Comparison of reporting of seat belt use by police and crash investigators: variation in agreement by injury severity
Introduction
In studies of seat belt use and traffic crash outcomes, errors in assessment of seat belt use may bias the estimates of any benefit afforded by wearing a belt in a crash. Seat belt misclassification is differential when the proportion misclassified is related to an outcome, such as death (Kelsey et al., 1996). Some authors have suggested that differential misclassification has produced exaggerated estimates of benefit in some studies, biasing the risk ratios toward 0; this might have occurred either because some dead crash victims were incorrectly classified as unbelted, or some survivors were incorrectly classified as belted, or both (Robertson, 1976, Partyka, 1988, Kahane, 2000, Robertson, 2002). Misclassification is non-differential when errors in assessing belt use are unrelated to the study outcome. Non-differential misclassification tends to bias risk ratios toward 1 (Kelsey et al., 1996), resulting in an underestimate of any benefit associated with seat belt use.
One of us has reported that both differential and non-differential misclassification of belt use occurs to some degree, but the amount of error in recent data suitable for a matched-cohort analysis was trivial, bias toward 1 and 0 was balanced, and the misclassification did not appreciably influence the risk ratio estimate (Cummings, 2002). Although that study was limited to front-seat occupant pairs in which one or both died, not all studies of seat belts are limited to those pairs.
In this paper, we compared the assignment of seat belt status as recorded by police with the assignment made by trained investigators who relied chiefly on vehicle inspection to determine belt use after a crash. We assessed how agreement on seat belt use varied by the injury severity of the crash victims. Several studies have estimated the association between seat belt use and death using a matched-pair cohort study design that was limited to front seat occupants in the same vehicle (Evans, 1986a, Evans, 1987, Kahane, 1986, Kahane, 2000, Cummings et al., 2003a); this design can estimate seat belt effectiveness for all pairs, using only information from pairs with at least one death (Evans, 1986b; Cummings et al., 2003b, Cummings et al., 2003c). We therefore limited our analysis to front seat occupant pairs in the same vehicle.
Section snippets
Methods
The National Highway Traffic Safety Administration collects information on a sample of all motor vehicle crashes reported to the police in the United States. Approximately 5000 crashes are investigated annually by trained investigators and the crash data are entered in the National Accident Sampling System Crashworthiness Data System (CDS) (NASS, 2001). The CDS contains data on seat belt use from two sources, police crash reports and CDS investigators. We evaluated the CDS data for all front
Results
The CDS data set contained 49,115 front seat occupants, aged 16 years or older involved in motor vehicle crashes with complete information on age, injury severity, and seat belt use as reported by police and CDS investigators. The mean age of occupants was 36.4 years and 55.4% were male. Most occupants were drivers (78.2%) and 70.1% of occupants were belted according to the CDS investigators. About one-fourth (28.7%) had no injuries, 54.6% had minor injuries (ISS 1–8), 7.5% had moderate
Discussion
We found that agreement between CDS investigators and police on the reporting of seat belt use and non-use in a motor vehicle crash increased with increasing injury severity among occupants of fatal and non-fatal crashes combined as measured by κ and the area under receiver operator characteristic curves. Agreement between CDS investigators and the police did not vary in a systematic way with injury severity among occupants in vehicles with a fatality.
Some investigators (Robertson, 1976, Evans,
Acknowledgements
This work was supported by grants R49/CCR002570 and R49/CCR019477-01 from the Centers for Disease Control and Prevention, Atlanta, GA. We would like to thank Christopher Mack for programming assistance with the CDS data files and Robert Kaufman, a CDS crash investigator for the past 12 years, for his advice on CDS procedures and variables, and comments on this manuscript.
References (20)
- et al.
Estimating seat belt effectiveness using matched-pair cohort methods
Accid. Anal. Prev.
(2003) - et al.
Matched-pair cohort methods in traffic crash research
Accid. Anal. Prev.
(2003) The effectiveness of safety belts in preventing fatalities
Accid. Anal. Prev.
(1986)Double pair comparison—a new method to determine how occupant characteristics affect fatality risk in traffic crashes
Accid. Anal. Prev.
(1986)- et al.
The injury severity score: a method for describing patients with multiple injuries and evaluating emergency care
J. Trauma Inj. Infect. Crit. Care
(1974) - Cameron, A.C., Trivedi, P.K., 1998. Regression Analysis of Count Data. Cambridge University Press, New York, pp....
Association of seat belt use with death: a comparison of estimates based on data from police and estimates based on data from trained crash investigators
Inj. Prev.
(2002)- et al.
Matched cohort methods in injury research
Epidemiol. Rev.
(2003) - et al.
Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach
Biometrics
(1988) Fatality risk reduction from safety belt use
J. Trauma
(1987)