The Relative Risks of School Travel: A National Perspective and Guidance for Local Community Risk Assessment -- Special Report 269 (2002)

There are two basic categories of safety data—exposure data and outcome data (BTS 1999). Exposure data, which measure the susceptibility of an individual or class of individuals to the undesired outcome, include number of trips taken, distance traveled (e.g., vehicle-miles or passenger-miles), and number of travelers; outcome data, which measure untoward or undesirable events, such as crashes, include numbers of accidents, fatalities, and injuries. The committee selected exposure and outcome datasets that could be used to shed light on the issues of concern. Because of the relatively infrequent occurrence of fatalities and severe injuries among children on trips during normal school travel hours, it was necessary to combine multiple years of statistics to obtain reasonable sample sizes. In addition, some otherwise promising datasets were not usable because they lacked the specificity needed to identify incidences or trips related to students going to and from school, or because they could not be paired with the corresponding exposure or outcome data. For example, as noted by Stutts and Hunter (1999, 505):

The various datasets that were examined for possible use in the committee’s analyses are briefly described in this chapter. A detailed description of the three datasets selected, the analyses conducted with each, and the limitations of each for the purposes of this study is then provided. Finally, conclusions are presented.

The committee identified a number of databases that could be used in assessing the safety of the various student transportation modes; unfortunately, most of the datasets are highly limited, contain data that cannot be used to generalize to the population of interest (i.e., all students in kindergarten through grade 12), or do not include the data categories needed to conduct the analyses of interest (e.g., purpose of trip or time of day). Only one dataset, the Nationwide Personal Transportation Survey (NPTS), provides any usable exposure data at the national level. The database of the National Association of State Directors for Pupil Transportation Services (NASDPTS), recorded annually in industry journals and School Bus Fleet and School Transportation News magazines, contains the most accurate data available on school bus ridership on a state-by-state basis.¹ However, the underlying data collected by the states, in many cases, is used to reimburse school districts for school bus services resulting in an overestimation of the number of student riders. In addition, the database does not contain data on the other modes used by children traveling to and from school. In addition, as with the other datasets, there is a lack of consistency in terminology. For example, some states report ridership based on actual head counts, others base their counts on the number of students “authorized” to use the school bus in their daily commute to and from school, and some also include private school ridership. In this dataset, it should be noted, ridership numbers are attached to trip purpose; however, this dataset could not be linked to others because it does not use the data selection criterion of normal school travel hours that is applied for the other modes. Therefore, the NASDPTS data could not be used for cross-mode comparisons.

Nine national outcome (or crash) datasets were considered by the committee: (a) the National Automotive Sampling System (NASS) General Estimates System (GES), (b) the NASS Crashworthiness Data System (CDS), (c) the NASS Pedestrian Crash Data System (PCDS), (d) the Crash Outcome Data Evaluation System (CODES), (e) the Highway Safety Information System (HSIS), (f) the Kansas Department of Education school bus loading/unloading fatality dataset, (g) the National Transit Database (NTD), (h) the National Electronic Injury Surveillance System (NEISS), and (i) the Fatality Analysis Reporting System (FARS).²

GES, which became operational in 1988, contains data from a nationally representative sample of police-reported motor vehicle crashes of all types. The reports are chosen from 60 areas that reflect the geography, roadway mileage, population, and traffic density of the United States. Data are collected weekly from approximately 400 police jurisdictions, from which about 50,000 police accident reports (PARs) are randomly sampled each year. PARs are completed by police officers investigating crashes that result in personal injury and/or property damage above the state’s reporting threshold. For example, in the state of Texas, if a vehicle involved in a crash is towed, a PAR must be completed, while in the state of California, one must be completed if property damage is estimated to be $500 or more. Each state, the District of Columbia, Puerto Rico, and the Virgin Islands has a unique crash report form. The reader is referred to the State Crash Report Forms Catalog 1999 Update (NHTSA 1999), which contains a copy of each state’s form, as well as the state crash reporting threshold. The National Highway Traffic Safety Administration (NHTSA) encourages uniformity across states in the data elements contained on the crash report form, and for this purpose encourages the use of American National Standards Institute (ANSI) D-16 (Manual on Classification of Motor Vehicle Traffic Accidents) and D-20 (Data Element Dictionary for Traffic Records Systems). NHTSA, in collaboration with the Federal Highway Administration (FHWA) and the National Association of Governors’ Highway Safety Representatives, has also developed a guideline for collection of crash data, referred to as the Model Minimum Uniform Crash Criteria (NHTSA 1998). Although states do not generally record data from crashes (either on- or off-highway) in which a motor vehicle is not involved, GES is the most complete injury database available.

