Recent epidemiologic studies have estimated little or no increased risk of automotive crashes related to cell phone conversations by the driver, whereas earlier case-crossover studies estimated the relative risk as close to 4. Did earlier studies introduce a positive bias in relative risk estimates by overestimating driving exposure in control windows? Driving exposures in a "control" window and a corresponding "case" window on the subsequent day were tabulated across 100 days for 439 GPS-instrumented vehicles in the Puget Sound area during 2005-2006. For control windows containing at least some driving, driving exposure was about one-fourth that of case windows. Adjusting for this imbalance reduces relative risk estimates in the earlier case-crossover studies from 4 to 1. Earlier case-crossover studies likely overestimated the relative risk for cell phone conversations while driving by implicitly assuming that driving during a control window was full-time when it may have been only part-time.
Abstract