Driving Consistency Errors Overestimate Crash Risk from Cellular Conversation in Two Case-Crossover Studies

The goal of this study is to help resolve the discrepancy in relative risk estimates between recent and early epidemiological studies of call-crash association. Recent epidemiological studies estimate a crash risk for cellular conversation near that of baseline driving – a relative risk of about one. In contrast, two early case-crossover studies estimated a relative crash risk of about four for cellular conversation while driving. One hypothesis to explain this fourfold discrepancy is that the early studies had less driving time in the control window on a day before the crash, than in the crash window just before the crash. This bias in driving exposure translated into relatively lower exposure to cellular conversation during control windows than during crash windows, thereby introducing an overestimate of the relative risk for cellular conversation while driving. To test this hypothesis, the present study developed a new driving consistency index (DCI), which measures the percentage overlap in driving times from one day to the next. The mean DCI for 240 vehicles in a Chicago GPS study with known driving times for two successive days was a surprisingly low 14.8%, substantially below the driving consistency estimates in the early case-crossover studies. After adjustment by the mean DCI, the relative risk estimates for cellular conversation while driving in the early case-crossover studies are about one, resolving the discrepancy with the more recent epidemiological studies