Crash severity is a key predictor of injury outcome in crashes. The best available measure of crash severity is delta-V, the nominal change in velocity experienced by a vehicle involved in a crash. Although delta-V is expressed as a velocity, it represents the energy absorbed by each vehicle and accounts for mass and stiffness differences between vehicles involved in multi-vehicle crashes. The National Automotive Sampling System (NASS) includes three major datasets that are sampled annually from crashes in the U.S. The General Estimates System (GES) is a complex stratified sample of 50,000 police-reported crashes in the U.S. This dataset includes only information available from the police report, which does not include delta-V. The Crashworthiness Data System (CDS) is a complex stratified sample of 5,000 towaway crashes involving light vehicles. This dataset includes details of in-depth crash investigations performed on vehicles selected for inclusion. Delta-V is included in CDS. The Fatality Analysis Reporting System (FARS) is a census of all fatal crashes that occur on public roads in the U.S. Most information in FARS comes from police reports, though there is some additional phone follow-up. No delta-V information is available in FARS. When injury data are based on police reports, they are limited to the categories of killed (K), incapacitating injury (A), non-incapacitating injury (B), possible injury (C), or no injury (O). Each dataset has a different data collection approach, and only CDS includes delta-V. Delta-V estimates are calculated by measuring vehicle crush and entering the measurements and other vehicle information into the WinSMASH (Sharma et al. 2007) program. WinSMASH uses vehicle mass and dimensions, as well as stiffness coefficients based on crash tests, to estimate the delta-V required to produce the observed level of crush. An alternative method of measuring delta-V is to download recorded acceleration from Event Data Recorders (EDRs). At this time, only some vehicles are equipped with EDRs, and only a small portion record lateral acceleration. Most crashes (other than fullfrontals) have a lateral component of delta-V. CDS includes EDR reports for any vehicles for which they have received permission. However, the EDR reports are only available for a nonrandom subset of CDS case vehicles. EDR reports are not available for GES or FARS. Although EDRs provide promise for collecting measured delta-V, they will not replace deltaV based on in-depth damage measurements in the near future. While CDS includes delta-V, the database is limited to towaway crashes and light vehicles. It would often be useful to know delta-V distributions for GES or FARS, which include different populations of crashes, since analysis of outcomes or safety benefits must account for crash severity. To this end, researchers have attempted to estimate crash severity for individual cases in GES using injury outcome. Farmer (2003) compared two commonly used measures of crash severity from police reports to corresponding measures from accident investigations in CDS. First, he compared posted speed limit to delta-V. Posted speed limit is often used as a proxy for crash speed (or crash severity) in analyses of GES data. Although Farmer found a relationship between posted speed and delta-V, it is too weak to reasonably use posted speed limit as a substitute for the probable delta-V in a given GES case. Farmer also investigated injury coding by police compared to outcomes recorded in medical records, but he did not explicitly explore injury outcome as an indication of crash severity. Farmer’s goal was to identify a substitute measure of crash severity for individual cases when delta-V is not available. Although it would be ideal to have an adequate delta-V surrogate for each case, an estimate of the delta-V distribution for a class of cases would be useful in several other circumstances. For example, evaluations of NHTSA’s crash testing program commonly compare the test speed to field distributions of delta-V for similar crashes to understand what real-world benefit might result from crash testing (Hackney et al (1996), Newstead et al (1996), Arbalaez et al, 2005). Delta-V distributions could also prove beneficial when estimating safety benefits for any countermeasure (e.g., forward collision warning, crash-imminent braking) that might reduce delta-V without eliminating a crash altogether. This paper presents a method that uses injury patterns in GES, combined with the relationship between delta-V and injury from CDS, to estimate the distribution of delta-V for a group of similar crashes. The method does not estimate delta-V for individual crashes, but instead identifies general patterns of the relationship between delta-V and injury that should hold for a larger population of similar crashes. (Author/publisher)
Samenvatting