Crash Modification Factors in the Highway Safety Manual: A Review (2023)

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.

13   Recommendations for Incorporating CMFs in the HSM and Possible Framework for Practitioner Use This chapter discusses how a new rating system for CMFs was developed. It also includes recommendations for how CMFs may be incorporated in the HSM and a proposed framework that is applicable for practitioner use. 4.1 Incorporating CMFs in the HSM2 In January 2016, the CMF subcommittee of TRB’s Highway Safety Performance Committee formed an ad hoc committee to investigate different options for Part D of the HSM2. The ad hoc committee was not charged with recommending a particular option but was asked to identify the advantages and disadvantages of different options to assist AASHTO and the NCHRP Proj­ ect 17­71 and 17­72 teams (and the project panels) in selecting the best option. This ad hoc committee was led by Raghavan Srinivasan, the principal investigator on NCHRP 17­72. The other members of the ad hoc committee included Daniel Carter of the University of North Carolina Highway Safety Research Center (HSRC), Raul Avelar of Texas A&M University, Geni Bahar of NAVIGATS Inc., and Alejandra Medina of the Virginia Tech Transportation Institute. The report from the ad hoc committee was submitted to TRB’s Highway Safety Performance Committee in April 2016. The ad hoc committee considered the following five options: • Option A: Printed Part D would include eligible CMFs, as in the 1st edition of the HSM (HSM1). • Option B: Part D would include eligible CMFs, as in the HSM1 but would be electronic. • Option C: Part D would not include CMFs but would present guidance for using and devel­ oping CMFs. CMFs would be available in the CMF Clearinghouse with an HSM stamp of approval. • Option D: Part D would not include CMFs but would present guidance for using and devel­ oping CMFs. CMFs would be available in the CMF Clearinghouse without an HSM stamp of approval. • Option E: Part D would include only a select set of CMFs. All CMFs would be available in the CMF Clearinghouse with an HSM stamp of approval. Following is a discussion of each option along with its advantages and disadvantages. 4.1.1 Option A 4.1.1.1 Definition/Assumptions Option A represents the status quo. Part D would be printed only and would include all CMFs that meet the inclusion criteria. The CMF Clearinghouse would continue to provide all C H A P T E R 4

14 Crash Modification Factors in the Highway Safety Manual: A Review published CMFs. With this option, a table showing the following information would be pro­ vided for each treatment: • Setting (intersection or roadway type), • Base condition, • Traffic volume conditions, • Indication of whether the CMF or CMFunction was developed with data from high­crash locations or from systemwide implementation, • Crash type and severity, • CMF or CMFunction, • Standard error of CMF, • Indication of whether the CMF is based on the combined value of multiple CMFs, and • Source(s) of the CMF or CMFunction. All CMFs would continue to be compiled in the CMF Clearinghouse with an HSM stamp of approval. 4.1.1.2 Advantages • Option A provides a curated list that takes into account new studies that replace or even conflict with previous CMFs documented in the HSM1. • A published manual may be viewed as a more reliable reference than a website, especially with regard to liability issues. • The consistency of keeping the existing format may be attractive to some users. • Among those who responded to the questionnaire conducted for NCHRP 17­72, the majority of the administrators, planners, and designers wished Part D to continue providing the eligible CMFs. Overall, 33% of the respondents wished Part D to continue providing the eligible CMFs. 4.1.1.3 Disadvantages • Unlike the CMF Clearinghouse, the HSM is not a free product. However, that is not necessarily a disadvantage of Part D, but of the HSM as a whole. • A printed manual cannot be as up to date as a website. Hence, the CMFs in Part D would not include the latest research. 4.1.2 Option B 4.1.2.1 Definition/Assumptions Option B is to include eligible CMFs in Part D, as in the HSM1, but to publish Part D as an electronic, or web­based, document. Part D would include all CMFs that meet the inclusion criteria. In terms of content, the information presented for each CMF would be the same as in Option A. All the CMFs would continue to be compiled in the CMF Clearinghouse, with an indication as to whether a particular CMF is included in the HSM2. 4.1.2.2 Advantages • Option B provides a curated list that takes into account new studies that replace or even conflict with previous CMFs documented in the HSM1. • The electronic format makes it easier to send updates, as compared with a printed Part D. • CMFs in the electronic Part D may be hyperlinked to related information in the CMF Clearinghouse. • Among those who responded to the questionnaire conducted for NCHRP 17­72, the majority of the administrators, planners, and designers wished Part D to continue providing the eligible CMFs. Overall, 33% of respondents wished Part D to continue providing the eligible CMFs.

