5.0     USE OF ACCIDENT HISTORY DATA TO HIGHLIGHT PARTICULAR INTERSECTION-LEG CONCERNS

Accident history data are used in IDRM to highlight or call special attention to concerns that are accompanied by a related accident pattern, as defined by specific traffic movements.  Accident history data are considered for intersection-leg concerns, but not for general intersection concerns (see Table 3), which are not necessarily associated with particular traffic movements.  The following discussion indicates how this is done.

5.1       Role of Site-Specific Accident History Data in IDRM

FHWA’s original concept, at the outset of this project, was to have a separate expert system to evaluate the accident history at an intersection, identify accident patterns, and recommend appropriate treatments.  The research team recommended, and FHWA agreed, that it would be better to have a single expert system that integrated both the diagnostic review of a design and the consideration of accident history data.  A geometric and operational diagnostic review of an intersection can be performed alone when no accident history data are available.  However, when accident history data are available, those data should be used to enhance the diagnostic review of the intersection.

IDRM will be used primarily by highway agency engineers and their consultants.  It is assumed that, in at least some cases, IDRM can be implemented in an environment where the agency’s accident history data are available online.  In other cases, accident history data may need to be obtained from an external system and entered manually by the user.

5.2       Conceptual Approach to Use of Site-Specific Accident History Data in IDRM

The conceptual approach to consideration of accident history data in IDRM is based on recognition that most of the concerns identified in a diagnostic review are associated with the risk of potential collisions involving specific pairs of vehicle movements.  For example, the concern Insufficient ISD to right (Case B1) for NB leg is intended as an indicator of the potential risk of collisions between northbound (NB) left-turning vehicles and westbound (WB) through vehicles.  While less common, there is also a risk of collision between NB left-turning vehicles and WB left-turning vehicles.  Therefore, when this concern is identified, it makes sense to check the accident history data to determine whether there is a pattern of collisions involving these two pairs of through or turning movements.  If a pattern of related crashes is evident, this indicates a need for heightened interest in improving ISD at the intersection because the ISD concern identified in the diagnostic review is supported by an existing accident pattern.  It should be noted that the conceptual approach, in effect, places primary emphasis on the geometric and operational reviews, using accident data in an important but supporting (not primary) role.  A designer need not abandon IDRM because she or he is lacking accident data of sufficient quality; on the other hand, having such data makes an enhanced and “richer” review possible.

The heightened interest in concerns that are accompanied by related accident patterns should be communicated to the IDRM user in the list of concerns identified by the diagnostic review either by presenting the ISD concern in a contrasting color, in a bold font, or perhaps with a special message.  This would contrast with other concerns, not confirmed by accident patterns, which would be presented in a normal color or font.

5.3       Identification of Patterns of Specific Collision Types

A key issue for implementation of this approach is the decision of how to identify accident patterns. This will be accomplished with a matrix of collision types and a threshold value for the minimum number of accidents that constitutes an accident pattern. 

The concerns identified by IDRM are related primarily to the risk of multiple-vehicle collisions (i.e., collisions involving two or more motor vehicles).  The traffic movement (or intended movement) being made by each vehicle will be used in classifying the collision types, just as an analyst does in preparing a collision diagram.  At a four-leg intersection, there are 12 possible traffic movements:

• Northbound (NB) leg (left turn, through, and right turn).
• Eastbound (EB) leg (left turn, through, and right turn).
• Southbound (SB) leg (left turn, through, and right turn).
• Westbound (WB) leg (left turn, through, and right turn).

Thus, for any accident involving two vehicles, there are a total of 12 x 12 = 144 possible combinations of traffic movements.  Each multiple-vehicle accident at the intersection being evaluated will be classified into one of these 144 cells.  At a three-leg intersection, there are only 6 x 6 = 36 combinations of traffic movements, so 108 of the 144 cells do not exist.  At multileg intersections, there are more than 144 combinations of movements, so specific procedures for handling multileg intersections will need to be developed.

In the case of one concern, Loss of control potential due to frequent braking for NB leg, single-vehicle accidents are also relevant.  This creates 12 new classifications of interest – single-vehicle accidents involving each of the 12 traffic movements at a four-leg intersection.  Thus, there are a total of 12 x 13 = 156 accident categories of interest, including single-vehicle accidents.

