Cohort Studies
~2 min read
Lesson 9 of 20
Notes
Cohort studies are a cornerstone of analytic epidemiology. They follow groups of participants defined by exposure status forward in time to observe who develops the outcome of interest, enabling calculation of incidence and measures of association.
In designing a cohort study, investigators first identify a source population, then recruit a sample. Ideally, participants are selected randomly, independent of their exposure status. For rare exposures (such as occupational exposure to a specific chemical), separate samples of highly exposed and unexposed individuals may be recruited — this is more efficient but risks selection bias if the groups differ in prognostic characteristics beyond the exposure of interest.
The healthy worker effect is a specific selection bias encountered in occupational cohort studies. Workers must be healthy enough to be employed, so occupational cohorts are systematically healthier than the general population. When an occupational cohort is compared to general population data, the occupational exposure appears falsely protective.
Measures calculated from cohort studies include: incidence proportion (IP/cumulative incidence), incidence rate (IR), relative risk (RR), and risk difference (RD). Both RR and RD require the time period to be reported.
Prospective cohort studies classify participants at the time of exposure assessment and follow them forward. Historical (retrospective) cohort studies use existing records to reconstruct a past follow-up period — useful for outcomes with long latency periods (e.g., mesothelioma after asbestos exposure), less expensive and time-consuming, but limited by data quality and the inability to control for confounders not measured in the original records.
Cohort study strengths: establishes temporal sequence; calculates incidence; can examine multiple outcomes from one exposure; good for rare exposures (recruit highly exposed group). Limitations: loss to follow-up causes selection bias if related to both exposure and outcome; misclassification of exposures; time-consuming and expensive; inefficient for rare or slow-developing outcomes (require very large sample sizes and long follow-up).