Guide for Quantitative Approaches to Systemic Safety Analysis (2020)

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.

20 Overview of Systemic Safety Management Approaches The general purpose and methodology for performing a systemic safety management analysis are explained in part 2.3 in Section 2, but agencies use a variety of implementation approaches for their systemic safety management programs. Differences in crash data availability, roadway characteristic inventory availability, and scope of the systemic safety program may affect how an agency applies a systemic safety management approach. This section describes three primary approaches that agencies have used to implement systemic safety management: • Application of the FHWA Systemic Tool methodology, customized to local data availability and program goals. • Application of SPFs, using in-house tools or Safety Analyst. • Application of the usRAP methodology, using the associated ViDA software. Each of the three approaches is described in more detail below. While describing the various applications of the systemic safety management approaches, data requirements are described using terminology consistent with the Model Inventory of Roadway Elements—MIRE 2.0 (Lefler et al., 2017) and the MMUCC Guideline—Model Minimum Uniform Crash Criteria— Fifth Edition (NHTSA, 2017). 3.1 Application of the FHWA Systemic Safety Project Selection Tool Methodology To support implementation of the systemic safety management approach to highway safety, FHWA published the Systemic Safety Project Selection Tool (Preston et al., 2013) guidance document to outline the systemic safety process and to support agency implementation of the approach. The Systemic Tool does not actually include any software components, but recently FHWA developed a tool that can be used to generate crash tree diagrams based on FARS or other crash datasets (FHWA, 2019B). The systemic safety management approach to traffic safety, as described in the Systemic Tool, is an “approach to safety [which] involves widely imple- mented improvements based on high risk roadway features correlated with specific severe crash types.” The approach “supplements and complements traditional site analysis” and proactively addresses locations that exhibit high-crash potential due to location attributes such as road- way geometry and cross-sectional design, roadside and area features, traffic control, and more (Preston et al., 2013). The Systemic Tool helps agencies identify crash contributing factors as well as identify target crash types, effective countermeasures, and priority facilities for implementa- tion of those countermeasures. Since its publication, the systemic safety management approach described in the Systemic Tool has been utilized by multiple state and local agencies. This section describes the various elements of the systemic safety management approach from the Systemic Tool (see Figure 2), summarizing some contemporary practices in implementing the process. S E C T I O N 3

Overview of Systemic Safety Management Approaches 21 This section of the guide primarily focuses on the first element of the Systemic Tool methodology—the systemic safety planning process. This process contains four basic steps: • Identify target crash types and crash contributing factors. • Screen and prioritize candidate locations. • Select countermeasures. • Prioritize projects. Each step in this process is discussed in depth in the following sections. Not all applications of a systemic safety management approach follow these steps, however, so agencies considering this tool should evaluate whether these steps will help them achieve their desired outcomes. 3.1.1 Target Crash Types and Contributing Factors To begin the systemic safety planning process, key crash types, facility types, and associated contributing factors are first identified. One option for an agency to identify target crash types is by conducting a systemwide analysis of its crash data, and as necessary, roadway inventory data. The analysis of several crash data elements is of particular interest, but linkages of the crash data to several roadway inventory data elements may also be useful when identifying target crash types. Linkages of the crash data to roadway inventory data elements as well as type of government ownership and rural/urban designation—both of which are Fundamental Data Elements (FDEs) as designated by FHWA as part of Model Inventory of Roadway Elements (MIRE)—are of primary interest to determine if a crash occurred on a state- maintained highway or a local/county road and to distinguish crashes separately by area type (i.e., rural and urban). The crash data elements of interest for identifying target crash types Figure 2. Systemic safety tool flow chart (Preston et al., 2013).

22 Guide for Quantitative Approaches to Systemic Safety Analysis will vary from agency to agency, but several of the common crash data elements of interest across agencies include: • Crash severity, • Manner of collision/collision impact, • First harmful event, • Speed-related, • Alcohol involvement, • Drug involvement, • Light conditions, and • Sequence of events. Another option for an agency to identify target crash types to address using a systemic safety management approach is to refer to a state or regional SHSP. These plans generally provide emphasis areas or target crash types for the state’s or region’s safety program. Referring to a state or regional SHSP to identify target crash types for systemic safety analysis does not require the use of any specific type of data (i.e., crash or roadway inventory). Although a systemwide analysis of crash and roadway inventory data is beneficial to identify potential target crashes for a systemic safety management approach, the results of a system- wide analysis may not be sensitive to specific geographical regions. Therefore, in addition to a systemwide analysis, it is also useful to analyze the same crash and roadway inventory data elements, focusing on certain geographical areas or regions within a state or jurisdiction. Comparing the results of a systemwide analysis and regional analysis can help identify target crash types of interest in certain geographical areas. Target crash types commonly identified as being widely dispersed across a network, rather than occurring frequently at specific locations, and often identified as target crash types to be addressed as part of a systemic safety management approach include: • Lane departure, • Rollover, • Fixed object, • Head-on, • Angle, • Speed-related, • Younger driver involvement, • Impaired driving, • Pedestrians, • Bicyclists, and • Nighttime. When identifying target crash types to be addressed through a systemic safety management approach, often the analyses focus on fatal and serious-injury crashes. It may also be beneficial to compare the proportion of fatal and serious-injury crashes to total crashes to identify target crash types. Once target crash types are identified, the next step is to identify where the crashes are occurring. To do so, agencies should identify similar facilities and document crash histories and patterns as well as physical and traffic characteristics based on the crash types being considered. A common way of doing this is using a crash tree diagram that provides a visual way of dis- aggregating crash data by facility attributes, such as jurisdiction, setting (i.e., rural versus urban), intersection versus non-intersection, and more. To generate a crash tree diagram for a target crash type, an analyst usually begins by consider- ing fatal and serious-injury crashes or perhaps a range of crash severity levels (e.g., all severity

Overview of Systemic Safety Management Approaches 23 levels combined, fatal and all-injury crashes, and/or fatal and serious-injury crashes). For the severity levels of interest, crashes are then subdivided by roadway characteristics of interest, which may again require linkages of the crash data to roadway inventory data elements and in some cases traffic volume data if the data are not available from the crash dataset. Potential combinations for classifying or categorizing facility types by roadway and traffic volume characteristics for visualization in crash tree diagrams are as follows: • Segment versus intersection, • Segment type: – State-maintained highway versus local road, – Freeway versus non-freeway, – Number of lanes, – Divided versus undivided, – One-way versus two-way, – Tangent versus horizontal curve, – High speed versus low speed, – Paved versus unpaved, – Traffic volume levels. • Intersection control type: – Signalized versus unsignalized. • Area type. Figure 3 illustrates a sample crash tree diagram of roadway departure crashes by area type, number of lanes, and median type. For this sample diagram, roadway departure crashes are overrepresented on rural two-lane and rural multilane divided highways. The common crash data elements used for generating crash tree diagrams to identify facility types of interest include: • Crash severity, • Relation to junction, • Type of intersection, • Trafficway description, • Total lanes in roadway, • Roadway alignment and grade, • Roadway functional class, State System 15,040 mi 8,889 FS crashes Rural 9,776 mi (65%) 6,933 FS crashes (78%) Multilane Undivided 150 mi (1%) 88 FS crashes (1%) Multilane Divided 602 mi (4%) 267 FS crashes (3%) Two Lane 3,760 mi (25%) 1,333 FS crashes (15%) Urban 5,264 mi (35%) 1,956 FS crashes (22%) Multilane Undivided 902 mi (6%) 356 FS crashes (4%) Multilane Divided 752 mi (5%) 800 FS crashes (9%) Two Lane 8,874 mi (59%) 6,045 FS crashes (68%) Roadway Departure Crashes Severity: Fatal and Serious Years: 2009 - 2018 Figure 3. Sample crash tree diagram of roadway departure crashes.

24 Guide for Quantitative Approaches to Systemic Safety Analysis • Annual average daily traffic, • Motor vehicle posted/statutory speed limit, • Width of lane(s) and shoulder(s), and • Access control. Often, a crash tree diagram can be created with information available from crash data alone. However, if a crash dataset is not comprehensive or data are missing, common roadway inventory data elements used for generating crash tree diagrams to identify facility types of interest include: • Type of government ownership (FDE), • Functional class (FDE), • Rural/urban designation (FDE), • Access control (FDE), • Surface type (FDE), • Number of through lanes (FDE), • Median type (FDE), • One-/two-way operations (FDE), • Intersection/junction geometry (FDE), • Annual average daily traffic (FDE), • Right shoulder total width, • Right paved shoulder width, and • Speed limit. Several facility types that agencies have selected to use systemic safety management approaches to program safety improvements include: • Low-volume local roads, • Rural two-lane roads, • Rural local roads, • Rural roads with pavement width less than 24 ft, • Horizontal curves on rural two-lane roads, • Unpaved roads, and • Signalized and stop-controlled intersections. The Contributing Factors for Focus Crash Types and Facility Types: Quick Reference Guide (Porter et al., forthcoming) provides a list of common target crash type and facility type combinations potentially addressed through a systemic safety management approach based on an analysis of national and state data. These combinations are provided in Table 1. Using historic crash data, the target crash type(s) and facility type(s) are then assessed to identify potential contributing factors that should be considered in the analysis. Analysts look for common characteristics among facilities where crashes have occurred that may correlate to specific road user behaviors or reduced safety performance. Though the correlation between crash types and identified contributing factors does not imply a causal relationship, the road- way characteristics are useful for identifying and prioritizing locations that may be expected to perform similarly and for distinguishing locations that may have greater potential for severe crashes based on a systemwide analysis. Table 2 provides a list of crash types commonly addressed using a systemic safety approach and contributing factors frequently associated with each crash type. Site characteristics and data elements commonly selected as potential contributing factors in a systemic safety analysis are provided in Table 3. Table 3 lists contributing factors for both roadway segments and intersections and specifies the data elements from MIRE and Model

Overview of Systemic Safety Management Approaches 25 Facility Type Crash Type R un -o ff- ro ad La ne de pa rt ur e H ea d- on A ng le R ol lo ve r Roadway Segments Rural, two-lane roads on horizontal curve sections X X X X Rural, two-lane roads on tangent sections X X X X X Intersections Four-leg minor-road stop-controlled intersections on rural two-lane roads X Four-leg minor-road stop-controlled intersections on urban two-lane roads X Three-leg minor-road stop-controlled intersections on rural two-lane roads X Four-leg signalized intersections on urban, multilane, divided roads X Four-leg signalized intersections on urban, multilane, undivided roads X Four-leg minor-road stop-controlled intersections on rural, multilane, divided roads X Table 1. Common target crash type and facility type combinations for systemic safety management (Porter et al., forthcoming). Target Crash Types Potential Contributing Factors Roadway departure Lane width Shoulder width/type Median width/type Horizontal curvature, delineation, or advance warning devices Superelevation Horizontal curve density Horizontal curve and tangent speed differential Presence of a visual trap at a curve or combination of vertical grade and horizontal curvature Roadway gradient Roadside or edge hazard rating (potentially including sideslope design and frequency of fixed objects) Driveway presence, design, and density Sight distance from access location Presence of shoulder rumble strips Presence of centerline rumble strips Posted speed limit or operating speed Presence of lighting Quality of roadway surface (or pavement condition and friction) Average daily traffic volumes Rollover Shoulder width/type Median width/type Roadside or edge hazard rating (potentially including sideslope design and frequency of fixed objects) Quality of roadway surface (or pavement condition and friction) Posted speed limit or operating speed Table 2. Common target crash types for systemic safety management approach and potential contributing factors. (continued on next page)

