reported health status EQ-5D dimensions the China
study reports the probability of having a problem in any of the EQ-5D dimensions
among the insured was 49% less than among the uninsured. In the Costa Rica
study health insurance improves an individuals self-perception of health status,
with the insured having a higher probability of perceiving their health status to be
good or very good. Moreover, the probability of declaring good or very good health
status was 10 percentage points higher for a subsample of insured people with at
least one diagnosed chronic disease than for those who are uninsured. Yet for dia-
betics, insurance reduced their self-perceived health even though it improved their
treatment signicantly over uninsured diabetics.
Health insurance reduces out-of-pocket payments, thus reducing
vulnerability to having to pay at times of illness or injury
In Ghana uninsured patients had out-of-pocket spending three times that of the
insured for formal care, three times that for informal care, and twice that for hospi-
tal care. In Indonesia health insurance has a signicant negative eect on the likeli-
hood of out-of-pocket spending, showing larger eects for rural over urban adults
and the largest eects for rural women. It also appears to have a greater impact
on reducing expenditures for those who are poor, more isolated, and with lower
availability of health services all variables correlated with residence in rural areas.
In Peru the probability of any spending on health services among those receiving
formal care in the four weeks prior to the survey was only 13% for those covered by
SIS but 86% for those without it . SIS almost eliminated spending for its aliates.
And in Costa Rica health insurance had a signicant impact on reducing per capita
spending on health both as a proportion of total per capita spending and as a pro-
portion of capacity for payment. As expected, however, the quantitative impact of
health insurance on out-of-pocket spending varies considerably depending on the
health goods and services covered and the schedule of copayments or deductibles.
Consistently across the studies, insurance tends to increase the probability of no
expenditures, while those incurring costs tend to spend as much as before or more.
Health insurance reduces the incidence of catastrophic nancial loss due to
high costs associated with serious illness or injury
Studies examining the relationship between health insurance and catastrophic nan-
cial loss tend to rely on proxies rather than on measures that actually gauge whether a
nancial catastrophe occurred at times of serious illness or injury. From the Namibia
case, for example, it is clear that families with a health problem may rely on coping
mechanisms such as borrowing from family, selling assets, and changing consump-
tion patterns, which most surveys cannot account for. Four of the studies in this vol-
ume proxied catastrophic nancial loss as a rising share of household nonsubsistence
expenditures on health. Te China case study denes catastrophic health spending as
out-of-pocket spending greater than 10%, 20%, 30%, and 40% of household income
(net of food expenditures). It found that having health insurance reduced rates of
catastrophic spending by signicant margins and reduced medical impoverishment
by about one-fth for those in the lowest income quartile. In Colombia both the
subsidized and contributory regimes reduce catastrophic expenditures, but their pro-
tection decreases, and out-of-pocket expenditures increase for the most serious, costly
illnessesso while there is catastrophic protection, its impact drops o just when it is
most needed. In both plans costs for uncovered or partially covered services cause the
problem. Costa Ricas program provides a signicant margin of nancial protection
to everyone, whether insured or not. Out-of-pocket health expenditures represent
only 2% of total expenditures for the poorest third of households and 4% for the
richest third. In the absence of nancial protection, this pattern is typically reversed,
with much higher percentages for all households, but especially for the poor.
Research issues
Chapter 2 concludes that methodological problems undermine the robustness of
more than half of past health insurance studies, such that causality cannot reliably
be established. Accordingly, chapter 2 provides guidelines to detect and correct
such problems. Nine issues cropped up to varying degrees in the case studies in this
volume (gure 10.1). Without appropriate adjustments, especially in retrospective
analysis, one or more of these issues can diminish the measured impact of health
insurance, as explained below.
Endogeneity
Because health insurance tends to be plagued by adverse selection, failure to fully
control for endogeneity means that people with poor health self-select to join health
insurance. Tus we might infer empirically that insurance causes poor health
because of self-selection of the sick into insurance pools. In Ghana, for example,
186 Chapter 10
people reporting chronic conditions were more likely to enroll in voluntary insur-
ance, and among richer households pregnant women were more likely to enroll.
Te positive impact of health insurance on health would be underestimated as a
result. Alternatively, perceived health status could appear to worsen with health
insurance, whereas in fact, health insurance may make people more knowledgeable
through more contacts with professionals, resulting in better health outcomes. Tis
certainly seems to be the case for diabetics in Costa Rica, who clearly get much
better care if insured but consider themselves sicker than uninsured diabetics do.
Poorly specied causal chain
Health insurance does not act directly on mortality-based measures (such as child
or adult mortality or life expectancy) or measures of stock or stature (such as
Diminished
impact
Endogeneity
Poorly
specied
causal chain
Spillover
eects on
provider
performance
Weak
instrumental
variables
Limited and
poor data
quality
Varying
provider
access
Design
elements that
limit nancial
access to
services or
reimbursement
Inadequate
timeline for
eects to
manifest
Inconsistent
exposure to
health insurance
over time
FIGURE 10.1
Factors diminishing the measured impact of health insurance
height-for-age or weight-for-age) it acts on improving health through access to
clinically proven medical services known to improve health. An attempt to connect
health insurance causally to more general health outcome measures that are slow to
change or rarely observed and are aected by many other factors will likely under-
state or diminish its impact. Tus, researchers should concentrate on measures of
health status that can be reasonably well connected to access to health services
oered under the health insurance arrangement being analyzed.
Inadequate timeline for eects to manifest
Scaling up health insurance requires advances on many fronts before signicant
and consistent eects emerge. Enrollments may well double in a short period, but
this does not mean provider networks are fully functioning, that quality issues have
been resolved, that patient entitlements are clear, or that insured households have
built trust in new medical schemes. Chapter 2 suggests that at least 1218 months
may be required to detect eects. Anything shorter than that may under estimate
the longer term eects of health insurance. Tus the study of the social experiment
in China is instructive, but it may understate impacts that accumulate over time
or start to show up after providers and patients gain additional years of experience
with the plan.
Inconsistent exposure to health insurance over time
In medical terms, when people are enrolled in health insurance as part of an
experiment to determine how they fare against those not enrolled, the enrollees
are regarded as the treatment group. Further, we know from in-depth studies of
compliance from the pharmaceutical industry that a major problem in studies of
drug eectiveness is that many people in the treatment group do not comply with
prescription instructions or drop out of treatment. Without appropriate controls,
this results in inconsistencies in exposure to drug treatments that can understate
estimated drug impacts. In health insurance studies, the term we use to describe
such individuals can be adopted from the Indonesia study in chapter 7, called
switchers. Te less they are controlled for, the more eects of health insurance may
be underestimated. Switching is an even more signicant problem if it is driven by
self-selection in when sick, out when well.
Design elements that limit nancial access to services or reimbursement
Policies on copayments, deductibles, and coinsurance by health insurance plans
tend to vary widely across countries and even within countries. Insurance may
induce households to use more medical services as nancial barriers fall, or the
188 Chapter 10
insurance plan may have a benet cut-out, as in Colombia, or limits may reduce
catastrophic protections, as in China. Failure to appreciate such features may lead
to the supercial conclusion that health insurance has far less impact on access and
out-of-pocket spending among poor households than expected. Without appreci-
ating the nature of the design issue, researchers may unintentionally understate
or diminish the eect that health insurance could have on nancial access and
nancial risk protection, because they typically cannot incorporate specic plan
attributes in their analysis.
Varying provider access
Increasingly, health insurance funds act as purchasers of services for their mem-
bers by contracting with networks of providers public, private, nongovernmental
organizations, or a mix. In some countries, contracted networks of providers may
be fairly extensive, providing medical entitlements to health insurance members in
both urban and rural areas. But in many rural and remote areas provider networks
are likely to be much thinner. Te more that provider networks are uneven, the less
the estimated eects of health insurance are likely to be. Tis consideration may
diminish the measured impact of health insurance more than expected, because of
inadequate recognition of key organizational features of delivery.
Limited and poor quality data
When data limitations prevent disaggregating the insured and uninsured into dif-
ferent subgroups, the possible impact of health insurance on groups of particular
interest, such as women and children in rural areas, will be obscured. Moreover,
small impacts of health insurance based on highly aggregated data may obscure
large impacts among subgroups which tend to cancel out when aggregated. In
such cases impacts of health insurance would be underestimated for such groups.
Weak instrumental variables
Confronted by unobservable variables that may be distorting the estimated rela-
tionship between the dependent variable (health outcomes) and the independent
variable (health insurance), researchers often resort to replacing health insurance
with an instrumental variable. Te strength of this approach is based on the plau-
sibility of the instruments, which in turn must comply with three assumptions
