Using CMFs for Road Safety Evaluation


Section A. Using CMFs Correctly in a Crash Prediction Model

The documents in this section were developed for NCHRP Project 17-78: Understanding and Communicating Reliability of Crash Prediction Models.

Implementation Product

Procedures for Quantifying the Reliability of Crash Prediction Model Estimates -- with a Focus on Mismatch between CMFs and SPF Base Conditions.  This document describes procedures for quantifying the safety effect of two factors that influence the reliability of the estimate obtained from a crash prediction model (CPM). The document demonstrate how the procedures can be used to quantify and interpret CPM estimate reliability.  PROCEDURE GUIDE

A CPM is defined by the Highway Safety Manual (HSM) to include a safety performance function (SPF), one or more crash modification factors (CMFs), and a local calibration factor (C).

The procedures address the following cases:

Research Report

Development of Procedures for Quantifying the Reliability of Crash Prediction Model Estimates. This report documents the research undertaken to develop the procedures described in the previous section.  RESEARCH REPORT

 

Section B. Using CMFs to Estimate Treatment Effect at a Site of Interest

The documents and tools provided in this section were developed for NCHRP Project 17-63: Guidance for the Development and Application of Crash Modification Factors. The final report for this project was published as NCHRP Report 991. Most of the documents on this webpage can be found in Appendix A and Appendix D of Report 991.

Implementation Products

 Procedure for Estimating the Safety Effect of a Treatment at a Site of Interest. This Procedure Guide discusses the following issues:

Underlying these issues is the premise that the value of an aggregated CMF can vary among jurisdictions due to differences in crash type and severity distribution. A case is made that a treatment's effect on specific combinations of crash type and severity is relatively constant among jurisdictions, even though its effect on total crash frequency can vary significantly among jurisdictions. This wide variation in total-crash CMFs among jurisdictions can be partly explained by jurisdictional differences in crash type and severity distribution.

The long-term solution is to move safety evaluation practice in the direction of disaggregated analysis (i.e., specifically evaluating various combinations of crash type and severity, as opposed to evaluating only "total" crash frequency). In the short term, it is likely that many evaluations will continue to be based on total crash frequency. In recognition of this short-term trend, the procedures described in this document can be used to mitigate the aforementioned issues. Specifically, these procedures can be used to compute a "locally calibrated" total-crash CMF based on the local crash type and severity distribution.  PROCEDURE GUIDE 

Some steps described in the Procedure Guide require calculations and statistical analysis. These calculations and analyses are automated in the CMF Regression Software tool. An example application of this tool is provided in Appendix E.  SPREADSHEET

Note: A 2022 security change by Microsoft prevents Excel files with macros (i.e., with the .xlsm extension) from running unless the analyst changes the file properties to unblock the file macros. Right-click on the filename, select Properties, select General, in the Security section (at the bottom) check the "Unblock" checkbox, and select OK.

Appendix A - Procedure for Disaggregating Aggregate CMFs. Currently, there are limited numbers of CMFs and prediction models available to support disaggregated analysis. Hence, Appendix A describes a procedure for using existing aggregate CMFs to estimate disaggregate CMFs. Appendix A is included in the Procedure Guide.  Example applications of this procedure are provided in Appendix B. 

Supplemental Tools and Appendices

Appendix B - Procedure for Disaggregating Aggregate CMFs: Example Applications. This appendix presents four example applications of the procedure for disaggregating aggregate CMFs. This procedure is described in Appendix A. A spreadsheet for implementing this procedure is described in Appendix E.  APPENDIX B

Appendix C - Procedure for Combining Multiple CMFs for a Common Treatment. This appendix describes a procedure for determining whether multiple CMFs values can be combined and, if appropriate to combine, a procedure for determining the overall average safety effect of the treatment. The CMF values used with this procedure may be obtained from either (1) a study that quantifies a CMF value for each of several treated locations, or (2) the published findings from several independent studies of a specific treatment. APPENDIX C

Appendix D - A Comparison of Two Estimators of Combined Average CMF. This appendix describes the findings from an evaluation of two alternative estimators of the combined average CMF. The combined average CMF is used to describe the overall average safety effect of a given treatment based on either (1) its application to several locations, or (2) the published findings from several independent studies of this treatment. The recommended estimator is included in the procedure described in Appendix C.  APPENDIX D

Appendix E - Example Application of CMF Regression Spreadsheet. This appendix describes an example application of the CMF Regression Software tool. This tool can be used to evaluate two or more CMFs associated with a common treatment. The three evaluation types supported by the tool consist of: (1) computing the overall average CMF, (2) computing CMFs by crash type or severity category using reported aggregate CMFs, or (3) computing CMFs as a function of site characteristics.  APPENDIX E