Some IDRM concerns are associated with only 1 of the 156 combinations of traffic movements.  For example, the concern Insufficient ISD to right (Case B1) for NB leg is associated specifically with the NB left-turn and WB through movement.  Other IDRM concerns are associated with multiple combinations of movements.  For example, the concern Increased crossing distance for NB leg deals with potential conflicts between a NB through vehicle and the following six traffic movements:

• EB through movement.
• EB left-turn movement.
• SB left-turn movement.
• WB through movement.
• WB left-turn movement.
• WB right-turn movement.

Thus, in assessing the presence of an accident pattern for this concern, the accident frequencies for 6 of the 156 cells should be combined (i.e., summed).  These are the six cells that represent the combination of the NB through movement with each of the six other movements identified above.

Accidents will be classified by the direction of travel and intended movement of the one vehicle for single-vehicle accidents and of the two vehicles for two-vehicle collisions. For collisions involving more than two vehicles, the movements associated with Vehicles 1 and 2 in the accident data will be used to classify the accident.  Some States always put the offending vehicle first; others number vehicles in the order in which they became involved in the accident.  In either of these cases, the use of Vehicles 1 and 2 would appear to be appropriate.

IDRM will recommend that accident patterns be identified from 3 to 5 years of accident history data.  Most analysts agree that 3 years is a desirable minimum for an intersection analysis.  Five years is a practical maximum because data are often not available for longer periods and, in many cases, modifications to the intersection make it inappropriate to use accident history data older than 5 years.

A threshold value will be used to determine what minimum number of accidents constitutes an “accident pattern.”  Clearly, the term “accident pattern” implies that the threshold value should be greater than one accident.  One accident, by itself, can hardly constitute a “pattern.” An appropriate threshold would relate to the overall expected accident frequency given traffic volume, location, type of traffic control, etc. The current application of IDRM is for intersections on rural two-lane highways, which experience relatively few accidents.  The threshold value, therefore, cannot be substantially greater than one, or very few “accident patterns” would be identified for normal conditions.

A review of expected accident rates at signalized and unsignalized intersections on two-lane highways was undertaken using the safety prediction models in Report No. FHWA-RD-99-207.   This review found that very few two-lane highway intersections, even those with relatively high volumes, are likely to have three or more accidents during a 3- to 5-year period in a pattern associated with a specific IDRM concern.  Therefore, for application to intersections on two-lane rural highways, the threshold number of accidents to identify an accident pattern should be two.  (NOTE:  If the IDRM concept should, at some future time be applied to higher volume roadways, such as multilane urban/suburban arterials, a threshold higher than two accidents will be appropriate.  Furthermore, a threshold that varies with traffic volume level might also be appropriate in that situation.)

5.4       Algorithm for Use in IDRM

1. Standard IHSDM crash data capabilities are available to IDRM if the user elects to consider actual history data for a particular intersection.
2. As IDRM evaluates the intersection, each time an intersection-leg-related concern is identified based on the geometric design review, check the associated accident data as follows:
   • Consult Table 25 to determine the specific accident classifications associated with that concern.  Use of Table 25 is described in more detail below.
   • Sum the number of accidents for all classifications shown in Table 25 for the concern in question.
   • If the total number of accidents for all classifications associated with the concern in question is two or more, then highlight the concern to indicate that it has been confirmed from accident data.  If the total number of accidents is zero or one, then do not highlight the concern.

5.5       Use of Table 25 to Identify Accident Patterns of Interest

Table 25 presents the following data for each IDRM intersection-leg concern:

• Description of concern.
• Type of leg to which the concern applies.
• List of traffic movement pairs to which the concern applies (i.e., accident patterns).
• Number of traffic movement pairs to which the concern applies.

The table, as presented, applies to four-leg unsignalized intersections, but it can be readily adapted to signalized and three-leg intersections.

The type of leg column indicates the type of leg to which specific concerns apply at unsignalized intersections.  For example, the table shows that, for the concern Insufficient ISD to right (Case B1) for Leg A, Leg A would normally be a minor approach.  At signalized intersections, all approaches are major approaches, so if this concern is identified, it must pertain to a major approach (probably for the case of flashing operation of a signal).