(Wooldridge 2001). First, the instrument must substantially explain aliation with
the insurance plan. Second, it should not have any direct eect on the outcome
variable of interest. Tird, it should not have an indirect eect on the outcome
through other variables left out of the outcome equation. At the simplest level,
in most datasets it is dicult to nd instrumental variables that aect insurance
status but not the variable of interest, such as use of health services or health status.
Judgments and tradeos are always made in trying to solve this problem.
Spillover eects on provider performance
Scaling up a well designed health insurance program will have wide ranging
impacts on the nance, organization, regulation, and behavior of the health system
as a whole. Typically the policy change involves partial or full separation of public
nance and provision with increased eciencies through contracting, improved
targeting of public subsidies for health care to the poor, and a better demarcation
between public spending for acute care and public spending for population-based
health care. In other words, health insurance may have positive spillover eects
that also improve the lot of the uninsured relative to the insured. When this hap-
pens, the impact of health insurance would tend to be underestimated as the rising
tide of greater eciency and equity of public subsidies lifts all boats.
Policy implications
Policymakers hope that in introducing or scaling up health insurance, the health
and well-being of citizens will be enhanced. Yet the uneven evidence in the past
has exposed policymakers in some countries to far-ranging debate, if not cyni-
cism, about the intended eects of health insurance. China is a case in point, with
some studies showing little or no consistent eects on important health outcome
parameters and others showing positive eects on access, use, and nancial risk
protection.
Realistically, this is a matter of designing the intervention to address the prob-
lem. Scaling up health insurance in low- and middle-income countries can, indeed,
deliver the kinds of eects that appeal to policymakers but only if the design of
health insurance explicitly embodies features that can yield those eects, an ade-
quate provider network is in place to serve the insured (or the intervention works
on providers too), and the details of implementation and execution of the policy
are managed well so the reform performs as designed. Five of the case studies
Ghana, China, Peru, Colombia, and Costa Rica help illustrate the art of the
possible in these dimensions.
Ghana saving nearly 1.5 million disability-adjusted life years
Te toll of mortality and morbidity in Ghana was about 12 million disability-
adjusted life years in 2004. About 71% of this toll was attributable to commu-
nicable diseases, maternal and perinatal conditions, and nutritional deciencies.
190 Chapter 10
Four major killers including infectious/parasitic diseases, respiratory diseases,
and maternal and prenatal conditions are responsible for 68% of the countrys
disease burden. Combating communicable diseases has been a major policy goal
of government for the last decade. Obstacles to reducing the disease burden in the
country have included a publicly funded health system providing low quality care,
nancial barriers to care for poor households, a strong propensity for people to
self-treat at times of illness or injury, and a lack of knowledge and use of preventive
practices among households.
Ghana started to reform its approach to battling disease in 2003 by embarking
on an ambitious policy to scale up health insurance and achieve universal cover-
age by 201520. Te initial health insurance platform consisted of three district
level health mutual organizations in 1999. Te number of mutuals spontaneously
expanded to 47 in 2000. In anticipation of universal health insurance, the mutuals
expanded to 159 in 2002, and to 258 in 2003. Government goals of 40% national
enrollment by 2004 were surpassed, and current enrollment is about 60%. Features
of Ghanas health insurance program relevant to combating major diseases include:
Local and community involvement to educate and enroll households in health
insurance.
Emphasis on prevention to combat communicable diseases, such as HIV/AIDS,
malaria, and tuberculosis.
Entitlement to cost-eective quality services locally, attuned to the epidemio-
logical challenges of the populace.
Cross-subsidizing premiums of the relatively rich with those of the poor and
relying on a value-added tax to subsidize premiums of the poorest households.
An estimate of the possible impact of health insurance on the disease burden
in Ghana enlists the following assumption: enrollment in health insurance draws
people into care at a modern health facility and away from self-treatment and moti-
vates them to use clinically proven interventions known to prevent or cure priority
communicable diseases. Based on data for 2007 the empirical analysis in chapter
4 shows that 88% of those enrolled in health insurance sought care at a modern
facility at times of illness, against 42% for those not enrolled. Te prevalence of
self-treatment declined by 29 percentage points, from 62% among nonenrollees to
33% for enrollees. Whether enrollees make best use of cost-eective preventive and
curative interventions that reduce the incidence of communicable diseases cannot
yet be quantied. It seems reasonable to assume, however, that the motives to do
so would be high, fueled by provider reimbursement policies that emphasize pre-
vention as a way of reducing more costly curative cases and by clients now entitled
access to demand services for premiums paid.
A crude estimate of the impact of Ghanas health insurance program on prior-
ity diseases might therefore be based on the following.
By 2007, 60% of the population is enrolled in health insurance.
For the 60% enrolled the propensity to use modern care and not self-treat is
about one-third higher than for the nonenrolled, or 18% (60% 30%).
Te four major communicable disease killers are responsible for 68% of Ghana
disability-adjusted life years.
Eective use of clinically proven interventions to combat the major killers by
18% of health insurance enrollees reduces the disease burden of the four major
killers by 12.2% (68% 18%).
About 1.46 million disability-adjusted life years would have been saved by
health insurance (12.2% 12 million) in 2007.
To put this estimate of 1.46 million disability-adjusted life years saved into per-
spective, imagine that a fully successful, vertical disease control program to elimi-
nate malaria was implemented in Ghana. Te result would be to reduce the overall
burden of disease by 8.2%, or 984,000 disability-adjusted life years. Eliminating
both malaria and tuberculosis through vertical disease control programs would
reduce the overall burden of disease by 11.1%, or 1.33 million disability-adjusted
life years. Such comparisons show that tackling todays disease control priorities
through health insurance has promise. Admittedly, these back-of-the-envelope
estimates are imprecise and require careful evaluation. Te eects on maternal
and child health would also need to be incorporated. Te point is to illustrate the
possible quantitative impact of health insurance relative to other common health
interventions.
China reducing copayments and coinsurance and building trust
Starting in 1979 China transformed the agricultural production system from col-
lective farming to the household responsibility system. Its Cooperative Medical
System, which provided community-based insurance to up to 90% of the rural
population, collapsed with the end of the communes. Later, it adopted a policy
of benign neglect and let market forces take over. Township health centers and
county hospitals began to rely on prots from drugs, laboratory tests, and surger-
ies for their incomes. Close to 60% of Chinas total health expenditures were soon
consumed by drugs, compared with about 20% in other low-income countries. At
the same time, the majority of villagers were not covered by any form of organized
health care nancing, subjecting them to major nancial risk.
Te 2003 National Health Survey found that 46% of rural Chinese who were
ill did not seek health care, and among them 40% cited cost as the main reason.
192 Chapter 10
Another 22% of those advised by physicians to be hospitalized refused to do so
because they could not aord it. Of those who became hospitalized, about 35%
discharged themselves against their doctors advice because of cost. When faced
with life-threatening conditions, many Chinese were either driven into poverty or
had to borrow money at usurious rates, reduce their nutritional intake, and discon-
tinue their childrens schooling to pay for care. Studies have found that medical
spending accounts for 30%40% of poverty.
To improve this situation, and to recapture the benets of the Cooperative
Medical System, the government launched the New Cooperative Medical Scheme
(NCMS) in 2003, with all rural county-level jurisdictions to be covered by 2008.
Under this policy the central government provided a subsidy of 10 RMB per
enrollee, to be matched by the local government, with additional premium con-
tributions from the villagers. In 2006 the subsidies were increased to 20 RMB
for both the central and local governments. By late 2006 more than 400 million
people were enrolled in the scheme, which was functioning in more than half of
Chinas rural counties.
But the NCMS has fallen short of expectations. Although outpatient and inpa-
tient use increased by 20%30%, enrollment was lower among poor households,
there was no impact on out-of-pocket spending among the poor, and increased
ownership of expensive equipment among central township health centers was not
associated with any impact on the cost per case (Wagsta and others 2007). Ana-
lyzing the design of the NCMS helps explain these impacts (or lack of them). In
particular, the NCMS budget was likely too small to reduce households out-of-
pocket spending, and copayments in the scheme were high, reecting large deduct-
ibles, low ceilings, and high coinsurance rates. Te high copayments were also
likely to have discouraged use of services among poor households, perhaps even
discouraging them from enrolling.
Dissatisfaction with several design elements of the NCMS prompted a modi-
ed approach. Working with the same rate of central and local government sub-
sidy, the NCMS approach has been recast as an experimental community-based
prepayment scheme targeted to rural populations the Rural Mutual Health Care
(RMHC) and applied in two low-income counties in the western region of the
country. Te RMHC provided a broader benet package that included a wide array
of outpatient services that enrollees wanted, in addition to more traditional hospi-
tal benets. It also featured cost-eective drugs from a reduced formulary at nego-
tiated and controlled prices. It reduced copayments and coinsurance. It selected
doctors competitively and put them on a salary. And it shifted more responsibility
and involvement to villagers.
During the 200406 piloting stage, the RMHC accomplished the following:
