OPPORTUNITIES FOR REDUCTION OF FATALITIES IN VEHICLE-GUARDRAIL COLLISIONS

Transcription

1 OPPORTUNITIES FOR REDUCTION OF FATALITIES IN VEHICLE-GUARDRAIL COLLISIONS Hampton C. Gabler Douglas J. Gabauer Virginia Tech Center for Injury Biomechanics Blacksburg, VA ABSTRACT In the United States in 2005, there were 1,189 fatal crashes and 35,000 injurious crashes into guardrails. Current efforts to reduce fatalities occurring in guardrail collisions have focused on frontal oblique collisions of cars and light trucks into guardrail. These crashes however represent a diminishing target population for fatality reduction. This paper examines the current opportunities for reducing fatalities in guardrail collisions in the United States. The analysis was based upon crash data from the ity Analysis Reporting System (FARS) and the National Automotive Sampling System General Estimates System (GES) for the years The greatest opportunity for fatality reduction is the protection of motorcyclists, who now account for 32% of guardrail fatalities, and car and light truck occupants in side impact, who now comprise 14% of all guardrail fatalities. Together, protection of motorcycle riders and protection of car and light truck occupants in side impacts account for nearly half of all fatalities (46%) which occur in vehicle-guardrail collisions. Additional targets for fatality reduction include light truck rollover and collisions with guardrail ends. Guardrails are designed to protect vehicle occupants from trees, poles, side slopes and other hazards their vehicle may encounter in run-off road crashes. Unfortunately, a guardrail is not always a forgiving object to strike. In the United States in 2005, there were 1,189 fatal crashes and 35,000 injurious crashes into guardrails [NHTSA, 2006a]. 51 st ANNUAL PROCEEDINGS ASSOCIATION FOR THE ADVANCEMENT OF AUTOMOTIVE MEDICINE October 15 October 17, 2007

2 Existing research on the risk of guardrail crash injury and fatality has been largely based on highway crash experience of a nonairbag fleet. Hunter et al. (1993) used detailed longitudinal barrier crash investigation data to examine occupant injury differences between guardrail end treatment and midsection impacts. A major finding was that end terminal hits were more likely to result in driver injury. Viner (1995a) examined the risk of rollover in 16,000 run-off-road crashes using linked crash and roadway data from Illinois. Although side slopes and ditches were found to be the largest cause of vehicle rollover, guardrails were found to have a higher than average overturn rate compared to other fixed objects in rural areas. Other researchers have found that approximately half of longitudinal barrier fatal crashes involved vehicle rollover [Bligh and Mak, 1999] and that light truck-type vehicles were more likely to rollover [Viner, 1995b]. In side impacts with roadside hardware, Troxel et al. (1991) found that guardrail ends produced more serious injuries than side impacts into guardrail midsections. For guardrail crashes in general, Michie and Bronstad (1994) estimated fatality risk based on unreported collision estimates. This research suggested that only 6 percent of all guardrail collisions involved any occupant injury or fatality. Elvik (1995) presented a metaanalysis including 32 previous longitudinal barrier studies that suggested that both median and roadside barriers reduce the probability of fatal injury by 20 and 45 percent, respectively. None of these studies, however, have included data from a primarily airbag-restrained fleet. CRASH TESTING The performance of guardrails in passenger vehicle collisions is evaluated in crash tests conducted according to National Cooperative Highway Research Program (NCHRP) Report 350 [Ross et al, 1993]. In its most severe test of passenger vehicle impacts with guardrail, NCHRP 350 crash tests prescribe a 100 km/hr impact into guardrail at an impact angle of 25 degrees. Test vehicles include both a small car (820 kg) and a large pickup truck (2000 kg). Guardrail crash performance is evaluated using several factors: (1) acceptable levels of occupant deceleration, (2) guardrail integrity, i.e., the vehicle must not penetrate through the rail, (3) the vehicle must remain upright and stable after the test, i.e. no rollover, and (4) the guardrail must safely redirect the vehicle back onto the highway, i.e. the post-impact exit angle must be a prescribed fraction of the pre-crash impact angle. NCHRP 350 also prescribes crash tests of heavy trucks into longitudinal barrier. Heavy truck test vehicles include a single unit truck (8000 kg), a tractor trailer (36,000 kg), and a tanker-trailer (36,000 kg). The goal of these tests is to evaluate guardrail integrity, as opposed to occupant risk, under the extreme loads applied by heavy vehicles. The only test of occupant injury risk in NCHRP 350 is a frontal or angled frontal oblique crash test of a passenger vehicle colliding with 32