Recommendations for Incorporating CMFs in the HSM and Possible Framework for Practitioner Use 15 4.1.2.3 Disadvantages • Unlike the CMF Clearinghouse, the HSM is not a free product. However, that is not necessarily a disadvantage of Part D, but of the HSM as a whole. • Although an electronic format is easier to update than a printed document, updating would still require some effort to implement. 4.1.3 Option C 4.1.3.1 Definition/Assumptions Under Option C, Part D of the HSM would not contain any CMFs but would present guid­ ance for using and developing CMFs. CMFs would be available in the CMF Clearinghouse with an HSM stamp of approval. 4.1.3.2 Advantages • About 50% of the respondents to the NCHRP 17­72 questionnaire preferred that the HSM not include CMFs in Part D but that it provide a stamp of approval to CMFs in the CMF Clearinghouse that meet the HSM reliability requirements. • An online format is more accessible. This was clear in the results of the focus group conducted as part of NCHRP 17­72. • There is no cost to access the CMF Clearinghouse. • The CMF Clearinghouse is continuously updated with emerging studies. • The majority of users (especially safety­specific personnel) are currently using the CMF Clearinghouse as their CMF source (NCHRP 17­72 focus group). • This option allows experienced users to get a broad view of all CMFs for a countermeasure, so that they then can select the most appropriate CMF (several participants in the NCHRP 17­72 focus group expressed this opinion). 4.1.3.3 Disadvantages • Among those who responded to the NCHRP 17­72 questionnaire, the majority of the admin­ istrators, planners, and designers wished Part D to continue containing the eligible CMFs. However, this represented only about 33% of the respondents overall. • Those currently using only HSM Part D for CMFs might lose their source of CMFs and possibly stop using CMFs. • Inexperienced or novice users may be more comfortable with a published manual. • Inexperienced or novice users may not know how to navigate the CMF Clearinghouse to find an appropriate CMF. • A manual may provide reassuring backing for safety and design decisions, especially in liability cases. • The CMF Clearinghouse is not an AASHTO product. Hence, AASHTO would rely on FHWA cooperation, approval to expand the HSM stamp of approval, and continued funding. • The incentive to purchase the HSM would be reduced, although it is a manual for safety work other than CMF searches. 4.1.4 Option D 4.1.4.1 Definition/Assumptions Under Option D, Part D of the HSM would not include CMFs but would present guidance for using and developing CMFs. CMFs would be available in the CMF Clearinghouse without an HSM stamp of approval.

16 Crash Modification Factors in the Highway Safety Manual: A Review 4.1.4.2 Advantages • There is no cost to access the CMF Clearinghouse. • The CMF Clearinghouse is continuously updated with emerging studies. • The majority of users (especially safety­specific personnel) use the CMF Clearinghouse as their CMF source (NCHRP 17­72 focus group). • This option allows experienced users to get a broad view of all CMFs for a countermeasure, so that they then can select the most appropriate CMF on the basis of the star rating. 4.1.4.3 Disadvantages • This option would create an even more difficult search for high­quality CMFs than that already experienced by some users. • Only a small portion (less than 10%) of the respondents to the NCHRP 17­72 questionnaire indicated that they would not use or want an HSM stamp of approval. 4.1.5 Option E 4.1.5.1 Definition/Assumptions Under Option E, Part D would include only a select set of CMFs; that is, the CMFs included in Part D would be a subset of all eligible CMFs. This subset might be defined as (1) only planning or broad level, (2) only aggregate CMFs (i.e., all crash types, all crash severities), or (3) national CMFs (i.e., not specific to region, state, or condition). The full set of CMFs, including those with and without HSM approval, would be available in the CMF Clearinghouse. 4.1.5.2 Advantages This option would cater to inexperienced users by providing an easy list of CMFs for a particular countermeasure or design element. 4.1.5.3 Disadvantages Users might not realize that type­ and severity­specific CMFs exist and might • Misapply a total CMF to a specific crash type or severity or • Not look for further CMFs that better address their site condition. 4.2 Development of CMF Rating Process for Before–After and Cross-Sectional Studies In the context of developing a new rating system, three study types were considered: • Individual CMFs and CMFunctions from a single before–after study, • Individual CMFs and CMFunctions from a single cross­sectional study, and • CMFs and CMFunctions from multiple studies. Prior to the interim meeting, the project team developed a rating system for CMFs estimated from single before–after and cross­sectional studies. Rating systems for CMFs and CMFunctions from multiple studies (also called meta­analysis or meta­regression studies) were developed after the interim meeting. To develop a rating/inclusion process for CMFs, the project team used the results of the two reviews discussed earlier—(1) different methods for rating CMFs, including the process used to include CMFs in the HSM, and (2) the star rating system in the CMF Clearinghouse—and other

