Report # MATC-MU: 276     

Final Report

WBS# 25-1121-0003-276

Safety Evaluation of Diverging Diamond
Interchanges in Missouri
Praveen Edara, Ph.D., P.E., PTOE

Associate Professor
Department of Civil and Environmental Engineering
University of Missouri
Carlos Sun, Ph.D., P.E., J.D.
Associate Professor

Boris Claros, MSCE
Research Assistant

Henry Brown, MSCE, P.E.
Research Engineer

2015
A Cooperative Research Project sponsored by
U.S. Department of Transportation-Research and Innovative
Technology Administration

“The contents of this report reflect the views of the authors, who are responsible for the facts and the accuracy of the information presented herein. This document
is disseminated under the sponsorship of the U.S. Department of Transportation’s University Transportation Centers Program, in the interest of information
exchange. The U.S. Government assumes no liability for the contents or use thereof.”

 Safety Evaluation of Diverging Diamond Interchanges in Missouri

Praveen Edara, Ph.D., P.E., PTOE
Associate Professor
Department of Civil and Environmental Engineering
University of Missouri - Columbia
Carlos Sun, Ph.D., P.E., J.D.
Associate Professor
Department of Civil and Environmental Engineering
University of Missouri – Columbia
Boris R. Claros, MSCE
Research Assistant
Department of Civil and Environmental Engineering
University of Missouri – Columbia
Henry Brown, MSCE, P.E.
Research Engineer
Department of Civil and Environmental Engineering
University of Missouri - Columbia

A Report on Research Sponsored by

Mid-America Transportation Center
University of Nebraska-Lincoln
&
Missouri Department of Transportation

January 2015

 Technical Report Documentation Page 1
1. Report No.
2. Government Accession No.
25-1121-0003-276
4. Title and Subtitle
Safety Evaluation of Diverging Diamond Interchanges in Missouri

7. Author(s)
Edara, P., Sun, C., Claros, B., Brown, H.
9. Performing Organization Name and Address
Mid-America Transportation Center
2200 Vine St.
PO Box 830851
Lincoln, NE 68583-0851
12. Sponsoring Agency Name and Address
Research and Innovative Technology Administration
1200 New Jersey Ave., SE
Washington, D.C. 20590

3. Recipient’s Catalog No.
5. Report Date
January 2015
6. Performing Organization Code
25-1121-0003-276
8. Performing Organization
10. Work Unit No.
11. Contract or Grant No.

13. Type of Report and Period Covered
14. Sponsoring Agency Code
MATC TRB Rip No. 34778

15. Supplementary Notes

16. Abstract
The Diverging Diamond Interchange (DDI) has gained in popularity in the United States during the last decade. The
operational benefits and lower costs of retrofitting a conventional diamond with a DDI have contributed to its increased use.
Existing research on DDIs has focused primarily on the assessment of operational benefits. Unfortunately, formal safety
evaluations of DDIs are lacking. This study filled the knowledge gap by conducting a safety evaluation at the project-level
(interchange) and the site-specific level (ramp terminals) of DDIs using three types of before-after evaluation methods: Naïve,
Empirical Bayes (EB), and Comparison Group (CG). Three evaluation methods were used since the methods involved different
trade-offs, such as data requirements, complexity, and regression-to-the-mean. The safety evaluation at the project-level
accounts for the influence of the DDI treatment in the entire footprint of the interchange. On the other hand, the site-specific
approach focused on the influence at the ramp terminals only. All three methods showed that a DDI replacing a conventional
diamond decreased crash frequency for all severities. At the project-level, the highest crash reduction was observed for fatal
and injury (FI) crashes – 63.2% (Naïve), 62.6% (EB), and 59.3% (CG). Property damage only crashes were reduced by 33.9%
(Naïve), 35.1% (EB), and 44.8% (CG). Total crash frequency also decreased by 41.7% (Naïve), 40.8% (EB), and 47.9% (CG).
Similarly, in the site-specific analysis, the highest crash reduction was observed for fatal and injury (FI) crashes – 64.3%
(Naïve), 67.8% (EB), and 67.7% (CG). Property damage only crashes were reduced by 35.6% (Naïve), 53.4% (EB), and 47.0%
(CG). Total crash frequency also decreased by 43.2% (Naïve), 56.6% (EB), and 53.3% (CG). A collision type analysis revealed
that the DDI, as compared to a diamond, traded high severity for lower severity crashes. While 34.3% of ramp terminal-related
FI crashes in a diamond occurred due to the left turn angle crashes with oncoming traffic, the DDI eliminated this crash type.
In summary, the DDI offers significant crash reduction benefits over conventional diamond interchanges.

17. Key Words
Diverging Diamond, Safety, Alternative Designs, Empirical
Bayes, DDI
19. Security Classif. (of this report)
Unclassified

18. Distribution Statement

20. Security Classif. (of this page)
Unclassified

i

21. No. of Pages
22. Price
46
Form DOT F 1700.7 (06/98)

 Technical Report Documentation Page 2
1. Report No.

2. Government Accession No.

3. Recipient’s Catalog No.

4. Title and Subtitle
Safety Evaluation of Diverging Diamond Interchanges in Missouri

5. Report Date
January 2015
6. Performing Organization Code

7. Author(s)
Edara, P., Sun, C., Claros, B., Brown, H.

8. Performing Organization
University of Missouri-Columbia

9. Performing Organization Name and Address
University of Missouri-Columbia
E 2509 Lafferre Hall, Columbia, MO 65211

10. Work Unit No.

12. Sponsoring Agency Name and Address
Missouri Department of Transportation
105 W. Capitol Avenue
Jefferson City, MO 65102

13. Type of Report and Period Covered

11. Contract or Grant No.

14. Sponsoring Agency Code

15. Supplementary Notes

16. Abstract
The Diverging Diamond Interchange (DDI) has gained in popularity in the United States during the last decade. The
operational benefits and lower costs of retrofitting a conventional diamond with a DDI have contributed to its increased use.
Existing research on DDIs has focused primarily on the assessment of operational benefits. Unfortunately, formal safety
evaluations of DDIs are lacking. This study filled the knowledge gap by conducting a safety evaluation at the project-level
(interchange) and the site-specific level (ramp terminals) of DDIs using three types of before-after evaluation methods: Naïve,
Empirical Bayes (EB), and Comparison Group (CG). Three evaluation methods were used since the methods involved different
trade-offs, such as data requirements, complexity, and regression-to-the-mean. The safety evaluation at the project-level
accounts for the influence of the DDI treatment in the entire footprint of the interchange. On the other hand, the site-specific
approach focused on the influence at the ramp terminals only. All three methods showed that a DDI replacing a conventional
diamond decreased crash frequency for all severities. At the project-level, the highest crash reduction was observed for fatal
and injury (FI) crashes – 63.2% (Naïve), 62.6% (EB), and 59.3% (CG). Property damage only crashes were reduced by 33.9%
(Naïve), 35.1% (EB), and 44.8% (CG). Total crash frequency also decreased by 41.7% (Naïve), 40.8% (EB), and 47.9% (CG).
Similarly, in the site-specific analysis, the highest crash reduction was observed for fatal and injury (FI) crashes – 64.3%
(Naïve), 67.8% (EB), and 67.7% (CG). Property damage only crashes were reduced by 35.6% (Naïve), 53.4% (EB), and 47.0%
(CG). Total crash frequency also decreased by 43.2% (Naïve), 56.6% (EB), and 53.3% (CG). A collision type analysis revealed
that the DDI, as compared to a diamond, traded high severity for lower severity crashes. While 34.3% of ramp terminal-related
FI crashes in a diamond occurred due to the left turn angle crashes with oncoming traffic, the DDI eliminated this crash type.
In summary, the DDI offers significant crash reduction benefits over conventional diamond interchanges.

17. Key Words
Diverging Diamond, Safety, Alternative Designs, Empirical
Bayes, DDI
19. Security Classif. (of this report)
Unclassified

18. Distribution Statement

20. Security Classif. (of this page)
Unclassified

ii

21. No. of Pages
22. Price
49
Form DOT F 1700.7 (06/98)

 Table of Contents
Acknowledgements ......................................................................................................................... vi
Disclaimer ......................................................................................................................................vii
Abstract ...................................................................................................................................................... viii

Chapter 1 Introduction .....................................................................................................................1
Chapter 2 Methodology ...................................................................................................................6
2.1 Site Selection and Data ........................................................................................................6
2.1.1 Treatment Sites ...........................................................................................................7
2.1.2 Comparison Sites ......................................................................................................12
2.2 Crash Reports Review........................................................................................................16
2.3 Crash Type Analysis ..........................................................................................................18
2.4 Safety Effectiveness Evaluation ........................................................................................19
2.4.1 Naive Method............................................................................................................20
2.4.2 Empirical Bayes Method...........................................................................................21
2.4.2.1 EB Project-Level ...........................................................................................22
2.4.2.2 EB Site-Specific ............................................................................................25
2.4.3 Comparison Group ....................................................................................................26
Chapter 3 Results ...........................................................................................................................28
3.1 Introduction ........................................................................................................................28
3.2 Crash Type Analysis ..........................................................................................................28
3.3 Safety Effectiveness Evaluation ........................................................................................31
3.3.1 Crash Severity Analysis ............................................................................................31
3.3.2 Naïve Method............................................................................................................33
3.3.3 Empirical Bayes Method...........................................................................................34
3.3.4 Comparison Group Method ......................................................................................37
3.4 Results of Site-Specific Analysis .......................................................................................39
3.4.1 Naive Method............................................................................................................39
3.4.2 Empirical Bayes Method...........................................................................................40
3.4.3 Comparison Group ....................................................................................................41
Chapter 4 Conclusions ...................................................................................................................43
References ......................................................................................................................................45

