Cover Image

Not for Sale



View/Hide Left Panel
Click for next page ( 18


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 17
17 and/or combinations of crash types with other crash-related "treatment strategy," "treatment," and "strategy" are used variables (e.g., crash type by time of day). As further de- interchangeably in the following text. Here, depending on the scribed in Appendix D of NCHRP Report 501 (18), a more emphasis area and strategy being addressed, the term "popu- detailed procedure in which categories within a given variable lation" may refer to humans (e.g., older drivers, pedestrians); (e.g., crash type) can be shown to be "over-represented" is vehicles (e.g., large trucks, motorcycles); or roadway sites provided in the CARE (Critical Analysis Reporting Environ- (i.e., individual roadway features, segments, corridors, ment) (23). In general, and as described in NCHRP Report intersections, and interchanges). High-level safety planning 501, the analyst will conduct a series of "drill down" analyses requires that the limited available safety funds be used in the to gain details on which issues/emphasis areas are most most effective ways. Funds should not be spent on treatments important for his/her jurisdiction. whose effects are small if those same funds could be used for other treatments that would provide greater benefits. Thus, as a minimum, the goal should be to only implement treat- Stage 2 Set a Crash, Injury or Death ments whose benefits exceed their costs, and the ultimate goal Reduction Goal for an Emphasis Area should be to implement the treatments with the highest ratio The emphasis-area team will then use a series of factors to of benefits to costs. However, to base safety planning deci- define a reduction goal for deaths and injuries in each of the sions on benefits and costs requires that the effectiveness of emphasis areas chosen. This decision will be based on outputs each potential treatment be defined (e.g., treatment "X" will of the Stage 1 analyses (i.e., the problem-size, total crash cost, reduce run-off-road crashes by 15 percent or treatment "Y" over-representation, and related outputs of the problem- will reduce older driver crashes by 20 percent). identification/drill-down analyses), on some estimate of the However, a review of any of the NCHRP Report 500 guides possible effects of strategies, and on the budget established for will indicate that there the emphasis area. NCHRP Report 501 (18) defines both an are many treatments Crash Reduction Factor vs. initial process and a revised process for establishing these that have been tried Accident Modification Factor goals. The initial process will most likely be based on "best in some cases used very judgment" of the factors above. The revised process is much widely and are gener- The level of effectiveness of a more iterative and analysis-driven where initial goals are ally considered to have treatment is referred to in much modified based on analyses that indicate what is realistic a positive effect on of the current safety literature as given the nature and size of the problem, the known or safety, but have never a Crash Reduction Factor (CRF) assumed countermeasure effectiveness for the final list of been formally evalu- or Accident Modification Factor chosen countermeasures, and the optimization of the exist- ated in a well-designed (AMF). The two terms are just ing budget either within a given emphasis area or across study from which an different ways of expressing emphasis areas. acceptable quantitative treatment effectiveness levels. A level of effectiveness CRF provides the expected (i.e., a specific CRF or proportional reduction in crash Stage 3 Define Treatment Strategies AMF) has been devel- frequency, for all crashes or for and Target Populations oped. specific crash severity levels, so a Having now defined the issue/emphasis area to be treated For this reason, the 15 percent reduction in crashes (e.g., run-off-road crashes, crashes involving drinking driv- process of choosing would correspond to a CRF = ers) and having defined the crash/injury reduction1 goal (e.g., treatments and choos- 0.15. Likewise, a 20 percent a 10 percent reduction in the number of fatal and serious in- ing targets for each reduction in fatal and serious in- jury run-off-road crashes on two-lane rural roads), the next treatment will be cov- jury crashes would correspond step is to define the treatment strategies to be employed and ered in four different to a CRF = 0.20. An AMF is de- the target population for each strategy. Note that the terms procedures: veloped by subtracting the CRF from 1.00, with an AMF of 1.00 Procedure 1 for representing no effect on safety. 1 Note that the term crash/injury reduction will be used instead of crash application to road- Thus, a treatment with a 15 per- reduction throughout this guide. In most cases, the two terms can be thought of as interchangeable. However, because the ultimate goal of safety programs way-based treat- cent effectiveness would have an is to reduce death and injury, and since this can be accomplished both by ments with a known AMF of 0.85 (i.e., 1.00 - 0.15). reducing crash frequency and by reducing the severity of crashes, the former effectiveness level AMFs above 1.00 indicate that term is considered more appropriate than the latter. In addition, it should be where a complete set the treatment can be expected to understood that the term "injury" includes both fatal and nonfatal injuries. Thus, the crash/injury reduction from a treatment may represent a reduc- of data (e.g., crash, result in an increase in crashes. tion in crash frequency, a reduction in crash severity, or both. roadway inventory,

OCR for page 17
