Observational Studies

Observational studies form the backbone of epidemiological research, providing insights into disease aetiology and the health of populations without the ethical or logistical constraints of experimental designs. The two major analytic observational study types are cohort studies and case-control studies, each with complementary strengths.

In a cohort study, investigators recruit a sample from the source population, classify participants by exposure status, and follow them prospectively to observe outcomes. Crucially, participants must be free of the outcome at baseline. Incidence proportion (cumulative incidence) and incidence rate can both be calculated. Relative risk (RR) is the ratio of the incidence in the exposed group to that in the comparison group; a risk difference (RD) expresses the absolute difference. Both metrics are interpretable only with a stated time period.

Prospective cohort studies follow participants from the time of exposure measurement. Historical (retrospective) cohort studies reconstruct past follow-up using existing records — useful for outcomes that take many years to develop but at the cost of relying on data collected for other purposes.

Case-control studies work backward from outcome. Cases — ideally incident (newly diagnosed) rather than prevalent — are identified, and controls drawn from the same source population are assigned an index date matching the date on which their paired case developed the outcome. Because incidence cannot be calculated, the measure of association is the odds ratio (OR). For rare diseases, the OR approximates the RR. Exposure measurement must be performed identically in both groups to prevent information bias.

Selection bias in case-control studies arises from how controls are selected; in cohort studies, from loss to follow-up that is related to both exposure and outcome, or from the healthy worker effect. Information bias occurs when there are systematic differences in the accuracy of exposure data between groups — common in case-control studies through differential recall (recall bias).

Confounding distorts the true relationship between exposure and outcome. A confounding variable must be associated with both the exposure and the outcome independently, and must not lie on the causal pathway. Confounding is controlled by restriction, matching, or multivariable analysis in observational studies; randomisation is the gold standard in experimental designs.

Heterogeneity in the context of epidemiology and systematic reviews refers to any kind of variability among studies — differences in participants, exposures, outcomes, or methods. High heterogeneity limits the validity of pooling results in a meta-analysis.