Recommendations for Incorporating CMFs in the HSM and Possible Framework for Practitioner Use 17 rating systems. The rating systems included a list of factors and levels. For before–after and cross­sectional studies, three factors were included: • Data (maximum of 55 points), • Confounding and appropriateness of statistical analysis (maximum of 75 points), and • Statistical significance (maximum of 20 points). Each level within a factor was associated with points. Any CMF could receive a maximum of 150 points. The rating system developed for before–after and cross­sectional studies was pilot tested with CMFs developed in the following eight studies: • Abuzwidah, M., and Abdel­Aty, M. “Safety Assessment of the Conversion of Toll Plazas to All­Electronic Toll Collection System.” Accident Analysis & Prevention, 80:153–161, 2015. • CH2M Hill. Development of Crash Modification Factors for Combined Treatments on Rural Two-Lane Roads. Missouri Department of Transportation, Jefferson City, 2015. • Fink, J., Kwigizle, V., and Oh, J.­S. “Impact of Adaptive Traffic Control Systems on Crash Frequency and Severity.” Presented at 94th Annual Meeting of the Transportation Research Board, Washington, DC, 2015. • Fitzpatrick, K., and Park, E. S. “Safety Effectiveness of HAWK Pedestrian Treatment.” Trans- portation Research Record: Journal of the Transportation Research Board, No. 2140, pp. 214–223, 2009. • Hoye, A. “Safety Effects of Fixed Speed Cameras: An Empirical Bayes Evaluation.” Accident Analysis & Prevention, 82:263–269, 2015. • Islam, M. T., and El­Basyouny, K. “Full Bayesian Evaluation of the Safety Effects of Reducing the Posted Speed Limit in Urban Residential Area.” Accident Analysis & Prevention, 80:18–25, 2015. • Peng, Y., Geedipally, S. R., and Lord, D. “Investigating the Effect of Roadside Features on Single­Vehicle Roadway Departure Crashes on Rural Two­Lane Roads.” Transportation Research Record: Journal of the Transportation Research Board, No. 2309, pp. 21–29, 2012. • Sun, X., Das, S., Zhang, Z., Wang, F., and Leboeuf, C. “Investigating Safety Impact of Edgelines on Narrow, Rural Two­Lane Highways by Empirical Bayes Method.” Transportation Research Record: Journal of the Transportation Research Board, No. 2433, pp. 121–128, 2014. One goal of the pilot test was to determine whether there should be a specific cutoff value to qualify for an HSM stamp of approval. As part of the pilot test, project team members were asked to rate each CMF in the study and provide a point score. They were also asked to indicate whether the CMFs from a particular study should get the HSM stamp of approval. On the basis of this review, a point score of 100 out of 150 seemed to be a reasonable threshold for identifying CMFs that would get an HSM stamp of approval. Following the pilot test, the rating system was modified. The proposed rating system for all study types is provided in Appendix C. The proposed rating system was used to rate all the CMFs in the CMF Clearinghouse. 4.2.1 Review of the New CMF Rating System The new CMF rating system was presented at many meetings and conferences, starting in 2016. In late 2019, the panel for the NCHRP 17­71 project team provided the NCHRP 17­72 project team with questions regarding the new rating system developed in NCHRP 17­72. The NCHRP 17­72 project team initially responded to these questions in late 2019 and provided an update in May 2020. During the 2020 TRB Annual Meeting, the TRB Highway Safety Per­ formance Committee recommended a task force of stakeholders to review the rating system developed by NCHRP 17­72 and suggest possible improvements. The task force included