iii

 List of Figures
Figure 1.1 Traffic movements at a DDI (FHWA 2014) ..................................................................2
Figure 1.2 Conflict points at DDI and TUDI interchanges (SI 2004) .............................................3
Figure 2.1 Steps involved in the research methodology ..................................................................6
Figure 2.2 RT-13 and I-44, Springfield, MO ...................................................................................9
Figure 2.3 I-270 and Dorsett Rd, Maryland Heights, MO ...............................................................9
Figure 2.4 James River Exp. and National Ave., Springfield, MO ...............................................10
Figure 2.5 US 65 and MO248, Branson, MO ................................................................................10
Figure 2.6 I-435 and Front Street, Kansas City, MO .....................................................................11
Figure 2.7 Chestnut Exp. and Route 65, Springfield, MO .............................................................11
Figure 2.8 Interchange footprint ....................................................................................................12
Figure 2.9 US 60 and US 160, Springfield, MO ............................................................................13
Figure 2.10 IS 170 and Page Ave., Overland, MO ........................................................................13
Figure 2.11 US 65 and Division St., Springfield, MO...................................................................14
Figure 2.12 US65 and Branson Hills Pkwy., Branson, MO ..........................................................14
Figure 2.13 IS 435 and 23rd Trfy., Kansas City, MO ...................................................................14
Figure 2.14 US 65 and Battlefield Rd., Springfield, MO ..............................................................15
Figure 2.15 Area of interest for ramp terminal related crashes .....................................................18
Figure 3.1 Before/after collision diagrams for fatal and injury crashes ........................................30
Figure 3.2 Crash frequencies before/after DDI implementation by facility ..................................32
Figure 3.3 Crash frequencies before/after DDI Implementation all facilities ...............................33

iv

 List of Tables
Table 2.1 DDI site characteristics ....................................................................................................8
Table 2.2 Comparison group sites description ...............................................................................16
Table 3.1 DDI before/after duration ..............................................................................................29
Table 3.2 Naïve method results: project-level ...............................................................................34
Table 3.3 Project-level EB results .................................................................................................36
Table 3.4 Comparison of treatment and control sites ....................................................................37
Table 3.5 Project-level safety effectiveness...................................................................................39
Table 3.6 Naïve method results: site-specific analysis ..................................................................40
Table 3.7 Site-specific EB results ..................................................................................................41
Table 3.8 Site-specific safety effectiveness ...................................................................................42

v

 Acknowledgments
The authors are thankful for the assistance provided by MoDOT staff John Miller, Mike
Curtit, Drew Williford, Ashley Reinkemeyer, Jen Harper, Myrna Tucker, Michelle Neuner, and
other district-level MoDOT staff. The authors wish to acknowledge the contributions of Paige
Martz, Calvin Fales, Dylan Hackman, and Lily Riehl for their assistance with the crash analysis.

vi

 Disclaimer
The contents of this report reflect the views of the authors, who are responsible for the facts
and the accuracy of the information presented herein. This document is disseminated under the
sponsorship of the U.S. Department of Transportation’s University Transportation Centers
Program, in the interest of information exchange. The U.S. Government assumes no liability for
the contents or use thereof.

vii

 Abstract
The Diverging Diamond Interchange (DDI) has gained in popularity in the United States
during the last decade. The operational benefits and lower costs of retrofitting a conventional
diamond with a DDI have contributed to its increased use. Existing research on DDIs has focused
primarily on the assessment of operational benefits. Unfortunately, formal safety evaluations of
DDIs are lacking. This study filled the knowledge gap by conducting a safety evaluation at the
project-level (interchange) and the site-specific level (ramp terminals) of DDIs using three types
of before-after evaluation methods: Naïve, Empirical Bayes (EB), and Comparison Group (CG).
Three evaluation methods were used since the methods involved different trade-offs, such as data
requirements, complexity, and regression-to-the-mean. The safety evaluation at the project-level
accounts for the influence of the DDI treatment in the entire footprint of the interchange. On the
other hand, the site-specific approach focused on the influence at the ramp terminals only. All
three methods showed that a DDI replacing a conventional diamond decreased crash frequency for
all severities. At the project-level, the highest crash reduction was observed for fatal and injury
(FI) crashes – 63.2% (Naïve), 62.6% (EB), and 59.3% (CG). Property damage only crashes were
reduced by 33.9% (Naïve), 35.1% (EB), and 44.8% (CG). Total crash frequency also decreased by
41.7% (Naïve), 40.8% (EB), and 47.9% (CG). Similarly, in the site-specific analysis, the highest
crash reduction was observed for fatal and injury (FI) crashes – 64.3% (Naïve), 67.8% (EB), and
67.7% (CG). Property damage only crashes were reduced by 35.6% (Naïve), 53.4% (EB), and
47.0% (CG). Total crash frequency also decreased by 43.2% (Naïve), 56.6% (EB), and 53.3%
(CG). A collision type analysis revealed that the DDI, as compared to a diamond, traded high
severity for lower severity crashes. While 34.3% of ramp terminal-related FI crashes in a diamond
occurred due to the left turn angle crashes with oncoming traffic, the DDI eliminated this crash

viii

 type. In summary, the DDI offers significant crash reduction benefits over conventional diamond
interchanges.

ix

 Chapter 1 Introduction
Recently in the U.S., the Diverging Diamond Interchange (DDI) has become a popular
alternative to other forms of interchange designs. Since the first DDI installation in Springfield,
Missouri, in 2009, there have been more than 30 locations across the U.S. where DDIs have been
installed. Three factors have contributed to this rapid adoption of the DDI in the U.S. First, the
operational benefits of the DDI, including lower overall delay and higher left turn movement
capacity compared to a conventional diamond, have made it an attractive alternative (Bared et al.
2006; Edara et al. 2005). Second, the lower costs of retrofitting an existing diamond interchange
with a DDI have also played an important role in its adoption. For example, a cost comparison
between the DDI and the Tight Urban Diamond Interchange (TUDI) in Kansas City, Missouri,
showed the DDI cost approximately 50% less (Hughes et al. 2010; MoDOT 2014). Third, fewer
conflict points compared to a conventional diamond, along with positive safety results from limited
safety evaluations, have provided further encouragement about the merits of the design evaluations
(Edara et al. 2005).
The operation of a DDI is shown in figure 1.1. In this figure, the freeway runs north and
south, while the crossroad runs east and west. Two ramp terminals are shown – west crossover and
east crossover. Proceeding from west to east, at the west crossover, the eastbound through and left
turn traffic crisscrosses the westbound through traffic. At the east crossover the eastbound left turn
movements diverge from the through traffic while the through traffic proceeds and crisscrosses the
westbound through and left turn traffic. Thus, vehicles drive on the ‘wrong’ side on the crossroad
between the two crossovers. The traffic that travels from east to west experiences the same
crossovers. Four movement types are shown in figure 1.1. Eastbound and westbound through
movements are shown in figure 1.1 (a), left turn movements from the crossroad onto the freeway

1

 are shown in figure 1.1 (b), left turn movements from freeway exit ramps to the crossroad are
shown in figure 1.1 (c), and all right turn movements are shown in figure 1.1 (d).

A

B

C

D

Figure 1.1 Traffic movements at a DDI (FHWA 2014)

The main impetus behind the initial research on DDI was to evaluate its operational
benefits as compared to other designs. While the seminal study of Chlewicki (2003) illustrated the
delay savings resulting from a DDI, the follow-up studies by Edara et al. (2005) and Bared et al.
(2006) further confirmed its operational benefits, specifically the doubling of left turn movement
2

 capacity. Several subsequent studies have agreed with these early studies on the operational
benefits of DDIs (Chlewicki 2013; Chilukuri et al. 2011). Because the motivation behind the initial
research into the DDI was improving operational benefits, there has been a gap in the existing
knowledge pertaining to the safety performance of the DDI. A preliminary assessment of the safety
of an intersection or interchange design was examined using conflict points. Figure 1.2 shows the
conflict points for both a DDI and a conventional diamond interchange. The DDI has 18 conflict
points (2 crossing, 8 merging, and 8 diverging), while the conventional diamond interchange has
30 conflict points (10 crossing, 10 merging, and 10 diverging) (Hughes et al. 2010; Chlewicki
2013; FHWA 2004). Fewer conflict points across all conflict types reduce the exposure of traffic
to crashes. More importantly, 8 out of 10 crossing conflict points are eliminated by the DDI design.
Crossing conflicts typically result in right angle collisions that have a higher potential for injuries
(Hughes et al. 2010).

Figure 1.2 Conflict points at DDI and TUDI interchanges (FHWA 2004)

3

 Typically, empirical safety evaluations of new alternative designs are not possible until a
few years after they are introduced into practice due to the lack of sufficient crash data. One study
(Chilukuri 2011) reviewed crash data for a one-year period after the first DDI was constructed in
Springfield, Missouri. The study concluded that the DDI was operating safely based on a
comparison of before and after crash frequencies. But the small sample size did not allow for a
rigorous statistical safety evaluation.
Due to the crossover of traffic at the two ramp terminals in a DDI, there was some initial
apprehension about the potential for wrong-way crashes (Chlewicki 2003). Some of these concerns
were alleviated through human factor studies conducted by the Federal Highway Administration
(FHWA). Using driver simulator studies, FHWA showed that wrong-way maneuvers were
minimal and not statistically different from those at a conventional diamond interchange (Inman
2007). There are no empirical studies using real-world crash data either confirming or denying the
higher frequency of wrong-way crashes at a DDI. There are also no empirical studies analyzing
differences in the types and frequencies of crashes between a DDI and a conventional diamond.
The current study filled the knowledge gap in the safety of the DDI. Data from six sites in
Missouri were used to conduct a before-after evaluation of the DDI. Missouri was the first state to
have built a DDI and has the largest number of DDIs built or under construction (15 as of the
writing of this report). Thus, Missouri offers a rich dataset for conducting a safety evaluation of
DDIs. The safety evaluation consisted of three types of observational before-after evaluation
methods: Naïve, Empirical Bayes (EB), and Comparison Group (CG). The approach with the three
evaluation methods consisted of project-level analysis (complete interchange footprint crashes)
and site-specific analysis (ramp terminal related crashes). Collision diagram analysis was also

4

 conducted to determine differences in crash types between a DDI and a conventional diamond
interchange.
This study made a few key contributions to the body of literature on DDI performance.
First, this was the first study to conduct a system-wide safety evaluation using multiple DDI sites.
Second, this study presented the first extensive safety evaluation of DDIs, at both project and sitespecific levels, using three before-after analysis methods. Third, crash modification factors (CMF)
at both project and site-specific levels were developed for the first time for the DDI for total, fatal
and injury, and property damage only crashes. The CMF values provide the expected reduction in
crashes achieved by a DDI as compared to a conventional diamond interchange. Fourth, an
extensive review of the collision diagrams was conducted to derive trends in the types of crashes
before and after a DDI was installed at the study sites.

5

 Chapter 2 Methodology
This chapter discusses the research methodology involving crash type and crash severity
analysis, and three different before-after statistical methods. Figure 2.1 shows a schematic of the
different steps involved in the methodology including site selection process, data collection,
review of crash reports, analysis of crashes, and application of before-after statistical evaluation
methods.

