Network Screening 101: Identifying High-Risk Road Segments with the Highway Safety Manual

Network screening is the systematic process of evaluating every segment and intersection in a road network to identify locations with the greatest potential for safety improvement. It’s the foundation of data-driven roadway safety—and the starting point for building a High-Injury Network, developing a safety action plan, or preparing an SS4A grant application.

This guide covers what network screening is, how the Highway Safety Manual (HSM) defines it, the performance measures used to rank locations, and how modern tools are making the process faster and more accessible.

What is Network Screening?

Network screening is the first step in the Highway Safety Manual’s roadway safety management process. Its purpose is straightforward: evaluate the entire road network and rank locations by their potential for crash reduction.

Rather than responding only to citizen complaints or high-profile crashes, network screening takes a proactive, data-driven approach. It systematically identifies the locations where safety investments will yield the greatest return.

Why Network Screening Matters

Federal programs like SS4A and the Highway Safety Improvement Program (HSIP) increasingly expect network screening as part of safety analysis. Using HSM-compliant methods signals to grant reviewers that your agency follows established best practices.

Performance Measures: How Locations Are Ranked

The HSM defines several performance measures for ranking locations during network screening. Each captures a different aspect of safety performance:

Understanding the Key Concepts

EPDO Weighting

Not all crashes are equal. A fatal crash represents a fundamentally different safety failure than a minor fender-bender. EPDO scoring assigns weights to each severity level to reflect this reality:

These weights are typically derived from crash cost data—the comprehensive economic cost of crashes at each severity level (medical costs, lost productivity, quality-of-life impacts). This means EPDO scoring directly reflects the societal cost of crashes at each location.

Regression to the Mean

One of the most important statistical concepts in network screening is regression to the mean (RTM). Locations with unusually high crash counts in one period may naturally return to normal levels in the next period, even without intervention.

If you select locations purely based on recent crash history, some of your “worst” locations may not actually have a systemic safety problem—they just had an unlucky period. The HSM addresses this through the Empirical Bayes (EB) method, which combines observed crash data with predicted crash frequency from Safety Performance Functions to produce a more reliable estimate.

Safety Performance Functions (SPFs)

SPFs are statistical models that predict the expected crash frequency for a road segment or intersection based on its characteristics: traffic volume (AADT), number of lanes, road type, speed limit, and other features. They answer the question: “How many crashes would we expect at a location like this?”

When used with the EB method, SPFs help distinguish between locations that are genuinely dangerous and those that merely experienced a temporary spike in crashes.

The Network Screening Process

Step 1: Define the Network

Establish the set of road segments and intersections to evaluate. This typically includes all public roads in your jurisdiction, segmented into consistent analysis units (e.g., 0.1-mile segments for roadways, individual intersections for nodes).

Step 2: Assemble Data

Gather crash records (minimum 3–5 years), traffic volumes (AADT), and roadway characteristics. Link crashes to their corresponding network locations using spatial analysis or linear referencing.

Step 3: Select Performance Measures

Choose which measures to apply based on your data availability and analytical goals. At minimum, calculate crash frequency and EPDO. If SPFs are available for your road types, apply the EB method for more reliable results.

Step 4: Calculate and Rank

Compute the selected performance measures for every location in the network. Rank locations from highest to lowest potential for safety improvement.

Step 5: Review and Validate

Examine the top-ranked locations. Do they align with local knowledge? Are there data quality issues inflating certain locations? Validation ensures the screening results are actionable.

Step 6: Prioritize for Further Analysis

The output of network screening is a prioritized list of locations warranting detailed safety investigation—not a final project list. These locations move to the next phase: diagnosis, countermeasure selection, and economic appraisal.

Network Screening for Segments vs. Intersections

Network screening applies differently depending on the location type:

• Segments — Evaluated per unit length (e.g., crashes per mile). Segment characteristics include number of lanes, shoulder width, speed limit, median type, and access point density. SPFs for segments use AADT and segment length as primary predictors.
• Intersections — Evaluated as point locations. Key characteristics include intersection type (signalized, stop-controlled, roundabout), number of approaches, turning movement volumes, and sight distance. Intersection SPFs typically use entering AADT from major and minor approaches.

Running separate analyses for segments and intersections is standard practice, as they have different crash patterns, risk factors, and applicable countermeasures.

Common Network Screening Pitfalls

• Relying solely on crash frequency — High-volume roads will always have more crashes. Without normalizing by traffic volume or using severity weighting, you’ll prioritize busy roads over genuinely dangerous ones.
• Ignoring regression to the mean — Short analysis periods and small crash counts amplify random variation. Use the EB method when possible, or at minimum use 5+ years of data.
• Inconsistent segmentation — Segment lengths affect crash density calculations. Very short segments inflate crash rates; very long segments mask localized problems. Use consistent, analytically appropriate segment lengths.
• Not separating severity levels — A location with 50 PDO crashes is fundamentally different from one with 5 fatal crashes. Always include severity-based analysis alongside total crash counts.
• Treating results as final — Network screening identifies candidates for further investigation. It does not definitively prove that a location needs a specific treatment. Always follow screening with detailed diagnosis.

Network Screening and Federal Grant Programs

Network screening directly supports multiple federal safety programs:

• SS4A (Safe Streets for All) — Network screening results feed directly into the comprehensive safety analysis required for Action Plan grants. Learn more about SS4A.
• HSIP (Highway Safety Improvement Program) — States are required to use a data-driven process to identify safety projects. Network screening is the standard approach.
• Local Road Safety Plans — FHWA encourages local agencies to adopt systemic safety analysis, with network screening as the starting point.

Using HSM-compliant network screening in your grant applications demonstrates analytical rigor and alignment with federal safety policy—both of which strengthen your competitive position.