18 Crash Modification Factors in the Highway Safety Manual: A Review individuals from the TRB Highway Safety Performance Committee and AASHTO HSM: Raul Avelar, Geni Bahar, Michael Dimaiuta, Karen Dixon, Dennis Emidy, Doug Harwood, John Ivan, Jeffrey Pulver, Stephen Read, Karen Scurry, and Derek Troyer. Before the meeting on May 13, 2020, the task force was provided with the following materials: • A memorandum summarizing the goals of the meeting. • An Excel file providing the list of CMFs rated with the NCHRP 17­72 rating system. Each row in this file represents one CMF. The file also included CMFs from the HSM1 that were developed by an expert panel and could not be rated with the NCHRP 17­72 rating system. The file contained two worksheets: “CMF Ratings,” which provides the ratings for each CMF along with other details, and “Column Descriptions,” which describes each column in the CMF Ratings worksheet. • Responses to questions from the NCHRP 17­71 project team (updated May 12, 2020). • Documents describing the NCHRP 17­72 rating system for before–after, cross­sectional, and meta­analysis/meta­regression studies. At the time of the meeting, CMFs with a rating of 100 or greater would be recommended for inclusion in the HSM2. The threshold for a rating of 100 was eight studies. However, this threshold can be changed. • Task 1 and Task 2 reports from NCHRP 17­72. These reports provided a review of CMF rating methods from 2016. • PowerPoint slides to be presented at the May 13, 2020, meeting. 4.2.2 Brief Summary of the May 13, 2020, Meeting with the Task Force Following is a summary of the comments during the meeting: 4.2.2.1 Points Based on Categories In the NCHRP 17­72 ratings system, the points for the number of miles/sites and the number of crashes are based on categories. For example, in a before–after study, the points for the number of miles/sites of the reference/comparison group are based on the following criteria: a. 50 or more miles/sites: 10 points, b. 25 to 49 miles/sites: 5 points, and c. Fewer than 25 miles/sites or cannot be determined: 0 points. Some in the group indicated that instead of assigning points on the basis of categories, it might be better to have a continuous scale—for example, if the number of comparison sites was between 25 and 49, the number of points should be between 5 and 10 instead of 5. 4.2.2.2 Points for “Unknown” Versus “No” For many factors, if information is unknown, the number of points awarded is the same as if the answer were “No.” For example, one of the factors is “accounts for changes in traffic volume during the study period.” If the answer is “Yes,” the factor is given 10 points; if it is “No” or “Unknown,” the factor is given 0 points. Some in the group felt that an answer of “Unknown” should receive between 0 and 10 points; that is, “Unknown” should be given some benefit of the doubt. 4.2.2.3 Ratings for CMFs Pertaining to Pedestrian or Other Rare Crash Types There was a concern that CMFs pertaining to pedestrian or other rare crash types might not receive a high enough rating, because the sample size associated with this crash type tends to

Recommendations for Incorporating CMFs in the HSM and Possible Framework for Practitioner Use 19 be smaller. Although this was considered a legitimate issue, it was also recognized that smaller sample sizes are also associated with less­reliable CMFs. So, the overall recommendation was not to modify the rating system specifically for pedestrian or other rather rare crash types. 4.2.2.4 Weight for Standard Error Some in the group wondered whether standard error should be given more weight than the allotted 20 points. However, others observed that increasing the weight for the standard error would lead to potentially even lower ratings for CMFs pertaining to pedestrian and other rare crash types. So, the overall recommendation was to keep the weight for the standard error the same. 4.2.2.5 CMFs with a Higher Current Star Rating and Lower Points in NCHRP 17-72 System Some CMFs that have a higher rating in the current star rating system got lower points with the NCHRP 17­72 rating system. This was concerning to some in the group. The NCHRP 17­72 project team agreed to investigate these CMFs to ensure that there were no errors. 4.2.2.6 Scores for Individual Categories as well as the Overall Score Some in the group stated that it would be useful to provide the scores (data, methodology, and statistical significance) for the individual categories in addition to the overall score out of 150. Everyone agreed this would be useful, and the NCHRP 17­72 project team agreed that this would be feasible in the CMF Clearinghouse. 4.2.2.7 Cutoff for Inclusion in the HSM2 The NCHRP 17­72 project team originally proposed a score of 100 points out of 150 as a cutoff for a CMF to receive the HSM2 stamp of approval. Since it had been decided that Part D of the HSM2 would not include CMFs but instead include information on identifying, apply­ ing, and developing CMFs, the group did not feel the need for a cutoff for HSM2 inclusion. The group suggested that practitioners should be encouraged to use the CMF with the highest possible star rating but also consider other important factors, such as the applicability of the CMF to the specific facility. 4.2.2.8 Standard Deviation of CMF Versus Standard Error of the CMF Estimate There was some discussion about the distinction between the standard deviation of the CMF versus the standard error of the CMF estimate and how this distinction should be com­ municated to the practitioner. The NCHRP 17­72 project team feels that this information should probably be included in Part D of the HSM. 4.2.3 Discussions Subsequent to the May 13, 2020, Meeting 4.2.3.1 Meeting of Project Team After the meeting, the NCHRP 17­72 project team conducted a phone call primarily to dis­ cuss the issues in Sections 4.2.2.1 and 4.2.2.2, namely, points based on categories of crashes and sites/miles and the distinction between “Unknown” and “No.” To investigate the first issue further, the number of crashes and sites/miles needs to be recorded. This detailed information was recorded by one group of reviewers, but not by the others. To investigate this issue, the NCHRP 17­72 project team needed to review hundreds of studies and thousands of CMFs to

