Highway Safety Manual User Guide (2022)

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.

3 2 Highway Safety Manual Overview The HSM provides analytical tools and techniques for quantifying the potential effects on crashes because of decisions made in planning, design, operations, and maintenance. The information provided in the manual will assist agencies in their efforts to integrate safety into their decision-making processes. HSM users should have a safety knowledge base that includes familiarity with general highway safety principles, basic statistical procedures, and interpretation of results, along with suitable competence to exercise sound traffic safety and operational engineering judgment. The HSM can be used for the following actions: • Identify sites with the most potential for crash frequency or severity reduction • Identify factors contributing to crashes and associated potential mitigation measures • Conduct economic appraisals of safety countermeasures and project prioritization • Evaluate the crash reduction benefits of implemented treatments • Calculate the effect of various design alternatives on crash frequency and severity • Estimate potential crash frequency and severity on highway networks • Estimate the potential effect on crash frequency and severity of planning, design, operations, and policy decisions The HSM can be used to consider safety in planning, design, construction/implementation, operations, and maintenance activities. The project development process was developed to discuss the stages of a project from planning to post-construction operations and maintenance activities. The HSM is organized into four parts: HSM Part A – Introduction, Human Factors, and Fundamentals; HSM Part B – Roadway Safety Management Process; HSM Part C – Predictive Methods; and Part D – Crash Modification Factors. 2.1 HSM Part A: Introduction, Human Factors, and Fundamentals HSM Part A has three chapters: HSM Chapter 1 - Introduction and Overview, HSM Chapter 2 – Human Factors, and HSM Chapter 3 – Fundamentals. HSM Chapter 1 – Introduction and Overview describes the purpose and scope of the HSM, describes the basics of highway safety, and explains the relationship of the HSM to planning, design, operations, and maintenance activities. This chapter summarizes the different elements included in the manual, provides a general description of the purpose and scope of the HSM, and explains the relationship of the HSM to the project development process. HSM Chapter 2 – Human Factors describes the core elements of human factors that affect the interaction of drivers and roadways and provides an introduction to human factors to support the application of information presented in HSM Parts B, C, and D. Good understanding of this interaction allows highway agencies to plan and construct highways in a manner that minimizes human error and crashes. The NCHRP Report 600A: Human Factors Guidelines for Road Systems provides more detailed information and insights about driver’s characteristics allowing analysts to deliberately consider the road users’ capabilities and limitations in roadway design and operational decisions. HSM Chapter 3 – Fundamentals describes a variety of analysis approaches and methodologies as well as the background information needed to apply the predictive method, crash modification factors (CMFs), and evaluation methods provided in Parts B, C, and D of the HSM.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 4 2.2 HSM Part B: Roadway Safety Management Process HSM Part B discusses the process of monitoring and reducing crash frequency on existing roadway networks. The roadway safety management process consists of six steps: network screening (HSM Chapter 4), diagnosis (HSM Chapter 5), safety countermeasure selection (HSM Chapter 6), economic appraisal (HSM Chapter 7), project prioritization (HSM Chapter 8), and safety effectiveness evaluation (HSM Chapter 9). HSM Part B allows users to: • Identify and rank sites based on the potential for reducing average crash frequency • Identify crash patterns with crash data, historical site data, and field conditions • Identify the crash contributing factors at a site • Select possible appropriate safety countermeasures to reduce the average crash frequency • Evaluate the benefits and costs of the possible safety countermeasures • Identify individual projects that are cost-effective or economically justified • Identify improvement projects at specific sites and across multiple sites • Evaluate effectiveness of a safety countermeasure in reducing crash frequency or severity The roadway safety management process can be applied in different stages of the project development process, as shown in Table 1. TABLE 1 Application of HSM Part B on Different Stages of Project Development Process HSM Chapter Sy st em P la nn in g Pr oj ec t P la nn in g Pr el im in ar y D es ig n Fi na l D es ig n C on st ru ct io n/ Im pl em en ta tio n O pe ra tio n M ai nt en an ce Chapter 4 – Network Screening  Chapter 5 – Diagnosis   Chapter 6 – Select Countermeasures    Chapter 7 – Economic Appraisal    Chapter 8 – Prioritize Projects  Chapter 9 – Safety Effectiveness Evaluation  Key concepts discussed in HSM Part B include: • Performance measure is used to evaluate the potential to reduce crash frequency at a site. • A collision diagram is a two-dimensional plan view representation to simplify the visualization of crash patterns that have occurred at a site within a given time. • A countermeasure is a roadway strategy intended to decrease crash frequency or severity, or both, at a site.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 5 • The Haddon Matrix is used to identify crash contributing factors before, during, and after a crash from the perspective of human, vehicle, and roadway. • Regression-to-the-mean (RTM) or selection bias refers to the bias created by the natural fluctuation of crash frequencies, which may lead one to draw incorrect conclusions about countermeasure effectiveness or sites with potential for improvement. • The net present value (NPV) method is used to express the difference between discounted costs and discounted benefits of an individual improvement project in a single amount. The monetary costs and benefits are converted to a present value using a discount rate. • A benefit-cost ratio (BCR) is the ratio of the present value benefits of a project to the implementation costs of the project. The following sections summarize the theoretical framework together with some important concepts and procedures for applying HSM Part B in the roadway safety management process. Refer to the relevant chapters in the HSM for more detailed information about roadway safety management. 2.2.1 HSM Chapter 4: Network Screening HSM Chapter 4 provides a process for reviewing a transportation network to identify and rank sites based on the potential for reducing average crash frequency and/or crash severity. The network screening process is comprised of five steps: establish the focus of network screening, identify the network and reference population, select the performance measures, select screening method, and screen and evaluate the results. The intended purpose of network screening can be either to identify sites with potential to reduce the average crash frequency or severity or focus on reducing a particular crash type, severity, frequency, or contributing factor. The selected network elements can then be identified and organized into different reference populations based on the roadway site characteristics (such as intersections, roadway segments). HSM Part B Section 4.2.2 (HSM p. 4-3) lists some potential characteristics that can be used to establish reference populations for intersections and roadway segments. The third step in the network screening process is to select one or more performance measures to evaluate the potential for reducing the number of crashes or crash severity at a site. The performance measures can be selected based on data availability, RTM, or other statistical bias, and how the performance threshold is established (Figure 1). Figure 1 presents different performance measures in relative order of complexity, from least to most complex. For example, crash rate near the top of the list. Crash rate is often used because the data are readily available, but the results are not statistically stable. Excess Expected Average Crash Frequency with Empirical Bayes (EB) adjustments is more reliable but requires more data than for analysis based on crash rate. Each of the performance metrics are described in HSM Part B Section 4.2.3 (HSM p. 4-6) along with the strengths and limitations of different performance measures. Refer to HSM Part B Section 4.4.2 for more details on data needs and calculation procedures for intersection performance measures.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 6 Source: HSM, 1st Edition Figure 1: Stability of Performance Measures The selected performance measure can be applied to roadway segments, intersections, and facilities using different screening methods. Generally, roadway segments can be screened using either a sliding- window or peak-searching method, while intersections can be screened using only a simple ranking method. Facilities that are a combination of intersections and roadway segments can be screened with a combination of screening methods. Only those screening methods that are consistent with the performance measures can be selected. Users can refer to HSM Part B Table 4-3 (p. 4-19) to determine the consistent screening method for the selected performance measure. Finally, the performance measure and the screening method can be applied to one or more of the roadway segments, intersections, or facilities. A list of sites ordered according to the selected performance measure can be generated for the next step to identify locations for further review. 2.2.2 HSM Chapter 5: Diagnosis The second step of the roadway safety management process, known as diagnosis, is to identify the contributing factors to the crashes; crash patterns; crash types; weather; potential road or roadside, vehicle, or human factors that may be relevant for the sites under investigation. Diagnosis is completed by reviewing existing crash data, assessing supporting documentation about the site conditions, and conducting an onsite field review. It is recommended to use 3 to 5 years of crash data to evaluate crash locations, crash type, and crash severity to identify patterns. The crash data can be displayed using geographic information system (GIS) tools, linear graphs, bar charts, pie charts, or tabular summaries to better interpret and understand the data. Tools such as collision diagrams, condition diagrams, and crash mapping are described in HSM Part B Section 5.2.2 (HSM p. 5-4). In addition to the safety data review, supporting documentation of site geometrics, traffic operations, site conditions, and uses should be evaluated. Documented information and personal testimony from local transportation enforcement and emergency services professionals may be useful for identifying potential crash contributing factors or to verify information gained from earlier data evaluations and

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 7 analysis. HSM Part B Section 5.3 (HSM p. 5-8) lists examples of possible supporting documentation to be used during a site safety assessment, and HSM Appendix 5B (HSM p. 5-24) provides a list of questions and data to consider when reviewing past site documentation. A site review is helpful to understand the area and potential issues better. Information gathered onsite might include geometric and traffic control information, as well as observation of traffic. A comprehensive field assessment involves travel through the site from all possible directions and modes, visiting the site during different times of the day and under different lighting/weather conditions. HSM Appendix 5C provides guidance on how to prepare for assessing field conditions. HSM Appendix 5D provides examples of field review checklists for different types of roadway environments. After the field assessment, crash data review, and supporting documentation review are completed, the information can be compiled and used to identify trends or crash patterns. If trends or patterns are identified, safety countermeasures can be selected to mitigate or address the contributing factor(s) for crash occurrence. 2.2.3 HSM Chapter 6: Select Countermeasures The contributing factors to observed crash patterns or types need to be identified before selecting appropriate safety countermeasures to address them. Multiple factors may be contributing to each identified crash pattern or types of crashes. To minimize the probability that a major contributing factor is overlooked, a broad range of possible contributing factors should be identified. Engineering judgment and statistical assessment are commonly applied to identify those factors that are expected to be the greatest contributors to each particular crash type or type after considering a broad range of contributing factors. The Haddon Matrix (which divides the crash contributing factors into human, vehicle, and roadway categories) can be used to identify contributing factors for observed crash types or patterns. Potential contributing factors before, during, and after a crash are identified to determine the possible reasons of a crash. HSM Part B Section 6.2.2 (HSM p. 6-3) lists the most common contributing factors associated with a variety of crash types. Users can also refer to NCHRP Report 500: Guidance for Implementation of the AASHTO Strategic Highway Safety Plan for more details about the contributing factors for specific crash types. Each site and corresponding crash history are unique, and identification of crash contributing factors can only be completed by careful consideration of all the facts gathered during the diagnosis process. Appropriate safety countermeasures can be selected after contributing factors have been identified. Countermeasure selection is used to develop potential engineering, education, enforcement, or emergency response treatments to address the contributing factors under consideration. Only crash- based countermeasures are covered in this edition of the Highway Safety Manual User Guide. The FHWA CMF Clearinghouse contains a comprehensive list of CMFs (FHWA, 2013). Engineering judgment and local knowledge are required when comparing contributing factors to potential safety countermeasures. When selecting countermeasures, users should also consider why the contributing factor(s) might be occurring, what could address the factor(s), and what is physically, financially, and politically feasible in the jurisdiction. For each specific site, one countermeasure or a combination of countermeasures could be considered to address the contributing factor. Users can refer to HSM Part D for countermeasures with quantitative CMFs. In some cases, contributing factors may not be easily identifiable, even when there is a clear crash pattern. In such cases, a review of the road environment upstream or downstream of the site may provide some insights to whether there is any influence at the project location.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 8 2.2.4 HSM Chapter 7: Economic Appraisal The main objectives for the economic appraisal of a safety countermeasure or combination of countermeasures are to determine whether a project is economically justifiable, and determine which project or alternative is the most cost-effective. There are two methods for conducting economic appraisals, benefit-cost analysis, and cost-effectiveness analysis. Both methods quantify the benefits of the proposed countermeasure(s). For benefit-cost analysis, the change in crash frequency or severity is converted to monetary values and compared to the cost of implementing the safety countermeasure. Additional project benefits such as savings in travel time or fuel consumption are common considerations during project evaluation, but the HSM only considers changes in crash frequency or severity. Users can refer to the AASHTO publication, A Manual of User Benefit Analysis for Highway and Bus-Transit Improvements (AASHTO Redbook) for considering other project benefits. For cost- effectiveness analysis, the change in crash frequency is compared directly to the project cost and is not quantified as monetary value. This approach provides a method to understand the value of countermeasure(s) implementation when the agency does not support the monetary crash costs values used to convert benefits to dollar value. The HSM suggests that the change in average crash frequency caused by the application of a safety countermeasure should be estimated using the HSM Part C predictive method. The expected change in average fatal, injury, and property damage only (PDO) crash frequency can be converted to a monetary value using the societal crash costs. Users can apply the accepted state state/local societal crash cost by crash severity and collision type, if available. They can also refer to the FHWA report, Crash Cost Estimates by Maximum Police-Reported Injury Severity within Selected Crash Geometries for other relevant values. HSM Table 7-1 (HSM p. 7-5) provides societal crash cost estimates by crash severity. The annual monetary value can be further converted to a present value using a discount rate and the service life of the safety countermeasures. The project costs include the present value of right-of-way acquisition, construction, operation, and maintenance costs throughout the service life of the project. Users can refer to Chapter 6 of the AASHTO Redbook for additional guidance regarding the categories of costs and their proper treatments in an economic appraisal. The NPV or BCR can be used to determine if a project is economically justifiable, and the cost- effectiveness index can be used to determine which project or alternative is most cost-effective. Users can refer to HSM Section 7.6 (HSM p. 7-8) for step-by-step instructions for each of these methods. After the economic appraisal is completed, the safety countermeasures for a given site can be ranked in descending or ascending order by project costs, BCR, cost-effectiveness index, and so forth. 2.2.5 HSM Chapter 8: Prioritize Projects Project prioritization begins by reviewing potential projects for construction/implementation and sorts them based on the results of ranking and optimization processes. Project prioritization methods are primarily applicable to the development of optimal improvement programs for an entire roadway system or across multiple sites, but they can also be applied for alternative evaluation of a single site. Chapter 8 provides three prioritization methods: ranking by economic effectiveness measures, incremental benefit-cost analysis, and optimization methods. The first two provide a list of projects prioritized based on specific criterion (refer to HSM Chapter 8.2 for additional details). Optimization methods are used to prioritize projects, which are already determined to be economically justified. The prioritization is based on determining the most cost-effective project or set of projects that fit a given budget and other constraints. The HSM includes three specific optimization methods to be

