San Francisco systematically under-reports cycling accidents

trauma room

Since it's behind a paywall, below is my transcription of the important study on bicycle injuries in San Francisco I wrote about here and here. The New York Times referred to the study in a story in October. 

I left out only tables, figures, and the endnotes. Emphasis added.

Using trauma center data to identify missed bicycle injuries and their associated costs

Dahianna S. Lopez, RN, MS, MPH, Dharma B. Sunjaya, BS, Shirley Chan, Sarah Dobbins, MPH, and Rochelle A. Dicker, MD, San Francisco, California

Bicycling and walking have become increasingly important modes of transportation in urban environments. The popularity of biking has grown in the last decade for various environmental, economic, and public health reasons, including the challenges posed by the cost of fuel, threat of greenhouse emissions, and diseases such as obesity, diabetes, and heart disease. The transportation, public health, and urban planning communities have come to view active transportation as a critical part of the solution to these challenges.

As we collectively work to strengthen sustainability in transportation, we must pay particular attention to the health and safety of those who use these modes of transport. Injuries from bicycle crashes are not only potentially life altering or deadly for individuals but are also costly. In an analysis of the medical and social consequences of bicycle injury, it was shown that hospitalized bicyclists experienced issues including persistent disability, cognitive and behavioral changes, and permanent work disability. To continually improve the safety and accessibility of alternative transportation for all communities, there must be a systematic way to view and track accurate surveillance data.

According to the National Highway Traffic Safety Administration, in the United States in 2009, 630 bicyclists were killed and 51,000 were injured after colliding with a vehicle. These, like many crash statistics, rely primarily on police reports and are typically underestimated. For example, in Germany, the number of bicycle crashes determined by hospital data exceeded the number of crashes reported to police by nearly two times.

Underestimation of crash rates results partly from a lack of recorded data on cyclist-only (CO) injury crashes, that is, crashes in which no contact is made with an automobile. Moreover, there is bias in both police-reported and hospital-reported bicycle crash data because less severe injuries are reported and treated less frequently. In an attempt to elucidate data on CO crashes, a study conducted in New Zealand reported that 22% of CO injuries could be linked to police reports, compared with 54% of injuries resulting from an auto-versus-bicycle (AVB) collision.

No study to date has compared CO injuries to AVB injuries with regard to their health outcomes and direct medical costs in the United States. The overall goal of this study is to better identify CO injuries and understand the severity and cost of bicycle injuries in San Francisco. We hypothesized that CO injuries carry an equal or greater burden of injury per patient and greater overall cost when compared with AVB injuries. 

This article addresses the following study aims: (1) test the hypothesis that CO injuries are underreported in police records compared with AVB injuries; (2) compare CO and AVB injuries in their incidence, injury severity, and admission rate; and, (3) report the medical cost of bicycle-related injuries from our trauma center database.

Study Design and Population

We conducted a 10-year retrospective cohort study of AVB and CO crash injuries treated at San Francisco General Hospital, an urban Level I trauma center, between January of 2000 and December of 2009. San Francisco General Hospital (SFGH) is the only trauma center for the city and county of San Francisco. The University of California San Francisco's Institutional Review Board granted our group permission to conduct this research study.

Data Sources

We used injury E-codes from the DRG International Classification of Diseases---9th Rev. (ICD-9) to select bicycle-related injuries from the hospital's trauma registry. We selected individuals with injuries who were either treated in the emergency department and released within 24 hours or admitted to the hospital. Using the San Francisco Department of Public Health Billing Information System, we collected hospital charges, professional fee charges, and payments for all patients. 

Police records were drawn from the Statewide Integrated Traffic Records System (SWITRS). SWITRS is a database that serves as a means to collect and process data gathered from a collision scene and is accessible by the public. Patients injured outside the city and county of San Francisco were excluded from the sample...

Police- and Hospital-Reported Bicycle Injuries

Police reported 3,717 bicycle injuries from 2000 to 2009; SFGH reported 2,504 bicycle injuries from 2000 to 2009...

Match between Police and SFGH Data

After matching the SFGH to SWITRS data using a probabilistic linking approach, we identified three categories : matched data, SFGH-only data, and SWITRS-only data. Thirty-one percent of SWITRS data were matched, and 45% of SFGH data was matched. The remaining 55% of bicycle injuries treated at SFGH were not associated with a police report...

A more detailed analysis of the matched data from 2000 to 2009 showed that 90.9% (1,033) of the cases were AVB injuries and 9.1% (104) were CO injuries. Of all injuries identified at SFGH (2,504), 41.5% were AVB injuries and 58.5% were CO injuries. Hence, among matched data, we saw substantially fewer CO injuries (9.1% vs. 41.5%), suggesting that CO injuries may be underreported in police data. Our data also showed that crashes with severe injuries were matched more frequently than crashes with minor injuries...

Patient Outcomes from SFGH-Identified Injuries

SFGH-Identified Bicycle-Associated Injuries

Of all injuries treated in the hospital from 2000 until 2009 (2,504), 37% arrived by ambulance and 3.2% self-transported to the hospital. Transport data were missing for the remaining population. Helmet use data were available for 1,178 bicycle associated injuries. Among those, 80.8% of injured patients were not wearing a helmet at the time of the injury. In addition, 28.2% of injured patients required admission to the hospital...

