ASSESSMENT ON FACTORS THAT AFFECT THE CARDIOVASCULAR PATIENTS

USING COX PH MODEL

Hankarso Harka Oromo1, Selmawit Serka Moja2, Kenenisa Abdisa Kuse3, Teshita Uke Chikako4,
John Elvis Hagan Jr.5,6, Abdul-Aziz Seidu7,8, Bright Opoku Ahinkorah9, Olana Angesa Dabi10,
Yohannis Desse Robe11, Reta Habtamu Bacha12

1. Road authority, arbagona, Sidama region, Ethiopia

2.
3. Department of Statistics, Bule Hora University, Bule Hora, P.O.Box 144, Ethiopia
abdisakenenisa40@gmail.com.
4. Wondo Genet College of Forestry and Natural Resource, Hawassa University, Hawassa, P.O.
Box 05, Ethiopia; teshitauke@hu.edu.et.
5. Department of Health, Physical Education, and Recreation, University of Cape Coast, Cape
Coast, Ghana; elvis.hagan@ucc.edu.gh.
6. Neurocognition and Action-Biomechanics-Research Group, Faculty of Psychology and Sport
Sciences, Bielefeld University, Bielefeld-Germany
7. College of Public Health, Medical and Veterinary Sciences, James Cook University,
Townsville, QLD 4811, Australia; abdul-aziz.seidu@stu.ucc.edu.gh.
8. Centre for Gender and , Takoradi Technical University, Takoradi P.O. Box 256, Ghana
9. School of Public Health, Faculty of Health, University of Technology Sydney, Sydney,
Australia; bright.o.ahinkorah@student.uts.edu.au
10. Lecturer, Assistant Professor, Department of Statistics, Bule Hora University, Bule Hora,
P.O.Box 144, Ethiopia: oangesa@gmail.com
11. Expert prevention and control infectious animal disease at Dimtu Vete Clinic
12. Department of Statistics, College of Natural Sciences, Jimma University Jimma, Ethiopia.
reta.habtamu@ju.edu.et.

Abstract
Background: Cardiovascular diseases (CVDs) are group of disorders of heart and blood
vessels. It is a major health problem across the world, 82% of cardiovascular death is
contributed by countries with low and middle income. These diseases affect younger populations
and lead to premature mortality in developing countries like Ethiopia due to lack of prevention
or effective management of CVD risk factors. The aim of this study was to assess the factors that
affect the survival of cardiovascular patients and choose appropriate models. Methods: This
study was conducted among patients on follow-up period. The source of data was the study
included patients that have taken from EDHS. Total 4712 patients, a sample of size were taken
for this study by using EDHS data. Semi-parametric survival models were performed to analysis
survival data by using R software. Results: Among 4712 patients, 1760(37.4.%) are alive during

i
the study period and 2952(62.6%) of patients are died which were deaths or events. The
prevalence of cardiovascular diseases was high. Higher prevalence of cardiovascular diseases
was found among patients who improperly use risk factors like alcohol and smoking.
Conclusions: In this study, being an uneducated, smoking cigarette, being diabetic patients and
blood pressure were significantly associated with cardiovascular diseases . Therefore, the health
organization should be implemented on a sentinel basis in Ethiopia, in order to inform policy
and guide strategies and programmers for the prevention and control of these risk factors of the
survival of cardiovascular patients.

Key Words: CVDs, Survival analysis, semi-parametric, Risk factors of CVD

ii
CHAPTER ONE

Background of the study

Cardiovascular diseases (CVDs) are group of disorders of heart and blood vessels. CVDs are a
major health problem almost half of all deaths across the world. The estimated 17.9 million
people died from CVDs in 2016 (WHO, 2018). Heart attacks and strokes accounted for 85.1% of
these deaths were the leading causes of disability-adjusted life years (DALYs) worldwide in
2015.

Cardiovascular disease, diabetes, and cancers – share a few common risk factors that are related
to diet and lifestyle behaviors (Klein,et al 2018). These include high blood pressure, high
cholesterol, tobacco use, excessive alcohol use, being overweight, obese or physically inactive,
all of which are on the rise in many African countries. The contribution of these risk factors to
CVD is consistent in Africa and other parts of the world (Kannel W.B, et al., 2009).

Although the effect of risk factors on CVD in Africa is similar to that in other populations, the
risk of CVD morbidity and mortality associated with hypertension may even be higher in
Africans. These diseases affect younger populations and lead to premature mortality in
developing countries due to lack of prevention or effective management of CVD risk factors.
Various studies have reported changing patterns of CVD risk factors, especially in urban areas in
Africa, due to unhealthy lifestyle behavior (Yohanes H., 2011).

The CVD risk profile of African populations is said to be consistent with early stages of the
epidemiological transition. However, there are no comprehensive studies that have adequately
described the transition in the African continent (Fuster V., 2014).

In Ethiopia, as in many developing countries, there are no complete or reliable records of births
and deaths at the national level. In the absence of reliable data, national health and development
policies and strategies lack a firm ground for the design and implementation of effective
programme. Instead, they are bound to rely on conventional views about disease control
priorities. Such policies and strategies fail to respond to emerging and re-emerging health
problems across different sections of the population (Nkomo V.T., 2017).