Site
Selection

Crash
Reports
Review

Crash Type
Analysis

• DDI sites (treatment group)
• Comparison group sites
• Data collection

• Crash location identification
• Contributing circumstances
• Assignment to interchange facilities

• Crash frequency comparison
• Generate collision diagrams

• Project-level and site-specific analysis
Safety
Evaluation
Methods

•Naïve, Empirical Bayes, Comparison Group methods

Figure 2.1 Steps involved in the research methodology

2.1 Site Selection and Data

6

 The before-after safety analysis of DDI designs implemented in Missouri was conducted
using data from six DDI sites (treatment sites). Six additional sites were used for comparison group
analysis (comparison sites).
2.1.1 Treatment Sites
Although there were ten operational DDI sites in Missouri at the time of this research, four
sites were recently opened to traffic and did not have enough crash data for the afterimplementation period. Therefore, these four sites were not included in the treatment group.
The duration of before and after periods, as shown in table 2.1, was determined by taking
into account seasonality and construction effects. Five years of crash data were processed for the
before period but some were unused in order to match the after period data. The after period
duration varied depending on the opening date of the DDI. The after period ranged from one year
to four years for the six sites. All six DDI designs replaced conventional diamond interchanges.
Table 2.1 contains the following characteristics of the six DDI locations: traffic volume, date
opened to traffic, the duration of before and after periods, geometric characteristics, pedestrian
crossings, traffic control for left turn movements from the crossroad to the entrance ramp, and the
right turn movements from the exit ramp to the crossroad. Figures 2.2 to 2.7 are aerial photographs
of the six DDI sites.

7

 Speed (mph)1
Crossroa
AADT2
d
Lanes3
Speed (mph)1
Freeway AADT2
Lanes
Configuration Type

Chestnut Exp. and
Route 65
Springfield, MO

After

I-435 and Front
Street Kansas City,
MO

Before

US 65 and MO248
Branson, MO

Periods
(Months
)

James River Exp.
and National Ave.
Springfield, MO

Opening Date

I-270 and Dorsett Rd
Maryland Heights,
MO

Site Location

RT-13 an d I-44
Springfield, MO

Table 2.1 DDI site characteristics

6/21/200
9
51

10/17/20
10
35

7/12/201
0
38

11/20/20
11
44

11/6/201
1
44

11/10/20
12
40

51

35

38

22

22

10

40
27082
4
60
47734
4
Overpas
s

35
29275
6
60
151923
8
Underpas
s

40
26891
6
60
68179
4
Overpas
s

35
19842
3
65
32604
4

40
16087
4
65
75276
6
Underpa
ss

40
24513
4
60
62207
6
Underpas
s

Pedestrian
Median Roadside
Accommodation
Ramp Terminal Spacing
530
480
(ft.)
Dist. to Adjacent Street
320/685 265/635
(ft.)
Notes: 1 Posted speed
2
AADT of 2013 for reference purpose only
3
Lanes between ramp terminals

8

Overpass

Median

Median

Median

Roadside

630

740

420

370

530/580

580/1795

530/195
5

160/475

 Figure 2.2 RT-13 and I-44, Springfield, MO

Figure 2.3 I-270 and Dorsett Rd, Maryland Heights, MO

9

 Figure 2.4 James River Exp. and National Ave., Springfield, MO

Figure 2.5 US 65 and MO248, Branson, MO

10

 Figure 2.6 I-435 and Front Street, Kansas City, MO

Figure 2.7 Chestnut Exp. and Route 65, Springfield, MO

The data necessary for conducting the before-after analysis were obtained from several
sources. Aerial photographs from Google Earth were used to measure distances and determine
geometric characteristics. The Automated Road Analyzer (ARAN) viewer from the MoDOT
Transportation Management System (TMS) database allowed for facilities to be viewed for
different years and at specific log miles, which enabled the estimation of short distances such as
lane widths and median widths. Computer Aided Design tools were used to measure horizontal

11

 curve distances and radii of ramps and freeway facilities on aerial photographs. Traffic data was
obtained from the MoDOT TMS database for different locations and years within the study period.
Crash data was collected for the entire interchange footprint for the study periods reported
in table 2.1. The footprint included the influence areas of all interchange components. For
freeways, crashes were included from the beginning of speed change lanes to the end of speed
change lanes in both directions of travel. For the crossroad, the influence area included 250 ft. (76
m) from the ramp terminals, and crashes were collected for the ramp terminals and the crossroad
segment in between the terminals. This footprint is recommended in the Highway Safety Manual
(HSM) (AASHTO 2010). Figure 2.8 illustrates in detail the footprint of the interchange and the
facilities considered.

Figure 2.8 Interchange footprint

2.1.2 Comparison Sites

12

 One comparison site was chosen for each treatment site. Each comparison site was
carefully selected by matching the traffic, geometric characteristics, and crash frequency (during
the before period) of the treatment site. The set of comparison sites is called the comparison group,
and figures 2.9 to 2.14 are aerial photographs of these sites. Seasonality and construction effects
were accounted for here just as in the collection of after treatment data as described in section
2.1.1.

Figure 2.9 US 60 and US 160, Springfield, MO

Figure 2.10 IS 170 and Page Ave., Overland, MO

13

 Figure 2.11 US 65 and Division St., Springfield, MO

Figure 2.12 US65 and Branson Hills Pkwy., Branson, MO

Figure 2.13 IS 435 and 23rd Trfy., Kansas City, MO
14

 Figure 2.14 US 65 and Battlefield Rd. Springfield, MO

The basic characteristics of the comparison group sites are presented in table 2.2. The
geometric features considered were the number of lanes, horizontal curves, left turn lanes on the
crossroad, presence of a median, and signal control. The geometric features and the AADTs of the
comparison facilities were tracked over the study period to ensure that they did not vary
significantly or witness high traffic volume fluctuations over the years. The comparison samples
were selected considering all the aforementioned features, but interchanges with identical features
are difficult to find. Therefore, for each DDI site, the best matched interchange, one which was
similar with respect to most of the important traffic and geometric features, was selected. A yoked
comparison was employed, which is a special case of comparison group analysis where a single
comparison site is matched to one treatment site based on similar conditions (Gross et al. 2010).
For example, US 65 and Chestnut Exp. was yoked to US 65 and Battlefield Rd., both located in
Springfield. As shown in the last column of tables 2.1 and 2.2, the AADTs for crossroad and
freeway are similar, as are the speed limits and the number of lanes. However, one was an overpass,
while the other was an underpass.

15

 Configuration Type

50
18461
5
70
23902
4
Overpas
s

40
45
34358
11178
6
4
60
60
120770
58988
6
6
Underpas Overpas
s
s

US 65 and Battlefield
Rd. Springfield,
Greene, MO

IS 435 and 23rd Trfy.
Kansas City, Jackson,
MO

US65 and Branson
Hills Pkwy. Branson,
Taney, MO

US 65 and Division St.
Springfield, Greene,
MO

IS 170 and Page Ave.
Overland, St. Louis,
MO

Site Location
Speed (mph)1
Crossroad AADT2
Lanes3
Speed (mph)1
Freeway
AADT2
Lanes

US 60 and US 160
Springfield, Greene,
MO

Table 2.2 Comparison group sites description

35
45
40
16767
22497
22725
5
6
4
60
65
60
29562
79635
65260
4
6
6
Overpas Underpas Overpas
s
s
s

Spacing Ramp Terminals
680
400
440
680
310
475
(ft.)
Distance to Public Road
290/100
530/550 220/440 430/430 890/225 575/800
(ft.)
0
Left Turn Signal IN4
PO/PO
PO/PO
PP/PP
PP/PP
PO/PO
PP/PP
Exit Ramp Right Turn
Y/Y
Y/Y
Y/Y
Y/Y
Y/Y
SC/SC
Signal5
Notes: 1 Posted speed
2
AADT of 2013 for reference purpose only
3
Lanes between ramp terminals
4
IN = Left turns on crossroad segment between ramp terminals, PP = Protective Permissive,
PO = Protected Only
5
Y = Yield, SC = Signal Control

2.2 Crash Reports Review
Crash reports in Missouri use the statewide Missouri Uniform Crash Report (MUCR)
format. The Missouri State Highway Patrol is the state depository for traffic crash reports with the
responsibility of training their officers to complete the reports following the Statewide Traffic
Accident Records System standards (STARS 2012). All crashes within the footprint of the
interchange were queried for both before and after periods.
16

 Some inconsistencies were found in the crash data obtained from the electronic crash
database. One inconsistency was the inaccurate placement of crashes occurring within the footprint
of an interchange, the so-called crash landing problem. Crashes occurring on the freeway were
sometimes placed on the crossroad and vice versa. Additionally, interchange terminal crashes were
often placed between the two terminals when the crash occurred at one or the other terminal. Other
inconsistencies included errors in the orientation or direction of travel. Crash reports had to be
manually reviewed in order to correct the crash landing problem. This manual review of crashes
ensured that accurate crash data was used for HSM site-specific analysis, collision diagram
generation, and HSM calibration of ramp terminal facilities.
The project-level safety evaluation includes all facilities within the interchange footprint,
while the site-specific analysis of DDI focuses only on ramp terminal related crashes. Ramp
terminal related means that a crash occurred due to the ramp terminal geometric design, operational
performance, and the influence of these factors on driver behavior. According to common crash
reporting practices, crashes that are within 250 ft. on the roadways away from the center of the
intersection in the approaching direction of the crossroad legs and exit ramp segment, are
considered intersection-related crashes (Vogt 1999; Bonneson, Geedipally and Pratt 2012).
However, there are some specific exceptions to this threshold. For instance, a crash that occurs
beyond 250 ft. in the exit ramp segment or crossroad legs, that was caused by queuing at the ramp
terminal, is ramp terminal related. Rear end and sideswipe crashes occurring on the freeway due
to the accumulation of traffic from the ramp terminal are considered ramp terminal related crashes
since the contributing circumstances were generated by the ramp terminal congestion (Bauer and
Harwood 1998). Figure 2.15 shows the possible locations of ramp terminal related crashes (areas
highlighted in blue). These locations include the ramp terminal itself, the crossroad approach legs,

17

 the exit ramps, part of the entrance ramps, and a small section of the freeway adjacent to the exit
ramps. Thus, all crash reports were carefully reviewed to account for all ramp terminal related
crashes for conducting the site-specific analysis.

