Are Rear-end Crashes Caused Mainly by an Interaction between Glance Duration and Closure Rate?

Victor and colleagues recently analyzed rear-end crash and near-crash data from the Strategic Highway Research Program Phase 2 naturalistic driving study. They measured the last off-path glance duration just before a crash, and the closure rate (the change in looming rate during the last glance). They concluded that the predominant cause of rear-end crashes was a “mismatch”– a short glance with a high closure rate or a long glance with a low closure rate. The current study independently tested this “mismatch” hypothesis using two epidemiological methods to estimate odds ratios for combined crashes and near-crashes (events). First, the glance and closure rates were stratified, and compared to a “just following,” no-crash baseline (short glances <0.5 s and closure rates near 0). Second, a logistic regression analyzed the data on continuous scales. Both tests confirmed a strong interaction between glance duration and closure rate. At closure rates <0.056 s, glances decreased event risk relative to baseline, proportional to glance duration. At closure rates >0.056 s, glances increased event risk, proportional to glance duration. However, a major data limitation that potentially upwardly biased the interaction OR estimate is an inherent dependency of closure rate on glance duration, simply because of the way closure rate was defined. The mismatch hypothesis for rear-end events must be tested with other rear-end event datasets not subject to this limitation before being considered fully validated.