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 9 used to prioritize safety projects including Linear Programming optimization, Integer Programming Optimization and Dynamic Programming Optimization. HSM Appendix 8A (HSM p. 8-13) provides more detailed information about these methods. Most recently, the Integer Programming Optimization has become the most widely used method for project optimization. All the project prioritization methods are directly applicable when crash reduction is the only consideration. However, typical highway projects involve many other factors that influence project selection and prioritization. The HSM provides a reference to a class of decision-making algorithms known as multi-objective resource allocation, which can be used to quantify the effect of multiple factors such as safety in terms of reduction of crashes, traffic operations in terms of vehicle hours of delay reduced, air quality benefits in terms of the emissions reduced, etc. Users can refer to HSM Table 8-1 (HSM p. 8-6) for selecting the appropriate project prioritization method. Computer software programs are available to prioritize projects or project alternatives efficiently and effectively. Results from these prioritization methods can be incorporated into the decision-making process. 2.2.6 HSM Chapter 9: Safety Effectiveness Evaluation Safety effectiveness evaluation is the final step of the roadway safety management process. It is the assessment of how crash frequency or severity has changed because of a specific treatment or safety countermeasure, or a set of treatments or projects, and how well funds have been invested in reducing crashes. When one treatment is applied to several similar sites, the safety effectiveness evaluation could also help estimate a CMF for the treatment. The safety effectiveness evaluation could be performed with the following objectives: • Evaluate a single project at a specific site to document the safety effectiveness of that specific project • Evaluate a group of similar projects to document the safety effectiveness of those projects • Evaluate a group of similar projects for the specific purpose of quantifying a CMF for a countermeasure • Assess the overall safety effectiveness of specific types of projects or countermeasures in comparison to their costs Safety effectiveness evaluations may use several different types of performance measures, such as a percentage reduction in crashes, a shift in the proportion of crashes by collision type or severity level, a CMF for a treatment, or a comparison of the crash reduction benefits achieved in relation to the cost of a project or treatment. The evaluation is more complex than simply comparing before and after crash data at treatment sites because consideration should also be given to what changes in crash frequency would have occurred at the evaluation sites between the periods before and after the treatment, even if the treatment had not been implemented. To consider these impacts, most evaluations use data for both treatment and non-treatment sites and for periods both before and after implementation of the treatments. Three basic study designs are used for safety effectiveness evaluation: observational before/after studies, observational cross-sectional studies, and experimental before/after studies. Selection of the appropriate study design for safety effectiveness evaluation depends on the nature of the treatment, the types of sites at which the treatment has been implemented, and the periods for which data are available for those sites. Refer to HSM Table 9-4 (HSM p. 9-6) for selecting the observational before after evaluation method. Detailed procedures for implementing different safety evaluation methods including

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 10 data needs and input, pre-evaluation activities, and computational procedures are provided in HSM Part B Section 9.4 (HSM p. 9-7). 2.3 HSM Part C: Predictive Method 2.3.1 Overview of the Predictive Method HSM Part C provides a predictive method for calculating the predicted and/or expected average crash frequency of a network, facility, or individual site and introduces the concept of safety performance functions (SPFs). These methods focus on the use of statistical models to address the inherent randomness in crashes. The chapters in HSM Part C provide the predictive method for roadway segments and intersections for the following facility types, as listed in Table 2. TABLE 2 HSM Part C Chapters HSM Chapter Undivided Roadway Segments Divided Roadway Segments Intersections Stop Control on Minor Leg(s) Signalized Three- Leg Four- Leg Three- Leg Four- Leg 10 – Predictive Method for Rural, Two- Lane, Two-Way Roads     11 – Predictive Method for Rural Multilane Highways      12 – Predictive Method for Urban and Suburban Arterials       Predictions of average crash frequency as a function of traffic volume and roadway characteristics can be used for making decisions relating to designing, planning, operating, and maintaining roadway networks. The approach is applicable for both safety-specific studies and as an element of a more traditional transportation study or environmental analysis. The predictive method has been outlined in 18 steps in a flowchart format and discussed in detail in HSM Part C, Section C.6 (HSM p. C-12). This method provides detailed guidance on dividing a facility into individual sites; selecting the period of analysis; obtaining geometric data and observed crash data; and applying the predictive models and EB adjustment method. Where a facility consists of several contiguous sites, or crash estimation is desired for a period of several years, some steps may be repeated. Depending on the roadway or roadside conditions proposed by an alternative, the use of the EB method may not be appropriate. The predictive method can be used to assess crashes for existing conditions, alternatives to existing conditions, or proposed new roadways. The predicted average crash frequency can be modeled with the geometric design, traffic control features, and traffic volumes of that site. When observed crash frequency is available, the expected average crash frequency could be determined with the EB method. Figure 2 lists common scenarios in which the HSM predictive method or EB method could be used to model the predicted or expected average crash frequency. There are situations when the expected average crash frequency cannot be computed, such as when crash data is not available or is considered unreliable; when a project on new alignment or new location is contemplated; and when a substantial change to a location or facility is being considered such that the observed crash data are irrelevant. An example of this is a two-lane rural road being reconstructed as a four-lane divided highway. A

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 11 detailed explanation of observed crash frequency, predicted average crash frequency, and expected average crash frequency is provided in Section 2.3.3 of this guide. Figure 3 describes the facility type definitions included in each HSM Part C chapter. Figure 2: Scenarios for HSM Predictive Method Application Figure 3: HSM Part C Chapters and Facility Types Scenarios for HSM Predictive Method Application • Existing traffic under past or future traffic volume • Alternative designs for an existing facility under past or future traffic volumes • Designs for a new facility under future (forecast) traffic volumes • Estimated effectiveness of countermeasures after a period of implementation • Estimated effectiveness of proposed countermeasures on an existing facility (prior to implementation) HSM Part C Chapters and Facility Site Types Part C Chapter Facility Types Chapter 10 - Predictive Method for Rural Two-Lane, Two-Way Roads • All rural highways with two lanes and two-way traffic operation. This includes two-lane highways with center two-way left-turn lanes (TWLTL) and sections with passing or climbing lanes. • Three- and four-leg intersections with minor-road stop control and four-leg signalized intersections. Chapter 11 - Predictive Method for Rural Multilane Highways • All rural multilane highways without full access control with four travel lanes, except for two-lane highways with side-by-side passing lanes. • Three- and four-leg intersections with minor-road stop control and four-leg signalized intersections. Chapter 12 - Predictive Method for Urban and Suburban Arterials • All arterials without full access control with two or four through lanes in urban and suburban areas. • Three- and four-leg intersections with minor-road stop control or traffic signal control.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 12 2.3.2 HSM Part C Relationship to HSM Parts A, B, and D HSM Part A – Introduction, Human Factors, and Fundamentals. This section presents background information to understand the methods provided in the HSM to analyze and evaluate crash frequencies. It also includes information related to SPFs and CMFs. Good understanding of the fundamentals of SPFs and CMFs is recommended before using HSM Part C. HSM Part B – Roadway Safety Management Process. Material presented in this section is used for monitoring, improving, and maintaining an existing roadway network. Applying methods from HSM Part B can help identifying sites that exhibit more crashes than what would be expected; diagnosing crash patterns at specific sites; selecting appropriate safety countermeasures to mitigate crashes; benefits and costs of potential alternatives; establishing projects prioritization; and assessing projects effectiveness after implementation. The predictive method in HSM Part C provides tools to estimate the predicted and/or expected average crash frequency for application in HSM Chapter 4, Network Screening, and HSM Chapter 7, Economic Appraisal. HSM Part D – CMFs. The CMFs in HSM Part D present information regarding the effects of various safety treatments that are used to quantify the change in average crash frequency and the statistical reliability of those countermeasures. Although some HSM Part D CMFs are included in HSM Part C for use with specific SPFs, only the CMFs included in HSM Part C are intended to be used with the models in HSM Part C. 2.3.3 Predicted versus Expected Crash Frequency The HSM predictive method can calculate both the predicted crash frequency and the expected crash frequency under different scenarios. The predicted average crash frequency of an individual site is the crash frequency calculated with the SPFs and CMFs based on the geometric design, traffic control features, and traffic volume of the site. This method will be used to estimate the crash frequency for a past or future year, or when the observed crash frequency is not available. The observed crash frequency refers to the historical crash data observed/reported at the site during the period of analysis. When the observed crash frequency is available, the expected crash frequency can be calculated. The expected crash frequency uses the EB method to combine the observed crash frequency with the predicted average crash frequency to produce a more statistically reliable measure. A weighted factor is applied to both estimates; this reflects the statistical reliability of the SPFs. The expected crash frequency is the long-term average crash frequency that would be expected from the specific site and is more statistically reliable compared with the predicted crash frequency. Figure 4 illustrates the observed, predicted, and expected average crash frequencies for a site.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 13 Figure 4: Illustration of Observed, Predicted, and Expected Crash Frequency Estimates 2.3.4 Safety Performance Functions SPFs are regression models for estimating the predicted average crash frequency of individual roadway segments or intersections. SPFs are developed through statistical regression techniques using historical crash data collected over several years at “base” sites with similar characteristics. The regression parameters are determined with the assumption that crash frequencies follow a negative binomial distribution, which is an extension of the Poisson distribution typically used for count data. The negative binomial regression allows the variance to differ from the mean through the incorporation of an additional parameter called the dispersion parameter. In cases where the variance is greater than the mean, the data is said to be over dispersed. The overdispersion parameter has positive values. This value is used to compute a weighted adjustment factor that is applied in the EB method described in HSM Section C.6.6. (HSM p. C-18) The dependent variable is the predicted average crash frequency for a facility type under base conditions. The independent variables are the segment length and average annual daily traffic (AADT) (for roadway segments) or the AADT on the major and minor roads (for intersections). Figure 5 shows a sample SPF developed for the Colorado Department of Transportation.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 14 Figure 5: Sample SPF – Colorado Department of Transportation (Source: Kononov, 2011) Multivariate models, or Level II SPFs, incorporate a variety of variables other than traffic volume only. Variables such as roadway geometry elements, access density, and weather can be used to estimate the dependent variable. The SPFs are developed for total crash frequency including all crash severity levels and, in some cases, collision types. However, SPFs for specific collision types and/or crash severity levels are also developed in some cases (see Table 3 for the list of SPFs included in HSM Part C). The user should select the appropriate SPFs when calculating the crash frequency for a specific site. TABLE 3 List of SPFs in HSM Part C Chapter Facility Type SPF for Collision Type SPF for Crash Severity Level Chapter 10 Roadway Segment • All collision types • All severity levels Intersection • All collision types • All severity levels Chapter 11 Roadway Segment • All collision types • All severity levels • Fatal-and-injury crashes Intersection • All collision types • All severity levels • Fatal-and-injury crashes Chapter 12 Roadway Segment • Single-vehicle crashes • All severity levels • Fatal-and-injury crashes • PDO crashes

