Abstract
This study evaluated the safety effectiveness of the systemic implementation of centerline rumble strips on New York State-owned roads. In New York State, rumble strips installed on the centerline are referred to as centerline audible roadway delineators (CARDs). The effect of CARDs on head-on sideswipe (HO/SS), run-off-road (ROR), injury crashes, and fatal crashes was analyzed. This was an observational, naïve before–after with comparison group study. The treatment group included all segments where CARDs were installed in 2013, 2014, and 2015 in New York State Department of Transportation Regions 1 and 4, and Dutchess County in Region 8. Eligible segments without CARDs in those regions constituted the control group. The crashes comprised 2009 to 2017 HO/SS and ROR fatal, all injury, and property damage only crashes. The study showed that CARDs are an effective treatment to reduce ROR and HO/SS crashes. There was a decrease of 24 ROR and 9 HO/SS crashes on CARDs segments installed in 2013, a decrease of 12 ROR and 27 HO/SS crashes on CARDs segments installed in 2014, and a decrease of 40 ROR and 35 HO/SS crashes on CARDs segments installed in 2015. Crash modification factors (CMFs) developed for CARDs included ROR crashes (a CMF of 0.774 [expected reduction of 22.6%] in all ROR crashes); HO/SS crashes (a CMF of 0.623 [expected reduction of 37.7%] in HO/SS total crashes); and HO/SS injury crashes (a CMF of 0.572 [expected reduction of 42.8%] in injury HO/SS crashes).
There were 985 total fatalities because of motor vehicle crashes in 2017 in New York State (data obtained from New York State Department of Transportation [NYSDOT] Safety Information Management System). Of those fatalities, 266 were a result of lane departure crashes (27%). The FHWA roadway departure website defines a lane departure crash as “a crash which occurs after a vehicle crosses an edge line or a centerline, or otherwise leaves the traveled way” ( 1 ).
In New York State, rumble strips installed on the centerline are referred to as centerline audible roadway delineators (CARDs). CARDs are a countermeasure that is used to reduce run-off-road (ROR) and head-on sideswipe (HO/SS) crashes. CARDs create noise and vibration inside the vehicle as the driver crosses the centerline. The noise is often loud enough to alert a distracted or drowsy driver who can then return the vehicle to the traveled way ( 1 ).
The goal of this study was to evaluate the effectiveness of CARDs via the quantification of a crash modification factor (CMF).
Literature Review
There have been multiple similar studies conducted to evaluate the effectiveness of rumble strips and the following paragraphs summarize the results of a subset of these studies. Persaud et al. reviewed 210 mi of two-lane rural roads at 98 treatment sites in seven states using an empirical Bayes (EB) before–after analysis ( 4 ). Persaud et al. concluded that centerline rumble strips reduce overall crashes by 14% and head-on and opposite-direction sideswipe crashes by 21%. The analysis showed that centerline rumble strips reduce all injury crashes by 15% and all injury head-on and opposite-direction sideswipe crashes by 25% ( 4 ). According to Harkey et al., the level of predictive certainty of the CMFs reported by Persaud et al. is “medium-high” ( 5 ). Harkey et al. stated that the estimated safety effects only apply to rural two-lane roads and not to other types of roads.
Torbic et al. identified 20 states that had conducted studies to analyze the effectiveness of CARDs in reducing crashes. Data from these studies showed head-on crashes were reduced from 34% to 95% owing to CARDs. The simple average percent reduction in head-on crashes from these studies was 65% ( 6 ).
Torbic et al. conducted an EB study on 627 mi of treatment sites and 439 mi of nontreatment sites across three states. On urban two-lane roads, the study showed a 40% reduction in total head-on and sideswipe opposite-direction crashes and a 64% reduction in fatal and injury target crashes owing to centerline rumble strips ( 6 ). For rural two-lane roads, the combined results from this study and that by Persaud et al. show that centerline rumble strips were responsible for a 9% reduction in total crashes, a 12% reduction in fatal and injury crashes, a 30% reduction in total head-on and sideswipe opposite-direction crashes, and a 44% reduction in fatal and injury head-on and sideswipe opposite-direction crashes ( 6 ). All results were significant at the 95% confidence level.