18 intersection inven- In many states, the achievement of a statewide goal for Why Perform the Higher Level tory, traffic counts) crash reduction will involve consideration of treatments both Analysis Shown in Procedure 1? are available for with and without known effectiveness and consideration of Selects treatments better planning purposes roads under both state and local jurisdiction, which are likely More detail on target road- Procedure 2 for ap- to involve crash datasets that are both mileposted and not ways or intersections for each plication to roadway- mileposted. Thus, many states may need to use more than treatment based treatments one of the four procedures. The application of each of these Provides better assurance that with known effec- procedures is described below. established goals can be met tiveness, but where Provides better allocation of only crash data, but Procedure 1 Choosing Roadway-Based limited safety dollars not inventory or traf- Treatments and Target Populations When fic volume data, are Treatment Effectiveness Is Known, and Both Crash available (e.g., for local jurisdictions without inventory data) and Non-Crash Data are Available Procedure 3 for application to both roadway treatments that do not have specified levels of effectiveness As indicated above, the highest level of treatment selection/ and driver-based treatments with or without known targeting analysis represented by Procedure 1 is based on an effectiveness economic analysis procedure. This procedure is applicable to Procedure 4 a hybrid of Procedures 1, 2, and 3 for consid- treatments intended to improve safety on roadway segments or ering treatments with known effectiveness and treatments at intersections. If required data are available or can be devel- without known effectiveness in the same process oped, Procedure 1 will provide the user with best selection of treatments and with the most detail on where the treatments Procedures 1 and 2 are based on an economic (benefit- should be targeted. It will also provide better assurance to the cost) analysis. Procedure 3 is not based on economic analy- user in the planning process that the established crash/injury sis, while Procedure 4 combines economic and non- reduction goal can be met. The increase in the precision of the economic procedures. The selection and targeting of analysis, in comparison to the other procedures discussed driver- and vehicle-based treatments is not based on an below, will result in much better allocation of safety dollars. For economic analysis. The following discussion provides an these reasons, Procedure 1, where feasible, is strongly recom- overview of each of these four procedures. Details of each mended to the user. Detailed examples of the application of of the procedures will be presented in the individual Procedure 1 in which multiple potential treatments and treat- sections concerning specific treatment types. ment targets are examined to produce a final recommended Procedures 1 and 2 generally result in plans whose program are presented at two FHWA web sites: http:// expected safety benefits are more accurately known than Pro- safety.fhwa.dot.gov/roadway_dept/docs/lanedeparture/index. cedures 3 and 4, because Procedures 1 and 2 are applicable to htm and http://safety.fhwa.dot.gov/intersections/intersection treatments whose effectiveness has been well documented. sap.htm (24,25). Therefore, the safety planning process should generally Procedure 1 is suitable for roadway-based treatments that exhaust funding opportunities under Procedures 1 and 2 address specific crash types and are aimed at modifying road- before proceeding to Procedures 3 and 4. way segments or intersections. However, it should be noted Exhibit III-1 will guide the user to the appropriate proce- that Procedure 1 can also be applied to driver- or vehicle- dure for each analysis situation. The key considerations in related treatments if the effectiveness of the treatment is choosing one of the four procedures are whether the effective- known, and if the treatment is to be targeted to specific loca- ness of the treatment(s) is known and whether the available tions on the highway. Examples include red-light-running crash data are "mileposted." enforcement targeted to intersections with high numbers of Inventory Data Available and Linkable to Crashes? Treatment Yes No Effectiveness Mileposted Crashes Mileposted Crashes Unmileposted Crashes Known? Yes Procedure 1 Procedure 2A Procedure 2B No Procedure 3 Procedure 3 Procedure 3 Some known, Procedure 4 Procedure 4 Procedure 4 some unknown Exhibit III-1. Guide to choice of procedures based on knowledge of treatment effectiveness and crash data quality.

OCR for page 17
19 such crashes or truck inspection programs targeted to While the detailed procedure will be presented in the indi- locations or corridors with high numbers of crashes associ- vidual strategy-based sections, the steps include the following: ated with truck speeding or mechanical failure. It should also be noted that, while Procedure 1 is driven by 1. Specify the types/classes of roadway segments that are hard data (e.g., information on costs, benefits, crash counts, potential targets for the treatments. roadway mileage), there are other policy issues that must also 2. Develop critical crash frequencies for each candidate treat- be considered before implementation, including user accept- ment the frequency of target crashes per mile or per ability. The FHWA Sample Plans (24, 25) noted above also intersection that, if treated, will result in crash/injury provide thoughts on such issues. For example, the first plan reductions whose economic benefit will exceed imple- recommends limiting "the installation of shoulder rumble mentation costs by some specified factor. In the examples strips to rural highways where there are no concentrations of provided in the two FHWA Sample Plans (24, 25), the homes due to the noise issue," and conferring "with state or benefit-to-cost ratios used were 2.0 or greater. local pedestrian/bicycle coordinator concerning potential 3. Using the inventory file, stratify potentially treatable application of shoulder rumble strips on bicycle routes." roadway segments by roadway class. Since these are policy decisions, each user will need to decide 4. Link target crashes with roadway segments or intersections what criteria are necessary. However, the point is that careful from the appropriate inventory data file, and then perform consideration of such issues is important. a computer screening of all potential locations (segments To implement Procedure 1, which provides