Diagnostic analysis of the logistic model for pedestrian injury severity in traffic crashes

Nang Ngai Sze, S. C. Wong

Research output: Journal article publicationJournal articleAcademic researchpeer-review

171 Citations (Scopus)

Abstract

This study attempts to evaluate the injury risk of pedestrian casualties in traffic crashes and to explore the factors that contribute to mortality and severe injury, using the comprehensive historical crash record that is maintained by the Hong Kong Transport Department. The injury, demographic, crash, environmental, geometric, and traffic characteristics of 73,746 pedestrian casualties that were involved in traffic crashes from 1991 to 2004 are considered. Binary logistic regression is used to determine the associations between the probability of fatality and severe injury and all contributory factors. A consideration of the influence of implicit attributes on the trend of pedestrian injury risk, temporal confounding, and interaction effects is progressively incorporated into the predictive model. To verify the goodness-of-fit of the proposed model, the Hosmer-Lemeshow test and logistic regression diagnostics are conducted. It is revealed that there is a decreasing trend in pedestrian injury risk, controlling for the influences of demographic, road environment, and other risk factors. In addition, the influences of pedestrian behavior, traffic congestion, and junction type on pedestrian injury risk are subject to temporal variation.
Original languageEnglish
Pages (from-to)1267-1278
Number of pages12
JournalAccident Analysis and Prevention
Volume39
Issue number6
DOIs
Publication statusPublished - 1 Nov 2007
Externally publishedYes

Keywords

  • Binary logistic regression
  • Hosmer-Lemeshow statistic
  • Injury severity
  • Interaction effect
  • Logistic regression diagnostics
  • Pedestrian casualty

ASJC Scopus subject areas

  • Human Factors and Ergonomics
  • Safety, Risk, Reliability and Quality
  • Public Health, Environmental and Occupational Health
  • Law

Cite this