ARSRPE Conference Paper Database

Using statistical modelling to predict crash risks, injury outcomes and compensation costs in Victoria

Schuster, R, Nieuwesteeg, Michael, Northrop, A, Lucas, C, Smith, D

Data / Research Methods


Background: In 2011, Victoria’s Transport Accident Commission (TAC) built a rich linked crash database to explore the research question: “What are the significant variables in predicting crash risk, injury outcomes and compensation costs when controlling for all other variables”? Aims: The core aims of the TAC Road Safety Risk Models project were to conduct sophisticated analyses of available data to identify key drivers of road trauma, injury severity and compensation costs, as well as identify key target markets. Method: The project began with an intense data build involving the sourcing, linking and cleansing of road safety and related data. This included crash and compensation data, as well as exposure data on Victorian licence holders and registered vehicles. Detailed injury data was also obtained. A series of statistical models were then developed to examine the relationship between person, vehicle and crash variables, along with injury severity and compensation costs. Results: A number of pre-crash variables were found to be significant predictors of crash risk and severity including vehicle, person and geo-demographic variables. Injury severity was found to be the most significant variable at predicting compensation costs. Conclusions: The established database provides a benchmark for future Road Safety policy analysis, particularly with consideration given to the cost of injury to society. With the prospect of new and improved data availability for key input datasets, the TAC has begun to update the linked dataset and refresh the models to identify new relationships.