Increased enrollment rates from 60% in the rst year of the study to 90% in
the last year.
Increased monthly outpatient visits by 70%, with the greatest increases in vil-
lages (versus townships) and among those with chronic conditions.
Reduced self-medication by 42%.
Reduced those not seeking care for nancial reasons from 12.7% in the base-
line year to 3.3% in 2006; reduced those hospitalized and discharging them-
selves because of nancial diculties from 57% in the baseline year to 40% in
2006.
Increased use in villages most for those in the bottom and top income groups.
Did not increase more costly inpatient use.
Reduced catastrophic health spending.
Reduced those who perceive themselves to be in poor or fair health by 37%.
Encountered high levels of reported satisfaction, with 70%90% satised or
very satised.
Te RMHC now has the potential of contributing to the governments dis-
ease control priorities, like tuberculosis. At present, national tuberculosis strategies
include faster and more accurate diagnostic tests, xed-dose combinations to treat
cases, case monitoring with mobile phones, new inputs such as laboratory net-
works, and improved coordination of tuberculosis service delivery to combat mul-
tidrug resistance. But analysts warn these necessary inputs may not be sucient
to achieve optimal results. Patient delay in seeking diagnosis for tuberculosis in
China has been attributed to lack of ability to pay (actual and perceived), particu-
larly among rural residents (Tang and Squire 2005; Liu and others 2007; Zhang
and others 2007). Financial costs also enter as a barrier to compliance since many
patients decide to stop treatment due to costs. While the internationally recom-
mended strategy for tuberculosis control (DOTS) treatment is theoretically free in
China, providers often supplement DOTS with additional high cost drugs for
example, for liver functionbecause provider incomes are tied to prots, creat-
ing incentives to push high priced drugs the poor cannot always aord (Tang and
Squire 2005; Liu and others 2007).
Te RMHC can complement national tuberculosis strategies in three ways:
by reducing the nancial barrier for patients to seek initial diagnosis; by dimin-
ishing the propensity of individuals to self-treat or seek no treatment (up to 30%
of cases); and by combating perverse incentives among health care providers who
deter (or delay) referrals of suspected tuberculosis patients to the appropriate level
of care, to capture fees associated with their illness. By contracting village doctors
194 Chapter 10
and compensating them with a salary plus a bonus, based on performance mea-
surements, the RMHC increased use of established protocols of treatment for com-
mon diseases such as tuberculosis-related upper respiratory infection and diarrhea.
At the conclusion of the RMHC experiment in 2007 the government of Gui-
yang, one of the intervention sites, immediately replicated the scheme to cover
around 1.7 million rural Chinese. In 2008 Shaanxi province followed suit and
began to replicate the RMHC throughout the province, which has more than
30 million rural inhabitants. When the ndings of this social experiment were
presented to Chinas top ocials, the government revised its policies for national
health insurance, which emphasized coverage for hospital care. It decided that
national health insurance benet packages should cover both primary care and
hospitalizations. Bulk purchasing and central distribution of drugs would be estab-
lished. And community governance would be greatly encouraged, all substantially
the result of the pilot and analytical work summarized in chapter 8.
Peru rapid inroads into the health of the poor
Health insurance in Peru illustrates the art of the possible because it represents a
purposive, strategic intervention to target health insurance to the poorest house-
holds (chapter 6). It does so by consolidating two pro-poor schemes initiated in
2001 into a program called Seguro Integral de Salud (SIS) and scaling up enroll-
ment from 3.6 million in 2001 to more than 10 million today. SIS beneciaries are
mainly poor families. Its benet package focuses on maternal-child interventions,
and its membership is largely children.
By no means does SIS match anyones conventional idea of a health insurance
plan. Historically, patients have enrolled at the point of service when they seek
care, if they qualify through a means test, and remain enrolled for a year. Ali-
ates are tied to Ministry of Health service providers, and the same services they
consume through SIS have always been free or heavily subsidized. Yet SIS entitles
aliates to a clearly dened package of services, at no cost at the point of service,
and providers receive a fee for service covering variable costs when they can show
the service was provided. Te three major changes are the explicit targeting to the
poor (mainly poor mothers and their children), an entitlement to specic benets,
and a nancial benet that provides a pecuniary incentive to a public facility to
seek out SIS aliates and provide services to them.
Perhaps more than any other study in this book, SIS has made rapid inroads
on health conditions of the poor in the short timeframe of six years. Te program
can be credited with improvements that feed directly into Perus goal of achieving
the Millennium Development Goals because preventable communicable diseases
among low-income households represent such a large share of the countrys disease
burden. To this end, SIS is helping put the country on a fast track to:
Improve rates of immunization among children ages 1859 months. Te probabil-
ity of children ages 1859 months being immunized is 65% for low-income
health insurance members, compared with 50% for low-income nonmembers,
an absolute gain of 15 percentage points and a relative increase of 30%.
Improve rates of treatment for diarrhea among children under age 5. Te prob-
ability of children under age 5 treated for diarrhea is 50% for low-income
health insurance members, compared with 29% for low-income nonmembers,
an absolute gain of 21 percentage points and a relative increase of 72%.
Improve rates of treatment for acute respiratory infection. Te probability of low-
income children under age 5 with health insurance being formally treated
for acute respiratory infection is 73%, compared with 51% for those with-
out health insurance, an absolute gain of 22 percentage points and a relative
increase of 43%.
Colombia a big bang reform
For those seeking to scale up universal health insurance in record time and to
combat age-old inequalities in access to health and other social services, Colom-
bia provides inspiration. Prior to the introduction of an ambitious health reform
in the early 1990s, Colombia had a vertically integrated social insurance system
based on payroll taxes for formal workers basically covering only the employed
contributor. It had a tax-nanced system of public providers serving the poor
and the not so poor, the latter especially for hospital and surgical services. And
it had a private provider system for all those with the ability to pay or those dis-
satised with the services provided in the traditional social insurance and public
systems.
Following the reforms, the government pursued universal health insurance,
mobilizing the private sector to provide coverage for a publicly determined benets
package, injecting competition among public and private providers and insurers,
and designing and implementing an explicit basic benets package, with client
choice of provider. In the process the government established a solidarity fund to
channel resources from individuals whose payroll contributions were greater than
the premiums for themselves (the contributory regime) toward individuals whose
contributions are less (the subsidized regime). Tis equalization fund made sure
that payroll contributions based on income were transformed into risk-adjusted
premiums for all insured, rich and poor. Trough this integrated risk-pooling,
those with low risk of disease subsidize those with high risk, those with the ability
196 Chapter 10
to pay subsidize those without the ability to pay, and those in productive ages sub-
sidize the young and the elderly.
With 65% of the population below the national poverty line, between 1993
and 2003, coverage of households in the bottom income quintile rose almost eight-
fold, from 6.1% to 46.5%. Compare this with a near tripling of coverage among
the population as a whole, from 23.8% to 61.8%. By 2005, 70% of the total popu-
lation was covered, with close to 60% coverage of the bottom two quintiles. By
2007 enrollment was estimated to be about 90%. In just 16 years Colombia did
what took many European countries 100 years.
Tis reform was particularly eective in improving outcomes for the poor, their
dependents, and the self-employed, the groups least likely to be covered by insur-
ance prior to the reform. For poor citizens insurance coverage reduces the probabil-
ity by 21 percentage points that they will incur an out-of-pocket payment greater
than 10% of their income (after subsistence expenses are deducted) even though
they are 41% more likely to use ambulatory care during the previous 12 months.
Insurance fundamentally alters the care of children in poor families as well; relative
to the uninsured, if they have a cough or diarrhea, they are 17% and 23%, respec-
tively, more likely to be taken to a health care facility than an uninsured child.
Immunization coverage is higher for the insured; births are far more likely to take
place in a facility and to be attended by a doctor or skilled professional. Te pattern
of higher use and greater nancial protection prevails for the self-employed who
are now insured. As Colombia approaches 100% coverage of the population, these
impacts will be extended to those who remain excluded.
Costa Rica boosting coverage from 15% to 90%
Costa Rica provides another example of rapid expansion of health insurance and
inclusion of the poor (chapter 5). In only 39 years coverage rose from 15.4% in
1961 to about 90% in 2000. At rst, only industrial workers were covered. A rst
extension added their dependents. By 1975 agricultural workers were covered, and
by 1984 the self-employed were included. By 2000 coverage across income quin-
tiles was fairly even, while remaining shares of the population not covered were
relatively young, single, and educated that is, individuals least likely to pursue
health insurance due to good health.
Te question addressed almost accidentally because the authors did not
expect the result by the Costa Rica study is does it matter how you cover the
last mile? It is well known in immunization programs that the cost of getting
coverage above 80% rises astronomically, yet vaccines so reduce incidence of the
immunizable disease that herd immunity may be achieved even if the last 20% are
not vaccinated. Is it the same with insurance? In our 2006 data, 81% were covered
by the national insurance system, and this has been the case for almost 20 years. To
guarantee access to the 19% who were not aliated, Costa Rican law prevents the
insurance system from denying care to anyone and subsidizing those who cannot
