Reduced Conflict Intersection Safety Evaluation in MN

A Reduced Conflict Intersection (RCI), also known as a Restricted Crossing U-Turn (RCUT) or J-Turn, is an at-grade intersection design used on high-speed, multi-lane expressways. The goal of an RCI is to improve intersection safety for vehicles by reducing the number and severity of angle crashes.

At standard expressway intersections, vehicles on any leg may turn directly onto any other leg.

At an RCI, vehicles on the mainline may turn onto any other leg while vehicles on the crossroad may only turn right onto the mainline. Crossroad vehicles going straight or turning left need to utilize the u-turn in the median to do so. The goal of this diverted path is to reduce the likelihood of angle crashes, commonly referred to as T-bone crashes, by making drivers on the crossroad only have to contend with one direction of mainline traffic at a time.

Below are aerial images of an RCI showing the median u-turns and the channelized left turns on the mainline.

Below is the view from the turn lane for a median u-turn at an RCI.

The first RCI in Minnesota was installed in 2010 in Willmar. As of 2020, there were 49 RCIs in place with more planned over the next several years.

The locations of these 49 RCIs can be seen in the map below. Click on the RCI locations to view their construction years.

Note that the following four locations with existing RCIs were not included in any analysis in this evaluation. Those locations are:

Note that the following four locations with existing RCIs were not included in any analysis in this evaluation. Those locations are:

• MN 65 & Viking Boulevard in East Bethel. This is a signalized RCI with dual right turn lanes and dual U-turn lanes. This is not included in the analysis because the layout and operations of this RCI are significantly different from the other RCI locations. This RCI was constructed in 2019.

• MN 62 & Carmen Lane in Mendota Heights. This intersection is located within 1/3 of a mile of signalized intersections on either side along MN 62. Because of the overwhelming influence of the adjacent signals on the crashes at this location, it is not included in analysis. No fatal (K) or serious injury (A) crashes have been reported at this intersection since 2015.

• MN 371 & County Road 112, MN 371 & County Road 168/107 in Pequot Lakes. Both RCIs are located on a newly constructed stretch of MN 371 that bypasses downtown Pequot Lakes. Since these are on a new stretch of highway, there is no “before” data at these locations. Because of this lack of data for comparison, these locations are not used in this analysis.

For comparison purposes in the evaluation, crashes that occurred during the year of construction at each location were not included in the analysis. The analysis was conducted in 2020, so the most recent year of data analyzed was 2019.

Using the crash data, three types of analysis were conducted - a before-after analysis, a cross-sectional analysis, and a comparison analysis.

The before-after analysis compares crash data at RCI locations before the RCI was installed and after the RCI was installed. Since the analysis for this evaluation only includes crash data through 2019, there is no “after” data for the RCIs constructed in 2019 or 2020. Those locations were therefore not utilized in this analysis. This leaves 25 RCI locations that have at least one site-year of before and one site-year of after data which totals to 89 site-years of before data and 89 site-years of after data

Using the before-after crash data as well as before-after volume data, a Wilcoxon Signed Rank Test was used. The analysis and testing were focused on six crash severities/types. These are based on both the expected benefits of RCIs as well as commonly heard concerns about RCIs. These focus types are listed below.

• Fatal (K) and suspected serious injury (A) crashes. RCIs are an alternative intersection that are intended to improve safety by reducing crashes with these serious outcomes.

• Angle crashes. This is the target crash type RCIs are intended to reduce.

• K and A angle crashes. These are the most severe outcome types of the target crash type that RCIs are intended to reduce.

• Rear end crashes. It is commonly heard that RCIs will contribute to an increase in rear end crashes.

• Sideswipe crashes. It is commonly heard that RCIs will contribute to an increase in sideswipe crashes.

• Total crashes. RCIs are intended to reduce the most severe types of crashes at the intersections they are installed at, but not necessarily intended to reduce overall crashes.

