SCHOOL OF POSTGRADUATE STUDIES

MEB 632 – STATISTICAL METHODS IN EPIDEMIOLOGY

FINAL EXAMINATION

2ND JUNE 2021

14:00 - 17:00 HOURS

TIME ALLOWED: 3 HOURS PLUS 5 MINUTES READING TIME

INSTRUCTIONS TO CANDIDATES:

1. Read the instructions very carefully.

2. Check that you have the correct examination paper in front of you.

3. There are FIVE (5) questions in this examination paper. Answer ANY FOUR (4).

4. All questions must be answered in the answer booklet provided only.

5. Write down the number of questions that you have answered on the cover of
the examination answer booklet.

6. Begin answering each question on a new page in the answer booklet provided
only.

7. Electronic calculators are permitted in this examination.

8. There shall be no communication among students during the examination.

Any students found doing so will be disqualified.

Page 1 of 9
 DO NOT TURN THIS PAGE UNTIL YOU ARE TOLD TO DO SO.

QUESTION ONE
The following output tables were produced from a logistic regression model analyzing a
study to investigate association between HIV infection (outcome) and number of
injections in the past year (exposure) taking into account the potential confounding
effect of blood transfusion received in the past 5 years, and number of sexual partners
ever. In the first stage, a univariable logistic regression model was fitted regressing HIV
infection on number of injections in the past year. The output is in Table 1.1:

Answer all the questions that follow under each output table.
Key to variables:

HIV: 0 = negative 1 = positive

Injections (number in past 12 months): 0 = none 1=1 2 = 2-4 3 = 5-9 4 = 10+

Blood (blood transfusion): 0 = yes 1 = no

Partners (number of sexual partners): 0 = 0-1 1 = 2-4 2 = 5-9 3 = 10+

Table 1.1: Univariable Logistic Regression Model Output

HIV Odds Std. Err Z P>|z| [95% Conf. Interval]

Ratio

Injections
0 1 (base)
1 1.15705 0.372845 0.45 0.651 0.61526, 2.17591
2 1.35176 0.304544 1.34 0.181 0.86921, 2.10218
3 2.01949 0.472498 3.00 0.003 1.27669, 3.19445
4 2.68656 0.807544 3.29 0.001 1.49051, 4.84235

Page 2 of 9
 Constant 0.24460 0.033092 -10.41 0.000 0.18763, 0.31888

Required:
A. Estimate the odds for HIV infection among those that did not receive an injection
in the past year. (2 marks)
B. Estimate the odds for HIV infection among those that received 10 or more
injections in the past year. (3 marks)

In the second stage, a multivariable logistic regression model was fitted that included:
receipt of blood transfusion in the past 5 years, and number of sexual partners ever.
The output is in Table1.2:

Table 1.2: Multivariable Logistic Regression Model Output

HIV Odds Std. Err z P>|z| [95% Conf. Interval]

Ratio

Injections
0 1 (base)
1 1.27783 0.436477 0.72 0.473 0.65422, 2.49588
2 1.40248 0.331600 1.43 0.153 0.88235, 2.22923
3 2.07869 0.506492 3.00 0.003 1.28940, 3.35114
4 2.44019 0.781211 2.79 0.005 1.30292, 4.57015

Blood
0 1 (base)
1 2.05684 0.805684 1.84 0.066 0.95450, 4.43225

Page 3 of 9
 Partners
0 1 (base)
1 2.12082 0.513232 3.11 0.002 1.31982, 3.40794
2 3.00791 0.864200 3.83 0.000 1.71280, 5.28230
3 7.58348 2.873964 5.35 0.000 3.60814, 15.93875

constant 0.11185 0.026661 -9.19 0.000 0.07010, 0.17846

Required:
C. Interpret the odds ratio of 0.11185 reported for the constant. (4 marks)

D. Calculate the odds of HIV infection among those that did not receive an injection
in the past year but have received a blood transfusion in the past 5 years and
have ever had 10 or more sexual partners. (7 marks)

E. Calculate the odds ratio of HIV infection comparing those that have had 0 or 1
sexual partners, 1 injection in the past year and have had no blood transfusion,
to those who have had 0 or 1 sexual partners, 1 injection in the past year and
have had blood transfusion. (9 marks)

[TOTAL: 25 MARKS]

Page 4 of 9
QUESTION TWO

Coronavirus disease 2019 (COVID-19) is a pandemic disease that was first identified in
the city of Wuhan in China with an outbreak of pneumonia without a clear cause in
December 2019 (Wang et al. 2020). However, many clinical trials and research works
are taking place to find effective vaccines or anti medications to lessen the dangerous
consequences of COVID-19. Additionally, several observational studies have also been
conducted to understand the epidemiological characteristics of this pandemic disease.