1
In Ethiopia health complications related to cardiovascular, especially for the urban communities,
are becoming the major health threat (Abdissa, et al.,2014) .

Currently in Ethiopia among adults in Addis Ababa reported more than 16% among males and
16.5 % among females are caused from CVD risk factor (WHO, 2018). It could be scaled up in
Ethiopia at a modest budget increase and that combination drug treatment to individuals having
more than 35 % absolute risk of CVD.

However, findings of this study was valuable for raising awareness, among policy makers and
the public at large, on the current magnitude of CVD risk factors and the need for surveillance
undertakings. This work sets the foundation for future efforts of prevention and control of CVD
and related chronic diseases (White MJ, et al., 2015).

Statement of the Problem

Now a day cardiovascular disease has been a major health problem, concern Addis Abeba
Ethiopia. The estimates more than 9 % of all deaths in 2012 were caused by CVD in Ethiopia. In
Ethiopia, cardiovascular is the second most common leading causes of all deaths. This is one of
the challenging problems that the country needs to address. What many people don’t know is that
unlike diseases such as cancer, cardiovascular disease (CVD) is not an inevitable part of human
life. In fact, CVD is associated with many preventable and treatable risk factors, such as high
cholesterol, smoking, hypertension, diabetes, physical activity, and obesity (Tolla M. T, et al.,
2016).
Most of the previous studies on different part of Ethiopia by considered hypertension, diabetic
and elder cardiac patients mentions CVD as serious problem in Ethiopia as well as in the world.
It is a major health problem across the world, 82% of cardiovascular death is contributed by
countries with low and middle income. Currently in Ethiopia among adults in Addis Ababa
reported more than 16% among males and 16.5 % among females are caused from CVD risk
factor (munroe PB et,al 2013). Consequently, this study focused on the risk factors for the death
of cardiovascular patients in Ethiopia by considering the following research questions. The
general objective of this study was to identify the risk factors for the survival of cardiovascular
patients in Ethiopia.

2
Methodology

Data Description
These studies were incorporated secondary data, from EDHS 2016. This study used survival
data. The event is death of cardiovascular patients and who alive until the end of the study time,
lost and dropped before death are considered as censored.

Dependent Variable
The dependent variable in this study was “The survival time of cardiovascular patients” or it
can be defined as “The time to death of cardiovascular patients”. It is duration of time from
date of treatment until date of death. Cardiovascular patients, who are alive during
the study time or dropped before death, are considered also as censored.

Independent Variables (Covariates)

The independent variables that used in this study are based on literature reviews on the factors
determining time to cardiovascular patients at the global level and in the country. Therefore,
based on the reviewed literature the independent variables which are assumed to influence the
survival of cardiac patients are stated as follows; Sex , Age of the patients, Educational status,
Economic level , Blood pressure , Body Mass index , Smoking, Alcohol use, Diabetes (High
glucose), Pulse rate, Region, Family history of cardiac disease

Table 3.1 Description and codes of covariates included in this study stated below

Covariates Categories and coding

Sex 1=female 2=male
Age of patients 0=10-20 1=20-30 2=30 and above
Blood pressure 0= Low (Below 90/60 mmHg)
1= Normal (90/60-120/80 mmHg)
2= High (above 120/80 mmHg)
Body mass index 0=Under -weight (<18.5), 1=Normal (18.5-
24.9)
2= Over-weight(>25)
Smoking 0=do not smokes 1=smokes
Alcohol use 0=yes 1=no
Diabetes 0=yes 1=no

3
 Pulse rate 0= regular(60-80) 1=irregular (≥ 80)
Region 1=Tigray 2=Afar 3=Amhara and 4=others
Economic level 5=poorest 2=poorer 3=middle 4=richer
1=richest
Family history of CVD 0=yes 1=no
Educational status 0=yes1=no

Survival data Analysis

Data collected on the time to an event-such as the death of a patient in a study-is known as
survival data. The use of survival analysis as opposed to the use of different statistical methods is
most important when there are censoring data (Aalen, 2016). It involves the modeling and
analysis of data that have a principal end and the time until an event occurs (time-to-event data).
By time, we mean years, months, weeks or days from the beginning of follow-up of an
individual until an event occurs. Survival data or time to event data measure the time elapsed
from a given origin to the occurrence of an event of interest. In survival analysis the researcher
usually refer to the time variable as survival time because it gives the time that an individual has
'survived' over some follow-up period (Kleinbaum. and Klein., 2011).

Most of the time censoring is the main feature of survival analysis. Censoring occurs when we
have some information about individual survival time, but we do not know the survival time
exactly. There are generally three reasons why censoring may occur. A person does not
experience the event before the study ends; a person is lost to follow-up during the study period
and a person withdraws from the study because of death (if death is not the event of interest) or
some other reason called censor.

The survival function is defined as the probability that the survival time is greater or equal to t is
given as;

S(𝑡) = 𝑃(𝑇 > 𝑡

This means the probability of being in cardiac patients at time T or the probability that there
were non cardiac patients after time t. Where 𝑇 be a non-negative random variable that describes
the length of time until events.(𝑡) is a monotonically decreasing function of 𝑡 with the properties.

