Vehicles in fatal crashes can be studied for indications of potential defects using an "early warning" surveillance statistic previously suggested for screening reports of adverse drug reactions. This statistic is illustrated with time series data for fatal, tire related and fire related crashes. Geographic analyses are used to augment the tire related statistics. A statistical criterion based on the Poisson distribution that tests the likelihood of an expected number of events, given the number of events that actually occurred, is a promising method that can be readily adapted for use in injury surveillance. Use of the demonstrated techniques could have helped to avert a well known injury surveillance failure. This method is adaptable to aid in the direction of engineering and statistical reviews to prevent deaths and injuries associated with potential motor vehicle safety defects using available databases. (Author/publisher).
Abstract