Purpose: To evaluate agreement between police and trained investigators regarding seat belt use by crash victims, according to injury severity.
Methods: We used data from the National Accident Sampling System Crashworthiness Data System (CDS) for front seat occupants, 16 years and older, in crashes during 1993-2000. Crashworthiness Data System investigators determined belt use from vehicle inspection, interviews, and medical record information; their assessment was considered the gold standard for this analysis. Occupant severity of injury was categorized in five levels from no injuries to death. We estimated the sensitivity, specificity, and area under receiver operating characteristic curves for police reports of belt use.
Results: Among 48,858 occupants, sensitivity of a police report that a belt was used was 95.8% overall and varied only modestly by injury severity. Specificity of a police report that a belt was not used was 69.1% overall; it was the lowest among the uninjured (53.2%) and greatest among the dead (90.4%). The area under the curve was 0.82 (95% confidence interval 0.82-0.83) overall; this was lowest among those not injured (0.75, 95% confidence interval 0.74-0.76) and increased with injury severity to 0.91 (95% confidence interval 0.90-0.93) among those who died.
Conclusion: Police usually classify belted crash victims as belted, regardless of injury severity. But they often classify unbelted survivors as belted when they were not. This misclassification may result in exaggerated estimates of seat belt effectiveness in some studies.