4
The hazard function is the probability that if a cardiac patient survives to time t, he/she died to
the event in the next instantaneous and given by:

Mathematically, denoted as:

f (t )
h(t) = S (t)

The cumulative hazard function is defined as the total number of failures or CVD over an
interval of time and it is obtained as

H(t) = ∫ h ( u ) du
0

Where ℎ(𝑢) is hazard risk and 𝑢 is accumulated risk. The existence of covariates that change
over time is also a distinguishing feature in survival analysis.

Non parametric Survival Model

Nonparametric analyses are more widely used in situations where there is doubt about the exact
form of distribution. The Kaplan-Meier estimator is the most common non-parametric or
distribution-free estimate of S(t)=P(T>t) without resorting to parametric models. The Kaplan-
Meier method is based on individual survival times and assumes that censoring is independent of
survival time (Kaplan E.L. and Meier., 1958).

Log-Rank Test

When comparing groups of subjects, it is always a good idea to begin with a graphical display of
the data in each group. A number of statistical tests have been proposed to answer this question
such as Log-rank, (Hosmer et al,.2008). Log-rank test is good for this to compare group of
subjects.

Regression Models for Survival Data

To know that how the survival experience of a group of individuals depends on the values of one
or more covariates, whose values have been recorded for each individual at the time origin.

To study this dependency, the researcher used semi-parametric survival model.

5
Semi-parametric Survival Model

Cox Proportional Hazards Regression Model

To explore the relationship between CVD patients times and explanatory variables, Cox
Proportional Hazard regression model proposed by (Cox, 1972), was implored.

The Cox PH model was used in modeling the survival data and analyzing the effect of risk
factors on survival time. Cox PH model is a semi-parametric model since there is no assumption
concerning the nature or shape of the underlying survival distribution. Since Kaplan Meier and
Log-rank test are commonly used for univariable analysis, for the multivariate analysis, the Cox
Proportional Hazard (CPH) models are used. The CPH model is essentially a regression model
used for investigating the association between the survival time of cardiac people and one or
more predictor/explanatory variables (Cox, 1972). In addition, the CPH regression model
simultaneously assesses the effects of several risk factors on survival time.

The hazard function in Cox model is called semi-parametric function since it does not explicitly
describe the baseline hazard function, h0 (t). The survival function is given by:

S (t, x, β) =e− H (t , x , β)

Where H (t, x, β) is the cumulative hazard function at time t for a subject with covariate x. Since
we have assumed that survival time is absolutely continuous, the value of the cumulative hazard
function is expressed as:

t t

H (t, x, β) =∫ h ( u , x , β ) du=exp(x , β¿)∫ h 0(u)¿ = exp (x, β) H0(t)

0 0

Under the Cox model, the survivorship function is

S (t, xi, β) =[S 0(t)¿¿ exp ⁡(x , β )]¿

Proportional hazards regression model is the most popular and commonly used regression model.
The estimation of the baseline hazard function is not required, time-dependent.

Graphical method to Check PH Assumptions

6
The proportional hazards (PH) assumption can be checked graphical diagnostics based on the
martingale residuals. To assessing the validity of PH assumption, By plotting estimated -log (-
log (survival)) versus survival time for two groups. The Researcher would see parallel curves if
the hazards are proportional assumptions are satisfied.

Model Adequacy Checking

After the model has fit the researcher has checked the adequacy of the model to assess which are
usually performed using model residuals. In linear regression methods, residuals are defined as
the difference between the observed and predicted values of the dependent variable.

The Cox-Snell residual is given by Cox and Snell, which is used for assessing the fitness of PH
model (Cox, D.R and Oakes D., 1984).

Martingale Residuals

To check the linearity of continuous variables the researcher would plot hazard against the
midpoint of the class and use plot of martingale residuals. For this particular study the plot of
martingale residuals against continuous covariates use to check linearity.

RESULTS OF THE STUDY

This chapter presents the analysis of data collected results and discussion of Risk Factors for
death of patients with Cardiovascular Diseases in Ethiopia. A total of 4712 samples of CVD
were included in the study from EDHS data. From the total, 2952(62.6%) of them uncensored
and the rest 1760(37.4.%) are censored. The functional status of patient is recorded in two
categories: Alive (censored) and Not alive (death or event). According to descriptive output
among 4712 patients, 1760(37.4%) are Alive during the study period and 2952(62.6%) of
patients are died or not alive which were deaths. This median value indicated that the half of the
cardiac patients died with probability 0.5.

As we have seen from Table 4.1 above a total of 4712 patients were considered for the analysis
Of these 62.6% were uncensored and 37.4% are still in alive(censored) , the median time is 6
month follow their cases, having a 95% confidence interval of lower bound 5 and upper bound 6.

7
Out of 4712 of cardiac patients 2577(54.6%) were females and 2135(45.4%) were males. Among
those patients the death proportion for female is 1692(57.31%) which is greater than that of male
patients which is 1260(42.69%).

About 695(14.74%) of patients were from Tigray region, 288(6.11%) of patients

were from Afar region and 504(10.69%) of patients were from Amhara regions and
3235(68.46%) patients were from others region.

