ascertainmentbias

Ascertainment bias, in statistics and epidemiology, is a systematic distortion that arises from the way subjects or data are identified, diagnosed, or recorded. It occurs when the method of ascertainment is related to the variable under study, leading to over- or under-representation of certain individuals, outcomes, or measurements. Unlike sampling bias introduced at the stage of selection, ascertainment bias specifically concerns the detection or identification process.

Common sources include selective testing or screening programs that enroll only people with certain symptoms, disparities in access to care, diagnostic tools with variable sensitivity, recall or reporting biases in self-reported data, and registry or publication biases where certain cases are more likely to be recorded or published. In genetics, ascertainment bias can occur when study cohorts are chosen from readily available data, making some populations overrepresented and inflating effect estimates. In epidemiology, surveillance systems that routinely capture certain outcomes but miss others can produce biased incidence or prevalence estimates.

Consequences include distorted measures of association, prevalence, or risk, which can mislead policy or clinical decisions. Mitigation approaches include designing studies with population-based sampling, broad and consistent eligibility criteria, using multiple independent data sources, standardizing measurement protocols, blinding assessors, and applying statistical adjustments such as weighting or sensitivity analyses. preregistration, replication in diverse datasets, and transparent reporting also help reveal and reduce ascertainment bias.