Development and Application of Crash Severity Models for Highway Safety: User Guidelines (2023)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

1   This user guide provides instructions for implementing the crash severity models developed under NCHRP Project 17-85. This project focused on estimating crash prediction models for the facilities covered in the first edition of the Highway Safety Manual (HSM) (AASHTO, 2010) and the already published supplements. However, freeway ramp segments and ramp terminals were ultimately excluded from the analysis due to data unavailability. The team investigated three candidate model approaches for each facility type—namely, negative binomial ordered probit fractional split (NB-OPFS) models, univariate count models (univariate), and multilevel discrete outcome models (multilevel). Several driver and vehicle variables—namely the proportion of female drivers, the proportion of young drivers (age between 16 and 24), the proportion of older drivers (age above 69), the proportion of heavy vehicles, and the proportion of motorcycles or other vehicles—are considered as covariates in addition to roadway and traffic characteristics commonly used in safety performance functions (SPFs). The existing HSM and NCHRP Project 17-62 (Ivan et al., 2021) models were considered to be benchmarks for evaluating the newly estimated approaches, so it is useful to compare the predic- tive performance of the new approaches relative to the existing approaches. This comparison is documented in Section 5.6 of the Conduct of Research Report for the project, which can be found in NCHRP Web-Only Document 351: Development and Application of Crash Severity Models for Highway Safety. After comparing the HSM models with the new methods proposed in this research, the team recommends using the existing HSM approach for rural and urban freeways. The performance of the HSM models was consistently better than the performance of the new alternative approaches. The authors note that the HSM approach for these facilities already includes SPFs for different severity levels, which were developed with considerably greater detail and with more input vari- ables than the HSM models for other facility types. Therefore, facility types finally discussed in this guide are rural two-lane, two-way roadways, rural multilane highways, and urban and suburban arterials, as listed in Table 1-1, which also identifies the modeling approach suggested for each facility type. The remainder of this guide includes (1) a summary of the model recommendations by facility types, (2) data requirements and preparation, (3) introduction of a web-based implementation tool, (4) discussions and summaries, and (5) an appendix. A flowchart is presented in Figure 1-1 to show steps to take to predict crashes by injury severity using the recommended methods. These steps are described in detail in the rest of this document. More generally, the procedure is as follows: 1. Select the facility type of interest. 2. Gather the required roadway and crash data and format them as specified for the predictive software tool. C H A P T E R 1 Introduction

2 Development and Application of Crash Severity Models for Highway Safety: User Guidelines 3. If using the Quasi-Induced Exposure (QIE) tool to generate the required demographic and vehicle proportion data, prepare the input file required to do this following the instructions. If demographic and vehicle composition data are available, add them to the roadway and crash data file according to the instructions. 4. Estimate crash counts and calibrate them to your jurisdiction. Facility Group Facility Type Model Type 1. Two-lane rural highways 1.1 – Segments Univariate 1.2 – 3-leg stop-controlled intersections Univariate 1.3 – 4-leg stop-controlled intersections Univariate 1.4 – 4-leg signalized intersections Univariate 2. Multilane rural highways 2.1 – Undivided segments Multilevel 2.2 – Divided segments Multilevel 2.3 – 3-leg stop-controlled intersections Multilevel 2.4 – 4-leg stop-controlled intersections Multilevel 2.5 – 4-leg signalized intersections HSM 3. Urban/suburban arterial segments 3.1 – Two-lane segments NB-OPFS 3.2 – Three-lane segments (with TWLTL) NB-OPFS 3.3 – Four-lane undivided segments NB-OPFS 3.4 – Four-lane divided segments NB-OPFS 3.5 – Five-lane segments (with TWLTL) NB-OPFS Urban/suburban arterial intersections 3.6 – 3-leg stop-controlled intersections NB-OPFS 3.7 – 3-leg signalized intersections NB-OPFS 3.8 – 4-leg stop-controlled intersections NB-OPFS 3.9 – 4-leg signalized intersections NB-OPFS Note: TWLTL = two-way left-turn lane. Table 1-1. Modeling approaches recommended for each facility type.

Introduction 3   Note: AADT = annual average daily traffic. Figure 1-1. Steps to take to predict crashes by injury severity using the recommended methods.