Applying the latent class growth model into a longitudinal analysis of traffic crashes

One of the most important and meaningful tasks in traffic safety is to describe how traffic crash risk changes over time. Over the last 20 years, a lot of work has been done on this topic. However, with the recent introduction of latent class models for analyzing crash data, there is a need to examine how this new type of models could be used for longitudinal data analysis. Latent class models dictate that part of the heterogeneity can be attributed by grouping distinct subpopulations into a common dataset. Investigating the commonalities among the different subgroups can be useful for targeting specific safety interventions. This paper consequently describes the application of the latent class growth models (LCGM) that is specifically tailored for longitudinal data. The analysis was accomplished using data collected between 1997 and 2007 on rural two-lane highways in Texas. Trends for all crash severities and injury crashes were examined and it was determined that the crash data could be drawn from three population subgroups: low crash risk (but not equal to zero), medium crash risk and high crash risk. The results of this study show that average shoulder width and speed limit has a stronger effect for the sites that were classified as high crash risk, whereas traffic flow had a stronger influence for sites classified as low risk. As expected, higher speed limits increased crash risk, while wider shoulder width reduced the risk. In conclusion, the LCGM offers good potential for analyzing longitudinal data, but further work is needed on this topic.