Analyses of Multiyear Statewide Secondary Crash Data and Automatic Crash Report Reviewing. Issue 1 (January 2015)

• Record Type: Journal Article
• Authors: Zheng, Dongxi; Chitturi, Madhav V.; Bill, Andrea R.; Noyce, David A.
• Is Part Of: Transportation research record. Volume 2514:Issue 1(2015)
• Page Start: 117
• Page End: 128
• Publication Date: 2015-01
• Subjects: Transportation -- Periodicals; Roads; Transport -- Périodiques; Routes -- Périodiques; Routes -- Conception et construction -- Périodiques; Roads; Transportation; 388.05
• DOI: 10.3141/2514-13
• Languages: English
• ISSNs: 0361-1981

Abstract

Secondary crashes are undesired consequences of highway incidents. Previous studies used different methods to identify secondary crashes. However, because of the expense of accurate identification, most studies focused only on small-scale highway networks (e.g., up to urban arte-rials). Some studies considered statewide networks, but the accuracy of their secondary crash identification methods was not justified. Recently, the authors proposed an efficient method to identify secondary crashes on statewide freeway networks with reasonable accuracy. A 1-year case study was conducted for preliminary analysis. As a continuing effort, this research included four more years of secondary crash data and found that (a) rear-end and sideswipe crashes in the same direction were the top two secondary crash types; (b) road debris, construction zones, and obscured visibility were three major potential highway contributing factors; ( c ) following too closely, inattentive driving, losing vehicle control, and speeding were four major potential driver contributing factors; ( d ) temporal distributions by hour and month were different between secondary crashes and general crashes; and ( e ) secondary crash hot spots clustered around urban areas and were within 1 mi from major freeway interchanges. In addition, an algorithm was proposed to detect secondary crashes on the basis of police narratives in the crash reports. This algorithm was evaluated by using 5-year secondary crash data (4 years for training and 1 year for testing). By choosing an optimal threshold, this algorithm identified all true secondary crashes while keeping false positives at a low number.