We designed this questionnaire-based study to determine the risk of auto crashes among Alzheimer's disease (AD) patients who continued to drive after the onset of AD, compared with normal age-matched control subjects and other drivers' statistical records. While ultimately all AD patients will become incapable of driving it is not known whether, under current licensing regulations and self-imposed limitations, patients with AD present a definably increased risk of being involved in crashes, and if so, the relative magnitude of the risk and at what point in the course of the disease the risk may become significantly increased. We administered a brief questionnaire to the caregivers of 130 AD patients and to 112 age-matched, nondemented control subjects and their spouses. Annual rates of occurrence and severity of all crashes, and of crashes reported to the authorities, were determined from spousal or other caregiver responses. For all years of driving following the onset of dementia, AD patients had a mean of 0.091 reported crashes per year compared with matched controls, who had an average of 0.040 reported crashes per year in the same period of time. The average number of crashes per year changed with each year of driving following the onset of AD, with considerably lower reported crash rates during the initial years of dementia: in year 1, the crash rate was 0.068; in year 2, 0.097; in year 3, 0.093; in year 4, 0.159; in year 5 and beyond, 0.129. For comparison, registered drivers of all ages have an average of 0.067 reported crashes per year, registered drivers 65 years and older have an average of 0.037 reported crashes per year, while registered 16- to 24-year-old males have an average of 0.148 reported crashes per year. We conclude that (1) the existing evidence suggests that AD patients who drive present a slightly increased risk for crashes compared with drivers of all ages but a lower risk than young unimpaired drivers, especially males. (2) During the first 2 to 3 years after the onset of AD, the magnitude of risk of crashes is well within the accepted risk for other registered drivers. (3) There is marked variability in the degree of disability due to AD and its rate of progression. Because of this, direct tests of driving competence-rather than the diagnosis of AD per se-should be considered as the criterion for continued licensure to drive, with sufficiently frequent retesting to anticipate the expected decline over years.