the highest or intersections) to determine which have crash frequen- level analysis of roadway-based treatments, the following data cies that exceed the critical crash frequencies. are required: 5. If performing a roadway segment analysis, correct the output for "treatment gaps" along the same route result- A specified effectiveness level (Crash Reduction Factor or ing from the network-screening computer program. Accident Modification Factor) for each treatment to be 6. Estimate the expected crash injury reductions on all the considered identified target locations. A computerized crash data file which includes sufficient 7. Repeat the above steps for each potential treatment type. crash details to identify crash types that will be affected by 8. Correct for multiple treatments on the same segment. each treatment ("target crashes" such as run-off-road 9. Sum all expected crash injury reductions for all chosen crashes, head-on crashes, run-off-road on curves), and treatment types and chosen target locations and compare which includes crashes for all potential target populations that total to the established goal. Computerized roadway inventory data and/or intersection 10. Add either new treatments or new targets or new inventory data that can be linked to the crash data by loca- approaches (e.g., inclusion of safety treatments in normal tion of the crash maintenance or rehabilitation efforts) until the goal is met. A network screening computer program which will exam- ine each segment of roadway of a given length (e.g., 1 mile) The following flow chart shown in Exhibit III-2 illustrates this or each intersection and calculate the number of target procedure. crashes that have occurred on each segment in the past 3 to A key advantage of having effectiveness measures for each 5 years (see further discussion below) treatment, as is the case in Procedure 1, is that the full set of Computerized traffic count data that is part of the roadway treatments needed to reach the crash/injury reduction goal or intersection inventory data or can be linked to it can be determined. The user can also determine the cost of Unit cost for each treatment both original implementa- implementing those treatments and, thus, determine the cost tion costs and annual maintenance costs of meeting the established crash/injury reduction goal. The cost of meeting the goal should then be compared to the If these data are available, Procedure 1 will lead the user available budget for safety improvement. If the cost of achiev- through a series of steps that will allow the user to choose a ing the goal is within the available budget, the required funds set of treatments and a set of targets (e.g., locations or should be programmed; depending on the size and nature of subpopulations) for each chosen treatment that will both the improvement program needed to reach the goal, the pro- meet the established crash/injury reduction goal, and will be gramming of these treatments may be in a single year or over cost-beneficial at some prescribed level. To do this, the user a multi-year period. If the cost of achieving the goal exceeds must analyze each potential treatment separately (choosing the available budget, the user can request an increased budget targets for each) and sum potential crash injury reductions from agency management, which may in turn inform higher across all treatments to determine if the established goal can political authorities (executive and legislative) of the funding be met. If not, new treatments or strategies should be added. level needed to meet the goal. If the funds required to meet

OCR for page 17
20 The reason for using a network screening program is that crashes are not uniformly distributed along a route; the network screening program identifies locations where crashes of particular types are concentrated, which represent poten- tial improvement locations. Many highway agencies have developed computer programs to screen the highway network for high crash fre- quencies and new software tools to perform such screening are under development. Existing highway agency screening programs typically use a "sliding window" approach in which a "window" of specified length is moved forward along the road in steps of specified length and the crash frequency within the "window" is checked at each step. Such existing network screening programs can be used to implement Procedure 1. A window length of 1 mile is recommended for use with existing network screening programs. A more sophisticated network screening approach will be possible with FHWA's SafetyAnalyst software tools which are currently under development. SafetyAnalyst (see http://www. safetyanalyst.org) uses a new screening approach, known as "peak searching" and a variation of the "sliding window" approach based on improved statistical methods. The "peak searching" algorithm in SafetyAnalyst uses a variable win- dow length based on homogeneous roadway segments. The SafetyAnalyst "sliding window" approach uses window lengths that can be selected by the user; this will allow the use of window lengths shorter than 1 mile and will allow the win- dow length to move forward in steps that are shorter than the window length. Network screening has typically been per- formed by highway agencies in computer database programs, but techniques for performing network screening in a Geo- graphic Information Systems (GIS) environment are being developed. Procedure 2 Choosing Roadway-Based Treatments and Target Populations When Treatment Effectiveness Is Known and Crash Data Are Available, but Detailed Inventory Data Are Not Available Procedure 2 for roadway-based treatments is intended for application on roadway segments or at intersections in "limited-data" situations. Here, crash data are required, but a formal roadway segment or intersection inventory database is not required. Thus, Procedure 2 is intended for application by highway agencies that do not have detailed computerized in- Exhibit III-2. Flow chart illustrating procedure 1. ventory data on the characteristics of each roadway segment and/or intersection. The goal is the same to select a set of the goal are not available, the user should proceed with the treatments with target locations identified for each treatment portion of the program that can be implemented and should that, if treated, will ultimately allow the user to meet the es- inform agency management that the program will proceed as tablished crash/injury reduction goal, and to do so such that far as practical, but that the established goal cannot be met. the economic benefits of the crash-injury reductions exceed