Figure 2.15 Area of interest for ramp terminal related crashes

2.3 Crash Type Analysis
Fatal and injury crash reports occurring at ramp terminals were reviewed to identify
differences in the types of crashes occurring at a conventional diamond versus a DDI. The reports
consisted of identical before and after period durations. A total of 356 months of crash data, 178
months for the conventional diamond interchanges (before period), and another 178 months for
the Diverging Diamond Interchanges (after period), were analyzed.
Although crashes occurring at all interchange facilities were reviewed, only the crashes
occurring at the ramp terminals or related to the ramp terminals were analyzed using the collision
diagrams. This focus on ramp terminals was due to the fact that the primary difference between
18

 the conventional diamond and a DDI is the configuration of ramp terminals and the interaction
between traffic movements at the terminals.
A collision diagram showing the location, type of crash, and relative frequency of
occurrence was created using before and after data. The use of a collision diagram to analyze crash
types is often used for analyzing intersection safety, including roundabout safety (FHWA 2010).
The collision diagram facilitated the identification and visualization of trends, locations, geometric
influence, and vehicle trajectories before the collision. As a result, collision diagrams with the
crash type and frequencies at both conventional diamond interchanges and DDIs were developed.
2.4 Safety Effectiveness Evaluation
Safety effectiveness evaluations use quantitative estimates of how a treatment, project, or
a group of projects affected crash frequencies or severities. The effectiveness estimate is useful for
future decision-making and policy development (AASHTO 2010). The observational before and
after evaluation methods used in this study compared the anticipated safety of a site without the
treatment in the after period to the actual safety of the entity with the treatment in the after period
(Hauer 1997). Two approaches, project-level and site-specific level, with three different methods,
were selected to evaluate the safety effectiveness of the DDI. The three methods were Naive,
Empirical Bayes (EB), and Comparison Group (CG). These methods were selected due to their
different approach and use in previous safety research (AASHTO 2010; Hummer et al. 2010).
With project-level analysis, an interchange is considered the entire facility or project by
aggregating the various facilities within its footprint. The facilities within the interchange footprint
include ramp terminals, ramp segments, speed-change lanes, crossroad, and freeway segment, as
previously shown in figure 2.8. On the other hand, site-specific analysis focuses on an individual
facility type such as the ramp terminals as in this study. When a DDI replaces an existing diamond

19

 interchange, the ramp terminals and crossroads undergo the most significant changes. The projectlevel analysis produced the safety effect over the entire interchange footprint while the site-specific
analysis produced the effects at the ramp terminals of a DDI.
2.4.1 Naive Method
The main impetus behind the Naïve before-after evaluation method is that the change in
safety from the before period to the after period is the result of all the changes that may have
occurred at the site, including the effect of treatment. The treatment may not be the only change
that occurs at a site and thus attributing the change in safety to the applied treatment alone may
not be accurate. Instead, the Naïve method assumes that the change in safety is caused by all factors
that may have changed from the before period to the after period (Hauer 1997). The safety
effectiveness is calculated using the expected number of crashes and the actual observed number
of crashes for the after period as discussed in Hauer (1997).
The key steps of the procedure are presented here. The expected number of crashes for the
after period (π) is calculated as

𝜋 = ∑ 𝑟𝑑 (𝑗)𝐾(𝑗)

(2.1)

where
𝜋

= expected crashes in the after period;

𝐾(𝑗)

= observed crashes in the before period at facility j;

𝑟𝑑 (𝑗) = ratio of duration of after period to before period for facility j.

𝑟𝑑 (𝑗) =

𝐷𝑢𝑟𝑎𝑡𝑖𝑜𝑛 𝑜𝑓 𝑎𝑓𝑡𝑒𝑟 𝑝𝑒𝑟𝑖𝑜𝑑 (𝑗)
𝐷𝑢𝑟𝑎𝑡𝑖𝑜𝑛 𝑜𝑓 𝑏𝑒𝑓𝑜𝑟𝑒 𝑝𝑒𝑟𝑖𝑜𝑑 (𝑗)

20

(2.2)

 The safety effectiveness (SE) is calculated using the odds ratio (𝜃), which is a function of
the expected crashes (𝜋), observed crashes (𝜆), and the variance of expected crashes (𝑣𝑎𝑟(𝜋)),
using the following equations:

𝑆𝐸 = 100 × (1 − 𝜃)
where

𝜆
𝜋
𝜃=
𝑣𝑎𝑟(𝜋)
1+
𝜋2

(2.3)

(2.4)

The variance is used to express the precision or statistical significance of the odds ratio. Thus,
highly variable data is less precise.
2.4.2 Empirical Bayes Method
The second before-after method, Empirical Bayes (EB), has been used in previous studies
to evaluate the safety effectiveness of alternative intersection designs (Hummer et al. 2010). The
EB method is also recommended by the HSM (2010) for conducting safety evaluations. HSM
discusses many safety effectiveness performance measures, such as percent reduction of crashes,
shift in crash type and severity, and crash modification factors (CMF) (HSM 2010). For
observational before-after studies, it is important to understand the underlying reasons for
implementing a certain treatment. Sites chosen for implementing a DDI typically have either
congestion or safety problems, or both. Thus, a selection bias is introduced into the sample. To
account for this bias and the resulting regression to the mean, the HSM (AASHTO 2010)
recommends using the EB method.
The EB method utilizes safety performance functions (SPF) to estimate the average crash
frequency for treated sites during the after period as though the treatment had not been applied
(AASHTO 2010). This estimated average crash frequency is then compared with the actual crash
21

 frequency during the after period. The expected crash frequency is calculated as the weighted
average of the observed crash frequency and the SPF-predicted crash frequency. The weights are
determined using the overdispersion parameter of the SPF and are not dependent on the observed
crash frequency. The comparison of the expected crash frequency and the observed crash
frequency for the after period forms the basis for deriving safety effectiveness (AASHTO 2010).
The SPF and the associated CMFs for each facility type were used to predict crashes for
each year of the study periods. The general form of the SPF used is shown in equation 2.5 as
𝑁𝑝𝑟𝑒𝑑𝑖𝑐𝑡𝑒𝑑 = 𝑁𝑆𝑃𝐹 × 𝐶𝑖 × (𝐶𝑀𝐹1 × 𝐶𝑀𝐹2 × … × 𝐶𝑀𝐹𝑖 )

(2.5)

where
𝑁𝑝𝑟𝑒𝑑𝑖𝑐𝑡𝑒𝑑

= predicted crash frequency for a specific year of a site type (crashes/year);

𝑁𝑆𝑃𝐹

= predicted crash frequency for base SPF of a site type (crashes/year)

𝐶𝑖

= calibration factor;

𝐶𝑀𝐹𝑖

= crash modification factor specific to a site type characteristic i.

For the site-specific analysis, the calibration factors for ramp terminals were developed
following the HSM approach. Developing calibration factors for other interchange facilities such
as speed change lanes and ramp segments was outside the scope of this study. Thus, for the projectlevel analysis, a calibration factor of 1.0 was used for all interchange facilities.
2.4.2.1 EB Project-Level
There are some differences in the way the EB method is applied at the project-level and
site-specific level, which are described in this section and the next section. For project-level EB
analysis, the predicted crash frequency for the whole interchange was obtained by summing the
predicted values for all interchange facilities as shown in equation 2.6 (AASHTO 2010). Equation
2.6 is formatted as
22

 2

4

4

𝑁𝑖𝑛𝑡𝑒𝑟 = ∑ (∑ 𝑁𝑟𝑎𝑚𝑝 𝑡𝑒𝑟,𝑖 + ∑ 𝑁𝑟𝑎𝑚𝑝,𝑖 + 𝑁𝑓𝑟𝑒𝑒𝑤𝑎𝑦 + ∑ 𝑁𝑠𝑐𝑙,𝑖 )
𝑦𝑒𝑎𝑟𝑠

𝑖=1

𝑖=1

(2.6)

𝑖=1

where
𝑁𝑖𝑛𝑡𝑒𝑟

= predicted crash frequency for all years of an interchange (crashes/year);

𝑁𝑟𝑎𝑚𝑝 𝑡𝑒𝑟,𝑖

= predicted crash frequency for ramp terminal i (crashes/year);

𝑁𝑟𝑎𝑚𝑝,𝑖

= predicted crash frequency for ramp segment i (crashes/year);

𝑁𝑓𝑟𝑒𝑒𝑤𝑎𝑦

= predicted crash frequency for freeway segment (crashes/year);

𝑁𝑠𝑐𝑙,𝑖

= predicted crash frequency for speed-change lane i (crashes/year).

The expected crash frequency was calculated using a weighted average of all the facilities
of an interchange, taking into account correlations among the facilities as recommended by the
HSM (AASHTO 2010). According to Hauer et al. (1997, 2002), there are two bounds of
correlation: perfectly correlated and independent facilities. The weight adjustment factors and
expected crashes following the bounds of correlation are:

𝑤𝐼 =

1.0
∑𝑎𝑙𝑙 𝑘𝑖 × (𝑁𝑝𝑟𝑒𝑑,𝑖 )
1.0 + 𝑖 𝑁
𝑝𝑟𝑒𝑑,𝑎𝑙𝑙

(2.7)

2

1.0

𝑤𝐶 =

2

1.0 +

√𝑘𝑖 × (𝑁𝑝𝑟𝑒𝑑,𝑖 ) ]
[∑𝑎𝑙𝑙
𝑖
𝑁𝑝𝑟𝑒𝑑,𝑎𝑙𝑙

23

2

(2.8)

 𝑁𝑒𝑥𝑝,𝐼,𝑎𝑙𝑙 = 𝑤𝐼 × 𝑁𝑝𝑟𝑒,𝑎𝑙𝑙 + (1 − 𝑤𝐼 ) × 𝑁𝑜𝑏𝑠,𝑎𝑙𝑙

(2.9)

𝑁𝑒𝑥𝑝,𝐶,𝑎𝑙𝑙 = 𝑤𝐶 × 𝑁𝑝𝑟𝑒,𝑎𝑙𝑙 + (1 − 𝑤𝐶 ) × 𝑁𝑜𝑏𝑠,𝑎𝑙𝑙

(2.10)

where
𝑤𝐼

= weighted adjustment factor assuming independence for all sites;

𝑤𝐶

= weighted adjustment factor assuming perfect correlation for all sites;

𝑘𝑖

= overdispersion parameter for facility i;

𝑁𝑝𝑟𝑒𝑑,𝑖

= predicted crash frequency for facility i;

𝑁𝑝𝑟𝑒𝑑,𝑎𝑙𝑙

= total predicted crash frequency for all sites;

𝑁𝑒𝑥𝑝,𝐼,𝑎𝑙𝑙

= total expected crashes with independent correlation for all sites;

𝑁𝑒𝑥𝑝,𝐶,𝑎𝑙𝑙

= total expected crashes with perfect correlation for all sites;

𝑁𝑜𝑏𝑠,𝑎𝑙𝑙

= total observed crashes for all sites.