Sayed et al. used an EB technique to review the crash patterns of 14 undivided, rural two-lane arterials that contained centerline rumble strips (303.1 mi), and 19 comparison group sites. Arterials with centerline rumble strips showed a statistically significant reduction of 29.3% in off-road left and head-on collisions combined, which was significant at the 95% level ( 7 ).
Outcalt compared traffic records on 17 mi of CARDs in Colorado for two 44-month periods: before the installation of CARDs from July 1, 1992 to March 1, 1996 and after the installation of CARDs from July 1, 1996 to March 1, 2000. The study determined head-on crashes decreased from 18 to 14, and sideswipe opposite direct crashes decreased from 24 to 18 despite a considerable increase in average daily traffic ( 8 ).
Datta and Savolainen conducted a before–after safety performance analysis on 4,214 mi of two-lane high-speed state roads that had CARDs treatments applied during the years 2008 to 2010. The analysis revealed a total of 2,488 “before” and 1,306 “after” target crashes. The crash analysis indicated statistically significant reductions at the 95% confidence level in all target crashes, including a 51% reduction in head-on crashes, a 56% decrease in sideswipe opposite crashes, and a 46% reduction in single vehicle ROR crashes ( 9 ).
Guin et al. conducted an EB analysis on 126.46 mi of CARDs deployed in 2015 and 2016. The reference crash data were all crashes along rural, undivided two-lane highways in Georgia from 2003 to 2008. The study used a seasonally adjusted 24-month predeployment (before) period (June 1, 2013 to May 31, 2005) and a 24-month postdeployment (after) period for comparison (June 1, 2016 to May 31, 2008). A CMF of 0.58 was calculated for total crashes ( 10 ).
Review of Current Practices in New York State
Identification of sites in New York State for rumble strip installation is based on meeting a set of criteria established by NYSDOT. The following are the NYSDOT Engineering Instruction 13-021 ( 11 ) criteria for installation of CARDs:
Pavement: Pavement surface score is 7 or better.
Median: There is no raised median, two-way left-turn lane, or median barrier. CARDs are appropriate for flush medians.
Length: The total quantity of CARDs in a project is 1,500 ft (500 m) or more. Because of the cost of mobilizing the equipment to mill in the CARDs, projects that would result in the total placement of less than 1,500 ft (500 m) may be exempted.
Speed: The posted speed is 45 mph or greater.
Volume: A current annual average daily traffic (AADT) of 2,000 vehicles per day or more.
Roadway width: The combined width of the lane(s) and shoulder, in each direction, must be at least 13 ft. (3.9 m).
More details are provided in NYSDOT’s US Customary Standard Sheet 649-03 ( 12 ).
Data Collection
Crashes
Crash data were collected from 2009 to 2017 by type (HO/SS and ROR) and by severity (fatal, all injury, and PDO) crashes. NYSDOT’s Accident Location Information System (ALIS) was used to obtain crash locations for the data used in this study. Target crashes were HO/SS and ROR crashes where the vehicle crosses the centerline because those are the crashes that could be affected by CARDs. Definitions for HO/SS and ROR crash types are presented in Table 1. Certain crash types were excluded from the study as they were not considered relevant to the types of crashes affected by CARDs. Those crashes included crashes that occur at intersections or in a parking lot, crashes resulting from mechanical problems, and crashes from animal action. Temporary state police crash records and nonreportable crashes were also excluded from the study. Police reports (officers’ notes and crash diagrams) for 14,025 ROR and 2,050 HO/SS crashes were reviewed in this study.
Table 2 shows the original number of crashes, the HO/SS crashes that were removed from the study, and the remaining crashes.
Criteria Used to Determine Target Crashes
Head-on Sideswipe (HO/SS) and Run-Off-Road (ROR) Crashes Included in the Study
Note: NYSDOT = New York State Department of Transportation.
CARDs Segments
The CARDs segments included the CARDs installed in NYSDOT Regions 1 and 4, and Dutchess County Region 8 in 2013, 2014 and 2015. These regions were selected to represent the upstate, downstate, and capital regions of New York State.