pay, extending the umbrella of catastrophic protection to everyone. Te authors
nd that yes, indeed, the insured and uninsured seem to have similar access to
health care and that they are both spared catastrophic expenditures.
Yet there are some troubling dierences in behavior and health-related results.
For example, health insurance contributes to completed schedules of immuniza-
tions among children ages 18 and younger. Health insurance is also conducive to
more regular referred hospitalizations, rather than admissions through the emer-
gency room, with an accompanying reduction in hospital length of stay among the
insured. Health insurance substantially improves the care of diabetics insured
diabetics are far less likely than their uninsured counterparts to end up in the
emergency room or inpatient ward of a hospital, and they use fewer medications.
Generally, insurance coverage reduces the probability that conditions more appro-
priately treated in an outpatient setting result in a hospital visit. Te uninsured are
more likely to require intensive care and to die at the hospital. Imputed savings if
the average length of stay for the uninsured could be reduced to that of the insured
are estimated at about $8.5 million in 2006. Additional savings could come from
the positive impacts of health insurance on timely treatment of diabetics, lower use
of intensive care, and so on, raising the question of whether there might be health-
ier ways to cover the last 19%, even if the current approach adequately handles
their nancial protection and access risks.
Conclusion
We hope that because of this book policymakers interested in scaling up health
insurance in low- and middle-income countries will be more informed about the
likely impacts of health insurance on health outputs and outcomes, the important
methodological factors that can obscure the measured impacts of health insurance,
and the ways well designed health insurance might be harnessed to improve condi-
tions facing poorer households within countries.
In reviewing the literature and conducting the case studies here, one thing is
clear: health insurance is a complex subject. Its eects are conditional on initial design
features and the ecacy of implementing them. If those design elements include
sensible medical entitlements and cross-subsidies for the poor, health insurance will
likely have equity-improving nancial and health impacts. Te goals of a health
insurance reform should be clear and the design elements consistent with them.
198 Chapter 10
In addition, measuring the eects of health insurance requires high-quality
data as well as creative statistical modeling to assure robust empirical estimates.
Tis applies particularly to retrospective analysis, which uses available historical
data and is less demanding of time and resources than prospective analysis.
As mentioned earlier, European countries took 100 years and more to intro-
duce, scale up, and repeatedly revise health insurance to achieve equitable uni-
versal coverage. Te cases collected in this book suggest that low- and middle-
income countries have the capability to shorten this gestation period substantially.
With clear goals, designs compatible with the goals, and reasonable implementa-
tion, success can be achieved and generate tangible benets for the population.
But designs and implementation will have to be modied constantly to adjust to
changing circumstances, technology, demands, and behavior. As an insurance pro-
grams impacts on equity, cost eectiveness, nancial protection, provider behavior,
and patient behavior become better understood, other modications to the basic
design and incentives created by the insurance system will have to be made. Know-
ing impacts and estimating the eects of changes requires an ability to monitor
the performance of all elements of the system, including providers, patients, and
administrators. We have only looked at patients in this book; the evaluation agenda
is much broader. It is far better to build monitoring and evaluation systems into an
insurance reform from the start so reform itself generates knowledge that automati-
cally feeds back through institutional mechanisms to improve it.
References
Liu, X., R. Tomson, Y. Gong, F. Zhao, S.B. Squire, R. Tolhurst, X. Zhao, F. Yan, and S. Tang.
2007. How aordable are tuberculosis diagnosis and treatment in rural China? An analysis
from community and tuberculosis patient perspectives. Tropical Medicine and International
Health 12(12): 146471.
Tang, S., and S. B. Squire. 2005. What Lessons Can Be Drawn from Tuberculosis (TB) Control
in China in the 1990s? An Analysis from a Health System Perspective. Health Policy 72(1):
93104.
Wagsta, A., M. Lindelow, G. Jun, X. Ling, and Q. Juncheng. 2007. Extending Health Insur-
ance to the Rural Population: An Impact Evaluation of Chinas New Cooperative Medical
Scheme. Policy Research Paper 4150. World Bank, Washington, DC.
Wooldridge, J. 2001. Econometric Analysis of Cross Section and Panel Data. Boston, MA: MIT
Press.
Zhang, T., S. Tang, G. Jun, and M. Whitehead. 2007. Persistent problems of access to appropri-
ate, aordable TB services in rural China: experiences of dierent socio-economic groups.
BMC Public Health 7: 19.
199
Editors and Authors
An email address is included for the corresponding editor or chapter
author for each group of authors.
Editors
Maria-Luisa Escobar is lead health economist and health systems pro-
gram leader at the World Bank Institute. On leave from the World Bank
as a senior fellow at the Brookings Institution, she worked on the Global
Health Initiative from 2006 through 2008, when research for this book
began. During the past 20 years at the Inter-American Development Bank
and the World Bank, she has worked on health nancing reform across
Latin America, and she was one of a small group of designers and imple-
menters of the Colombia reforms in the 1990s, led by Juan Luis Londoo.
Charles C. Grin is senior advisor in the Europe and Central Asia
Regional Oce of the World Bank. Over the past 25 years he has
worked on health nancing policy in low- and middle-income countries
around the world. Most recently he was director for human develop-
ment at the World Bank in South Asia, then in Eastern Europe, before a
leave of absence to join the Brookings Institution as a senior fellow from
2006 through 2008 (cgrin@worldbank.org).
R. Paul Shaw is a former World Bank lead economist and previously
an economist in the Canadian government, the International Labour
200 Editors and Authors
Organization, the United Nations Population Fund, and academia. At the World
Bank he created the Health Sector Reform Flagship Program, which has supported
more than 20,000 policymakers around the world in applying rigorous analytical
techniques to support improved health policies and health system performance. He
currently advises the Bill & Melinda Gates Foundation on health economics.
Authors
Chapter 2Literature Review
Ursula Giedion is a health economist with 20 years experience in Colombia,
Latin America, and other developing countries. Her areas of expertise include the
study of health sector reform and health sector nancing strategies, the design
and implementation of health prioritization policies, and the implementation of
econometric health-related impact evaluation studies. In recent years she has car-
ried out consultancies for the Inter-American Development Bank, the Economic
Commission for Latin America and the Caribbean, the World Bank, the Interna-
tional Development Research Centre, and the Brookings Institution (ugiedion@
gmail.com).
Beatriz Yadira Daz is a PhD student at the University of Essex and holds an MS in
economics. She has experience using quantitative methods for the design, analysis,
and evaluation of the impact of public policy. She has 11 years professional experi-
ence at several of Colombias government agencies, including the National Plan-
ning Department, the National Statistics Institution, and the Ministry of Finance.
As part of research teams for projects in developing countries, she has worked as a
consultant for the World Bank, the Pan-American Health Organization, and the
Brookings Institution.
Chapter 3Namibia
Emily Gustafsson-Wright is a research fellow at the Brookings Institution and the
Amsterdam Institute for International Development. Her work involves analysis
of health risk and low-cost health insurance programs in Africa. Her professional
experience includes consulting for the World Banks Human Development Net-
work, where she focused on education and child labor, conditional cash transfer
programs, and risk and vulnerability in Latin America. She holds a PhD in eco-
nomics from the University of Amsterdam/Te Tinbergen Institute and an MS
in applied economics and nance from the University of California (egustafsson-
wright@brookings.edu).
Wendy Janssens, a research fellow at the Amsterdam Institute of International
Development and the Vrije Universiteit Amsterdam, holds a PhD in economics
from the University of Amsterdam/Te Tinbergen Institute. Her research interests
include health risk, nancing and HIV/AIDS, early child development, and social
capital. Her regional experience includes Africa, Southeast Asia, and the Caribbean
(wjanssens@feweb.vu.nl).
Jacques van der Gaag is co-founder and co-director of the Amsterdam Institute
for International Development and a distinguished visiting fellow at the Brookings
Institution. Before serving as dean of the faculty of economics and business at the
University of Amsterdam, he held various positions at the World Bank in Washing-
ton, DC, including chief economist of the Human Development Network. He has
published widely in refereed journals and books and served on the editorial boards
of the Journal of Human Resources, the Journal of Health Economics, and the World
Bank Economic Review (jvandergaag@brookings.edu).
Chapter 4Ghana
Slavea Chankova is an associate at Abt Associates, Inc., specializing in global
health program evaluation and research, focusing on health systems and reproduc-
tive, maternal, and child health. She has worked on assignments for the United
Nations Childrens Fund, the United Nations Development Programme, and
Analysis Group (slavea_chankova@abtassoc.com).
Chris Atim is a senior health economist working with the World Banks Health
Systems Strengthening program in Senegal and is executive director of the African
Health Economics and Policy Association. He has worked as a senior health econo-
mist for the HLSP Institute and Abt Associates, Inc.
Laurel Hatt is a health economist and senior associate at Abt Associates, Inc.,
specializing in health nancing, maternal and child health, and survey research.
She has consulted for the Johns Hopkins School of Public Health and the World
Bank.
Chapter 5Costa Rica
James Cercone, founder and president of Sanigest International, is a mathemati-
cal economist from the University of Michigan with 20 years experience in pub-
lic policy and public sector reform in Latin America, the Caribbean, and Eastern
Europe (jcercone@sanigest.com).
toile Pinder, consultant at Sanigest Internacional, holds a masters degree in the
