Distortion from how cases are detected or reported, not their true occurrence
Ascertainment bias is a systematic difference in how cases are identified, diagnosed, or reported across the groups, sites, or time periods being compared. Because the completeness or sensitivity of case-finding varies, the resulting counts reflect detection effort as much as, or more than, true disease occurrence. It is distinct from selection bias in that the distortion arises from the surveillance or diagnostic process itself, not from who is enrolled.
When appraising a study, this bias threatens the internal validity of any comparison of incidence, prevalence, or trends over time or between settings. A study can report a falling rate of injury or disease that reflects declining case capture (fewer clinics reporting, stricter diagnostic criteria) rather than a genuine reduction in risk, leading readers to draw the wrong causal conclusion about an intervention or policy.
It typically surfaces in long-running registries, surveillance databases, or multi-site cohorts where diagnostic technology, coding definitions, reporting mandates, or clinician awareness change across the study period or differ between sites. Appraisal involves checking whether case definitions and ascertainment methods were held constant or documented across time and sites, whether sensitivity analyses stratify by era or centre, and whether the authors triangulate against an independent, stable data source (e.g. imaging-confirmed subgroup) to see if trends persist.
Ascertainment bias can be difficult to fully rule out even with careful reporting, because case-finding infrastructure often changes for reasons not documented in the dataset itself. It is sometimes invoked loosely as a generic caveat without specific evidence that detection actually varied, which can be used to dismiss inconvenient findings as readily as to genuinely qualify them; readers should look for concrete evidence of a detection-practice change rather than accepting the label at face value.
This guide was auto-drafted and is pending editorial review.