BACKGROUND: A multitest approach is optimal for the identification of at-risk driving among older adults. This study examined the predictive validity of a combination of office-based screening tests for on-road driving performance in older adults with and without mild cognitive impairment (MCI)/dementia. METHODS: Forty-four normal control, 20 participants with MCI, and 20 participants with dementia completed a battery of office-based assessments. On-road driving evaluation classified participants as not at-risk (n = 65) or at-risk drivers (n = 19). RESULTS: Logistic regression revealed age and 2 tests of visual attention abilities (Useful Field of View [UFOV] Divided Attention and Neuropsychological Assessment Battery [NAB] Driving Scenes) best predicted at-risk drivers ( C statistic = 0.90); no cutoff score had both sensitivity and specificity >80%. CONCLUSIONS: Future research on larger and more clinically representative neurological samples will improve understanding of the utility of the UFOV Divided Attention and NAB Driving Scenes in detecting at-risk older adult drivers in the clinic.