For partial correlation conditions, Bonneson et al. (2012) recommend averaging the
expected crash estimate of the perfect correlation and independent conditions. The average
expected crash frequency of partial correlation conditions is:

𝑁𝑒,𝑃𝐶 =

𝑁𝑒𝑥𝑝,𝐼 + 𝑁𝑒𝑥𝑝,𝐶
2

(2.11)

where
𝑁𝑒𝑥𝑝,𝑃𝐶 = expected number of crashes assuming partial correlation among interchange
facilities;
𝑁𝑒𝑥𝑝,𝐼 = expected number of crashes assuming independence among interchange
facilities;
24

 𝑁𝑒𝑥𝑝,𝐶 = expected number of crashes assuming perfect correlation among interchange
facilities.
2.4.2.2 EB Site-Specific Analysis
For site-specific EB analysis, the predicted crash frequency for the each ramp terminal was
obtained by predicting crashes using the prediction methodology described in the HSM (AASHTO
2010). All ramp terminals prior to the treatment were full diamond interchanges (D4 type in HSM).
The prediction model for D4 ramp terminals is formatted in equation 2.12 as

𝑁𝐷4,𝑖 = 𝑁𝑆𝑃𝐹 𝐷4,𝑖 × 𝐶𝐷4,𝑖 × (𝐶𝑀𝐹1 × 𝐶𝑀𝐹2 × … × 𝐶𝑀𝐹𝑖 )

(2.12)

where
𝑁𝐷4,𝑖

= predicted crash frequency for a specific year of a D4 ramp terminal with i lanes
(crashes/year);

𝑁𝑆𝑃𝐹 𝐷4,𝑖 = predicted crash frequency for base SPF of a D4 ramp terminal with i lanes
(crashes/year)
𝐶𝐷4,𝑖

= calibration factor for D4 ramp terminal with i lanes;

𝐶𝑀𝐹𝑖 = crash modification factor specific to a site type characteristic i.

The expected crashes were calculated based on the weighted value of each facility. The
weight is a function of the predicted crashes and the SPF overdispersion parameter (k), and it is
formatted in equation 2.13 as

𝑤𝐷4,𝑖 = 1+𝑘

1
𝐷4,𝑖 ×∑ 𝑁𝐷4,𝑖

(2.13)

𝑁𝑒𝑥𝑝 𝐷4,𝑖 = 𝑤𝐷4,𝑖 × 𝑁𝐷4,𝑖 + (1 − 𝑤𝐷4,𝑖 ) × 𝑁𝑜𝑏𝑠,𝑖 (2.14)
25

 where
𝑤𝐷4,𝑖 = weighed value for D4 ramp terminal;
𝑘𝐷4,𝑖

= overdispersion parameter for D4 ramp terminal with i lanes;

𝑁𝐷4,𝑖

= predicted crash frequency for a specific year of a D4 ramp terminal with i lanes
(crashes/year);

𝑁𝑒𝑥𝑝 𝐷4,𝑖 = expected crashes at D4 ramp terminal type with i lanes;
𝑁𝑜𝑏𝑠,𝑖 = observed crashes at ramp terminal i.

2.4.3 Comparison Group
A before and after comparison group method compares the after period crash frequency of
treatment sites (DDI) with the crash frequency of a set of control (or comparison) sites. One
comparison site was chosen for each treatment site. Each comparison site was carefully selected
by examining the traffic, geometric characteristics, and crash frequency (during the before period)
of the treatment site. The same procedure applies to both project-level and site-specific analysis.
The suitability of the comparison group was verified using the sample odds ratio test
presented by Hauer (1997). This test compares crashes over a specified time period for the
comparison and treatment groups during a period before the treatment was implemented. If the
mean of the sequence of odds ratios is sufficiently close to 1.0 and the confidence interval includes
the value of 1.0, then the candidate comparison group is considered a good candidate (Gross et al.
2010; Hauer 1997). The sample odds ratio is calculated as

𝑇1 × 𝐶2
𝑇2 × 𝐶1
𝑆𝑎𝑚𝑝𝑙𝑒 𝑂𝑑𝑑𝑠 𝑅𝑎𝑡𝑖𝑜(𝑆𝑂𝑅) =
1
1
1+𝑇 +𝐶
2
1
26

(2.15)

 where
𝑇1

= total crashes for treatment group in year 1;

𝑇2

= total crashes for treatment group in year 2;

𝐶1

= total crashes for comparison group in year 1;

𝐶2

= total crashes for comparison group in year 2.

The CG safety effectiveness is calculated using both observed crash data and predicted
values. In the first step, SPFs are used to determine the predicted crashes for both before and after
periods, and for treated and comparison sites. An adjustment factor by severity for each period is
then calculated for each pair of treatment and comparison sites by dividing the total number of
predicted crashes for the treatment site and the total number of predicted crashes for the
comparison site. Each treated site is compared to all the comparison sites, thus there are adjustment
factors for each pair of treatment and comparison sites. The expected crashes for comparison and
treatment sites are then calculated using the adjustment factors and observed crashes. The safety
effectiveness values for each site and for the entire treatment group are computed using the
expected and observed crashes. The HSM provides the necessary equations and an illustrative
example for computing the adjustment factors, expected crashes, and the safety for the CG method
(AASHTO 2010).

27

 Chapter 3 Results
3.1 Introduction
This chapter presents the results for the crash type analysis and the safety effectiveness
evaluation discussed in the previous chapter. The final collision diagram is presented, describing
the crash types for conventional diamond interchanges and Diverging Diamond Interchanges. The
safety effectiveness evaluation results are presented, showing crash frequencies by severity, and
the results from the other three methods: 1) Naïve, 2) Empirical Bayes (EB), and Comparison
Group (CG).
3.2 Crash Type Analysis
Since crash type analysis examines the total number of crashes in addition to percentages,
the same data duration, before and after DDI implementation, was used for each site. Sites 1, 2,
and 3 had the same duration of before and after periods. However, for sites 4, 5, and 6, the duration
of the after period was shorter than the before period. Thus, the duration of the before period for
sites 4, 5, and 6 was reduced to match the shorter after period. It is important to note that this
adjustment in duration was only performed for the collision diagram analysis. The crash frequency
analysis and the safety evaluation (Naïve, EB, CG) procedures used the actual durations listed in
table 3.1. The collision diagrams for the before and after period are shown in figure 3.1. Crashes
were classified into 14 different types for the before and after periods. Although the total number
of crash types was 14 in both periods, the distribution and ranking of the types of crashes was
different.

28

 Periods
(Months)

US 65 and MO248
Branson, MO
I-435 and Front
Street Kansas City,
MO
Chestnut Exp. and
Route 65
Springfield, MO

Before
After

James River Exp.
and National Ave.
Springfield, MO

Opening Date

I-270 and Dorsett Rd
Maryland Heights,
MO

Site Location

RT-13 and I-44
Springfield, MO

Table 3.1 DDI before/after duration

6/21/200
9
51
51
10/17/20
10
35
35
7/12/201
0
38
38
11/20/20
11
44
22
11/6/201
1
44
22
11/10/20
12
40
10

29

  

BEFORE PERIOD
Conventional Diamond Interchange (D4 Ramp Terminals}

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

   
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

BEFORE
Crash Type 
1 Crossing left turn inside crossroad leg 5? 34.3
2 Rear end at exit ramp 5] 30.?
3 Crossing left turn at exit ramp 14 8.4
mm 4 Rear end inside crossroad through lane 14 8.4
5 Rear end outside erossroad 8 4.8
6 Rear end due to queuing from ramp terminal 8 4.8
7 Rear end inside crossroad left turn lane 5 3.
8 Sideswipe exit ramp left turn 2 1.2
9 Run off exit ramp right turn 2 1.2
10 Loss of control left turn lane inside crossroad 0.6
1 1 Loss of control outside crossroad bay 1 0.6
12 Run off exit ramp left tum 0.6
13 Loss of control exit ramp bay 1 0.6
14 Pedestrian on 0.6
AFTER PERIOD
Diverging Diamond Interchange (DDI)
AFTER
Crash type 
1 Rear end at exit ramp right turn 23 371.1
2 Rear end outside crossroad 6 9.7
3 Rear end inside erossroad 5 8.1
4 Sideswipe on erossroad 4 6.5
5 Rear end due to queuing from ramp terminal 4 6.5
6 Wrong direction inside crossroad 3 4.8
Crossing at ramp terminal 3 4.8
Notes: 8 Loss of control left turn at exit ramp ha},r 3 4.8
a. Fatal and Injury crashes were considered for analysis; Pedestrian 0n crosswalk 2 3.
In. Ramp terminal related crashes only. . ID Sideswipe on exit ramp 2 3.2
e. hiiotoreyeles and bicycles were consrdered as velueles; 1 1 Rear end exit ramp le? tum 2 3.2
d. Cyclists on crosswalks were consedered as pedestrians; . 
e. All loss of control categories were single vehicle crashes; 12 Sidesufipe exrt ramp 13ft turn and Insrde crossroad 2 3'2
f. Same before-?after period duration for each site; 13 Run 0? ramp 2 3'2
]4 Loss of control left tum inside crossroad bay 1 1.6

. 356 months of crash data (178 months each period).

