Standard methods of investigating real-world crashes are hampered by the rapid rate at which the vehicle fleet changes as well as logistical hurdles involved in collecting sufficient quantities of data regarding specific vehicle and crash conditions to draw useful conclusions regarding injury causation. This degrades the ability of real-world crash data to contribute in a timely fashion to the assessment and improvement of vehicle and occupant protection systems.

The University of Michigan Health System, General Motors and OnStar are collaborating on a project to collect real-world crash data using the OnStar system to identify and screen crash cases from around the US. For crash events of interest, informed consent is obtained, medical interviews are conducted and the vehicle is inspected for photographic documentation. Medical records and digital medical imaging data files are also obtained for determination of injury mechanism and outcome.

Most real-world crash data collection systems have limitations. Systems in which a small subset of crashes is randomly sampled have very limited numbers of crashes from specific vehicle models and crash conditions. Geographically based census collection systems can have the same limitation. Medically based crash data collection systems provide optimal detail and insight regarding injury causation factors, but are also biased by being outcome-sampled and expensive. The novel use of advanced automatic collision notification technology for screening allows researchers to very efficiently identify the subset of real-world crash cases that hold most value for assessment of injury risk or evaluation of vehicle safety performance. Cost effectiveness will increase even further once photographic documentation of crash damage is no longer necessary. The involvement of independent, academically based medical researchers significantly enhances subject enrollment and enables the collection of sensitive medical records and digital imaging data.