A comparative study of newly developed Kansas-specific safety performance functions with HSM models for rural four-lane divided highway segments

Abstract

For increased accuracy of the results of the procedures in the Highway Safety Manual (HSM), states are encouraged to develop jurisdiction-specific safety performance functions (SPF), if necessary. After performing calibration of four-lane divided rural-multilane segments using the HSM methodology, underprediction was observed for total crashes and overprediction was identified for fatal and injury crashes, indicating a need to examine the development of local SPF for Kansas conditions. Accordingly, negative binomial regression was applied to obtain the most suitable model for local conditions. Several additional variables were considered and tested in the new SPFs, which were run through validation set of locations to confirm accuracy. The newly developed SPFs were also compared to the HSM-given SPF and adjusted SPF using statistical parameters such as mean prediction bias, mean absolute deviation, and mean squared prediction error, leading to the conclusion that the newly developed Kansas-specific SPF for four-lane divided rural highway segments reliably predicts total crashes and fatal and injury crashes in rural Kansas. The Kansas-specific SPFs are capable of more accurately predicting total crashes and fatal and injury crashes on multilane segments compared to the HSM and the modified HSM models.