26 Guide for Quantitative Approaches to Systemic Safety Analysis Target Crash Types Potential Contributing Factors Pedestrians (Roadways) Lane width Shoulder width/type Presence of sidewalk Driveway presence, design, and density Presence of transit stops Presence of lighting Posted speed limit or operating speed Bicyclists (Intersections) Intersection skew angle Type of intersection traffic control Number of lanes crossed Presence and type of bicycle facility Presence of transit stops Presence of lighting Presence of left-turn or right-turn lanes Left-turn phasing Allowance of right-turn-on-red Bicyclists (Roadways) Number of lanes Lane width Presence and type of bicycle facility Shoulder width/type Presence of on-street parking Driveway presence, design, and density Presence of shoulder rumble strips Quality of roadway surface (or pavement condition and friction) Presence of lighting Posted speed limit or operating speed Average daily traffic volumes Proportion of commercial vehicles in traffic stream Nighttime Presence of lightingPosted speed limit or operating speed Presence of left-turn or right-turn lanes Fixed object Shoulder width/type Median width/type Horizontal curvature, delineation, or advance warning devices Superelevation Horizontal curve density Horizontal curve and tangent speed differential Presence of a visual trap at a curve or combination of vertical grade and horizontal curvature Roadway gradient Roadside or edge hazard rating (potentially including sideslope design and frequency of fixed objects) Driveway presence, design, and density Sight distance from access location Presence of shoulder rumble strips Presence of lighting Quality of roadway surface (or pavement condition and friction) Pedestrians (Intersections) Intersection skew angle Type of intersection traffic control Number of lanes crossed Presence of pedestrian refuge island Pedestrian crosswalk presence, crossing distance, signal head type Presence of pedestrian activated flashers or beacons Presence of transit stops Presence of lighting Left-turn phasing Allowance of right-turn-on-red Adjacent land-use type, such as schools, commercial, or alcohol sales establishments Table 2. (Continued).

Overview of Systemic Safety Management Approaches 27 Contributing Factor MIRE Data Element MMUCC Data Element Roadway Segments Number of lanes Number of through lanes (FDE) Total Lanes in Roadway Lane width Outside through-lane widthInside through-lane width Width of Lane(s) and Shoulder(s) Shoulder width/type Right shoulder type Right shoulder total width Right paved shoulder width Width of Lane(s) and Shoulder(s) Median width/type Median type (FDE)Median width Median Width Horizontal curvature, delineation, or advance warning devices Horizontal curve degree or radius Roadway Curvature Roadway Alignment and Grade Superelevation Curve superelevation No relevant variable available Horizontal curve density No relevant variable available No relevant variable available Horizontal curve and tangent speed differential No relevant variable available No relevant variable available Presence of a visual trap at a curve or combination of vertical grade and horizontal curvature No relevant variable available No relevant variable available Roadway gradient Vertical alignment feature typePercent of gradient Grade Roadway Alignment and Grade Roadside or edge hazard rating (potentially including sideslope design and frequency of fixed objects) Roadside clear zone width Right sideslope Right sideslope width Left sideslope Left sideslope width Roadside rating No relevant variable available Driveway presence, design, and density Major commercial driveway count Minor commercial driveway count Major residential driveway count Minor residential driveway count Major industrial/institutional driveway count Minor industrial/institutional driveway count Other driveway count Relation to Junction Sight distance from access location No relevant variable available No relevant variable available Presence of shoulder rumble strips Right shoulder rumble strip presence/type Left shoulder rumble strip presence/type No relevant variable available Presence of centerline rumble strips Centerline rumble strip presence/type Pavement Markings, Longitudinal Quality of roadway surface (or pavement condition and friction) Surface friction Surface friction date International roughness index International roughness index date Pavement condition (present serviceability rating) Pavement condition date Roadway Surface Condition Weather Condition Presence of lighting Roadway lighting Light Condition Length of passing zone Passing zone percentage No relevant variable available Sight distance in passing zone No relevant variable available No relevant variable available Opportunity for passing No relevant variable available No relevant variable available Presence of on-street parking On-street parking presenceOn-street parking type Location of first harmful event relative to the trafficway Proximity to adjacent traffic No relevant variable available No relevant variable available Presence of sidewalks Sidewalk presence No relevant variable available Presence of transit stops No relevant variable available Contributing circumstances—roadway environment Table 3. Potential contributing factors and associated MIRE and MMUCC data elements. (continued on next page)

28 Guide for Quantitative Approaches to Systemic Safety Analysis Posted speed limit or operating speed Speed limit Nighttime speed limit 85th percentile speed Mean speed Motor vehicle posted/statutory speed limit Presence and type of bicycle facility Presence/type of bicycle facility Presence/type of bicycle facility Average daily traffic volumes Annual Average Daily Traffic (AADT) (FDE) AADT year (FDE) AADT Proportion of commercial vehicles in traffic stream Percent single unit trucks or single truck AADT Percent combination trucks or combination truck AADT Percentage trucks or truck AADT No relevant variable available Intersections Intersection skew angle Intersecting angle No relevant variable available Number of intersection approaches Intersection/junction number of legsIntersection/junction geometry (FDE) Type of intersection Type of intersection traffic control Intersection/junctional traffic control (FDE) Type of intersection Number of signal heads versus number of lanes No relevant variable available No relevant variable available Presence of backplates No relevant variable available No relevant variable available Visibility of traffic control devices No relevant variable available No relevant variable available Presence of advanced warning signs No relevant variable available No relevant variable available Intersection located in/near horizontal curve No relevant variable available No relevant variable available Presence of left-turn or right-turn lanes Left-turn lane type (each approach) Right-turn channelization (each approach) Number of exclusive left-turn lanes (each approach) Number of exclusive right-turn lanes (each approach) No relevant variable available Left-turn phasing Approach left-turn protection (each approach) Left-/right-turn prohibitions No relevant variable available Allowance of right-turn-on-red Right-turn-on-red prohibitions (each approach) No relevant variable available Overhead versus pedestal mounted signal heads No relevant variable available No relevant variable available Contributing Factor MIRE Data Element MMUCC Data Element Roadway Segments Pedestrian crosswalk presence, crossing distance, signal head type Crosswalk presence/type (each approach) Pedestrian signal activation type (each approach) Pedestrian signal presence/type (each approach) No relevant variable available Presence of pedestrian-activated flashers or beacons No relevant variable available No relevant variable available Presence of transit stops No relevant variable available Contributing circumstances—roadway environment Presence of lighting Intersection/junction lighting Light condition Presence of nearby railroad crossing No relevant variable available No relevant variable available Presence of automated enforcement No relevant variable available No relevant variable available Adjacent land-use type, such as schools, commercial, or alcohol sales establishments No relevant variable available No relevant variable available Average daily entering vehicles Approach AADT (each approach) Approach AADT year (each approach) Total volume of entering vehicles Table 3. (Continued).

Overview of Systemic Safety Management Approaches 29 Minimum Uniform Crash Criteria (MMUCC) related to each contributing factor. Analysts may find these datasets helpful in conducting an analysis of contributing factors. Descriptive statistics can be generated to compare the proportion of sites where the contributing factors are present with the percentage of crashes occurring at sites where the contributing factors are present for the severity level of interest. Contributing factors found to be overrepresented at locations experiencing crashes should be considered for inclusion in network screening procedures. For example, while rural roads may account for only a minority of mileage on an agency’s roadway network and a minority of the vehicle miles traveled, roadway departure crashes are generally more common on these rural roads, potentially even making up a large portion of an agency’s total crashes. Because these crashes are overrepresented on specific roadway types, analysts can focus evaluations and treatments on the roadway types where they are occurring. In addition, these crash types tend to be distributed around the network, rather than clustered at specific locations, and are often difficult to predict based on crash history. Therefore, they can some- times be missed when using a traditional crash-history-based, site-specific, safety management approach. This is where the systemic safety management approach is valuable to proactively find network locations which, regardless of crash history, are believed to have a high potential for future crashes based on the presence or absence of contributing factors. While high-quality crash data are not necessary to implement a systemic safety manage- ment approach, some states use recent crash history as an additional contributing factor that incorporates some elements of the traditional crash-history-based safety management approach to their systemic safety screening approach. 3.1.2 Screen and Prioritize Candidate Locations Once crash types, focus facility types, and crash contributing factors are identified, the next step is to conduct a network screening to prioritize locations for potential safety improvement. An agency can develop a network screening ranking methodology based on the presence or absence of the relevant contributing factors. Based on the assumption that target crashes will likely occur at locations where more contributing factors are present, sites with more contrib- uting factors present are locations with greater potential to reduce future crashes. A network screening ranking methodology can be developed that weighs each contributing factor equally or assigns relative weights to each of the contributing factors. If a weighting procedure is used, preferably it should be based on some measurable difference (e.g., CMF comparison or statistical correlation) as opposed to engineering judgment. This information helps agencies prioritize the sites where countermeasures that may be most appropriate to address the target crashes should be installed. 3.1.3 Countermeasure Selection To address the contributing factors at identified candidate locations, a selection of counter- measures should be identified for potential deployment. These countermeasures should be appropriate for the identified target crash types and facility types being considered in the systemic safety analysis. Additionally, countermeasures should be evidence based, proven to be applicable, and reliable within the context of the analysis. Ideally, the countermeasures should be low cost, feasible to be deployed widely across a system, and proven to provide significant crash reduction. Selections may also include some higher cost countermeasures to be deployed sparingly at locations that meet advanced thresholds. Countermeasure options should be evaluated for each target crash type and facility type based on effectiveness, imple- mentation costs, and unique agency policies, practices, and experiences, as these commonly influence project decision making.