3 a guardrail. The concern is that, with the widespread availability of frontal airbags, frontal impacts may no longer be the predominant source of fatality in guardrail crashes. Needed is a new analysis of U.S. crash data to identify the priorities for fatality reduction in vehicle-guardrail collisions. OBJECTIVE The objective of this paper is to determine the opportunities for reducing fatalities in guardrail collisions in the current US fleet. METHODS DATA SOURCES The data for this study was extracted from the ity Analysis Reporting System (FARS) database and the National Automotive Sampling System General Estimates System (GES) for the years FARS is a comprehensive census of all US traffic related fatalities which occur within 30 days of a traffic crash [NHTSA, 2006a]. GES is a database containing information on approximately 60,000 randomly sampled police reported accidents each year [NHTSA, 2006b]. Cases from GES are assigned weights that can be used to estimate the number of similar non-sampled crashes that may have occurred that year. FARS was analyzed to determine the frequency and characteristics of fatal guardrail crashes. The GES sample included both fatal and non-fatal crashes into guardrail and was analyzed to determine the number of persons who were exposed to guardrail crashes. FARS and GES data from the calendar years were aggregated for much of the analysis which follows in order to increase the number of cases available for analysis. FARS data were used to examine historical fatality trends over this fifteen year time period by vehicle type. CASE SELECTION Only cases for which a guardrail collision was denoted as the most harmful event were included in the analysis. The most harmful event corresponds to the single event which was most likely to have caused fatality or the most severe injury in a crash [NHTSA, 2006c]. In the event of a crash which does not involve injury, the most harmful event is that event which caused the most property damage. Most harmful event was not missing for any of our FARS cases. In cases where the GES value for most harmful event was missing, GES provides an imputed value for this data element using hotdeck imputation, a form of single imputation [Shelton, 1993]. In our sample, only 0.4% (37 of 9264) of the GES cases had a missing harmful event code requiring use of an imputed value. DATA Crash mode was determined using the general area of the subject vehicle damaged in the most harmful event. In FARS, general area of damage was determined using the Principal Impact Point 33

4 field. In GES, the crash mode was determined by use of the General Area of Damage field in the Event table which corresponds to the most harmful event. Frontal impacts were defined as crashes in which damage in the most harmful event was to the front, left front corner or right front corner of the vehicle. Side impacts were defined as crashes in which the damage was to the left or right side of the vehicle. Rear impacts were defined as crashes in which damage in the most harmful event was to the rear, left rear corner or right rear corner of the vehicle. Rollovers were defined as any crash in which the vehicle overturned no matter where damage to the vehicle was located. The FARS datasets for 2004 and 2005 contain additional data elements which permit an expanded analysis of vehicle-guardrail crashes. The new data elements include a six event crash sequence, which describe chronology of events in the crash, and an improved description of the guardrail component struck in a crash. Prior to 2004, FARS aggregated guardrail length of need (e.g. the portion between the end treatments) and guardrail end treatments into a single guardrail category. Beginning in 2004, FARS separated guardrail impacts into two components: guardrail face impacts and guardrail end impacts. Guardrail face impacts have the same code as the pre-2004 guardrail object struck code. As of 2005, however, not all states had adopted the new coding scheme. New Jersey, for example, did not modify its police accident report to use these new categories until A list of states using the new coding scheme was not available for this study. As a conservative alternative, guardrail impacts for any state not having a guardrail end impact code in were arbitrarily listed as Componentunspecified. Because only two years from our year dataset distinguished between the guardrail face and the guardrail end, both guardrail categories from FARS data were aggregated into a single guardrail group for the FARS analysis. FARS was used to estimate the distribution of guardrail impacts by component struck. This dataset was also used to determine if rollover happens before or after a guardrail impact. Exposure data for this analysis was not available as GES does not contain similar data on guardrail component struck. ANALYSIS We conducted a descriptive analysis of occupants exposed and fatally injured in vehicle-guardrail collision. The analysis was conducted by vehicle type and guardrail crash characteristics first for all vehicles and then for cars, light trucks, and vans only. The analysis of vehicle-guardrail collisions by vehicle type was conducted to identify opportunities for fatality reduction in vehicle types, e.g. heavy trucks, buses, and motorcycles, for which occupant risk is not measured in NCHRP 350 crash testing. Because the composition of the vehicle fleet may change with time, this analysis will examine FARS to determine the trends in guardrail fatalities by vehicle type over this 15 year time span. 34

