Crash Heterogeneity: Implications for Naturalistic Driving Studies and for Understanding Crash Risks

Abstract

Motor vehicle crashes are heterogeneous in their conditions of occurrence, risk factors, and causal scenarios. Horizontal heterogeneity refers to the many distinct scenarios within any crash severity level. Vertical heterogeneity is seen in the different proportions of characteristics at different severity levels. Evidence is presented for the causal heterogeneity of crashes involving all motor vehicles and specifically for large trucks. If horizontally or vertically defined crash subsets are not representative of other subsets, then findings from them cannot be validly generalized to other populations. Furthermore, crash heterogeneity contradicts a key assumption of the Heinrich Triangle, the assumption that crashes within the triangle have identical or highly similar causal factors regardless of outcome severity. The Heinrich Triangle assumption is explicit in naturalistic driving studies (NDS) capturing mainly noncrash dynamic events and minor crashes below conventional reporting thresholds. NDS causal prevalence estimates have little likely validity in relation to fatal and injury crashes where the preponderance of human harm occurs. NDS external validity could perhaps be improved by post hoc mathematical indexing of captured events to the objective profiles of target populations.