Explaining regional disparities in traffic mortality by decomposing conditional probabilities.

Background: In the USA, the mortality rate from traffic injury is higher in rural and in southern regions, for reasons that are not well understood.

Methods: For 1754 (56%) of the 3142 US counties, we obtained data allowing for separation of the deaths/population rate into deaths/injury, injuries/crash, crashes/exposure and exposure/population, with exposure measured as vehicle miles travelled. A 'decomposition method' proposed by Li and Baker was extended to study how the contributions of these components were affected by three measures of rural location, as well as southern location.

Results: The method of Li and Baker extended without difficulty to include non-binary effects and multiple exposures. Deaths/injury was by far the most important determinant in the county-to-county variation in deaths/population, and accounted for the greatest portion of the rural/urban disparity. After controlling for the rural effect, injuries/crash accounted for most of the southern/northern disparity.

Conclusions: The increased mortality rate from traffic injury in rural areas can be attributed to the increased probability of death given that a person has been injured, possibly due to challenges faced by emergency medical response systems. In southern areas, there is an increased probability of injury given that a person has crashed, possibly due to differences in vehicle, road, or driving conditions.