OBJECTIVES Obstructive sleep apnea (OSA) is associated with an increased motor vehicle accident (MVA) risk. Conventional measures of OSA severity do not predict individual risk. Cognitive function tests have failed to incorporate outcomes in risk prediction. We aimed to identify markers of cognitive function for MVA risk prediction in OSA. METHODS OSA patients [n = 114, 75% male, median age 51 (43-61) years, body mass index (BMI) 30 (27-33) kg/m2, apnea-hypopnea index 25 (6-49) n/h, and Epworth Sleepiness (ESS) score 11 (8-16)] were recruited from a sleep laboratory. Two cognitive function tests, the Attention Network Test (ANT) and a modified Oxford Sleep Resistance Test (OSLER) test (GOSLING), were assessed. RESULTS OSA patients with (n = 11) or without (n = 103) a MVA record in the Swedish traffic accident registry were identified. In patients with a MVA, 64% were commercial drivers. In patients with a MVA history, more lapses [42 (5-121) vs. 5 (1-25), P = 0.02] and fewer responses [238 (158-272) vs. 271 (256-277), P = 0.03] to stimuli in the ANT were found. In the GOSLING, the number of lapses was higher (29 (10-97) vs. 7 (2-19), P = 0.01) and the reaction time was longer [462 (393-551) vs. 407 (361-449) ms, P = 0.05]. OSA severity and ESS score poorly predicted MVAs (P > 0.2). CONCLUSIONS We have demonstrated that deficit in sustained attention, assessed by daytime neurocognitive function tests, was associated with MVA risk in OSA patients. We were unable to detect an association between MVA history and severity of OSA or the ESS score. The findings provide a rationale for further development of objective MVA risk assessment tools in OSA.