V Eksler1, S Lassarre, I Thomas. 1. French National Institute for Road Safety Research, Group for the Analysis of Road Risk and its Governance, 77420 Champs-sur-Marne, France. elksler@inrets.fr
Abstract
BACKGROUND: Road accidents are the tenth leading cause of premature death worldwide and, based on current trends, are likely to become the third leading cause of disability-adjusted life years lost by 2020. Road mortality varies in time and space both between countries and also between regions within the same country. Identifying and understanding the background of regional differences may lead to better understanding of the sources of road accidents, and enable the application of more effective road safety policies. METHODS: A Bayesian ecological regression model based on a unified generalized linear mixed model framework is introduced. Population density and country (affiliation) were used as covariates and were fitted into the model at the four levels of spatial aggregation known as 'nomenclature of statistical territorial units (NUTS) regional classification'. RESULTS: Population density has a significant influence on road mortality. For all countries together, the elasticity estimate is -0.32, meaning that a 10% increase in population density is linked to a 3.2% decrease in road fatalities. A multi-level model defined at the NUTS-3 level, taking into account the NUTS-2 aggregation, enables infraregional variances in road mortality to be taken into account and produces the most reliable estimates of the model parameters. Variation in the Bayes relative risk (the mortality ratio 'standardized' by population density and country effect) is highest at the NUTS-3 level, but is lower at country level and NUTS-2 level, which suggests that other important underlying factors are responsible for the variations in road mortality between regions. Mapping the Bayes relative risk enables the identification of regions that should be targeted by national and regional policies. Last but not least, a new ranking of European countries according to their road mortality risk, adjusted for population density, is presented.
BACKGROUND: Road accidents are the tenth leading cause of premature death worldwide and, based on current trends, are likely to become the third leading cause of disability-adjusted life years lost by 2020. Road mortality varies in time and space both between countries and also between regions within the same country. Identifying and understanding the background of regional differences may lead to better understanding of the sources of road accidents, and enable the application of more effective road safety policies. METHODS: A Bayesian ecological regression model based on a unified generalized linear mixed model framework is introduced. Population density and country (affiliation) were used as covariates and were fitted into the model at the four levels of spatial aggregation known as 'nomenclature of statistical territorial units (NUTS) regional classification'. RESULTS: Population density has a significant influence on road mortality. For all countries together, the elasticity estimate is -0.32, meaning that a 10% increase in population density is linked to a 3.2% decrease in road fatalities. A multi-level model defined at the NUTS-3 level, taking into account the NUTS-2 aggregation, enables infraregional variances in road mortality to be taken into account and produces the most reliable estimates of the model parameters. Variation in the Bayes relative risk (the mortality ratio 'standardized' by population density and country effect) is highest at the NUTS-3 level, but is lower at country level and NUTS-2 level, which suggests that other important underlying factors are responsible for the variations in road mortality between regions. Mapping the Bayes relative risk enables the identification of regions that should be targeted by national and regional policies. Last but not least, a new ranking of European countries according to their road mortality risk, adjusted for population density, is presented.