The traffic movement pairs column lists the traffic movement pairs that should be considered in assessing the concern.  The number of traffic movement pairs to be considered for particular concerns ranges from 1 to 23, out of the 156 combinations of interest.  The intersection legs for a four-leg intersection are read as follows:

• Leg A is the leg for which the concern in question was identified.
• Leg A+1 is the next leg clockwise from Leg A (i.e., the roadway to the left of Leg A).
• Leg A+2 is the second leg clockwise from Leg A (i.e., the roadway opposite Leg A).
• Leg A+3 is the third leg clockwise from Leg A (i.e., the roadway to the right of Leg A).

Thus, if Leg A were the NB leg, then Legs A+1, A+2, and A+3 are the EB, SB, and WB legs, respectively.

At a three-leg intersection, the same rules apply, except that in applying the rules, the “missing” minor-road leg must be counted.  For example, if Leg A is the minor-road leg at a three-leg intersection, then:

• Leg A+1 is the major-road leg to the left.
• Leg A+2 is the non-existent minor-road leg opposite Leg A.
• Leg A+3 is the major-road leg to the right.

Thus, if Leg A were the NB leg, then Legs A+1 and A+3 are the EB and WB legs, respectively, and Leg A+2, the SB leg, does not exist.

By contrast, if the concern applied to the major-road leg to the left of the minor-road leg at a three-leg intersection, then Leg A is a major-road leg and the movement pairs in the table would be interpreted as follows:

• Leg A+1 is the non-existent major-road leg.
• Leg A+2 is the major-road leg to the right.
• Leg A+3 is the minor-road leg.

In this situation, if Leg A were the EB leg, then Legs A+2 and A+3 would be the WB and NB legs, respectively, and the SB leg would not exist.

When applied to three-leg intersections, Table 25 includes some traffic movement pairs that do not exist.  However, this should not cause any problem because, by definition, the number of accidents classified for each of these traffic movement pairs should be zero.

5.6       Summary

In summary, the approach to using site-specific accident history data in IDRM is as follows:

• Accident history data for 3 to 5 years at each intersection evaluated will be used whenever available.  It is desirable, but not essential, that the accident data should be accessible to IDRM in electronic form.
• Only accidents that occur within 76 m (250 ft) of the intersection and are classified as “at intersection” or “intersection-related” are considered.
• The evaluation will focus on single- and multiple-vehicle collisions, each of which will be classified into one of 156 cells based on the movements of the involved vehicles.
• Whenever an intersection-leg concern is identified in the diagnostic review, IDRM will check the accident history for collisions involving traffic movements related to that concern.  If the number of total collisions for those related traffic movements is two or more, greater attention will be directed toward that concern by presenting it in the list of concerns in a contrasting color, or a bold font, or with a special message.
• Table 25 defines which traffic movement pairs are associated with each intersection-leg concern.
• For four-leg unsignalized intersections, all concerns shown in Table 25 are considered.  For signalized intersections, only the concerns for which Leg A Type is "major" or "either" are considered.  For three-leg intersections, any traffic movement pair that involves a non-existent traffic movement should be ignored.  However, if an approach has flashing red operation at night, while a conflicting approach has flashing yellow operation at night, the concerns for which Leg A Type is "minor" should also be considered.

Table 25.  Accident Patterns for Increased Emphasis on Particular Intersection-Leg Concerns at Four-Leg Intersections

Table 25.  Accident Patterns for Increased Emphasis on Particular Intersection-Leg Concerns at Four-Leg Intersections (continued)

Table 25.  Accident Patterns for Increased Emphasis on Particular Intersection-Leg Concerns at Four-Leg Intersections (continued)

Table 25.  Accident Patterns for Increased Emphasis on Particular Intersection-Leg Concerns at Four-Leg Intersections (continued)

Table 25.  Accident Patterns for Increased Emphasis on Particular Intersection-Leg Concerns at Four-Leg Intersections (continued)

¹SV – single-vehicle

Previous | Table of Contents | Next

TFHRC Home | FHWA Home | Feedback United States Department of Transportation - Federal Highway Administration