The CARDs segments from all regions are provided in Excel spreadsheets to the NYSDOT Main Office on an annual basis by the NYSDOT Safety Evaluation Engineers (SEEs). The data provided were imported into geographic information systems (GIS) and validated against 2017 Fugro data. Fugro (Leidschendam, the Netherlands) is a company that provides pavement management and asset inventory data to NYSDOT. Fugro staff visually inspect pavement imagery every one hundredth of a mile and create a geodatabase that lists each segment and whether it has CARDs installed.
Eligible Segments without CARDs
The eligible segments were queried from the 2017 NYSDOT Roadway Inventory System (RIS) geodatabase based on the criteria listed in NYSDOT Engineering Instruction 13-021 and combined with the Fugro CARDs data to determine the eligible segments without CARDs.
Table 3 and Figures 1 to 5 illustrate the miles of CARDs segments and eligible segments without CARDs used in the study by NYSDOT region, AADT range, functional class, posted speed limit, and by number of lanes.
Miles of Centerline Audible Roadway Delineators (CARDs) Segments and Eligible Segments without CARDs Used in the Study
Note: NYSDOT = New York State Department of Transportation.

Centerline audible roadway delineators (CARDs) segments and eligible segments without CARDs used in the study.

Percent of miles of Centerline audible roadway delineators (CARDs) segments and eligible segments without CARDs used in the study, by AADT range.

Percent of miles of enterline audible roadway delineators (CARDs) segments and eligible segments without CARDs used in the study, by functional class.

Percent of miles of centerline audible roadway delineators (CARDs) segments and eligible segments without CARDs used in the study, by posted speed limit.

Percent of miles of centerline audible roadway delineators (CARDs) segments and eligible segments without CARDs used in the study, by number of lanes.
Data Processing
The data was processed via the following tasks:
The 2017 RIS geodatabase was provided by the NYSDOT Highway Data Services Bureau (RISFile0). RIS data (RISFile0) were queried to determine which segments were eligible. An indicator field was added to the RIS geodatabase indicating whether a segment was eligible for CARDs (RISFile1).
CARDs data (location and year of installation) were provided by NYSDOT SEEs as Excel spreadsheets and imported into GIS. The data were verified against 2017 Fugro data to determine whether CARDs were present. If there was a mismatch, CARDs that were identified by Fugro were included.
RISFile1 was spatially joined to the verified CARDs data in GIS. The spatial join appended the attributes of the verified CARDs data to RISFile1. RISFile2 was RISFile1 plus three columns: whether the CARDs were present, the year the CARDs were installed, and whether the segment was eligible for CARDs but did not have them installed.
Crash data (HO/SS and ROR) were downloaded from ALIS in separate Excel files. The data were manually filtered to only include crashes correctable by the presence of CARDs, and were imported into GIS as point layers.
RISFile2 was joined to HO/SS crashes correctable by CARDs in GIS by a spatial join. The spatial join appended the attributes of the crash data to RISFile2. RISFile3 was RISFile2 plus several columns (by year): number of HO/SS crashes, number of HO/SS injury crashes, number of HO/SS fatal crashes. RISFile3 was spatially joined to ROR crashes correctable by CARDs in GIS. RISFile4 was RISFile3 plus several columns (by year): number of ROR crashes, number of ROR injury crashes, number of ROR fatal crashes.
RISFile4 was exported from GIS and analyzed in Excel.
Figure 6 illustrates the data process diagram for this study.

Centerline audible roadway delineators (CARDs) before and after study process diagram.
Data Analysis
Methodology
An observational, naïve before–after with comparison group study was used to evaluate the CARDs. This analytical approach is one of the evaluation methods identified in AASHTO’s HSM. Gross et al. showed that this type of study can be used to derive CMFs from a before–after study ( 13 ). The treatment group included segments where CARDs were installed in 2013, 2014, and 2015 in NYSDOT Regions 1 and 4, and Dutchess County Region 8. The comparison group included eligible segments where CARDs were not installed in these regions. Table 3 and Figure 1 provide information related to the number of miles of roadway in the treatment group versus the comparison group by region.