AVB versus CO Injuries

Those who had CO injuries were three times more likely to have worn a helmet at the time of injury...Nineteen percent of AVB injuries and 53% of CO injuries required admission to the hospital intensive care unit. Based on SFGH data, we found that those with CO injuries were four times as likely to require intensive care unit admission...

The median number of hospital days was 3 days for both admitted AVB and CO injury patients. The median Injury Severity Score for both AVB and CO injuries was 10, but a higher proportion of AVB victims died compared with CO victims (3.7% vs. 2.0%). This distribution however was not statistically significant. Among injuries for which we had hospital disposition data (39% of the total), there was no difference between those with AVB and CO injuries.

Cost of Bicycle Injury in the City and County of San Francisco

After adjusting for economic inflation, the total cost for bicycle-related injuries treated in the hospital in 10 years was $36.4 million (expressed in 2009 dollars). Admitted patient costs accounted for 77% of the total cost of injury. In addition, 74% of the total cost of injury was charged to public funds, including Medicaid and Medi-Cal. Eight million dollars are omitted from this calculation given that the insurance status variable was incomplete for many non-admitted patients.

From 2000 to 2009, the cost of injury per year more than doubled, from $2.6 million in 2000 to $6.8 million in 2009. Across all years, the median injury costs for admitted and non-admitted incidents were $19,207 and $4,088, respectively.

Discussion

Surveillance Gaps in Bicycle Injury Data

We found that police-generated bicycle injury data underreport injuries. Furthermore, this underreporting was much more pronounced for CO[cyclist-only] than for AVB[auto-versus-bicycle] injuries. Our data showed a higher proportion of matched data from the SFGH database compared with the SWITRS match ratio (45% vs. 31%). This suggests that there are fewer CO crashes in which the police are called and the victim went to the hospital. This trend is also demonstrated in the proportions of matched data for SFGH and SWITRS during all years, leading us to assume that not all victims who report the incident to police are then treated at the hospital. This reporting bias is important to consider when looking at injury severity data.

Injury Severity

In our comparison of AVB and CO injuries, we found that CO injuries four times more likely to be admitted to the hospital after presentation in the emergency department. Despite this increased likelihood of admission, our data indicated that length of hospital stay, hospital disposition, and the Mean Injury Severity Scores were not different among AVB injury patients and CO injury patients.

In addition, we found notable data on helmet use among those injured, which shows that a significantly higher proportion of AVB injury patients report not wearing a helmet. Owing to limitations in the retrospective data, we were not able to fully explore the scope of bicycle injury in San Francisco. Further studies are necessary to understand the differences in injury severity, predictors of injury, comorbidities, the role of alcohol use, particular demographics of bicycle riders, and the sequelae of AVB and CO bicycle injuries.

Costs of Bicycle Injury

In reporting the medical cost of bicycle related injuries, we found that AVB patients incurred a significantly higher cost of treatment than do CO patients. The cost for both AVB and CO grew tremendously from 2000 to 2009, nearly tripling for cost of AVB medical care and doubling for cost of CO medical patient care. Of note, we found that 74% of the cost of injury was charged to public funds. Given this information, it is clear that improved bicycle injury prevention efforts may result in reduced state and federal health care expenditures.

CONCLUSION

Traditionally, SWITRS data are used as the criterion standard to inform policy and set priorities for urban projects involving the built environment and other traffic safety measures. However, SWITRS data, as previously reported by Sciortino et al, are only one piece of the surveillance puzzle. Trauma centers can play a key role in future collaborations to define issues and develop prevention strategies for cyclist crashes as well as pedestrian crashes.

In this work, we also unroofed a previously underappreciated population of cyclist victims, those who are injured in the apparent absence of involvement with an automobile. Further work needs to be done to understand risks involved in these incidents and to characterize those who are most often involved. Possible issues may include crashes to avoid cars, tires caught in light rail tracks, swerving to avoid pedestrians, or avoidance of unforeseen road hazards.

Creating a surveillance tool that matches data from injured patients who come to the hospital emergency department, ambulance data, and police report data, trauma centers may be able to provide access to information that can improve research and influence city policy and planning initiatives.

Access to bicycling and other nonmotorized modes of transportation is a significant need, especially in light of the public health issues resulting from inactivity. To encourage bicycling, improved roadways and management of risk and risk perception is key. 

Surveillance data from both police records and hospital records will enable a better understanding of the needs for cycling infrastructure and public health education through evidence based research on both national and international scales. By failing to implement effective and sustainable surveillance strategies for these types of injuries, our efforts to prevent bicycle-related injuries may remain limited.

AUTHORSHIP

D.S.L., S.C., and R.A.D. designed the study. D.S.L., D.B.S., and S.C. collected the data. D.B.S. and S.D. analyzed the data. S.C. managed the data. S.D. prepared the figures. R.A.D. contributed in the private investigation of the study. D.S.L., S.D., and R.A.D. wrote the manuscript. All authors interpreted the data.

DISCLOSURE

This research was supported by the Centers for Disease Control and Prevention Injury Center Grant R49 CE001178.

Address for reprints: Rochelle A. Dicker, MD, Department of Surgery, Ward 3A, San Francisco General Hospital, 1001 Portrero Ave., San Francisco, CA 94110. email: DickerR@sfghsurg.ucsf.edu