| Institution: | Kung-Jong Lui, PhD, Daniel McGee, PhD, Phil Rhodes, MS, and Daniel Pollock, MD, are affiliated with the Biometrics Branch, Division of Injury Epidemiology and Control, Center for Environmental Health and Injury Control, Centers for Disease Control, Public Health Service, U.S. Department of Health and Human Services, 1600 Clifton Road, N.E., Koger, Room 1055, Atlanta, GA 30333, USA |
| Abstract: | In addition to experimental trials in automotive factories, there is a fundamental need to monitor real people involved in real motor vehicle collisions to determine the impact of automotive design characteristics on injury mortality. To aid in designing future safety features for drivers' seats, data from the Fatal Accident Reporting System (FARS) were used to assess the effects of safety belts, directions of crash impacts, age, sex, and car weights on motor vehicle injury fatalities. Furthermore, because the FARS includes only accidents in which there was at least one fatality, this paper introduces a multivariate approach — a logistic regression conditioned on each accident —to avoid the sampling biases inherent in the FARS. The resulting model is used to quantify the relations among the safety belts, the directions of crash impacts, and vehicle weights and their effects on fatalities. Recause the proposed approach allows researchers to study many important variables simultaneously and eliminates the biases resulting from many possible confounders, the estimates presented in this paper are considered to be more precise and more firmly established than earlier estimates. The new approach to analyzing the FARS data will also be useful for investigating the effects of other risk factors or automotive characteristics on crash fatalities. |
