U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
|This report is an archived publication and may contain dated technical, contact, and link information|
Publication Number: FHWA-RD-99-078
Injuries to Pedestrians and Bicyclists: An Analysis Based on Hospital Emergency Department Data
CHAPTER 2. METHODS
The current study was carried out to provide the Federal Highway Administration (FHWA) and the National Highway Traffic Safety Administration (NHTSA) with more complete information on the full spectrum of situations and events causing injury to pedestrians and bicyclists in order to increase awareness of the problem and to help guide program and countermeasure development. FHWA was especially interested in obtaining more detailed information on the location of the injury events with respect to the roadway and on the particular characteristics of road-related events not involving a motor vehicle. Both FHWA and NHTSA were also interested in any additional information that could be gathered on alcohol as a precipitating factor in bicyclist and pedestrian injuries.
The general study approach coupled prospective data collection at hospital emergency departments with retrospective analyses of statewide hospital discharge and motor vehicle crash file data. These databases were analyzed independently and in conjunction with one another to address the study's key research questions.
Three geographically dispersed States were identified and invited to participate in the study--California, New York State, and North Carolina. California and New York State were targeted because they each mandate recording of an External Cause of Injury or "E-Code" (U.S. Department of Health and Human Services, 1991) for each hospital discharge. Although North Carolina does not require E-coding, E-codes are used in the North Carolina Trauma Registry, which incorporates data from approximately a dozen hospitals, including all of the State's Level I and Level II trauma centers. Also, limited E-coded hospital discharge data were available from a North Carolina Hospital Inpatient Discharge Database, formerly maintained by the North Carolina Medical Database Commission.
In each of the three States, two or three hospital emergency departments were identified that were willing to participate in the data collection. For this phase of the study, a special survey form was developed for use in recording information about pedestrian and bicyclist cases to be included in the study (see appendix A). Emergency department data were collected over approximately a 1-year time period at each of the hospitals.
The emergency department survey forms were all forwarded to the Highway Safety Research Center for entry into a computerized datafile and were analyzed using SAS statistical software (SAS, Incorporated, Cary, NC). Project staff also obtained computer files of the hospital discharge data from California and New York State, as well as the Trauma Registry and Medical Database Commission data from North Carolina. Finally, motor vehicle crash data were obtained from each of the States corresponding to the available hospital data. A more detailed description of the data and study methodology follows.
Data Collection Procedures
Emergency Department Data Collection
In New York State, the hospitals were located in a large urban setting and its surrounding suburbs; in California, in smaller urban settings; and in North Carolina, in smaller urban settings that also pulled from large rural areas. Prior to initiating data collection activities, project personnel met with emergency department staff at each hospital to explain the project, provide in-service training, and finalize the specific data collection procedures.
Appendix B contains a copy of the case identification guidelines that were developed for the in-service training of the emergency department data collection staff. Figure 2 highlights the key definitions adopted for the study.
The actual process of collecting the emergency department data varied across the sites. Data collection activities in the Buffalo area were coordinated through a physician who was also
Bicyclist: Any person riding or being carried on a bicycle or other two- or three-wheeled vehicle operated solely by pedals.
Includes: bicycle, tricycle, big wheel, pedal scooter
Excludes: mopeds, other motor-assisted bicycles, motorized scooters
Pedestrian: Any person traveling from one location to another, not in or on a motor vehicle or other road vehicle. Also includes persons working or playing in roadways or other areas generally open to vehicular traffic.
(1) all persons injured as a result of being struck by a motor vehicle, regardless of where the collision took place.
(2) other persons injured as the result of a fall or other mishap while walking, running, standing, working, playing, lying, etc. on a public street or highway or in a public vehicular area (PVA). A PVA is any area that is generally open to and used by the public for vehicular traffic, including entrances to public buildings, parking lots and garages, service stations, stores, restaurants, businesses, etc.
(3) persons injured on other public transportation-related facilities not generally open to vehicular traffic, including, but not limited to, public walkways (sidewalks), alleyways, multi-purpose trails, etc.
(1) persons injured on private property unless a motor vehicle is involved.
(2) persons injured on public property not serving a transportation function (playgrounds, ballfields, parks, etc.) unless a motor vehicle is involved.
(3) any injury incurred while inside a building, residence, or other structure, with the exception of parking garages and similar facilities.
Figure 2. Case identification definitions for emergency department data collection.
