EPI7-3:STUDY DESIGN - A Simpler Approach - Thinking From The 2 By 2 Table

Consider the by now familiar 2 by 2 table:

When thinking about study design it is useful to bear the above table in mind.

One approach is to think of what you are going to be observing and measuring and whether you will be starting out with different groups based on disease status and on which you will be measuring different past exposures, or by following those exposed differently over time to measure health outcomes, or simply to measure everyone in a population of interest.

For example, if one starts out with two groups or cohorts (labelled in red in the table above) of which one is exposed and the other not, the study design is a cohort or follow up study where the disease outcome is measured over time. This can be prospective in which case two groups with different exposures are followed up over time and health outcomes measured. It could also be retrospective, where there are good exposure records (in the company files for instance) and it is possible to conduct a mortality study for all workers who worked at that company for more than a month since 1940 and then to find out from vital statistics who has died and who not, and to trace back their exposures while they were at work.

For example, in a study of workers in a petroleum refinery, one identifies all those who were exposed to benzene at work and goes back in the records to work out how much they were exposed to over the years they worked and then looks up in the national cancer registry or registry of births and deaths to see what they died of (assuming they have died). Death from leukemia is the outcome, and because one is collecting information retrospectively (exposure and also perhaps mortality data) this is called a retrospective or historical cohort study. This is a very typical type of occupational health study in developed countries where the data infrastructure is good.

If on the other hand one starts out measuring the disease status (shown in red in the above table) and then proceeds to determine exposure in the past, this would be a case control study.

Staying with this example, because the numbers of people at work who are exposed to benzene in any one workplace may be small, it may not practical to conducted a cohort study as above. Instead one could go to a cancer registry or hospital and collect leukemia cases and take exposure histories from all cases and match them up to a number of controls (people in the registry or hospital who don't have leukemia) and compare the exposures to benzene in the two groups.

If one measures both disease status and exposure status simultaneously then this is typical of the prevalence or cross-sectional study. In this instance one is beginning with all the people, or the entire population of interest, T, who are studied. The outcome of the study will determine who has the disease and who is disease free along with their exposures.

General Introduction to Occupational Health: Occupational Hygiene, Epidemiology & Biostatistics by Prof Jonny Myers is licensed under a Creative Commons Attribution-Noncommercial-Share Alike 2.5 South Africa License