Abstract:Driving at different times of day within 1 day and over several days is associated with different levels of accident risk. Analyses of accident and nonaccident data from a less-than-truckload carrier representing 6 months of operation in 1984 are used to explore changes in daily and multiday accident risk. Cluster analysis is used to extract a distinct pattern of driving over a 7-day period from a sample of 1,066 drivers (including those with accidents and nonaccidents on the eighth day). The analyses yielded clear interpretable driving patterns that could be associated with levels of relative accident risk. Higher risk was generally, but not exclusively, associated with extensive driving in the 2 to 3 days before the day of interest. The two patterns with the highest risk of an accident were those that contained heavy driving during the preceding 3 days and consisted of driving from 3:00 p.m. to 3:00 a.m. (Pattern 1) and from 10:00 p.m. to 10:00 a.m. (Pattern 8). The lowest risk was associated with driving from 8:00 p.m. to 6:00 a.m. but with limited driving on the preceding 3 days. Given the virtually limitless possible combinations of driving schedules, it is encouraging that interpretable distinct multiday patterns could be extracted from a data base of more than 1,000 observations. Within each pattern, drivers experienced similar duty hours: cumulative driving during the 7 days ranged from 47 to 49 hr. Continuous driving (between mandatory 8-hr off-duty periods) ranged from 7.8 to 8.4 hr. Individual drivers also experienced a cycle of on-duty and off-duty time that ranged from 22.3 to 23 hr, closer to the 24-hr period that is desirable from the perspective of human performance theories. The findings suggest that it is possible to identify and extract patterns of multiday driving and that these patterns are associated with different levels of accident risk. Additional empirical tests and the development of refined accident risk models are suggested for future research.

Note:This is a condensed version of RR-90-10 (PubID 1022).