30 Guide for Quantitative Approaches to Systemic Safety Analysis 3.1.4 Project Prioritization To determine which candidate locations receive different treatments, a decision-making framework should be established. It should be based on the countermeasures selected as well as contributing factors and other characteristics of each candidate location that impact the applicability, effectiveness, and viability of each treatment. The next step is to apply the decision-making framework to identify one or more specific countermeasures for potential implementation at each candidate location and then prioritize the projects for implementation by considering such factors as available funding, individual project costs, other programmed projects, time to develop project plans, expected crash reduction, public outreach, environ- mental constraints, right-of-way constraints, agency policies, and program goals. 3.2 Application of Safety Performance Functions (SPFs) for Systemic Safety Management A second variation to applying systemic safety management is through the use and application of SPFs for network screening (i.e., to prioritize locations for potential safety improvement). An SPF is an equation that predicts the number of crashes on a given roadway or at a specific intersection based on characteristics of the site. Agencies can use SPFs developed from their own data or existing SPFs calibrated to local conditions. The FHWA guidebook titled, Safety Performance Function Decision Guide: SPF Calibration vs SPF Development (Srinivasan et al., 2013), provides guidance to agencies concerning how to develop jurisdiction-specific SPFs or calibrate existing SPFs. SPFs used for network screening analyses within a systemic safety management approach do not necessarily require the same amount of data to develop or calibrate as SPFs intended for project level analysis such as the SPFs included in HSM Part C. With SPFs for network screening, the required inventory data elements associated with either SPF development or calibration are primarily used to define the facility types to which the SPFs apply. For SPFs used in project level analyses, typically the data requirements are greater since the data elements are used to both define the facility types to which the SPFs apply and address base conditions associated with CMFs used in conjunction with the SPFs. With SPFs for network screening, generally base conditions are undefined, which reduces the data requirements for SPF development or calibration. For example, several roadway and intersection inventory elements, which may be used to define the facility types and associated SPFs for network screening, are provide in Table 4. Description MIRE Data Element Roadway Segments Segment length Segment length (FDE) Area type Rural/urban designation (FDE) Number of through lanes (by direction) Number of through lanes (FDE) Median type (divided/undivided) Median type (FDE) Two-way versus one-way operation One-/two-way operations (FDE) Access control (freeway/non-freeway) Access control (FDE) Intersections Area type No relevant variable available Number of intersection legs Intersection/junction number of legsIntersection/junction geometry (FDE) Type of traffic control Intersection/junctional traffic control (FDE) Table 4. Potential inventory data elements for network screening SPFs.

Overview of Systemic Safety Management Approaches 31 Network screening SPFs can be used to calculate multiple performance measures such as predicted, expected, or excess crash frequencies. One or more performance measures are then selected to rank sites for their potential to reduce future crashes. When using SPFs in the context of systemic safety management, the purpose is not to calculate the predicted, expected, or excess crash frequencies for all crash types combined; that would essentially be implementing a crash-history-based safety management approach. Rather, the SPFs are used to calculate predicted, expected, or excess crash frequencies of target crash types such as lane departure or roadway departure crashes. Implementing a systemic safety management approach using SPFs requires only a few inventory data elements as presented in Table 4. If calculating expected or excess crash frequencies, high-quality crash data with accurate location information throughout the roadway network are required, and high-quality crash data are preferred for calibration when using existing SPFs to calculate predicted crashes. Traffic volume data are also needed for sites throughout the network. Some agencies may not have robust roadway inventory, traffic volume, and crash datasets for all of the roads within their jurisdiction. As a result, agencies may choose to calculate pre- dicted, expected, or excess crash frequencies using SPFs for only a portion of the roadways and intersections within their jurisdiction for network screening purposes. Other portions of the network for which they do not have high-quality roadway inventory, traffic volume, and crash datasets, cannot be addressed through a systemic safety management approach using SPFs. It is also possible that agencies may develop SPFs or calibrate existing SPFs for certain portions of their network for which they know they have a documented history of crashes. In such instances, agencies make a conscious decision to implement a systemic safety manage- ment approach using SPFs, focusing on a specific facility type or crash type. For example, an agency may identify through data analysis that a majority of the crashes that are occurring within its jurisdiction occur on rural two-lane roads, so it may make a decision to develop SPFs for rural two-lane roads and no other facility types. In this instance, the agency can implement a systemic safety management approach using SPFs addressing rural two-lane roads and then choose to implement another safety management approach (if any) for the other roads within its jurisdiction. In terms of the six-step safety management process and implementing a systemic safety management approach using SPFs, agencies use SPFs for network screening to prioritize sites for potential safety improvement based on predicted, expected, or excess crash frequencies of target crash types. The diagnosis and countermeasure selection process is relatively streamlined as the target crash type has already been identified; and, typically, the agency has already identi- fied one or more proven, low-cost countermeasures for potential implementation to address the target crash type, so diagnosis and countermeasure selection usually involves minimal effort. Ultimately, sites selected for safety improvement are based on economic measures such as benefit-cost ratios and available funding. Other factors, such as other programmed projects, time to develop project plans, expected crash reduction, public outreach, environmental constraints, right-of-way constraints, agency policies, and program goals, should also be considered when programming sites for safety improvement. To implement a systemic safety management approach using SPFs, agencies with the necessary technical expertise can develop their own SPFs and in-house tools for data management and network screening, or they can hire universities and/or consultants with the technical expertise to do so. Alternatively, agencies can use existing safety management software that utilizes SPFs to implement a systemic safety management approach. The most direct and comprehensive existing software that implements a systemic safety management approach using SPFs is the AASHTOWare Safety Analyst software. Safety Analyst implements the six main steps of the safety management process as outlined in Part B of the HSM (see Figure 1), incorporates

32 Guide for Quantitative Approaches to Systemic Safety Analysis SPFs and EB procedures in most analytical procedures, and includes a systemic site selection module that provides the capability to select an individual countermeasure and identify the most cost-effective sites for implementation. The remainder of this section highlights key features of the Safety Analyst software and the systemic site selection module. 3.2.1 Capabilities of Safety Analyst The Safety Analyst software incorporates the six main steps of the roadway safety manage- ment process into five modules as follows: • Module 1. Network Screening Tool. • Module 2. Diagnosis Tool and the Countermeasure Selection Tool. • Module 3. Economic Appraisal Tool and the Priority Ranking Tool. • Module 4. Countermeasure Evaluation Tool. • Module 5. Systemic Site Selection. A brief description of the first four modules within Safety Analyst is presented below to provide a general overview of the functionality and capabilities of the software as a whole, while a more detailed description of Module 5 is presented because it most directly relates to systemic safety management. In Module 1, network screening can be performed on an entire network or portion of the network to identify and rank individual sites based on the potential for improving safety. Two primary performance measures that can be used for ranking sites for potential safety improvement include average expected crash frequencies and excess crash frequencies. These performance measures combine predicted crash frequencies from SPFs and observed crashes to rank sites for potential safety improvement. Average expected crash frequencies and excess crash frequencies can be calculated for total, fatal and serious-injury, fatal and all-injury, property-damage-only, and equivalent property-damage-only crash severities. They can also be calculated for all collision types combined or for specific collision types. Network screening can be performed using a peak searching, sliding window, or simple ranking approach. Sites with higher than average expected crash frequencies or excess crash frequencies are considered to have greater potential for safety improvement and are given a higher priority for further evaluation with the other modules. Sites can also be screened and prioritized based on: • Modified LOSS (i.e., Mod LOSS), which measures the number of standard deviations above or below the mean of the average expected crash frequency (see Figure 4), C ra sh es Traffic Volume (AADT) Figure 4. Illustration of modified LOSS concept.

Overview of Systemic Safety Management Approaches 33 • A test of proportions that measures if the observed proportion of a given collision type at a given site is above the average proportion of the given collision type for similar sites, • Steady or sudden increases in mean crash frequencies, and • Crash frequencies and rates for extended corridors. In Module 2, as part of the diagnostic process, Safety Analyst provides several options for identifying crash patterns of interest for further diagnosis, including: • Crash summary reports, • Collision diagrams, • Safety performance reports, and • Statistical tests. The crash summary reports and collision diagrams present information based on observed crashes only. However, the safety performance reports and statistical tests provide additional information including several performance measures based on predicted and expected crash frequencies calculated using SPFs. The safety performance reports provide crash summaries for individual sites including: • Observed crash count, • Average observed crash rate, • Average observed crash frequency, • Average predicted crash frequency, • Average expected crash frequency, • Average excess crash frequency, and • Modified level of service of safety. Similarly, the statistical tests provide details for the average observed crash frequency and the average expected crash frequency. The combinations of crash summaries help the user to identify crash patterns of interest for further diagnosis and countermeasure selection. For common crash patterns of interest, Safety Analyst includes a series of diagnostic scenarios that guide the user through a series of questions. Depending upon the response input by the user, Safety Analyst outputs a list of recommended countermeasures to be considered for further economic analysis. In Module 2, analyses are performed on a site-by-site basis or groups of sites can be analyzed together as a single project. Module 2 is normally more critical for the crash-history-based safety management approach when network screening is performed looking at all crash types combined. Module 2 is not utilized as often for systemic safety management. In Module 3, Safety Analyst provides the capability to evaluate the benefits and costs of potential countermeasures and identify individual projects that are cost-effective or economi- cally justified. The procedures begin with Safety Analyst calculating the expected crash frequency for each year included in the history period. Then, Safety Analyst predicts the crash frequency at the site for the analysis period, for both the condition in which the counter- measure selected for analysis is implemented and the condition in which no change is made. Finally, Safety Analyst calculates the expected number of crashes and injuries reduced as the difference between predicted crash frequency for the countermeasure implementation condition and the no-change condition. With this information, Safety Analyst provides the option for evaluating the economic value of implementing a countermeasure based on several measures that include: • Cost-effectiveness index (expressed in terms of dollars spent per crash reduced), • EPDO cost effectiveness (expressed in terms of dollars spent per EPDO crash reduced this measure provides a weighting based on crash severity),

34 Guide for Quantitative Approaches to Systemic Safety Analysis • Benefit-cost ratio (expressed as ratio of the present safety benefit of a countermeasure to its construction costs), • Net benefit (expressed as the difference between safety benefits in monetary terms and construction costs), • Construction costs, • Safety benefits (expressed as the number of crashes reduced in monetary terms), • Number of total crashes reduced (i.e., all severity levels combined), • Number of fatal and serious-injury crashes reduced, and • Number of fatal and all-injury crashes reduced. Safety Analyst also provides the capability to maximize the net benefits across all sites and countermeasures under evaluation, taking into consideration budgetary constraints. The opti- mization process outputs the recommended countermeasures and implementation sites that will maximize the net benefits (i.e., difference between safety benefits and construction costs) given the available budget. Module 4 provides the capability to conduct before-after evaluations of implemented safety improvements to increase the knowledge of project effectiveness and supplement or improve the safety effectiveness measures used for economic analyses such as in Module 3. Safety Analyst uses the EB approach to compensate for regression to the mean and properly account for changes in safety that may be due to changes in other factors, such as traffic volume, to estimate the percent change in crash frequency. The safety effectiveness of a countermeasure can be based on all crash types combined or specific crash types and range of severity levels. Safety Analyst also provides the capability to evaluate shifts in proportions of crash types. Analyses can be performed to evaluate the effectiveness of individual countermeasures (or combinations of countermeasures) and construction projects. Benefit-cost analyses can also be conducted to assess the economic benefits of a countermeasure or construction project. Safety Analyst provides the capability to perform one or more steps of the roadway safety management approach without going through the entire six-step process. For example, if there is a known safety concern at a site that is not necessarily identified through network screening procedures, Safety Analyst can be used to identify recommended countermeasures and conduct economic analyses to select the most cost-effective countermeasure to remedy the safety concern. In this way, Safety Analyst is capable of performing the six main steps of the roadway safety management process in sequential order or as individual steps. Based on its functionality, Safety Analyst can be used to implement both the crash-history- based and the systemic safety management approaches. By implementing Modules 1, 2, and 3 in sequence and focusing on a site’s overall crash experience (i.e., all crash types combined), Safety Analyst can be used to conduct hot-spot analyses. Module 1 can be used to prioritize locations with potential for safety improvement based on overall crash experience. Module 2 can then be used to conduct detailed investigations of the high-priority locations from Module 1 to identify specific countermeasures to remedy the crash patterns of interest at the individual locations. Then, Module 3 can be used to perform an economic analysis of the proposed changes to determine the recommended mixture of countermeasures and sites to be improved that should result in the most cost-effective use of funds to improve safety. The operational workflow of Modules 1, 2, and 3 can also be used to implement more of a systemic safety management approach; however, Module 5 was recently incorporated within the software to more efficiently provide the capability to select an individual countermeasure that is intended to remedy a target crash type and identify the most cost-effective sites for implementation of the countermeasure. For example, if there is a known safety concern such as a pattern of run-off-the-road or head-on crashes along certain facility types, an appropriate

