Older drivers have a high crash rate per vehicle mile of travel. Coupled with the growth of the number of older drivers on the road, this has generated interest in the identification of factors which place older drivers at increased risk. However, much of the existing research on medical and functional risk factors for crash involvement has generally been inconsistent. Methodological differences between studies have been hypothesized as being partly responsible for such inconsistencies. The source of information used to identify crash-involved drivers has been identified as one such difference. This paper reports on the agreement between self-report and state record for identifying crash involved-older drivers. We also sought to determine whether the prevalence of visual and cognitive impairment differs across crash-involved drivers identified by either or both sources. Finally, we assessed whether risk factors for crash involvement differed when crash-involved drivers were identified by either self-report or state records. Results indicated that there was a moderate level of agreement between self-reported and state-recorded crash involvement (kappa = 0.45). However, we did find significant differences between crash-involved drivers identified via state records and/or self-report with respect to demographic (age, race), driving (annual mileage, days per week driven), and vision impairment (acuity, contrast sensitivity, peripheral visual field sensitivity, useful field of view). We also found that the possibility for biased measures of association is real. Useful field of view impairment was associated with both self-reported and state-recorded crash involvement; however, the magnitude of the associations was disparate. Moreover, glaucoma was identified as a significant risk factor when considering state-recorded crashes but not self-reported crashes. While validation of these findings is required, research designed to identify risk factors for crash involvement among older drivers should carefully consider the issue of case definition, particularly if self-report is used to identify crash-involved older drivers. © 1998 Elsevier Science Ltd. All rights reserved.