The Wilcoxon Signed Rank Test results in a p-value which is compared to a predetermined threshold significance level of 0.05 in this case. When the p-value is below the significance level, that suggests there is a significant difference in the before-after results. The results are shown in the chart and table below.

As shown, the conversion to RCIs resulted in statistically significant decreases in fatal and serious injury crashes (KA crashes), angle crashes, and fatal and serious injury angle crashes. Additionally, there was found to be a statistically significant increase in rear end crashes. Though there were increases in sideswipe crashes and decreases in total crashes, these changes were not found to be statistically significant.

It was also found that the crash rates for all injury type crashes decreased with the installation of RCIs, but the crash rates for non-injury crashes (property damage only crashes) increased. With no statistically significant change in total crashes shown, the data is suggesting the installation of RCIs result in a severity shift of crashes from higher to lower severities.

It is noted that the crash reporting system behind the crash data in Minnesota underwent changes in the beginning of 2016. This upgrade improved the crash data system in many ways but, due to changes in the labels of crash severities, it resulted in a change in the percentage of injury severity crashes. As the result of these label changes, Minnesota experienced a dramatic increase in A and B severity crashes from 2015 to 2016 (increasing by 83% and 51% for A and B crashes, respectively). Based on this change, some of the locations in the before-after analyses may have been impacted. However, the results show that A and B severity crashes both experienced large decreases at the RCI locations. This emphasizes the decreases seen at RCI locations.

The cross-sectional analysis takes a group of locations that have RCIs at them (treatment sites) and compares the before-after crash data there against the before-after crash data at a group of similar intersections without RCIs (control sites). For this comparison, only RCI locations that had at least three years of “after” data were included. There are 13 locations that have RCIs during this 2017 through 2019 period.

For the control group, these locations should be similar to the treatment sites but cannot have had an RCI at them between 2017-2019. The sites that are included in this group at the seven locations where RCIs were constructed in 2020 and 35 locations with future RCIs planned/considered. That totals to 42 locations for the control group.

At the treatment sites, the before period is the three-year period before an RCI was installed, and the after period is the three years from 2017 through 2019 when an RCI was in place. At the control sites, the before period is the three years from 2013 through 2015 and the after period is the three years from 2017 through 2019.

For this cross-sectional analysis, all the before periods in this analysis are before 2016 and all the after periods in this analysis are after 2016. These time periods were selected to allow for a comparison between the treatment and control groups so that neither group is disproportionally impacted by the 2016 statewide changes to the crash data.

Using the cross-sectional crash data, a Mann-Whitney U-Test was used. This test resulted in a p-value which was compared to a significance level of 0.05 in this case. When the p-value is below the significance level, that suggest there is a significant difference in the results. The results are shown below.

As shown, the RCI sites showed statistically significant decreases in fatal and serious injury (KA) crashes, angle crashes, and fatal and serious injury (KA) angle crashes. These results line up with the goals of RCIs and are similar to what was seen in the before-after analysis. The installation of RCIs also showed a statistically significant increase in rear end crashes with no statistically significant changes at the 0.05 significance level for sideswipe or total crashes. 

RCIs typically replace side-street, stop-controlled intersections on high-speed expressways. One alternative to the RCI would be a grade separated intersection, or an interchange. Interchanges require more right-of-way and have significantly higher costs associated with them as compared to an RCI. Another alternative to the RCI would be a signalized intersection. This analysis compares the crash data at interchanges with volumes similar to what would be found at an RCI as well as at signalized intersections with volumes and characteristics similar to what would be found at an RCI to the crash data at RCIs.

There are over 700 interchanges in Minnesota including many that serve very high volumes of traffic. To be able to get a set of interchanges that would be able to be meaningfully compared to RCIs, the volumes had to be considered. High volume interchanges, such as those that serve the meeting of two Interstate Highway System routes, would not be locations where an RCI would ever be considered. Because of that, only low volume interchanges were selected. Low volume, in this case, means daily volumes of 45,000 or less on the mainline with average daily volumes of 6,000 or less on the minor approaches. These volumes represent the upper end of the volumes seen at RCIs in Minnesota. Using those filters, 225 interchanges were selected and crash data from 2017 through 2019 was used.