20 Crash Modification Factors in the Highway Safety Manual: A Review extract this information. Similarly, the project team needed to review the studies to distinguish between “Unknown” and “No.” This re­review of the studies was not included in the original scope of work. It was clear that if this re­review needed to be undertaken, the project would not be able to accomplish Task 9 (CMF gap analysis) to the same level of detail, and this decision would have to be made by the panel. The project team had a meeting with the panel for NCHRP 17­72 on June 8, 2020, to discuss this issue and other issues. 4.2.3.2 Meeting with FHWA Clearinghouse Focus Group On May 20, 2020, HSRC and Vanasse Hangen Brustlin conducted a call with the CMF Clearinghouse focus group, which included the following individuals from FHWA and state DOTs: Kelly Campbell, Douglas Cobb, Ana Maria Eigen, Dennis Emidy, Parry Frank, Cory Hopwood, Kohinoor Kar, Jeffrey Pulver, Karen Scurry, Carrie Simpson, Carol Tan, and Derek Troyer. This meeting was conducted as part of FHWA’s HSIS project but is discussed here because it is relevant to NCHRP 17­72. The meeting posed three sets of questions to the par­ ticipants. The questions and the participants’ responses were as follows: Issue 1: Converting the NCHRP 17-72 Rating System to a New Star Rating System for the CMF Clearinghouse. Star ratings for some CMFs would change, and some HSM1 CMFs would not meet the inclusion criteria for the HSM2. What impact would this have on researchers and practitioners? What would be the best way to communicate this change? DOTs are using the current star ratings to allocate their safety funding. There was concern about the impact the new star ratings would have on funding that had already been allocated to projects on the basis of a higher star rating under the current rating system. The changes to the star ratings of specific CMFs should be communicated to the DOTs, perhaps in an email or document to explain why the rating was lower for some CMFs—this could be an educational tool. There was discussion about the possibility of suggesting alternates to CMFs that would see a rating drop under the new rating system (e.g., going from a five­star to a two­star rating). States have their own CMF lists and do not access the CMF Clearinghouse every day. There was concern as to how the change would be reflected on the CMF description page to show what had changed. It was suggested that a decrease in a CMF’s rating under the new system might be identified in the form of a footnote or in the comments. Practitioners need to be informed about the reason for the change in the rating system. Perhaps FHWA could send practitioners a memo and also make this information available as a link from the CMF Clearinghouse. This memo could perhaps indicate that, given the large number of CMFs in the clearinghouse, factors that are critical to the quality of a CMF have been identified. The importance of the applicability of the CMF to the specific type of facility under consideration could also be illustrated—that is, the star rating of the CMF may not always be the most important factor. Some attendees felt that both the old and new star rating systems should be provided for 1 year as a transition. The old star rating could perhaps be shown as text rather than as stars. Issue 2: Consistency Between the HSM2 Inclusion Criteria and the New Star Rating System for the CMF Clearinghouse. Would you prefer a fixed­point cutoff for inclusion in the HSM2? Would you prefer the HSM2 inclusion criteria to be consistent with the star ratings? The attendees were undecided about the need for cutoff for HSM2 inclusion, since the star rating and the actual number of points out of 150 would be available and practitioners could make decisions on the basis of these parameters and the applicability of a CMF. The question of whether the HSM1 inclusion criteria were still relevant was also raised. These concerns suggest