OCR for page 17
21 the treatment implementation costs. The lack of detailed in- Procedure 2A, which applies a "medium level" analysis of ventory data on individual roadway segments or intersections roadway-based treatments, the following data are required: means that estimates concerning treatment effects are not as accurate as in Procedure 1, and that the targets identified are A specified effectiveness level (CRF or AMF) for each treat- not as likely as those in Procedure 1 to provide the intended re- ment to be considered sults. Thus, the user cannot have as much faith in the benefit- A computerized crash data file which includes sufficient to-cost results as was the case with Procedure 1. crash details to isolate target crash types (run-off-road, Procedure 2 has two variations. In the first, designated below head-on crashes, and run-off-road on curves) and potential as Procedure 2A, while an inventory database is not required, target populations that will be affected by each treatment, each crash must be "mileposted" (i.e., given a specific "ad- and which is "mileposted" such that the location of each dress" on a specific route) with sufficient accuracy such that crash is included clusters of crashes can be found and the locations at which A network screening computer program which can read an those clusters occur can be identified as candidate improve- input file composed of target crash records sorted by route ment locations. This is likely to be the case for many state high- and milepost, and can count the number of target crashes way agencies when intersection treatments are being analyzed, within a given specified length (e.g., 1 mile for segment- since most highway agencies will have mileposted all crashes based treatments and 500 ft for intersection treatments) but do not have an "intersection inventory" that provides the that have occurred in the past 3 to 5 years (mileposted) location of each intersection on each route. This Unit cost for each treatment both original implementa- may also be the case for local agencies (e.g., counties, cities, tion costs and annual maintenance costs towns) that can provide a "milepost" or other "address" for their crashes. Currently, more highway agencies have de- If these data are available, Procedure 2A will lead the user veloped roadway-segment inventory data files than have de- through a series of steps that will allow the user to choose a set veloped intersection-inventory data files. Therefore, some of treatments and a set of targets for each chosen treatment highway agencies may find that they can apply Procedure 1 to that will both meet the established crash/injury reduction roadway segment improvements, but that they must apply this goal, and will be cost-beneficial at some prescribed level. As first variation of Procedure 2 to intersection improvements. with Procedure 1, one must analyze each potential treatment The second variation of Procedure 2, designated below as separately (choosing targets for each), implement a procedure Procedure 2B, requires crash data that is addressed to a given which estimates "combined effectiveness" for segments or street or route within the jurisdiction (e.g., a route within a intersections where multiple treatments for the same crash given county), but the crash data does not have to be "mile- types will be applied, correct for "treatment gaps" along the posted" to a specific location on that route. The disadvantage same route if a segment-based program, and sum potential of this procedure as compared to either Procedure 1 or the crash injury reductions across all treatments to determine if "mileposted-crash version" of Procedure 2A is that only full the established goal can be met. If not, new targets or strate- routes within a jurisdiction (e.g., a county or township) can be gies should be added. analyzed. Thus, a route will not pass the benefit-cost screen While the detailed procedure will be presented in the indi- unless there are sufficient crashes (and thus potential vidual strategy-based sections (e.g., see Section IV, "Roadway crash/injury reductions) on the full route to overcome the Segment Programs"), the steps in the process are essentially cost of treating the full route, and the user will not be able to identical to the steps in Procedure 1, with the exception that determine which segments of a given route would produce the network screening program is used somewhat differently. results that have a higher B/C ratio. The steps include the following: Since Procedure 1 is likely to provide better and more accurate safety improvement plans than either of the versions 1. Develop critical crash frequencies for each candidate treat- of Procedure 2, it is important to the future expansion of ment the frequency of target crashes per mile or per inter- highway safety programs that more agencies move toward the section that, if treated, will result in crash/injury reductions development of roadway segment and intersection inventory whose economic benefit will exceed implementation costs data files. Since the information on safety improvements pro- by some specified factor. vided by Procedure 2B will not be as accurate or as detailed as 2. Sort crashes by route/milepost in ascending order, and that for Procedure 2A, it is also important that agencies move then perform a computer screening of all segments on toward a system where all crashes are "mileposted." all routes that are potential treatment locations to determine which segments have target crash frequencies Procedure 2A For Mileposted Crashes. To implement that exceed the critical crash frequencies calculated in the mileposted-crashes version of Procedure 2, designated as Step 1.

