Background Clinical test batteries to predict safe vs unsafe drivers are not accurate enough to exempt the on-road test for medically at-risk drivers. Applying serial trichotomization to clinical test batteries may reduce the need to test all at-risk drivers.

Aims To examine whether serial trichotomization predicts pass/fail outcomes of a comprehensive driving evaluation (CDE) in medically at-risk drivers with physician-referrals for a CDE.

Methods CDE data was collected retrospectively from two driver assessment clinics in Canada (n=143; mean age 69.3±14.1 years). Clinical tests included the Montreal Cognitive Assessment (MoCA), Trail Making A and B tests, and Useful Field of View subtests 1–3 (UFOV1–3), and a pass/fail or indeterminate (i.e., fail with lessons and retest) outcome on the CDE. Serial trichotomization involved performing a receiver operating characteristics curve for each clinical test to determine cut-points with 100% accuracy in predicting pass/fail outcomes. A funnel was created arranging the clinical tests in order of accuracy (i.e., most-least) for predicting pass/fail outcomes, and with each clinical test’s cut-points, determining pass/fail or indeterminate outcomes.

Results Compared to participants’ CDE outcomes, serial trichotomization of the UFOV3, UFOV2, UFOV1, Trails B, MoCA, and Trails A predicted more pass (44% vs 33%) and fail outcomes (34.3% vs 27%) with fewer indeterminate outcomes (21.7% vs 40%).

Conclusion Serial trichotomization more accurately identifies safe vs unsafe medically at-risk drivers, reducing the number of unnecessary on-road tests, driving assessor caseloads, and client wait times.

Learning Outcome Describe the process of serial trichotomization of clinical tests predicting CDE outcomes.