Citation

Xin C, Wang Z, Lee C, Lin PS, Chen T, Guo R, Lü Q. Accid. Anal. Prev. 2019; 123: 51-59.

Affiliation

Department of Civil and Environment Engineering, University of South Florida, 4202 E. Fowler Avenue, ENB118, Tampa, FL, 33620, USA. Electronic address: qlu@usf.edu.

Copyright

(Copyright © 2019, Elsevier Publishing)

DOI

10.1016/j.aap.2018.11.008

PMID

30465990

Abstract

Single-motorcycle crashes are overrepresented on horizontally curved segments of rural, two-lane, undivided (RTU) highways. However, the relationship between single-motorcycle crash risk and the design features of horizontal curves on RTU highways is not well-studied in existing literature. This study aims to quantify the effect of horizontal curve type and radius on the risk of single-motorcycle crashes with a matched case-control study that can address the issues of the low sample mean, aggregation bias, and uncontrolled confounders existing in the traditional cross-sectional study. In the matched case-control study, three matching factors-year, annual average daily traffic (AADT), and segment length-were selected to match controls (RTU segments without crash records) with cases (RTU segments with crash records). A total of 1601 cases and 16,010 matched controls over 11 years (2005-2015) were identified as matched-strata. A conditional logistic model was fitted on the matched-strata data to estimate the crash modification factors (CMFs) of horizontal curve design features for single-motorcycle crashes. The modeling results highlighted the interaction effects between curve type and radius on the risk of single-motorcycle crashes. Sharp (radius ≤ 1500 ft) non-reverse curves were identified as the riskiest curve design for motorcyclists, followed by sharp reverse curves and moderate (1500 ft < radius ≤ 3000 ft) reverse curves. The study also revealed that motorcyclists might take safety-compensation behaviors on sharp curves, narrow shoulders, and poor pavement conditions. Engineering and education countermeasures are suggested for comprehending curve presence and associated risk level, reducing curve entry speed, and improving safety awareness. Finally, the limitations of the study and possible solutions are discussed.

Copyright © 2018 Elsevier Ltd. All rights reserved.

Language: en

Keywords

Aggregation-bias issue; Conditional logistic model; Crash modification factor; Low mean problem; Matched case-control study; Motorcycle crash