Overview of Systemic Safety Management Approaches 35 countermeasure (e.g., shoulder rumble strips or centerline rumble strips) can be selected for potential implementation, and through minimal input from the user, Safety Analyst can identify the most cost-effective sites to implement the countermeasure. Module 5, designated as the systemic site selection module, enables users to initiate a systemic safety analysis and prioritize sites for implementation of a specific countermeasure by efficiently working through, in a single module, modified inputs similar to Module 1 (network screening) and Module 3 (economic appraisal and priority ranking). The general workflow of the systemic site selection in Module 5 is as follows: Step 1. Select Site List for Analysis. The user selects a list of sites to analyze, which may consist of the entire roadway network under the jurisdiction of the agency or a portion of the network and include roadway segments, intersections, and ramps. Step 2. Select Option to Perform Systemic Site Selection. The user selects to conduct a Module 5 analysis to identify locations most appropriate for implementation of a selected countermeasure. Step 3. Select Site Type, Site Subtype, and Countermeasure for Analysis. The user selects one site type (i.e., roadway segments, intersections, or ramps) for inclusion in the systemic safety analysis and then selects a countermeasure (e.g., shoulder rumble strips or widen paved shoulder) to identify the most appropriate locations for implementation. Upon selection of a countermeasure, Safety Analyst automatically identifies the site subtypes to which the countermeasure is applicable and for some countermeasures provides default CMFs and crash type attributes associated with the countermeasure. The user then selects the site subtype(s) to include in the analysis. If default CMFs are not provided within Safety Analyst for the respective countermeasure and site subtype(s), the user must input a CMF(s) for analysis purposes. Step 4. Select Network Screening Methodology. The user selects the type of network screening to be performed to identify sites with crash patterns that can be remedied by the countermeasure selected for implementation. The user is given the option of selecting between “Conventional Network Screening” and “Percentage Report.” The Percentage Report provides the option to specify a maximum percentage of sites to include from the network screening results. The user then selects to conduct network screening based either on peak searching or sliding window approaches. Both network screening approaches use observed crash data and SPFs in combination with the EB approach to calculate expected crash frequencies and excess crash frequencies for individual sites or windows. When specifying the type of network screening to be performed, the user inputs the following for analysis: • Crash severity levels: – Total crashes, – Fatal and all-injury crashes, – Fatal and serious-injury crashes, – Property-damage-only crashes, – Equivalent property-damage-only crashes. • Analysis period, • Area weights, • Crash frequency limiting values, • EPDO weights by severity (as applicable), • Coefficient of variation (for peak searching approach), and • Sliding window parameters (for sliding window approach). When the CMF for the countermeasure is specific to a particular crash type or types, the network screening analysis ranks the sites with the highest expected or excess crash frequency

36 Guide for Quantitative Approaches to Systemic Safety Analysis of that particular crash type or types. If the CMF for the countermeasure is not for a specific crash type, the network screening analysis ranks the sites with the highest expected or excess crash frequency for total crashes (i.e., all crash types combined). Step 5. Specify Sites for Economic Analysis. The user has the option to select all ranked sites from the network screening, all screened sites regardless of whether they met the screening criteria specified in Step 4, or a specified group of sites for economic analysis. The user also selects whether the countermeasure is to be applied to the entire site (i.e., roadway segment) or the window with the highest potential for safety improvement. Step 6. Specify Parameters for Economic Analysis. The user specifies select input parameters or threshold values for performing an economic analysis to identify the most cost-effective locations to implement the countermeasure, such as the following: • Expected implementation year, • Number of years to analyze, • Economic and ranking criteria: – Cost effectiveness, – EPDO-based cost effectiveness, – Benefit-cost ratio, – Net benefits, – Construction costs, – Safety benefits, – Number of total crashes reduced, – Number of fatal and serious-injury crashes reduced, – Number of fatal and all-injury crashes reduced. • Crash costs and weights by severity, • Optimize net benefits within construction budget, ranking by: – Total crashes reduced per mile, – Total crashes reduced per site, – Net benefits per mile, – Net benefits per site. • Total budget for countermeasure construction. When the user selects optimization based on net benefits and specifies the budget amount, the primary output is a list of the sites where the countermeasure is economically justi- fied for installation, presented in rank order by total crashes reduced per mile, total crashes reduced per site, net benefits per mile, or net benefits per site. With the general workflow of the systemic site selection module, the diagnosis and counter measure selection steps of the roadway safety management approach are unnecessary because of previous decisions. Therefore, these steps are not included in the workflow of Module 5. 3.2.2 Input Data Needs for Safety Analyst Table 5 summarizes the required data elements for Safety Analyst. The required data elements include inventory data for roadway segments, intersections and/or ramps; traffic volume data; and crash data. Lack of a required data element will invalidate a site or crash or prevent it from being analyzed. Most of the required data elements are FDEs as designated in MIRE. Table 5 indicates if the required Safety Analyst data element is an FDE. States are required to have access to a complete collection of the MIRE FDEs on all public roads by September 30, 2026 (23 CFR 924.11b).

Overview of Systemic Safety Management Approaches 37 Data Element Description Roadway Segment Data Segment ID (MIRE FDE) This item is a unique, agency-specific identifier for the roadway segment. Route type (MIRE FDE) The value of this item is the category of the route where the site is located. This item should be included whether it is part of the location identifier or not, as searches may be conducted separately on this item. Default values include: • Interstate, • U.S. route, • State route, • Business route, • Business loop, • Spur route, • County road, • Township road, • Local road, • Other, and • Unknown. Route name (MIRE FDE) The value of this item is the number or name of the route where the site is located. Where routes overlap, the more important route type and the corresponding lower route number normally take precedence. For routes without numbers, the road or street name should be used. Area type (MIRE FDE) This item characterizes the area in which the site is located. This item is required for the automatic assignment of a subtype to the site. Default values include: • Urban, • Rural, and • Unknown. Start location (MIRE FDE) The location of the starting point of the roadway segment, represented in one of the four location systems supported by Safety Analyst. Represents a location reference value in one of the four supported reference systems: Route/Milepost, Route/County/Milepost, Section/Distance, or Route/Section/Distance. End location (MIRE FDE) The location of the ending point of the roadway segment, represented in one of the four location systems supported by Safety Analyst. Represents a location reference value in one of the four supported reference systems: Route/Milepost, Route/County/Milepost, Section/Distance, or Route/Section/Distance. Segment length (MIRE FDE) The value of this item is the length of the segment. In some cases, the segment length may be computed directly from the milepost values by finding the difference between Ending Milepost and Beginning Milepost. In other cases, this length may have to be derived by other means, depending upon the location system being used. Median type (MIRE FDE) The value of this item is the indication of the type and characterization of the area separating opposing traffic lanes. Default values include: • Divided—opposing traffic lanes divided, • Undivided—opposing traffic lanes undivided, and • Unknown—unknown if traffic lanes are separated. Access control (MIRE FDE) The value of this item is the degree that access to abutting land in connection with a highway is fully, partially, or not controlled by public authority. Default values include: • Full access control—public authority has full access control; • Partial access control—public authority has partial access control; • No access control—public authority has no access control; and • Unknown—public authority has unknown access control. Two-way versus one-way operation (MIRE FDE) The value of this item is an indication of whether or not a roadway serves one-way or two-way traffic. Default values include: • One-way road or street—roadway serves one-way traffic; • Two-way road or street—roadway serves two-way traffic; • One direction of travel for a divided highway—roadway serves one direction of travel for a divided highway; and • Unknown—roadway serves unknown traffic. Table 5. Required Safety Analyst data elements (adapted from AASHTO, 2016). (continued on next page)

38 Guide for Quantitative Approaches to Systemic Safety Analysis Data Element Description Roadway Segment Directional Attributes Direction (MIRE FDE) This item identifies one of two opposing directions on the roadway segment to which the subelements and items contained in the directional element apply. Safety Analyst uses this item to distinguish the two opposing directions on a roadway segment without inferring any specific direction of travel. Number of through lanes (MIRE FDE) The value of this item is the total number of through lanes in this direction of travel. Turn lanes and auxiliary lanes are not included in this count. Intersection Data Intersection ID (MIRE FDE) This item is a unique, agency-specific identifier for the intersection. Location (MIRE FDE) The location of the intersection, represented in one of the four location systems supported by Safety Analyst: Route/Milepost, Route/County/Milepost, Section/Distance, or Route/Section/Distance. Area type (MIRE FDE) This item characterizes the area in which the site is located. This item is required for the automatic assignment of a subtype to the site. Default values include: • Urban, • Rural, and • Unknown. Intersection type (MIRE FDE) The type of intersection at which two or more roadways intersect at grade (i.e., Tee intersection, Y intersection, four-leg intersection, traffic circle/roundabout, multileg—five or more legs—intersection). Default values include: • Tee intersection—two or more roadways intersect at grade in a Tee intersection, • Y intersection—two or more roadways intersect at grade in a Y intersection, • Four-leg intersection—two or more roadways intersect at grade in a four- leg intersection, • Traffic circle/roundabout—two or more roadways intersect at grade in a traffic circle or roundabout, • Multileg intersection, five or more legs—two or more roadways intersect at grade in a multileg intersection of five or more legs, • Other—two or more roadways intersect at grade in another intersection type, and • Unknown—two or more roadways intersect at grade in an unknown intersection type. Type of traffic control (MIRE FDE) The type of traffic control at the intersection. Default values include: • Stop signs on cross street only (minor-road stop)—traffic control at intersection consists of stop signs on cross street only; • All-way stop signs (all-way stop)—traffic control at intersection consists of all-way stop signs; • Signals; • No control (unsupported)—no traffic control at intersection; • Yield signs on cross street only (unsupported)—traffic control at intersection consists of yield signs on cross street only; • Roundabout (unsupported)—traffic control at intersection consists of roundabout; and • Unknown (unsupported)—unknown traffic control at intersection. Interchange influence area on mainline freeway The value of this item indicates whether or not a roadway segment is within an interchange influence area. In general terms, the limits of mainline freeway segments within interchange areas are defined to extend approximately 0.3 mi upstream from the gore (i.e., painted nose of the gore area) of the first ramp of a particular interchange to approximately 0.3 mi downstream from the gore (i.e., painted nose of the gore area) of the last ramp of the given interchange. Conversely, all mainline freeway segments that extend beyond these defined limits for interchange areas are, by definition, mainline freeway segments outside an interchange area. Default values include: • Yes, roadway is within interchange influence area; • No, roadway is not within interchange influence area; and • Unknown, unknown whether roadway is within interchange influence area. Roadway Segment Data (continued) Table 5. (Continued).

