Enhancing road safety through early detection of outbreaks in the frequency of motor vehicle crashes

Methods for detecting outbreaks in the frequency of particular human-related phenomena have typically monitored daily counts for geographical regions. However, age can also be a significant factor in the frequency distribution of particular phenomena. Using data relating to motor vehicle crashes on public roads, this paper offers a methodology for detecting outbreaks that are age group clustered. The transitional Poisson regression model is used to provide day-ahead forecasts (expected values) for daily crash counts across different age groups. Standardized smoothed count departures from their smoothed day-ahead forecasts across all age groups are used to detect systematic outbreaks. The CUSUM of sequential age group standardized scores is used to signal outbreaks that are age-clustered. Potential applications of the developed methodology include early detection of age-related epidemics and unusual increases in work-related accidents.