CDS, which covers about 5,000 accidents in depth annually, contains data on vehicle occupants and is used to study injury mechanisms; precrash events are not examined in detail. This shortcoming is moot for purposes of the committee’s analysis, however, as there are not enough cases involving school and transit vehicles for this dataset to be useful.

PCDS contains data on pedestrian crashes. It is essentially an update (or continuation) of earlier pedestrian data files, such as the Pedestrian Injury Causation Study of 1977 and the special pedestrian data collected for NASS from 1979 through 1987. PCDS data for 1995–1997 relate to only 280 cases, too few for the purposes of this study. Although there is a code for school buses and other buses, none of the 280 cases in the file involved these types of vehicles.

CODES contains linked statewide crash and injury data for 20 states that match vehicle, crash, and human characteristics with final medical and financial outcomes. The purpose of the dataset is to improve decision making related to highway safety and injury control. CODES studies often tend to be in-depth examinations of special problems; however, they usually cover highly limited geographical areas.

HSIS, which contains selected data for eight states (California, Illinois, Maine, Michigan, Minnesota, North Carolina, Utah, and Washington), is a roadway-based system that includes data on a large number of accident, roadway, and traffic variables. The data are collected annually from the participating states,

processed, documented, and prepared for analysis. The dataset contains little information specifically about travel to and from school, and thus could not be used to address the issues of concern to the committee.

The Kansas Department of Education conducts an annual national survey of school bus loading and unloading fatalities. As with the NASDPTS dataset, however, it does not contain data for the other modes, making comparison with other studies difficult, if not statistically invalid. For example, there are no comparable loading and unloading data for the other modes.

NTD, which is maintained by the Federal Transit Administration (FTA), contains performance, operational (such as miles traveled, passenger-miles, and passengers carried), and financial information, as well as injury data stratified by mode [motor bus, trolley bus, light rail (passenger rail operating on exclusive or separated guideways)] reported by all transit agencies receiving federal assistance.³ This database, which is used for management and planning by transit systems, as well as for policy analysis and investment decision making, also contains data on numbers of fatalities and injuries by transit system; however, data on age of victim and purpose of trip are not included. The data collected make it possible to compute injury rates for each transit mode for each agency, as well as national rates by mode. However, the data are not reported by passenger age, and this greatly restricts use of the data in school transportation analyses.

NEISS, which has been maintained for approximately 30 years by the Consumer Product Safety Commission (CPSC), is a representative sampling of U.S. hospital emergency departments (Stutts and Hunter 1999). Its primary purpose has been to provide data on consumer product–related injuries occurring in the United States. In 2000, CPSC initiated an expansion of the system to collect data on all injuries. Thus, although not useful to the committee for the present study, NEISS may provide highly useful information in the future, depending on the data elements that are added.

FARS, established in 1975, provides a census of all highway fatalities (including deaths of school-age children by time of day and vehicle type) in all 50 states, the District of Columbia, and Puerto Rico. The data in FARS are obtained from existing state data: PARs, state vehicle registration files, state driver licensing files, state highway department data, vital statistics, death certificates, coroner/medical examiner reports, hospital medical reports, emergency medical service reports, and other state records. From these documents, data for more than 100 FARS data elements are obtained. Each year, specific data elements are modified in response to changing user needs, vehicle characteristics,

and highway safety emphasis areas. As specific as the data are, however, no personal identifying information (such as names or addresses) is recorded in FARS. Similar to GES, FARS was developed to provide an overall measure of highway safety, to identify traffic safety problems, and to provide a single objective basis for evaluating the effectiveness of motor vehicle safety standards and highway safety incentives. NHTSA annually provides descriptive statistics for traffic crashes of all degrees of severity in its Traffic Safety Facts series, based on GES and FARS data (see NHTSA 2000 for a recent report). (FARS data may also be accessed on the Internet at www.fars.nhtsa.dot.gov.)