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 15 TABLE 3 List of SPFs in HSM Part C Chapter Facility Type SPF for Collision Type SPF for Crash Severity Level • Multiple-vehicle nondriveway collision • All severity levels • Fatal-and-injury crashes • PDO crashes • Multiple-vehicle driveway-related collision • All severity levels • Vehicle-pedestrian collision • All severity levels • Vehicle-bicycle collision • All severity levels Intersection • Multiple-vehicle collision • All severity levels • Fatal-and-injury crashes • PDO crashes • Single-vehicle crashes • All severity levels • Fatal-and-injury crashes • PDO crashes • Vehicle-pedestrian collision • All severity levels • Vehicle-bicycle collision • All severity levels 2.3.5 Crash Modification Factors HSM Part C base models are developed using a given set of site characteristics and are used to estimate the predicted average crash frequency. The Part C CMFs are used to adjust the base models to local conditions. A CMF represents the relative change in estimated average crash frequency due to differences for each specific condition and provides an estimate of the effectiveness of the implementation of a particular countermeasure. For example, paving gravel shoulders, adding a left-turn lane, or increasing the radius of a horizontal curve. Part D includes all CMFs in the HSM. Some Part D CMFs are included in Part C for use with specific SPFs, since they are specific to the SPFs developed in those chapters. The remaining Part D CMFs can be used with the outcomes of the predictive method to estimate the change in crash frequency for a given countermeasure under the conditions described in HSM Section C.7 (HSM p. C-19). See also section 2.3.9 of this guide. All CMFs included in the HSM were selected through an expert panel review process and contain a combination of base conditions; setting and road type; AADT range in which the CMF is applicable; crash type and severity addressed by the CMF; CMF value; standard error; CMF source; and attributes of the original studies (if available). Part C CMFs have the same base conditions as their corresponding SPFs in Part C. 2.3.6 Weighting Using the Empirical Bayes Method The EB method can be used to calculate the expected average crash frequency for past and future periods and applied at either the site or the project level. Application at the project level is done when users do not have location-specific observed crash data for the individual roadway segments or intersections that are part of the project and when data is aggregated across all sites. The EB method combines the observed crash frequency with the predicted average crash frequency. This adjustment is only applied when observed crash data for a minimum of 2 years are available for either the specific site or the entire facility.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 16 The EB method uses a weighted factor (w) which is a function of the SPF’s overdispersion parameter (k) to combine the two estimates. As the value of the overdispersion parameter increases, the weighted adjustment factor decreases; thus, more emphasis is placed on the observed/reported crashes rather than the SPF predicted crash frequency. This estimate depends on the data characteristics (dispersed versus small overdispersion) used to develop the prediction models. Additional details can be found in HSM Part C, Appendix A.2 (HSM p. A-15) 2.3.7 Calibration versus Development of Local SPFs The predictive models in HSM Part C are composed of three basic elements: SPFs, CMFs, and a calibration factor. The HSM SPFs were developed using data from a subset of states. Difference in crash data quality, roadway inventory, traffic counts, crash reporting thresholds, and weather conditions are some of the factors that vary among states that may affect the prediction of the number and severity of crashes. Therefore, for the predictive method to provide results that are reliable for each jurisdiction that uses them, it is important that the SPFs in HSM Part C are calibrated to account for local conditions. Several DOTs have calibrated or are in the process of calibrating the HSM default SPFs. Some agencies are developing jurisdiction-specific SPFs using their own data to further enhance the reliability of the HSM Part C predictive method. The sophistication of state-specific SPFs may vary and require additional statistical analysis expertise. Calibration and SPF development are prepared by the agency rather than by individual users. During the calibration development period, HSM users can still use the HSM Part C to assess relative differences among alternatives within the same facility type and control type. However, the output from an HSM SPF cannot be used to describe an actual prediction, as it lacks the necessary calibration factor. 2.3.8 Crash Severity and Collision Type Distribution for Local Conditions Application of the HSM SPFs results in total predicted crash frequency or by specific severity. The HSM also provides distributions of crash frequency by severity and collision type. These tables may be used to separate the crash frequencies into different severity levels and collision types. These distributions can be used in cases where there is concern regarding certain collision types or crash severity levels. Users can refer to SPFs for specific injury levels or SPFs for total crashes combined with crash severity and type distribution to estimate specific injury levels. The crash severity and collision type distribution tables in the HSM were developed using specific state data. Agencies may provide jurisdiction-specific tables to be used instead of the HSM default tables. Application of agency-specific tables may provide predictions that are more accurate. 2.3.9 Methods for Estimating the Safety Effectiveness of a Proposed Project The following are the four HSM methods for estimating change in predicted average crash frequency for a project, listed in order of predictive reliability: • Method 1: Apply the HSM Part C predictive method to calculate the predicted average crash frequency of existing and proposed conditions. • Method 2: Apply the HSM predictive method to calculate the predicted average crash frequency of existing conditions, and application of appropriate HSM Part D CMFs to calculate the safety performance of the proposed condition.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 17 • Method 3: For cases where HSM Part C predictive method is not available, but an SPF for a facility not included in the HSM is available. Apply the SPF to calculate the predicted average crash frequency of existing conditions and apply an appropriate HSM Part D CMF to estimate the safety performance of the proposed condition. A locally derived project CMF can also be used as part of this method. • Method 4: Apply the observed crash frequency to calculate the expected average crash frequency of existing conditions and apply the appropriate HSM Part D CMF to the existing conditions expected average crash frequency to obtain the expected average crash frequency of the proposed condition. In all four methods, the delta between existing and proposed predicted average crash frequencies is used as the project effectiveness estimate. Depending on the project, if the observed crashes are also available, the expected crash frequency could be calculated using the predicted and observed crash frequencies and the project effectiveness estimate could be adjusted accordingly. 2.3.10 Limitations of the HSM Predictive Method The HSM predictive method has been developed using U.S. roadway data. The predictive models incorporate the effects of several geometric design elements and traffic control features. Variables not included in the predictive models were not necessarily excluded because they have no effect in crash frequency; it may merely mean that the effect is not fully known or has not been quantified at this time. In addition to the geometric features, the predictive method incorporates the effect of nongeometric factors in a general sense. One example of this limitation is the variation in driver populations. Different sites experience significant variations in demographics and behavioral factors including age distribution, years of driving experience, seatbelt usage, and alcohol usage. The calibration process accounts for the statewide influence of such crash factors on crash occurrence; however, these factors are not considered in site-specific variations, which may be substantial. The case is similar for the effect of weather, which might be incorporated through the calibration process. Another factor not included in the predictive method is the effect of traffic volume variations throughout the day or proportions of different vehicle types. This is mainly because these effects are not fully understood. Lastly, the predictive method treats the effects of individual geometric design and traffic control features as independent of one another and does not account for potential interactions between them. It is likely that such interactions exist, and, ideally, they should be accounted for in the predictive models. At present, such interactions are not fully understood and are difficult to quantify.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 18 2.3.11 HSM Part C Summary HSM Part C provides the basic methodology for calculating the predicted and/or expected crash frequency for selected highway facilities under given traffic and geometric conditions. The following concepts (Figure 6) were incorporated in the procedure: • Safety performance functions: SPFs are regression equations that are used to calculate the predicted crash frequency for a specific site (with specified base conditions) as a function of annual average daily traffic, and (in the case of roadway segments) the segment length. • Base condition: A specific set of geometric design and traffic control features, under which the SPFs were developed. • Crash modification factors: HSM Part C CMFs are used to account for the safety effects of differences between the base conditions and the site conditions of the highway facilities under investigation. • Local calibration factor: It is used to account for the differences between jurisdictions for which the SPFs were developed. Differences could be associated to factors such as driver population, climate, weather, and/or crash reporting thresholds. • Empirical Bayes Method: The EB method is used to combine the predicted average crash frequency with the observed crash frequency to obtain the expected average crash frequency for the selected highway facilities. • Crash severity and collision type distributions: These distributions are applied in the predictive method to determine the crash frequency under specific crash severity and collision types. The crash severity and collision type distribution tables were derived from HSM-related research projects. Some of these distributions can be replaced with locally derived values. Predictive Method Concepts Predictive method incorporates the following concepts: • SPFs • Base condition • CMFs • Local calibration factor • EB method • Crash severity and collision type distributions Figure 6: Predictive Method Main Concepts

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 19 2.3.12 HSM Chapter 10: Predictive Method for Rural Two-Lane, Two-Way Roads HSM Chapter 10 provides a methodology to estimate the predicted and/or expected average crash frequency, crash severity, and collision types for rural two-lane, two-way facilities. Crashes involving vehicles of all types, bicycles, and pedestrians are included, except for crashes between bicycles and pedestrians. The predictive method can be applied to existing sites, design alternatives to existing sites, or new sites. This chapter is applicable to all rural highways with two-lane and two- way traffic operation that do not have access control and are outside of cities or towns with a population greater than 5,000 people (HSM Section 10.3, p. 10-2). Additionally, it can be used on two-lane, two-way highways with center two-way left-turn lanes (TWLTLs); and with two-lane highways with passing lanes, climbing lanes, or short segments of four-lane cross sections —up to 2 miles in length—where additional lanes are provided to enhance passing opportunities. Longer sections can be addressed with the rural multilane highway procedures outlined in HSM Chapter 11. Figure 7 shows a typical example of a rural two-lane, two-way roadway. This chapter also addresses three- and four-leg intersections with minor-road stop control and four-leg signalization on all the roadway cross sections. Table 4 includes the site types on rural two-lane, two- way roads for which SPFs have been developed for predicting average crash frequency, severity, and collision type. Figure 8 lists the facility types and definitions provided in HSM Chapter 10. TABLE 4 Roadway Segment and Intersection Types and Descriptions for Rural Two-Lane, Two-Way Roads Facility Type Site Types with SPFs in Chapter 10 Roadway Segments Undivided rural two-lane, two-way roadway segments (2U) Intersections Unsignalized three-leg (stop control on minor-road approaches) (3ST) Unsignalized four-leg (stop control on minor-road approaches) (4ST) Signalized four-leg (4SG) Figure 7: Rural Two-Lane, Two-Way Road

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 20 HSM Chapter 10 also provides guidance on how to define roadway segments and intersections (HSM Section 10.5, p. 10-11). A roadway segment is defined as a section of continuous traveled way that provides two-way operation of traffic uninterrupted by an intersection and comprises homogeneous geometric and traffic control features. A segment begins and ends at the center of bounding intersections or where there is a change in homogeneous roadway characteristics. When a roadway segment begins or ends at an intersection, the length of the roadway segment is measured from the center of the intersection. An intersection is defined as the junction of two or more roadway segments. The intersection models estimate the average crash frequency that occurs at the intersection (Region A in Figure 9), and intersection-related crashes that occur on the intersection legs (Region B in Figure 9). Figure 9: Rural Two-Lane, Two-Way Roads – Definition of Roadway Segments and Intersections Rural Two-Lane, Two-Way Roads Facility Type Definitions Facility Type Definition Undivided roadway segment A roadway consisting of two lanes with a continuous cross-section providing two directions of travel in which the lanes are not physically separately by either distance or a barrier. Additionally, segments with a TWLTL or passing lanes are included as part of this definition. Unsignalized three- leg intersection with stop control An intersection of a rural two-lane, two-way road and a minor road. A stop sign is provided on the minor road approach to the intersection only. Unsignalized four- leg intersection with stop control An intersection of a rural two-lane, two-way road, and two minor roads. A stop sign is provided on both minor road approaches to the intersection. Signalized four-leg intersection An intersection of a rural two-lane, two-way road and two other rural two-lane, two-way roads. Signalized control is provided at the intersection by traffic lights. Figure 8: Rural Two-Lane, Two-Way Roads Facility Types and Definitions