Overview of Systemic Safety Management Approaches 39 Data Element Description Ramp Data Ramp configuration (MIRE FDE) Describes the characterization of the design of the ramp. Default values include: • Diamond, • Partial clover leaf loop (i.e., parclo loop), • Free-flow loop, • Free-flow outer connection, • Direct or semi-direct connection, • Collector-distributor (C-D) road or other connector, • Other, and • Unknown. Ramp length (MIRE FDE) Total length of ramp, including horizontal curves and tangents. Agencies should use a fixed convention for measuring ramp length, such as starting the measurement at the gore area. Traffic Data Roadway segment AADT (MIRE FDE) The value of this item is the AADT for the associated inventory element. For roadway segments, this is the average number of vehicles passing through a segment from both directions of the mainline route for all days of a specified year. Major road AADT (MIRE FDE) The value of this item is the AADT for the associated inventory element. For intersections, this is the average number of vehicles passing through an intersection from both directions of the major roadways for all days of a specified year. Minor road AADT (MIRE FDE) The value of this item is the AADT for the associated inventory element. For intersections, this is the average number of vehicles passing through an intersection from both directions of the minor roadways for all days of a specified year. Ramp AADT (MIRE FDE) The value of this item is the AADT for the associated inventory element. For ramps, this is the average number of vehicles traversing the ramp in one direction for all days of a specified year. Year (MIRE FDE) The value of this item is the calendar year for which the associated traffic data are applicable. Crash Data Crash ID The length and format of this agency-specific identifier varies among states and is used to link subfiles of vehicles and occupants to crashes. All states assign crashes a code number of this type (often known as the crash report number or case number). If case identifiers may be repeated in subsequent years, the two-digit year should be added as a prefix to make the identifier unique. Location The location of the crash, represented in one of the four location systems supported by Safety Analyst. Represents a location reference value in one of the four supported reference systems: Route/Milepost, Route/County/Milepost, Section/Distance, or Route/Section/Distance. Ramp type Indicates whether the ramp is used to enter or exit a freeway or connect two freeways. Default values include: • Off ramp—exit freeway, • On ramp—enter freeway, • Freeway-to-freeway ramp—connect two freeways, • Other—other type of ramp, and • Unknown—unknown type of ramp. Ramp ID (MIRE FDE) This item is a unique, agency-specific identifier for the ramp. Area type (MIRE FDE) This item characterizes the area in which the site is located. This item is required for the automatic assignment of a subtype to the site. Default values include: • Urban, • Rural, and • Unknown. Start location (MIRE FDE) The location of the ramp, represented in one of the four location systems supported by Safety Analyst. Represents a location reference value in one of the four supported reference systems: Route/Milepost, Route/County/Milepost, Section/Distance, or Route/Section/Distance. Table 5. (Continued). (continued on next page)

40 Guide for Quantitative Approaches to Systemic Safety Analysis Data Element Description Crash Data (continued) Relationship to junction Identifies the type of related cross street to the crash site. Default values include: • Non-junction, • At intersection, • Intersection related, • At driveway or driveway related, • Entrance/exit ramp, • Other part of interchange, • Railroad/highway grade crossing, • Crossover related, • Other, and • Unknown. Crash type and manner of collision The type of first harmful event in a single-vehicle crash or in a multiple-vehicle collision, manner in which two vehicles in transport initially came together without regard to the direction of force, or the type of object with which a single vehicle collided. Default values include: • Single-Vehicle Collisions: – Collision with parked motor vehicle, – Collision with railroad train, – Collision with bicyclist, – Collision with pedestrian, – Collision with animal, – Collision with fixed object, – Collision with other object, – Other single-vehicle collision, – Overturn, – Fire or explosion, and – Other single-vehicle non-collision, and – Unknown. • Multi-Vehicle Collisions: – Rear-end, – Head-on, – Rear-to-rear, – Angle, – Sideswipe, same direction, – Sideswipe, opposite direction, – Other multiple-vehicle collision, and – Unknown. Number of vehicles involved The count of motor vehicles (e.g., automobiles, single-unit trucks, truck combinations that are in motion or on a roadway) involved in the crash. (Note: Parked vehicles are not included in this vehicle count, nor are bicycles and pedestrians.) Crash date The date (year, month, and day) on which the crash occurred. Crash severity The severity of the crash based on the most severe injury to any person involved. Default values include: • Fatal injury, • Serious injury, • Minor injury, • Possible injury, • Property damage only, and • Unknown. Number of fatalities The number of all fatalities (drivers, occupants, pedestrians) resulting from this specific motor vehicle crash. Number of non-fatal injuries The number of all injured persons (drivers, occupants, pedestrians), excluding fatalities, resulting from this specific motor vehicle crash. Table 5. (Continued).

Overview of Systemic Safety Management Approaches 41 Agencies can still use Safety Analyst if they do not have all of the required data elements. For example, if an agency has a roadway inventory database, but not an intersection or ramp inventory database, the agency can still bring its roadway data into the software and use Safety Analyst to analyze roadway segments but not intersections and ramps. Similarly, most state agencies only have comprehensive inventory data for their statewide network and/or inter sections of state routes but not comprehensive inventory data for local routes or intersections of state and local roads. Therefore, most state agencies initially develop their Safety Analyst dataset for their statewide network so they can analyze their statewide network with the goal to incorporate and analyze data for local roads and intersections at a later date. In addition to the required data elements, other data can be brought into Safety Analyst for varying purposes. These data elements can be categorized as follows. • Conditionally Required: These data elements are necessary to meet particular conditions. For example, if an agency is using section-distance locations, then the section ID becomes a mandatory data item; it is not required for route-milepost or route-county-milepost loca- tion systems. Also, the travel direction for a segment is required conditional to the operation way (i.e., if the segment represents one side of a divided highway, then the travel direction is required). • Functionally Required: These data elements are necessary to support a particular func- tionality in the software. For example, the number of auxiliary lanes is required by some countermeasure functions, so it may be needed to support some types of analyses but not others. This also applies to some data items used to support collision diagrams (e.g., direction of increasing mileposts or leg type and direction). • Optional: These data elements do not support any significant functionality in Safety Analyst. Generally, these items are displayed somewhere in the geographic user interface or can be used when querying sites for the site list. Querying for sites is a “function” in Safety Analyst, but it is not significant to the analysis of a site. Several conditionally required, functionally required, and optional data elements are MIRE FDEs. Safety Analyst provides the capability to analyze a wide range of crash types. As long as the crash types of interest are included in the Safety Analyst datasets, Safety Analyst provides the capability to identify and prioritize sites where the crash type is expected. In particular, the functionality to be able to conduct network screening for particular target crashes is critical for Safety Analyst to be used for systemic safety analysis. The Safety Analyst software specifies default crash types for which the software is designed to analyze; however, agencies have the capability to modify the values or categories of each crash data element to more closely match their crash type definitions and categories. Agencies may also incorporate agency-defined data elements into Safety Analyst for analysis purposes. Safety Analyst includes information for 417 countermeasures for use in systemic site selec- tion, diagnosis and countermeasure selection, and economic appraisal and priority ranking of potential countermeasures for implementation. Of these, only a small percentage of the countermeasures have information related to their expected safety effectiveness in reducing crashes and the related crash types they are intended to remedy, but the user has the capability to edit the default safety effectiveness information provided within Safety Analyst and/or input the expected safety effectiveness of a countermeasure for analysis purposes. Most of the countermeasures within Safety Analyst for which CMFs are provided are included in the first edition of the HSM.

42 Guide for Quantitative Approaches to Systemic Safety Analysis 3.2.3 Results Provided by Safety Analyst The primary output reports provided by Safety Analyst for the systemic site selection module (Module 5) are described in this section. Output reports are provided for viewing and use as hypertext markup language (HTML), rich text format (RTF), portable document format (PDF), and comma-separated values (CSV) files. Two types of output reports are generated from Module 5 analyses. The first type of output report presents the network screening results based on the initial steps of the systemic site selection process. The network screening output report lists, in priority order, the sites that were analyzed, met the screening criteria, and have the highest potential for safety improvement. Entire sites are presented in priority order based on the expected crash frequency, excess crash frequency, modified LOSS of the peak window for roadway segments, or based on a simple ranking criteria for intersections and ramps. The number of expected and excess fatalities and injuries (i.e., at the person level) is also included on the output report. Table 6 presents sample network screening results that identify the highest ranked sites along rural two-lane highways for potentially installing shoulder rumble strips to reduce the frequency of run-off-the-road crashes. The analysis is based on fatal and all-injury crashes. Sites meet the screening criteria if the peak window has at least 0.36 fatal and all-injury, run-off-the-road crashes/mi/yr. The sites are ranked based on expected crash frequency. Twenty sites met the screening criteria. Site 4148 is the highest ranked site and has an expected crash frequency of 0.61 fatal and all-injury run-off-the road crashes/mi/yr based on the peak window from milepost (MP) 19.985 to 20.085 on Route 142570 in County 18. The entire length of Site 4148 is from MP 19.485 to 24.148. Site 6204 is the second highest ranked site and has an expected crash frequency of 0.57 fatal and all-injury run-off-the road crashes/mi/yr based on the peak window from MP 5.016 to 5.116 on Route 142301 in County 18. Of the sample sites considered in the analysis, the 20 sites in Table 6 represent the sites where it potentially makes the most sense to install shoulder rumble strips to reduce fatal and all-injury run-off-the-road crashes based on expected crash frequency. Further economic analyses should be conducted as part of the systemic site selection process to determine where it is most economical to install the shoulder rumble strips along the network. The second type of output report from Module 5 is generated from the economic analysis. Three types of economic output reports can be generated: economic appraisal, priority ranking, and optimization results. The economic appraisal table provides a simple means for comparing and evaluating the proposed countermeasure based on up to four economic criteria, including cost effectiveness, EPDO-based cost effectiveness, benefit-cost ratio, and net benefits. In this table, the economic results are not presented in a particular prioritized order. The second group of economic output reports, referred to as priority ranking tables, present the proposed countermeasure in prioritized order based on the user-selected ranking criteria, including cost effectiveness, EPDO-based cost effectiveness, benefit-cost ratio, net benefits, construction costs, safety benefits, number of total crashes reduced, number of fatal and serious-injury crashes reduced, and number of fatal and all-injury crashes reduced. A separate table is provided for each ranking criterion. The final type of economic output report provides the optimization results in a table that presents the mixture of sites in priority order where the countermeasure could be implemented to provide the most economic benefit for a given specified budget based on either total crashes reduced per mile, total crashes reduced per site, net benefits per mile, or net benefits per site. Table 7 presents a sample output of optimization results considering implementation of continuous milled-in shoulder rumble strips along the 20 rural two-lane highways with the highest expected frequency of run-off-the-road crashes (see Table 6). Sites are presented in