(I:

 

 

 

 

 

 

Figure 3.1 Before After collision diagrams for fatal and injury crashes

30

 

 

 

As seen in figure 3.1, the top two crash types in the before period at the conventional
diamond ramp terminals were: 1) collision of left turn movements from inside the crossroad and
the oncoming through movement, and 2) rear end collisions on the exit ramp at the intersection.
In the after period for the DDI design, the top two crash types were: 1) rear end collisions between
right turning movements on the exit ramp at the intersection, and 2) rear end collisions on the
outside crossroad approach leg to the ramp terminal. It was also observed that some other types of
crashes distributed across the different legs of the DDI ramp terminal increased, but all these
crashes were of lower severity. For instance, sideswipes at the different merging and diverging
locations, and the loss of control in the bays while making turning movements, increased with the
DDI; however, none of these types of crashes resulted in any severe injuries. Thus, the DDI design
traded a severe crash type, a right angle left turn crash, with less severe rear end, sideswipe, and
loss of control crash types. The wrong way crashes inside the crossroad between the two ramp
terminals accounted for 4.8% of the crashes occurring at the DDI ramp terminals.
3.3 Safety Effectiveness Evaluation
3.3.1 Crash Severity Analysis
The severity of crashes was studied during the before and after periods. The annual crash
frequency was calculated for each treated facility, and it was classified into three severity
categories: total crashes (TOT), fatal and injury (FI), and property damage only (PDO). Figure 3.2
shows the results of the calculations. The crash frequency for most of the facilities decreased for
all severity categories.

31

 34.1
24.2

9.9

Chestnut Exp. and Route 65 Springfield, MO

8.7

Before PDO

14.6

I-435 and Front Street Kansas City, MO

Before TOT

32.3
23.6

6.0

62.7

77.3

34.6
28.6

Before FI
After TOT

27.5
19.4

8.1
12.9
9.5
3.4

US 65 and MO248 Branson, MO

James River Exp. and National Av. Springfield,
MO

After PDO

19.4
7.0

34.7

43.3
36.3
149.6
32.7

I-270 and Dorsett Rd Maryland Heights, MO
10.1
18.6

RT-13 and I-44 Springfield, MO
6.9

0.0

After FI

54.1

20.0

36.0

116.9
75.5
65.4
54.6

34.5
27.6

40.0

60.0 80.0 100.0 120.0 140.0 160.0
Crashes/year

Figure 3.2 Crash frequencies before/after DDI implementation by facility

The crash data was aggregated across all six sites by severity type, and the annual crash
frequency was calculated as shown in figure 3.3. The crash data was classified into four severity
categories: minor injury, disabling injury, fatal, fatal and injury (FI), property damage only (PDO),
and total crashes (TOT). Figure 3.3 shows that the crash frequency decreased for minor injury,
disabling injury, and PDO crashes. There were no fatal crashes at any of the six sites before the
installation of DDI. There was one pedestrian fatality that occurred during the after period at one
site, but the details of that crash were unknown since it was a hit and run that occurred late at night.
Since the fatal crash occurred within the footprint of the DDI, it was still included in the safety
evaluation in this study. Figure 3.3 also presents the aggregate crash frequency of all injury crashes
32

 denoted by FI (fatal and injury) and the total number of crashes denoted by TOT. The percentage
reductions in crash frequency after DDI implementation were 57.7% for FI (16.8 to 7.1), 26.5%
for PDO (47.1 to 34.6), and 34.7% for TOT (63.9 to 41.7).

70.0
Before

63.9

After

60.0
47.1

Crashes/year

50.0

41.7
40.0

34.6

30.0
20.0
10.0

16.8

16.0

7.1

6.8
0.8 0.3

0.0 0.1

Disabling
Injury

Fatal

0.0
Minor
Injury

FI

PDO

TOT

Severity

Figure 3.3 Crash frequencies before/after DDI implementation all facilities

3.3.2 Naïve Method
The odds ratio and safety effectiveness were computed for three categories of crashes –
fatal and injury only crashes (FI), property damage only crashes (PDO), and total crashes (TOT).
As shown in table 3.2, the safety effectiveness results showed a 41.7% (2.9%) reduction in total
crash frequency after DDI implementation. The value in the parenthesis denotes the standard error
of the estimated safety effectiveness. The FI crash frequency experienced the greatest reduction of
63.2% (4.1%), while the PDO crash frequency decreased by 33.9% (3.7%). All reductions were
statistically significant at the 95% confidence level.
33

 Table 3.2 Naïve method results: project-level
Parameters
Observed Crashes After
Period
Expected Crashes After
Period
Expected and Observed
Difference
Odds Ratio
Safety Effectiveness (%)
95% Statistical Significance

FI

Estimates
PDO
TOT

104.00 506.00

610.00

St. Error
FI
PDO TOT
10.20 22.49 24.70

282.02 764.10 1046.12 15.99 25.99 30.51
178.02 258.10
0.37
63.2
Yes

0.66
33.9
Yes

436.12
0.58
41.7
Yes

18.96 34.37 39.25
0.04
4.1

0.04
3.7

0.03
2.9

As previously discussed, the Naïve method can only estimate the cumulative effect of all
changes that have occurred at the treatment sites during the study period. However, it is not
possible to ascertain the individual effects of the safety treatment using the Naïve method.
Variability of traffic, road user behavior, weather, and many other factors could change over time
(Hauer 1997). Nevertheless, the Naïve method still serves as a good starting point for the safety
analysis due to its statistical accuracy, and it has been frequently used in safety evaluations as it
provides a precise upper bound (Hauer 1997).
3.3.3 Empirical Bayes Method
The project-level EB method involved three choices for correlations previously discussed:
independent, fully correlated, and partially correlated. The results for the three crash severity
categories are shown in table 3.3. In table 3.3, the observed crashes, the EB expected crashes, and
the safety effectiveness values for each site are reported in different rows. The standard error values
are also reported in parenthesis next to each safety effectiveness value. The right-most column
provides the results for the entire treatment group (combination of all six sites).
Since the actual correlation among the interchange facilities is not known, the safety
effectiveness values obtained assume partial correlation can be used for determining the crash
34

 modification factors for the DDI (Bonneson 2012). The safety effectiveness values for partial
correlation are highlighted in red bold text in table 3.3, although the results from the three
correlation choices did not differ very much. For the entire treatment group (‘All Sites’ column in
table 3.3), the percentage reduction in crashes was the greatest for FI crashes at 62.6% compared
to the 35.1% for PDO and 40.8% for TOT crashes. These findings are consistent with the results
of the crash severity analysis and the Naïve method. The left turn angle crashes that were
predominant in the traditional diamond design (before period) were completely eliminated in the
DDI design (after period), which accounts for the reduction in severe crashes.
The EB results for individual sites (see table 3.3) showed that the DDI was effective at
decreasing the FI crashes at all six sites, although the reduction at the sixth site was not statistically
significant at the 95% confidence level. The PDO crashes also decreased at all six sites with the
reductions being statistically significant except for sites 3 and 6. The TOT crashes also decreased
at all six sites, and all of the reductions were statistically significant except for site 6. The lack of
statistical significance of the EB results for site 6 was due to two reasons. First, the duration of the
after period for site 6 was the smallest among all six sites at 10 months. Thus, the lack of statistical
significance can simply be the result of the small sample size. Second, the observed crash
frequencies per year before DDI (10 FI, 24 PDO, 34 TOT) and after DDI (9 FI, 24 PDO, 32 TOT)
were not considerably different.

35

 TOT

PDO

FI

Observed
29
29
22
6
11
7
Crashes
EB Expected
74
82
61
15
27
9
I1 Crashes4
SE (St.E.)5
61.0(8.1) 64.8(7.2) 63.9(8.5) 60.8(16.3) 59.6(12.4) 20.3(30.4)6
EB Expected
83
88
64
16
26
9
C2 Crashes
SE (St.E.)
65.1(7.5) 67.0(6.8) 65.4(8.4) 63.4(15.4) 57.5(13.3) 18.2(31.4)6
EB Expected
79
85
62
16
27
9
P3 Crashes
SE (St.E.)
63.2(7.8) 65.9(7.0) 64.7(8.4) 62.1(15.8) 58.6(12.8) 19.3(30.9)6
Observed
116
188
114
17
52
19
Crashes
EB Expected
164
302
119
37
98
18
Crashes
I
SE (St.E.)
29.3(9.0) 37.8(5.6) 4.4(12.5)6 53.9(11.7) 47.2(7.7)
-3.0(24.1)6
EB Expected
198
326
126
41
106
20
C Crashes
SE (St.E.)
41.5(7.7) 42.4(5.2) 9.7(12.3)6 58.4(10.7) 51.1(7.1)
3.0(22.8)6
EB Expected
181
314
123
39
102
19
P Crashes
SE (St.E.)
36.0(8.3) 40.2(5.4) 7.1(12.4)8 56.3(11.2) 49.2(7.4)
0.1(23.5)6
Observed
145
217
136
23
63
26
Crashes
EB Expected
233
383
163
52
126
27
Crashes
I
SE (St.E.)
37.9(6.6) 43.3(4.6) 16.6(9.1)6 55.8(9.6)
49.9(6.6)
4.7(19.1)6
EB Expected
274
412
172
57
132
28
C Crashes
SE (St.E.)
47.2(5.8) 47.4(4.3) 20.8(9.0) 59.7(8.9)
52.3(6.3)
7.8(18.6)6
EB Expected
254
398
167
55
129
28
P Crashes
SE (St.E.)
42.9(6.2) 45.4(4.5) 18.8(9.0) 57.8(9.2)
51.1(6.4)
6.2(18.8)6
1
Notes: I denotes independent correlation
2
C denotes full correlated
3
P denotes partial correlation
4
The expected crash values are rounded (up) to facilitate comparison with observed crash values
5
SE denotes Safety Effectiveness (%). ST.E. denotes Standard Error (%).
6
Not significant at the 95% confidence level

36

All Sites

Chestnut Exp. and
Route 65
Springfield, MO
(Site 6)

I-435 and Front Street
Kansas City, MO
(Site 5)

US 65 and MO248
Branson, MO
(Site 4)

James River Exp. and
National Ave.
Springfield, MO
(Site 3)

I-270 and Dorsett Rd
Maryland Heights, MO
(Site 2)

RT-13 and I-44
Springfield, MO
(Site 1)

Parameter

Correlation

Severity

Table 3.3 Project-level EB results

104
269
61.4(4.2)
286
63.7(4.1)
277
62.6(4.1)
506
739
31.6(3.8)
818
38.2(3.5)
779
35.1(3.7)
610
984
38.1(3.0)
1076
43.4(2.8)
1030
40.8(2.9)

 3.3.4 Comparison Group Method
The sample odds ratio is used for determining the suitability of the comparison group. For
computing the sample odds ratio, a time frame of five years was chosen (2004 to 2009) before any
DDIs in the treatment group were implemented. The mean, the standard error, and the 95%
confidence interval of the sample odds ratio were computed, and the results are shown in table 3.4.
The mean value for FI, PDO, and TOT crashes were 0.97 (0.31 standard error), 1.01 (0.20), and
1.00 (0.22), respectively, all close to 1.0. All 95% confidence intervals also included 1.0. Based
on the sample odds ratio results and confidence intervals, the comparison group was deemed to be
suitable for comparison with the treatment group following the FHWA guidelines for developing
crash modification factors (Gross et al. 2010).

