Epidemiology and Statistics

This document discusses the features of a cohort study, advantages and disadvantages of routinely collected data, the association between drinking very hot tea and cancer of the oesophagus in an Iranian Province, and more.

Epidemiology and Statistics 1
Epidemiology and Statistics
Name
Course Number
Date
Faculty Name

Epidemiology and Statistics 2
Global public Health, Epidemiology and Statistics
Question 1
a) Features of a cohort study with descriptions of how they are applied
i. Study participants are observed over a specified time range, either prospectively or
retrospectively. In this study, singleton births were followed for 7 years while data
was collected using the nationwide registries.
ii. The sample size is determined prior the start of the study. 58 841 was the number of
singleton birth and this number of determined before the study was started. There
were no additional study participants during the study period, which means that the
sample may only decrease because of drop-outs, loss of follow-up among other
reasons.
iii. Study endpoints are clearly stated before commencing the study. In the study, the
researchers clearly defined that an endpoint or observation of interest would be at
least one hospitalisation because of asthma or prescription for asthma before the age
of 7 years. Any observation of interest after the defined time would not be of
importance to the study.
b) Advantages and disadvantages of routinely collected data
Advantages
- Causation can be determined by linking exposure and disease
- Patients data can be independently analysed to understand their progression
Disadvantages
- There is a risk of loss to follow-up which leads to decrease in the study sample size
- Some patients/study participants might miss clinics, hence missing data
c) The risk ratio of high exposures versus no expose
risk ratio=0.045
0.033 =1.364
The relative risk of developing asthma for high exposure to maternal smoking compared
to no exposure was 1.36. Therefore, maternal smoking increased the risk of asthma
among the infants and children by approximately 36%.

Epidemiology and Statistics 3
d) The meaning of adjusted odds ratio for high exposure to maternal smoking on emergence
of asthma
The adjusted odd ratio is greater than 1 and the lower bound for the confidence interval is
greater than one. Having an odds ration with a confidence interval that does not include 1
means that the odds ratio is statistically significant, hence one group is either more or less
likely to experience the phenomenon. Therefore, after adjusting for the other covariates,
the highly exposed group of infants and children on maternal smoking were more likely
to develop asthma compared to their non-exposed counterparts by approximately 35%.
e) The odds ratio has been adjusted for gender, birth order, maternal age, marital status and
index of social economic status. What was it adjusted?
All these covariates which were adjusted for are potential confounding factors. Failing to
adjust for potential confounders might bias the result. Therefore, since the effect of these
variables on development is known or has been reported in other studies, including them
to the model reduces the variation, hence improving the confidence of the estimates (El-
Masri, 2013).
Question 2
Rapid HIV ELISA for sub-Saharan Africa with a sensitivity of 95% and a specificity of 96%
based on western blot assay
a) The mean of the sensitivity and specificity of the ELISA kit in detecting HIV in sub-
Saharan Africa
The ELISA kit would have a higher chance of detecting a true negative than a true
positive by 1%. Given a 100 HIV positive samples confirmed by western blot assay, there
is a 95% chance that a sample would be correctly determined as HIV positive. Similarly,
given 100 negative sample tested by western blot assay, the ELISA kit would correctly
categorise the sample 96% of the time.
b) 2 x 2 contingency table for 1,000 pregnant women given 40% HIV prevalence
ELISA Test

Epidemiology and Statistics 4
Positive Negative Total
HIV Status Positive 380 20 400
Negative 24 576 600
Total 404 596 1,000
c) Positive and negative predictive values
Positive predictive value= true positive
true positive+false positive = 380
404 =0.94
Negative predictive value= true positive
true positive+false positive = 576
596 =0.966
d) A rural are in Zimbabwe with a prevalence of 0.2%.
In this rural population of elderly people in Zimbabwe with an HIV prevalence of 0.2%,
the sensitivity and specificity will remain constant because they are kit-specific.
However, the negative and positive predictive values will change since the counts of the
false positive and negative are affected. The positive predictive value will decrease
significantly while the negative predictive value will increase. Therefore, as the
prevalence increases, a similar shift will be observed where the negative predictive rate
decreases as the positive predictive value increases. Sub-Saharan is assumed to have high
prevalence rates of HIV, which has informed the development of the kit. However, the
rural population with a very low prevalence does not match the characteristics of sub-
Saharan Africa. Applying this kit in this population will not produce reliable results for
the testing of HIV (MacFarland, 2014).
e) Criteria to be taken into account when deciding whether to a screening programme is
justified.
The criteria to determining whether a screening programme is justified is based on the
capability of the kit in use to detect true and positive and negatives. Any kit to be
implemented should be specific and designed based on the prevalence of the disease
within that area. A kit which has not been specifically developed for the population will
be erroneous and unable to detect the condition effectively. For instance, a kit that has
been developed for high prevalence population would only work effectively in this
population. The positive and negative predictive values among other diagnostic measure

