A causal inference analysis of injury severity in motorcyclist crashes in Brazilian highway databases

Abstract

Motorcyclist fatalities have risen in Latin and Asian countries, highlighting the necessity for a systematic approach through the Safe System (SS) to achieve zero fatalities. Traditional studies often rely on data-driven methodologies, which may lead to biased results and may fall short of capturing the systematic SS perspective. This paper develops a causal analysis of crash severity involving motorcyclists based on the SS approach and Pearl’s causal inference paradigm, using observational data. Initially, the main sources of bias related to traditional categorical modeling applied to crash severity analysis were illustrated through a simulation example. Subsequently, a literature review was conducted to establish a conceptual framework rooted in the SS. Finally, causal hypotheses were formulated and tested using structural equation modeling applied to data collected from motorcyclist crashes on federal highways in the state of Ceará/Brazil. The outcomes of the causal model support the initial hypothesis that alcohol use significantly contributes to the severity outcome of motorcycles-related crashes. The model unveils a noteworthy association between weekends and alcohol consumption and an inclination among younger individuals toward newer motorcycles with more powerful engines. The adoption of a causal approach enhances result reliability by controlling for confounding variables and incorporating a theoretical framework.