Measures of Disease Frequency

November 7, 2024

ER 600.01: Epidemiology and Research Methods

Mucho Mizinduko, MD, MPH

1
 Learning Objectives
1. Define study population, and explain the difference between
fixed and dynamic populations
2. Explain what is meant by the term ”at risk”
3. Define and explain the differences between proportion, ratio,
and rate
4. Define and calculate prevalence
5. Explain what is meant by “person-years of observation” and
calculate the number of person-years of observation from
epidemiologic data
6. Define and calculate cumulative incidence and incidence rate
7. Explain the mathematical relationship between prevalence and
incidence rate (P=IRxD)
2
 Goals of Epidemiology
• Determine the extent of disease in a population

• Identify patterns and trends of disease occurrence

• Identify the causes of disease

o Does exposure “X” increase/decrease the risk of outcome “Y”?
o Remember: Usually expressed as a hypothesis

• Evaluate effectiveness of interventions (prevention efforts and

treatments)

3
 Measuring Disease Frequency
Four fundamental components to measure the frequency of
disease in a population:

1. Source Population (Which Group of People)

2. Cases of Disease (Numerator)

3. Size of Source Population (Denominator)

4. Time (Implicit or Explicit)

4
 Measuring Disease Frequency (cont.)
Four fundamental components to measure the frequency of
disease in a population:

1. Source Population (Which Group of People)

2. Cases of Disease (Numerator)

3. Size of Source Population (Denominator)

4. Time (Implicit or Explicit)

5
 Source Population
How do we define and select a population?
Common Usage:
Group of people living in a specific geographic area

Epidemiologic Usage:
Group of people with a common characteristic
Can be almost any characteristic:
• Workers at Urafiki Textile Mill
• Women at MNH who had a live birth in 2016
• Diners at Tessa’s restaurant last weekend
• People listed in a cancer registry
• Patients who use medical clinic at Mloganzila
Hospital
• Survivors of Ebola outbreak in West Africa
6
 Types of Populations
FIXED DYNAMIC
A population whose membership A population whose membership
is defined on the basis of some is defined by being in a state or
event; membership is permanent condition; membership is
transient
• Survivors of 2014 – 2016 WA • Residents of Mwanza City
Ebola Outbreak • People with NHIF, AAR, Jubelee
• Women enrolled in the Black Health Insurances
Women’s Health Study

Why it matters: Type of population impacts which type of measure

of disease frequency (and study design) is most appropriate

Test: Once in this population, can you ever leave it? Will the
event/state that made you eligible for the population ever change?
7
 Types of Populations:
Example From the Literature

Surveillance and outbreak reports

The proof of the pudding is in the eating: an outbreak of emetic
syndrome after a kindergarten excursion, Berlin, Germany,
December 2007
G O Kamga Wambo, F Burckhardt, C Frank, P Hiller, H Wichmann-Schauer, I Zuschneid, J Hentschke, T Hitzbleck, M Contzen, M
Suckau, K Stark

An outbreak of food poisoning (emetic syndrome) occurred in three kindergartens (A, B and C) in Berlin, Germany, on 3
December 2007 after an excursion during which food was served. We conducted a retrospective cohort study among the
kindergarten children and personnel who participated in the trip. The overall attack rate among the 155 participants was
30%. It was 31% among the 137 children (aged two to six years) and 17% among adults (n=18). The consumption of rice
pudding was significantly associated with disease. Among those who ate rice pudding, the attack rate was 36%, compared
with 0% for non-eaters (relative risk: infinite, p<0.001, aetiological fraction: 100%), but differed significantly between
kindergartens A (43%), B (61%) and C (3%), probably because groups were served from different pots. Bacillus cereus sensu
stricto was identified from one vomit sample. The clinical and epidemiological characteristics suggest that B. cereus emetic
toxin (cereulide) was the causative agent, although it could not be proven in the single vomit isolate. Inadequate food
handling most probably led to the outbreak. Single-portion ready-to-eat rice pudding was recommended for subsequent
excursions and no further cases of food poisoning occurred.