Given the study objectives and the data available in the various datasets, the committee elected to use the following three datasets for its analyses:

• NPTS, the only national dataset that contains exposure/travel information, used to estimate the number of trips taken and miles traveled by school-age children;

• FARS, used to analyze student fatalities; and

• GES, used to analyze student injuries.

Information from all three datasets had to be filtered and grouped to extract relevant inputs for the committee’s risk analyses. Three main generalizations had to be made: (a) time of day was used as a surrogate for school travel trips; (b) age groupings were created from individual age groups; and (c) general transportation mode categories were delineated.

The FARS and GES datasets do not reliably record purpose of trip for each incident. As a surrogate for purpose of trip, the committee defined school travel hours (i.e., hours during which most school-related trips would occur), as follows:⁴

• Months of year: September 1 through June 15^5,6

• Days of week: Monday through Friday

• Hours of day: 6:00 a.m. to 8:59 a.m. and 2:00 p.m. to 4:59 p.m.⁷

Because of the limited amount of data in the two output datasets (FARS and GES), an analysis for each age group was not possible. Therefore, the committee created the following four age groupings: 5–10 years of age, 11–13, 14–15, and 16–18. This categorization is associated with the ages at which children change schools (the break around 10–11 years is when children generally move from elementary to middle school; 11–13 years of age maps to middle-school ages, and 14–18 is the rough equivalent of high school ages). These groupings also correspond roughly to transition points in the developmental and behavioral characteristics of children. Further, while the last two age groupings both include high school students, those aged 16–18 are more likely to have a driver’s license (or friends that have one), resulting in a different distribution of travel modes used. In addition, those aged 16–18 are more likely to hold after-school jobs or to participate in after-school activities that require transportation after regularly scheduled school bus service times. Therefore, the committee believed this last age group would be significantly different in both modes of travel used and number/length of trips during normal school travel hours, and thus should be examined separately.

There are many ways to classify injury and fatality data. Table 2-1 shows a classification consisting of 16 categories. FARS and GES provide data for 14 of these categories; categories 6 and 8 do not include motor vehicles. For much of the analysis, these 16 categories were combined into six more general categories:

• Categories 1 and 2—school bus–related crashes;

• Categories 3, 4, and 5—passenger vehicle crashes;

• Categories 6 and 7—pedestrian crashes;

• Categories 8, 9, and 10—bicyclist crashes;

• Categories 11 and 12—other bus crashes; and

• Categories 13, 14, and 15—motorcycle crashes.

NPTS serves as the nation’s inventory of daily personal travel. It is a computer-assisted telephone interview survey of households in the United States. For each trip made during a preselected 24-hour time period, data are gathered regarding purpose of trip, mode of travel, length of trip, day of travel, vehicle occupancy, driver characteristics (e.g., age, gender, worker status, education level), and vehicle attributes (e.g., make, model, model year, annual miles driven, odometer readings). These data are gathered for all areas of the country, all days of the week, and all months of the year. Data are collected only for the civilian, noninstitutionalized population in the United States aged 5 years and older. NPTS does not include responses from military personnel living on base or overseas, or from residents of nursing homes, assisted-living facilities, long-term medical institutions, college dormitories, or prisons.

NPTS has been conducted five times—in 1969, 1977, 1983, 1990, and 1995; the sixth survey was initiated in late 2001. Since the survey contents and method-

ology have been modified each time, data from multiple years cannot be used in one analysis without some type of data manipulation. Therefore, the committee used data from the most recent survey available at the time (the 1995 survey) for its analyses. In 1995, NPTS consisted of a national sample of 21,020 households and an additional 21,013 households in five add-on areas (New York; Massachusetts; Oklahoma City and Tulsa, Oklahoma; and Seattle, Washington). The data were collected from May 1995 through July 1996 (Chen et al. 2000). [See RTI and FHWA (1997) and Hu and Young (1999) for an in-depth description of the 1995 NPTS survey procedures and methodology.]

NPTS is a random sample of the nine census regions stratified by population size and other factors. The 1995 survey began with 160,000 telephone numbers, approximately 45 percent of which were dropped when there was no response. Because the stratification was done using telephone numbers, all households without a telephone—a group in which households at the poverty level are over-represented—were excluded.