From the total sample, the death proportion for Tigray region which is 14.74%) seems larger
than Amhara and Afar region which are 10.69% and 6.11%respectively. Regarding to
educational status, 1591(53.8%) of cardiac patients were uneducated and 840(28.4%) patients
were primary and 243(8.2%) patients were higher.

From this, the death proportion was higher for those patients who were uneducated which is
1591(53.8% and cardiac patients who were primary, secondary, higher which are 840(28.4%),
278(9.4%), 243(8.2%) respectively. Cardiovascular patients having age 30 and above years old
have death proportion of 2436(82.5%) were highest death than the cardiovascular patients having
age 10-20 and 20-30 with death proportion of 41(1.3%) and 475(16%) respectively.

Out of the total cardiovascular patients included in this study, 3199(67.89%) of patients were
smokers and 1513(32.21%) were non-smokers. The death proportion was lower for those
patients who were non-smoker which is 1346(45.6%) compared to smokers with death
proportion of 1606(54.4%)

Out of the total cardiovascular patients 3636(77.1%) of cardiac patients were alcohol users and
1076(22.8%) were non-alcohol users. The death proportion was lower for those patients who
were non-alcohol users which is 659(22.3%) compared to alcohol users with death proportion of
2293(77.6%).

Survival Analysis

In order to investigate each individuals estimate of the survival time researcher used estimation
techniques such as Kaplan Meier curve function

Table 4. 3 Kaplan Meier survival estimate

8
 time n.risk n.event survival std.err lower 95% CI upper 95% CI
1 4712 845 0.8207 0.00559 0.8098 0.8317
2 3391 404 0.7229 0.00671 0.7099 0.7362
3 2748 395 0.6190 0.00751 0.6044 0.6339
4 2078 163 0.5704 0.00783 0.5553 0.5860
5 1773 203 0.5051 0.00816 0.4894 0.5214
6 1480 213 0.4324 0.00837 0.4163 0.4491
7 1118 131 0.3818 0.00848 0.3655 0.3987
8 913 191 0.3019 0.00845 0.2858 0.3189
9 596 156 0.2229 0.00827 0.2072 0.2397
10 366 108 0.1571 0.00789 0.1424 0.1734
11 219 48 0.1227 0.00757 0.1087 0.1384
12 115 95 0.0213 0.00453 0.0141 0.0323

Here there is another output up to n, since n=4712

As we have seen from the above Table 4.3 the survival time of cardiovascular patients risk
becomes decreasing order with starting one.

Fig 4.1 The plot of the overall estimate of Kaplan-Meier survivor function of cardiac patients in
Ethiopia.

Fig.4.1 illustrates the Kaplan-Meier curve (KM) graph of the survival times for CVD patients
ensuring for censored (alive) or uncensored (death) data in the study.

9
It shows that most of the cardiac deaths occurred in the earlier months of dots initiation and it
declined in the later month of follow up and the estimate of the overall Kaplan-Meier Survivor
function revealed that most of the deaths occurred in the earlier month of time and it declined in
the later months of follow-up.

From this we have seen that the largest observations are uncensored.

This implies that the survival time is not converging to zero.

Log Rank test

By using the Log Rank test, test of equality was done along the probabilities across the different
groups. The null hypothesis to be tested is that there is no difference between the probabilities of
an event occurring at any time point for each population. It is considered that the predictor would
be included in a model if the Log Rank test has a p-value of < 0.05. However, if the predictor has
a p-value greater than 0.05, it is highly unlikely that it would contributed anything to a model
which includes other predictors. The p-values of the log-rank test shows difference in survival
experience between two or more levels of the covariates. The hypotheses being tested for each
variable are:

H0: There is no difference between survival curves

H1: There is a difference between survival curves

Comparison of Survival curves among covariates (Log- rank test)

The results of the log-rank test for the equality of survivor functions are presented as follows.
Table 4.4 log-rank test Different groups of covariates.

Covariates Chi- Df P value

square
value
Sex 25.6 1 4e-07
Age 0.2 2 0.9
Educational status 11.3 3 0.01
Economic level 6.7 4 0.2

10
 Diabetes 13.7 1 2e-04
BMI 2.3 2 0.3
Pulse rate 6.2 3 0.1
Region 23.6 11 0.051
Family history of 0.2 1 0.7
cardiac
Smoking 317 1 2e-16
Alcohol use 1.2 1 0.03
Blood pressure 49.1 2 2e-11
The researcher used log-rank test to look into the significance of difference in survival
experience among different factors. The null hypothesis to be tested has been no difference
between the probabilities of an event occurring at any time point for each categorical
covariate.

The variables are included in the study if the p-value from the log-rank test is less than 0.05,
which at 5 percent significant levels. The log-rank test results in Table 4.4 above show that
covariates Sex, Educational status, Smoking, alcohol use, blood pressure, and diabetes mellitus
are a significant covariates at 5% level of significance, the p-values are 4e-07, 0.01, 2e-16, 0.03,
2e-11 and 0.0002 respectively, which are less than 0.05 at 5 percent significant level. Whereas
Region, economic level, pulse rate, BMI, family history of cardiac and age have not significance
impact at 5% level of significant. The Kaplan-Meier estimator survival curve can be used to
estimate survivor function among different categorical covariates so that, the researcher can
made comparison among categorical covariates.

