Segmenting and investigating pedestrian-vehicle crashes in Ghana: A latent class clustering approach

Abstract

In low- and middle-income countries (LMIC) pedestrians and cyclists account for approximately 26 % of the road traffic deaths, which is a considerable amount as it is well known that the majority (90 %) of the world's road traffic deaths occur in these countries. In Africa however, pedestrian and cyclist deaths account for 44 % of their yearly road related deaths. Ghana is no exception to this trend; in fact, it has been estimated that pedestrian crashes alone account for 36.7 % of road related deaths in the country. Therefore, the objective of this study is to use historical crash records from 2018 to 2020 to explore pedestrian-vehicle crashes in Ghana, to identify the groups of pedestrians represented in pedestrian-vehicle crashes by use of a latent class analysis (LCA) model, then conduct injury severity analyses using a mixed logit approach on each pedestrian group found in the LCA modeling. Results indicate that by segmenting the pedestrian crash data into homogenous groups, some variables were found to only be significantly associated with injury severity within some classes. Other variables were found to be significant across multiple classes yet experience different trends within each. For example, no traffic control was found to be significant within three subgroups but affect severity levels differently across classes. Further the darker hours of the day were more likely to be associated with fatal and major injury outcomes across multiple classes. This study provides new direction for studying different types of pedestrian crashes, particularly in LMICs and provides targeted interventions.