OCR for page 17
22 3. If performing a segment-based program, correct the out- separately (choosing targets for each), implement a procedure put for "treatment gaps" along the same route resulting which estimates "combined effectiveness" for segments or from the network screening computer program. intersections where multiple treatments for the same crash 4. Estimate the expected crash injury reductions on all the types will be applied, and sum potential crash injury reduc- identified target locations. tions across all treatments to determine if the established goal 5. Repeat the above steps for each potential treatment type. can be met. If not, new targets or strategies should be added. 6. Correct for multiple treatments on the same segment or at While the detailed procedure will be presented in the indi- the same intersection. vidual strategy-based sections, the steps include the following: 7. Sum all expected crash/injury reductions for all chosen treatment types and chosen target locations and compare 1. Develop critical crash frequencies for a candidate treat- that total to the established crash/injury reduction goal. ment the frequency of target crashes per mile or per 8. Add either new treatments or new targets or new ap- intersection that, if treated, will result in crash-injury proaches (e.g., inclusion of safety treatments in normal reductions whose economic benefit will exceed imple- maintenance or rehabilitation efforts) until the crash/ mentation costs by some factor. injury reduction goal is met. 2. Link target crashes with each route or intersection in each jurisdiction (but not to a specific point on the route). This Procedure 2B For Unmileposted Crashes. To imple- will require computer sorting of crashes by each named ment the version of Procedure 2 without mileposted crashes, route or at each named intersection. Some manual effort which applies a "medium level" analysis of roadway-based will be required to correct misspelled names and to group treatments, the following data are required: routes or streets that have multiple names. 3. Count target crashes for each route or intersection and A specified effectiveness level (CRF or AMF) for each treat- enter the total counts into a spreadsheet (one route or ment to be considered. intersection per row), along with the route mileage for A computerized crash data file which includes sufficient that route (if using a segment-based program). Repeat crash details to identify crash types that will be affected by this step for each route and/or intersection in the juris- each treatment ("target crashes" such as unsignalized in- diction under study. tersection crashes, run-off-road crashes, head-on crashes, 4. Use the spreadsheet to calculate annual crash frequencies and run-off-road on curves), and which includes crashes for each potential route or intersection. for all potential target roadways. Each crash record must 5. Define possible target routes or intersections by deter- contain a county or jurisdiction name where the crash oc- mining which have calculated annual frequencies that curred, and a "route/road on" variable the name of the exceed the developed critical crash frequencies. route or road where the crash occurred. If intersection 6. Estimate the expected crash injury reductions on all the treatments are being considered and the crash record con- chosen target locations by adding up target crashes and sistently includes the name of the crossing roadway, then it multiplying by the treatment effectiveness level. may be possible to treat the intersection crashes as mile- 7. Repeat the above steps for each potential treatment type. posted, and Procedure 2A may be used, even though the 8. Correct for multiple treatments on the same route or at "mileposts" are not numerical. However, in this situation, the same intersection. it may not be possible to consider intersection-related 9. Sum all expected crash injury reductions for all chosen crashes that occur on the intersection approaches at some treatment types and chosen target locations and compare distance from the intersection. that total to the established goal. "Route length" information that will provide the length in 10. Add either new treatments, new targets or new approaches miles of each road or route within a county that is a potential (e.g., inclusion of safety treatments in normal mainte- target for any treatment, or at least the approximate length. nance or rehabilitation efforts) until the goal is met. Unit cost for each treatment both original implementa- tion costs and annual maintenance costs. Procedure 3 Choosing Driver, Vehicle or Roadway Treatments and Target Populations If these data are available, Procedure 2B will lead the user When Treatment Effectiveness in Terms of through a series of steps that will allow the user to choose a set Crash/Injury Reduction Is Not Known of treatments and a set of targets for each chosen treatment that will both meet the established crash/injury reduction The two procedures described above allow the user to goal, and will be cost-beneficial at some prescribed level. As choose treatments and treatment targets for a given problem with Procedure 1, one must analyze each potential treatment while ensuring that the economic value of the crash-injury

OCR for page 17
23 reductions will exceed the cost of implementing the program. measures that may be related to safety (e.g., behavioral Both procedures require that the treatments being examined changes that may be related to crash/injury reduction). The each have a known level of effectiveness expressed in terms of treatments classified as experimental are treatments that have an expected crash/injury reduction a defined CRF or AMF. the potential to be beneficial, but have had only limited use Unfortunately, many of the roadway-oriented treatments and cannot be considered either proven or extensively tried. and many of the non-roadway-oriented treatments (i.e., As stated in each of the NCHRP Report 500 guides, these driver- and vehicle-oriented strategies) in the NCHRP Report experimental treatments should be used with care, only in 500 guide series do not have defined levels of effectiveness. pilot programs, and should be evaluated carefully before wide- Thus, economic analyses like those that are the basis for spread implementation. The Sample Plans presented at the Procedures 1 and 2 are not possible for these treatments. FHWA web sites (24, 25) include example plans for such pilot Despite the inability to perform formal economic analyses, programs. Thus, the process described below is intended for the user can still make an educated choice of which treat- application only to proven and tried treatments. ments will be most effective in their jurisdiction, and can de- Procedure 3 is intended for application to tried or experi- velop a targeting strategy for the treatment in cases where it mental treatments for which the analyst has decided that