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 21 2.3.13 Calculating the Crash Frequency for Rural Two-Lane, Two-Way Roads HSM Chapter 10 provides the methodology for calculating the predicted and/or expected crash frequency for roadway segments and intersections on rural two-lane, two-way roads. The calculation is for a given period during which the geometric design and traffic control features are unchanged and traffic volumes are known. The entire process could be divided into the following steps: 1. Predicted crash frequency under base conditions 2. Predicted crash frequency under site conditions 3. Expected crash frequency with EB method 4. Crash frequency under different collision types and crash severity levels Step 1: Predicted Crash Frequency under Base Conditions The predicted average crash frequency for the roadway segments and intersections under base condition could be determined by replacing the AADT and segment length (for roadway segments) or the AADTs for major and minor roads (for intersections) in SPFs with site-specific values. Table 5 lists the SPFs for different facility types included in HSM Chapter 10 and the applicable AADT ranges for the SPFs. Only application to sites within the AADT ranges could provide reliable results. TABLE 5 Rural Two-Lane, Two-Way Roads SPFs in HSM Chapter 10 Facility Type HSM Equation AADT Range Rural two-lane, two-way roadway segments Equation 10-6 0 to 17,800 vpd Three-leg stop-controlled intersection Equation 10-8 AADTmajor: 0 to 19,500 vpd AADTminor: 0 to 4,300 vpd Four-leg stop-controlled intersection Equation 10-9 AADTmajor: 0 to 14,700 vpd AADTminor: 0 to 3,500 vpd Four-leg signalized intersection Equation 10-10 AADTmajor: 0 to 25,200 vpd AADTminor: 0 to 12,500 vpd Notes: AADTmajor = average annual daily traffic on the major route AADTminor = average annual daily traffic on the minor route vpd = vehicles per day Step 2: Predicted Crash Frequency under Real Conditions Each SPF listed in Table 5 is used to estimate the predicted crash frequency of a roadway segment or intersection under base conditions, which is later adjusted to site-specific conditions. The base conditions are a specific set of geometric design and traffic control features under which the SPFs were developed and are not necessarily the same for all facilities. The base conditions for roadway segments and intersections on rural two-lane, two-way roads are listed in Figure 10.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 22 Figure 10: Rural Two-Lane, Two-Way Roads Base Conditions CMFs are applied to account for the differences between the specific site under investigation and the base condition for the facility type. CMFs are used to adjust the SPF estimate of predicted average crash frequency for the effect of individual geometric design and traffic control features. The CMF for the SPF base condition of each geometric design and traffic control feature has a value of 1.00. CMF values less than 1.00 indicate the treatments reduce the predicted average crash frequency in comparison to the base condition. Similarly, CMF values greater than 1.00 indicate the treatments increase the predicted crash frequency. The CMFs presented in HSM Chapter 10 and the specific site types to which they apply are listed in Table 6. TABLE 6 CMFs for Rural Two-Lane Highway Segments and Intersections Facility Type CMF CMF Description HSM CMF Equations and Tables Roadway Segments CMF1r Lane width Definition (HSM p. 10-23 to 10-25) Table 10-8 (HSM p. 10-24) Equation 10.11 (HSM p. 10-24) CMF2r Shoulder width and type Definition (HSM p. 10-25 to 10-27) Table 10-9 (HSM p. 10-26) HSM Equation 10-12 (HSM p. 10-27) CMF3r Horizontal curves: length, radius, and spiral transitions Definition (HSM p. 10-27) HSM Equation 10-13 (HSM p. 10-27) CMF4r Horizontal curves: superelevation Definition (HSM p. 10-28) HSM Equations 10-14, 10-15, and 10-16 (HSM p. 10-28) CMF5r Grades Definition (HSM p. 10-28) Rural Two-Lane, Two-Way Roads Base Conditions Roadway Segments Intersections • Lane width: 12 feet • Shoulder width: 6 feet • Shoulder type: Paved • Roadside hazard rating: 3 • Driveway density: 5 driveways per mile • No horizontal curvature • No vertical curvature • No centerline rumble strips • No passing lanes • No two-way left-turn lanes • No lighting • No automated speed enforcement • Grade level: 0% • Intersection skew angle: 0 degrees • No intersection left-turn lanes on approaches without stop control • No intersection right-turn lanes on approaches without stop control • No lighting

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 23 TABLE 6 CMFs for Rural Two-Lane Highway Segments and Intersections Facility Type CMF CMF Description HSM CMF Equations and Tables HSM Table 10-11 (HSM p. 10-28) CMF6r Driveway density Definition (HSM p. 10-28 to 10-29) HSM Equation 10-17 (HSM p. 10-28) CMF7r Centerline rumble strips Definition (HSM p. 10-29) CMF8r Passing lanes Definition (HSM p. 10-29) CMF9r Two-way left-turn lanes Definition (HSM p. 10-29 to 10-30) HSM Equations 10-18 and 10-19 (HSM p. 10-30) CMF10r Roadside design Definition (HSM p. 10-30) HSM Appendix 13A (HSM p. 13-59 to 13-63) HSM Equation 10-20 (HSM p. 10-30) CMF11r Lighting Definition (HSM p. 10-30) HSM Equation 10-21 (HSM p. 10-31) HSM Table 10-12 (HSM p. 10-31) CMF12r Automated speed enforcement Definition (HSM p. 10-31) Intersections CMF1i Intersection skew angle HSM Equation 10-22 (HSM p. 10-31) HSM Equation 10-23 (HSM p. 10-32) CMF2i Intersection left-turn lanes HSM Table 10-13 (HSM p. 10-32) CMF3i Intersection right-turn lanes HSM Table 10-14 (HSM p. 10-33) CMF4i Lighting HSM Equation 10-24 (HSM p. 10-33) HSM Table 10-15 (HSM p. 10-33) The SPFs were developed in HSM-related research from the most complete and consistent available data sets. However, the predicted crash frequencies may vary substantially from one jurisdiction to another for a variety of reasons. Calibration factors provide a method for incorporating local data to improve the estimated crash frequencies for individual locations. The local calibration factor accounts for the differences between the jurisdiction under investigation and the jurisdictions that were used to develop the default HSM SPFs. The local calibration factor is calculated using local crash data and other roadway characteristic data. The process for determining calibration factors for the predictive models is described in HSM Part C, Appendix A.1 (HSM p. A-1). The predicted crash frequency under real conditions can be calculated using Equation 1: 𝑁 = 𝑁 × 𝐶 × (𝐶𝑀𝐹 × 𝐶𝑀𝐹 × 𝐶𝑀𝐹 × … × 𝐶𝑀𝐹 ) (Eq. 1) where: 𝑁 = predicted average crash frequency for a specific year for site type x 𝑁 = predicted average crash frequency determined for base conditions of the SPF developed for site type x

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 24 𝐶𝑀𝐹 = CMFs specific to site type x and specified geometric design and traffic control features y 𝐶 = calibration factor to adjust SPF for local conditions for site type x Step 3: Expected Crash Frequency with Empirical Bayes Method This step can be omitted if no recorded crash data for the specific site under investigation were available or the data are considered unreliable. When historical crash data are available, the EB method (either site-specific or project-level) can be used to combine the HSM Chapter 10 predicted average crash frequency with the observed crash frequency. The expected average crash frequency is a more statistically reliable estimate. The expected average crash frequency can be determined using Equation 2: 𝑁 =𝑤 × 𝑁 + (1 − 𝑤) × 𝑁 (Eq. 2) where: 𝑤 = the weighted adjustment to be placed on the predictive model estimate. This value can be calculated using Equation 3: 𝑤 = ×∑ (Eq. 3) where: 𝑘 = the overdispersion parameter of the associated SPF used to estimate 𝑁 . Table 7 lists the 𝑘 values for SPFs of different facility types. TABLE 7 Overdispersion Parameters for SPFs in HSM Chapter 10 Facility Type Overdispersion Parameter (k) Rural two-lane, two-way roadway segments 0.236 per length of the roadway segment Three-leg stop-controlled intersection 0.54 Four-leg stop-controlled intersection 0.24 Four-leg signalized intersection 0.11 Step 4: Crash Frequency under Different Collision Types and Crash Severity Levels HSM Chapter 10 provides the crash severity and collision type distribution table for all the facility types included, as listed in Table 8. The crash frequency under different severity levels and collision types could be determined based on the distribution table after the predicted or expected crash frequencies were determined. These proportions can be updated based on local data for a particular jurisdiction as part of the calibration process. TABLE 8 Crash Severity and Collision Type Distribution Table for Different Facility Types Facility Type Crash Severity Distribution Collision Type Distribution Rural two-lane, two-way roadway segments HSM Table 10-3 HSM Table 10-4 Three-leg stop-controlled intersection HSM Table 10-5 HSM Table 10-6 Four-leg stop-controlled intersection HSM Table 10-5 HSM Table 10-6 Four-leg signalized intersection HSM Table 10-5 HSM Table 10-6

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 25 Figure 11 shows the HSM Chapter 10 predictive method flowchart for calculating the predicted and expected average crash frequency for rural two-lane, two-way roads. Figure 11: Flowchart for Calculating Expected Crash Frequency on Rural Two-Lane, Two-Way Roads 2.3.14 Data Requirements for Rural Two-Lane, Two-Way Roads For the study period, it is important to determine the availability of AADT volumes and, for an existing roadway, the availability of observed/reported crash data to determine whether the EB method is applicable. A good understanding of the SPFs’ base conditions will help determine relevant data needs and avoid unnecessary data collection. The base conditions for rural two-lane, two-way roads are defined in Section 2.3.12, and in HSM Section 10.6.1 (HSM p. 10-14) for roadway segments and HSM Section 10.6.2 (HSM p. 10-17) for intersections. General data for intersections and segments can be collected from different sources. Examples of data sources include commercial aerial maps, design plans, and states’ roadway inventory system. Data needed for this example are summarized in the following sections. Intersection Data Generally, the effect of major and minor road traffic volumes (AADT) on crash frequency is incorporated through an SPF, while the effects of geometric design and traffic controls are incorporated through the CMFs. Data required to apply the predictive method for intersections are listed in Table 9.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 26 TABLE 9 Intersection Data Requirements for Rural Two-Lane, Two-Way Roads Intersections Units/Description Intersection type Unsignalized three-leg (3ST), unsignalized four-leg (4ST), and signalized four-leg (4SG) Traffic flow major road AADTmajor (vpd) Traffic flow minor road AADTminor (vpd) Intersection skew angle degrees Number of signalized or uncontrolled approaches with a left-turn lane From 0 to 4 Number of signalized or uncontrolled approaches with a right-turn lane From 0 to 4 Intersection lighting Present or not present Calibration factor (Ci) Derived from calibration process Observed crash data Applicable only with the EB method; crashes that occur at the intersection or on an intersection leg, and are related to the presence of an intersection during the period of study Note: Ci = intersection calibration factor vpd = vehicles per day Roadway Segment Data The effect of traffic volume in crash frequency is incorporated through an SPF, while the effects of geometric design and traffic control features are incorporated through the CMFs. There is no minimum roadway segment length when applying the predictive method. However, when dividing the facility into small homogeneous sections, it is recommended to keep the minimum roadway segment length as 0.10 mile to minimize the calculation efforts and avoid modifying the results. Table 10 includes data requirements for roadway segment locations. TABLE 10 Roadway Segment Data Requirements for Rural Two-Lane, Two-Way Roads Roadway Segments Units/Description Segment length miles Traffic volume AADT (vpd) Lane width feet Shoulder width feet Shoulder type Paved, gravel, composite, or turf Length of horizontal curve miles Radius of curvature feet Spiral transition curve Present or not present Superelevation variance feet/feet Grade percent (%) Driveway density Driveways per mile Centerline rumble strips Present or not present