Table 3.4 Comparison of treatment and control sites
Sample Odds Ratio
Mean
Standard Error
95% Confidence
Interval

FI
0.97
0.31
[0.36 1.58]

Severity
PDO
1.01
0.20
[0.62 1.40]

TOT
1.00
0.22
[0.57 1.43]

The safety effectiveness was then calculated using the comparison group (CG) method
previously discussed. The CG method produced safety effectiveness values (and standard errors)
of 60.6% (4.6%) reduction in FI crashes, 49.0% (3.0%) reduction in PDO crashes, and 52.9%
(2.5%) reduction in TOT crashes, all significant at the 95% confidence level.
The project-level safety effectiveness results from the Naïve, EB, and CG methods are
compared in table 3.5. The safety effectiveness values for each category (FI, PDO, TOT) are shown
in different rows for the three methods. Again, the standard error values are reported in parenthesis
37

 next to each safety effectiveness value. The overall safety effectiveness values for the entire
treatment group are also shown in the right-most column. The Naïve results for individual sites
shown in table 3.5 revealed that the DDI was effective at decreasing FI crashes at all six sites, PDO
crashes at five out of six sites (one site witnessed an increase that was not statistically significant),
and total crashes at all six sites. The variation in the safety effectiveness values for FI crashes
across the sites was not high. However, PDO and TOT crashes showed higher variation across the
six sites. The EB results for individual sites were previously discussed. The CG results for
individual sites, shown in table 3.5, indicated statistically significant reductions in FI crashes for
sites 1, 2, and 3 only. Site 6 actually showed an increase in FI crashes, although it was not
statistically significant. For the CG method, statistically significant reduction in PDO and TOT
crashes were observed for the first five sites. Again, site 6 showed increases in PDO and TOT
crashes that were statistically significant. In addition to the short duration of the after period and
the lack of considerable variation in the observed crash frequency before and after DDI for site 6,
one additional reason may have contributed to the CG results for site 6. The comparison site used
for site 6 witnessed higher crash reductions for FI and TOT crashes. For comparison site 6, the
observed crash frequencies per year in the before period were: 12 FI, 31 PDO, 42 TOT and in the
after period were: 2 FI, 34 PDO, 36 TOT crashes. Is summary, able 3.5 shows that DDI decreased
FI, PDO, and TOT crashes, and the results are similar across all three methods.

38

 Chestnut Exp. And
Route 65
Springfield, MO
(Site 6)

I-435 and Front Street
Kansas City, MO
(Site 5)

US 65 and MO248
Branson, MO
(Site 4)

James River Exp. and
National Ave.
Springfield, MO
(Site 3)

I-270 and Dorsett Rd.
Maryland Heights, MO
(Site 2)

RT-13 and I-44
Springfield, MO
(Site 1)

Method

Severity

Table 3.5 Project-level safety effectiveness

Naïve 63.3 (7.9) 69.5 (6.4)
64.5 (8.7)
60.0 (17.3)
59.3 (13.2)
15.1 (34.4) 1
FI
EB
63.2 (7.8) 65.9 (7.0)
64.7 (8.4)
62.1 (15.8)
58.6 (12.8)
19.3 (30.9) 1
1
1
CG
70.7 (6.6) 71.6 (6.3)
69.9 (7.8) 37.9 (29.2)
22.6 (26.7)
-195.7 (142.5) 1
1
Naïve 23.7 (9.4) 44.2 (5.1) -3.6 (13.8)
51.5 (13.0)
54.6 (6.9)
3.7 (24.3) 1
1
PDO EB
36.0 (8.3) 40.2 (5.4)
7.1 (12.4)
56.3 (11.2)
49.2 (7.4)
0.1 (23.5) 1
CG
60.9 (5.1) 60.6 (3.8)
37.4 (8.7)
44.0 (15.7)
32.4 (10.9)
-169.4 (76.7)
Naïve 37.0 (6.7) 49.7 (4.2)
20.5 (9.1)
53.6 (10.6)
55.4 (6.2)
6.2 (20.3) 1
TOT EB
42.9 (6.2) 45.4 (4.5)
18.8 (9.0)
57.8 (9.2)
51.1 (6.4)
6.2 (18.8) 1
CG
64.0 (4.0) 63.3 (3.3)
49.2 (6.1)
44.7 (13.3)
32.3 (9.9)
-163.8 (64.5)
Notes: Standard error values are shown in the parenthesis next to the safety effectiveness
1
Not significant at the 95% confidence level

All Sites
(in %)
63.2 (4.1)
62.6 (4.1)
60.6 (4.6)
33.9 (3.7)
35.1 (3.7)
49.0 (3.0)
41.7 (2.9)
40.8 (2.9)
52.9 (2.5)

3.4 Results of Site-Specific Analysis
The statistical sample size for conducting site-specific analysis was two times the one used
for project-level analysis since each interchange has two ramp terminals. Thus, a total of 12 ramp
terminals from the DDI sites and another 12 ramp terminals from the CG sites were included in
the analysis. As previously discussed, the Empirical Bayes method uses a different set of equations
for the site-specific analysis. The Naïve and CG approaches for site-specific analysis are identical
to those used for project-level analysis.
3.4.1 Naive Method
The safety effectiveness results of the Naïve method are shown in table 3.6. The FI crash
frequency experienced the greatest reduction of 64.3% (5.4%), while the PDO crash frequency
decreased by 35.6% (4.8%), and the total crash frequency decreased by 43.2% (3.8%) after DDI
implementation. The values in the parenthesis denote the standard error of the estimated safety
effectiveness. All reductions were statistically significant at the 95% confidence level.

39

 Table 3.6 Naïve method results: site-specific analysis
Parameters
Observed Crashes After Period
Expected Crashes After Period
Expected and Observed
Difference
Odds Ratio
Safety Effectiveness (%)
95% Statistical Significance

Estimates
St. Error
FI
PDO
TOT
FI
PDO TOT
57.00 280.00 337.00 7.55 16.73 18.36
158.76 433.79 592.55 12.15 19.95 23.36
101.76 153.79 255.55 14.31 26.04 29.71
0.36
64.3
Yes

0.64
35.6
Yes

0.57
43.2
Yes

0.05
5.4

0.05
4.8

0.04
3.8

3.4.2 Empirical Bayes Method
The safety effectiveness was calculated following the site-specific EB methodology
previously described. The results of EB method for individual sites and all sites combined are
shown in Table 3.7. The EB method produced safety effectiveness values (and standard errors) of
67.8% (4.7%) reduction in FI crashes, 53.4% (3.5%) reduction in PDO crashes, and 56.6% (2.8%)
reduction in total crashes, all significant at the 95% confidence level. Additionally, the results for
individual sites (ramp terminals) showed significant reduction at a majority of the sites. The ramp
terminal of the DDI at Chestnut Exp. and Route 65, in Springfield, Missouri, had negative safety
effectiveness results for PDO and TOT, but the results were not significant.

40

 Table 3.7 Site-specific EB results

9
5

29
21

68.8 (11.2)
76.5 (11.1)

W

9

34

73.7 (9.6)

S

6

18

67 (14.3)

N

12

26

E

2

5

54.1 (14.6)
58.0
(29.7)4

W

1

7

85.6 (14.1)

E

1

5

79.8 (19.8)

W

0

5

100.0 (0.3)

65
103
81

47

166

44
43

36
67

1

9

9
5

22
23

4

17

19.9 (15.3)4
69.1 (6.6)
60.6 (8.4)
71.8 (4.8)
-22.6
(26.6)4
36 (12.4)
88.8 (10.9)
59.7 (14.6)
78.0 (10.1)
76.6 (12.0)

SE (St.E.)

N
E

52
32
32

EB Expected Crashes

54.6 (15.8)

Observed Crashes

22

TOT

SE (St.E.)

10

EB Expected Crashes

I-435 and Front
Street
Kansas
City, MO

S

Observed Crashes

US 65 and MO248
Branson, MO

SE (St.E.)3

I-270 and Dorsett Rd
Maryland Heights,
MO
James R. Exp. and
Nat. Ave.
Springfield, MO

EB Expected Crashes2

RT-13 and I-44
Springfield, MO

PDO

Observed Crashes

Location

FI

Ramp Terminal1

Severity

62
41
37

87
132
102

28.7
(11.7)
69.0 (5.6)
63.9 (7.0)

56

201

50
55

54
93

3

14

10
6

29
28

4

22

72.1 (4.3)
7.5 (17.8)4
41.1 (9.6)
78.1
(12.8)
65.9
(11.6)
78.3 (9.1)
82.1 (9.1)

67.5
(23.0)
14.1
-100.5
-61.4
W
2
2
(60.7)4
9
4
(75.8)4
11
7
(54.1)4
All Sites Analysis
57 177 67.8 (4.7) 280 599
53.4 (3.5)
337 776 56.6 (2.8)
Notes: The expected crash values are rounded (up) to facilitate comparison with the observed crash values increase in
crashes
1
The ramp terminal designation was according to the location with respect to the center of the interchange and the
orientation. N = North, S = South, E = East, and W = West
2
SE denotes Safety Effectiveness (%). ST.E. denotes Standard Error (%).
3
Not significant at the 95% confidence level
Chestnut Exp. and
RT 65 Springfield,
MO

E

0

2

100.0 (0.0)

2

5

55.9 (31.1)

2

6

3.4.3 Comparison Group
The mean value of odds ratio for FI, PDO, and TOT crashes were computed and found to
be close to 1.0. All 95% confidence intervals also included 1.0. These values mean that the CG
was suitable. The safety effectiveness for site-specific analysis was then calculated using the CG
method previously discussed. The CG method produced safety effectiveness values (and standard
errors) of 67.7% (5.39%) reduction in FI crashes, 47.0% (4.52%) reduction in PDO crashes, and

41

 53.3% (3.5%) reduction in TOT crashes, all significant at the 95% confidence level. A comparison
of the results of the Naïve, EB, CG methods for the site-specific analysis is presented in table 3.8.
Again, similar to table 3.5 (project-level), the site-specific analysis also shows that all three
methods (Naïve, EB, and CG) resulted in similar improvements in interchange safety due to DDI.