there is not to be applied jurisdiction-wide (e.g., where specific is likely to be a crash/injury reduction benefit, but for which counties, communities, roads, or driver groups are to be tar- the analyst does not have sufficient evidence to estimate a spe- geted). In general, the choice of treatment will be based on cific CRF or AMF value. In the event that the analyst is able to "what is likely to work best for the target group" and the estimate a specific CRF or AMF value, even if the estimate is choice of targets will be based on identification of locations only an approximation, it is recommended that Procedure 4 "where the crash/injury problem of interest is found." From be used rather than Procedure 3. the description of Procedure 3 presented below, it should be Unlike Procedures 1 and 2, when crash/injury reduction apparent to the user that this procedure could produce results effectiveness of a treatment is not known, guidance on how that are far less likely to fulfill the established safety goal than to choose among several potential treatments and how to tar- the results of Procedures 1 and 2. Nevertheless, when data to get the treatments selected is much more general in nature, use Procedures 1 and 2 are lacking, Procedure 3 is the best and will require "best judgment" on the part of the user. The available substitute. The obvious drawbacks of Procedure 3 following general approach to Procedure 3 is suggested for a highlight the need for better data systems to implement single issue or emphasis area (e.g., head-on crashes on two- analysis procedures that are expected to be more accurate. lane roads or older driver crashes), but can be repeated for The analysis of treatments without documented crash/ multiple emphasis areas. injury reduction effectiveness should focus on those treat- Procedure 3 should be applied in two steps. First, choose ments for which the user decides that there is some evidence of the "best treatments" (i.e., the treatments most likely to be a crash/injury reduction benefit, even if a documented CRF or applicable in a given jurisdiction) from among the set of all AMF based on research using crash data is not available. The treatments presented in the NCHRP Report 500 guide appli- NCHRP Report 500 guide series classifies treatments into three cable to the emphasis area in question. Second, choose the types proven, tried, and experimental based on whether target locations or populations to which the selected treat- credible evaluations based on crash data have been performed ments should be applied. The choice of the "best treatments" (i.e., a CRF or AMF has been defined). The proven treatments from a listing of many potential treatments can be based on are those that have been shown through credible evaluations the following factors: to be effective in reducing crashes. However, not all treatments classified as proven will have a CRF, since some evaluations a. Which of the many potential treatments is judged to be the are based on outcomes that are related to crash/injury reduc- most effective? Even though the exact crash/injury reduc- tion but do not actually have a measured and quantified value tion effects of each potential treatment are not known, for the expected reduction in crash/injury frequency. For ex- there may be some knowledge of which treatments are ample, the effect of occupant restraint laws is stated in terms more likely to be the most effective. To develop an imple- of "restraint use increase" rather than crash-injury reduction. mentation plan in the absence of treatment effectiveness The treatments classified as tried include those treatments that estimates, the user must exercise judgment about which have been used by agencies (in some cases used often), where treatments are best suited to a particular problem. The there is little possibility of negative impacts on crash/injury detailed information in the NCHRP Report 500 guide frequency, and where there is an expectation (but not scien- series, particularly in the section of each strategy discussion tific proof) that the effect of the treatment on safety is likely to labeled "Expected Effectiveness," will provide the user with be a positive one. The evidence could include poorly designed a summary of current knowledge on which judgment can or executed crash/injury evaluations and indirect or surrogate be based. The knowledge in the NCHRP Report 500 guides

OCR for page 17
24 may be supplemented with the past experience of the This is related to Factor b in the above list. Here, for agency with the treatments under consideration. There are the roadway-departure area, the issue is whether to treat also some general approaches that can be based on past run-off-road, head-on, tree-related or other roadway examinations of different strategy types over time. For segment crash types, and on which roadway systems example, for programs aimed at drivers and vehicles, (e.g., two-lane rural roads, four-lane divided roads, free- treatments that incorporate legal sanctions will usually be ways). For driver-related programs, the issue might be more effective than education, and education linked with whether to treat younger or older drivers, or even which enforcement will usually be more effective than either crash types affecting older drivers to target. Whatever the alone. For roadway-based treatments, the best of multiple issue, the prioritization will be based on the frequency choices without known effectiveness will often be the treat- and severity of the specific types of target crashes occur- ment that can potentially affect the crash types and sever- ring in a user's jurisdiction. For each crash type, the user ity levels that are most prevalent at the locations or on the could begin the process by analyzing 3 to 5 years of crash portions of the roadway system under consideration for data to determine the frequency of each type. However, treatment (see factor "b" below). since some crash types are more severe than others (e.g., b. The relative magnitude of the crash types and severity levels head-on crashes are more severe than run-off-road that the treatment will affect. In general, the user should focus crashes), total crash frequency alone does not provide on treatments that could affect the largest portion of the ex- the complete answer. While an alternative is to restrict isting crash experience at the locations or on the portion of the analysis to only fatal and serious-injury crashes, this the roadway system under consideration for treatment. The will severely limit the crash sample, and will also omit a use of a "drill down" approach to crash data analysis to large component of the crash problem non-serious in- determine as many specifics of crash type and severity dis- jury and no-injury crashes. A better solution is to weight tributions related to the issue under study is recommended. each crash by an economic cost based on its type and Thus, if one is