Overall fit for the Coxph regression model

There are several different tests of fit of the Coxph regression model:

Table 4.5 Several different tests of fit of the Coxph regression model:

Method Primary Use

1 Cox-Snell residuals Overall Fit
2 Martingale residuals Form of covariates (should covariates be
transformed)

11
3. Deviance residuals Outliers
4. Schoenfeld residuals Proportional Hazards assumption
5 Score residuals Influential points

The proportional hazard assumption is shown by p-value and is tested at 5% significant level.
Thus any result with p-value of less than 0.05 means the proportionality assumption is violated.
Table 4.6 Output of Proportional Hazard assumptions
Covariates Chi-square Df P-value
Sex 0.32296 1 0.0570
Smoking 0.48474 1 0.486
Diabetes 0.00225 1 0.962

Blood pressure 0.43318 2 0.510

Alcohol use 2.56724 1 0.109
Region 4.47760 3 0.064
Economic level 0.12068 4 0.728
BMI 2.69895 2 0.100
Educational status 1.42419 2 0.233

Pulse rate 1.25357 1 0.263

Family history of 1.23254 1 0.267

cardiac disease
Age 4.56917 1 0.053
GLOBAL 19.21640 12 0.083

The results shown that, the test is not statistically significant for each of the covariates, and the
global test is also not statistically significant. None of the covariates violated the assumptions of
Cox Proportional Hazard. Hence in this study Cox Proportional Hazard model is used.

As we have seen the above Table 4.6 the result is shows the p-value of rho statistic with
corresponding covariates are greater than 0.05 level of significant. This indicates that we not
reject the null hypothesis of the proportionality of cox proportional hazard model; this shows
that the assumption of proportional hazard is satisfied for those variables.

12
Results of Cox Proportional Hazards Model

Table 4.7 Results of Univarate cox PH regression Analysis

Covariates Coef exp(coef) se(coef) Z P

Smoking 0.664774 1.944052 0.037695 17.635 < 2e-16

Region -0.003323 0.996683 0.005965 -0.557 0.57752

Sex -0.150281 0.860466 0.037431 -4.015 5.95e-05

Economic -0.010898 0.989161 0.011572 -0.942 0.34632

level
Diabetes 0.117661 1.124863 0.038951 3.021 0.00252

Educational -0.039460 0.961308 0.019381 -2.036 0.04174

status
Alcohol use -0.062099 0.939790 0.044392 -1.399 0.016185

Pulse rate -0.021591 0.978641 0.019998 -1.080 0.28031

Family CVD 0.006588 1.006610 0.041113 0.160 0.87269

Age of -0.014778 0.985330 0.043536 -0.339 0.73427
patients
Blood 0.181830 1.199410 0.032673 5.565 2.62e-08
pressure
Body mass 0.021786 1.022025 0.020215 1.078 0.28116
index

As we have seen from the univarate Cox PH regression analysis Table 4.7 above, the
covariates Sex, Educational status, Blood Pressure, Alcohol, Smoking use, Diabetes mellitus, are
statistically significant at 20%-25% level.
Then covariates Sex, Educational status, Blood Pressure, Alcohol use, Smoking, and Diabetes
mellitus, are statistically significant at 20%-25% level of significance and selected as significant
risk factors for the death of cardiovascular patients from univariate cox PH regression analysis.
But the other four covariates such as, region, economic level, age, pulse rate and BMI are not
significant at 20%-25% of significant level.

13
Multivariable Cox PH Regression Analysis

The multivariable Cox PH regression analysis conducted by including all the potential risk
factors that had P-value less than 20%- 25% significant level in univariable Cox PH regression
analysis. To select the most appropriate subgroup of covariates in our model, the approach of
stepwise was applied.

It means only covariates with P-value less than or equal to 0.05 will be tested in the model, and
to keep it in the model, its P-value should be less than or equal to 0.05.

The result is shown as the following Table 4.8

Table 4.8 Results of the Multivariable Cox PH regression Analysis

Covariates Categories Coef Exp(coef) Se(coef 95%CI

P-value Lower Upper
Sex Female -
Male 0.1321165 0.8762389 0.038058 0.000518 0.8133 0.9441

No education
Primary 0.0198685 1.0200672 0.043110 0.644892 0.9374 1.1100
Education

Secondary -0.0653261 0.936762 0.065646 0.319680 0.8237 1.0654

Higher 0.1569656 0.8547335 0.069355 0.023624 0.7461 0.9792

Alcohol No - - - -
use Yes -0.0544140 0.9470400 0.044911 0.022567 0.8672 1.0342

Low - - - - - -
Blood Normal -0.0427697 0.9581320 0.052371 0.052371 0.8647 1.0617
pressure
High 0.33306 1.39523 0.06224 8.77e-08 1.2350 1.5763

No - - - - - -
Diabetes Yes 0.1231410 1.1310439 0.039479 0.001814 1.0468 1.2220

No
Smoking Yes 0.68681 1.9873678 0.0378 2e-16 1.8451 2.1406

14
These variables are the most factors of time to death for cardiovascular patients in Ethiopia.

A positive sign indicates that the hazard (risk of cardiovascular) is higher while a negative sign
indicates that the hazard (risk of cardiovascular) is lower. As we have seen the final model is
generated by including the variables Sex, educational status, Blood Pressure, Alcohol use,
diabetes and smoking are significant covariates at 5% level of significance.

