Analyzing naturalistic driving behavior recorded with in-car cameras is an ecologically valid method for measuring driving errors, but it is time intensive and not easily applied on a large scale. This study validated a semi-automated, computerized method using archival naturalistic driving data collected for drivers with mild Alzheimer’s disease (AD; n = 44) and age-matched healthy controls (HC; n = 16). The computerized method flagged driving situations where safety concerns are most likely to occur (i.e., rapid stops, lane deviations, turns, and intersections). These driving epochs were manually reviewed and rated for error type and severity, if present. Ratings were made with a standardized scoring system adapted from DriveCam®. The top eight error types were applied as features to train a logistic model tree classifier to predict diagnostic group. The sensitivity and specificity were compared among the event-based method, on-road test, and composite ratings of two weeks of recorded driving. The logistic model derived from the event-based method had the best overall accuracy (91.7%) and sensitivity (97.7%) and high specificity (75.0%) compared to the other methods. Review of driving situations where risk is highest appears to be a sensitive data reduction method for detecting cognitive impairment associated driving behaviors and may be a more cost-effective method for analyzing large volumes of naturalistic data.

Background: Controlled naturalistic driving for examining impacts of cognitive impairment on driving safety is rare. Objective: Evaluating the safety among drivers with mild cognitive impairment based on near collision incidents using naturalistic driving, and investigating its correlation with cognitive measures. Methods: Frequency of near collisions of 44 cognitively impaired [Age = 75.1(±6.7), MMSE = 25.5(±2.5)] and 19 control group drivers [Age = 72.5(±7.8), MMSE = 29.3(±0.8)] were obtained from two weeks of recorded driving. Survival time free of predicted collision based on a previously established near-collision to collision estimate ratio of 11 : 1, for 140 hours of driving exposure was calculated. Participants were also tested using Mini-Mental Status Examination (MMSE), Trail A, and Trail B. Spearman correlation and Cox survival analysis were conducted. Results: Near collision frequency per driving hour was correlated with MMSE (r = -0.258, p = 0.041). Survival analyses showed that cognitively impaired drivers might be prone to higher probability of having collision (p = 0.056) with a hazard ratio of 5.78 (p = 0.092). When all participants were combined, there was a significant difference (p < 0.017) in all the three cognitive measures between drivers with and without predicted collision, which were not significant within patient or control group alone (p > 0.186). Cox regression analysis showed MMSE as the only significant factor (p < 0.025) for survival time of predicted collision, but not age, gender, or driving experience. Conclusion:The association between driving critical events and cognitive measures suggests that some drivers with mild cognitive impairment might have an elevated driving collision risk compared to control drivers. Standard clinical cognitive measures may be reasonable predictors.