Combining Data Sets for Analysis of Cycling Safety: How Detailed Exposure Data Can Help Explain Crash Data

Paper i proceeding, 2015

Crash databases are commonly queried to infer crash causation, prioritize countermeasures to prevent crashes, and evaluate safety systems. However, crash databases fail to capture road user behavior before the crash and are not sufficient to estimate crash risk. In Sweden, as in many other countries, crash databases are particularly sterile when it comes to bicycle crashes. In fact, not only are bicycle crashes underreported in police reports, they are also poorly documented in the hospital report. Nevertheless, police and hospital reports an unreplaceable source of information that clearly highlights the surprising prevalence of single‐bicycle crashes and hints to some cyclist behavior, such as alcohol consumption, that may increase crash risk. In this study, we use exposure data from 11 road site stations measuring cyclist flow in Gothenburg to help explain crash data from police and hospital reports and estimate risk. For instance, our results show that crash risk is greatest in weekends at nights, and that this risk is larger for single‐bicycle crashes compared to crashes between a cyclist and another motorist. This result suggests that the population of night‐cycling riders in weekends is particularly prone to specific crash types, which may be influenced by specific contributing factors (such as alcohol), and may require specific countermeasures. More in general, our results demonstrate that exposure data can help select, filter, aggregate, highlight, and normalize crash data to obtain a sharper view of the cycling safety problem for a more fine‐tuned intervention.