Director of Research for the Department of Emergency Medicine at the University of Buffalo. At the three participating Buffalo hospitals, emergency department staff were trained to identify prospective cases and either completed a supplementary checklist or a draft version of the survey form for each case identified. Once every 1-2 weeks, a data collector supported by the project would visit the emergency department and, working from the information sheets and patient cover sheets, fill out the final survey forms. When information was incomplete or unclear, the data collector was usually able to contact the patient by telephone to obtain the required information. The data collector also played a key role in providing feedback to the emergency department staffs and motivating them to maintain interest in the data collection over the 1-year study period.
In California, the project worked through a local subcontractor to help identify and solicit hospitals to participate in the study, train hospital staff, and oversee the data collection activities. However, at each of the three California hospitals, emergency department personnel completed the actual survey forms themselves. At the Goleta Valley and Modesto sites, the survey forms were completed by hospital staff at the time of the emergency department visit. At the Oxnard site, cases were identified on a weekly basis from a computerized record of all injury cases, and the survey forms were completed by a team of three emergency department nurses. Although no follow-up telephone calls were made to the patients, information recorded in the medical files was generally sufficient to complete the survey form.
At New Hanover Regional Medical Center in Wilmington, N.C., the data collection procedure included a combination of survey forms completed by emergency department staff at the time of treatment and a retrospective examination of case logs to capture any missed cases. All survey forms were completed by the hospital staff. The other North Carolina site, Pitt County Memorial Hospital in Greenville, was the only emergency department where patient injuries are routinely E-coded. Because of this, it was possible to identify cases electronically from the hospital's computerized emergency department records. In order to ensure that patient records contained the necessary information to complete all questions on the survey, including the detailed location of the injury event, emergency department staff were trained to record these particular details in their case documentation.
The actual dates of data collection also varied among the hospitals, but generally spanned a 1-year time period. All data were collected between January 1, 1995 and May 1, 1996.
Hospital Discharge Data
As already noted, California and New York State were selected as data collection sites because each mandates recording of an E-code for all persons discharged from the hospital, and because the State hospital discharge database was centrally maintained and accessible for research purposes. For the current study, project staff developed a list of E-codes that could be used to identify each of the various categories of pedestrian and bicyclist injury events (see table 1).
Table 1. E-code groupings for identification of pedestrian and bicyclist injury cases.
The list is relatively straightforward with respect to bicycle-motor vehicle and pedestrian-motor vehicle events. Bicycle-only events can be identified, but their place of occurrence cannot be identified unless a second E-code (E849) is provided that specifically identifies Place of Occurrence. For example, an E-code of 849.5 identifies an event that occurs on a street or highway. Unfortunately, this second E-code is not typically used with motor vehicle and other road-vehicle events. Second E-codes are recommended, however, when the primary event is a fall, so that pedestrian-only falls that occur on a street or highway can be identified if the primary E-code is a fall and the Place of Occurrence is coded as 849.5. However, pedestrian falls that occur in other off-road locations, including parking lots, sidewalks, and driveways, generally cannot be differentiated from falls occurring on stairs, inside homes, on playgrounds, etc.
With these caveats, New York and California each provided their most recent year(s) of hospital discharge data (1994 in California and 1994-1995 in New York State) on a computerized datafile. The data included all of the E-codes listed in table 1, except for pedestrian falls where no place of injury was recorded. Both States also provided summary tables of fall accidents to be utilized in the analyses.
As noted above, North Carolina does not require that E-codes be reported on hospital discharge records. However, all of the State's Level I and Level II trauma centers include E-codes on the data they submit to the North Carolina Trauma Registry (NCTR). For the current study, a computerized dataset was obtained of all motor vehicle traffic injury discharges during 1994 and 1995. In addition, the project obtained a computerized dataset of bicycle and pedestrian injury cases identified in the North Carolina Medical Database Commission files for fiscal years 1993 and 1994 (the two most recent years available). Although this is a statewide database, E-codes were only reported for an estimated 43 percent of the injury cases and not all hospitals contributed, so the numbers obtained are not an accurate accounting of all bicycle and pedestrian hospitalizations in the State.
