Open AccessFactors related to highway crash severity in Brazil through a multinomial logistic regression model
Author(s) -
Lucas Franceschi,
Luciano Kaesemodel,
Vera do Carmo Comparsi de Vargas,
Andréa Cristina Konrath,
Luiz Ricardo Nakamura,
Thiago G. Ramires,
Camila Belleza Maciel Barreto,
Amir Mattar Valente
Publication year - 2022
Publication title -
transportes
Language(s) - English
Resource type - Journals
eISSN - 2237-1346
pISSN - 1415-7713
DOI - 10.14295/transportes.v30i1.2566
Subject(s) - multinomial logistic regression , crash , logistic regression , transport engineering , poison control , injury prevention , occupational safety and health , environmental health , pedestrian , geography , regression analysis , human factors and ergonomics , medicine , statistics , computer science , engineering , mathematics , pathology , programming language
Reducing the number of deaths by road crashes is an important priority for many countries around the world. Although focusing on the occurrence of crashes can provide safety policies that help reduce its numbers, studying their severity can provide different measures that may help further reduce the number of deaths by focusing on the most severe problems first. In this paper, a multinomial logistic regression model is fitted to nationwide highway crash data in Brazil from 2017 to 2019 to identify and estimate the associated factors to crash severity. Severity is classified as without injury, with injured victims or with fatal victims. Amongst other observations, results indicate that pedestrian involvement in highway crashes increase dramatically their severity. Also, conditions that favor greater speeds (clear weather, times when there are fewer vehicles on the road) are also related to an increase in crash severity, pointing to a proportional relation with traffic fluidity. Moreover, some observed limitations on the data may indicate that Brazil would benefit greatly from national crash records standards and unified databases, especially crossmatching crash reports with health data.