It was determined that approximately 72 percent of households that received the travel log filled it out. The overall response rates were 55.3 percent for household-level data and 34.3 percent for person-level data.

Given the data fields available in NPTS, it is possible to extract data on trips made and miles traveled by transportation mode, age, date, time of day, purpose of trip, and type of geographic region (i.e., rural versus urban).⁸ On the basis of this dataset, it was estimated that approximately 50 million students (Census Bureau 2000) made 9.7 billion trips (5.9 billion urban and 3.8 billion rural) to school during normal morning school travel hours and traveled approximately 44.0 billion student-miles (20.0 billion urban and 24.0 billion rural). Table 2-2 provides data on number of trips, and Table 2-3 shows student-miles traveled during normal morning and afternoon school travel hours. Tables containing detailed trip and student-mile data are provided in Annex 2-1. The data suggest that, on average, each student made approximately 194 trips during normal morning school travel hours throughout the school year—an average of 1.1 school trips per student each school morning.⁹ The definition of a trip is one-way travel from one address to another. Thus, if a person travels from home to school to drop off a student, and then goes to work, the driver has made two trips and the student one. If a child walks to the bus stop and rides a bus to school, this is considered one trip.

It was estimated that during normal afternoon school travel hours, these same students made 13.8 billion trips (8.6 billion urban and 5.2 billion rural) and traveled approximately 69.3 billion student-miles (34.7 billion urban and 34.6 billion rural). The data suggest that, on average, each student made approximately 276 trips during normal afternoon school travel hours throughout the school year, which translates to approximately 1.5 trips per student each afternoon. Given that it is known, at least anecdotally, that more trips are taken in the afternoon than in the morning travel period (e.g., to run errands, go to the library, go to lessons, meet friends), one would expect there to be more afternoon than morning trips on average.

For each analysis, the sample size (N), the estimated statistic (trips or student-miles) based on the sample size, and the standard error are reported in the tables in Annex 2-1. These data are the basis for the analyses in Chapter 3 that compare the risk to students of traveling to and from school during normal school travel hours using the various modes.

The committee’s analyses were hampered by problems inherent in the quantity and quality of the data, especially difficulties in obtaining accurate and reliable enrollment and ridership data. This was especially the case when the committee tried to gather data at other than the aggregate level. As noted, the NPTS data are based on a sampling of some 42,000 households in the United States, and

they cannot be used reliably for analyses at the state and community levels with any degree of confidence. Thus the committee’s ability to gauge the effects of state-by-state or local differences was constrained. While it was recognized that substantial variation exists in the modes of student travel according to location and such factors as climate, infrastructure, and local economic and demographic characteristics, the committee was unable to measure this variation directly. Without such data, conclusions concerning the direct effects of local conditions on the numbers of fatalities and injuries could not be drawn since rates and risk ratios could not be computed. In addition, the data in the NPTS database are self-reported, a feature associated with many well-known limitations. For example, there may be an undercount of trips because, in the case of pupil transportation, the bus trip to school may be documented, but not the trip between home and the bus stop. All travel data for children aged 5–13 are reported by their parents, while travel data for teens aged 14–17 may be either self-reported or reported for them by household adults. Teenagers may not report all afternoon trips they made, or parents responding for their older children may not be aware of all the trips the children made.

The NPTS survey is intended to gather information on personal travel of U.S. households, including why, how, when, where from, where to, how frequently, how long, and with whom trips were made. The survey is not limited to schoolage children and not focused on school-related transportation—small facets of daily travel that are captured by the dataset only to a limited extent. This can be seen by the extremely small sample size (N) shown in the tables in Annex 2-1.

Number of fatalities was one of two outcome measures examined by the committee. These data are contained in FARS, a database first developed in 1975 that contains only data on all fatal traffic crashes that occur on public roadways in the United States.¹⁰ Data in this national database are extracted from medical examiner, coroner, emergency medical, and police accident reports, as well as from driver, vehicle, and roadway classification records. There is detailed information in the database on crash, vehicle, driver, and occupant characteristics.