State Motor Vehicle Crash Data
To complete the picture, State motor vehicle crash data for all reported crashes involving either a pedestrian or a bicyclist were obtained from each of the participating States--California, New York, and North Carolina. Analysis files were created for each State, containing key variables such as pedestrian/bicyclist age, gender, injury severity, date of crash, and time of day. The datafiles were each examined individually and in conjunction with the hospital and emergency department datafiles to explore issues of reporting and to provide a basis for estimating the relative frequencies of the various categories of pedestrian and bicyclist injury events.
Description of the Data
The tables presented in this section provide an overview of the data obtained from each of the three sources--hospital emergency departments, hospital discharge databases, and State motor vehicle crash files. Additional descriptive tabulations are presented in the body of the report and in the appendices.
Hospital Emergency Department Data
A total of 2,802 pedestrian and bicyclist injury cases were reported by the 8 participating hospitals: 50 percent by the 3 Buffalo sites, 35 percent by the 3 California sites, and 15 percent by the 2 North Carolina sites. Table 2 presents a comparison of the numbers of cases reported by the participating hospital emergency departments and the size of these emergency departments, as measured by their total annual visits. These results suggest relatively higher levels of reporting for Erie County Medical Center and Millard Fillmore Hospital in Buffalo, and for St. John's Medical Center in Oxnard. Certainly, the level of participation by these three hospitals appears to have been strong. However, without more specific information on the numbers and types of injury cases treated at each of the hospitals, it is not possible to draw conclusions about their relative levels of participation or how representative the data are of the total number of treated cases.
The distribution of types of cases reported by the participating hospital emergency departments grouped by State is presented in table 3. Overall, one-third (33 percent) of the reported cases were pedestrian-only events and just over a fourth (27 percent) were bicycle-only events. Motor vehicles were involved in less than a third (30 percent) of the reported incidents. The two North Carolina hospitals reported higher percentages of pedestrian-motor vehicle and bicycle-motor vehicle events, and a much lower percentage of pedestrian-only events. Part of this may be due to the manner in which the data were collected at these sites, particularly at PittTable 2. Comparison of reported pedestrian and bicyclist injury cases with total
Hospital emergency department visits at the eight participating hospitals.
1 Percentage of column total.
County Memorial Hospital, where cases were primarily identified from the recorded E-code(s) on a patient's record. As noted earlier, whereas specific E-codes are available for identifying motor vehicle-related events, falls and other non-collision events cannot be as easily identified. Also, the especially high percentage of pedestrian-only events among the New York (Buffalo) cases is probably the result of an unusually cold winter marked by numerous snow and ice storms: just over a third (35 percent) of the pedestrian-only cases reported by the three Buffalo hospitals involved slips on ice or other weather-related falls.
The study also sought information on injury events involving two or more bicycles colliding with one another (bicycle-bicycle) and bicycle collisions with pedestrians (bicycle-pedestrian). Both events were relatively rare, each accounting for only about 1 percent of the reported cases. However, the two event types together accounted for more than 4 percent of the total number of bicycle cases identified.
Table 3. Distribution of pedestrian and bicyclist injury case types by reporting site.
1 Percentage of column total.
Just under 4 percent of the reported cases were identified as "other" or "uncertain" events, with the highest percentage of these being from North Carolina (Pitt County). The majority of these cases arose either from E-codes that could not be directly mapped to a specific category or from cases where hospital personnel simply could not determine whether an individual had been struck by a motor vehicle or not. The latter situation might involve, for example, an injured bicyclist or a drunk pedestrian found lying alongside a roadway. The "other" category includes events such as fingers getting caught in a closing car door, a car running over the foot of a disembarking passenger, a bicyclist riding into the back of a parked vehicle, or other such events that do not fit the usual definition of a pedestrian-motor vehicle or bicycle-motor vehicle collision. Finally, the "non-case" category includes events that, by the case definitions adopted in the current study, were not considered pedestrian or bicyclist events. Examples include a child injured when his sled runs into a lamp post, or a fall from a moped or other motorized two-wheel vehicle.
In addition to the type of event, a second key variable collected for the study was the location where the event occurred, whether on the roadway or in an off-road location such as a sidewalk, driveway, yard, multi-use path, etc. This information is summarized in table 4. Just under half (48 percent) of the reported events occurred in a roadway; 21 percent occurred on a sidewalk; and 9 percent occurred in some type of parking lot. Off-road trails and parks, and private driveways or yards accounted for most of the remaining event locations. The precise location of the event was unknown for just under 10 percent of the reported cases. Whereas roadway locations predominated for events involving a motor vehicle, sidewalks and other off-road locations featured prominently in those events that did not involve a motor vehicle.
