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.
97  The use of CMFs to produce quantitative estimates of the benefits of safety improvements has grown in recent years. The motivation for this research was centered on questions regarding the application and development of CMFs. For example, will a CMF that is developed using data from one region of the country and at sites with specific geometric design and traffic control fea- tures be equally applicable to a different region for slightly different design and control features? If several treatments are applied to the same site, will their combined effectiveness be found to equal the value obtained by multiplying their respective CMF values? Considering these questions, this project produced guidelines for safety practitioners and researchers who work to apply or develop CMFs to accurately estimate the effect of a safety treatment. The guidelines produced under this research include ⢠A procedure for estimating the effect of a proposed treatment on a site of interest through a process that identifies potential CMFs, matches them on major site characteristics, and adjusts or combines them to produce a final CMF to be used at the site of interest ⢠A procedure to estimate the combined safety effect of two treatments at the same location using a method that hinges on the magnitudes of the individual treatment effects and the determination of how much overlap there is between the treatments ⢠Guidelines on developing CMFunctions either from cross-sectional regression models or from a set of CMF point estimates. These guidelines are documented in Appendixes A, B, and C, providing step-by-step instruc- tions for the prescriptive procedures (or a list of topics for the more general guidelines) and example applications of the procedures or guidelines. These appendixes also document the sup- porting research that was conducted under NCHRP Project 17-63 to develop these guidelines. Other appendixes (D through G) provide user manuals for accompanying tools and discussions on directions for future CMF research. The research conducted under Project 17-63 addressed many questions regarding the appli- cation and development of CMFs, but as knowledge grows, so grows the recognition of what is left unknown. The remainder of this section presents ideas for future research that would build on the results of this project. Combined Effect of Multiple Treatments This research has supported the development of guidelines to estimate the combined safety effect of multiple treatments. While ground truth CMFs were developed for both individual and combined treatment effects for 19 scenarios, there is a need for additional testing and devel- opment of the recommended methods. The associated guidelines document in Appendix B C H A P T E R  4 Conclusions and Suggested Research
98 Guidelines for the Development and Application of Crash Modification Factors provides a framework for other researchers to follow to produce CMFs for individual and com- bined treatment effects. As additional information is generated on the individual and combined treatment effects, it will be possible to validate the recommended methods and modify the rec- ommendations as needed. Assessing the 19 scenarios from this research shows that consideration of the standard error can affect decisions about treatment effectiveness. As such, there is a need for additional research to develop more accurate methods to estimate the standard error of the combined treatment effect, particularly when there is overlap among the treatment effects. Based on the results of this study, it is apparent that overlapping effects are common for multiple treatments. It is also apparent that the proposed method for estimating the standard error of the combined treat- ment tends to overestimate the standard error. This method provides a reasonable estimate of the upper bound of the standard error, but it would be more appropriate to account for the over- lapping treatment effects. There may be interest in further exploring whether to implement a single treatment or multiple treatments to address a given safety issue. This question should be answered before estimating the combined treatment effect. In some cases, redundant treatments can help to enhance safety under different conditions. For example, longitudinal pavement markings and post-mounted delineators both provide information to the driver regarding the roadway alignment. While they may seem redundant, the pavement markings are better aligned with the driverâs field of view, and the post-mounted delineators are beneficial when the pavement markings become faded or are obscured by rain or snow. In other cases, redundant treatments may not provide any added value, only added cost. If this is the case, then it may be more appropriate to implement a single treatment and allocate the remaining resources to another location. Enhancing Future CMF Research The field of road safety research has produced much knowledge of the safety effect of infra- structure improvements, but there is potential for better and more reliable results. Appendix F presents a strong argument that continued attempts to extract reliable CMFs by fitting single- equation models to cross-sectional data are unlikely to bring about consensus. It is crucial to accurately attribute an observed safety effect to the cause related to the CMF of interest, and researchers ought to strive to create conditions in which nuisance influences (causes) are mini- mized and well accounted for. To achieve this goal, the following concrete points need to be realized: ⢠Randomized controlled experiments are the engine of progress toward evidence-based prac- tice and ought to be conducted in future CMF research. Appendix F identifies circumstances where this is possible, and other circumstances may exist now and are likely to develop in the future. ⢠Research programs such as NCHRP should develop and test the propinquity study design to approximate randomized trials by minimizing influences other than the cause related to the CMF of interest. Should it offer promise, a research program should be established to widely implement it to exploit available cross-sectional data. ⢠Causal inference methods have seldom been used in road safety research but are common in other disciplines facing a similar reality. A research program should be established to study, examine, and adapt the various causal inference methodologies used in epidemiology, sociol- ogy, and econometrics to evaluate their promise for the estimation of trustworthy CMFs.
Conclusions and Suggested Research 99  ⢠The success of CMF research depends on how well one can predict what would have been the safety of units had they not been manipulated. Many methods for generating predictions are in use, but it is not clear which performs best in what circumstances. A comprehensive program of research should be established to investigate. Closer integration of CMF research and practice will enhance the trustworthiness of research results and promote their use in practice. A way must be found to make research an integral part of operating agencies and break down the barriers separating research and practice.
