Apart from injury underreporting (i.e. casualties unavailable in police records, but possibly available in hospital records), it is also acknowledged that there exists an injury severity reporting inaccuracy problem (“misreporting”), covering in many countries for over 50% of all injuries, especially slight ones. The objective of this research is the analysis of injury severity misreporting in European countries, on the basis of in-depth fatal accident investigation data collected within the SafetyNet integrated research project. In this dataset, two distinct classifications are available concerning injury severity: “Police injury severity” and “SafetyNet medical outcome”, i.e. as validated or corrected on the basis of additional data sources. After a thorough exploratory analysis of the data, logistic regression models were developed, in which the dependent variable indicated whether injury severity matched between the two classifications or not. The probability of misreporting injury severity was correlated with a number of road user, vehicle, road and accident characteristics. Overall, the data included several cases presenting injury severity misreporting. A general trend could be identified, according to which, the more complex the accident (e.g. higher traffic) and the accident site (e.g. junction, daytime), and the more vulnerable the road user (e.g. children, elderly, pedestrians), the higher the probability of injury severity score to be different between the police and SafetyNet. The results also suggest that score differences may be due either to recording bias (e.g. the Police tending to misreport injury severity incorrectly under certain conditions), or to a general difficulty in identifying the correct severity score in some cases. The results of such analyses may be a very first step towards the development of correction coefficients for injury severity misreporting.
|Tags||accident analysis, statistical modelling|