Crashes between large trucks and passenger vehicles are always "high-risk event[s]," particularly for the occupants of the passenger vehicles, who "account for 94 percent of deaths" that result from these crashes. A crash known as an underride "increase[s] the probability that death or injury will occur." An underride is generally defined as "a type of crash in which a portion of a passenger vehicle slides under another vehicle." Underrides can occur between two passenger vehicles, but are more common and more dangerous between a large truck and a passenger vehicle.

The Fatal Accident Reporting System (FARS), "the federal government's census of fatal crashes occurring on public roads in the United States," records large truck crashes involving fatalities. Between 1988 and 1993, "deaths in large truck-car crashes coded as underrides in FARS averaged 179 per year . . . , or about 5 percent of all deaths in truck-car crashes." However, the accuracy of this count has been questioned. For example, studies in the 1970s found that "about 80 percent of fatal truck-car crashes in the United States involved underrides." Other countries have also reported higher numbers of underrides among fatal truck-car collisions than those found in FARS.

These differences prompted a study that compared the number of fatal truck underride crashes in FARS with those of another national data base called the National Accident Sampling System Crashworthiness Data System (NASS/CDS). Elisa R. Braver, Michael X. Cammisa, Adrian K. Lund, Nancy Early, Eric L. Mitter, and Michael R. Powell discussed the comparison in "Incidence of Large Truck-Passenger Vehicle Underride Crashes in Fatal Accident Reporting System and National Accident Sampling System" (Transportation Research Record 1595). Data from NASS/CDS showed that fatal underrides of large trucks by passenger vehicles "occur[ed] much more frequently than was indicated in FARS." The authors attributed the inaccuracy to various factors, and they recommended "improvements . . . in FARS data collection and coding procedures to address the problem of undercounting underrides." They also found that while NASS/CDS was more accurate, it too had faults that should be addressed.

BACKGROUND AND METHODS

Each year, FARS records about 41,000 deaths from motor vehicle crashes, and police reports are the primary information source for this crash data. NASS/CDS investigates about five thousand police-reported crashes each year. FARS defines an underride as "'a vehicle sliding under another vehicle during an accident. The classic example is an automobile striking the rear end or the side of a tractor-trailer and coming to a stop under the trailer.'" Before 1995, the NASS/CDS definition of an underride was "one vehicle bumper sliding under another vehicle's bumper, as indicated by an uneven damage pattern caused by unequal amounts of crush in different vertical regions in the front or rear planes of the passenger vehicle." Because this definition did not cover side impacts to passenger vehicles or trucks, "in 1995, NASS/CDS started to code underrides involving the sides of large trucks" but not the sides of passenger vehicles.

The present study defined an underride "as the sliding of a passenger vehicle at least partially under a large truck at some time during a crash." This definition expanded the NASS/CDS definition "to evaluate the presence of underrides involving side planes of vehicles." The study compared "underride coding in FARS and NASS/CDS . . . for 275 fatal truck-car crashes that were contained in both data bases between 1988 and 1993." In each crash, at least one passenger vehicle occupant died. The authors noted that because "more resources and personnel are devoted to each NASS/CDS crash than to each FARS crash, . . . they should be more accurate than FARS records for the same crash."

RESULTS

The study's findings verified this assumption: "Of the 275 fatal truck-car crashes included in both data bases between 1988 and 1993, NASS/CDS coded 75 (27 percent) and FARS coded 18 (7 percent) as involving underrides." The "NASS/CDS coding conventions excluded most cases of underride involving the sides of the truck or passenger vehicle," but "when those crashes were included . . . , it is estimated that underrides were involved in half of all fatal truck-car crashes." Annually, this would amount to an estimated 1,108 fatal underride crashes, as opposed to the 136 indicated by FARS.

The authors attributed the undercounting of fatal underride crashes in FARS to a variety of factors. Police reports for about half the underride crashes lacked the information necessary for analysts to identify an underride. Even when the necessary information was available, analysts still "did not necessarily code underrides." Failing to recognize the indicators of underrides may have been one cause; the fact that underrides "were not a separate variable in pre-1994 FARS" may have been another. In 1994, underrides became a separate variable and analysts were trained in its coding. Both measures increased the number of fatal truck-car crashes coded as underrides. Despite these changes, "the undercount is still substantial." In addition, these changes do not address "the poor quality of police report data in many cases."

The front of the truck was the most common impact point, and half the frontal truck underrides involved the sides of passenger vehicles. The rest of the underrides "were almost equally distributed between rear and side truck impacts." Analysis of the truck types showed that "tractor-trailers had the highest number of fatal underride crashes per year at 770, but fatal underrides of single-unit trucks were also common at 261 per year, including 69 rear truck impacts." Bobtails accounted for only 53 fatal crashes per year, with most involving the front of the truck and none involving the rear.

CONCLUSIONS AND RECOMMENDATIONS

According to the researchers, "fatal underride crashes are much more common than indicated by FARS and should be given higher priority for prevention," including "potential highway safety initiatives" such as "front and side truck underride protective requirements, as well as improved rear underride guards for single-unit trucks." In 1995, NASS/CDS "changed its coding protocol . . . to include underrides involving the sides of large trucks," but more changes "are needed because underrides involving the sides of passenger vehicles are still not identified. NASS/CDS investigators can and should code underrides whenever they occur, but particularly in any crash between large trucks and passenger vehicles."

The undercounting of underrides in FARS should be addressed by improving the interview skills of FARS analysts and guiding them in coding underrides. For example, analysts "should routinely ask the investigating police officers if passenger vehicles involved in large-truck crashes slid under the large trucks." In addition, if the police report lacks information, "FARS analysts should code underrides . . . as unknown rather than not present."