Epidemiology and Statistics 5
values will not be compromised. However, transferring this particular kit to another
population which differ by characteristics of the diseases, the results obtained will not be
of high precision.
A population with low prevalence rate would require high sensitive kit, hence reducing
the false positive rates. For populations with high prevalence rates, a highly specific kit
will produce the best results. An average population would require a kit with a balance
between the sensitivity and specificity. Maintaining this balance will stabilize the
negative and positive predictive rates.
Question 3
The association between drinking very hot tea and cancer of the oesophagus in an Iranian
Province
a) The study design and why it was used
The researcher as interested to answer the question of relationship between drinking very
hot tea and having oesophagus cancer. Therefore, the best study design would include
exposure to drinking hot tea. The study design used was a case-control study which
included oesophagus cancer cases and a matched control population. All of the study
participants ought to be drinking tea. A case-control study would establish odds of having
the disease for the exposed and non-exposed groups. Therefore, this kind of design would
provide evidence of whether exposure to the identified phenomenon is associated with
significantly higher or lower odds of the disease (Yin, 2009).
b) The odds of ratio for drinking hot tea is 7.7 [95% confidence interval 4.4 to 13.9]
Drinking hot tea is associated with almost 8 fold chance of developing oesophagus cancer
compared to not taking hot tea. This statistic is statistically significant at 95% confidence
level. Based on the study, there is a 95% confidence that the odds of getting oesophagus
for people who drink hot tea will range between 4 fold to 14 fold compared to no taking
hot tea.
c) Matching in case control studies
Since this study design aim at comparing groups, it is crucial to select participant who
have similar characteristics that might confound the outcome. Therefore, matching is
selecting controls that are similar to the cases in all possible characteristics. For instance,

Epidemiology and Statistics 6
having cases who are male, it would strengthen the study while having male controls and
so on. Developing cancer can be confounded by age, sex, place of residence among other
potential confounders and that is the reason the cases whether matched to the cases based
on these factors (Srikanth and Doddamani, 2013).
d) An adjusted odds ratio for drinking hot tea was 8.16 [95% confidence interval 3.93 to
16.91] compared to the controls
The adjusted odds ratio estimate is slightly higher compared to the unadjusted while the
95% confidence interval has widened. It indicates that adjusting the potential confounders
has modified the effect of drinking hot tea and developing oesophagus cancer. As
discussed above, there might be potential confounders which might introduce error on the
estimate. Therefore, performing a conditional logistic regression would cater for the
effect of these potential confounders, while reducing the error in the odds ratio estimate.
Age, sex or place of residence or a combination of either might have been used in the
model for adjustment (Hosmer, Lemeshow and Sturdivant, 2013).
e) Do I believe the results
In research, there is differences in strength of evidence based in the study design. These
results does not display the entire truth. Having better study designs such as cohort
studies would provide stronger evidence (Cook, Sheikh and Netuveli, 2004).

Epidemiology and Statistics 7
References
Cook, A., Sheikh, A. and Netuveli, G. (2004) Basic skills in statistics: a guide for healthcare
professionals. 1st edn, Class Health. 1st edn. London: Class Publishing. Available at:
https://books.google.co.uk/books?
hl=en&lr=&id=xd_rwm7rFwEC&oi=fnd&pg=PR7&dq=basic+skills+in+statistics+cook+A&ots
=-ub0jwsYOl&sig=bvHDQ2922QcXcbQWhb5akge6HjY.
El-Masri, M. M. (2013) ‘Odds ratio.’, The Canadian nurse, 109(6), p. 14. doi:
10.1054/ebog.2000.0196.
Hosmer, D. W., Lemeshow, S. and Sturdivant, R. X. (2013) Applied Logistic Regression Third
Edition, Applied Logistic Regression. doi: 10.1002/0471722146.
MacFarland, T. W. (2014) Introduction to Data Analysis and Graphical Presentation in
Biostatistics with R. doi: 10.1007/978-3-319-02532-2.
Srikanth and Doddamani, P. K. (2013) ‘Overview of study designs’, International Journal of
Pharmacy and Pharmaceutical Sciences, 5(3), pp. 1011–1015. doi:
10.1093/acprof:oso/9780195092424.003.0003.
Yin, R. K. (2009) Case Study Research: Design and Methods, Applied social research methods
series ; doi: 10.1097/FCH.0b013e31822dda9e.