Table 4. Distribution of emergency department-reported pedestrian and bicyclist injury cases by location of injury event.
1 Percentage of column total.
Detailed injury event type and event location results based on the hospital emergency department data are contained in chapters 3 and 4 of this report. In addition, appendix C contains additional basic descriptive tables for the emergency department data, including information on the age, gender, race, and disposition status of the injured pedestrians and bicyclists.
Hospital Discharge Data
As noted earlier, computerized hospital discharge data were obtained for each of the three States where emergency department data were collected. For California and New York, these data were available statewide. For North Carolina, where statewide E-coding of hospital discharges is not mandated, two sources of information were examined: computerized data from the North Carolina Trauma Registry and available E-coded data from the North Carolina Hospital Discharge Database. Although the latter is a statewide database, as noted earlier, not all hospitals participated and not all reported injury cases contained a valid E-code for identifying event types. Thus, like the NC Trauma Registry, this data source underestimates the number of hospitalized bicyclists and pedestrians in the State. (See chapter 6 for weighted North Carolina estimates.)
Table 5. Distribution of pedestrian and bicyclist injury cases reported in hospital discharge datafiles.
1 Percentage of column total. 2 No cases identified.
Table 5 summarizes the available hospital discharge data from each State. For California and New York, data were obtained for all cases identified by the E-code listing in table 1. However, the table only includes those pedestrian-only cases that could specifically be identified as occurring on a street or highway. Cases occurring in other locations, or those for which place of occurrence was either missing or unknown, are excluded, since they could also include fall events that would be outside the scope of the study (e.g., falls inside homes or falls occurring at recreational or sports facilities). For North Carolina, no attempt was made to capture pedestrian-only data because the second place of occurrence E-code is not routinely reported. Also, for all States, bicycle-only cases have been grouped into a single category that does not differentiate between roadway and non-roadway events. Again, this is because the place-of-occurrence E-code was not routinely reported for these cases.
Based on the data in table 5, pedestrian-motor vehicle crashes are by far the most frequent injury-causing event, with the vast majority of these occurring in the roadway. Pedestrian-only falls that occur in the roadway are also quite common. For the New York State data, where place of occurrence was routinely coded, pedestrian-only cases made up nearly a third of the database. In California, the percentage was lower, but this is probably an underestimate, since a significant portion of the California cases had missing place-of-occurrence information. Except for the North Carolina Trauma Registry data (which captures more severe injury cases), bicycle-only cases outnumbered bicycle-motor vehicle cases by a factor of nearly two to one.
Additional cross-tabulations of the hospital discharge data are contained in appendix D, with separate tables for bicyclists and pedestrians.
State Motor Vehicle Crash Data
The final data source examined was State motor vehicle crash data. For these data, no attempt was made to identify cases that may not have involved a motor vehicle. The data were used primarily in developing overall projections of pedestrian and bicyclist injuries, and for examining potential underreporting of pedestrian and bicyclist injury events. Table 6 presents the overall case distributions for the data obtained. For New York, more than 70 percent of the pedestrian and 55 percent of the bicycle crashes occurred in one of the five counties defining New York City; and in California, 62 percent of the pedestrian and 51 percent of the bicycle crashes occurred in either the Los Angeles or San Francisco Bay areas. Compared to these two States, North Carolina is much more rural in character.
Additional cross-tabulations of interest for the State crash file data are contained in appendix D.
Table 6. Summary of State pedestrian- and bicyclist-motor vehicle crash data.
1 Percentage of column total.
The remaining sections of the report provide specific data tabulations that address the primary research questions for this study, namely:
1. What are the frequency and characteristics of bicycle injury events that occur in non-roadway locations and/or those that do not involve a motor vehicle, and how do they differ from bicycle-motor vehicle crashes that occur on the roadway? (chapter 3)
2. What are the frequency and characteristics of pedestrian injury events that occur in non-roadway locations and/or those that do not involve a motor vehicle, and how do they differ from pedestrian-motor vehicle crashes that occur on the roadway? (chapter 4)
3. What role does alcohol play in each of these events? (chapter 5)
4. What are the estimated frequencies of motor vehicle and non-motor vehicle, and roadway and non-roadway events causing injury to pedestrians and bicyclists? (chapter 6)