So we interpreted as these covariates are more risk factors for the death of cardiovascular patient,
those variables have significant effect on the survival of cardiovascular patients.

From the above table that means results of the multivariable Cox PH model, discussion was
based on the hazard ratios. Comparison is made with the reference category and in between
groups for the categorical covariates. The reference category for the sex is female .The estimated
hazard ratio for male is 0.876 This implies that survival time to patients who are males were at a
rate for the risk of cardiovascular patients increases about 0.876 times higher than those who are
females.

This means that for the covariate males, the risk of cardiovascular increases about 0.876 times
over all covariates controlled. The 95% confidence interval shows that the hazard ratio is as low
as 0.8133 and as high as 0.9441.

The reference category for the educational status group is no education. The estimated hazard
ratio for educational status in primary is 1.020. This implies that the time to survive
cardiovascular patients who are in the primary group are a rate the risk of CVD increases about
1.020 times as the compared with no education.

The 95% confidence interval suggests that an increase in the rate of CVD due to educational
status categories of primary could actually be as low as 0.9374 and as high as 1.1100 times.

The estimated hazard ratio for secondary school is 0.936. On the other hand, the estimate for
secondary had a negative value indicating that the hazard ratio decreases over time. It can be

15
discussed that for the covariate, secondary school, the risk of CVD decreases about 0.936 times
as the compared with the no education.

The 95% confidence interval suggests that an increase in the rate of CVD due to educational
status categories of secondary could actually be as low as 0.8237 times and as high as 1.0654
times. This suggesting that there is a difference in the survival of CVD between those who are
primary, secondary group and who are uneducated group in this variable. The estimated hazard
ratio for Blood pressure in normal blood pressure is 0.951. The confidence interval (0.8647,
1.0617) suggests that the rate could actually be as much as high.

This suggesting that there is a difference in the survival of CVD between those who have low
blood pressure and who have normal blood pressure. The reference category for the smoking
group is who don’t smokes. The estimated hazard ratio for who uses smoking is 1.987. On the
other hand, the smoke users had a positive value indicating that the hazard ratio increases over
time.

The estimated hazard ratio for patients who have diabetes is 1.1310. This implies that patients
people who are in the categories of diabetes attacked are died at a rate the risk of CVD
increases about 1.13 times as the compared with the those who are in the categories of no
diabetes.

The estimated hazard ratio for time to death in alcohol use is 0.947. On the other hand, the
estimate for alcohol use had a negative value indicating that the hazard ratio decreases over time.
This implies that patients who are in the alcohol use group are died at a rate the risk of CVD
decreases about 0.947 times as the compared with those who are in the group of no alcohol use.

Discussion

This study is aimed to determine the prognostic factors that affect the survival of cardiovascular
patients in Ethiopia. For these purposes, the researcher has used survival analysis such as; semi-
parametric survival models are used. Analysis revealed that some demographic, socioeconomic
and environmental factors had a statistically significant effect on the survival of cardiovascular
patients (White MJ, et al., 2015).

16
According to this study Sex of patients with cardiovascular diseases has a significant effect on
the survival of cardiovascular patients a similar study at Washington stated that women patients
had higher survival time as compared to men patients (Yuanxin Hu., 2013) and (Taggu W, Lloyd
G., 2007).

The present study revealed that, Educational status has statistically significant effect on the
survival of cardiovascular patients. Similarly the finding of a study by (Dichiara T, et al., 2015),
Educational status adverse health effects and has been reported as an important predictor of
CVDs. Educated cardiac patients are more survive than that of uneducated cardiac patients.

The present finding is contradicted with the finding of a study by Tamiru, et al., 2010) who
investigated that the risk factors that determine the patients with cardiovascular diseases
mortality in Ethiopia. His results suggested that Age and body mass index are significant effect
on the death of cardiovascular patients, but pulse rate and economic level have not significant
effect on the survival of cardiovascular patients. His result stated that the patients having
irregular pulse rate were much more likely to die than the patients having regular (normal) pulse
rate. Another contradicts results by (Weber, et al., 2014), also stated that the cardiovascular
system is strongly affected by ageing.

This study revealed that blood pressure significant factor for the survival of cardiovascular
patients, thus patients with cardiac access to normal blood pressure had less risk of death (high
survival) as compared to those cardiac patients having abnormal blood pressure (high or low
blood pressure).

A similar findings by (Collett, D., 2003), and suggested that Blood pressure is a well-known risk
factor for cardiovascular diseases. And another similar study on cardiac patients at Tikur Anbesa
Specialized Tertiary Referral Hospital used the cox regression mode. Blood Pressure is one of
the major significant factor that affect the survival of cardiac patients (Hosmer, H et al,2008).
Blood pressure was the major factors that affect time to the cardiovascular (Melaku T, et al.,
2017).

A similar finding by (Munroe PB, et al., 2013) stated that cardiac patients inherit low tendency
toward heart disease than people who do have family history of CVDs. This finding is consistent

17
with the reports of Munroe PB, et al which indicated that family histories of cardiac diseases
were the not main cause of cardiovascular disease.