8
 Types of Populations:
Example From the Literature
PHARMACOEPIDEMIOLOGY AND DRUG SAFETY 2008; 17: 144-150
Published online 18 October 2007 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/pds.1515

ORIGINAL REPORT

Risk of breast cancer in relation to antibiotic use†

Hani M. Tamim PhD1,2, James A. Hanley PhD3, Ali H Hajeer PhD2, Jean-Francois Boivin MD, ScD3,4
and Jean-Paul Collet MD, PhD5

METHODS

We carried out a population-based nested case-control study using data compiled from the
cancer registry of the Saskatchewan Cancer Agency and the outpatient prescriptions drug and health
insurance registration databases of Saskatchewan Health. The source population was the XXXX
cohort defined by registration with Saskatchewan Health from 1 January 1981 to 31 December 2000 and
eligible for prescription drug benefits. It included all people aged between 5 and 82.5 years, living in
Saskatchewan and registered with the Saskatchewan Health at least 5 years earlier and eligible for the
outpatient prescription drug benefits, and with no prior history of cancer (except for non-melanoma skin
cancer).

9
 Types of Populations:
Examples for Practice
What type of population is this?

A study of the incidence of headaches among students

while they are enrolled in ER 600.01?

A study of the incidence of study related stress among

students who were enrolled in ER 600.01 on Nov 7, 2024?

10
 Types of Populations:
Special Considerations
Healthcare catchment populations:
• People who would use a specific healthcare facility when ill

• Very difficult to define

11
 Measuring Disease Frequency (cont.)
Four fundamental components to measure the frequency of
disease in a population:

1. Source Population (Which Group of People)

2. Cases of Disease (Numerator)

3. Size of Source Population (Denominator)

4. Time (Implicit or Explicit)

12
 Cases of Disease

Cases of disease = numerator of all measures of frequency

• Disease = ANY health outcome

• Disease, defect, injury

• You MUST have a clear definition

o Can be based on one criterion, or many criteria
o May be straightforward, or very difficult
o May choose to identify sub-populations: men only, children <5
years, etc.

13
 Cases of Disease (cont.)

Methods of disease ascertainment:

Clinical information Other recorded information

• Signs and symptoms • Disease registries
• Physical examination • Surveillance programs
• Diagnostic tests • Death certificates
• Pathological investigation • Insurance records
• Medical records

Self-report *Definitions may change over

• Some reliable, some not time

14
 Practice Question

Community A has 50 new cases of heart disease in 2022

Community B has 150 new cases of heart disease from 2022 to

2023

15
 Measuring Disease Frequency (cont.)
Four fundamental components to measure the frequency of
disease in a population:

1. Source Population (Which Group of People)

2. Cases of Disease (Numerator)

3. Size of Source Population (Denominator)

4. Time (Implicit or Explicit)

16
 Size of Source Population

Size of source population = denominator for all measures

• Size is based on source population that you identified

o Full population
o Sample of the population

• Necessary for comparison of disease across populations

o Cannot compare number of cases alone

17
 Measuring Disease Frequency (cont.)
Four fundamental components to measure the frequency of
disease in a population:

1. Source Population (Which Group of People)

2. Cases of Disease (Numerator)

3. Size of Source Population (Denominator)

4. Time (Implicit or Explicit)

18
 Time

Time is necessary for all measures of frequency

• Time = Time period when disease occurred, or was resolved

• Can be a single point in time

o At study enrollment, at birth, a particular date

• Can be an interval of time

o The course of follow-up, from birth to 10 years of age, while
living in the city of Boston

• Can be explicit or implicit

19
 Measuring Disease Frequency:
Types of Parameters
RATIO
Division of one number by another
Numbers do not need to be related

PROPORTION
Division of two related numbers
Numbers must be related: the numerator is SUBSET OF the denominator
Often expressed as a percentage

RATE
Division of one number by another in which TIME is intrinsic part of denominator
Most frequently misunderstood parameter

20
 Measuring Disease Frequency:
Types of Measures

3 Types of Measures:

1. Prevalence
2. Cumulative Incidence
3. Incidence Rate

ALL MEASURES SHARE 3 CHARACTERISTICS:

• Number of cases (numerator)
Measured in
• Size of source population (denominator) source population