Analyses were performed using 9 years of FARS data (1991–1999). In these 9 years, a total of 51,350 children between the ages of 5 and 18 were killed in all traffic crashes in the United States (Table 2-4). Of these, 7,470 were killed during normal school travel hours,¹¹ 2,719 were killed during school session

hours, and 40,655 were killed during all nonschool hours. Table 2-5 breaks these values down by the 15 fatality categories used by the committee. Details on the school session time and non–school time categories can be found in Annex 2-2.

The FARS variable “roadway functional class” was used to distinguish between “rural” and “other” (basically urban) crash sites for mapping to the rural/ urban classification used in the NPTS analyses. Table 2-6 shows the distribution of normal school travel time fatalities for urban versus rural geographic locations. Table 2-7 shows the distribution of fatalities by individual ages, and Table 2-8 shows the distribution by age group.

The FARS database, being limited to fatalities, is likely to overstate or understate the incidence of uncommon events, such as fatalities not involving passenger vehicles, when only a single year of data is considered. An extremely rare event, such as an incident resulting in multiple fatalities to pupils aboard a school bus, can skew the data by inflating the risk for that mode during the year of occurrence and can change the interpretation or ranking of risk for that mode. For these reasons, fatality data were examined for a longer period.

As noted, moreover, the national databases used to examine the safety of children traveling to and from school—FARS for fatality counts and GES for non-fatal injury estimates—provide data only for incidents in which a motor vehicle is involved. There is no national database to record pedestrian and bicycling fatalities and injuries not involving a motor vehicle. The result is underestimation of the number of fatalities and injuries involving the non–motor vehicle modes, which hampered the committee’s analyses of the relative safety of these modes. However, FARS and GES do provide some insight into the safety of these modes when they interact with the motor vehicle modes.

The committee encountered two other difficulties in attempting to analyze fatalities that occurred when students were being transported to and from school and school-related activities. First, it was not possible to determine that a trip was, in fact, a school trip (going either to or from school or to or from a school-sponsored activity), especially for the non–school bus modes. Second, pedestrian fatalities resulting from crashes involving other buses could not be identified. For example, if a student were fatally injured crossing the road to get to a transit stop, this would not be recorded as a transit bus–related fatality in the database.

Data in the GES database are subdivided into three levels of nonfatal injury— levels A, B, and C—and one fatal injury category. This ANSI-developed injury severity rating scale is used by most states:

• Level A: incapacitating injury—Any nonfatal injury that prevents the injured person from walking, driving, or normally continuing the activities he or she was able to perform before the injury occurred. Included are injuries such as severe lacerations, broken or distorted limbs, skull or chest injuries, abdominal injuries, unconsciousness at or when taken from the accident scene, and an inability to leave the accident scene without assistance. Momentary unconsciousness is excluded.

• Level B: nonincapacitating evident injury—Any injury, other than a fatal or incapacitating injury, that is evident to observers at the scene of the accident where the injury occurred. Included are injuries such as lumps on the head, abrasions, bruises, and minor lacerations. Limping (the injury cannot be seen) is excluded.

• Level C: possible injury—Any injury reported or claimed that is not fatal, incapacitating, or nonincapacitating evident. Included are such injuries as momentary unconsciousness; claims of injuries not evident; limping; and complaints of pain, nausea, or hysteria.

These categories are quite subjective, meaning that different individuals (e.g., police officers) who apply the scale may interpret the definitions differently. Further, because some states do not use this classification, a state’s classification of injuries may not correlate with that in GES or with those used by other states. Finally, because only three categories are used, injuries of vastly different severity must, at times, be grouped at the same severity level: “For example, under the ANSI D16.1 scale, injuries ranging from broken arms to quad-riplegia are all classified as incapacitating injuries” (TRB 1989, 54).

GES provides information on a nationally representative (stratified) sample of all severities of police-reported traffic crashes within 60 geographic sites across the United States. It is a probability sample of approximately 45,000 annual U.S. police-reported crashes on public roads that result in property dam-

age, injury, or death. GES estimates are intended to provide information on a national level about motor vehicle crashes and the vehicles and people involved. The purpose of GES is to track trends in these national-level estimates so that highway safety problem areas can be identified, to provide a basis for regulatory and consumer initiatives, and to form the basis for cost–benefit analyses of highway safety initiatives. [See NHTSA (1991) for information on the set of crashes described by GES estimates, the sample selection procedures, the estimation procedure, and the reliability of those estimates in terms of sampling error.]