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 27 TABLE 10 Roadway Segment Data Requirements for Rural Two-Lane, Two-Way Roads Roadway Segments Units/Description Passing lanes Present (1 lane), present (2 lanes), or not present Two-way left-turn lane Present/not present Roadside hazard rating Scale: 1 to 7 (1 = the safest, 7 = the most dangerous) Segment lighting Present or not present Auto speed enforcement Present or not present Calibration factor (Cr) Derived from calibration process Observed crash data Applicable only with the EB method; crashes that occur between intersections and are not related to the presence of an intersection during the period of study Note: vpd = vehicles per day More information about the roadside hazard rating can be found in HSM Part D, Appendix 13A (p. 13-59). 2.3.15 HSM Chapter 11: Predictive Method for Rural Multilane Highways HSM Chapter 11 provides a method to estimate the predicted and/or expected average crash frequency, crash severity, and collision types for rural multilane highway facilities. Crashes involving vehicles of all types, bicycles, and pedestrians are included, except for crashes between bicycles and pedestrians. The predictive method can be applied to existing sites, design alternatives to existing sites, new sites, or for alternative traffic volume projections. Estimates of crash frequency can be made for a period that occurred in the past or will occur in the future. This chapter is applicable to all rural multilane highways without full access control that are outside urban areas that have a population less than 5,000 persons. This comprises all rural nonfreeways with four through travel lanes, except for two-lane highways with side-by-side passing lanes. Moreover, this chapter addresses three- and four- leg intersections with minor-road stop and four-leg signalized intersections on all the roadway cross sections. Figure 12 shows typical examples of undivided and divided rural multilane highways. Table 11 includes the different site types for which SPFs have been developed for estimating expected average crash frequency, severity, and collision type. Figure 13 lists the facility types and definitions provided in HSM Chapter 11. Figure 12: Rural Multilane Highways

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 28 TABLE 11 Roadway Segment and Intersection Types and Descriptions for Rural Two-Lane, Two-Way Roads Facility Type Site Types with SPFs in HSM Chapter 11 Roadway Segments Rural four-lane undivided segments (4U) Rural four-lane divided segments (4D) Intersections Unsignalized three-leg (stop control on minor-road approaches) (3ST) Unsignalized four-leg (stop control on minor-road approaches) (4ST) Signalized four-leg (4SG) a Note: a The four-leg signalized intersection models do not have base conditions; therefore, these models can be used only for generalized predictions of crash frequency. Figure 13: Multilane Rural Roads Facility Types and Definitions Multilane Rural Roads Facility Type Definitions Facility Type Definition Undivided four-lane roadway segment (4U) A roadway segment consisting of four lanes with a continuous cross- section that provides two directions of travel in which lanes are not physically separated by either distance or a barrier. Multilane roadways where opposing lanes are separated by a flush/ nontraversable median or similar means are considered undivided facilities. However, HSM Chapter 11 predictive methods do not address multilane highways with flush separators. Divided four-lane roadway segment (4D) Divided highways are nonfreeway facilities (such as facilities without full control access) that have lanes in two directions of travel separated by a raised, depressed, or flush median that is not designed to be traversed by a vehicle; this may include raised or depressed medians with or without physical median barrier, or flush medians with physical median barriers. Three-leg intersections with stop control (3ST) An intersection of a rural multilane highway (such as four-lane divided or undivided roadway) and a minor road. A STOP sign is provided on the minor-road approach to the intersection only. Four-leg intersection with stop control (4ST) An intersection of a rural multilane highway (such as four-lane divided or undivided roadway) and two minor roads. A STOP sign is provided on both minor-road approaches to the intersection. Four-leg signalized intersection (4SG) An intersection of a rural multilane highway (such as four-lane divided or undivided roadway) and two other rural roads, which may be two- lane or four-lane rural highways. Signalized control is provided at the intersection by traffic lights.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 29 To apply the predictive method, the roadway within the defined study area limits must be divided into homogenous individual sites, segments and intersections. Roadway segment boundaries begin at the center of an intersection and end at either the center of the next intersection, or where there is a change in the segment’s cross-section (homogeneous segment). The length of the roadway segment is measured from the center of the intersection. An intersection is defined as the junction of two or more roadway segments. The intersection predictive models estimate the predicted average crash frequency of crashes within the intersection limits (Region A in Figure 14) and intersection-related crashes that occur on the intersection legs (Region B in Figure 14). Figure 14: Rural Multilane Highways – Definition of Roadway Segments and Intersections 2.3.16 Calculating the Crash Frequency for Rural Multilane Highways HSM Chapter 11 provides the methodology for calculating the predicted and/or expected crash frequency for roadway segments and intersections on rural multilane highways. The calculation is for a given period during which the geometric design and traffic control features are unchanged and traffic volumes are known. The whole process could be divided into the following steps: 1. Predicted crash frequency under base conditions 2. Predicted crash frequency under site conditions 3. Expected crash frequency with EB method 4. Crash frequency under different collision types and crash severity levels Step 1: Predicted Crash Frequency under Base Conditions The predicted average crash frequency for the roadway segments and intersections under the base condition may be determined by replacing the AADT and segment length (for roadway segments) or the AADTs for major and minor roads (for intersections) in SPFs with site-specific values. Table 12 lists the SPFs for different facility types included in HSM Chapter 11 and the applicable AADT ranges for the SPFs. Only application to sites within the AADT ranges is likely to provide reliable results. NOTE: SPFs for 4SG on rural multilane highways have no specific base conditions and, therefore, can only be applied for generalized predictions. No CMFs are provided for 4SG intersections, and predictions of average crash frequencies cannot be made for intersections with specific geometric design and traffic control features.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 30 TABLE 12 Rural Multilane Highways SPFs in HSM Chapter 11 Facility Type Equation in HSM AADT Range Rural four-lane undivided segments (4U) HSM Equation 11-7 Up to 33,200 vpd Rural four-lane divided segments (4D) HSM Equation 11-9 Up to 89,300 vpd Unsignalized three-leg (stop control on minor- road approaches) (3ST) HSM Equation 11-11 AADTmajor 0 to 78,300 vpd AADTminor 0 to 23,000 vpd Unsignalized three-leg (stop control on minor- road approaches) (4ST) HSM Equation 11-11 AADTmajor 0 to 78,300 vpd AADTminor 0 to 7,400 vpd Signalized four-leg (4SG) HSM Equations 11-11 and 11-12 AADTmajor 0 to 43,500 vpd AADTminor 0 to 18,500 vpd Notes: AADTmajor = average annual daily traffic on the major route AADTminor = average annual daily traffic on the minor route vpd = vehicles per day Highway agencies may wish to develop their own jurisdiction-specific SPFs derived from local conditions and crash experience. These SPFs may be substituted for models presented in HSM Chapter 11. The HSM provides criteria for development of SPFs and is presented in HSM Part C, Appendix A.1.2 (HSM p. A-9). Step 2: Predicted Crash Frequency under Site Conditions The crash frequency calculated using the SPFs shown in the previous section is the predicted crash frequency for the roadway segments or intersections under base conditions. Base conditions are the prevalent conditions under which the SPFs were developed and are not necessarily the same for all facilities. The base conditions for roadway segments and intersections on rural multilane highways are listed in Figure 15. Figure 15: Rural Multilane Highway Base Conditions Rural Multilane Highways Base Conditions Undivided Roadways Divided Roadways Intersections • Lane width: 12 feet • Shoulder width: 6 feet • Shoulder type: Paved • Sideslopes: 1:7 (vertical: horizontal) or flatter • No lighting • No automated speed enforcement • Lane width: 12 feet • Right shoulder width: 8 feet • Median width: 30 feet • No lighting • No automated speed enforcement • Intersection skew angle: 0 degrees • No intersection left-turn lanes except on stop-controlled approaches • No intersection right-turn lanes except on stop-controlled approaches • No lighting

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 31 CMFs are applied to account for the differences between the specific site under investigation and the base condition for the facility type. CMFs are used to adjust the SPF estimate of predicted average crash frequency for the effect of individual geometric design and traffic control features. The CMF for the SPF base condition of each geometric design and traffic control feature has a value of 1.00. CMF values less than 1.00 indicate the treatments reduce the predicted average crash frequency in comparison to the base condition. Similarly, CMF values greater than 1.00 indicate the treatments increase the predicted crash frequency. The CMFs presented in HSM Chapter 11 and the specific site types to which they apply are listed in Table 13. TABLE 13 CMFs for Rural Multilane Highway Segments and Intersections Facility Type CMF CMF Description HSM CMF Equations and Tables Undivided Roadway Segments CMF1u Lane width on undivided segments Definition (HSM p. 11-26 to 11-27) HSM Table 11-11 (HSM p. 11-26) HSM Equation 11-13 (HSM p. 11-26) CMF2u Shoulder width and type Definition (HSM p. 11-27 to 11-28) HSM Tables 11-12 and 11-13 (HSM p. 11-27) HSM Equation 11-14 (HSM p. 11-27) CMF3ru Side slopes Definition (HSM p. 11-28) HSM Table 11-14 (HSM p. 11-28) CMF4ru Lighting Definition (HSM p. 11-28 to 11-29) HSM Equation 11-15 (HSM p. 11-28) HSM Table 11-15 (HSM p. 11-29) CMF5ru Automated speed enforcement Definition (HSM p. 11-29) See text (HSM p. 11-29) Divided Roadway Segments CMF1d Lane width on undivided segments Definition (HSM p. 11-29 to 11-30) HSM Table 11-16 (HSM p. 11-30) HSM Equation 11-16 (HSM p. 11-29) CMF2d Right shoulder width on divided roadway segment Definition (HSM p. 11-30 to 11-31) HSM Table 11-17 (HSM p. 11-31) CMF3rd Median width Definition (HSM p. 11-31) HSM Table 11-18 (HSM p. 11-31) CMF4rd Lighting Definition (HSM p. 11-31 to 11-32) HSM Equation 11-17 (HSM p. 11-31) HSM Table 11-19 (HSM p. 11-32) CMF5rd Automated speed enforcement Definition (HSM p. 11-32) See text (HSM p. 11-32)

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 32 TABLE 13 CMFs for Rural Multilane Highway Segments and Intersections Facility Type CMF CMF Description HSM CMF Equations and Tables Three- and Four- Leg Stop- Controlled Intersections CMF1i Intersection angle (3ST and 4ST) Definition (HSM p. 11-33 to 11-34) 3ST: HSM Equations 11-18 and 11-19 (HSM p. 11-33) 4ST: HSM Equations 11-18 and 11-19 (HSM p. 11-33) CMF2i Left-turn lane on major road Definition (HSM p. 11-34) HSM Table 11-22 (HSM p. 11-34) CMF3i Intersection right-turn lanes Definition (HSM p. 11-34 to 11-35) HSM Table 11-23 (HSM p. 11-35) CMF4i Lighting Definition (HSM p. 11-35) HSM Equation 11-22 (HSM p. 11-35) HSM Table 11-24 (HSM p. 11-35) The SPFs were developed in HSM-related research from the most complete and consistent available data sets. However, the predicted crash frequencies may vary substantially from one jurisdiction to another for a variety of reasons. Calibration factors provide a method for incorporating local data to improve the estimated crash frequencies for individual locations. The local calibration factor accounts for the differences between the jurisdiction under investigation and the jurisdictions that were used to develop the default HSM SPFs. The local calibration factor is calculated using local crash data and other roadway characteristic data. The process for determining calibration factors for the predictive models is described in HSM Part C, Appendix A.1 (HSM p. A-1). The predictive crash frequency under real conditions can be calculated using Equation 4: 𝑁 = 𝑁 × 𝐶 × (𝐶𝑀𝐹 × 𝐶𝑀𝐹 × 𝐶𝑀𝐹 × … × 𝐶𝑀𝐹 ) (Eq. 4) where: 𝑁 = predicted average crash frequency for a specific year for site type x 𝑁 = predicted average crash frequency determined for base conditions of the SPF developed for site type x 𝐶𝑀𝐹 = crash modification factors specific to site type x and specified geometric design and traffic control features y 𝐶 = calibration factor to adjust SPF for local conditions for site type x Step 3: Expected Crash Frequency with Empirical Bayes Method This step can be omitted if recorded crash data for the specific site under investigation were unavailable or data are considered unreliable. When historical crash data is available, the EB method (either site- specific or project-level) is used to combine the HSM Chapter 11 predicted average crash frequency with the observed crash frequency. The expected average crash frequency is a more statistically reliable estimate. The expected average crash frequency can be determined using Equation 5: 𝑁 =𝑤 × 𝑁 + (1 − 𝑤) × 𝑁 (Eq. 5)