• Measure of time
21
 Prevalence
• Measures presence of EXISTING CASES of disease in a
population during a specified time period

• Involves BEING IN A STATE

o Someone has disease at the time you are measuring it
– They may be cured from this disease in the future
– They may have had the disease for awhile, or recently developed the
disease

• Denominator includes TOTAL SOURCE POPULATION

o Healthy
o Sick
o “At risk”
o Not “at risk”

22
 Prevalence
POINT PREVALENCE = single point in time
• On Dec 31
• At enrollment
• At birth
PERIOD PREVALENCE = longer time period
• Jan 1 to Dec 31
• 20 – 25 years of age

x x xx x x x x xx x x x x x x x x xxx x
x 2001 2002 2003 2004 2005 2006 2007
TIME

23
 Prevalence
• Point Prevalence
Prevalence = # EXISTING cases at time point
# in TOTAL POPULATION at time point

• Period Prevalence
Prevalence = # EXISTING cases during time period
# in TOTAL POPULATION during time period

UNITS = # of people NO UNITS

# of people
 Point Prevalence

What is the prevalence of arthritis in City A on January 1, 2022?

• City A has 7,000 people with arthritis on January 1, 2022

• The total population of City A is 70,000 on that same day

Prevalence = # EXISTING cases at time point

# in TOTAL POPULATION at time point

25
 Period Prevalence
What is the prevalence of arthritis in City A during 2022?

• City A has 14,000 people with arthritis during

2022.
• Prevalence
The total= population of City
# EXISTING cases A time
during is 75,000
period during
# in TOTAL POPULATION during time period
2022.
 Special Types of Prevalence

AUTOPSY “RATE:” # people with a particular finding at autopsy

# of people who received an autopsy

BIRTH DEFECT “RATE:” # infants born with birth defects

10,000 births

*Note that these are NOT true rates…

27
 Incidence
• Measures occurrence of NEW CASES of disease in a population
during a specified time period
o Usually first time occurrence

• Involves TRANSITION from one state to another

o Incidence of disease: non-diseased  diseased
o Incidence of death: alive  dead
o Incidence of cure: diseased  non-diseased

• Denominator only includes SOURCE POPULATION “AT RISK”

o Exclude people who already have the disease (not ANY disease)
o Exclude people who are immune
– Vaccination = immunity from measles
– Male sex = immunity from uterine cancer

28
 Prevalence vs. Incidence
Prevalence Incidence
What Existing Disease New Disease
(1) Assess burden of disease (1) Etiologic research
•Administration and planning
•Allocate resources for (2) Evaluation of prevention
interventions and treatment
interventions
Why (2) When incidence is difficult
to measure
• Birth defects
• Conditions with uncertain
onset (MS, depression)
• Cross-sectional studies

29
 Prevalence or Incidence?
Examples for Practice
Period Prevalence? Or Point Prevalence? Or Incidence?

Students in ER 600.01 who wore eyeglasses today.

Students in ER 600.01 who wear eyeglasses at any time during

the semester.

Students in ER 600.01 who start to wear eyeglasses in the next

three months.

30
 Incidence

CUMULATIVE INCIDENCE
INCIDENCE RATE

SIMILARITY: NEW cases of disease

DIFFERENCE: How TIME is incorporated

Cumulative Incidence: Time is described in words that go along

with the number
Incidence Rate: Time is an intrinsic part of the denominator
31
 Cumulative Incidence
• Measures occurrence of NEW CASES of disease in a
population during a specified time period
• Involves TRANSITION from one state to another
o Incidence of disease: non-diseased  diseased
o Incidence of death: alive  dead
o Incidence of cure: diseased  non-diseased

• Denominator only includes SOURCE POPULATION “AT RISK”

• Critical Assumption: All people in the source population have
been followed for the entire specified time period
o No loss to follow-up (LTF)
o No competing risks
32
 Cumulative Incidence (cont.)
Proportion of a population at risk that develops disease
over a specified time period

Cumulative = # NEW cases during time period

Incidence # in TOTAL POPULATION AT RISK during time period

UNITS = # of people NO UNITS

# of people

33
 Cumulative Incidence (cont.)

What is the cumulative incidence of breast cancer?