The committee performed analyses on 9 years of GES nonfatal injury data (1991–1999) to determine the number of nonfatal injuries involving each mode of interest for the age groupings of interest during the defined normal school travel hours. Given that portions of a month (e.g., the first 2 weeks but not the last 2) could not be segregated, the analyses were conducted for the entire months of September through June.

Although the injuries in GES are classified into three levels, the small number of accidents and injuries for some of the age categories within a mode resulted in large standard error estimates. In addition, some category C injuries are severe (e.g., whiplash and concussion) and the committee wanted to be sure not to miss serious but unobservable injuries. Finally, for the travel modes that the committee used, the percentage of (A+B) injuries when compared to (A+B+C) were significant. For the school bus, other bus, and the two private vehicle categories, (A+B) was between 37 and 40 percent of the total. The effect of including C categories would not change the relative comparison among these modes. For the other two transportation categories (walking and bicycling), (A+B) was 68 to 70 percent of the total injuries. This is because of the higher lethality of accidents involving children not protected by a vehicle. An injury rate analysis based on categories (A+B) would look very similar to the one in the report with the exception of walking and bicycling. These two modes would have their injury rates increase relative to the other modes by approximately 70 percent. Therefore, although the range of injuries that are included in the A, B, and C categories is quite broad, the three injury levels were collapsed to ensure sufficient sample sizes within each category for the committee’s analyses. Given the already significant differences in the risk rates and the committee’s desire to capture all injuries, the committee felt that doing separate analyses based on (A+B) injuries were not warranted.

From 1991 through 1999, an estimated 5,714,048 school-age children were injured, 1,379,394 of whom received their injuries during normal school travel hours (see Tables 2-9 through 2-11).¹² Of the latter injuries, 40 percent

occurred to pupils traveling in a passenger vehicle driven by an operator 19 years of age or older, and 32 percent to students aged 16–18 who were driving a motor vehicle. Just over 3.5 percent were student passengers on a school bus, and only 0.25 percent were student pedestrians injured in school bus–related crashes. Of the total student injuries, 5 percent are estimated to be to bicyclists in crashes not involving school buses. Table 2-10 shows the estimated total number of injuries to school-age children during normal school travel hours for the years 1991 through 1999, inclusive, broken down by age and mode categories.

It is interesting to note that for each of the four age groupings, numbers of injuries sustained in passenger vehicle crashes are consistently highest relative to the other modal categories. For example, of the injuries sustained by students 5–10 years of age, 72 percent occurred when they were riding in a passenger vehicle driven by an operator 19 years of age or older. For those aged 11–13, this category also represents the largest proportion of injuries—49 percent. Of the injuries sustained by students aged 14–15, 39 percent occurred when they were riding in a passenger vehicle driven by someone under age 19. And for those aged 16–18, the majority of injuries (62 percent) occurred when they themselves were driving a motor vehicle. Table 2-11 shows for comparison the estimated total number of injuries for the same breakdowns for non–school travel hours for the same years.

Given the sampling procedures for the GES database, there are some limitations due to sampling error and standard error; there is also a potential lack of representativeness. For example, Traffic Safety Facts (NHTSA 2000) reports that there were 7,500 school bus passenger injuries, with a standard error of approximately 1,300, and there were 350 pedestrian injuries sustained in school bus–related crashes, with a standard error of 300. These standard error estimates are taken from the published generalized broad error estimates and are not derived individually for each individual estimate. This lack of statistical precision diminishes the confidence in the dataset. Again, however, GES is the best database available for analyzing transportation-related injuries to students and addressing the issues of concern to the committee.

According to NHTSA (2000), the GES data elements may be modified yearly, leading to some inconsistencies in the dataset. These inconsistencies may in turn limit the usability of the data for answering particular questions, especially if data for multiple years are being examined. Moreover, as with FARS, GES does not capture purpose of trip; therefore, analysis of the data for preselected time periods captures all types of trips occurring during those times, which may include non-school-related trips, especially in the afternoon time period.