5 The remaining analyses included only cars and LTV collisions with guardrail. When grouped together, cars, light trucks, and vans will be referred to as passenger vehicles. First, the distribution of exposed occupants and fatalities in passenger vehicles will be examined by crash mode, i.e., frontal, side, rear, or rollover crashes. The distribution by crash mode will then be further disaggregated by vehicle type to identify whether any crash modes are more dangerous for particular vehicle types. Contingency tables were used to statistically test for differences in occupant fatality risk. Table 1. Vehicle-Guardrail Crash Data Set (FARS, GES ) Variable ities Occupants Exposed (weighted) Occupants Exposed (unweighted) All 2,979 3, ,864 9,264 Calendar Year ,268 1, ,657 1, ,009 1, ,325 1, ,360 1, ,245 1,485 Crash Mode Front 1,501 1, ,662 5,608 Side Impact ,339 2,448 Rollover , Rear , Other , Vehicle Body Type Cars 1,232 1, ,430 5,651 LTVs ,176 2,566 Motorcycles 980 1,003 8, Heavy Trucks , Buses 1 5 3, Other/Unknown ,

6 RESULTS The final data set consisted of 3,133 cases from FARS , and a total of 855,864 weighted cases from GES Table 1 shows the distribution of data for FARS and GES by vehicle type, crash mode and calendar year. LTV refers to light trucks and vans, a category which includes pickup trucks, sport utility vehicles, minivans, and full sized vans. The other vehicle category includes heavy trucks, buses, and special purpose vehicles. DISTRIBUTION OF FATALITIES BY VEHICLE TYPE Figure 1 presents the distribution of fatalities by vehicle body type in collisions in which a guardrail impact was the most harmful event. 80% 73% 70% 60% % Occupants Exposed to Guardrail collisions % Guardrail s % 50% 40% 30% 20% 42% 22% 22% 32% 10% 0% 4% 4% 0.7% Cars LTV Motorcycles Other Figure 1. ities vs. Occupants Exposed to Guardrail Collisions by Vehicle Body Type (FARS , n=3,133; GES , n=855,864) As shown in Figure 1, car occupants comprised 42% of all fatalities resulting from a guardrail collision in Following car occupants were motorcycle riders with 32% of all fatalities in this crash mode. This was surprising as motorcycle riders were involved in less than 1% of all guardrail crashes, but accounted for over 30% of all guardrail crash fatalities in In terms of fatalities per person exposed to guardrail crashes, motorcycle riders were dramatically overrepresented in number of fatalities resulting from guardrail impacts. 36

7 As shown in Figure 2, the problem of motorcycle fatalities in guardrail collisions is a growing problem. From , the number of car occupants who were fatally injured in guardrail collisions declined by 31% from 251 deaths in 2001 to 171 deaths in In contrast, the number of fatally-injured motorcyclists increased by 73% from 129 to 224 fatalities during the same time period. In 2001, fatalities from motorcycle-guardrail collisions exceeded the number of deaths from LTV-guardrail collisions. In 2005, motorcycle rider fatalities (224) resulting from guardrail collisions surpassed car fatalities (171) for the first time Cars Motorcycles LTVs Other Vehicles 250 ities Year Figure 2. Guardrail Crash ities as a Function of Vehicle Body Type (FARS ) DISTRIBUTION OF FATALITIES BY CRASH MODE Figure 3 presents the distribution of passenger vehicle-guardrail crash fatalities by crash mode. Passenger vehicles include cars, light trucks, and vans. Frontal impacts, of the type tested in NCHRP 350 accounted for nearly half of all passenger vehicle-guardrail crash fatalities. However, guardrail collisions involving rollover comprised 27% of all passenger vehicle fatalities. Side impacts into guardrails, such as nontracking vehicles colliding with guardrail ends, accounted for 22% of all fatalities. The characteristics of fatal rollover and side impact collisions, two opportunities for fatality reduction, will be further investigated in this analysis. 37