• 1,000 women without breast cancer were followed for 5 years

• 30 of these women developed breast cancer

Cumulative = # NEW cases during time period

Incidence # in TOTAL POPULATION AT RISK during time period

34
Cumulative Incidence: Practice Question

What is the cumulative incidence of coronary heart disease?

• You conduct a study with 1394 participants whom you follow

for 10 years in order to measure the cumulative incidence of
CHD in this population.
• At the time of enrollment, 42 participants had a medical
history of CHD.
• During the 10 year follow-up period, 265 participants were
diagnosed with CHD.

35
 Special Types of Cumulative Incidence
CRUDE MORTALITY “RATE:” # Deaths due to any cause during specified time
100,000 population

INFANT MORTALITY “RATE:” # Infant deaths due to any cause during specified time
1,000 population

MORBIDITY “RATE:” # Illnesses due to any cause during specified time

100 population

LIVE BIRTH “RATE:” # Live births during specified time

1,000 pop’n OR childbearing women

*Note that these are NOT true rates…

36
 Special Types of Cumulative Incidence
CASE FATALITY “RATE:” # Deaths due to disease X during specified time period
# People with disease X

# New cases of disease during specified time period

ATTACK “RATE:” # Population at risk at beginning
* Infectious of time period
disease
outbreaks “At risk” = exposed to the infectious agent

*Note that these are NOT true rates…

37
 Limitation of Cumulative Incidence

Suppose we want to estimate

the incidence of tuberculosis in
Boston during the time interval
1/1/22 - 2/28/22

Cumulative incidence assumes that the group of people being

observed has been followed for the entire time period of
observation.
38
 Limitation of Cumulative Incidence (cont.)

Difficult to know the details of

these changes,we
Suppose must
want make the
to estimate
assumption that all members of
the incidence of tuberculosis in
the population are followed for
Boston during the time interval
the full time period and that they
1/1/07
remain - 2/31/07
at risk.

In reality, some people will move in and out of Boston, and

some will die due to competing risks.
39
 Limitation of Cumulative Incidence (cont.)

X X
Suppose we want to estimate
the incidence of tuberculosis in
Boston during the time interval
1/1/07 - 2/28/07

CI is not a perfect measure in a (1) fixed population that loses

members over time, or in a (2) dynamic population.
40
 Limitation of Cumulative Incidence (cont.)
2 drugs evaluated for pain relief (marked as “X”)
X o
X o
Drug A o
X o
(n=10) X
X X

X o
X o
Drug B o
X o
(n=10) X
X X
1 2 3 4 5 6 7 8 9 10
Hours

41
 Incidence Rate
• Measures SPEED of occurrence of NEW CASES of disease in a
population
• Involves TRANSITION from one state to another
o Incidence of disease: non-diseased  diseased
o Incidence of death: alive  dead
o Incidence of cure: diseased  non-diseased

• Denominator only includes SOURCE POPULATION “AT RISK”

• Does NOT assume the group of people being observed has
been followed for the entire time period of observation

42
 Incidence Rate (cont.)
Number of people who develop disease per number of people
in a population at risk*time at risk (person-time)

Incidence = # NEW cases during time period

Rate Total person-time of observation in population at risk

UNITS = # of people 1/time, or

# of people*time cases/person-years

43
 Person-Time

Amount of time each study participant is under observation

• Accrued only among persons at risk

• Accrued only while being followed (follow-up time)
• Accrual ends when:
o Participant is no longer part of population at risk
– Develops disease under study (if first-time occurrence)
– Dies
– Is lost to follow-up (moves away, quits study)
– No longer meets eligibility criteria (becomes too old)
o Study ends

44
 Person-Time (cont.)

Example of How Person-Time is Accrued

ta rt d
S En
Jan Jan Jan Jan
2000 2001 2002 2003 Total P-Y
Subject 1 ---------------- 1 person-year
D
Subject 2 ---------------- 1 person-year
X
Subject 3 ------------------------------------------------ 3 person-years
Subject 4 --------------------------------D 2 person-years
Subject 5 ----------- 0 person-years
X

