A recent study found that drowsy driving prevalence in U.S. national crash databases is substantially higher than previously estimated, especially for fatal crashes. The aims of the current study are to merge this result with a new estimate of the prevalence and odds ratio (OR) of drowsy driving in the 100-Car naturalistic driving study (NDS), and investigate interactions with secondary tasks, particularly cell phone conversation. A 2010 NDS study matched baseline video clips to crash/near-crash video clips for driver demographics, time of day, and GPS location. Using that matched baseline to remove bias from those variables, the current study estimates the drowsy driving OR for crashes to be 63, substantially higher than previous estimates. In addition, observable moderate to severe drowsiness causes an estimated 20% of all crashes, while non-observable microsleeps likely elevate that percentage. A logistic regression analysis on the 100-Car data found no interaction between drowsy driving and secondary tasks as a whole. However, the moderately-difficult task group (which includes cell phone conversation) reduced the drowsy driving crash/near-crash OR, as did cell phone conversation alone. These new NDS analyses provide preliminary evidence that curtailing drowsy driving will reduce more crashes than curtailing secondary tasks while driving. (Author/publisher)
Abstract