ID Site Type Site Subtype County Route Site Start Location Site End Location Average Observed Crashes for Entire Site1 Location with Highest Potential for Safety Improvement Rank Additional Windows of Interest Average Observed Crashes1 Predicted Crash Frequency1 Expected Crash Frequency Excess Crash Frequency Modified LOSS Start Location End LocationExpected Frequency1 Variance2 No. of Fatalities No. of Injuries Excess Frequency1 Variance2 No. of Fatalities No. of Injuries Cat 4148 Seg Seg/Rur; 2-lane 18 1425710 19.485 24.148 0.23 3.00 0.22 0.61 0.01 0.02 0.68 0.40 0.03 0.01 0.44 0.81 III 19.985 20.085 1 '20.285_20.385; 20.385_20.485; 21.285_21.385; 21.685_21.785; 22.385_22.485; 23.885_23.985' 6204 Seg Seg/Rur; 2-lane 18 1427301 2.516 6.204 0.29 3.09 0.20 0.57 0.01 0.02 0.64 0.37 0.03 0.01 0.41 0.80 III 5.016 5.116 2 '2.616_2.716; 3.716_3.816; 4.716_4.816; 5.816_5.916; 6.016_6.116' 2516 Seg Seg/Rur; 2-lane 18 1427301 0.000 2.516 0.48 2.84 0.20 0.56 0.01 0.02 0.62 0.37 0.03 0.01 0.41 0.80 III 2.000 2.100 3 '0.500_0.600; 0.700_0.800; 1.000_1.100; 1.600_1.700; 1.800_1.900; 2.100_2.200' 8177 Seg Seg/Rur; 2-lane 17 551706 4.826 8.177 0.18 2.79 0.19 0.55 0.01 0.02 0.61 0.36 0.03 0.01 0.40 0.79 III 6.926 7.026 4 '' 6628 Seg Seg/Rur; 2-lane 17 551310 13.576 16.628 0.17 2.43 0.18 0.51 0.01 0.02 0.56 0.33 0.02 0.01 0.37 0.75 III 13.976 14.076 5 '' 0306 Seg Seg/Rur; 2-lane 14 502809 8.991 10.306 0.71 4.65 0.12 0.51 0.01 0.02 0.56 0.38 0.01 0.01 0.42 1.06 III 9.691 9.791 6 '' 6275 Seg Seg/Rur; 2-lane 16 933209 5.027 6.275 0.26 1.54 0.27 0.49 0.01 0.02 0.54 0.22 0.04 0.01 0.24 0.40 III 5.127 5.227 7 '5.727_5.827' 1166 Seg Seg/Rur; 2-lane 17 551706 8.177 11.166 0.34 1.46 0.19 0.47 0.02 0.02 0.52 0.29 0.03 0.01 0.32 0.64 III 9.277 9.477 8 '' 7932 Seg Seg/Rur; 2-lane 15 899407 4.302 7.932 0.32 2.69 0.16 0.46 0.01 0.02 0.51 0.30 0.02 0.01 0.34 0.74 III 5.702 5.802 9 '7.402_7.502' 3065 Seg Seg/Rur; 2-lane 15 899004 0.000 3.065 0.20 1.42 0.25 0.46 0.01 0.02 0.51 0.20 0.04 0.01 0.23 0.38 III 2.000 2.100 10 '2.200_2.300; 2.500_2.600; 2.800_2.900' 0265 Seg Seg/Rur; 2-lane 16 933209 6.275 10.265 0.12 1.53 0.25 0.45 0.01 0.02 0.50 0.20 0.04 0.01 0.22 0.38 III 6.875 6.975 11 '8.175_8.275; 9.575_9.675' 9443 Seg Seg/Rur; 2-lane 14 503510 6.128 9.443 0.17 2.58 0.15 0.45 0.01 0.02 0.49 0.29 0.02 0.01 0.32 0.73 III 9.328 9.428 12 '9.343_9.443' 2453 Seg Seg/Rur; 2-lane 12 565810 1.207 2.453 0.41 2.58 0.15 0.44 0.01 0.02 0.48 0.29 0.02 0.01 0.32 0.72 III 2.107 2.207 13 '' 5832 Seg Seg/Rur; 2-lane 16 932308 0.859 5.832 0.06 2.90 0.15 0.43 0.01 0.02 0.48 0.29 0.02 0.01 0.32 0.73 III 1.859 1.959 14 '' 6556 Seg Seg/Rur; 2-lane 18 1425710 9.381 16.556 0.25 2.78 0.15 0.43 0.01 0.02 0.48 0.29 0.02 0.01 0.32 0.73 III 12.281 12.381 15 '15.181_15.281' 3524 Seg Seg/Rur; 2-lane 18 1431908 0.000 3.524 0.18 1.49 0.22 0.41 0.01 0.01 0.45 0.19 0.03 0.01 0.21 0.38 III 1.000 1.100 16 '1.200_1.300; 2.300_2.400; 2.400_2.500' 7164 Seg Seg/Rur; 2-lane 17 551310 0.000 7.164 0.12 1.37 0.14 0.38 0.01 0.01 0.42 0.24 0.02 0.01 0.26 0.62 III 1.900 2.100 17 '2.000_2.200' 5172 Seg Seg/Rur; 2-lane 15 899310 13.009 15.172 0.17 2.20 0.13 0.37 0.00 0.01 0.41 0.25 0.01 0.01 0.27 0.67 III 13.909 14.009 18 '' 7445 Seg Seg/Rur; 2-lane 12 566510 1.497 7.445 0.20 1.48 0.14 0.37 0.01 0.01 0.41 0.23 0.02 0.01 0.26 0.62 III 3.997 4.197 19 '' 1778 Seg Seg/Rur; 2-lane 12 565703 0.303 1.778 0.21 2.84 0.12 0.36 0.00 0.01 0.40 0.24 0.01 0.01 0.27 0.68 III 1.503 1.603 20 '' 1. Units for Observed, Predicted, Expected, and Excess Crash Frequency: – Roadway segments: crashes/mi/yr, – Intersections: crashes/yr, and – Ramps: crashes/mi/yr. 2. Units for Variance: – Roadway segments: (crashes/mi)2/yr, – Intersections: crashes2/yr, and – Ramps: (crashes/mi)2/yr. Note: The final column, Additional Windows of Interest, shows those windows that meet the screening criteria, but have an expected (or excess) crash frequency less than the peak window. Table 6. Sample Safety Analyst systemic site selection: network screening results.

Proposed Site-CM Site ID Site Type County Route Beginning Location Ending Location Countermeasure CM Start Location CM End Location Construction Cost for Single Implementation Safety Benefit Present Value of Construction Cost for Analysis Period Net Benefits per Site Net Benefits per Mile Total Crashes Reduced per Site* Total Crashes Reduced per Mile* 19 7445 Seg/Rur; 2-lane 12 566510 1.497 7.445 Install continuous milled-in shoulder (or edgeline) rumble strips 1.497 7.445 $8,922 $3,157,561 $14,949 $3,142,611 $528,348 36.21 6.09 16 3524 Seg/Rur 2-lane 18 1431908 0.000 3.524 Install continuous milled-in shoulder (or edgeline) rumble strips 0.000 3.524 $5,286 $2,803,583 $8,857 $2,794,726 $793,055 17.85 5.07 15 6556 Seg/Rur; 2-lane 18 1425710 9.381 16.556 Install continuous milled-in shoulder (or edgeline) rumble strips 9.381 16.556 $10,763 $2,254,955 $18,033 $2,236,921 $311,766 14.22 1.98 17 7164 Seg/Rur; 2-lane 17 551310 0.000 7.164 Install continuous milled-in shoulder (or edgeline) rumble strips 0.000 7.164 $10,746 $1,680,840 $18,006 $1,662,834 $232,110 13.71 1.91 14 5832 Seg/Rur; 2-lane 16 932308 0.859 5.832 Install continuous milled-in shoulder (or edgeline) rumble strips 0.859 5.832 $7,460 $728,645 $12,499 $716,146 $144,007 9.33 1.88 2 6204 Seg/Rur; 2-lane 18 1427301 2.516 6.204 Install continuous milled-in shoulder (or edgeline) rumble strips 2.516 6.204 $5,532 $1,305,400 $9,269 $1,296,131 $351,445 7.92 2.15 4 8177 Seg/Rur; 2-lane 17 551706 4.826 8.177 Install continuous milled-in shoulder (or edgeline) rumble strips 4.826 8.177 $5,026 $564,834 $8,422 $556,412 $166,043 6.15 1.84 1 4148 Seg/Rur; 2-lane 18 1425710 19.485 24.148 Install continuous milled-in shoulder (or edgeline) rumble strips 19.485 24.148 $6,994 $869,347 $11,720 $857,627 $183,922 5.90 1.27 10 3065 Seg/Rur; 2-lane 15 899004 0.000 3.065 Install continuous milled-in shoulder (or edgeline) rumble strips 0.000 3.065 $4,598 $890,929 $7,703 $883,226 $288,165 5.56 1.82 3 2516 Seg/Rur; 2-lane 18 1427301 0.000 2.516 Install continuous milled-in shoulder (or edgeline) rumble strips 0.000 2.516 $3,774 $906,394 $6,324 $900,071 $357,739 5.35 2.13 6 0306 Seg/Rur; 2-lane 14 502809 8.991 10.306 Install continuous milled-in shoulder (or edgeline) rumble strips 8.991 10.306 $1,972 $482,598 $3,305 $479,293 $364,481 4.40 3.35 7 6275 Seg/Rur; 2-lane 16 933209 5.027 6.275 Install continuous milled-in shoulder (or edgeline) rumble strips 5.027 6.275 $1,872 $540,478 $3,137 $537,341 $430,562 4.07 3.26 9 7932 Seg/Rur; 2-lane 15 899407 4.302 7.932 Install continuous milled-in shoulder (or edgeline) rumble strips 4.302 7.932 $5,445 $492,633 $9,123 $483,510 $133,198 3.68 1.01 8 1166 Seg/Rur; 2-lane 17 551706 8.177 11.166 Install continuous milled-in shoulder (or edgeline) rumble strips 8.177 11.166 $4,484 $302,767 $7,512 $295,254 $98,780 3.60 1.20 11 0265 Seg/Rur; 2-lane 16 933209 6.275 10.265 Install continuous milled-in shoulder (or edgeline) rumble strips 6.275 10.265 $5,985 $404,996 $10,028 $394,968 $98,990 3.13 0.79 12 9443 Seg/Rur; 2-lane 14 503510 6.128 9.443 Install continuous milled-in shoulder (or edgeline) rumble strips 6.128 9.443 $4,972 $104,990 $8,332 $96,658 $29,158 1.62 0.49 20 1778 Seg/Rur; 2-lane 12 565703 0.303 1.778 Install continuous milled-in shoulder (or edgeline) rumble strips 0.303 1.778 $2,212 $205,119 $3,707 $201,412 $136,551 1.56 1.06 13 2453 Seg/Rur; 2-lane 12 565810 1.207 2.453 Install continuous milled-in shoulder (or edgeline) rumble strips 1.207 2.453 $1,869 $231,050 $3,132 $227,918 $182,920 1.53 1.23 5 6628 Seg/Rur; 2-lane 17 551310 13.576 16.628 Install continuous milled-in shoulder (or edgeline) rumble strips 13.576 16.628 $4,578 $89,658 $7,671 $81,987 $26,863 0.56 0.18 18 5172 Seg/Rur; 2-lane 15 899310 13.009 15.172 Install continuous milled-in shoulder (or edgeline) rumble strips 13.009 15.172 $3,245 $15,862 $5,436 $10,426 $4,820 0.11 0.05 Totals: $105,735 $18,032,639 $177,166 $17,855,473 $4,862,923 146.48 38.73 * These are estimates of the number of crashes reduced and not a guarantee. CM = countermeasure. Table 7. Sample Safety Analyst systemic site selection: optimization results (ranked by total crashes reduced per site).

