recall bias

**Definition:** Systematic error arising from differences in the accuracy or completeness of recalled past exposures between comparison groups. It is a subtype of information bias (measurement bias).

**Mechanism / Pathophysiology of the bias:** Cases (people with the disease) are often highly motivated to search for explanations and therefore recall or report exposures more thoroughly than healthy controls. For example, mothers of children with congenital anomalies may recall minor illnesses or medication use during pregnancy more readily than mothers of healthy children. This produces differential misclassification of exposure (error in exposure assessment differs by outcome status). Most often this inflates the observed association (bias away from the null), though it can go in either direction.

**Where it occurs:** - Classic in retrospective case-control studies (exposure is assessed after the outcome is known). - Also relevant in retrospective cohort studies and cross-sectional surveys that rely on self-report. - Less of a problem in prospective cohort studies and randomized controlled trials, where exposure is recorded before the outcome occurs.

**Ways to reduce it:** - Use objective records (medical records, pharmacy databases, biomarkers, prescription registries) rather than self-report. - Use a control group with a different disease (a comparable motivation to recall), such as siblings or patients with other conditions. - Blind interviewers and participants to the study hypothesis; use standardized, structured questionnaires. - Shorten recall periods and use memory aids (calendars, photos). - Choose a prospective design when feasible.

**Differentiation from similar biases (high-yield on boards):** - **Recall bias:** differential accuracy of remembering exposures (case-control). - **Observer/interviewer (detection/ascertainment) bias:** the investigator's knowledge of the participant's status or hypothesis affects measurement or questioning. Reduced by blinding. - **Reporting (social desirability) bias:** participants underreport or overreport behaviors (e.g., alcohol, drug use, sexual history) because of stigma. - **Selection bias** (e.g., Berkson, Neyman/prevalence-incidence, healthy worker effect, loss to follow-up): occurs in how subjects are enrolled or retained, not how data are measured. - **Lead-time bias:** early diagnosis falsely appears to lengthen survival. - **Length-time bias:** screening preferentially detects slowly progressing disease. - **Hawthorne effect:** subjects change behavior because they know they are being observed. - **Pygmalion effect:** the researcher's expectations influence the outcome. - **Procedure/measurement bias:** the group being studied is measured or treated differently.

**Clinical vignette pattern (Step 1/Step 2):** "A case-control study examines the association between cell phone use and brain tumors. Patients with brain tumors report more past cell phone use than controls, but the objective records show no difference." Answer: recall bias.

**Key point:** Recall bias is a threat to internal validity, and it cannot be fixed by increasing sample size or by statistical adjustment after data collection. It must be addressed through study design.