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 33 where: 𝑤 = the weighted adjustment to be placed on the predictive model estimate. This value can be calculated using Equation 6: 𝑤 = ×∑ (Eq. 6) where: 𝑘 = the overdispersion parameter of the associated SPF used to estimate 𝑁 . Table 14 lists the 𝑘 values for SPFs of different facility types. TABLE 14 Chapter 11 SPFs Overdispersion Parameters Facility Type Overdispersion Parameter (k) Rural four-lane undivided segments (4U) 1/e (c + ln (L) ) Rural four-lane divided segments (4D) 1/e (c + ln (L) ) Unsignalized three-leg (stop control on minor-road approaches) (3ST) Coefficients listed in HSM Table 11-7 Unsignalized three-leg (stop control on minor-road approaches) (4ST) Coefficients listed in HSM Table 11-7 Signalized four-leg (4SG)1 Coefficients listed in HSM Table 11-8 Note: 1 The four-leg signalized intersection models do not have base conditions and, therefore, can be used only for generalized predictions of crash frequency Step 4: Crash Frequency under Different Collision Types and Crash Severity Levels HSM Chapter 11 SPFs provide regression coefficients to estimate not only total crashes, but also fatal- and-injury crashes for segments and intersections. These coefficients can be found in HSM Table 11-3 (HSM p. 11-15), and Table 11-5 (HSM p. 11-19) for undivided and divided roadway segments, and HSM Tables 11-7 and 11-8 (HSM p. 11-22) for intersections. PDO crashes are computed as the difference between total and fatal-and-injury crashes. In addition, collision type distribution tables are included for all the facility types, as listed in Table 15. The crash frequency for different collision types can be determined based on the distribution table after the predicted or expected crash frequencies are calculated. These proportions can be updated based on local data for a particular jurisdiction as part of the calibration process. TABLE 15 Rural Multilane Highway Collision Type Distributions Facility Type Collision Type Distribution Rural four-lane undivided segments (4U) HSM Table 11-4 Rural four-lane divided segments (4D) HSM Table 11-6 Unsignalized three-leg (stop control on minor-road approaches) (3ST) HSM Table 11-9 Unsignalized three-leg (stop control on minor-road approaches) (4ST) HSM Table 11-9 Signalized four-leg (4SG) HSM Table 11-9

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 34 Figure 16 shows the HSM Chapter 11 predictive method flowchart for calculating the predicted and expected crash frequency for rural multilane highways. Figure 16: Flowchart for Calculating Predicted and Expected Crash Frequency on Rural Multilane Highways 2.3.17 Data Requirements for Rural Multilane Highways For the study period, it is important to determine the availability of AADT volumes, and for an existing roadway, the availability of observed crash data to determine whether the EB method is applicable. A good understanding of the SPFs’ base conditions will help determine relevant data needs and avoid unnecessary data collection. The base conditions for rural multilane highways are defined in Section 2.3.16, as well as in HSM Sections 11.6.1 (HSM p. 11-14) and 11.6.2 (HSM p. 11-17) for roadway segments and HSM Section 11.6.3 (HSM p. 11-20) for intersections. General data for intersections and segments can be collected from different sources. Examples of data sources include commercial aerial maps, design plans, and states’ roadway inventory systems. Data needed for this example are summarized in the following sections. Intersection Data Generally, the effect of major and minor road traffic volumes (AADT) on crash frequency is incorporated through SPFs, while the effects of geometric design and traffic controls are incorporated through the CMFs. Data required to apply the predictive method for intersections are listed in Table 16.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 35 TABLE 16 Intersection Data Requirements for Rural Multilane Highways Intersections Units/Description Intersection type Unsignalized three-leg (3ST), unsignalized four-leg (4ST), and signalized four-leg (4SG) Traffic flow major road AADTmajor (vpd) Traffic flow minor road AADTminor (vpd) Intersection skew angle degrees Number of signalized or uncontrolled approaches with a left-turn lane From 0 to 4 Number of signalized or uncontrolled approaches with a right-turn lane From 0 to 4 Intersection lighting Present or not present Calibration factor (Ci) Derived from the calibration process Observed crash data Applicable only with the EB method; crashes that occur at the intersection or on an intersection leg, and are related to the presence of an intersection during the period of study Notes: Ci = intersection calibration factor vpd = vehicles per day Roadway Segment Data The effect of traffic volume on crash frequency is incorporated through an SPF, while the effects of geometric design and traffic control features are incorporated through the CMFs. There is no minimum roadway segment length when applying the predictive method. However, when dividing the facility into small homogeneous sections, it is recommended to keep the minimum roadway segment length as 0.10 mile to minimize the calculation efforts and avoid affecting the results. Table 17 includes data requirements for roadway segment locations. TABLE 17 Roadway Segment Data Requirements for Rural Multilane Highways Roadway Segments Units/Description Segment length miles Traffic volume AADT (vpd) Lane width feet Shoulder width feet Shoulder type – right shoulder for divided Paved, gravel, composite, or turf Median width (Divided Only) feet Side slopes (Undivided Only) miles Segment lighting Present or not present Auto speed enforcement present or not present

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 36 TABLE 17 Roadway Segment Data Requirements for Rural Multilane Highways Roadway Segments Units/Description Calibration factor (Cr) Derived from the calibration process Observed crash data Applicable only with the EB method; crashes that occur between intersections and are not related to the presence of an intersection during the period of study Notes: AADT = average annual daily traffic Cr = roadway segment calibration factor vpd = vehicles per day 2.3.18 HSM Chapter 12: Predictive Method for Urban and Suburban Arterials HSM Chapter 12 provides a structured methodology for estimating the predicted and/or expected average crash frequency, crash severity, and collision types for urban and suburban arterial facilities. Crashes involving all vehicle types, bicycles, and pedestrians are included, except for crashes between bicycles and pedestrians. The method is applicable to existing sites, design alternatives to existing sites, new sites, and alternative traffic volume projections. This chapter is applicable to all arterials that are inside urban boundaries where the population is greater than 5,000 people (HSM Section 12.3, p. 12-2). The term suburban refers to outlying portions of an urban area. This chapter includes arterials without full access control, other than freeways, with two- or four-lane undivided facilities, four-lane divided and three- and five-lane roads with center TWLTLs in urban and suburban areas. HSM Chapter 12 includes three- and four-leg intersections with minor-road stop control or traffic signal control on all the roadway cross sections to which the chapter applies. Table 18 contains the site types on urban and suburban arterials for which SPFs have been developed for predicting average crash frequency, severity, and collision type. TABLE 18 Roadway Segment and Intersection Types and Descriptions for Urban and Suburban Arterials Facility Type Site Types with SPFs in HSM Chapter 12 Roadway Segments Two-lane undivided arterials (2U) Three-lane arterials with a center TWLTL (3T) Four-lane undivided arterials (4U) Four-lane divided arterials (4D) Five-lane arterials including a center TWLTL (5T) Intersections Unsignalized three-leg (stop control on minor-road approaches) (3ST) Signalized three-leg intersections (3SG) Unsignalized four-leg (stop control on minor-road approaches) (4ST) Signalized four-leg (4SG) Figure 17 lists the facility types and definitions provided in HSM Chapter 12.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 37 Figure 17: Urban and Suburban Arterials Facility Types and Definitions Commonly, a roadway consists of a contiguous group of sites (intersections and roadway segments). On each roadway, multiple site types may exist, including divided and undivided segments and signalized and unsignalized intersections. To apply the predictive method, the roadway is divided into individual homogeneous segments and intersections. HSM Chapter 12 provides guidance on how to define roadway segments and intersections (HSM Section 12.5, p. 12-9). A roadway segment is defined as a section of continuous traveled way that provides two-way operation of traffic uninterrupted by an intersection and consists of homogeneous geometric and traffic control features. A segment begins and ends at the center of bounding intersections, or where there is a change in homogeneous roadway characteristics. When a roadway segment begins or ends at an intersection, the length of the roadway segment is measured from the center of the intersection. Urban and Suburban Arterials Facility Type Definitions Facility Type Definition Two-lane undivided arterials Two-lane roadway with a continuous cross-section providing two directions of travel in which the lanes are not physically separately by either distance or a barrier. Three-lane arterials Three-lane roadway with a continuous cross-section providing two directions of travel, with a TWLTL in the center. Four-lane undivided arterials Four-lane roadway with a continuous cross-section providing two directions of travel in which the lanes are not physically separated by either distance or a barrier. Four-lane divided arterials Two-lane roadway with a continuous cross-section providing two directions of travel in which the lanes are physically separated by either distance or a barrier. Roadways with raised or depressed median are also included in this category. Five-lane arterials including a center TWLTL Five-lane roadway with a continuous cross-section providing two directions of travel in which the center lane is a TWLTL. Unsignalized three-leg intersection with stop control Intersection of an urban/suburban arterial with a minor road. Stop sign is present on the minor road approach. Signalized three-leg intersections Intersection of an urban/suburban arterial with a minor road. Traffic light is provided at the intersection. Unsignalized four-leg (stop control on minor- road approaches Intersection of an urban/suburban arterial with two minor roads. Stop sign is present on both the minor road approaches. Signalized four-leg intersection Intersection of an urban/suburban arterial with two minor roads. Traffic light is provided at the intersection.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 38 An intersection is defined as the junction of two or more roadway segments. The intersection predictive models estimate the predicted and/or expected average crash frequencies within the intersection limits (Region A in Figure 18) and intersection-related crashes that occur on the intersection legs (Region B in Figure 18). Figure 18. Urban and Suburban Arterials – Definition of Roadway Segments and Intersections 2.3.19 Calculating the Crash Frequency for Urban and Suburban Arterials HSM Chapter 12 provides the methodology for calculating the predicted and/or expected crash frequency for roadway segments and intersections on urban and suburban arterials. The calculation is for a given period during which the geometric design and traffic control features are unchanged and traffic volumes are known. The entire process could be divided into the following steps: 1. Predicted crash frequency under base conditions 2. Predicted crash frequency under site conditions 3. Expected crash frequency with EB method 4. Crash frequency under different collision types and crash severity levels Step 1: Predicted Crash Frequency under Base Conditions The predicted crash frequency for the roadway segments and intersections under the base condition can be determined by replacing the AADT and segment length (for roadway segments) or the AADTs for major and minor roads (for intersections) in SPFs with site-specific values. Table 19 lists the different facility types included in HSM Chapter 12 and the applicable AADT ranges for the SPFs. Only application to sites within the AADT ranges would provide reliable results. TABLE 19 Urban and Suburban Arterials Facility Types and AADT Ranges Item Facility Type AADT Range Roadway Segments Two-lane undivided arterials (2U) Up to 32,600 vpd Three-lane arterials with TWLTL (3T) Up to 32,900 vpd