attempting to reduce older driver crashes, severity, and then accumulate the total crash cost within then a preliminary analysis should be performed using each target crash type. Information on economic cost per available crash data to determine the frequencies and pro- severity level within 22 different crash types can be found portions of different types of crashes involving older drivers. in Crash Cost Estimates by Maximum Police-Reported In- Then, treatments that target the most prevalent crash types jury Severity Within Selected Crash Geometries (22). This should be selected. Problem analysis of this type is an analysis of crash costs will provide the user with overall iterative process; the results of one analysis of crash data may information on which lane-departure crash type has the suggest another. A sequence of analyses may eventually lead greatest economic impact. to the identification of a treatment or a set of treatments that It is noted that the 22 crash types covered in this FHWA appear to address the most prevalent crash types. It cannot report are types defined by what is hit (vehicle-pedestrian, be ensured that the treatments that address the most preva- vehicle-vehicle, vehicle-object, vehicle rollover) and the lent crash types will, in fact, produce the largest crash nature of the impact (angle, rear-end, head-on, etc.); the reductions, but in the absence of treatment effectiveness type of crash location (intersection or non-intersection); estimates, a focus on treatments that address the most and two speed limit groups ( 45 mph or 50 mph). The prevalent crash types is a logical choice. types do not include crashes for specific vehicle types (e.g., c. The cost of the potential treatments per target unit (either per truck and motorcycle crashes), specific driver populations person or per mile or per intersection). (older vs. younger drivers), or for certain crash situations d. Other technical or policy considerations there may be reasons (e.g., crashes in work zones). However, since the eco- you can't implement a potential strategy in your jurisdiction, nomic costs are derived for each level of crash severity, the even if it's potentially the "best" (e.g., some jurisdictions may user can develop a weighted per-crash cost for any crash not allow the removal of roadside trees, or some treatments scenario (e.g., work zone crashes, truck-car crashes, older require enabling legislation such that municipalities cannot driver crashes) by using the severity distribution of that use the treatment without authorization under state law). crash scenario in their own data sample. The user may further refine this analysis by examining These factors must be combined in some fashion to first crash frequency or total crash cost within roadway classes. decide which treatment to choose. While there are multiple If the crash data are mileposted and linkable inventory data ways of making this choice, the following represents one such are available, details of roadway types can be linked to each procedure. crash record (e.g., number of lanes by divided/undivided). If inventory data are not available, there may be variables 1. Prioritize the specific crash types (within your chosen on the crash record itself that can be used in a less-detailed emphasis area) to be addressed. analysis (e.g., number of lanes, rural vs. urban, route type).

OCR for page 17
25 This analysis will then produce a listing of potentially treatment and other technical and policy considerations. treatable crash types (perhaps by road class) that can be Unfortunately, there are no good guidelines for how to sorted by crash frequency or total crash cost, thus "weight" the different factors. While problem size (total providing a ranked listing. For the higher-ranked crash crash cost) and assumed treatment effectiveness are key fac- types, the user can then conduct additional analyses to tors, there may be technical, policy, and cost considerations determine more of the specifics of the crash circum- that will remove certain treatments from consideration even stances (e.g., nighttime vs. daytime distributions of total if they are felt to be effective. The user will have to choose the crash costs). These additional "drill-down" analyses final treatments based on best judgment. The procedure should be designed to provide additional information outlined above will at least ensure that the major factors in that could lead to the choice of one treatment over the decision are clearly defined. The output of this step will another (e.g., raised pavement markers are primarily be one or more chosen treatments, with the nature of the effective at night or in rainy weather). treatment defining the specific crash types more likely to be 2. Identify possible treatments for use for each high- affected (e.g., raised delineators will affect run-off-road priority crash type. crashes at night). The user will then review the pertinent NCHRP Report The user should be able to work backwards using the 500 guides and list treatments that would be most number of crashes likely to be affected by a given treat- appropriate for each of the high-priority crash types iden- ment and the cost of applying that treatment to a given tified in the above step. The choice should be limited to population or location (see items b and c described at the those treatment strategies that are classified as tried in the beginning of this procedure) to determine the treatment guides. (Proven treatments have known effectiveness lev- effectiveness needed to maintain a cost-benefit ratio els and can be analyzed in one of the three procedures greater than or equal to one. above.) If not already conducted in the "drill-down" analysis in the preceding step, more specific information B 1.0 Ct on the crash costs related to each potential treatment strat- egy could be developed by specifying the crash types that Where: are most likely to be affected by each strategy (e.g., night- B = economic benefit of applying a selected treat- time run-off-road-right crashes for raised pavement ment to a given location or population markers), producing crash frequencies for each specified Ct = the cost of applying that treatment to the crash type, and multiplying the frequencies by cost per selected location or population crash. For some strategies, the NCHRP Report 500 guide presents information concerning which crash types are B = N*Cc*Eff most likely affected by that treatment strategy. Where: 3. Rate the possible treatments based on estimated N = Number of target crashes for the subpopula- effectiveness. tion or location where the treatment is to be Since this procedure deals with treatment strategies with applied unknown effectiveness, this appears to be impossible. How- Cc = average economic cost per target crash ever, for a given set of possible treatments for a particular Eff = treatment effectiveness, or the percent reduc- crash type/road class combination, it