8 Rollover 27% Frontal Impact 45% Side Impact 22% Rear Impact 4% Other Impact 2% Figure 3. Distribution of ities resulting from Passenger Vehicle-Guardrail Collisions (FARS , n = 2008) GUARDRAILS IMPACTS AND ROLLOVER Over onefourth (27%) of all fatalities in passenger vehicle-guardrail collisions involved a rollover. As shown in Figure 4, rollovers only occurred in 2% of all car-guardrail collisions, but accounted for 20% of all car-guardrail fatalities in

9 70% 64% 60% Occupants Exposed ities 50% 46% 40% 30% 28% 27% 20% 20% 10% 0% 6% 4% 2% 2% 0.3% Frontal Side Rear Rollover Other ities: Exposed: n = 608 n = 336,784 n = 354 n = 150,538 n = 57 n = 29,618 n = 260 n = 11,152 n = 30 n = 1,749 Figure 4. Car-Guardrail Collisions (FARS and GES ) Light trucks and vans are particularly at risk of rollover. In collisions with guardrails, LTVs overturn three times as frequently as cars (7% vs. 2%) in police-reported crashes. A contingency table analysis indicates that this increased frequency is statistically significant for both fatal crashes and all crashes (p-value < 0.001). As shown in Figure 5, rollovers only occurred in 7% of all LTV-guardrail crashes, but accounted for 40% of all LTV-guardrail fatalities. 39

10 70% 60% 50% 58% Occupants Exposed ities 40% 42% 40% 30% 29% 20% 13% 10% 0% 7% 5% 3% 1% 2% Frontal Side Rear Rollover Other ities: Exposed: n = 297 n = 147,617 n = 90 n = 73,724 n = 18 n = 13,222 n = 280 n = 17,319 n = 14 n = 2,696 Figure 5. LTV-Guardrail Collisions (FARS and GES ) Relative ity Risk w.r.t. Frontal Impact Car Occupants LTV Occupants Frontal Side Rear Figure 6. Relative ity Risk by Crash Mode with respect to Frontal Impact (FARS ) SIDE IMPACTS Side impacts accounted for 22% of fatalities (444 of 2008) in passenger vehicle-guardrail crashes, and 14% of all guardrail crash fatalities (444 of 3133). Figure 6 presents the fatality risk 40

11 of guardrail collisions relative to the fatality risk in a frontal collision with guardrail. The occupant of a car which side impacts a guardrail has a 30% higher probability of being fatally injured than the occupant of a car involved in a frontal impact into a guardrail. In contrast, light truck occupants have a 30% lower probability of being fatally injured in a side impact than in a frontal impact. To further investigate the relative risk shown in Figure 6, 2 x 2 contingency tables were generated for each collision and vehicle type combination. Table 2 summarizes the data comparing fatal injury risk by vehicle type for each collision mode. The row proportions are shown in parentheses below each cell count. For each comparison, the chi-square value and associated p-value is listed in the rightmost columns. Based on the p-values in Table 2, only the increased fatality risk of car occupants compared to LTV occupants in side impact guardrail crashes is statistically significant (α = 0.05). A similar contingency table analysis was performed (data not shown) comparing fatal injury risk by collision type for cars and LTVs. For cars, there was a statistically significant increase in fatality risk for side impacts relative to frontal impacts (pvalue < 0.001). For LTVs, the decrease in fatality risk for side impacts relative to frontal impacts was statistically significant (p-value < 0.001). Table 2. Collision Type by Vehicle and Injury Type: Contingency Table Summary Collision Type Front Side Rear Vehicle Type Car LTV Car LTV Car LTV Non- χ 2 P (99.82) (0.18) (99.80) (0.20) (99.76) (0.24) (99.88) (0.12) (99.81) (0.19) (99.86) (0.14) < GUARDRAIL CRASH SEQUENCE The FARS dataset was analyzed to determine the guardrail crash sequence. This 41