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 39 TABLE 19 Urban and Suburban Arterials Facility Types and AADT Ranges Item Facility Type AADT Range Four-lane undivided arterials (4U) Up to 40,100 vpd Four-lane divided arterials (4D) Up to 66,000 vpd Five-lane arterials with TWLTL (5T) Up to 53,800 vpd Intersections Three-leg intersection with stop control on minor approach (3ST) AADTmajor 0 to 45,700 vpd AADTminor 0 to 9,300 vpd Three-leg signalized intersection (3SG) AADTmajor 0 to 46,800 vpd AADTminor 0 to 5,900 vpd Four-leg intersection with stop control on minor approach (4ST) AADTmajor 0 to 58,100 vpd AADTminor 0 to 16,400 vpd Four-leg signalized intersection (4SG) AADTmajor 0 to 67,700 vpd AADTminor 0 to 33,400 vpd 4SG intersections pedestrian models AADTmajor 0 to 82,000 vpd AADTminor 0 to 49,100 vpd Pedestrianvol 0 to 34,200 ped/day Note: Pedestrianvol = pedestrians per day crossing all four legs combined SPFs are provided for different collision types: multiple-vehicle nondriveway, single-vehicle, multiple- vehicle driveway-related, and vehicle-pedestrian collisions. Adjustment factors are provided for vehicle- bicycle and stop-controlled intersection vehicle-pedestrian collisions. Table 20 summarizes the different SPFs by collisions type for roadway segments and intersections. TABLE 20 Urban and Suburban Arterials SPFs in HSM Chapter 12 Facility Type SPF Components by Collision Type HSM Equation Roadways Segments Multiple-vehicle nondriveway collisions HSM Equations 12-10, 12-11, and 12-12 (HSM p. 12-18 and 12-20) Single-vehicle crashes HSM Equations 12-13, 12-14, and 12-15 (HSM p. 12-20 to 12-21) Multiple-vehicle driveway-related collisions HSM Equations 12-16, 12-17, and 12-18 (HSM p. 12-22 and 12-27) Vehicle-pedestrian collisions HSM Equation 12-19 (HSM p. 12-27) Vehicle-bicycle collisions HSM Equation 12-20 (HSM p. 12-27) Intersections Multiple-vehicle collisions HSM Equations 12-21, 12-22, and 12-23 (HSM p. 12-29) Single-vehicle crashes HSM Equations 12-24, 12-25, 12-26, and 12-27 (HSM p. 12-32 to 12-33 and p. 12-36) Vehicle-pedestrian collisions HSM Equations 12-28, 12-29, and 12-30 (HSM p. 12-36 and 12-38) Vehicle-bicycle collisions HSM Equation 12-31 (HSM p. 12-38)

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 40 Step 2: Predicted Crash Frequency under Real Conditions Each SPF listed in Table 20 is used to estimate the predicted crash frequency of a roadway segment or intersection under base conditions, which is later adjusted to site-specific conditions. Base conditions are a specific set of geometric design and traffic control features under which the SPFs were developed and are not necessarily the same for all facilities. The base conditions for roadway segments and intersections on urban and suburban arterials are listed in Figure 19. Figure 19: Urban and Suburban Arterials Base Conditions CMFs are applied to account for the differences between the specific site under investigation and the base conditions. CMFs are used to adjust the SPF estimate of predicted average crash frequency for the effect of individual geometric design and traffic control features. The CMF for the SPF base condition of each geometric design and traffic control feature has a value of 1.00. CMF values less than 1.00 indicate the treatments reduce the predicted average crash frequency in comparison to the base condition. Similarly, CMF values greater than 1.00 indicate the treatments increase the predicted crash frequency. The CMFs presented in HSM Chapter 12 and the specific site types to which they apply are listed in Table 21. TABLE 21 CMFs for Urban and Suburban Arterials Roadway Segments and Intersections Facility Type CMF CMF Description CMF Equations and Tables Roadway Segments CMF1r On-street parking Definition (HSM p. 12-40) HSM Table 12-19 (HSM p. 12-40) HSM Equation 12-32 (HSM p. 12-40) CMF2r Roadside fixed objects Definition (HSM p. 12-41) Urban and Suburban Arterials Base Conditions Roadway Segments Intersections • Absence of on-street parking • Absence of fixed objects • For divided facilities: median width of 15 feet • Absence of lighting • Absence of Automated Speed Enforcement • Absence of left-turn lanes • Permissive left-turn signal phasing • Absence of right-turn lanes • Permitting right turn on red (RTOR) • Absence of intersection lighting • Absence of RLR cameras • Signalized: vehicle-pedestrian collisions - Absence of bus stops within 1,000 feet - Absence of schools within 1,000 feet of intersection - Absence of alcohol sales establishments within 1,000 feet of intersection

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 41 TABLE 21 CMFs for Urban and Suburban Arterials Roadway Segments and Intersections Facility Type CMF CMF Description CMF Equations and Tables HSM Tables 12-20 and 12-21 (HSM p. 12-41) HSM Equation 12-33 (HSM p. 12-40) CMF3r Median width Definition (HSM p. 12-41) HSM Table 12-22 (HSM p. 12-42) CMF4r Lighting Definition (HSM p. 12-42) HSM Equation 12-34 (HSM p. 12-42) HSM Table 12-23 (HSM p. 12-42) CMF5r Automated speed enforcement Definition (HSM p. 12-43) See text (HSM p. 12-43) Multiple-Vehicle Collisions and Single-Vehicle Crashes at Intersections CMF1i Intersection left-turn lanes Definition (HSM p. 12-43) HSM Table 12-24 (HSM p. 12-43) CMF2i Intersection left-turn signal phasing Definition (HSM p. 12-43 to 12-44) HSM Table 12-25 (HSM p. 12-44) CMF3i Intersection right-turn lanes Definition (HSM p. 12-44) HSM Table 12-26 (HSM p. 12-44) CMF4i Right-turn-on-red Definition (HSM p. 12-44) HSM Equation 12-35 (HSM p. 12-44) CMF5i Lighting Definition (HSM p. 12-45) HSM Table 12-27 (HSM p. 12-45) HSM Equation 12-36 (HSM p. 12-45) CMF5i Red-light cameras Definition (HSM p. 12-45 to 12-46) HSM Equations 12-37, 12-38, and 12-39 (HSM p. 12-45) Vehicle-Pedestrian Collisions at Signalized Intersections CMF1p Bus stops Definition (HSM p. 12-46) HSM Table 12-28 (HSM p. 12-46) CMF2p Schools Definition (HSM p. 12-46) HSM Table 12-29 (HSM p. 12-46) CMF3p Alcohol sales establishments Definition (HSM p. 12-47) HSM Table 12-30 (HSM p. 12-47)

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 42 The SPFs were developed in HSM-related research from the most complete and consistent available data sets. However, the predicted crash frequencies may vary substantially from one jurisdiction to another for a variety of reasons. Calibration factors provide a method for incorporating local data to improve the estimated crash frequencies for individual locations. The local calibration factor accounts for the differences between the jurisdiction under investigation and the jurisdictions that were used to develop the default HSM SPFs. The local calibration factor is calculated using local crash data and other roadway characteristic data. The process for determining calibration factors for the predictive models is described in HSM Part C, Appendix A.1. The predicted crash frequency under real conditions can be calculated using Equation 7: 𝑁 = 𝑁 × 𝐶𝑀𝐹 × 𝐶𝑀𝐹 × 𝐶𝑀𝐹 × … × 𝐶𝑀𝐹 + 𝑁 + 𝑁 × 𝐶 (Eq. 7) Segments: 𝑁 = 𝑁 + 𝑁 + 𝑁 (Eq. 8) Intersections: 𝑁 = 𝑁 + 𝑁 (Eq. 9) where: 𝑁 = predicted average crash frequency for a specific year on site type x 𝑁 = base conditions predicted average crash frequency for site type x 𝑁 = base conditions predicted average crash frequency multiple-vehicle nondriveway collisions for site type x 𝑁 = base conditions predicted average crash frequency single-vehicle crashes for site type x 𝑁 = base conditions predicted average crash frequency multiple-vehicle driveway-related collisions for site type x 𝑁 = predicted average crash frequency of vehicle-pedestrian collisions per year for site type x 𝑁 = predicted average crash frequency of vehicle-bicycle collisions per year for site type x 𝐶𝑀𝐹 = CMFs specific to site type x and specified geometric design and traffic control features y 𝐶 = calibration factor to adjust SPF for local conditions for site type x Step 3: Expected Crash Frequency with Empirical Bayes Method This step can be omitted if no recorded crash data for the specific site under investigation were available or considered unreliable. When historical crash data is available, the EB method (either site-specific or project-level) is used to combine the HSM Chapter 12 predicted average crash frequency with the observed crash frequency. The expected crash frequency is a more statistically reliable estimate. The expected average crash frequency can be determined using Equation 10: 𝑁 = 𝑤 × 𝑁 + (1 − 𝑤) × 𝑁 (Eq. 10) where: 𝑤 = the weighted adjustment to be placed on the predictive model estimate. The value can be calculated using the following equation: 𝑤 = ×∑ (Eq. 11)

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 43 where: 𝑘 = the overdispersion parameter of the associated SPF used to estimate 𝑁 . Table 22 lists the overdispersion values for urban and suburban arterials. TABLE 22 SPFs Overdispersion Parameters in Chapter 12 Facility Type Overdispersion Parameter (k) Segments multiple-vehicle nondriveway collisions Coefficients listed in HSM Table 12-3 Segments single-vehicle crashes Coefficients listed in HSM Table 12-5 Segments multiple-vehicle driveway-related collisions Coefficients listed in HSM Table 12-7 Intersections multiple-vehicle collisions Coefficients listed in HSM Table 12-10 Intersections single-vehicle crashes Coefficients listed in HSM Table 12-12 Intersections vehicle-pedestrian collisions Coefficients listed in HSM Table 12-14 Step 4: Crash Frequency under Different Collision Types and Crash Severity Levels HSM Chapter 12 provides the collision type distribution tables based on the crash severity level for roadway segments and intersections (Table 23). The crash frequency under different severity levels and collision types can be determined based on the distribution table after the predicted or expected crash frequencies are calculated. These proportions can be updated based on local data for a particular jurisdiction as part of the calibration process. TABLE 23 Urban and Suburban Arterial Crash Severity and Collision Type Distributions Facility Type Collision Type Crash Severity and Collision Type Distribution Roadways Segments Multiple-vehicle nondriveway collisions HSM Table 12-4 Single-vehicle crashes HSM Table 12-6 Multiple-vehicle driveway-related collisions HSM Table 12-7 Vehicle-pedestrian collisions HSM Table 12-8 Vehicle-bicycle collisions HSM Table 12-9 Intersections Multiple-vehicle collisions HSM Table 12-11 Single-vehicle crashes HSM Table 12-13 Vehicle-pedestrian collisions HSM Table 12-16a Vehicle-bicycle collisions HSM Table 12-17 Note: a Pedestrian crash adjustment factors for stop-controlled intersections Figure 20 illustrates the HSM Chapter 12 predictive method flowchart for calculating the predicted and expected crash frequency for urban and suburban arterial roads.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 44 Figure 20: Flowchart for Calculating Expected Crash Frequency on Urban and Suburban Arterials 2.3.20 Data Requirements for Urban and Suburban Arterials For the study period, it is important to determine the availability of AADT volumes, and for an existing roadway, the availability of observed crash data to determine whether the EB method is applicable. To determine the relevant data needs and avoid unnecessary data collection, it is important to understand the SPFs’ base conditions. The base conditions for urban and suburban arterials are defined in Section 2.3.19, as well as in HSM Section 12.6.1 (HSM p. 12-17) for roadway segments and in HSM Section 12.6.2 (HSM p. 12-28) for intersections. General data for intersections and roadway segments can be collected from different sources. Examples of data sources include commercial aerial maps, design plans, and states’ roadway inventory systems. Data needed for this example are summarized in the following sections. Intersection Data Generally, the effect of major and minor road traffic volumes (AADT) on crash frequency is incorporated through SPFs, while the effects of geometric design and traffic controls are incorporated through the CMFs. Data required to apply the predictive method for intersections are listed in Table 24.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 45 TABLE 24 Intersection Data Requirements for Urban and Suburban Arterials Intersections Units/Description Intersection type Include unsignalized three-leg (3ST), signalized three-leg (3SG), unsignalized four-leg (4ST), and signalized four-leg (4SG) Traffic flow major road AADT (vpd) Traffic flow minor road AADT (vpd) Intersection lighting Present or not present Calibration factor Derived from calibration process Data for unsignalized intersections only Number of major-road approaches with left- turn lanes 0, 1, or 2 Number of major-road approaches with right- turn lanes 0, 1, or 2 Data for signalized intersections only Number of approaches with left-turn lanes 0, 1, 2, 3, or 4 Number of approaches with right-turn lanes 0, 1, 2, 3, or 4 Number of approaches with left-turn signal phasing 0, 1, 2, 3, or 4 Type of left-turn signal phasing for all legs Not applicable, permissive, protected, protected/permissive, or permissive/protected Number of approaches with right-turn-on-red prohibited 0, 1, 2, 3, or 4 Intersection red-light cameras Present or not present Sum of all pedestrian crossing volumes-only signalized intersection Sum of pedestrian volume Maximum number of lanes crossed by a pedestrian Number of lanes Number of bus stops within 1,000 feet (300 meters) of intersection Number Schools within 1,000 feet (300 meters) of intersection Number Number of alcohol sales establishments within 1,000 feet (300 meters) Number Observed crash data Applicable only with the EB method; crashes that occur at the intersection or intersection legs, and are related to the presence of an intersection during the period of study