This study revealed that cardiovascular diabetes mellitus had statistically significant effect on the
survival of cardiac patients. Contrary the study stated that Diabetes mellitus is an important
chronic disease on CVD morbidity and mortality (Tesfaye F, et al., 2009) and the current study
contradicts the study by Adem A, et al reports that Type 2 diabetes mellitus is the commonest
risk factor identified risk factor of patients with ischemic heart disease (Adem A, et al., 2011).
According to the reports of the two studies that type 2 diabetes mellitus were not the leading
cause of cardiovascular disease.

Conclusion
The burden of risk factors for CVDs is increasing in developing countries including Ethiopia.
The prevalence of cardiovascular diseases was high. Higher prevalence of cardiovascular
diseases was found among patients who improperly use risk factors like alcohol and smoking. In
this study, being an uneducated, smoking cigarette, being diabetic patients and blood pressure
were significantly associated with cardiovascular diseases.

People who drink heavily have high mortality and attacked by cardiovascular diseases, including
sudden death. In addition, they may suffer from psychological, social and other medical
problems related to high alcohol consumption. Illiterate or uneducated population could be
regarded as highly growing epidemic of CVD and associated risk factors, as most of
physiological risk factors are disproportionately concentrated among this section of population.

Lower levels of educational attainment were associated with greater prevalence of cardiovascular
risks. Those who were uneducated, alcohol user, smoker, who have diabetes mellitus and blood
pressure, were more likely to have high cardiovascular risks. Generally, results shows that semi-
parametric model that means cox proportional hazard model was found to be the most
appropriate and preferred model than parametric model by supported test and checking different
assumption. The cox PH model fits well the cardiac data set even from other models.

18
Based on the study risk factors were identified for time-to-death patients from cardiovascular
disease. The researcher highly suggest that clinicians to give training especially to illiterate
patients a more emphasis should also give to those who infected by RCHD. Further studies
should be conducted in the regions of Ethiopia and identify other risk of factors CVDs that are
not included in this study.

Reference

Aalen, O., Borgan, O. and Gessing, H., 2016. Survival and event history analysis: a process

point of view. Springer Science & Business Media. 5(2).pp.8-12.

Abdissa, S.G., Oli, K., Feleke, Y., Goshu, D.Y., Begna, D.M. and Tafese, A., 2014. Spectrum of
cardiovascular diseases among Ethiopian patients at Tikur Anbessa specialized university
teaching hospital, Addis Ababa. Ethiop Med J, 52(1), pp.9-17.

Adem, A., Demis, T. and Feleke, Y., 2011. Trend of diabetic admissions in Tikur Anbessa and
St. Paul's University Teaching Hospitals from January 2005-December 2009, Addis
Ababa, Ethiopia. Ethiopian medical journal, 49(3), pp.231-238.

Amin P.M.,hajizadeh E.,Moghimi B,D.,safaee A ,,and abadi A.,Reza M.Z.(2007) comparing cox
regression and parametric models for survival of patients with gastric carcinoma. Asian
pacific journal of cancer prevention, 8, pp. 412-416.

Bradburn, M.J., Clark, T.G., Love, S.B. and Altman, D.G., 2003. Survival analysis part II:
multivariate data analysis–an introduction to concepts and methods. British journal of
cancer, 89(3), p.431.

Collett, D., 2003. Modelling survival data in medical research(No. 610.72 C721m). Florida, US:
CRC Press.

Cox, D.R., 1972. Regression models and life‐tables. Journal of the Royal Statistical Society:
Series B (Methodological), 34(2), pp.187-202.

19
Cox, D.R. and Oakes, D., 2012. Analysis of survival data (Chapman and Hail, New York).

DiChiara, T., Scaglione, A., Corrao, S., Argano, C., Pinto, A. and Scaglione, R., 2015.
Association between low education and higher global cardiovascular risk. The Journal of
Clinical Hypertension, 17(5), pp.332-337.

Finks, S.W., Airee, A., Chow, S.L., Macaulay, T.E., Moranville, M.P., Rogers, K.C. and
Trujillo,T.C., 2012. Key articles of dietary interventions that influence cardiovascular
mortality. Pharmacotherapy: The Journal of Human Pharmacology and Drug Therapy, 32(4),
pp.e54-e87.

Fuster, V., 2014. Global burden of cardiovascular disease: time to implement feasible strategies
and to monitor results.

Gesese, M.T., 2017. Survival Analysis of Time to Cardiovascular Disease Complication of

Hypertensive Patients at Felege Hiwot Referal Hospital in Bahir-Dar, Ethiopia: A
Retrospective Cohort Study. Archives of Current Research International, pp.1-9.

Hosmer Jr, D.W., Lemeshow, S. and May, S., 2008. Applied survival analysis: regression
modeling of time-to-event data(Vol. 618). Wiley-Interscience. Makinde, O.D., Khan,
W.A. and Khan, Z.H., 2013. Buoyancy effects on MHD stagnation point flow and heat
transfer of a nanofluid past a convectively heated stretching/shrinking sheet. International
Journal of Heat and Mass Transfer, 62, pp.526-533.

