Patient-associated delay in seeking medical care in the setting of acute coronary disease is assuming increasing importance as the benefits of reperfusion therapies become more time dependent. Given the importance of accurate information concerning prehospital delay, we examined the extent of concordance between information reported by patients in structured interviews by hospital staff nurses compared with information about time of acute symptom onset as recorded in the medical record. Data were obtained from 1137 patients with a discharge diagnosis of coronary heart disease who were admitted to six coronary care units in the Minneapolis-St. Paul metropolitan area. The average and median durations of prehospital delay were similar as reported in the structured personal interviews and through the review of medical records for the respective disease groups. The extent of individual level of agreement of delay time was considerably poorer, however. The Pearson correlation coefficients on the logarithmically transformed data were 0.48, 0.50, and 0.59 for persons with acute myocardial infarction, unstable angina, and chronic coronary disease, respectively, in comparing data noted in the medical record with that obtained in the personal interviews concerning prehospital delay time. These results suggest good agreement between personal interviews and medical record accounts in characterizing the average length of prehospital delay at the aggregate level but considerably less agreement at the individual patient level. (Am Heart J 1998;135:293-9.)