evaluative clinical sciences from Dartmouth College, working on public sector
reform and modernization in Latin America, the Caribbean, the United States,
Eastern Europe, and Southeast Asia. Ms. Pinder developed the technical and regu-
latory framework for the legislation of the Bahamas National Health Insurance Act.
Jos Pacheco Jimnez is director of the consulting division of Sanigest Interna-
cional. He holds an MA in development economics from the Institute of Social
Studies (Te Hague) and in impact evaluation methods from the Massachusetts
Institute of Technologys Abdul Latif Jameel Poverty Action Lab. He has worked
on many aspects of health insurance in more than 20 countries in Latin America,
the Caribbean, Eastern Europe, and Southeast Asia.
Rodrigo Briceo holds a masters degree in economics from the National Univer-
sity of Costa Rica. He specializes in nancial and social sector data analysis of
resource allocation and primary health care.
Chapter 6Peru
Ricardo Bitrn is founder and president of Bitrn y Asociados in Santiago, Chile, a
leading research and consulting group specializing in health economics. He teaches
at the University of Chile and the World Bank Institutes Course on Health Sec-
tor Reform and Sustainable Financing. He graduated from the University of Chile
with a degree in industrial engineering and earned both a PhD in health economics
and an MBA in nance from Boston University (ricardo.bitran@bitran.cl).
Rodrigo Muoz is an engineer specializing in health economics, social metrics,
and evaluation. He has worked in Africa, Asia, Latin America, and the Caribbean.
His expertise includes economic assessments of social and health programs, impact
evaluations, surveys, and software design.
Lorena Prieto was a consultant at Bitran & Associates for six years and is an assis-
tant professor at ESAN University in Lima, Peru. A health economist, she focuses
on cost-eectiveness of health services and evaluation of reform programs, risk
adjustment, and health policy impacts on households.
Chapter 7Indonesia
Facundo Cuevas is an economist in the East Asia and Pacic Human Develop-
ment Department of the World Bank, working on social protection, poverty, and
migration. He has conducted research in health and crime, including work for an
article in the upcoming World Bank publication on crime and violence in Central
America.
Susan W. Parker is professor of economics in the division of economics at the
Center for Research and Teaching in Economics in Mexico City and an adjunct
professor at the Rand Corporation. She has extensive experience in the design,
implementation, and evaluation of programs, including both randomized con-
trolled trials and nonexperimental methods, with numerous publications in the
eld (susan.parker@cide.edu).
Chapter 8China
Winnie Yip is reader in economics for health policy in the Department of Public
Health, University of Oxford. She holds a PhD in economics from the Massachu-
setts Institute of Technology and was associate professor of international health
policy and economics at the School of Public Health, Harvard University. She leads
several projects on large-scale health system interventions and evaluations in China
(winnie.yip@dphpc.ox.ac.uk).
William Hsiao is K.T. Li professor of economics at the School of Public Health,
Harvard University, and he holds a PhD in economics from Harvard University.
He has published widely on U.S. health policy, comparative health systems, and
health policy in developing countries.
Chapter 9Colombia
Ursula Giedion is a health economist with 20 years experience on Colombia, other
countries in Latin America, and other developing countries. Her areas of expertise
include the study of health sector reform and health sector nancing strategies,
design and implementation of health prioritization policies, and implementation of
econometric health-related impact evaluation studies. In recent years she has car-
ried out consultancies for the Inter-American Development Bank, the Economic
Commission for Latin America and the Caribbean, the World Bank, the Interna-
tional Development Research Centre, and the Brookings Institution.
Carmen Elisa Flrez is an economist/demographer working on Colombia and
other Latin American and Caribbean countries during the last 28 years. She has
focused on social economics, particularly the impact of health system reform on
access and use of health services and health status. She has carried out national
consultancies for government institutions and international consultancies for the
Economic Commission for Latin America and the Caribbean, the International
Development Research Centre, and the World Bank.
Beatriz Yadira Daz is a PhD student at the University of Essex and holds an MS in
economics. She has experience using quantitative methods for the design, analysis,
and evaluation of the impact of public policy. She has 11 years professional experi-
ence at several of Colombias government agencies, including the National Plan-
ning Department, the National Statistics Institution, and the Ministry of Finance.
As part of research teams for projects in developing countries, she has worked as a
consultant for the World Bank, the Pan-American Health Organization, and the
Brookings Institution (yadiradiaz76@gmail.com).
Eduardo Alfonso is an economist engaged in research and the use of evidence to
improve public policy. His areas of expertise include the design and implementa-
tion of public policy impact evaluations, analysis of health insurance market fail-
ures and regulation, and estimation of risk-adjusted health insurance premiums.
In recent years he has worked for the Health Ministry and the National Planning
Department in Colombia and carried out consultancies for the World Bank.
Renata Pardo, an economist, has worked as an advisor for the Ministry of Social
Protection and for the Health Division at the National Planning Department on
the Colombian health system. She has also participated in research projects assess-
ing the impact of health insurance on nancial protection and the determinants of
catastrophic health expenditures. Currently, she works at the National Planning
Department on issues related to poverty measures.
Manuela Villar is a public health specialist with experience working in Latin
America and Africa. Her areas of expertise include health systems research, health
policy design and implementation, and monitoring and evaluation. In recent years
she has worked for the World Bank on health reform and health policy in Latin
America.
205
A
access and use
impact of health insurance on, 17
increased, 181
access to care
barriers to, 148, 163, 181
RMHC increasing probability of, 148
Act for the Workers Protection, in Costa Rica, 91
acute illness, reported prevalence in Namibia, 38
acute respiratory infection, in Peru, 112, 114, 195
adult weight loss. See weight loss
adverse selection
data reecting, 144146
in enrollment in NHIS, 68
health insurance plagued by, 185
as an NHIS concern, 78
into private insurance schemes, 41
aliation, with health insurance system in Costa Rica, 9293
African countries, epidemics, 33
age structure, of insured and uninsured in Costa Rica, 95
age-based exemptions to NHIS, 69
AIDS, accompanied by weight loss, 46
analytical methods, for NHIS study, 83, 85
Index
206 Index
anemia, in Ghana, 59
antiretroviral therapy, in Namibia, 34, 46
Asabri, for police, military, and dependents,
123
Askes, social health insurance program, 123
Askeskin health insurance program, 9, 123
asset score, lower for uninsured households,
48
B
balancing properties, propensity score match-
ing, 146, 147
barriers to access, 148, 163, 181
baseline sample, 81
beneciary identication system, 157
benets package(s)
established by the government in Colom-
bia, 158
of NHIS as nancially unsustainable, 79
of RMHC, 140
in subsidized regime, 170
bidirectional causality, 24
biomedical survey, with HIV test data, 33
Brazil, impact of supplemental private health
insurance on self-perceived health status, 24
C
Caja Costarricense del Seguro Social (the
Caja), 89, 9091, 95
care-seeking, in Ghana, 71, 78
catastrophic expenditure
incidence of, 168169
priority for NCMS in China, 137
reduction in, 151
RMHC reducing, 148
vulnerability increasing for poorest indi-
viduals, 168
catastrophic nancial loss, 185
catastrophic health care, 21
catastrophic insurance, in Costa Rica, 102
Catholic Diocese of Sunyani, 60
causal chain, poorly specied, 186187
children
with acute respiratory infections, 112
benets to insured, 164
with diarrhea, 112
DPT vaccinations of in Peru, 106
gaining insurance increasing use, 125
health insurance impacts on in Indonesia,
132134
immunizations, 112, 164
insurance impacts on, 130
insured dierences from uninsured, 124
premium exemption for NHIS more like
for, 69
China
case study, 10
health insurance reducing rates of cata-
strophic spending, 185
impact of rural mutual health care,
137153
national tuberculosis strategies, 193
observed impacts of health insurance, 183
reducing copayments and coinsurance
while building trust, 191194
social experiment understating impacts,
187
studies on, 16
summary data on, 7
uneven evidence in studies, 189
chronic conditions, in Costa Rica, 100
chronic disease, in Namibia, 3738
chronic illness, in Ghana, 68
claims payments, delays in Ghana, 63
cohabitation, with a Caja member, 96
coinsurance, policies on, 187
Index 207
Colombia
case study, 10
challenges remaining, 175
comparison of insured poor against unin-
sured poor, 11
gross national income, 157
health insurance reform, 155175
health sector reform, 157159
health system ows of funds and alia-
tions, 159
impacts of health insurance, 183
out-of-pocket expenditures increasing for
costly illnesses, 185
poor gaining benets from insurance
coverage, 181
scaling up universal health insurance, 195
self-employed in contributory scheme, 181
studies on, 16
summary data on, 7
Colombian Demographic and Health Sur-
veys (DHS), 156
communicable diseases
in Ghana, 59, 190
in Namibia, 34
community governance, used by RMHC,
142
community-based health insurance schemes
in Ghana, 60
positive eect on health status, 23
community-based prepayment scheme
experimental, 192
social experiment of, 137
contributory regime in Colombia, 164,
166167
benets package, 158
copayments, 171
for employed and self-employed, 155
growth in coverage of, 161162
impact on access barriers for self-
employed, 167
impact on health-related outcome vari-
ables, 156
impact on incidence of catastrophic
expenditure, 172
increased access to and use of health
services, 173
mitigating eect in face of catastrophic
spending, 170171, 172
Cooperative Medical System, collapse of in
China, 10, 138, 191
copayments
absence of, 90
policies on, 187
required under contributory regime, 171
Costa Rica
case study of, 89
eect of social health insurance on child
and infant mortality, 22
nancial protection to everyone, 185
guaranteeing access to uninsured, 197
health system, 9093
impact of health insurance, 89103, 182
reducing per capita spending on health,
184
studies on, 16
summary data on, 7
universal health coverage, 89, 196197
countries
selecting for case studies, 610, 178
