The National Academies

NCHRP 17-93 [Active]

Updating Safety Performance Functions for Data-Driven Safety Analysis

  Project Data
Funds: $500,000
Staff Responsibility: Leslie C. Harwood
Research Agency: University of North Carolina
Principal Investigator: Raghavan Srinivasan
Effective Date: 8/15/2019
Completion Date: 12/31/2022
Comments: Research In Progress


Data-driven safety analysis (DDSA) methods, including those presented in the AASHTO Highway Safety Manual (HSM), are becoming increasingly popular due to their ability to reliably and quantitatively assess the safety performance of existing and proposed roadways. Predictive analysis—a DDSA approach—is implemented using (1) safety performance functions (SPFs); (2) severity distribution functions (SDFs); and (3) crash modification factors (CMF).
SPFs are regression models that predict the crash frequency for a given set of geometric and operational conditions. For planning purposes, SPFs can be simple—using only the annual average daily traffic (AADT) data. They may also be detailed, having several additional input variables. SDFs (used for predicting the severity distribution of crashes), SPF adjustment factors (in HSM Part C), and CMF (discussed in HSM Part D) are used to determine the effectiveness of safety treatments by adjusting the SPF base model to represent additional site characteristics (e.g., traffic crashes, traffic volumes, roadway classification, roadway geometry, and cross-sectional features).


Many SPFs and CMFs used in current tools were developed several years ago based on research results and included in the HSM. For example, AASHTOWare Safety Analyst SPFs were developed using data collected from 1993-2001 from the Highway Safety Information System (HSIS) multistate database. Over time, the relationships represented by the models have changed due to improvements in vehicle design, changes in driver behavior, and new geometric configurations. Significant resources have been devoted to the development of these statistical models, and substantial resources are likewise required to update models using current methods.

The effectiveness of current methods is unclear due to (1) the use of outdated
functions and factors (SPFs, SPF adjustment factors, and SDFs) and (2) inefficient procedures to develop and update DDSA methods and models. 
Research is needed to (1) determine how to efficiently update models in the HSM, spreadsheets (derived from NCHRP Project 17-38), AASHTOWare Safety Analyst, and Interactive Highway Safety Design Model (IHSDM); and (2) examine the resources needed to develop new methods. 
The objectives of this research are to (1) determine how frequently SPFs, SPF adjustment factors (also called CMFs in Part C of the HSM), and SDFs should be updated; (2) determine under what conditions (e.g., changes in data and statistical properties) SPFs, SPF adjustment factors, and SDFs require updates; and (3) develop an implementable approach and guidance for maintaining and updating existing and future functions and factors. 
STATUS (February 2021)
Research in progress.

To create a link to this page, use this URL: http://apps.trb.org/cmsfeed/TRBNetProjectDisplay.asp?ProjectID=4575