A recent study was conducted by “The Novel Coronavirus Pneumonia Emergency

Response Epidemiology Team” in February, 2020 to understand the summary
characteristics of patients; age distributions and sex ratios; case fatality and mortality
rates; geo-temporal analysis of viral spread; epidemiological curve construction; and
subgroup analysis. Part of this analysis involved the use of Poisson regression
modeling. The results are given in Table 2.1.

Required:
Write a report on your interpretation and discussion of the results in Table 2.1.
[TOTAL: 25 MARKS]

Page 5 of 9
Page 6 of 9
QUESTION THREE
In survival analysis, we use information on event status and follow up time to estimate a
survival function. Table 3.1 comprises number of surviving units and those that fail.

Where:

ni: the number of surviving units just prior to ti;

di: the number of units that fail at ti;

ⅆ𝑖
𝑞= 𝑛𝑖
; and

𝑝=1 −𝑞

Table 3.1: Survival rate calculations

Time ni di q p Survival Rate

1 180 5
10 175 10
20 160 5
35 150 5
50 130 2

Required:
A. Calculate the missing values in Table 3.1. (18 marks)

B. Sketch the Kaplan-Meier Curve for the survival rates calculated in (a). (7 marks)

[TOTAL: 25 MARKS]

QUESTION FOUR

Page 7 of 9
A. Analysis of the data on two groups, smokers and non-smokers, in a study that
investigated survival (days) following a root canal using cox regression model is
given as follows:
stcox smoking
failure _d: status == 1
analysis time _t: days
Iteration 0: log likelihood = -34.773766
Iteration 1: log likelihood = -34.619583
Iteration 2: log likelihood = -34.619583
Refining estimates:
Iteration 0: log likelihood = -34.619583
Cox regression -- no ties
No. of subjects = 30 Number of obs = 30
No. of failures = 14
Time at risk = 3647
LR chi2(1) = 0.31
Log likelihood = -34.619583 Prob > chi2 = 0.5787

Table 4.1: Cox Regression for Smoking

-----------------------------------------------------------------------
_t | Haz. Ratio Std. Err. z P>|z| [95% Conf. Interval]
-------------+---------------------------------------------------------
smoking | 1.373438 .7870813 0.55 0.580 .4466919 4.222895
-----------------------------------------------------------------------

Required:
Based on Table 4.1, interpret the:
I. Hazard Ratio; (4 marks)
II. P-Value; and (3 marks)

Page 8 of 9
 III. Confidence Interval. (3 marks)
B. The following questions refer to the article titled: A matched, case–control study
of the association between Schistosoma japonicum and liver and colon cancers,
in rural China; Annals of Tropical Medicine & Parasitology, Vol. 99, No. 1, 47–52
(2005).
REQUIRED
Using the Liver Cancer data in Table 4.2, calculate the matched - Crude Odds
Ratio for the association between Schistosomiasis infection and Liver Cancer
and interpret your answer. (7 marks)

Table 4.2: Liver Cancer Data

Case Control Total

Schistosomiasis 17 16 33
No Schistosomiasis 34 106 140
Total 51 122 173

C. Contrast between Cohort and Case-control Studies. (8 marks)

[TOTAL: 25 MARKS]

QUESTION FIVE
A. Explain what the term censoring means, different ways in which it can occur,
and why it is an important issue in analyses of time-to-event or survival data.

Page 9 of 9
 (7
marks)
B. When no event times are censored, a non-parametric estimator of S(T) is 1−Fn(t),
where Fn(t) is the empirical cumulative distribution function. However, when some
observations are censored, we can estimate S(t) using the Kaplan-Meier
product-limit estimator.

Required:
Calculate and fill in the missing values in Table 5.1.

Table 5.1: Cumulative Survival Rates

(18 marks)

[TOTAL: 25 MARKS]

Page 10 of 9
END OF EXAMINATION PAPER

Page 11 of 9