Application of tablet-based cognitive tasks to predict unsafe drivers in older adults

Abstract

BackgroundDue to aging and medication interferences, a wide range of motor, sensory, and cognitive skills that are imperative for driving are affected in older adults. Though on-road tests are most indicative of driving ability, they are costly, stressful, time-consuming, and risky. Application of tablet-based cognitive tasks is investigated in identifying unsafe drivers in a population of healthy and at-risk for driving older adults. MethodForty-nine older adult participants aged 54 to 81 (M = 78.08, SD = 9.78) that were screened by their physicians as “at-risk for driving impairment”, and forty-eight control participants aged 54 to 81 years (M = 65.85, SD = 6.93) completed an on-road driving test designed specifically to evaluate cognitive decline related to driving, and a set of tablet-based cognitive tasks (composed of reaction speed, decision making, memory, and bi-manual perceptual-motor tasks) that measured the cognitive skills needed during driving. Accuracy and reliability of predicting unsafe drivers based on the cognitive tasks were investigated using different trichotomous classifiers (class outputs: safe, unsafe, undefined). ResultsTrichotomous naive Bayes demonstrated the highest overall accuracy performance of 73%, a sensitivity of 69%, and a specificity of 75%. The rate of misclassified unsafe drivers was 19%, and the rate of misclassified safe drivers was 8%. ConclusionHigh accuracy and reliable prediction of unsafe drivers using cognitive-only tasks in a sample of older adults population demonstrate the efficacy of a widely available screening tool that can be applied in other cognitively impaired populations such as drug users.