Estimating seat belt effectiveness using matched-pair cohort methods

https://doi.org/10.1016/S0001-4575(01)00087-2Get rights and content

Abstract

Using US data for 1986–1998 fatal crashes, we employed matched-pair analysis methods to estimate that the relative risk of death among belted compared with unbelted occupants was 0.39 (95% confidence interval (CI) 0.37–0.41). This differs from relative risk estimates of about 0.55 in studies that used crash data collected prior to 1986. Using 1975–1998 data, we examined and rejected three theories that might explain the difference between our estimate and older estimates: (1) differences in the analysis methods; (2) changes related to car model year; (3) changes in crash characteristics over time. A fourth theory, that the introduction of seat belt laws would induce some survivors to claim belt use when they were not restrained, could explain part of the difference in our estimate and older estimates; but even in states without seat belt laws, from 1986 through 1998, the relative risk estimate was 0.45 (95% CI 0.39–0.52). All of the difference between our estimate and older estimates could be explained by some misclassification of seat belt use. Relative risk estimates would move away from 1, toward their true value, if misclassification of both the belted and unbelted decreased over time, or if the degree of misclassification remained constant, as the prevalence of belt use increased. We conclude that estimates of seat belt effects based upon data prior to 1986 may be biased toward 1 by misclassification.

Introduction

Several studies have estimated the effects of seat belts on fatality or injury in road traffic crashes (Robertson, 1976, National Highway Traffic Safety Administration, 1984, National Highway Traffic Safety Administration, 1999, Evans, 1986b, Evans, 1990, Evans, 1996, Rivara et al., 2000). In 1984, the National Highway Traffic Safety Administration (1984) analyzed the results of several crash studies and concluded that when occupants wearing three-point restraints were compared with unbelted occupants, the relative risk of death was about 0.55; the agency still uses this estimate of effectiveness (1999). In 1986, Evans (1986b) estimated that the relative risk of death for a belted front-seat occupant was 0.58 compared with an unbelted front-seat occupant. More recently, Rivara et al. (2000) estimated that front-seat occupants wearing manual shoulder plus lap belts had a relative risk of death, compared with unbelted occupants, of 0.27; this is considerably different from the estimate in earlier studies.

One method for estimating effects of restraints is to take advantage of the fact that crashes often involve drivers and right front-seat passengers in the same vehicle. By estimating effects among these naturally matched-pairs, we can account for many potential confounding variables that otherwise might be costly, difficult, or impossible to measure; Evans (1986a) noted this when he introduced his double-pair method. Persons in the same vehicle at the time of a crash are matched in regard to all factors related to the vehicle and the crash: car make and model, speed, type of collision, response time of ambulance personnel, proximity to a hospital, and so on. Within these pairs of front-seat occupants, there may still be differences in regard to individual-level factors such as age, gender, and seat position. If we can measure these few potential confounding factors that are not accounted for by matching, we can control for these by using stratified or regression methods (Cummings et al., 2003). Thus, an analysis among matched-pairs offers the potential to efficiently control for many variables.

In this study we used matched-pair methods to assess the effectiveness of seat belt use in preventing death among front-seat occupants in crashes. In particular, we wished to understand why a recent study of seat belt use found them to be more effective than studies done in the 1980s.

Section snippets

Study sample

We used data from the Fatality Analysis Reporting System (FARS), a database maintained by the National Highway Traffic Safety Administration which contains information about all motor vehicle crashes on public roadways that result in the death of at least one occupant or non-motorist within 30 days of the crash. We selected records from calendar years 1975 through 1998. In some states the proportion of records that were missing seat belt data was large. We excluded from the study sample the

Main analysis, based upon calendar years 1986 through 1998

During 1986 through 1998 there were 53,311 study vehicles that crashed. Seat belt information was missing for 8.2% of the occupant records, leaving 47,580 vehicles for analysis.

Study occupants were usually male (61%). Their mean age was 38.2 years and median age was 29 years. Seat belts were used by 31.6% of the occupants. Among belted occupants, 51.5% were male, 51.1% were drivers, their mean age was 46.3 years, and 52.1% died. Among unbelted occupants, 65.5% were male, 49.5% were drivers,

Discussion

We estimated that among front-seat occupants in car crashes during 1986 through 1998, the relative risk of death was 0.39 among those using a seat belt compared with those not using a belt. This relative risk estimate is considerably different from the estimate of 0.55 based on crash data collected prior to 1986 (National Highway Traffic Safety Administration, 1984, Evans, 1986b).

The difference between the older and newer estimates was not explained by our analytic method; when we restricted

Acknowledgements

This work was supported in part by Grant R49/CCR002570 from the Centers for Disease Control and Prevention, Atlanta, GA. We thank Chris Mack for extracting the FARS data that we used for the analysis and Andrés Villaveces MD, MPH, for sharing his data regarding seat belt laws. The opinions and assertions contained herein are the private views of the authors and are not to be construed as official or as reflecting the views of the Department of Defense, the US Army, the US Army Medical

References (19)

There are more references available in the full text version of this article.

Cited by (79)

  • The Safe Systems Pyramid: A new framework for traffic safety

    2023, Transportation Research Interdisciplinary Perspectives
  • How would increasing seat belt use affect the number of killed or seriously injured light vehicle occupants?

    2016, Accident Analysis and Prevention
    Citation Excerpt :

    Cummings et al. (2003) have investigated several possible explanations for improved seat belt effectiveness over time (differences in analysis methods, changes of vehicle characteristics or crash characteristics), and concluded that the most likely explanation is that earlier estimates of seat belt effects may be underestimated because of misclassifications of seat belt use, especially prior to 1986. Seat belts in more recent cars models (the study by Cummings et al., 2003, is based on data from 1975 to 1998) may actually have become more effective, e.g. because of increased use of pretensioners and load limiters. Although it seems reasonable to find improved effects of seat belts over time, there is one factor that may have contributed to an overestimation of the effectiveness of seat belts among front seat occupants.

  • Injury risk for matched front and rear seat car passengers by injury severity and crash type: An exploratory study

    2015, Accident Analysis and Prevention
    Citation Excerpt :

    By matching front and rear seat passengers in the same vehicle in the same crash, a matched cohort study design is able to control for possible vehicle and crash-related confounding characteristics such as vehicle model and vehicle speed (Smith and Cummings, 2006). Many of the vehicle matched-cohort studies have examined risk of death between rear and front seat passengers (Cummings et al., 2003a,b; Smith and Cummings, 2004, 2006) or have examined risk of serious injury and mortality (Bilston et al., 2010). Further work is needed to compare rear and front seat car passengers for finer levels of injury severity (Brown and Bilston, 2014), such as by using detailed injury severity categories (e.g. using six severity categories) or by using broad-level injury severity categories (e.g. using three severity categories).

View all citing articles on Scopus
View full text