summary data on, 7
cream-skimming, 37
cross country studies, 16
curative health care
NHIS causing an increase in use of, 58, 78
probability of seeking, 115116
cut-out method, 158
208 Index
D
data, limited and poor quality, 188
data and methodology, for Costa Rica study, 95
data collection
for NHIS study, 8283
for RMHC experiment, 144, 145146
death of a family member
coping with, 4647
as a health shock, 43
deductibles, policies on, 187
delivery
attended by a skilled health professional,
114
organizational features of, 188
Demographic and Health Surveys (DHS)
in Colombia, 156
in Peru, 106
design elements, limiting nancial access,
187188
diabetics, in Costa Rica, 100, 184, 186
Diamond Health Services, 50
diarrheal diseases
in Ghana, 59
improving rates of in Peru, 195
probability of being formally treated for,
114
SIS children with, 112
disability-adjusted life years saved, 191
disease burden
in Costa Rica, 96
in Ghana, 190
distribution eects, of RMHC, 148
distributional impact, of health insurance,
18, 2021
district mutual schemes, organizing sensitiza-
tion exercises, 86
DPT vaccinations, of children in Peru, 106
drug prices, reduction in China, 151
E
earned income, not aected by terminally ill
in Ghana, 47
EBAIS (Equipos Bsicos de Atencin Integral
en Salud), 90
economic consequences, of health shocks, 45
economic variables, description of, 4244
elective insurees, 107
eligibility, under the Caja, 9192
emergency services, free to insured and
uninsured, 92
employed, more likely to be insured in
Namibia, 36
ENAHO (National Household Survey), 106,
115
endline household survey, 64
endline sample, 81
endogeneity
constraints on, 8586
controlling for, 5, 115
dealing with, 15
as a research issue, 185186
tackling by analyzing specic subpopula-
tions, 24
enrollment
equitable not achieved for NHIS in
Ghana, 58, 78
in NHIS, 6568
in RMHC, 144148
ENSA (National Health Survey), 95
ENSA 2006 database, estimated models,
104105
entidad promotora de salud. See EPS
EPS (health-promoting entity), selecting, 158
EQ-5D dimensions, 184
equalization fund, 159, 195
Equipos Bsicos de Atencin Integral en
Salud. See EBAIS
Index 209
EsSalud (social security-nanced sector),
106107, 109112
European countries, achieving equitable
universal coverage, 198
evidence, evaluating robustness of available,
1328
exemptions, in Ghana, 60
F
fee exemption policies, facing various dif-
culties, 60
nancial access, design elements limiting,
187188
nancial barriers, RMHC reducing, 150
nancial contribution, as a percentage of
income, 159
nancial loss, protecting from the risk of, 12
nancial protection
greater for low health-related expenses,
174
health insurance increasing, 26
impact of health insurance on, 1922, 171
NHIS positive impact on, 79
positive impact of NHIS on, 73
provided by health insurance, 167171, 172
of RMHC, 148
nancial sustainability
of NHIS threatened in Ghana, 58
of subsidy expansion in Colombia, 175
nancing and benet packages, of RMHC, 140
formal sector employees, paying NHIS
premiums, 78
formal workers, aliated with the Caja, 92
formally employed, contributory regime
improving access and use, 167
fraudulent claims, under NHIS, 63
free health care for all, as goal in Ghana, 59
full nancial coverage, by SIS, 108
G
Ghana
case study of, 8
health system, 59
National Health Insurance Scheme
(NHIS), 5886
saving disability-adjusted life years, 189191
summary data on, 7
uninsured patients with higher out-of-
pocket spending, 184
women with health insurance, 181
Ghana Health Service, levels of, 59
government employees, paying NHIS premi-
ums, 6869
government health facilities, in Namibia, 38
government health fund (PSEMAS), in
Namibia, 3536
government-subsidized universal health
insurance, in Colombia, 155
Greater Windhoek, in Namibia, 33
gross domestic product (GDP)
Costa Rica, 89
share devoted to health, 2
gross national income (GNI), 2
Colombia, 157
Indonesia, 122
Namibia, 34
Peru, 106
Sub-Saharan Africa, 34
growth monitoring visits, in Peru, 114, 115
H
HAART (highly active antiretroviral treat-
ment), 50
health and the health care sector, in Namibia,
3439
health care
nancing in Ghana, 5962
210 Index
infrastructure investment, 80
NHIS positive impact on use of modern, 78
health care providers, sending claims in
Ghana, 62
health care questionnaire, for NHIS study, 83
health care services
decision related to use of, 160
equal access to, 89
health care spending, soaring in China, 139
health care use, 3739
health expenditures
consumed by drugs in China, 191
inequality in, 3739
in Namibia, 33
small proportion of individuals using
large share of total, 140
health nancing reform, in the 1990s in
Ghana, 60
health insurance
benetting poorer households, 184
as a complex subject, 197
coverage in Colombia, 160162
coverage in Costa Rica, 9192
coverage in Peru, 109
eect varying across population groups,
160
eects on use by gender and location, 127
evaluating the impact of, 14
factors diminishing measured impact of,
185189
heterogeneity of, 25
impact in Indonesia, 128130, 131
impact in the countries examined,
180185
impact on access and use, 1719
impact on disease burden in Ghana,
190191
impact on nancial protection, 1922
impact on health-related outcomes in rich
countries, 5
impact on health status, 24, 2224, 26,
160
impact on measures of health status, 184
impact on out-of-pocket expenditures for
health, 4
impacts of, 99102
impacts on access and use, 181
impacts on adults in Indonesia, 125130
impacts on children in Indonesia, 130,
132134
impacts on indicators associated with
health status, 127128
improving access, 11, 25
improving care of diabetics in Costa Rica,
197
inconsistent exposure to over time, 187
increasing use of inpatient and outpatient
care, 134
inequities surrounding coverage, 39
limits of concept of use, 18
measuring eects of, 198
mitigating eects of, 40
mitigating out-of-pocket expenditures, 26
no signicant impact on out-of-pocket
health expenditures, 102
as not a homogeneous product, 179
not aecting access and use of health
services, 99101
not aecting many standard self-reported
measures, 135
observed impacts on selected indicators,
182183
positive association with being fully
immunized, 112
positive eect on medical care, 13
proportion covered by in Indonesia, 123
Index 211
purpose of, 14, 159160
reducing incidence of catastrophic nan-
cial loss, 185
reducing incidence of catastrophic pay-
ments, 26
reducing out-of-pocket payments, 2122,
184
reducing price of health care, 17
reducing probability of health care spend-
ing, 134
use and health status for children,
132133
use by expenditure distribution, 128
Health Insurance Fund, 50
health insurance industry, in Namibia, 35
health insurance program, in Ghana, 190
health insurance reform, in Colombia,
155175
health insurance schemes
community-based in Ghana, 60
heterogeneity across dierent, 17
heterogeneity of, 18, 25, 179
impact of in Zambia, 21
mandatory in Costa Rica, 91
multiple in Jordan, 25
Nkoranza district, 60
positive eect on health status of commu-
nity-based, 23
types of in Ghana, 61
health insurance system, in Costa Rica, 89
health outcomes, 152, 160, 173
Health Providing Institutions providers, 107
health services
barriers to access more related to supply,
163
demand indicators, 110112
health insurance encouraging overuse of,
1819
sectors in Ghana, 59
use for acute illness or injury and hospi-
talization, 38
health shocks
consequences for uninsured and insured, 55
economic consequences, 45, 51
experienced by those without insurance, 180
inequitable impact of, 50
reducing income-earning capacity, 44
relationship with economic variables, 41
strong correlation between being
un insured and negative consequences, 52
types of, 42
for uninsured and insured households, 43
uninsured coping with, 3948
health social security, in Peru, 107
health spending
in Namibia, 34
skewed distribution of, 168, 169
health status
dierent measures of, 24
impact of health insurance on, 24, 22,
101, 171, 173, 184
indicators showing little relationship with
insurance coverage, 130, 131, 135
inequality in, 3739
insurance eects by gender and location,
129
of insured and uninsured, 97
RMHC impact on, 148150
selecting measures of, 187
health systems, burden of dysfunctional, 1
healthcare services, coverage by health insur-
ance, 82
Heckman selection model, 56
heterogeneity
across dierent health insurance schemes,
17
212 Index
across dierent population groups and
insurance schemes, 15
of health insurance schemes, 25, 179
of NCMS implementation, 20
high-dedectible catastrophic insurance, 142
highly active antiretroviral treatment
(HAART), 50
HIV infection
biological markers for, 56
consequences of, 46
higher among uninsured, 50
incubation period, 46
more prevalent among employed, 50
as not a health shock per se, 46
rates among working-age adults, 49
results not showing substantial eects
related to, 48
HIV prevalence rates, in Namibia, 34, 4850
HIV/AIDS, in Namibia, 8, 33, 34
HIV/AIDS treatment, high medical costs, 51
HIV-positive individuals, more likely to
refuse HIV test, 42
homogeneous good, health insurance as not,
160
hospital discharge database, statistics from,
96
hospitalization(s)
average out-of-pocket spending decreased,
73
coping with consequences of, 4748
decrease in the likelihood of, 7273
health insurance conducive to more regu-
lar referred, 197
impact of RMHC, 150
likelihood avoidable less for insured, 101
more likely for individuals covered by
NHIS, 77
reduced expenditures under NHIS, 77
Household Responsibility System, 138
household survey data, covering Greater
Windhoek, 35
household/individual questionnaire, for
RMHC, 144
households without health insurance, coping
strategies, 44, 48
hypertension, in Ghana, 59
I
illness/injury
health care for, 7172
positive eects of NHIS on health care
for, 7677
immigrants, 94, 96, 99
immunizations. See also vaccinations
improving rates of in Peru, 195
insurance increasing coverage of, 173174
insured population with higher coverage
among children, 112
more likely completed for insured chil-
dren, 164
impact, of health insurance on access and
use, 17
imputed savings, of health insurance in Costa
Rica, 197
incentives, under NHIS to provide more than
necessary, 63
Income and Expenditure Survey (IES 2004),
95, 101102
income and health nance indicators, for
country groupings, 3
income loss, dealing with, 40
income source, for NHIS, 7980
Index 213
indicators
for country cases, 7
descriptive data for health service
demand, 110112
health insurance impacts on, 182183
of health status, 130, 131, 135
income and health nance for select coun-
try groupings, 3
indigents, NHIS means test for, 62
indirect members, in Costa Rica, 92
individuals, entering labor market for rst
time, 130
Indonesia
case study, 910
health insurance distributional eects,
181, 184
health system, 123124
impact of health insurance, 122, 183
summary data on, 7
Indonesian Family Life Survey (IFLS), 122,
126
inequality, of wealth in Namibia, 34
informal care
dened, 175
reduced prevalence of, 181
informal provider, proportion seeking care
from, 71
injury, reported prevalence of, 38
inpatient care, insured with higher rates of,
124
INS (National Insurance Institute), public
insurance monopoly in Costa Rica, 90, 93
