Question 1

Which of the following design options maximises the exchangeability of groups being compared to assess whether a given exposure causes a particular disease?&#10;A)A standard randomised controlled trial (RCT)&#10;B)A cross&#8208;over trial&#10;C)A carefully matched cohort study&#10;D)A counterfactual thought experiment&#10;E)A carefully matched case&#8208;control study

Accepted Answer

This is clearly the theoretical optimum, with each group being
simultaneously exposed and unexposed to the factor of interest.But in practice of
course we can't run this study since we don't have the necessary parallel universe.So in
the real world either A or B provide our best substitutes.B has the advantage of each
person being both exposed and not exposed, but not at the same time; and it can only
be used in certain situations, as discussed in the text.Matching in a case‐control study
(E) actually introduces new confounding; and while this is not so in a cohort study (C), it
is an inefficient and insufficient strategy for controlling for many confounders.So in
general the RCT rules, although there are many causal associations which it cannot (or
should not) be used to assess.

Question 2

What is/are the main disadvantage(s) of using a cross&#8208;sectional survey to assess the association between marijuana use and schizophrenia? (Select all that apply)&#10;A)Cross sectional studies are affected by recall bias&#10;B)Cross sectional studies are not generalisable to clinical populations&#10;C)It is not possible to determine whether marijuana use preceded the onset of schizophrenia&#10;D)Having schizophrenia can distort memories, thus leading to recall bias

Accepted Answer

Possibly the most important limitation is that it is not possible to
determine whether marijuana use preceded the onset of schizophrenia (C).In cross‐
sectional studies, exposure and outcome information are collected at the same time.In
this study, it is not possible to definitively establish the temporal ordering between
'marijuana use' and 'schizophrenia'.It may be that people who use marijuana are more
likely to subsequently receive a diagnosis of schizophrenia.Or, it may be that people
who have a diagnosis of schizophrenia are more likely to use marijuana.Recall bias is
possible in any cross‐sectional study (A), and perhaps especially in persons with mental
illness (D).The question is not a clinical one, so B is irrelevant.

Question 3

Which of the following is true of a case&#8208;control study?&#10;A)Good for studying rare diseases&#10;B)Generally less expensive than cohort studies&#10;C)Retrospective&#10;D)A, B and C are all true&#10;E)None of the above statements is true

Accepted Answer

A, B and C are all true.
Case‐control studies are good for studying rare
diseases and, because they can usually be done more quickly and only a relatively small
number of disease‐free people are studied, they are generally less expensive than
cohort studies.Their main theoretical drawback is that they are by necessity
retrospective in nature because the cases have already developed disease by the time
they are recruited and therefore are particularly susceptible to exposure
misclassification and recall bias.

Question 4

Which of the following is true in the context of interpreting the results of a case&#8208; control study compared to a cohort study? (select all that are true)&#10;A)Exposure information has to be obtained after diagnosis of disease so may be less accurate than in a cohort study&#10;B)Confounding is a bigger problem than in a cohort study&#10;C)Accuracy of disease diagnosis is a bigger problem than in a cohort study&#10;D)Recall bias is a bigger problem than in a cohort study&#10;E)Confounding is less of a problem than in a cohort study

Accepted Answer

In a case‐control study the exposure information is
collected after the development of disease and thus is often less accurate than in a
cohort study when it is collected before people develop disease.For example, it is
possible that an individual's exposure status may have changed with the onset of early
disease (e.g.they have lost weight or changed their diet), or that they will recall past
information differently because they know they have disease (recall bias).Confounding
(B, E) is a problem in all observational research and so is just as much a problem in
cohort studies as it is in case‐control studies.In a case‐control study it is usually possible
to verify the diagnosis of all cases so diagnosis is at least as accurate, if not more
accurate than in a cohort study (C).