Kannel, W.B. and Vasan, R.S., 2009. Is age really a non-modifiable cardiovascular risk factor?.
American Journal of Cardiology, 104(9), pp.1307-1310.

Kaplan, E.L.and Meier, P.,1958. Nonparametric estimation from incomplete observations.

Journal of the American statistical association, 53(282), pp.457-481.

Klein J, Moeschberger M. Survival Analysis: Techniques for Censored and Truncated Data,
New York: Springer-Verlag, 2018.

Kleinbaum, D.G. and Klein, M., 2011. Kaplan-Meier survival curves and the log-rank test. In
Survival analysis (pp. 55-96). Springer, New York, NY.

20
Mathers, C.D., Lopez, A.D. and Murray, C.J., 2016. The burden of disease and mortality by
condition: data, methods and results for 2001. Global burden of disease and risk factors, 45,
p.88.

Melaku T, M.A., Schiffrin, E.L., White, W.B., Mann, S., Lindholm, L.H., Kenerson, J.G., Flack,
J.M., Carter, B.L., Materson, B.J., Ram, C.V.S. and Cohen, D.L., 2014. Clinical practice
guidelines for the management of hypertension in the community: a statement by the
American Society of Hypertension and the International Society of Hypertension. The
journal of clinical hypertension, 16(1), pp.14-26.

Misganaw, A., Mariam, D.H., Ali, A. and Araya, T., 2014. Epidemiology of major
noncommunicable diseases in Ethiopia: a systematic review. Journal of health,
population, and nutrition, 32(1), p.1.

Munroe PB, S.W., Airee, A., Chow, S.L., Macaulay, T.E., Moranville, M.P., Rogers, K.C. and
Trujillo,T.C., 2013. Key articles of dietary interventions that influence cardiovascular
mortality. Pharmacotherapy: The Journal of Human Pharmacology and Drug Therapy,
32(4), pp.e54-e87

Nkomo, V.T., 2017. Epidemiology and prevention of valvular heart diseases and infective
endocarditis in Africa. Heart, 93(12), pp.1510-1519.

Pampel, F.C., Denney, J.T. and Krueger, P.M., 2012. Obesity, SES, and economic development: a
test of the reversal hypothesis. Social science & medicine, 74(7), pp.1073-1081.

Taggu W, Lloyd G. Treating cardiovascular disease in women. Menopause International.

2007;13: 159-64

Tamiru, S. and Alemseged, F., 2010. Risk factors for cardiovascular diseases among diabetic
patients in southwest Ethiopia. Ethiopian journal of health sciences, 20(2)

Tesfaye F M., Wang, H., Lozano, R., Davis, A., Liang, X., Zhou, M., Vollset, S.E., Ozgoren,
A.A., Abdalla, S., Abd-Allah, F. and Aziz, M.I.A., 2009. Global, regional, and national
age-sex specific all-cause and cause-specific mortality for 240 causes of death, 1990-
2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet.

21
Tolla, M.T., Norheim, O.F., Memirie, S.T., Abdisa, S.G., Ababulgu, A., Jerene, D., Bertram, M.,
Strand, K., Verguet, S. and Johansson, K.A., 2016. Prevention and treatment of
cardiovascular disease in Ethiopia: a cost-effectiveness analysis. Cost Effectiveness and
Resource Allocation, 14(1), p.10.\

Tunstall-Pedoe, H. ed., 2017. MONICA, monograph and multimedia sourcebook: world's

largest study of heart disease, stroke, risk factors, and population trends 1979-2002.
World Health Organization

Weber, E.N., Pitsavos, C., Yannakoulia, C.M. and Panagiotakos, D.B., 2014. Comparisons
between survival models in predicting cardiovascular disease events: application in the
Attica study (2002-2012). Journal of Statistics Applications & Probability, 4(2), p.203.

White, M., Duquette, D., Bach, J., Rafferty, A., Fussman, C., Sharangpani, R. and Russell, M.,
2015, December. Family history of sudden cardiac death of the young: prevalence and
associated factors. In Healthcare (Vol. 3, No. 4, pp. 1086-1096). Multidisciplinary Digital
Publishing Institute.

World Health Organization, 2018. The work of WHO in the African Region-report of the Regional
Director: Executive summary & illustrative report 2017-2018.

Yohannes, H, 2011. Survival Analysis of Time to Cardiovascular Disease Complication of

Hypertensive Patients at Felege Hiwot Referal Hospital in Bahir-Dar, Ethiopia: A
Retrospective Cohort Study. Archives of Current Research International, pp.1-9.

Yohannes H, P.M.,hajizadeh E.,Moghimi B,D.,safaee A ,,and abadi A.,Reza M.Z.(2011)

comparing cox regression and parametric models for survival of patients with gastric
carcinoma. Asian pacific journal of cancer prevention, 8, pp. 412-416.

YuanxinHu.,2013. Survival Analysis of Cardiovascular Diseases All Theses and Dissertations

(ETDs). Paper 1209 htp://openscholarship.wustl.edu/etd/120 Zühlke, L., Mirabel, M. and
Marijon, E., 2013. Congenital heart disease and rheumatic heart disease in Africa: recent
advances and current priorities. Heart, 99(21), pp.1554-1561.

22