Recommendations for Incorporating CMFs in the HSM and Possible Framework for Practitioner Use 21 that the PDF with HSM2­approved CMFs is not needed. However, this decision would have to be approved by AASHTO. Issue 3: What Other Information Would You Like to See in the CMF Clearinghouse?. How can we improve the functionality of the CMF Clearinghouse? Are there any other com­ ments or concerns that you would like to express about the CMF Clearinghouse? • CMF Clearinghouse search engine could be optimized. If users are not able find the exact wording used in a CMF, it is difficult to identify it. Perhaps the clearinghouse interface could include a pop­up function that would automatically suggest terms similar to the words being entered by the user. The HSM pooled fund is working on a data dictionary that could be a starting point for automatically suggesting similar key words. • Consistency in definitions of crash severity between studies. Depending on the study, injury crashes can sometimes include fatal crashes. These cases are study specific—the crash severities are identified in the CMF Clearinghouse on the KABCO scale. In the case of any anomalies, an explanation is also provided in the comments for the specific CMF. 4.2.4 Sensitivity Analysis of Continuous Rating Scale for the Number of Crashes Used to Develop a CMF Following the May 13, 2020, NCHRP 17­72 stakeholder meeting and the June 8, 2020, NCHRP 17­72 panel meeting, the research team was tasked with conducting a sensitivity analysis of the use of a continuous scale as opposed to a fixed scale for the number of crashes used to develop a CMF. The research team had the crash data recorded for all the before–after studies (the actual number of crashes per mile was only recorded by category and could not be included in the sensitivity analysis). In the panel meeting on June 8, 2020, it was decided that only the before– after studies would be used to conduct the sensitivity analysis, so that the results would then inform whether it would be worthwhile to re­review the cross­sectional studies to record the crash counts to be used to score on a continuous scale. A continuous rating scale was applied to assign scores for (1) the number of crashes in the reference/comparison group (10 points maximum) and (2) the number of crashes in the before period plus the expected number of crashes in the after period in the treatment group (10 points maximum). Tables 1 and 2 show the average number of points from the NCHRP 17­72 rating system by the current star rating in the CMF Clearinghouse as well as the number of CMFs with a score of Before–After Studies with Fixed Scale for Number of Crashes CMF Clearinghouse Star Rating CMFs Rated Using 17-72 Rating System Average Points CMFs with Points > 100 5 205 124.39 175 4 1,544 112.59 1,236 3 2,153 91.61 1,102 2 737 45.38 33 1 262 24.22 5 All 4,901 89.03 2,551 CMFs in HSM 134 100.72 77 Table 1. Average points from the NCHRP 17-72 rating system when a fixed scale is used.

22 Crash Modification Factors in the Highway Safety Manual: A Review 100 points or more. Table 1 shows the results when a fixed scale for the number of crashes was used and Table 2 the results when a continuous scale was used. Use of a continuous scale for rating the number of crashes accounted for a 0.88­point increase in the average ratings for all before–after CMFs (from 89.03 to 89.91), as well as 10 additional CMFs that would score 100 points or more. On the basis of this minor change in before–after ratings, the research team recommended that it would not be worthwhile to re­review the cross­ sectional studies at this point to record the information on the actual crash counts (a re­review of the cross­sectional studies would have taken resources away from Task 9, the CMF gap analysis, one of the important tasks in the project). After further discussion with the panel members and stakeholders, it was decided that the fixed rating scale would continue to be used and that crash numbers would be recorded for all CMFs entered (where available) in the CMF Clearinghouse after the transition in the CMF Clearinghouse rating system. The CMF Clearinghouse used the information developed in NCHRP 17­72 to develop a revised star rating system. Further details are available at https:// www.cmfclearinghouse.org/sqr.cfm. Before–After Studies with Continuous Scale for Number of Crashes CMF Clearinghouse Star Rating CMFs Rated Using 17-72 Rating System Average Points CMFs with Points > 100 5 205 124.89 175 4 1,544 113.22 1,238 3 2,153 92.56 1,110 2 737 46.55 33 1 262 25.18 5 All 4,901 89.91 2,561 CMFs in HSM1 134 101.54 77 Table 2. Average points from the NCHRP 17-72 rating system when a continuous scale is used.