Institute of Medicine, of the National Acad-
emies, 5
instrumental variables, selecting, 188189
insurance. See health insurance
insurance enrollment, in Greater Windhoek,
3539
insurance scheme, impact depending on
specics of, 25
insurance status
changes in in Indonesia, 123124
determinants of, 99, 100
unobserved characteristics aecting, 41
insured
appearing to have worse health status, 99
descriptive statistics in Indonesia, 124125
experiencing shorter hospital stays, 101
having higher burden of disease, 96
higher rates of outpatient care, 124
more likely to report chronic illness, acute
illness, and hospitalization, 37
paying more out of pocket, 39
by state regime, 90
use of government health facilities, 38
insured children, dierences from uninsured,
124
insured diabetics, self-perceived health status,
101
insured mothers, less likely to have babies
with low birthweight, 101
insured pregnant women, clear benet for, 163
integral health insurance, 107108
integrated risk-pooling, 195196
intensive care unit, uninsured more likely to
end up in, 98
international conferences, on social health
insurance, 1
international insurance companies, in Costa
Rica, 93
intervention, designing to address the prob-
lem, 189
interventions, independent eects of aspects
of, 152
214 Index
J
Jamsostek insurance, for private sector work-
ers, 123
Jordan, multiple health insurance schemes,
25
K
knowledge gaps, 24, 27
L
Law 100, in Colombia, 157, 158
life expectancy
in Colombia, 157
in Costa Rica, 89
in Indonesia, 122
in Peru, 106
Living Standards Measurement Survey 2003
(LSMS), 156
local circumstances, familiarity with, 27
loss of weight
consequences of, 44
in the previous 12 months, 42
as proxy for advanced state of AIDS, 46, 48
low cost insurance schemes, demand for, 51
low-income countries, lack of evidence on
eectiveness of health insurance, 14
M
malaria
in African countries, 33
in Ghana, 59
in Namibia, 34
mammograms, for fewer uninsured woman,
96
mammography result, dierent from descrip-
tive analysis, 101
mandatory health insurance scheme, in Costa
Rica, 91
mandatory health plan, in Colombia, 157
maternal health care, eects of NHIS, 7378
medical aid funds, in Namibia, 35, 36, 41
medical care, positive eect of health insur-
ance on, 13
medical expenditures, of HIV-infected
households, 46
medical impoverishment
NCMS not addressing a major cause of,
151
reduction in, 151
RMHC reducing, 148
medical insurance, Greater Windhoek house-
holds with, 37
medical savings account, 142
member indentication cards, delays in issu-
ing for NHIS, 63
membership growth of NHIS in Ghana, 62, 63
meningococcal meningitis, in Ghana, 59
methodological recommendations, 2627
migrants, aliated with the Caja, 92
Ministry of Health, in Ghana, 59
Ministry of health in Costa Rica, 90
MINSA, in Peru, 107, 108
modern providers, proportion seeking care
from under NHIS, 71
monitoring and evaluation systems, building
into insurance reform, 198
mortality rate, as too blunt a measure of
health, 23
mutual health organizations, in Ghana, 61, 86
N
Namibia
coping mechanisms of families with a
health problem, 185
dierences between insured and un insured,
11
Index 215
health insurance industry, 78
impacts of health insurance, 182
low-cost health insurance to protect poor,
3355
summary data on, 7
Namibia Health Plan, 36
Namibia Medical Care, 36
National Health Insurance Act of 2003, in
Ghana, 8, 61
national health insurance benet packages,
revised in China, 194
National Health Insurance Fund (NHIF)
in Ghana, 61
income, 79
reinsurance aspect of, 80
National Health Insurance Scheme (NHIS),
5886
benets out of reach for many poor
people, 62
benets package, 6162
challenges for the future, 7980
determinants of enrollment in, 6569
dierences in health care use and payment
comparing insured and uninsured, 7678
early successes and challenges, 6263
eects of implementation on health care
use and spending, 6976
establishment of, 6162
evaluation of, 6465
policy implications, 7879
politically driven process behind, 80
poorest wealth quintile less likely to be
exempt, 69
premium exemptions for enrollment,
6869
as a tax-funded social health insurance
system, 79
weak performance incentives for provid-
ers, 63
National Health Survey, in Costa Rica,
9394, 95
National Health Survey (2003), in China,
139, 191192
National Household Survey (ENAHO), 106
National Insurance Institute (INS), 90, 93
National Social Secubenefrity System, man-
dating social security schemes, 123
network of providers, belonging to the Caja, 90
New Cooperative Medical Scheme (NCMS),
137
benet packages revised, 152
in China, 10
enrollment lower among poor households,
192
evaluating impact of, 1920
impact on access and use among the poor,
18
launch of, 192
modality reducing out-of-pocket pay-
ments and incidence of catastrophic
payments, 20
modied approach, 192
premiums charged, 139
reducing incidence of catastrophic spend-
ing among poorest, 20, 21
in western and central regions of China,
142143
NHIS, benets package and subsidies as
unsustainable, 79
Nkoranza district, 64, 81
Nkoranza District Health Insurance Scheme,
60
no treatment, reduced prevalence of, 181
noncommunicable diseases, in Namibia, 34
noncontributory regime, 90
216 Index
nonenrolled, in RMHC, 144148
nonenrollment in the NHIS, due to unaf-
fordable premiums, 68
nonfood consumption, lower for uninsured
households with a hospitalization shock, 48
not-for-prot missionary health facilities, in
Namibia, 35
O
Onso district, 64, 81
Okambilimbili (buttery) project, 50
Okambilimbili Survey, 35
older people, suering from chronic pain, 150
open funds, in Namibia, 35
ordinary least squares regression, 41
out-of-pocket expenditures
average per capita annual, 39
by Costa Ricans, 102
does health insurance aecting, 101102
health insurance reducing, 184
impact of health insurance on, 4
improved nancial protection against in
Ghana, 58
inequities in, 39
reduced for maternal care under NHIS, 76
reduction under NHIS, 7172
outpatient care
insurance increasing probability of, 126
use of in Indonesia, 125
outpatient health expenditures, not protect-
ing chronic patients with, 151
overall impact, of health insurance on access
and use, 17
P
pap-smear exam, probability of receiving, 113
partial subsidy system, in Colombia, 175
payroll fees, by insurance scheme, 93
pensioners, aliated with the Caja, 92
Peru
case study, 9
insured required to use public system, 11
Integral Health Insurance (Seguro Inte-
gral de Salud, or SIS). See SIS (Seguro
Integral de Salud)
observed impacts of health insurance, 182
poor gaining substantial benets from
insurance coverage, 181
probability of spending on health services,
184
summary data on, 7
targeting health insurance to poorest
households, 194
Peruvian health sector, 106108
PharmAccess Foundation (PharmAccess), 50
plan obligatorio de salud. See mandatory
health plan
policy implications, 189197
poor citizens, insurance coverage improving
outcomes, 196
poor households
benetting from health insurance, 181, 184
coping strategies buering shock, 40
poorer households, health insurance benet-
ting, 181
population, not covered by any type of insur-
ance, 109
poverty
attributed to medical expenditures in
China, 139
medical spending accounting for in
China, 192
predictors, of enrollment in NHIS, 6667
pregnancy
in Ghana, 59
Index 217
NHIS enrollment related to, 68
premium and registration fees, for NHIS in
Ghana, 62
premiums
cross-subsidizing in Ghana, 190
established by the government in Colom-
bia, 158
for NHIS, 69, 86
prepayments, 21
pre-post, treatment-control study design,
143144
pre-post evaluation design, of Ghana case
study, 8
pre-post evaluation study, 64
primary care
in addition to hospitalization, 150
of the Caja, 90
primary care providers, 2
private health insurance, in Namibia, 5152
private health sector, in Namibia, 35
private providers, in Ghana, 59
private sector, in Peru, 106
private subsector, in Costa Rica, 91
prot-seeking entities, turning health provid-
ers into, 138
propensity score matching methods, limita-
tions on, 76
pro-poor design features, of health insurance, 179
pro-poor orientation, 6
prospective analysis, 198
provider access, varying, 188
provider performance, spillover eects on, 189
provision and provider payment, organization
of, 140142
PSEMAS (government health fund), in
Namibia, 3536
public delivery system, unied in Costa Rica, 91
public health care sector, in Namibia, 33
public health care system, poor and unin-
sured unprotected in Namibia, 51
public health facilities, at user fees in
Namibia, 35
public providers, retaining access to in
Colombia, 156
public system option, impact of a free or low
cost, 11
purchaser-provider split, 90
Q
quality, of health care services, 101, 175
quality evaluation protocol, 15
quasi-experimental methods, 2627
R
Rawlings, President Jerry, in Ghana, 60
rgimen contributivo. See contributory regime
rgimen subsidiado. See subsidized regime
regional conferences, on social health insur-
ance, 1
regular insurees, 107
reported health status, dierences in, 38
Republic of Namibia Okambilimbili Survey.
See Okambilimbili Survey
research community, a few countries receiv-
ing substantial attention, 27
research issues, in the case studies, 185189
respiratory tract infections, in Ghana, 59
retired insurees, 107
retrospective analysis, 198
revenue for NHIS, remaining constant over
time, 7980
rich countries
health nance systems, 2
impact of health insurance, 5
separating nancing from provision of
care, 2
218 Index
risk equalization fund (HEALTH-IS-VITAL),
supported by PharmAccess, 50
risk pooling, growth of, 6061
risk-pooled fund, 140
RMHC Fund Oce, selecting village doc-
tors, 140141
robustness, of the evidence base, 1516
robustness of empirical analysis, 179
rural areas, insurance reducing probability of
household health care spending, 126
rural health care
in China, 138139
new national policy for in China, 137
Rural Mutual Health Care (RMHC), 192
accomplishments of, 193
benet package compared to New Coop-
erative Medical Scheme, 143
community-based prepayment scheme, 10
complementing national tuberculosis
strategies, 193194
described, 139143
design of, 140142
enrollment rates, 144
evaluating impact of, 137
impacts of, 148150
implemented in three towns, 143
improving access to basic health care and
health status, 150
replication of, 151152, 194
rural population in Colombia
improvements for, 164
insurance improving access to and use of
care, 174
S
saliva-based HIV test, in Namibia, 42
sample characteristics, of NHIS study, 8485
sample selection, for NHIS study, 8182
secondary care, of the Caja, 90
SEG/SMI programs, eects on health care
results, 112115
Seguro Escolar Gratuito (SEG), 107
Seguro Materno Infantil (SMI), 107
self-employed
aliated with the Caja, 92
benetting from contributory regime, 174
in Colombia, 196
compared to formally employed in
Colombia, 167
contributory regime improving access and
use, 164, 166
facing worse basic access conditions, 167
self-selecting out of the insurance plan, 99
self-exclusion, in Costa Rica, 96
self-medication, RMHC reducing probability
of, 148
self-perceived health status, in Brazil, 24
self-reported health status, 24, 101, 184
self-selection
of the sick into insurance pools, 185186
