A prospective cohort study was conducted over a period of 10 years to investigate whether there is an association between smoking and stroke. The study population was 15,000 people selected at random from a national government funded health insurance database.
Table 1 shows the results of a pilot study comparing self-reported current smoking status and the current smoking status determined using a gold standard method.
Table 1: Comparison of a self-reported smoking status and gold standard test
Not current smoker
Not current smoker
a) From the results of the study shown in Table 1, calculate the sensitivity and specificity, and their 95% confidence intervals, of using self-reported smoking status compared to the gold standard test. Report and interpret the meaning of each of the measures you have calculated. [8 marks]
b) Would using self-reported smoking status impact on the validity of a study investigating the effect of smoking on stroke? Provide reasons to support your answer. [3 marks]
A clinical examination was conducted on all participants at baseline. The researchers collected information on demographic information, such as age and sex, health behaviours such as past and current smoking status and exercise, medications and participants’ medical history. The researchers had access to data on every visit to a doctor or hospital for each study participant, as well as the death records of study participants which they used to determine whether study participants had a stroke or died from a stroke. Page 2 of 9
PHCM9498 Epidemiology and Statistics for Public Health 2015 – Assignment 2
Of the 15,000 participants randomly selected for the study, 1,000 already had a history of stroke and were excluded. Of the remaining study participants, full data were available for 11,000 participants. The study investigators found that 2,200 of the remaining 11,000 study participants reported being current smokers at baseline. A total of 500 of the 11,000 study participants had a stroke during the study period, of whom 167 were current smokers at baseline.
a) Why did the investigators exclude the 1,000 people who had previously had a stroke at the beginning of the study? [1 mark]
b) Were there any important problems with selection bias in this study? Provide reasons to support your answer. [3 marks]
c) Draw a 2 x 2 table to display the study data of the association between current smoking and stroke. [2 marks]
d) Calculate the Relative Risk and its 95% confidence interval using a hand calculator. Please show your working and write a conclusion. [7 marks]
e) The investigators analysed the data further to assess whether there was an association between smoking and death due to stroke. They also wanted to know whether age was a confounder or effect modifier of the association between heavy smoking and death due to stroke. They calculated the following values:
i. All participants - heavy smoking and death due to stroke: RR = 2.5
ii. Male Participants - heavy smoking and death due to stroke: RR = 2.2
iii. Female participants - heavy Smoking and death due to stroke: RR = 3.5
Is gender is a confounder or effect modifier of the association between smoking and death due to stroke? Give reasons for your answer. [5 marks]