In road safety, treatment of crash locations (black spots) is typically addressed with a single treatment for the most common crash (accident) type. For instance, if there are multiple cross intersection crashes, you install traffic signals (an oversimplification). There is a crash reduction factor (CRF) for this (or CMF in some countries).
My question relates to treating a given crash location with multiple treatments to increase effectiveness (???). For the example above, it could be installation of traffic signals, high-level street lighting and reducing the speed limit. CRFs are available for each treatment type in isolation. How does one confidently estimate the overall crash reduction factor for such multiple treatment? Is there any robust research on such method?
We have published on this subject previously, but the method is very simple and not adequate for many situations.
Hi Chris, very good question as usually :-) There is a discussion going on in this field, starting from Elvik (2009) - https://www.researchgate.net/publication/26305766_An_exploratory_analysis_of_models_for_estimating_the_combined_effects_of_road_safety_measures_Accid_Anal_Prev
Some summary has been provided in OECD report Sharing Road Safety (2012) - http://www.oecd.org/publications/sharing-road-safety-9789282103760-en.htm - or in FHWA webinar in 2013 - http://www.cmfclearinghouse.org/collateral/Estimating_Combined_Effect_of_Multiple_Treatments.pdf
I am sure you know about the multiplicative method - and its limitations. It is still under research - a NCHRP project, finishing soon, will hopefully give some answers (http://apps.trb.org/cmsfeed/TRBNetProjectDisplay.asp?ProjectID=3421).
Article An exploratory analysis of models for estimating the combine...
Many thanks Jiri. Great answers as usual. I'm off to Indonesia in 24 hrs. I'll pass your detailed pointers to one of our researchers to follow up while I'm away.
We'll be seeking funding to review international literature on this, and if possible, propose a framework for practitioners. Might even do some analysis as we have access to a black spot program data set.
Could be worth asking the guys on NCHRP project, which finishes in a month, how far they have come and what they have achieved. Their objective number 2 goes "guidelines for how existing and future CMFs can be combined in a single location with multiple treatments".
There is also another report online that addresses this issue, including reviews of various methods for combining multiple CMFs:
www.cmfclearinghouse.org/collateral/Combining_Multiple_CMFs_Final.pdf
I'm in no way an expert on statistical analysis of crash data but it seems none of the post hoc methods of combining CMFs are satisfactory. There is no a priori way to know how different treatments might interact, and the only way to begin to know is to use interaction terms in the regression analysis studies underlying the development of CMFs. With respect to lighting, with which I am most familiar, just using multiple regression modeling (e.g., http://dx.doi.org/10.1016/j.ssci.2010.06.008) to include not only lighting but other safety-related factors that might be associated with the presence of lighting (markings, traffic signals, medians, speed limits, etc.) is a fairly recent and still not universally used analysis method (e.g., https://trid.trb.org/view/2015/C/1338722). Modeling that includes interactions among various treatments will ne needed to move toward the answers to this important question. Hopefully, the data sets to allow this type of modeling are or will be available.
Chris
I believe that causative factors (such as poor visibility, wet surface, trees close to edge line, human error etc ) need to be identified from the crash analysis. Based on these factors, options can be developed and compared using Cost Benefit Analysis and Value Engineering. Design should allow for errant driver. The community should be consulted prior to implementing treatments. Effectiveness can be verified by comparing crashes before and after the treatments.
Best regards
Ravi
Dear friends,
Apart from the interesting aspects seen, I share almost everything, in this case, I suggest you review the "Interactive effects" announced by Evans (Evans, L. (1985) Traffic Safety. Science Serving Society. Bloomfield Hills, Michigan. P. 347-353), especially the feedback parameter which characterizes the degree to which users respond to the safety change. This behavioural adaptation was recently supported by Elvik (Elvik, R. (2013). Impact of Behavioural Adaptation and Cost-Benefit Analyses, In Christina Rudin-Brown and Samantha Jamson (2013). Behavioural adaptation and road safety: Theory, evidence, and action. Chapter 20. Taylor & Francis, Boca Raton) as a conceptual framework for the treatment of behavioural adaptation in cost-benefit analyses.
- Badges
- Science topic