Highlights • Driving is essential for quality of life for older adults in the US. • Age-related declines lead older adults to self-restrict driving. • Self-reported and actual driving in challenging situations are mildly inconsistent. • Cognitive and physical declines are linked to less driving in challenging situations. • Older adults living alone drive more in challenging situations.

Abstract Older adults in the United States rely heavily on driving their own vehicles to commute to work, shop for groceries, and access public services. To effectively help older adults maintain mobility and independence,we need to better understand how thecognitive, visual functioning, and health declines influence their tendency to self-restrict their driving. The objective of this study is to develop a causal model to examine the effects of age, gender, household status (specifically living alone), physical, cognitive, visual abilities, and health status on older adults’ driving mobility in terms of driving exposure and avoidance. Driving exposure was measured by actual driving data, whereas driving avoidance was assessed by both self-report data and actual driving exposure to challenging situations. Structural equation modeling was used to analyze data collectedin the Second Strategic Highway Research Program Naturalistic Driving Study for establishing relationships between the selected factors and mobility. The structural equation model included a total of794 participants aged 65 and over (367 or 46.22%femalesand 427 or 53.78% males). Results indicate that poorer health is associated with less driving exposure; deteriorating cognitive and physical capabilities are associated with more self-reported driving avoidance and less actual driving in challenging situations; visual function is associated with self-reported avoidance; living alone is associated with higher driving exposure in general as well as in challenging situations; self-reported driving avoidance of challenging situations has a negative association with actual driving in those same situations. The final model could be applied to predict older adults’ mobility changes according to their age, gender, household status, as well as their visual, physical, cognitive and health status.