12 dataset contained 1008 fatal vehicle-guardrail crashes resulting in 1051 fatally-injured occupants. As shown in Table 3, the distribution of fatalities by crash mode in the FARS dataset is very similar to the FARS dataset. Table 3. Vehicle-Guardrail Crash Data Set for (FARS ) Variable ities %ities (n=1051) %ities (n=3133) All 1,008 1,051 Crash Mode Front % 50% Side Impact % 24% Rollover % 19% Rear % 3% Other % 4% Guardrail Crash Sequence No Overturn % - GR-Overturn % - Overturn-GR % - Over-GR-Over % - GR-Over-GR 8 9 1% - Unknown 9 9 1% - Motorcycles * % - Guardrail Impacts % % % % - Guardrail Component Struck Face % - End % - Not Specified % - * FARS does not allow a rollover designation for motorcycles. 42

13 In most fatal vehicle-guardrail crashes (90%), a guardrail was only struck once. In FARS , vehicles overturned in 167 crashes in which the vehicle-guardrail impact was the most harmful event. Note that this overturn count does not include motorcycles as, in FARS, motorcycle crashes are never coded as rollover regardless of whether the motorcycle overturns In 79% of the non-motorcycle crashes involving a rollover (132 of 167), the overturn occurred after the guardrail impact. GUARDRAIL COMPONENT STRUCK The guardrail component which was struck by the vehicle was coded in 821 of the FARS cases. In 81% of the fatal collisions (665 of 821), the vehicle struck the face of the guardrail rail. As shown in Table 4, the frequency of a fatal collision with a particular component varied by vehicle type. Cars struck the guardrail face in only 2/3 of the fatal crashes (193 of 291) while nearly all (96% or 311 of 325) fatal motorcycle collisions involved a collision with the guardrail face. Table 4. Distribution of by Guardrail Component Struck and Vehicle Body Type (FARS ) Vehicle Type into GR Face into GR End % of which hit GR face in this mode Car % LTV % Motorcycles % Other Vehicle % Total % As shown in Table 5 and Table, guardrail end treatments are a particular hazard in side impacts and as a rollover tripping mechanism. Over half of all cars in fatal side crashes with a guardrail struck the guardrail end. In both cars and LTVs, approximately one-third of all fatal guardrail collisions with a rollover involved an impact with the guardrail end. 43

14 Table 5. Distribution of Car by Guardrail Component Struck and Crash Mode (FARS ) Crash Mode into GR Face into GR End % of which hit GR Face Front % Side Impact % Rollover % Rear % Other % Total % Table 6. Distribution of LTV by Guardrail Component Struck and Crash Mode (FARS ) Crash Mode into GR Face into GR End % of which hit GR Face Front % Side Impact % Rollover % Rear % Other % Total % DISCUSSION In the US, motorcycle riders now account for more fatalities than the passengers of any other vehicle type involved in a guardrail collision. From , motorcycle riders accounted for 32% of all US guardrail crash fatalities, but less than 1% of all guardrail collisions.. Approximately 27% of all fatal passenger vehicle-guardrail collisions result in a rollover. Light trucks and vans are particularly at risk of rollover. In today s fleet, many light trucks have a center of gravity which is higher than the guardrail. When light trucks collide with guardrails, there is a significantly greater chance of guardrail vaulting 44