Signalized intersections are utilized on a wide variety of intersection types, so to get a meaningful comparison site for RCIs, only signalized intersections that are on high-speed, rural roadways with the same volume constraints as the low volume interchanges were used. Signalized intersections that include interchange ramps were not included. Using those filters, 19 intersections were selected and crash data from 2017 through 2019 was used.

Like the cross-sectional analysis, the 13 RCI locations that were fully in place from 2017 through 2019 were used for comparison. Using only 2017 through 2019 data avoids any inconsistencies between the pre-2016 and post-2016 crash data due to the statewide changes previously discussed.

Using the crash and volume data for each group, any statistically significant differences between crash rates for these three intersection types were checked. A Kruskal-Wallis rank sum test was utilized with a Wilcoxon Rank Sum test used on the results to determine if they are significantly different from one another compared to a significance level of 0.05. The results of this testing are shown below. The crash types and severities analyzed that did not have statistical significance between intersection types are not shown.

Here are the results where there is a significance level of 0.05:

• The average crash rate for total crashes at interchanges is higher than at RCIs.

• The average crash rate for property damage only crashes at interchanges is higher than at RCIs.

• The average crash rate for intersection related crashes at signals is higher than at RCIs.

If a significance level of 0.10 were to be used rather than 0.05, the following conclusions could be drawn:

• The average crash rate for angle crashes at signals is higher than at RCIs.

• The average crash rate for sideswipe crashes at interchanges is higher than at RCIs.

• The average crash rate for sideswipe crashes at RCIs is higher than at signals.

With relatively small numbers of K and A crashes at chosen locations for RCIs and rural signals, there is not the ability to draw clear distinctions between the intersection types regarding severe crashes.

RCIs tend to have lower crash rates compared to low volume interchanges when it comes to overall crashes and lower crash rates compared to signals when it comes to angle crashes. The low rate of angle crashes at RCIs is in line with the other results from this study. Though the before-after and cross-sectional analyses showed increases in rear-end crashes at RCIs, they are not statistically significantly different than the other intersection types.

Intersection related crashes are crashes that the attending officer determined were impacted by the presence of an intersection. With RCIs, the area included to collect crashes is quite large due to the location of the median U-turns. Similarly, interchanges encompass large areas. Because of that at RCIs, a portion of the crashes that occur within that large envelope may not be related to the RCI but just happened to occur at that location. That is always the case with any intersection, but the large envelope of the RCI makes it potentially more so. The crash rate results for intersection related crashes show RCIs have lower intersection related crashes than signals, which could indicate even a lower portion of the total crashes occurring at RCIs are related to the RCI itself.

The following images show a breakdown of how and where rear end, angle, and sideswipe crashes are occurring at RCIs and standard intersections. The crash numbers shown in these figures are from the Before-After analysis with the crashes at standard intersections being from the Before period and crashes at RCIs being from the After period.

The results of the before-after and cross-sectional analyses conducted show the RCIs in Minnesota are exhibiting their intended safety benefits. The analyses showed the following impacts of RCIs:

• Reductions in fatal and serious injury crashes.

• Reductions in angle crashes.

• Reductions in fatal and serious injury angle crashes.

• Increases in rear end crashes.

• No changes to sideswipe crashes.

• No changes to total crashes.

These results are consistent with the safety goals of RCIs as well as with the previous evaluation of RCIs in Minnesota. Though the RCIs are not causing significant changes in total crashes, there is a severity shift that is resulting in a decrease in high severity crashes.

A comparison between RCIs, rural signals, and low volume interchanges show that RCIs appear to result in lower overall crashes than interchanges and lower angle and intersection related crashes than signals.

More details about this evaluation can be seen in the full report found here: http://www.dot.state.mn.us/trafficeng/safety/reportspubl.html