may be feasible to make a judgment concerning which treatment strategy tion in target crashes would be expected to be most effective. For example, for Since different severity levels have different crash costs, the run-off-road crashes on two-lane rural roads, one would as- value used for Cc can be a weighted average of the crash sume that rumble strips on two-lane rural roads would be costs associated with the crash types likely to be affected. more effective than wider edge lines or raised delineators. Solving for the treatment effectiveness, the equation reads: For alcohol-related crashes or occupant-restraint-related crashes, treatments involving enforcement coupled with Ct Eff = public information are more effective than either approach N * Cc by itself. At times, this will clearly be a very difficult judg- ment to make. The analyst can then determine whether the calculated 4. Choose "best" treatment(s) by considering estimated treatment effectiveness required to reach the breakeven effectiveness, unit cost and other technical and policy point is likely to be achievable. considerations. 5. Target the chosen treatments to the roadway segments The user will then combine the output of the steps above or subpopulations where the problem is found. with at least two other factors in making a final decision on Since this procedure concerns treatment strategies with- which treatment(s) to implement the cost per mile of the out known effectiveness, it will not be possible to target the

OCR for page 17
26 treatments based on any type of economic analysis such as measures are not available (tried treatments) should then be those in Procedures 1, 2A and 2B. Instead, the treatment considered. Experimental treatments may have a modest role will be targeted to roadway segments, intersections, or in a safety improvement program, particularly if the program vehicle or driver subpopulations showing the highest total is structured to evaluate the effectiveness of the experimental crash cost or frequency, coupled with user judgment con- procedure. cerning other characteristics of the potential target groups The recommended planning approach in this situation is a (e.g., the nature of the roadway and roadside at potential hybrid of Procedures 1, 2, and 3 described above and includes target locations, driver subpopulations most likely to be the following steps described below: reached by the treatment), and technical and political issues. More guidance on targeting using Procedure 3 is 1. Determine if proven treatments can meet the established included in the individual sections that follow. goal. Consider treatments with known effectiveness meas- 6. Decide what to do with multiple treatments on the same ures using either Procedure 1, 2A, or 2B, as appropriate, segments/routes or subpopulations. depending on the types of data available. Determine the The above steps could possibly produce roadway loca- crash/injury reduction achieved and compare it to the tions, intersections, and vehicle or driver subpopulations established crash/injury reduction goal. If the goal has not within a jurisdiction that could be treated with multiple yet been met, proceed to Step 2. treatments. Unlike the earlier procedures where it is possi- 2. Consider tried treatments to supplement the proven ones. ble to estimate combined effectiveness for multiple treat- Consider treatments without known effectiveness measures ments on the same segments or routes, since treatment ef- that have been used extensively by highway or driver/ fectiveness is not known here, the user will have to use other vehicle agencies. If effectiveness measures for these treat- factors in the final treatment choice for these locations. ments can be estimated based on imperfect information, Again, more guidance is given in the sections that follow. then proceed to Step 3; otherwise, proceed to Step 4. 7. Add new treatments, new targets or new approaches 3. Estimate the effectiveness of tried treatments if possible, and (e.g., inclusion of safety treatments in normal mainte- analyze them using the appropriate procedure above. This nance or rehabilitation efforts) until the available step involves attempting to estimate the effectiveness of funding is used. treatments without known CRFs or AMFs. Note that In Procedures 1, 2A, and 2B, an iterative process is used estimating treatment effectiveness is very difficult and until sufficient treatment types and locations are selected can lead to poor treatment choices unless the estimates such that the established crash reduction goal can be are realistic. This estimation was not suggested in Proce- reached. In Procedure 3, without effectiveness measures for dure 3 for this reason. It is only suggested at this point the treatments, it is not possible to verify whether or not a since the user has already considered all proven treatments specific set of treatment types and treatments will meet the before reaching this stage. It is suggested that the follow- established goal. Therefore, the best that can be done is to ing guidelines be used in making such estimates: proceed in selecting treatment types and treatments until the a) In general, be as conservative as possible. Very few available budget for safety improvement has been fully com- treatments can be expected to affect crash frequency by mitted. The total benefit of the selected program will not be more than 15 to 25 percent. forecastable, but the success of the program can be deter- b) When possible, formulate an effectiveness estimate mined by evaluations conducted after its implementation. that is applicable to particular target crash types only, not to total crashes. c) Base estimates for tried treatments on CRFs for similar Procedure 4 Choosing Treatments and treatments if they exist. For example, a CRF exists for Target Populations in Emphasis Areas for shoulder rumble strips. Other treatments that also try to which Some Candidate Treatments Have keep the driver from leaving the roadway by alerting him Known Effectiveness Estimates and (e.g., enhanced edgeline marking, raised profile mark- Other Treatments Do Not ing) but do not give the same level of warning would be In many situations, users considering a safety improvement expected to have somewhat similar, but lower, CRFs. program in a particular emphasis area will need to consider Once effectiveness is estimated, apply Procedure 1, 2A, or both treatments that have known effectiveness measures and 2B as appropriate, depending on the types of data available. treatments that do not. In this situation, it is recommended Determine the crash/injury reduction achieved in Steps 1 that the user give priority to treatments that have known and 3 combined and compare it to the established crash/ effectiveness measures (proven treatments). Treatments injury reduction goal. If the goal has not yet been met, that have been used extensively but for which effectiveness proceed to Step 4.