Objectives To examine the sensitivity and specificity of the Assessment of Driving-Related Skills (ADReS), a clinical tool recommended by the American Medical Association for identifying potentially unsafe older drivers that includes tests of vision, motor function, and cognition. Design Cross-sectional observation study. Setting Memory assessment outpatient clinic of a university hospital. Participants Drivers with normal cognition (n = 47) and cognitive impairment (n = 75). Measurements A neurologist completed the ADReS during an office visit. Additional cognitive tests of executive, visuospatial, and visuomotor function were also performed. On a separate day, participants completed a standardized on-road test, assessed by a professional driving instructor using a global safety rating and a quantitative driving score. Results In this sample of currently active older drivers with and without cognitive impairment, measures of cognition - particularly the Trail-Making Test Part B - were more highly correlated with driving scores than other measures of function. Using recommended scoring procedures, the ADReS had a sensitivity of 0.81 for detecting impaired driving on the road test, with a specificity of 0.32 and an area under the receiver operating characteristic curve (AUC) of 0.57. A logistic regression model that incorporated computerized maze task and Mini-Mental State Examination scores improved overall classification accuracy, yielding a sensitivity of 0.61, a specificity of 0.84, and an AUC of 0.80. Conclusion In its present form, the ADReS has limited utility as an office screen for individuals who should undergo formal driving assessment. Improved scoring methods and screening tests with greater diagnostic accuracy than the ADReS are needed for general office practice.