Observational study

An observational study measures variables without assigning participants to interventions or manipulating the conditions being studied. It can reveal associations, but confounding can complicate causal conclusions.

Confounding: A distortion in an estimated relationship caused by a third variable associated with both the exposure and outcome. A confounder can make an observed association differ from the causal effect.
Exposure (epidemiology): A factor or condition whose relationship with a health outcome is studied. Observational studies compare outcomes across naturally occurring exposure levels.
Randomized controlled trial: An experiment that randomly assigns participants to an intervention or control condition and compares outcomes. Random assignment can balance confounders, unlike ordinary observation.
Epidemiology: The study of health and disease patterns, causes, and effects in populations. Observational studies are a core way epidemiologists investigate population health.
Causal inference: The process of estimating whether and how a cause changes an outcome. Observational evidence supports causal claims only when assumptions and alternative explanations are addressed.
Selection bias: Systematic error arising when inclusion in a study or analysis depends on characteristics related to the variables being studied. Who enters or remains in a study can distort its measured associations.
Outcome (epidemiology): A health event, condition, or measurement selected for study. The outcome defines what researchers measure in relation to an exposure.
Natural experiment: A study exploiting an external event or policy that assigns exposures in a way resembling randomization. It uses naturally occurring assignment rather than researcher-assigned treatment.
Pharmacoepidemiology: The study of medication use and effects in large populations. It often examines drug safety using routine health records rather than assigned treatments.
External validity: The extent to which study findings apply beyond the conditions and participants studied. Real-world samples may reflect ordinary practice, but selection still limits generalization.

Linked from 24 pages