摘要：Objectives. We estimated the effects of snowfalls on US traffic crash rates between 1975 and 2000. Methods. We linked all recorded fatal crashes (1.4 million) for the 48 contiguous states from 1975 through 2000 to daily state weather data. For a subsample including 17 states during the 1990s, we also linked all recorded property-damage-only crashes (22.9 million) and nonfatal-injury crashes (13.5 million) to daily weather data. Employing negative binomial regressions, we investigated the effects of snowfall on crash counts. Fixed effects and other controls were included to address potential confounders. Results. Snow days had fewer fatal crashes than dry days (incidence rate ratio [IRR] = 0.93; 95% confidence interval [CI] = 0.90, 0.97), but more nonfatal-injury crashes (IRR = 1.23; 95% CI = 1.18, 1.29) and property-damage-only crashes (IRR=1.45; 95% CI=1.38, 1.52). The first snowy day of the year was substantially more dangerous than other snow days in terms of fatalities (IRR = 1.14; 95% CI=1.08, 1.21), particularly for elderly drivers (IRR=1.34; 95% CI=1.23, 1.50). Conclusions. The toll of snow-related crashes is substantial. Our results may help estimate the potential benefits of safety innovations currently proposed by meteorology and traffic safety experts. Poor weather-related driving conditions are associated with 7000 fatalities, 800 000 injuries, and more than 1.5 million vehicular crashes annually in the United States. 1 Adverse weather is present in 28% of total crashes and nearly 20% of highway fatalities. 2 Analysts estimate the economic toll of weather-related crashes at $42 billion. 3 Understanding the effects of adverse weather on motor vehicle crashes matters because experts have identified a number of communications and engineering innovations (largely technologies to collect and communicate real-time road condition information, such as sensors and dynamic message signs) that could significantly reduce the crash and injury rates, but at a potentially substantial cost. 1, 2 Previous studies have associated precipitation with markedly increased crash rates. 4– 7 Recent work also shows that the risk posed by precipitation rises dramatically with the time since last precipitation. 4 Less well studied to date are the effects of a specific form of precipitation, snowfall. Crash counts are not inevitably higher in snowy weather than in dry weather. On the one hand, snow makes driving more dangerous, by reducing tire adherence and impairing visibility. On the other hand, experienced drivers typically drive more slowly and carefully in snowy weather, and many people avoid or postpone unnecessary travel. Perhaps as a reflection of these offsetting factors, the handful of published studies addressing the crash consequences of snow has produced some conflicting results. The weight of the evidence suggests that less severe crashes (e.g., those producing only property damage) increase during snows, while more severe crashes (those resulting in fatalities) decrease. Significantly increased crash rates have been documented in snowy months in Canada, 8 on snowy days in the United Kingdom, 9 and during snowstorms in Iowa. 10 Perry and Symons 9 found increased rates of crashes involving injuries and fatalities on snowy days in the United Kingdom, but Brown and Baass 11 noted fewer crashes involving injuries in the winter months in Canada, as did Fridstrom et al. 7 in snowy months in Denmark and Finland. Eisenberg 4 found decreased rates of fatal crashes on snowy days in the United States, a finding echoed in analysis of winter months in Canada 11 and snowy months in Scandinavia. 7 To date, only 2 previous studies have examined the effects of the first snowfall of the season. Defining first snowfall as the first snow in a month following a month without snow, Fridstrom and Ingebrigtsen 12 found significant increases in both injury and fatal crashes in Norway. Subsequently, however, research by Fridstrom and colleagues 7 produced mixed findings: injury crashes rose significantly during the winter’s first month with snow (compared with other months with snow) in Denmark but not in either Finland or Norway. Fatality rates were no different in the first snowy month than in other snowy months. We present findings from the first detailed analysis of the impact of snowfall on crash rates by severity level (property damage only, injury, and fatal) in the entire United States (excluding Hawaii, Alaska, and the District of Columbia), using a data set combining daily data on traffic crashes and snowfall for the period 1975 through 2000. Of special interest are the effects of first snowfalls of the season, which might be expected to affect crashes and their outcomes differently from later snowfalls. Many people may be unprepared to avoid driving when the first snowfall occurs each year; others may not adapt driving procedures, such as reducing speed and braking earlier, as completely as they will later in the snow season. We also present the first analysis of the differential effects of snowfall on drivers of different ages.
