Estimating injury severity for motorized and non-motorized vehicle-involved crashes: Insights from random-parameter ordered probit model with heterogeneity in means and variances

The use of advanced models to investigate the determinants of injury severity outcomes for motorized and nonmotorized-involved crashes are sparse. Therefore, random-parameter ordered probit models with heterogeneity in means and variances were developed to estimate factors affecting injury severity for motorized and nonmotorized-involved crashes. Data covering a five-year period comprising 5976 and 634 cases for motorized and non-motorized-involved crashes respectively, was retrieved from the database of the National Road Safety Authority, State Insurance Company and Driver and Vehicle Licensing Authority in Ghana and used for model estimation. The results show that factors have varying significant effects on injury severity outcomes for motorized and non-motorized models. Marginal effects indicate that old age occupants, head-on-collision, exceeding a posted speed limit of 100 km/h and crash during weekends contributed greatly to the likelihood of severe injury outcomes in motorized model. Additionally, male non-motorists, non-use of helmet, rear-end collision, right-angle collision and crash on urban roads and during weekends, contributed significantly to the severe injury outcomes of non-motorized models. The direction of effect of the factors on severe injury was observed to have varying degrees of estimated coefficients. The difference in estimated coefficients shows that crashes involving non-motorized vehicles were more likely to result in severe injury compared to motorized vehicles. The motorized model had heterogeneity in means of five (5) random parameters observed, while the non-motorized model had heterogeneity in means of four (4) random parameters observed with two variables affecting the variance of three random parameters. Based on the results, various countermeasures were proposed to enhance road traffic safety.