Prediction of Driving Ability in People With Dementia- and Non- Dementia-Related Brain Disorders

Brain disorders can impair physical and cognitive functions necessary for safe driving. Two hundred people with brain disorders referred for a driving assessment were recruited and their performance on a computerized battery of sensory-motor and cognitive tests (SMCTests) and a blinded on-road assessment determined. Based on SMCTests performance, binary logistic regression (BLR) and nonlinear causal resource analysis (NCRA) models classified on-road pass or fail with 70% accuracy. Greater accuracy could be achieved by splitting referrals into two groups: (1) Dementia and (2) Non-dementia-related brain disorders. BLR models classified on-road driving outcome as pass or fail with accuracies of 76% (Dementia) and 75% (Non-dementia), while NCRA models had accuracies of 77% (Dementia) and 80% (Non-dementia). Measures of attention were most critical for predicting driving ability in the dementia group. In the non-dementia group, prediction of driving ability was most accurate with assessment of a broader range of sensory-motor and cognitive functions. Compared to BLR, NCRA was able to identify and use additional measures to improve accuracy. NCRA is also better able to accommodate outliers due to it being a non-linear modelling method based upon individual performance-limiting impairments. We propose three main factors underlying sub-optimal prediction of driving ability based on SMCTests performance: (1) there are one or more functions important for driving ability which are not currently assessed with SMCTests – these could be sensory-motor or cognitive or other (e.g., attitude, confidence, insight, road code knowledge); (2) suboptimal classification/prediction techniques or models; or (3) inaccuracies in the on-road driving assessments.