D = disease of interest
X = death 45
 Person-Time (cont.)
• Several ways to accrue person-time
o 100 people followed 1 year each = 100 person years
o 10 people followed 10 years each = 100 person years

• Equivalent rates
o 2.5 / 100 person-years
o 25 / 1,000 person-years
Most convenient
o 250/10,000 person-years

x =
46
 Incidence Rate: Practice Question
What is the incidence rate of hypertension in a cohort of
U.S. Black Women?
• Followed 30,330 women from 1997 – 2001  104,574 PY
• 2,314 cases of hypertension were ascertained during follow-up

Incidence = # NEW cases during time period

Rate Total person-time of observation in population at risk

Cozier et al., 2006 47

 Comparison of CI and IR
CI IR
Strengths • Easily calculated & understood • Does not assume complete
• Can be used to estimate follow-up
individual risk: “1 in 8 women • Can take into account when a
will develop breast cancer” disease occurs

Limitations • Assumes complete follow-up • Not as easily calculated

for all individuals • Can be difficult to ascertain
• Does not take into account person-time
WHEN disease occurs • Interpretation is not intuitive
• Cannot account for multiple
disease occurrences
Appropriate • Fixed populations with short • Dynamic populations
Use follow-up, or no loss to follow- • Fixed populations with long
up follow-up, or substantial loss to
follow-up

48
 Relationship between CI and IR
• CI and IR are 2 approaches to measuring the same thing
• Makes sense that they are related to one another:
CI = 1 – e [- Σ( IR * T) ]

When CI is considered
small, this simplifies to…

CI ≈ Σ (IRi * Ti)
When is CI considered small? CI ≤ 10%
49
 Relationship between CI and IR (cont.)
• Why should I care about this relationship?
o If you know the IR (+ T), you can estimate the CI
o If you know the CI (+ T), you can estimate the IR

• Assumes that IR is constant within each time interval, and

across sub-groups
o In reality, IRs change over time, and are different for different
sub-groups (e.g.; men vs. women, young vs. old)
o Close enough – formula allows us to make comparisons, and to
translate IR into CI, which can be easier to interpret/discuss

50
 Relationship between CI and IR (cont.)
CI ≈ Σ (IR * T)
What is the 5-year risk (CI) of death among a population of 1,000 people in
whom the incidence rate of death is 4/1,000 PY?

What is the 5-year risk (CI) of death among a population of 1,000 people in
whom the incidence rate of death is 1/1,000 PY in adults aged 45-49, and
5/1,000 PY in adults aged 50-54?

51
 Relationship between P and IR
To be a prevalent case, an individual has to:
o Have been an incident case at some point
o Still have the disease
Prevalence is affected by
incidence & duration

Incidence Cure

Prevalence Time

Death
52
 Relationship between P and IR (cont.)

P ≈ IR * D

• D = Average duration of disease; this is the time from

diagnosis to recovery or death

• Assumes
o Population is in “steady state” where inflow=outflow
o Prevalence is low (e.g., <10%)

53
 Relationship between P and IR (cont.)
How can you use this formula to estimate the average
duration of disease?
D ≈ P / IR
IR lung cancer = 45.9/100,000 PY
P lung cancer = 23/100,000 persons

54
 Relationship between P and IR (cont.)

Primary prevention efforts to reduce exposure to smoking

Incidence Prevalence
1. Increase 1. Increase
2. Decrease 2. Decrease
3. No identifiable effect 3. No identifiable effect
4. Don’t know 4. Don’t know

55
Relationship between P and IR (cont.)

Use of HAART medications by HIV-infected persons?