Question 5

Controls are needed in a case&#8208;control study because:&#10;A)They may be followed to determine if they have the disease in question&#10;B)They increase the sample size so that statistical significance can be achieved&#10;C)They provide a comparable estimate of the frequency of exposure in the absence of disease&#10;D)They are matched to cases on suspected risk factors

Accepted Answer

They provide an estimate of the frequency of exposure in the absence of
disease which can then be compared to the frequency of the same exposure in cases.
Without controls we would not know if an exposure was associated with the disease.

Question 6

Which of the following methods is the best method to control for confounding through study design?&#10;A)Randomisation&#10;B)Restriction&#10;C)Matching&#10;D)All of the above

Accepted Answer

While all of the above can be useful, randomisation is the most effective.
Random allocation of individuals to either an exposed or unexposed group (or
treatment or control group) ensures better than other approaches that exposed and
unexposed groups are exchangeable and that confounders (known and unknown) are
equally balanced.Restriction and matching will also make the exposed and unexposed
groups more similar, however, it is not as effective as random allocation and is limited
to those potential confounding variables that are known and selected a priori.

Question 7

In measuring an exposure in a prospective cohort study design, which biases are most readily avoidable?&#10;A)Misclassification&#10;B)Interviewer bias&#10;C)Recall bias&#10;D)Healthy worker effect&#10;E)None of the above

Accepted Answer

Recall bias cannot occur for measures of a participant's state at the
point of entry into a cohort study, as they are unaware of their future disease history so
cannot be influenced by it; and this is also true for an interviewer at baseline.In
contrast all measures may be misclassified (A) at any point in a study and the healthy
worker effect (D) influences the number of health events the cohort experiences.

Question 8

The following table shows data from an epidemiological study.What type of study was this most likely to be?    &#10;A)Case&#8208;control&#10;B)Cohort&#10;C)Randomised controlled trial (RCT)&#10;D)Cross&#8208;sectional&#10;E)Descriptive

Accepted Answer

As the results are presented in terms of the number of episodes of disease
in relation to the number of person‐years at risk of disease, this is most likely to have
been a cohort study.While RCTs are also prospective, person‐years are rarely used in
describing denominators.

Question 9

In assessing the relative merits of case&#8208;control and cohort studies, which one of the following statements is correct?&#10;A)A cohort design is more suited to the study of rare diseases&#10;B)Case&#8208;control studies are more costly&#10;C)Case&#8208;control studies take longer to complete&#10;D)Unlike case&#8208;control studies, cohort studies permit the direct calculation of incidence rates

Accepted Answer

Unlike case‐control studies, cohort studies permit the direct calculation of
incidence rates.In cohort studies information on person‐time over the period of follow‐
up can be used to calculate the incidence rate (i.e.incident cases divided by person‐
time).However, it is also worth noting that while incidence rates can't be calculated
directly from case‐control studies, it is sometimes possible to use the sampling rates of
cases and controls to estimate the incidence rate.In general, case‐control studies are
less costly and quicker to complete and more suited to studying a rare disease than a
cohort study.

Question 10

In a study designed to determine the incidence of hypertension (high blood pressure) in an inner‐city community, a sample of residents is selected for examination.Which
Of the following sampling methods is likely to lead to the least bias?

A)Residents who respond to a call for volunteers in the community's weekly newspaper
B)Residents living in a random selection of addresses from the community
C)Residents selected from records of the local general practitioner
D)Residents randomly selected from the check‐out lines at the local supermarket
E)Any of the above would be OK

Accepted Answer

You would definitely not want to use a call for volunteers in a local paper (A), this
will give a highly selected sample as it won't include people who don't get or don't read the
local paper.Similarly shoppers eager to get home will refuse, and those who don't use the
supermarket won't be included so D is not ideal.GP records (C) in most countries won't give
population coverage, so you would only get those who have presented because of illness and
have no idea of the prevalence in the rest of the population.B is the ideal sampling frame,
although it too is challenging because if you don't get a high participation rate your findings
may still be misleading.

Deck 4: Healthy Research: Study Designs for Public Health