The EB method was not pursued owing to the complexity of applying the method properly. According to Persaud and Lyon, “there are a number of difficulties which, if not properly resolved, will render this methodology just as invalid as the conventional methods, resulting in a misuse of precious resources and a general lack of faith in the method” ( 14 ). New York State does not have its own safety performance functions (SPF) for CARDs, which are critical to performing the EB method. SPFs are listed in the HSM, however, the HSM recommends its SPFs be calibrated to local conditions: “as these distributions will vary between jurisdictions, the estimation will benefit from updates based on local crash severity and collision type data” ( 3 ).
The comparison group methodology is an alternative to the EB method in certain cases. Gross et al. assert, “Where there is no regression-to-the-mean and where a suitable comparison group is available, the comparison group methodology can be a simple alternative to the more complex empirical Bayes approach” further qualifying, “this may be true in cases where … a blanket treatment is applied to all sites of a given type” ( 13 ). In New York State, CARDS are installed systemically based on the aforementioned criteria in NYSDOT Engineering Instruction 13-021.
The regression-to-the-mean effect can decrease if 3 years’ worth of crash data are available. According to Sharma and Datta, “before-and-after studies produce results similar to those of the EB method, and the RTM effect becomes insignificant when three or more years’ worth of traffic crash data are used in the evaluation of high-crash locations” ( 15 ). There were 4 years of before–after data for CARDs installed in 2013, 3 years’ data for CARDs installed in 2014, and 2 years’ before–after data for CARDs installed in 2015.
As stated in the HSM, “a valid comparison group is essential when implementing an observational before/after evaluation study using the comparison group method” ( 3 ). According to Gross et al., “the comparison group should be drawn from the same jurisdiction as the treatment group and be similar to the treatment group in relation to geometric and operational characteristics” ( 13 ). The comparison group included similar roadway segments to those locations treated with CARDs in the same regions. As shown in Figures 2 to 5, AADT, functional class, posted speed limit, and number of lane distributions are similar between the treatment and the comparison group.
Statewide Findings
The CMFs calculated are presented in Table 4. The results of this study showed:
CARDs segments installed in 2013 had a CMF of 0.572 (expected reduction of 42.8%) in all injury HO/SS crashes.
CARDs segments installed in 2014 had a CMF of 0.623 (expected reduction of 37.7%) in all HO/SS crashes, and a CMF of 0.551 (expected reduction of 44.9%) in all injury HO/SS crashes.
CARDs segments installed in 2015 had a CMF of 0.774 (expected reduction of 22.6%) in all ROR crashes, and a CMF of 0.492 (expected reduction of 50.8%) in all HO/SS crashes
Crash Modification Factors (CMFs) for Centerline Audible Roadway Delineators (CARDs) Installed in 2013–2015 in the Study Area
Note: The CMFs for fatal crashes have a small sample size and therefore should be used with caution. Bold indicates the numbers are statistically significant at the 95th percentile.
Figures 7 and 8 show the ROR and HO/SS CMFs including error bars. In reviewing these figures, the CMFs and error bars that fell completely below the red line were found to be statistically significant and showed an expected safety benefit at the 95% confidence level.

Run-off-road crash modification factors (CMFs) with 95% confidence interval for centerline audible roadway delineators (CARDs) installed in 2013 to 2015 in the study area.

Head-on sideswipe crash modification factors (CMFs) with 95% confidence interval for centerline audible roadway delineators (CARDs) installed in 2013 to 2015 in the study area.
The overall study results are shown in Tables 5 to 7. The results by region are presented in Tables 8 to 10.
Crash Modification Factors (CMFs) and 95% Confidence Intervals (CIs) for Centerline Audible Roadway Delineators (CARDs) Installed in 2013 in the Study Area
Crash Modification Factors (CMFs) and 95% Confidence Intervals (CIs) for Centerline Audible Roadway Delineators (CARDs) Installed in 2014 in the Study Area
Note: NA = not available.
Crash Modification Factors (CMFs) and 95% Confidence Intervals (CIs) for Centerline Audible Roadway Delineators (CARDs) Installed in 2015 in the Study Area
Crash Modification Factors (CMFs) for Centerline Audible Roadway Delineators (CARDs) Installed in 2013 to 2015—NYSDOT Region 1
Note: NA = not available; NYSDOT = New York State Department of Transportation.