Another limitation of this database is that there is an underreporting of injured passengers in some of the sampling units because only those crashes that resulted in a PAR and exceeded the particular state’s reporting threshold level had the potential to be included in the dataset. There could also be under-reporting of serious injury because internal injuries such as intra-abdominal and intracranial injury may not be detectable at the scene. On the other hand, given the imprecision of the category definitions, particularly Level C (possible injury), and the policy in some districts of transporting to the hospital all students involved in a bus crash, there could also be an overreporting of injuries in this category. Another limitation is that injury data for light rail transit, as well as for bicycling and walking, obtained from GES include only injuries occurring in crashes that involved a collision with a motor vehicle. For the purposes of the data, light rail vehicles are not considered to be motor vehicles because they do not operate on roadways; thus injuries sustained aboard that mode are not included in the dataset. Unfortunately, these data are largely absent in other sources as well.

Data problems occur with fatality and injury data for children traveling to and from school and school-related activities regardless of the mode used. For the issues addressed in this study, the available data have many limitations: needed data are not available, there are definitional inconsistencies across databases and across years for the same database, there are recording errors in the datasets, and there are unknowns and missing data in the datasets that need to be taken into account.

Although numerous datasets exist, few contain representative data in sufficient quantity to be used for the types of detailed analyses conducted by the committee. At the same time, sufficient data at the community level are not easily accessible, if they are available at all. This diminishes the completeness of assessments that may be conducted and in turn impedes the ability to manage the risks involved in school transportation appropriately.

Currently available data on fatalities and injuries associated with transportation to and from school and school-related activities are illuminating but incomplete. The committee also found it difficult to link the data from multiple databases, especially because of the lack of consistency in terminology and other limitations noted above. One of the primary responsibilities and contributions of the federal agencies whose mission encompasses issues related to school transportation is to collect good, accurate, reliable data. If done correctly, a consistent, comprehensive data collection effort would benefit all highway modes, including school transportation. However, obtaining more thorough and complete data is not without cost. Given the large numbers of fatalities and injuries that occur on highways in the United States and the fact that relatively few of these involve students during school travel hours, the benefits of any additional data collection efforts need to be fully considered before such efforts are recommended or implemented.

APTA American Public Transportation Association

BTS Bureau of Transportation Statistics

FHWA Federal Highway Administration

NHTSA National Highway Traffic Safety Administration

RTI Research Triangle Institute

TRB Transportation Research Board

APTA. 2002. 2002 Public Transportation Fact Book. Washington, D.C.

BTS. 1999. Transportation Statistics Annual Report 1999. U.S. Department of Transportation, Washington, D.C.

Census Bureau. 2000. Statistical Abstract of the United States. Washington, D.C.

Chen, L.-H., S. P. Baker, E. R. Braver, and G. Li. 2000. Carrying Passengers as a Risk Factor for Crashes Fatal to 16- and 17-Year-Old Drivers. Journal of the American Medical Association, Vol. 283, No. 12, pp. 1578–1582.

Hu, P. S., and J. R. Young. 1999. Summary of Travel Trends 1995 Nationwide Personal Transportation Survey. Oak Ridge National Laboratory, Oak Ridge, Tenn., Dec. www.fhwa.dot.gov/ohikm/nptspage.htm.

NHTSA. 1991. National Accident Sampling System General Estimates System Technical Note, 1988 to 1990. Washington, D.C.

NHTSA. 1998. Model Minimum Uniform Crash Criteria (MMUCC). Washington, D.C. www.nhtsa.dot.gov/people/ncsa/codes/mindata/guideline.pdf.

NHTSA. 1999. State Crash Report Forms Catalog 1999 Update. Washington, D.C. www.nhtsa.dot.gov/people/perform/trafrecords/.

NHTSA. 2000. Traffic Safety Facts 1999: A Compilation of Motor Vehicle Crash Data from the Fatality Analysis Reporting System and the General Estimates System. Washington, D.C.

RTI and FHWA. 1997. 1995 NPTS User’s Guide. Publication FHWA-PF-98-002; HPM-40/10-97 (2M)EW. North Carolina. www.cta.ornl.gov/npts/.

Stutts, J. C., and W. W. Hunter. 1999. Motor Vehicle and Roadway Factors in Pedestrian and Bicyclist Injuries: An Examination Based on Emergency Department Data. Accident Analysis and Prevention, Vol. 31, pp. 505–514.

TRB. 1989. Special Report 222: Improving School Bus Safety. National Research Council, Washington, D.C.