Changes in older drivers' risky driving behavior over time: Insights from a naturalistic study.

• Investigated the changes in the risky driving behavior of older drivers with age using naturalistic driving data. • There was heterogeneity in individual trajectories of risky driving behaviors. • The heterogeneity can be explained by the presence of three subgroups of drivers. • Initial levels of risky driving behaviors were much more important than time slopes in identifying subgroups of drivers. • Female, sensation-seeking, and low-mileage drivers were associated with higher driving risk. This study aimed to investigate changes in the risky driving behavior of older drivers with age using naturalistic driving data. The driving behaviors of 44 drivers aged 60 years or older at baseline were tracked for two to five years. Harsh events characterized by elevated g-forces were used as driving outcomes. A multilevel modeling approach was used to examine (1) the heterogeneity in individual trajectories of risky driving behaviors and (2) the factors associated with the average risky driving behavior over the study period and changes in risky driving behavior over time. The results indicated that there was heterogeneity in individual trajectories of risky driving behaviors. Older female drivers and a higher level of sensation seeking (i.e., the tendency to pursue new and different sensations, feelings, and experiences) were associated with higher rates of harsh events over the study period. A reduction in driving exposure was associated with an increase in harsh event rates, and driving exposure can partly explain the sex difference in risky driving behaviors. The cluster analysis indicated that the heterogeneity in individual trajectories of risky driving behaviors can be explained by the presence of three subgroups of drivers, with each group exhibiting specific trajectories. Initial levels of risky driving behaviors were much more important than time slopes in identifying subgroups of drivers, indicating that differences between drivers existed early and remained relatively consistent over the study period. A multinomial logistic regression was used to examine the predictors of class membership identified by the clustering algorithm, and the findings were similar to those from the multilevel regression analysis. The findings of this study enhance our understanding of how older drivers' risky driving behaviors change with age and indicate the distinctive individual processes of change in driving behaviors with age. [ABSTRACT FROM AUTHOR]