Table 3.8 Site-specific safety effectiveness

CG

EB

Naïve

TOT

CG

EB

Naïve

PDO

CG

EB

FI

Naïve

Ramp Terminal1

DDI Site

Severity

60.0
54.6
63.2
-10.6
19.9
22.1
12.7
28.7
36.5
(14.5) (15.8)
(14.0)
(21.9)2
(15.3)2
(15.9)2
(15.0)2 (11.7)
(11.2)
1
70.0
68.8
76.5
46.7
69.1
72.3
53.9
69.0
73.7
N
(11.1) (11.2)
(9.1)
(11.5)
(6.6)
(6.2)
(8.6)
(5.6)
(5.0)
61.5
76.5
63.0
45.8
60.6
54.2
47.9
63.9
56.0
E
(18.9) (11.1)
(19.9)
(11.7)
(8.4)
(10.3)
(10.4)
(7.0)
(9.1)
2
60.9
73.7
63.2
60.2
71.8
65.4
60.0
72.1
65.6
W
(14.8)
(9.6)
(14.8)
(6.8)
(4.8)
(6.1)
(6.3)
(4.3)
(5.6)
70.0
67
70.7
-41.9
-22.6
-35.3
0.0
7.5
9.0
S
(13.3) (14.3)
(13.9)
(32.5)2
(26.6)2
(32.3)2
(19.7)2 (17.8)2
(18.5)2
3
63.6
54.1
66.1
24.6
36
37.1
38.2
41.1
49.2
N
(11.9) (14.6)
(11.8)
(15.0)2
(12.4)
(12.9)
(10.5)
(9.6)
(8.9)
63.7
58.0
33.4
85.7
88.8
71.3
75.0
78.1
54.2
E
(25.6) (29.7)2
(52.6)2
(13.8)
(10.9)
(29.9)
(14.7)
(12.8)
(28.2)2
4
83.3
85.6
73.2
42.0
59.7
13.8
52.4
65.9
32.2
W
(15.9) (14.1)
(28.2)
(21.3)
(14.6)
(33.0)2
(16.4)
(11.6)
(24.1)2
81.8
79.8
61.3
79.2
78.0
55.6
79.3
78.3
58.4
E
(17.3) (19.8)
(41.0)2
(9.6)
(10.1)
(21.0)
(8.7)
(9.1)
(18.0)
5
100.0
100.0
100.0
75.0
76.6
46.1
78.4
82.1
56.1
W
(0.0)
(0.3)
(0.0)
(12.9)
(12.0)
(28.7)2
(11.1)
(9.1)
(23.2)
100.0
100.0
100.0
46.6
55.9
-75.0
55.5
67.5
-34.4
E
(0.0)
(0.0)
(0.0)
(37.7)2
(31.1)2
(135.0)2 (31.4)2 (23.0)
(102.3)2
6
11.0
14.1
-284.6
-125.2
-100.5
-899.8
-83.5
-61.4
-654.5
W
(62.5)2 (60.7)2 (320.5)2
(89.0)2
(75.8)2
(435.3)
(64.5)2 (54.1)2
(287.7)
All
64.3
67.8
67.7
35.6
53.4
47.0
43.2
56.6
53.3
Sites
(5.4)
(4.7)
(5.39)
(4.8)
(3.5)
(4.52)
(3.8)
(2.8)
(3.5)
Notes: Standard error values are shown in the parenthesis next to the safety effectiveness
1
Ramp terminal designation was according to the location with respect to the center of the interchange and the
orientation. N = North, S = South, E = East, and W = West
2
Not significant at the 95% confidence level
S

42

 Chapter 4 Conclusions
This research project addressed the safety evaluation of the Diverging Diamond
Interchanges using crash and operational data from 6 interchanges in Missouri. Missouri was ideal
for such a study because it was the first state to implement DDIs in the U.S., thus significant after
treatment data was available. Collision diagram analysis was conducted to determine the crash
type and frequency at conventional diamond interchanges and DDIs. A safety effectiveness
evaluation was performed using crash rates by severity, and three observational before and after
studies were used at the interchange (project-level) and ramp terminal (site-specific) levels: Naïve,
Empirical Bayes (EB), and Comparison Group (CG).
The collision diagram analysis revealed that right angle crashes were predominant in the
before period at the ramp terminals of a conventional diamond. Specifically, 34.3% of ramp
terminal-related fatal and injury crashes occurred due to collisions between the crossing left turn
from inside the crossroad and the oncoming through traffic. Due to the crossover design, the DDI
completely eliminated this crash type from occurring. One of the potential concerns of a DDI is
the possibility of wrong-way crashes. This study found that 4.8% of all fatal and injury crashes
occurring at the ramp terminal of a DDI were wrong-way crashes. The review of remaining crash
types found that the DDI exchanged high severity crash types, such as those occurring at a
conventional diamond, for lower severity crash types.
The three before-after safety evaluation methods produced consistent results. The DDI
design replacing a conventional diamond decreased crash frequency for all severities. At the
project-level, the most significant crash reduction was observed for fatal and injury crashes –
63.2% (Naïve), 62.6% (EB) and, 60.6% (CG). Property damage only crashes reduced by 33.9%
(Naïve), 35.1% (EB), and 49.0% (CG). The total crash frequency also decreased by 41.7% (Naïve),

43

 40.8% (EB), and 52.9% (CG). Similarly, in the site-specific analysis, the highest crash reduction
was observed for fatal and injury (FI) crashes – 64.3% (Naïve), 67.8% (EB), and 67.7% (CG).
Property damage only crashes were reduced by 35.6% (Naïve), 53.4% (EB), and 47.0% (CG).
Total crash frequency also decreased by 43.2% (Naïve), 56.6% (EB), and 53.3% (CG). The safety
effectiveness results for the individual sites also demonstrated that FI, PDO, and TOT crashes
decreased at most sites after DDI implementation. This study documented the safety benefits of
DDI, which complements the existing knowledge on the operational benefits of DDI.

44

 References
AASHTO, Highway Safety Manual (HSM). American Association of State Highway and
Transportation Officials, 2010.
Bared, J. G., P. Edara, and R. Jagannathan. Design and Operational Performance of Double
Crossover Intersection and Diverging Diamond Interchange. In Transportation Research
Record: Journal of the Transportation Research Board, No. 1912, Transportation Research
Board of National Academies, Washington, D.C., 2006, pp 31-38.
Bauer, K., and D. Harwood. Statistical Models of Accidents on Interchange Ramps and SpeedChange Lanes. FHWA-RD-97-106. Federal Highway Administration, Washington, D.C.,
1998.
Bonneson J.A., S. Geedipally, and M. P. Pratt. Safety Prediction Methodology and Analysis Tool
for Freeways and Interchanges. National Cooperative Highway Research Program, Project
17-45, National Cooperative Highway Research Program Transportation Research Board
of The National Academies, Washington, D.C., 2012.
Chilukuri, V., S. Smith, T. Michael, and T. Ryan, 2011. Diverging Diamond Interchange
Performance Evaluation (I-44 and Route 13). Report No. OR11-012, U.S. Department of
Transportation, Federal Highway Administration, Missouri Department of Transportation.
Chlewicki, G. New Interchange and Intersection Designs: The Synchronized Split-Phasing
Intersection and the Diverging Diamond Interchange. Presented at the 2nd Urban Street
Symposium, Anaheim, California, 2003.
Chlewicki, G., 2013. Learning from Eight Operational Diverging Diamond Interchanges in the
Unites States. Presented at the 92nd Annual Meeting of the Transportation Research Board,
Washington, D.C.
Edara, P., J.G. Bared, and R. Jagannathan. Diverging Diamond Interchange and Double Crossover
Intersection: Vehicle and Pedestrian Performance. Presented at the 3rd International
Symposium of Highway Geometric Design, 2005 Chicago, IL.
FHWA, Signalized Intersections: Informational Guide, Publication No. FHWA-HTR-04-091U.S.
Department of Transportation. Federal Highway Administration., Washington, DC, 2004
FHWA, Roundabouts: An Informational Guide (Second Edition). NCHRP Report 673, U.S.
Department of Transportation, Federal Highway Administration, Transportation Research
Board, 2010.
FHWA, Diverging Diamond Interchange Informational Guide. Report No. FHWA-SA-14-067,
U.S. Department of Transportation, Federal Highway Administration, Washington, DC,
2014.

45

 Gross, F., B. Persaud, and C. Lyon. A Guide to Develop Quality Crash Modification Factors.
Report No. FHWA-SA-10-032. FHWA, U.S. Department of Transportation, 2010.
Hauer, E. Observational Before-After Studies in Road Safety: Estimating the Effects of Highway
and Traffic Engineering Measures on Road Safety. Department of Civil Engineering,
University of Toronto, Emerald Group Publishing Limited, United Kingdom, 1997.
Hauer, E., D.W. Harwood, F. Council, and M. Griffith. Estimating Safety by the Empirical Bayes
Method: A Tutorial. In Transportation Research Record, No. 1784, TRB, National
Research Council, Washington, D.C., 2002, pp. 126-131.
Hughes, W., R. Jagannathan, D. Sengupta, and J. Hummer. Alternative Intersections and
Interchanges: Informational Report. Publication FHWA-HRT-09-060. FHWA, U.S.
Department of Transportation, 2010.
Hummer, S. E., J. E. Haley, R. L. Ott, R. S. Foyleand, and C. M. Cunningham. Superstreet Benefits
and Capacities. Report No. FHWA/NC/2009-06. North Carolina Department of
Transportation, U.S. Department of Transportation, Federal Highway Administration,
2010.
Inman, V.W., 2007. Driver’s Evaluation of the Diverging Diamond Interchange. Report No.
FHWA-HRT-07-048. Federal Highway Administration, U.S. Department of
Transportation.
MoDOT, Interstate 435 and Front Street Interchange Improvements. Missouri Department of
Transportation.http://www.modot.org/kansascity/major_projects/Front_Street_DDI.htm.
Accessed: May 19, 2014.
STARS, Missouri Uniform Crash Report Preparation Manual (Version 2.0). The Statewide
Traffic Accident Record System (STARS), 2012.
Vogt, A. Crash Models for Rural Intersections: Four-Lane by Two-Lane Stop-Controlled and
Two-Lane by Two-Lane Signalized. Report No. FHWA-RD-99-128. Federal Highway
Administration, McLean, Virginia, 1999.

46