AbstractBackgroundA reliable history is critical to establish accurate clinical diagnoses, prioritize investigations and select appropriate treatments; yet few studies have considered the factors that influence the reliability of medical histories. Recognizing this, we determined the accuracy of medical history in community‐dwelling older adults engaged in longitudinal studies of memory and aging.MethodThe association between participant‐specific factors and accurate reporting of stroke or diabetes were evaluated using multivariable logistic regression in 1,401 participants enrolled in studies at the Knight Alzheimer Disease Research Center (Saint Louis, Missouri), including 425 participants with dementia (30.3%). History was verified with a reliable collateral source in all participants. Stroke and diabetes were selected as index variables, as gold standard measures—brain MRI for stroke and hemoglobinA1c (HbA1c) for diabetes—were routinely obtained in participants.Result25/862 (2.9%) participants endorsed a history of stroke, which was confirmed on MRI in 8/25 (positive predictive value, 17.4%). Stroke was detected on MRI in 38/837 (4.5%) of those without reported stroke (negative predictive value, 95.5%). Older age (OR per decade=1.7, 95%CI: 1.2‐2.5) and use of an unrelated collateral source (OR=2.0, 95%CI: 1.03‐3.7) were associated with greater odds of inaccurate reporting of stroke history. 190/1322 (14.4%) participants reported a history of diabetes, which was confirmed with HbA1c in 105/190 (positive predictive value, 55.3%). HbA1c 6.5% was detected in 29/1132 (2.6%) of those who denied diabetes (negative predictive value, 97.4%). African Americans (OR=1.6, 95%CI: 1.02‐2.6) and those with hypertension (OR=2.6, 95%CI: 1.5‐4.2) or HbA1c compatible with impaired glucose tolerance (OR=4.2, 95%CI: 2.7‐6.4) had greater odds of reporting a history of diabetes. Cognitive impairment was not associated with inaccurate reporting of stroke or diabetes.ConclusionThe accuracy of reported history of diabetes and stroke was low in older community‐dwelling participants, independent of cognitive status. A history of diabetes or stroke should be objectively confirmed when relevant to decision making in clinical and research settings.