Overview of Systemic Safety Management Approaches 45 priority order for implementation of continuous milled-in shoulder rumble strips based on the total number of crashes reduced per site. Site 7445 represents the highest ranked site for implementation of continuous milled-in shoulder rumble strips. Installation of this counter- measure at Site 7445 is expected to result in a reduction of 36.21 crashes at this site over a 20-year analysis period (analysis period is based on user input). Similarly, Site 3524 represents the second highest ranked site for implementation of continuous milled-in shoulder rumble strips. Installation of this countermeasure at Site 3524 is expected to result in a reduction of 17.85 crashes at this site over a 20-year period. With a budget of a little less than $106,000, continuous milled-in shoulder rumble strips could be installed along the entirety of these 20 sites with an expected net benefit of $17,855,473 and total crashes reduced of 146.48, amounting to $4,862,923 per mile in savings over the analysis period (i.e., 20 years). 3.2.4 Key Strengths and Advantages of Safety Analyst Key strengths and advantages of the Safety Analyst software are as follows: • The software improves the effectiveness of decision making by automating state-of-the-art statistical approaches described in HSM Part B to improve the identification and program- ming of site-specific highway safety improvements. • The software improves the efficiency of decision support by integrating all parts of the safety management process into a single, modular software package. • The software includes default SPFs that are automatically calibrated to local conditions using an agency’s crash data. The software also provides the capability to enter alternative SPF functional forms and parameter values for use in analyses. • The software provides the capability to adjust input values quickly and rerun analyses in an efficient manner. • Most of the required data elements for the software are FDEs, for which all states are required to have access on all public roads by September 30, 2026. The software also makes use of existing roadway inventory, traffic volume, and crash databases when available. • Decisions on where to systemically place selected countermeasures are based on quantitative economic analyses. Sites for economic analysis can be based on all ranked sites from network screening, all screened sites, or selected sites from the site list. • Full crash history is reviewed for each site. • Statistical procedures explicitly account for traffic volume. • All of the default countermeasures included in the software, with safety effectiveness infor- mation provided, are considered reliable. If safety effectiveness information is missing for a countermeasure, the user has the capability to input the information as part of the analysis. • The software includes a large amount of support information, including detailed steps of all calculations. 3.2.5 Key Weaknesses or Limitations of Safety Analyst Key weaknesses or limitations of the Safety Analyst software are as follows: • The software requires mapping and importing of data elements, which can be quite burden- some to agencies. • Systemic site selection procedures require preliminary analyses to identify focus crash types and countermeasures, similar to Step 1 of the Systemic Tool, which must be performed outside of Module 5 (Systemic Site Selection). • Systemic site selection procedures only consider one countermeasure at a time. Multiple countermeasures cannot be considered in systemic site selection.

46 Guide for Quantitative Approaches to Systemic Safety Analysis • Limited information on reliable countermeasures is incorporated into the software due to the current state of knowledge on countermeasures. • The software requires purchasing an annual license. At the time of this publication, only a few states have limited experience using the systemic site selection module since this functionality was only recently added to the software. This should not be viewed as a weakness or limitation of Safety Analyst. It simply reflects the evolution and continued enhancement of the software. As indicated in part 3.2.1, the func- tionality and capabilities of Safety Analyst can be used to implement both the crash-history- based and the systemic safety management approaches. For years, several agencies have been using Safety Analyst for network screening, identifying potential countermeasures, and prioritizing and selecting infrastructure-related safety improvement projects; however, it has been in the context of using the crash-history-based safety management approach. Over time, as licensed agencies become more familiar with the systemic site selection module in Safety Analyst, it is anticipated that the functionality and capabilities of Module 5 will be more fully utilized to identify the most cost-effective sites for implementation of specific countermeasures. 3.2.6 Getting Access to Safety Analyst Agencies purchase an annual license to use Safety Analyst from AASHTO. AASHTO provides licensing details in its annual AASHTOWare Catalog and on its AASHTOWare website (https://www.aashtoware.org/). The catalog is typically available in May for the next fiscal year. Several licensing options usually exist. For example, for fiscal year 2020, the annual fee for a single workstation was $21,200, and the annual fee for a site license was $37,200. With the Safety Analyst license, an agency receives a designated number of technical support hours for engineering-related issues. In addition, engineering (and data management) support webinars are held for licensed agencies to attend. Additional support (e.g., for onsite training) can be purchased through service units in special fixed-fee increments. In fiscal year 2020, 12 state/provincial transportation agencies and one federal agency licensed Safety Analyst. 3.3 Application of U.S. Road Assessment Program (usRAP) Methodology and ViDA Software The ViDA software, available through the usRAP, is a safety management tool used to rate the safety of a roadway based on an assessment of the presence and condition of roadway, roadside, and intersection design elements and to identify cost-effective countermeasures to reduce fatal and serious-injury crashes (usRAP, 2019). ViDA was developed cooperatively by usRAP and its international partner, the International Road Assessment Programme (iRAP). The ViDA software is available for use by any public agency and its consultants at no cost. (For those familiar with the usRAP tools software, ViDA is its successor.) The usRAP ViDA software uses crash prediction models, rather than site-specific crash history data, to estimate the likelihood of fatal and serious-injury crashes. While crash data are not required to use ViDA, a limited amount of networkwide crash history data is useful for calibrat- ing the crash prediction models to local conditions. Where site-specific crash data are available, crash patterns at a location of interest can be considered during the design and implementation phases of the safety projects recommended in the ViDA results.

Overview of Systemic Safety Management Approaches 47 3.3.1 Capabilities of usRAP ViDA The ViDA software can be used to perform two types of analysis: • Develop star ratings and • Develop safer roads investment plans. Star ratings and safer roads investment plans are developed for 328-ft (i.e., 100-m) sections of roadway and are then combined to provide recommended improvements for specific road sections, entire routes, and entire road networks. Star ratings indicate the extent to which geometric design and traffic control features known to have positive effects on safety are present on each 328-ft segment of a road network. Star ratings range from one-star to five-stars. A one-star road is generally the most basic two-lane, undivided roadway with few safety features. For example, a one-star road may have sharp curves, many driveways, no shoulders, and poor roadside design. A five-star road has all, or nearly all, of the geometric design and traffic control features known to reduce crashes. A conventional highway may have the features of a five-star road over short roadway sections, such as an individual 328-ft roadway segment, but extended sections of five-star roads are generally found only on the best-designed freeways. Two-lane roads are generally one-, two-, or three-star roads, depending on the geometric design and traffic control features they incorporate. Multilane, divided roads are generally three- or four-star roads, and freeways are generally four- or five-star roads. Star ratings are assigned using a scoring system developed based on available research on the safety effects of road design features from around the world. The scoring system includes factors for both crash likelihood and crash severity for specific crash types. Separate star ratings are assigned to vehicle-occupant, motorcycle, bicyclist, and pedestrian crashes, because the factors that influence crash frequency and severity for each of these user groups vary widely. The resulting star ratings are provided in the form of maps, tables, and downloadable spread- sheets and represent the results on the first step in the roadway safety management process (i.e., network screening). Safer roads investment plans are infrastructure improvement programs consisting of cost- effective infrastructure improvements for specific locations across an entire road network. In developing a safer roads investment program, the ViDA software considers over 70 specific countermeasures. The development of a safer roads investment plan for each 328-ft road segment proceeds as follows: • Crash frequency and severity for the road segment are estimated from the scoring system used to develop the star ratings and are calibrated based on the entire network crash history. • Each of the 70 specific countermeasures is evaluated for each road segment. If there appears to be an engineering need for the countermeasure and that countermeasure is not already present on the roadway segment, the countermeasure is identified for consideration in an economic analysis. • All countermeasures identified in the previous step are considered in a benefit-cost evalu- ation. The software calculates the expected frequencies of fatal and serious-injury crashes before implementation of any countermeasure, the reduction in fatal and serious-injury crashes that would result from implementation of the countermeasure, and the benefits and costs of the countermeasure in monetary terms. Countermeasures are placed into the final safer roads investment plan if all of the following are true: – The benefit-cost ratio for the countermeasure exceeds a minimum benefit-cost ratio specified by the user.

48 Guide for Quantitative Approaches to Systemic Safety Analysis – The countermeasure is compatible with other cost-effective countermeasures for the same location. – The countermeasure is not overridden by a mutually exclusive countermeasure for the same location that is more cost effective. – The countermeasure is consistent with countermeasures recommended for adjacent road segments. The resulting safer roads investment plan provides maps, tables, and downloadable spreadsheets showing specific countermeasures recommended for implementation at specific locations. Since ViDA is a planning-level tool, final projects are developed for implementa- tion only after a detailed engineering study is conducted separately and not part of the ViDA software. These engineering studies consider site-specific engineering, cost factors, and crash patterns and may change the specific countermeasure recommended at any given location. The safer roads investment plans, in essence, represent the combined output of Steps 2 (diagnosis), 3 (countermeasure selection), 4 (economic appraisal), and 5 (project prioritization) of the roadway safety management system. 3.3.2 Input Data Needs for ViDA The primary input data for ViDA is a spreadsheet file of more than 50 roadway characteristics for each 328-ft roadway segment on the road network. Table 8 lists the roadway characteris- tics included as input data. These include all data needed to develop star ratings and fatal and serious-injury estimates for each roadway segment and determine which countermeasures are present or not present on the roadway segment. The input data can be coded from review of aerial photos and street-level photos of the site using highway agency photologs or web-based tools such as Google Earth or Bing Streetside. Technicians or graduate students can be trained as data coders. On average, it takes about 30 minutes of labor per mile for a trained coder to prepare the input data for a roadway. Coding tools that can be used to prepare input data are available from usRAP (see contacts at www.usrap.org). Roadway data are generally coded in one direction of travel for undivided roads and separately for each direction of travel for divided roads. If a highway agency already has particular input variables available in an appropriate form in a database, these can be used as input to ViDA without the need to recode those particular variables. 3.3.3 Results Provided by ViDA The results provided by ViDA are described in this section. For star ratings, the analysis results include: • Tables that show the distribution of roads on a road network by star rating level (see example in Table 9). For example, for the sample road network consisting of 5,386.1 miles, 5% of the roads that could be analyzed have five-star ratings; 13% of the roads that could be analyzed have four-star ratings; 42% of the roads that could be analyzed have three-star ratings; 23% of the roads that could be analyzed have two-star ratings; and 17% of the roads that could be analyzed have one-star ratings, taking into consideration motor vehicle crashes. However, if motorcycle crashes are of primary interest, 3% of the roads that could be analyzed have five-star ratings; 5% of the roads that could be analyzed have four-star ratings; 40% of the roads that could be analyzed have three-star ratings; 30% of the roads that could be analyzed have two-star ratings; and 22% of the roads that could be analyzed have one-star ratings. Similarly, star ratings can be calculated taking into consideration pedestrian and bicycle crashes. By default, the tables cover the entire road network, but the software can also create tables for individual routes and individual road sections.