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 46 Roadway Segment Data The effect of traffic volume (AADT) on crash frequency is incorporated through an SPF, while the effects of geometric design and traffic control features are incorporated through the CMFs. Table 25 includes data requirements for roadway segment locations. TABLE 25 Roadway Segment Data Requirements for Urban and Suburban Arterials Roadway Segments Units/Description Roadway type (2U, 3T, 4U, 4D, ST) Include two-lane undivided arterials (2U), three-lane arterials (3T) including a center TWLTL, four-lane undivided arterials (4U), four-lane divided arterials (4D), and five-lane arterials (5T) including a center TWLTL Segment length miles Traffic volume AADT (vpd) Type of on-street parking None, parallel, or angle Proportion of curb length with on-street parking percent Median width – for divided only Not present, or select from scale 10 feet to 100 feet Lighting Present or not present Auto speed enforcement Present or not present Major commercial driveways Number Minor commercial driveways Number Major industrial/institutional driveways Number Minor industrial/institutional driveways Number Major residential driveways Number Minor residential driveways Number Other driveways Number Speed category Posted speed 30 mph or greater Roadside fixed-object density Fixed objects per mile Offset to roadside fixed objects Length (feet) Calibration factor Derived from calibration process Observed crash data Applicable only with the EB method; crashes that occur between intersections and are not related to the presence of an intersection during the period of study Note: mph = miles per hour

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 47 2.4 HSM Part D: CMF Applications Guidance HSM Part D provides information on estimating how effective a treatment, geometric characteristic, and operational characteristic will be in reducing crashes or injuries at a specific location. The effectiveness is expressed in terms of CMFs, trends, or no effect. The CMFs can be used to evaluate the expected average crash frequency with or without a particular treatment or estimate the expected average crash frequency with one treatment versus a different treatment. CMFs are provided for roadway segments (HSM Chapter 13), intersections (HSM Chapter 14), interchanges (HSM Chapter 15), special facilities and geometric situations (HSM Chapter 16), and road networks (HSM Chapter 17). Part D includes all CMFs in the HSM. Some Part D CMFs are included in Part C for use with specific SPFs. The remaining Part D CMFs can be used with the outcomes of the predictive method to estimate the change in crash frequency described in HSM Section C.7 (HSM p. C-19). HSM Part D can be applied to the different stages of the project development process, as listed in Table 26. TABLE 26 Stages of the Project Development Process Sy st em P la nn in g Pr oj ec t P la nn in g Pr el im in ar y D es ig n Fi na l D es ig n C on st ru ct io n/ Im pl em en ta tio n O pe ra tio n M ai nt en an ce HSM Part D        Other Relevant Chapters in the HSM HSM Part B HSM Part C HSM Ch. 5 HSM Ch. 6 HSM Ch. 7 HSM Part C HSM Ch. 6 HSM Ch. 7 HSM Part C HSM Ch. 6 HSM Ch. 7 HSM Part C HSM Ch. 6 HSM Ch. 7 HSM Ch. 5 HSM Ch. 6 HSM Ch. 7 HSM Ch. 5 HSM Ch. 6 HSM Ch. 7 HSM Part D introduces the following concepts: • Crash modification factor – An index of how much crash experience is expected to change following a modification in design or traffic control. CMF is the ratio between the number of crashes per unit of time expected after a modification or measure is implemented and the number of crashes per unit of time estimated if the change does not take place. • Precision – The degree to which repeated measurements are close to each other. • Standard error – Indicates the precision of an estimated CMF. It is used as a measure of reliability of the CMF estimate. The smaller the standard error, the more reliable (less error) the estimate becomes. A CMF with a relatively high standard error means that a high range of results could be obtained with that treatment. It can also be used to calculate a confidence interval for the estimated change in expected average crash frequency. Refer to HSM Appendix 3C (HSM p. 3-44) for additional details about CMF and standard error.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 48 • CMF confidence interval – It can be used to consider the possible range of the CMFs. For CMFs with high standard errors, the upper end of the confidence interval could be greater than 1.0 even if the CMF itself is relatively small, which means that the treatment could potentially result in an increase in crashes. Some CMFs in Part D are accompanied by a superscript when special awareness of the standard error is required. • Trend – If the standard error was greater than 0.10, the CMF value was not sufficiently accurate, precise, and stable to be included in HSM Part D. In these cases, HSM Part D indicates a trend, if sufficient information is available. HSM Part D includes such information in the appendix at the end of each chapter. The HSM Appendix also lists treatments with unknown crash effects. • Accuracy – A measure of the proximity of an estimate to its actual or true value. Part D CMFs were evaluated by an expert panel for inclusion in the HSM based on their standard error. Standard error values were used to determine the level of reliability and stability of the CMFs to be presented in the HSM. A standard error of 0.10 or less indicates a CMF value that is sufficiently accurate, precise, and stable. Some CMFs are expressed as functions, and do not have specific standard errors that could be used. Understanding the standard error and reliability of the different CMFs will help analysts to build awareness of what can be expected from each safety treatment. A CMF with a high standard error does not mean that it should not be used; it means that if the CMF is used, the user should keep in mind the range of results that could be obtained. 2.4.1 HSM Chapter 13: Roadway Segments HSM Chapter 13 provides the information used to identify effects on expected average crash frequency resulting from treatments applied to roadway segments. A roadway segment is defined as a continuous portion of a roadway with similar geometric, operational, and vehicular characteristics. The more than 80 roadway segment treatments are classified based on the treatment characteristics. For each treatment category, the treatment CMF availability is provided in a table. The HSM table numbers for treatment summary information are listed in Table 27. TABLE 27 Roadway Segments – HSM Table Number for Information on Treatment Summary Treatment Category HSM Table Number for Treatment Summary Roadway element HSM Table 13-1 Roadside element HSM Table 13-17 Alignment element HSM Table 13-26 Roadway sign HSM Table 13-29 Roadway delineation HSM Table 13-34 Rumble strip HSM Table 13-43 Traffic calming HSM Table 13-47 On-street parking HSM Table 13-49 Roadway treatment for pedestrians and bicyclists HSM Table 13-54 Highway lighting HSM Table 13-55 Roadway access management HSM Table 13-57 Weather issue HSM Table 13-59

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 49 The CMFs for different treatments are usually provided in the format of figures, equations, or tables. Users may then determine the CMFs and relevant standard errors based on the treatment and the facility characteristics. When determining the CMFs for a specific treatment on a particular facility type, special attention should be paid to the AADT range, setting, and crash types for which the CMFs were developed. For treatments without CMF values, the user can refer to HSM Appendix 13A to obtain information about the trend in crashes or user behavior (if available). HSM Appendix 13A also lists some treatments with unknown crash effects at the time the HSM was developed. 2.4.2 HSM Chapter 14: Intersections HSM Chapter 14 provides information used to identify effects on expected average crash frequency resulting from treatments applied at intersections. An intersection is defined as the general area where two or more roadways join or cross, including the roadway and roadside facilities for traffic movements within the area. There are more than 50 intersection treatments included in the HSM Part D, and they are classified based on the treatment characteristics. CMFs are organized into the following three categories: CMF is available; information available was sufficient to present a trend but not a CMF; and quantitative information is not available. For each treatment category, the treatment CMF availability is provided in a table. The HSM table numbers for treatment summary information are listed in Table 28. TABLE 28 Intersections – HSM Table Number for Information on Treatment Summary Treatment Category HSM Table Number for Treatment Summary Intersection type HSM Table 14-1 Access management HSM Table 14-8 Intersection design elements HSM Table 14-9 Intersection traffic control and operational elements HSM Table 14-19 The CMFs for different treatments are usually provided in the format of figures, equations, or tables. The users could then determine the CMFs and relevant standard errors based on the treatment and the facility characteristics. Special attention should be paid to the AADT range, setting, and the crash types used to develop the CMFs. This is particularly important when determining a CMF for a specific treatment on a particular facility type. Treatments without CMF values indicate that research quantitative information was not enough to be included in the HSM. HSM Appendix 14A lists some treatments with unknown crash effects at the time the HSM was being developed. 2.4.3 HSM Chapter 15: Interchanges HSM Chapter 15 provides the information used to identify effects on expected average crash frequency resulting from treatments applied at interchanges and interchange ramp terminals. An interchange is defined as a system of interconnecting roadways in conjunction with one or more grade separations that provides for the movement of traffic between two or more roadways or highways on different levels, and an interchange ramp terminal is defined as an at-grade intersection where a freeway interchange ramp intersects with a nonfreeway cross street.

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 50 The crash effects of interchange design elements are included in this chapter. The list of treatments included under interchange design elements and availability of relevant CMFs for different facility types are presented in HSM Table 15-1 at the beginning of HSM Section 15.4 (see Table 29). TABLE 29 Interchanges – HSM Table Number for Information on Treatment Summary Treatment Category HSM Table Number for Treatment Summary Interchange design elements HSM Table 15-1 For treatments without CMF values, the user can refer to HSM Appendix 15A to determine whether sufficient information about potential trend in crashes or user behavior for the treatment could be found. HSM Appendix 15A lists some treatments with unknown crash effects at the time the HSM was being developed. 2.4.4 HSM Chapter 16: Special Facilities and Geometric Situations HSM Chapter 16 provides CMFs for design, traffic control, and operational elements at various special facilities and geometric situations including highway-rail grade crossings, work zones, TWLTLs, and passing and climbing lanes. For each special facility or geometric situation, the list of treatments included and the availability of relevant CMFs for different facility types are provided in tables. Information from this table can be used to check the availability of the CMF for a specific treatment on a particular facility type. The HSM table numbers for treatment summary information are listed in Table 30. TABLE 30 Special Facilities and Geometric Situations – HSM Table Number for Information on Treatment Summary Treatment Category HSM Table Number for Treatment Summary Highway-rail grade crossing traffic control and operational elements HSM Table 16-1 Work zone design elements HSM Table 16-4 TWLTL elements HSM Table 16-5 Passing and climbing lanes HSM Table 16-6 For treatments without CMF values, the user can refer to HSM Appendix 16A to determine whether sufficient information about potential trends in crashes or user behavior for the treatment could be found. HSM Appendix 16A lists some treatments with unknown crash effects at the time the HSM was developed. 2.4.5 HSM Chapter 17: Road Networks The information presented in HSM Chapter 17 is used to identify effects on expected average crash frequency resulting from treatments applied to road networks. Nearly 20 treatments for road networks are included in HSM Chapter 17. The treatments for road networks are classified into categories based on the treatment characteristics. For each treatment category, a summary of treatments related to the specific treatment category, including a list of treatment and the availability of relevant CMFs for different facility types is provided in a table. Information from this table can be used to check the availability of the CMF for a specific treatment

SECTION 2 – HIGHWAY SAFETY MANUAL OVERVIEW 51 on a particular facility type. The HSM table numbers for treatment summary information are listed in Table 31. TABLE 31 Road Networks – HSM Table Number for Information on Treatment Summary Treatment Category HSM Table Number for Treatment Summary Network planning and design approaches/ elements HSM Table 17-1 Network traffic control and operational elements HSM Table 17-2 Road-use culture network considerations and treatments HSM Table 17-4 For treatments without CMF values, the user can refer to HSM Appendix 17A to determine whether sufficient information about potential trend in crashes or user behavior for the treatment could be found. HSM Appendix 17A lists some treatments with unknown crash effects at the time the HSM was being developed.