15 or roll-over [Stephens, 1996; Eskandarian, 2003]. In collisions with guardrails, LTVs overturn three times as frequently as cars (7% vs. 2%) in police-reported crashes. Because the horizontal rail is near or below the center of gravity of many light trucks and cars, guardrails can act as a tripping mechanism for rollover. To address this issue, guardrail systems are under development that have higher horizontal rails. The Midwest Guardrail System, for example, has raised the rail height from 686-mm (27-in.) to 787 mm (31 in.) [Faller et al, 2004]. For car occupants, side impacts into guardrails were substantially more dangerous than frontal impacts into guardrails (p<.001). The occupant of a car which side impacts a guardrail has a 30% higher probability of being fatally injured than the occupant of a car involved in a frontal impact into a guardrail. Side impacts are less of a threat to LTV occupants than car occupants (p<.05). LTV occupants are seated higher than car occupants. In a side impact with a guardrail, the guardrail would impact below or low on the LTV passenger compartment. Ray et al (1998) have developed crash test procedures for evaluating the side impact performance of roadside hardware. NCHRP 350 however does not mandate an acceptable performance level for side impacts into guardrails. Particularly dangerous are impacts with the end treatment of a guardrail. Guardrail end treatments are designed to breakaway under the loads which are typical of a frontal impact. Because the side of a vehicle, unlike the front, has so little structure to protect an occupant, side impacts to a guardrail end treatment can be especially dangerous. Guardrail end treatments also appear to be a potential rollover hazard. Approximately 1 in 3 fatal passenger vehicle-guardrail crashes which resulted in a rollover struck the ends of the guardrail. As guardrail ends comprise only a tiny portion of the length of a typical guardrail system, this suggests that collisions with guardrail ends are overrepresented in terms of fatality risk per unit length. LIMITATIONS One limitation of this study is its reliance on the most harmful event data element. Our study defined a guardrail impact as a collision in which the guardrail was the most harmful event. This determination of most harmful event is based upon the judgment of the FARS or GES analyst, and may vary from analyst to analyst Different types of steel guardrail systems, e.g. w-beam and cable barrier, may have very different crash performance. Our study however cannot discriminate between these systems as neither FARS nor GES distinguish between different types of steel guardrail systems. W- beam barrier is the predominant barrier used on US highways and is likely to comprise the majority of the cases in our dataset. A small number of other steel guardrails, such as box beam barriers and cable barriers, however, are also likely to have been included in the analysis. 45

16 Because the guardrail component data element was only available in FARS , this part of the analysis could only be performed with the last two years of our six year data set. Because the number of cases is small, the results should only be considered as a preliminary indication of the findings that can be expected when this study is repeated with additional future years of data. Finally, this study relies on FARS and GES which describes crashes on US highways with a US highway vehicle fleet. The results of this study may not be applicable to other national highway environments. IMPLICATIONS This study has important policy implications for efforts underway to update guardrail crash test performance guidelines, e.g. NCHRP 350. Because all guardrail systems installed on federal highways must exhibit satisfactory performance in NCHRP 350 crash tests, this test has effectively become a design specification for guardrail systems installed on US highways. NCHRP 350 prescribes only frontal or angled frontal oblique crash testing of passenger vehicles into guardrail. In addition to other evaluation criteria, the guardrail system is considered to have failed if the vehicle overturns as a result of the impact. Again, this is limited to a rollover following a frontal or angled frontal crash and does not account for overturns following other crash modes such as side impacts. NCHRP 350 type crashes represent a diminishing target population for fatality reduction. Currently, cars and light truck occupants account for 64% of vehicle-guardrail fatalities (2008 of 3133). Of the fatally injured car and light truck occupants, 72% were either involved in a frontal collision with a guardrail or were occupants of a vehicle which overturned after collision with a guardrail. Therefore, fatalities which occurred in crashes similar to NCHRP 350, i.e., frontal impacts and rollovers of cars, light trucks and vans, represent only 46% (0.72 x 0.64) of all fatalities which occur as the result of a guardrail collision. Over half of all vehicle-guardrail crash fatalities are not included in the NCHRP 350 target population. In particular, motorcycleguardrail crashes (32% of all guardrail crash fatalities) and car and LTVguardrail side impacts (14% of all guardrail crash fatalities) are an unmet safety problem in the U.S. Guardrails are not designed to protect motorcyclists. Motorcycles represent less than 1% of all police-reported crashes into guardrail. Because motorcycle crashes are rare in comparison with car and LTV crashes, guardrail design has rightly focused on protecting the occupants of cars and LTVs. However, elimination of vehicle-guardrail fatalities will not be possible unless the issue of motorcycles-guardrail collisions is addressed. Although NCHRP 350 describes a side impact test, there is no requirement for satisfactory performance on the test. There is a critical need for the development and / or implementation of new safety programs, advanced barrier designs, 46