Incidence Prevalence
1. Increase 1. Increase
2. Decrease 2. Decrease
3. No identifiable effect 3. No identifiable effect
4. Don’t know 4. Don’t know

56
 Putting It All Together
P CI IR
Existing cases New cases New cases
Total population Population at risk Person-time at risk
Type Proportion Proportion Rate
Units None None 1/time
Range 0-1 0-1 0-
Numerator Existing cases New cases New cases
Denominator Total population Population at risk Person-time at risk
Specify time? Yes Yes Yes
Population Fixed or Dynamic Fixed with few exits Fixed or Dynamic;
Exits OK
Uses Disease burden, Causes, impact of Causes, impact of
incidence difficult interventions interventions

P ≈ IR * D
CI ≈ IR * T
57
PRACTICE

58
Which Type of Disease Frequency: P, CI, IR?
1. Percentage of Kinondoni residents who contracted Cholera during
the course of an epidemic (January 1 – 20, 2016)

2. Percentage of potential army recruits rejected at intake exam

because of poor vision

3. Number of common colds diagnosed in 1,000 person-years within

an elementary school

4. Percent of live-born infants with a cardiac malformation among

100,000 live-births

5. Percent of persons with malaria whose disease is resistant to

treatment with chloroquine
59
Ten factory workers were followed for 90 days to assess the development of
asthmatic-like symptoms after acute exposure to toluene-di-iso-cyanate as
the result of an industrial accident. Consider this table of data obtained
from the exposed factory workers:
Subject Follow-up (Days) Status at End of FUP
A 90 No symptoms
B 10 Symptoms
C 90 No symptoms
D 90 No symptoms
E 30 Symptoms
F 60 Symptoms
G 90 No symptoms
H 90 No symptoms
I 15 Symptoms
J 25 Symptoms

What type of population is described?

60
Ten factory workers were followed for 90 days to assess the development of
asthmatic-like symptoms after acute exposure to toluene-di-iso-cyanate as
the result of an industrial accident. Consider this table of data obtained
from the exposed factory workers:
Subject Follow-up (Days) Status at End of FUP
A 90 No symptoms
B 10 Symptoms
C 90 No symptoms
D 90 No symptoms
E 30 Symptoms
F 60 Symptoms
G 90 No symptoms
H 90 No symptoms
I 15 Symptoms
J 25 Symptoms

What is the 90-day cumulative incidence of asthmatic-like symptoms?

61
Ten factory workers were followed for 90 days to assess the development of
asthmatic-like symptoms after acute exposure to toluene-di-iso-cyanate as
the result of an industrial accident. Consider this table of data obtained
from the exposed factory workers:
Subject Follow-up (Days) Status at End of FUP
A 90 No symptoms
B 10 Symptoms
C 90 No symptoms
D 90 No symptoms
E 30 Symptoms
F 60 Symptoms
G 90 No symptoms
H 90 No symptoms
I 15 Symptoms
J 25 Symptoms

What is the incidence rate of asthmatic-like symptoms in these workers?

62
Ten factory workers were followed for 90 days to assess the development of
asthmatic-like symptoms after acute exposure to toluene-di-iso-cyanate as
the result of an industrial accident. Consider this table of data obtained
from the exposed factory workers:
Subject Follow-up (Days) Status at End of FUP
A 90 No symptoms
B 10 Symptoms
C 90 No symptoms
D 90 No symptoms
E 30 Symptoms
F 60 Symptoms
G 90 No symptoms
H 90 No symptoms
I 15 Symptoms
J 25 Symptoms

Assuming that the observation of all subjects began on the same calendar date and
that symptoms didn’t resolve, what was the prevalence of those symptoms on day
28?

63
 Common Mistakes
1. Failing to specify the time period for all measures
o You must specify the time period

- Reporting Point Prevalence: P = 0.00 = 0.0% at specified time point

- Reporting Period Prevalence: P= 0.00 = 0.0% over specified time
period
- Reporting CI: CI = 0.00 = 0.0% over specified time period
- Reporting IR: IR = 0.00 cases per person-time (time=day
week, month, year)

2. Forgetting to identify the population at risk

o You must remove the prevalent cases from your population at risk

3. Failing to answer all of the questions being asked

64
 Summary
• We measure the frequency of disease in a population to:
o Determine the extent of disease in a population
o Identify patterns and trends of disease occurrence
o Identify the causes of disease

• Four fundamental components are necessary to measure the

frequency of disease:
1. Source Population (Which Group of People)
2. Cases of Disease (Numerator)
3. Size of Source Population (Denominator)
4. Time (Implicit or Explicit)
• These components are used to estimate the frequency of
existing cases of disease (prevalence) and the occurrence of
new cases of disease (incidence)
65