Citation

Gandhi BVR, Lytton RL, Das B. Transp. Res. Rec. 1984; 960: 40-48.

Copyright

(Copyright © 1984, Transportation Research Board, National Research Council, National Academy of Sciences USA, Publisher SAGE Publishing)

DOI

unavailable

PMID

unavailable

Abstract

An improved safety index model is developed for narrow bridges by using data collected on 78 bridges in Texas by the Texas Transportation Institute. Cluster analysis was used to classify the bridges into more-safe and less-safe groups. Correlation and factor analyses were used to find the interrelationships between the independent variables of bridge and approach roadway characteristics. Stepwise regression was used to find the most important variables related to accident rate. On the groups developed by cluster analysis, logistic regression was used to develop a model that is a function of several variables that were found to be significantly related to accident rate. The enhanced safety index model consists of the following variables: bridge width, length, average daily traffic, and speed, as well as the subjective safety factors F6 (grade continuity), F7 (shoulder reduction), and F9 (traffic mix). The model was arrived at scientifically by using more objective procedures of classification and by having a stronger correlation with the accident rate than the previous models used by other researchers. The fewer variables used yield a much higher R2 when accident rate is used as the response variable in multiple regression. It is sensitive to changes or improvements in the constituent factors. The model yields the fraction of concordant pairs of predicted probabilities and responses as 0.91 and a high rank correlation of 0.81 between predicted probability and response, which indicates the goodness of the model. It also gives the safety index directly and can be used to identify a potentially hazardous narrow bridge.

Language: en

Keywords

BRIDGES, HIGHWAY; ACCIDENT PREVENTION