Forecasting mortality of road traffic injuries in China using seasonal autoregressive integrated moving average model

Purpose: Road traffic injuries have become a major public health problem in China. This study aimed to develop statistical models for predicting road traffic deaths and to analyze seasonality of deaths in China. Methods: A seasonal autoregressive integrated moving average (SARIMA) model was used to fit the data from 2000 to 2011. Akaike Information Criterion, Bayesian Information Criterion, and mean absolute percentage error were used to evaluate the constructed models. Autocorrelation function and partial autocorrelation function of residuals and Ljung-Box test were used to compare the goodness-of-fit between the different models. The SARIMA model was used to forecast monthly road traffic deaths in 2012. Results: The seasonal pattern of road traffic mortality data was statistically significant in China. SARIMA (1, I, I) (0, 1, 1)12 model was the best fitting model among various candidate models: the Akaike Information Criterion, Bayesian Information Criterion, and mean absolute percentage error were -483.679, -475.053, and 4.937, respectively. Goodness-of-fit testing showed nonautocorrelations in the residuals of the model (Ljung-Box test, Q = 4.86, P = .993). The fitted deaths using the SARIMA (1, 1, 1) (0, 1, 1)12 model for years 2000 to 2011 closely followed the observed number of road traffic deaths for the same years. The predicted and observed deaths were also very close for 2012. Conclusions: This study suggests that accurate forecasting of road traffic death incidence is possible using SARIMA model. The SARIMA model applied to historical road traffic deaths data could provide important evidence of burden of road traffic injuries in China. (C) 2015 Elsevier Inc. All rights reserved.