17 and enhanced vehicle-based countermeasures to protect motorists in collisions with guardrails. CONCLUSIONS This paper has examined U.S. crash statistics from to determine opportunities for fatality reduction in vehicle collisions with guardrails. Our analysis suggests four targets for fatality reduction: motorcycle crashes, LTV rollover collisions, side impacts of cars into barriers, and collisions with guardrail ends. Rollover and fatal guardrail end treatments collisions continue to be a problem despite evaluation of both crash modes in NCHRP 350 crash testing. The greatest opportunity for fatality reduction however is the protection of motorcyclists and side impact protection of passenger vehicle occupants in guardrail collisions. Together, motorcyclist and side impacts of passenger vehicles with guardrail account for nearly half of all vehicle-guardrail fatalities. Neither is the focus of current regular guardrail crash testing. ACKNOWLEDGEMENTS The authors gratefully acknowledge the New Jersey Department of Transportation for their sponsorship of this research program. REFERENCES Bligh RP, Mak KK. Crashworthiness of Roadside Features Across Vehicle Platforms. Transportation Research Record 1690: 68-77, Elvik R. The Safety Value of Guardrails and Crash Cushions: A Meta- Analysis of Evidence from Evaluation Studies. Accident Analysis and Prevention Vol. 27, No. 4, pp , Aug Eskandarian A, Bahouth G, Digges K, Godrick D, Bronstad M. Improving the Compatibility of Vehicles and Roadside Safety Hardware: Final Report, NCHRP Project 22-15, Transportation Research Board, National Research Council, Washington, D.C., Faller RK, Polivka KA, Kuipers BD, Bielenberg BW, Reid JD, Rohde JR, Sicking DL. Midwest Guardrail System for Standard and Special Applications. Transportation Research Record 1890: 19-33, Hunter WW, Stewart JR, Council FM. Comparative Performance of Barrier and End Treatment Types Using the Longitudinal Barrier Special Study File. Transportation Research Record 1419: 63-77,

18 Michie JD, Bronstad ME. Highway Guardrails: Safety Feature or Roadside Hazard? Transportation Research Record 1468: 1-9, National Highway Traffic Administration (NHTSA). Traffic Safety Facts 2005: A Compilation of Motor Vehicle Crash Data from the ity Analysis Reporting System and the General Estimates System, U.S. Department of Transportation Report, DOT HS , 2006a. National Highway Traffic Safety Administration (NHTSA). National Automotive Sampling System (NASS) General Estimates System (GES) Analytical Users Manual U.S. Department of Transportation, 2006b. National Highway Traffic Safety Administration (NHTSA). FARS Coding & Validation Manual, U.S. Department of Transportation, 2006c. Ray MH, Hargrave MW, Carney JF, Hiranmayee K. Side Impact Crash Test and Evaluation Criteria for Roadside Safety Hardware. Transportation Research Record 1647: , Ross, HE, Sicking DL, Zimmer RA, Michie JD. Recommended Procedures for the Safety Performance Evaluation of Highway Features. NCHRP Report 350, Transportation Research Board, National Research Council, Washington, D.C., Shelton T. Imputation in the NASS General Estimates System, U.S. Department of Transportation, Report DOT HS , Stephens B. Safety Appurtenance Design and Vehicle Characteristics. Transportation Research Circular 453: 16-25, Troxel LA, Ray MH, Carney JF. Side Impact Collisions with Roadside Obstacles. Transportation Research Record 1302: 32-42, Viner JG. Risk of Rollover in Ran-Off-Road. Transportation Research Record 1500: , 1995a Viner JG. Roadside Safety Problem. Transportation Research Circular 435: 17-29, 1995b. 48