Crash Modification Factors (CMFs) for Centerline Audible Roadway Delineators (CARDs) Installed in 2013 to 2015—NYSDOT Region 4
Note: NA = not available; NYSDOT = New York State Department of Transportation.
Crash Modification Factors (CMFs) for Centerline Audible Roadway Delineators (CARDs) Installed in 2013 to 2015—NYSDOT Region 8 (Dutchess County)
Note: NA = not available; NYSDOT = New York State Department of Transportation.
Regional Findings
Region 1
There were no significant CMFs for CARDs installed in 2013 because of the small sample size.
For CARDs installed in 2014, HO/SS crashes had a 45% reduction, which was significant at the 95% confidence level. This compares with the overall trend in which HO/SS crashes showed a 37.7% reduction.
For CARDs installed in 2015, ROR crashes had a 41.7% reduction, which was significant at the 90% confidence level. This compares with the overall trend in which ROR crashes showed a 22.6% reduction.
Region 4
For CARDs installed in 2013, HO/SS injury crashes had a 63.1% reduction, which was significant at the 95% confidence level. This compares with the overall trend in which HO/SS injury crashes showed a 42.8% reduction.
For CARDs installed in 2014, HO/SS injury crashes had a 56.5% reduction in injury crashes, which is significant at the 95% confidence level. This compares with the overall trend in which HO/SS injury crashes showed a 44.9% reduction.
For CARDs installed in 2015, HO/SS crashes showed a 50.9% reduction in crashes and a 53% reduction in injury crashes, both of which were significant at the 95% confidence level. This compares with the overall trend in which HO/SS crashes show a 50.8% reduction in crashes and a 34.5% reduction in injury crashes.
The results in Region 4 were most like the overall trend because almost 70% of the CARDs segments in the study were from this region.
Region 8 (Dutchess County)
For CARDs installed in Dutchess County in 2013 and 2014, there were no significant CMFs for CARDs installed in 2013 or 2014 because of the small sample size.
For CARDs installed in 2015, ROR crashes had a 65.8% reduction in crashes, which was significant at the 95% confidence level, and a 50.9% reduction in injury crashes, which was significant at the 90% confidence level. This compared with the overall trend in which ROR crashes show a 22.6% reduction in crashes and a 21.7% reduction in injury crashes.
Summary
CARDs were found to be an effective treatment for reducing HO/SS and ROR crashes. CMFs were calculated:
ROR—a CMF of 0.774 (expected reduction of 22.6%) in all ROR crashes.
HO/SS—a CMF of 0.623 (expected reduction of 37.7%) in all HO/SS crashes, and a CMF of 0.572 (expected reduction of 42.8%) in all injury HO/SS crashes.
The most conservative of the statistically significant CMFs are recommended for use.
Recommendations for Future Research
As stated, the EB method was not pursued in the current research owing to the complexity of applying the method properly.However, an EB before–after study could be conducted once NYSDOT has developed its own state-specific SPFs. According to Harkey et al., the EB method normalizes volume differences between the before and after periods, accounts for time trends, reduces the level of uncertainty in the estimates of safety effect, and properly accounts for differences in reporting practice from different jurisdictions ( 5 ). However, the observational, naïve before–after with comparison group study is documented in AASHTOs HSM as a method that can be used to evaluate treatment effectiveness.
The study could be improved by including 2018 crash data. Including 2018 crash data would provide 5 years’ before–after data for CARDs installed in 2013, 4 years’ before–after data for CARDs installed in 2014, and 3 years’ before–after data for CARDs installed in 2015.
The study could be expanded by including CARDs from the remaining regions of New York State. This would reduce the effect of regression-to-the-mean.
The study could also be advanced by conducting a disaggregate analysis to identify conditions under which the treatment is most effective: The RIS geodatabase contains variables such as AADT, functional class, posted speed limit, and lane width that could be analyzed.
Footnotes
Author Contributions
The author confirms sole responsibility for the following: study conception and design, data collection, analysis and interpretation of results, and manuscript preparation.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