of sicker individuals into health insur-
ance, 24
self-treatment, reduced prevalence of, 181
semiparametric methods, applying, 155, 156
sensitization exercises, 86
service industry, uninsured rate in Namibia, 36
service quality, improving, 175
services, access and use in Colombia,
162167
simultaneity, eecting inuence interpreta-
tions, 41
SIS (Seguro Integral de Salud), 106, 107, 194
as agency under MINSA, 108
beneciaries as mainly poor families, 109
coverage, 108, 116117, 118
eect increasing with income, 113
Index 219
eects reaching the less poor, 118
formation of, 107108
immunization rates for children, 112
impact on health care results using
ENAHO data, 115117
improvements on health conditions of the
poor, 194
increase in child growth monitoring visits,
114
increase in probability of a pap smear, 113
increase in receiving formal treatment for
acute respiratory infection, 114
as a limited program, 108
as a maternal and child health insurance
program, 108
plans, 108
population coverage, 109
positive eects of, 117
probability of formal treatment for diar-
rhea, 114
scaling up enrollment, 194
summary eects of, 117
skin disease and ulcers, in Ghana, 59
social experiment(s)
of a community-based prepayment
scheme, 137
informing policy actions, 153
suering from limitations, 27
social health insurance, in formerly commu-
nist countries, 23
social insurance system, vertically integrated, 157
social security, paying for hospitalization, 48
Social Security and National Insurance Trust
Fund (SSNIT), 61
Social Security Institute (Instituto de Seguros
Sociales or ISS), 162, 175
social security-nanced sector (EsSalud), in
Peru, 106
sociodemographic statistics, for insured and
uninsured in Costa Rica, 9394, 95
socioeconomic categories, enrollment across, 36
socioeconomic evaluation form (Ficha de
Evaluacin Socio Econmica), 108
solidarity fund, receiving excess contribu-
tions, 159
spillover eects, 15, 189
SSNIT contributors, exempt from NHIS
premiums, 62
state-covered members, aliation dened at
demand, 92
study design, of RMHC, 143144
study limitations, 8586, 152
Sub-Saharan Africa
Ghana characteristic of, 59
health spending, 34
subsidies, of NHIS, 79
subsidized health insurance products, in
Greater Windhoek, 33, 50
subsidized insurees, 107
subsidized regime in Colombia
benets package, 158, 170
data evaluating, 156
eligibility for, 175
impact on health status, 171, 173
increased access to and use of health
services, 173
increasing access for the poor, 162164
for the poor in Colombia, 155
propensity score matching, 169170
supply of care, in Ghana, 80
T
Taiwan
National Health Insurance, 2223
study on, 16
220 Index
tax-nanced public sector (the Ministry of
Health, or MINSA), in Peru, 106
tertiary care, of the Caja, 9091
timeline, inadequate for eects to manifest, 187
township health centers, purchasing drugs in
bulk, 141142
treatment group, controlling, 187
tuberculosis
in African countries, 33
controlling in China, 193194
in Namibia, 34
U
unearned income in Namibia
components of, 44
high due to hospitalization, 4748
increasing, 40
uneven evidence, far-ranging debate resulting
from, 189
uninsured
admitted to hospitals through emergency
room, 96
aected by out-of-pocket health expendi-
tures, 39, 40
as baseline for determining value added,
180
coping with health shocks in Namibia,
3948
in Costa Rica, 9399, 102103
covering death-related (medical) expendi-
tures, 47
descriptive statistics in Indonesia, 124125
diabetics, 100
disadvantaged in Costa Rica, 103
economic consequences of health shocks, 51
economic outcome and coping variables,
5354
experiencing longer stays in the hospital, 98
facing higher out-of-pocket expenditures
from health shocks, 44
foregoing care, 38
health shock eects, 44
higher chance of inpatient death, 99
hospitalization resulting in high medical
costs for, 47
inequalities for bottom quintiles, 39
lower asset score, 48
meaning of, 180
more likely to be immigrants, 94
negative consequences in rich countries, 5
paying more in relative terms, 39
reducing consumption of nonfood items, 47
tending to be relatively poor families, 180
use of government health facilities, 38
working for a small company, 94
Universal Coverage Act, in Costa Rica, 8
universal health care coverage, in Costa Rica,
89
universal health insurance, in Colombia, 157
universal insurance coverage, in rural areas,
138
universal social security coverage, in Indone-
sia, 123
unnecessary drugs, waste caused by, 142
unnecessary treatments, waste caused by, 142
urban-rural dierences, 164, 165
user fee exemptions, existing remaining in
limbo, 80
user fees, in Ghana, 60
utilization statistics, of insured and unin-
sured, 96, 97
V
vaccinations. See also immunizations, health
insurance improving likelihood of complet-
ing full series, 101
Index 221
variables
explaining aliation with the Caja, 95
weak instrumental, 188189
Vietnam, health insurance program, 21, 23
village clinics, in China, 140
village doctors
as private practitioners, 138
RMHC screening and regulating, 141
selecting, 140141
village management committee, checking
village clinics, 142
villagers, monitoring RMHC operations, 151
voluntary enrollment, for NCMS in China,
137
voluntary members, of insurance program in
Costa Rica, 92
W
wealth, dierences within Namibias popula-
tion, 34
wealth indexes, constructing for NHIS study,
83
wealth quintile, enrollment in NHIS increas-
ing with, 78
weight loss
consequences of, 44
in previous 12 months, 48
as proxy for advanced state of AIDS, 46, 48
World Bank Health VIII project in China
impact of, 19
mixed evidence of an impact on health
status, 23
reducing out-of-pocket payments among
the poorest, 20
World Health Organization (WHO)
advocating social health insurance, 2
estimates on nancial catastrophe for
expensive emergency care, 1
Z
Zambia, impact of health insurance schemes,
21
Zhangjai Town
in China, 141
The Brookings Institution is a private nonprofit organization
devoted to research, education, and publication on important
issues of domestic and foreign policy. Its principal purpose is to bring the highest
quality independent research and analysis to bear on current and emerging policy
problems. The Institution was founded on December 8, 1927, to merge the activities
of the Institute for Government Research, founded in 1916, the Institute of
Economics, founded in 1922, and the Robert Brookings Graduate School of Eco -
nomics and Government, founded in 1924. Interpretations or conclusions in
Brookings publications should be understood to be solely those of the authors.
Board of Trustees
Honorary Trustees
John L. Thornton
Chair
Strobe Talbott
President
Robert J. Abernethy
Liaquat Ahamed
Dominic Barton
Alan R. Batkin
Richard C. Blum
Geoffrey T. Boisi
Crandall Bowles
Paul L. Cejas
Abby Joseph Cohen
Arthur B. Culvahouse Jr.
Alan M. Dachs
Steven A. Denning
Vishakha N. Desai
Paul Desmarais Jr.
Kenneth M. Duberstein
Cheryl Cohen Effron
Alfred B. Engelberg
Bart Friedman
Ann M. Fudge
Brian L. Greenspun
Glenn Hutchins
Shirley Ann Jackson
Benjamin R. Jacobs
Kenneth M. Jacobs
Suzanne Nora Johnson
Richard Kauffman
Richard A. Kimball Jr.
Nemir Kirdar
Klaus Kleinfeld
Philip H. Knight
Thomas C. Ramey
Edgar Rios
David M. Rubenstein
James S. Rubin
Haim Saban
Sheryl K. Sandberg
Victoria P. Sant
Leonard D. Schaeffer
David F. Swensen
Lynn Thoman
Larry D. Thompson
Michael L. Tipsord
Andrew H. Tisch
Antoine W. van Agtmael
John W. Wilhelm
Tracy R. Wolstencroft
Daniel B. Zwirn
Leonard Abramson
Elizabeth E. Bailey
Zo Baird
Rex J. Bates
Louis W. Cabot
James W. Cicconi
William T. Coleman Jr.
Kenneth W. Dam
D. Ronald Daniel
Mario Draghi
Lawrence K. Fish
Cyrus F. Freidheim Jr.
David Friend
Henry Louis Gates Jr.
Jeffrey W. Greenberg
Lee H. Hamilton
William A. Haseltine
Teresa Heinz
Joel Z. Hyatt
James A. Johnson
Ann Dibble Jordan
Vernon E. Jordan Jr.
Herbert M. Kaplan
Breene M. Kerr
Marie L. Knowles
David O. Maxwell
Donald F. McHenry
Arjay Miller
Mario M. Morino
Maconda Brown OConnor
William A. Owens
Frank H. Pearl
Samuel Pisar
John Edward Porter
Charles W. Robinson
James D. Robinson III
Warren B. Rudman
B. Francis Saul II
Ralph S. Saul
Michael P. Schulhof
John C. Whitehead
Stephen M. Wolf
James D. Wolfensohn
Daniel Yergin
Ezra K. Zilkha
Trustees AG:trustees TR 8/16/10 4:18 PM Page 1
The Impact of
Health Insurance in
Low- and Middle-Income
Countries
Maria-Luisa Escobar
Charles C. Grifn
R. Paul Shaw
EDITORS
O
ver the past twenty years, many low- and middle-income countries have
experimented with health insurance options. While their plans have
varied widely in scale and ambition, their goals are the same: to make
health services more affordable through the use of public subsidies while also
moving care providers partially or fully into competitive markets.
Colombia embarked in 1993 on a fteen-year effort to cover its entire population
with insurance, in combination with greater freedom to choose among providers.
A decade later Mexico followed suit with a program tailored to its federal system.
Several African nations have introduced new programs in the past decade, and
many are testing options for reform. For the past twenty years, Eastern Europe
has been shifting from government-run care to insurance-based competitive
systems, and both China and India have experimental programs to expand
coverage. These nations are betting that insurance-based health care nancing
can increase the accessibility of services, increase providers productivity, and
change the populations health care use patterns, mirroring the development of
health systems in most OECD countries.
Until now, however, we have known little about the actual effects of these dramatic
policy changes. Understanding the impact of health insurancebased care is
key to the public policy debate of whether to extend insurance to low-income
populationsand if so, how to do itor to serve them through other means.
Using recent household data, this book presents evidence of the impact of
insurance programs in China, Colombia, Costa Rica, Ghana, Indonesia, Namibia,
and Peru. The contributors also discuss potential design improvements that
could increase impact. They provide innovative insights on improving the
evaluation of health insurance reforms and on building a robust knowledge
base to guide policy as other countries tackle the health insurance challenge.
Maria-Luisa Escobar is lead health economist and health systems program
leader at the World Bank Institute. Charles C. Grifn is senior adviser in the
European and Central Asia regional ofce of the World Bank. R. Paul Shaw,
a former World Bank lead economist, advises the Bill and Melinda Gates
Foundation on health economics.
Washington, D.C.
www.brookings.edu
Cover by Rich Pottern Design
E
s
c
o
b
a
r
/

G
r
i
f
n
/
S
h
a
w
T
H
E

I
M
P
A
C
T

O
F

H
E
A
L
T
H

I
N
S
U
R
A
N
C
E

I
N

L
O
W
-

A
N
D

M
I
D
D
L
E
-
I
N
C
O
M
E

C
O
U
N
T
R
I
E
S

Theimpactofhealthinsurance Fulltext

Uploaded by

Copyright:

Available Formats

Theimpactofhealthinsurance Fulltext

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Theimpactofhealthinsurance Fulltext

Uploaded by

Copyright:

Available Formats

The Impact of

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.