Toward Reducing the Possibility of False-Positive Results in Epidemiologic Studies of Traffic Crashes

44 Large databases, such as the U.S. Fatality Analysis Reporting System (FARS), are often used to study factors that influence traffic crashes that result in injuries or deaths, and how they may be affected by human activities or environmental conditions. This genre of epidemiologic research presents considerable study design challenges, since crash rates vary by season, day of the week, and time of day. To minimize the effects of these extraneous factors, many studies use a double control matched design as laid out by Redelmeier and Tibshirani in the Journal of Clinical Epidemiology in 2017. This design identified the particular days (or portions of a day) in which the condition/factor of concern was present. The concern might relate to a national election or holiday, or Super Bowl Sunday, or the change to/from daylight savings time (DST). For each such day of concern, and for each year for which data are available, two comparison days—seven days before and seven days after—are identified. The notation for the counts of interest on these three days is shown in Figure 1. The statistical methods used to convert the summed counts (C, B, A) in Figure 1 into inferential Toward Reducing the Possibility of FalsePositive Results in Epidemiologic Studies of Traffic Crashes