Overview of Systemic Safety Management Approaches 49 • Roadway type (divided/undivided) • Upgrade cost (extent of roadside development that would influence the cost of installing countermeasures) • Land use (separately for each side of the road) • Area type (rural/urban) • Speed limit • Truck speed limit (may be the same as or may differ from the general speed limit) • Traffic volume (AADT) • Median type • Centerline rumble strips • Shoulder rumble strips • Roadside severity (object type and distance to object, separately for each side of the road) • Paved shoulder width • Intersection type • Intersection channelization • Intersecting road volume (grouped into broad categories) • Intersection quality • Property access points (driveways) • Number of through traffic lanes • Lane widths for through traffic lanes • Curvature (categories based on the speed at which the curve may be traversed) • Quality of curve • Grade • Road condition • Road surface type • Delineation • Street lighting • Pedestrian crossing facility (major road) • Pedestrian crossing quality • Pedestrian crossing facility (side road) • Pedestrian fencing • Traffic calming • Vehicle parking • Presence of sidewalks (separately for each side of the road) • Presence of service road (frontage road) • Bicycle facility • Sight distance • Motorcycle percentage in traffic flow (grouped into broad categories) • Pedestrian peak hour flow across the road (grouped into broad categories, may be estimated) • Pedestrian peak hour flow along the road (separately for each side of the road, grouped into broad categories, may be estimated) • 85th percentile operating speed (may be measured or estimated) • Mean operating speed (may be measured or estimated) • School zone warning • School zone crossing guard Table 8. Input variables needed to create star ratings and safer roads investment plans in ViDA. Star Rating Vehicle Occupant Motorcyclist Pedestrian Bicyclist Length (mi) Percent Length (mi) Percent Length (mi) Percent Length (mi) Percent 5 Stars 259.6 5% 146.2 3% 126.2 2% 83.2 2% 4 Stars 681.1 13% 267.9 5% 174.6 3% 54.6 1% 3 Stars 2,253.4 42% 2,168.3 40% 248.5 5% 328.3 6% 2 Stars 1,252.5 23% 1,595.5 30% 202.1 4% 916.4 17% 1 Star 937.5 17% 1,206.2 22% 725.0 13% 773.3 14% Not Applicable 2.0 0% 2.0 0% 3,909.8 73% 3,230.2 60% Totals 5,386.1 100% 5,386.1 100% 5,386.1 100% 5,386.1 100% Table 9. Sample of usRAP ViDA star rating summary table for sample road network.

50 Guide for Quantitative Approaches to Systemic Safety Analysis • Maps that can be displayed on the computer screen that show all roads for the road network being analyzed, color coded by star ratings. By default, the maps cover the entire road network, but the software can also create maps for individual routes and individual road sections. • Downloadable spreadsheets that show the star rating scores and the star ratings for each 328-ft section of road. The spreadsheets also include the latitudes and longitudes for the beginning point of each 328-ft section. The spreadsheet can be imported directly into a geographic information system (GIS), so that the data can be managed and plotted on maps in a GIS environment. The results for safer roads investment plans include: • A table that summarizes the countermeasures in the safer roads investment plan by counter- measure type, including: – Countermeasure name, – Number of fatalities and serious injuries reduced, – Length of road or number of sites improved, – Present value of crash reduction benefits over the analysis period (typically 20 years), – Present value of countermeasure implementation cost over the analysis period (typically 20 years), – Cost-effectiveness (dollars spent per fatality or serious injury reduced), and – Benefit-cost ratio. The table summarizes the countermeasures in the safer roads investment plan for each 328-ft road segment. The benefit-cost ratio for each countermeasure within each 328-ft road segment must exceed the minimum benefit-cost ratio specified by the user. Table 10 shows a typical table summarizing a safer roads investment plan. By default, the table covers the entire road network, but the software can also create tables for individual routes and indi- vidual road sections. • Maps for any given countermeasure that can be displayed on the computer screen to show the locations where implementation of each countermeasure is recommended. Each 328-ft road segment where a given countermeasure is recommended is represented by a dot on the map. • Downloadable spreadsheets that show the recommended countermeasures by 328-ft road segments. The spreadsheets include all of the data listed above for the safer roads invest- ment plan summary table for individual countermeasures within individual 328-ft segments. The spreadsheets also show the latitude and longitude for the beginning point of each 328-ft segment. The countermeasure data in the spreadsheet can be imported directly into a GIS platform, so that the data can be managed and plotted on maps in a GIS environment. The spreadsheets also include information for every countermeasure that was considered for the safer roads investment plan but not recommended for implementation, including the reason the countermeasure was not recommended for implementation (e.g., benefit-cost ratio not high enough). This permits countermeasure alternatives to be readily considered. 3.3.4 Key Strengths and Advantages of ViDA Key strengths and advantages of the usRAP ViDA software are as follows: • The software is web-based and readily accessible. • Any public agency and its consultants may use the software free of charge. • Access to the software is password protected, so users can control access to their data. • The software is relatively easy to use. New users can acquire a basic familiarity with the software in one to two hours of training. Full understanding of the data needs and data acquisition for the software requires approximately 12 hours of training.

Overview of Systemic Safety Management Approaches 51 • Input data consists mostly of familiar design and traffic control parameters that can be coded from aerial and street-level photographic images, such as Google Earth or Bing Streetside and can be coded by trained technicians or students with an average of approximately 30 minutes of labor per mile of roadway. • Input data can be readily managed with commercially available spreadsheet software such as Microsoft Excel. • The user can customize analysis parameters including fatality and serious-injury costs, unit countermeasure costs, and minimum benefit-cost ratios to match their agency’s experience and practices. • The software can be calibrated for application to a particular roadway network if network- wide fatal and serious-injury crash history data are available. • The software processes data rapidly to develop safer roads investment plans. Most datasets can be uploaded to the software in a minute or two. Processing of data to develop a safer roads investment plan requires approximately 10 minutes for a typical county road system and approximately 20 minutes for state routes in a typical highway district. • All of the more than 70 countermeasures built into the software are considered for each 328-ft road segment, unless the user chooses to turn off consideration of any particular countermeasure. This reduces the possibility that any desirable countermeasure will be missed. Roadside barriers—right side Roadside barrier—left side Clear roadside hazard—left side Clear roadside hazard—right side Improve delineation Shoulder paving right side (> 3 ft) Shoulder paving left (> 3 ft) Bicycle lane (on-road) Improve curve delineation Lane widening (up to 1.5 ft) Delineation and signing (intersection) Sideslope improvement—right side Sideslope improvement—left side Shoulder rumble strips FSI = Fatal and serious injuries PV = Present value BCR = Benefit-cost ratio Table 10. Sample ViDA output of a safer roads investment plan summary table for a sample road network.

52 Guide for Quantitative Approaches to Systemic Safety Analysis • The results provide a program of cost-effective potential infrastructure improvements to reduce fatal and serious-injury crashes, with specific countermeasures recommended for implementation at specific locations. • The software includes a large amount of support information and transparency of the process. It is well documented, and the publications are easy to understand. 3.3.5 Key Weaknesses or Limitations of ViDA Key weaknesses or limitations of the usRAP ViDA software are as follows: • The software uses crash prediction models based on the best worldwide safety research, much of which are from countries other than the United States. Thus, the software does not use the crash prediction models most familiar to U.S. users, such as those in the HSM. However, the ViDA software provides results that are appropriate for U.S. application because the crash reduction effectiveness of specific countermeasures is likely to be similar among the United States and other developed countries, and the calibration process can incorporate local crash experience. • For many agencies (especially those that do not maintain freeways), most or all of their network will always rate as poor-to-fair simply because they are two-lane facilities. However, many low-volume and/or well-maintained roads may be perfectly adequate for their expected traffic volumes, even if the usRAP ViDA software indicates they are one- to three-star roadways. • The software currently displays results in metric units (i.e., distance in kilometers, speeds in kilometers per hour). The software has been partially customized to U.S. highway engineering terminology, and countermeasure descriptions can be displayed with U.S. customary units (i.e., lane and shoulder widths in feet). Further customization to U.S. customary units is expected in the future. 3.3.6 Getting Access to ViDA The usRAP ViDA software can be accessed at https://vida.irap.org. On the home page, anyone can register to use the software. Once registered, the user may request access to existing safer roads investment plan files relevant to their agency or, with the appropriate permissions in place, may create their own files for analysis. Access to specific files is controlled by usRAP on a request basis so that usRAP can assure each agency using the ViDA software that their files are private. There are three levels of access that can be requested from usRAP: • A user with Reader access can look at existing results in the software but cannot download or change any information. • A user with Analyst access can look at existing results and can download results files. • A user with Creator access can look not only at existing results and download results, but can also create new files and run analyses. 3.4 Summary Agencies generally implement systemic safety management using one of three approaches: • Application of FHWA’s Systemic Tool methodology, customized to local data availability and program goals. • Application of SPFs using in-house tools or Safety Analyst. • Application of the usRAP methodology, using the associated ViDA software. Table 11 summarizes the three systemic safety management implementation approaches, including the aspects of the safety management process for which they are used, the types of data required, the associated tools available, and their advantages and disadvantages.

Overview of Systemic Safety Management Approaches 53 Application of FHWA’s Systemic Tool Methodology Application of SPFs Application of usRAP Methodology Can be used for Identification of target crash type(s), facility type(s), and contributing factor(s); network screening and prioritization; countermeasure selection; project site selection. Network screening for site selection, evaluation of countermeasure impact on a site-by-site basis. Development of star ratings for network screening, development of safer roads investment plans. Data required Network crash data, inventory data related to crash types and crash contributing factors, effectiveness information on proven countermeasures and context for application. Site-specific crash data, inventory data and traffic volume data related to planning level SPFs, effectiveness information on proven countermeasures and context for application. Detailed roadway inventory data, traffic volume estimates, network crash data (if available). Tools available Guidance document, crash treediagram tool. In-house tools, Safety Analyst software. ViDA software. Advantages Highly customizable to agency needs and available data, identifies locations across the network where a given countermeasure may be appropriate. Based on reliable safety research and widely-accepted crash prediction methodology; Safety Analyst provides a systemic site selection module to identify sites where it is cost effective to implement a particular countermeasure. No crash data required, free software to develop safer road investment plans. Disadvantages Requires in-house data management and analysis; although data-driven, the approach may not be as statistically rigorous as the other approaches. In-house analysis and SPF development requires safety prediction modeling expertise; Safety Analyst requires purchasing an annual license, and it takes time to format an agency’s data for use in the software. Requires collection of roadway and roadside data from street- view images, which can be labor intensive; based on international research, which may not always reflect local conditions; is not focused on a specific crash type or countermeasure but instead provides suggested countermeasure packages across the network. Table 11. Overview of three systemic safety management implementation approaches.