Costs of Injuries Resulting from Motorcycle Crashes:
A Literature Review
Appendix A
Critical Reviews of Publications on Motorcycle Crash Costs

Table of Contents / NHTSA home / Next / Overview

Each review begins with a citation or citations to the reviewed article(s) in American Psychological Association format. Next comes the article summary, an expanded abstract edited for clarity and accuracy. The summary describes the study’s objectives, study population, data, methods, results, and conclusions, using these standard headings in this order for ease of use. The summary is followed by a critique, which indicates methodological strengths and weaknesses and assesses whether the study conclusions are merited based on the research reported. The reviewers might also note any important facts, such as the presence or absence of helmet laws in the time and place under study, that were not explicitly stated in the article. The reviewers might also highlight interesting methods, facts, or sources gleaned from an article under review. The criteria followed in writing the summary and critique of each article can be found in Appendix E.

Our comments in the “Weaknesses” section of a review do not necessarily imply fault on the part of the authors or the publications under review. Some weaknesses are unavoidable, given the current status of the data, methods, and confidentiality protections. Calling something a “weakness” does not necessarily imply that the authors could or should have done anything differently. Rather, it is intended to alert the reader to limitations on how the article’s results and conclusions should be interpreted and employed for purposes of motorcycle safety policy formulation.

Begg, D.J., Langley, J.D., & Reeder, A.I. (1994). Motorcycle crashes in New Zealand resulting in death and hospitalization. I: Introduction methods and overview. Accident Analysis and Prevention, 26(2), 157-164.

Langley, J.D., Begg, D.J., & Reeder, A.I. (1994). Motorcycle crashes resulting in death and hospitalization. II: Traffic crashes. Accident Analysis and Prevention, 26(2), 165-171.

Abstract
Objective. The authors have written a series of three papers that describe the epidemiology of motorcycle crashes resulting in death and hospitalization in New Zealand. The first paper describes the methods used for the study, provides an overview of all crashes, and, in particular, compares traffic crashes with nontraffic crashes. The second paper focuses on traffic crashes, and the third (not reviewed here) focuses on nontraffic crashes.

Study Populations, Data, & Methods. The source of the fatality data was national mortality data files for 1978-1987 (ten years). The source of the hospitalization data was the 1988 national morbidity file, which records all public hospital discharges in New Zealand. Motorcyclists were identified by E-code and text description of cause (which New Zealand uses to supplement E-codes). AIS scores were assigned using the ICDMAP program, based on the primary injury diagnosis. Inpatient costs were extrapolated from costs for patients at the Dunedin Public Hospital for April 1988 through March 1990 (two years).

Results. Fatalities. A total of 1,175 motorcyclist fatalities were identified for the period 1978-1987, resulting in a mortality rate of 3.6 per 100,000 persons per year. Males 15-19 and 20-24 years of age had very high motorcycle traffic mortality rates (25.2 and 26.4, respectively), especially laborers (40.0) and forestry workers (32). Maori and non-Maori had similar rates. Motor vehicle traffic crashes represented 96 percent of the fatalities, and the majority (63 percent) of traffic deaths were attributable to a collision with another motor vehicle. Drivers were the victims in 88 percent of fatalities.

Hospitalizations. During 1988, a total of 2,623 motorcycle crash victims (2,222, or 85 percent, in traffic crashes) were hospitalized, resulting in a hospitalization rate of 80.4 (68.1 traffic) per 100,000 persons per year. Males 15-19 and 20-24 years of age had very high morbidity rates (464 and 462 total, 409 and 416 traffic). Maori had a higher morbidity rate than non-Maori in traffic crashes (99 vs. 61). Drivers were the victims in 86 percent of hospitalizations. For hospitalized victims, the injury locations were the lower limb (43 percent), head (23 percent), upper limb (16 percent), trunk (9 percent), internal (5 percent), and other (2 percent). Fractures were the most common injury to the lower and upper limbs (71 percent and 80 percent, respectively), and intracranial injury was the most common head injury (71 percent). The head was the site for 65 percent of severe/critical injuries (AIS>3).

The most common (40 percent) traffic crash was a collision with another motor vehicle. Collision crashes were more likely to result in lower limb injury (48 percent vs. 38 percent), have severity of AIS-3 or higher (39 percent vs. 21 percent), and result in hospital stays longer than a week (46 percent vs. 28 percent) than noncollision crashes.Whereas 29 percent of the traffic crashes were AIS-3 or higher, the comparable figure for nontraffic crashes was 19 percent.

The median length of hospital stay was 4 days. 33 patients (2 percent) died in the hospital. Of these, 24 (73 percent) had a principal condition of head injury.

There was a significant linear increase in the fatality rate between 1978 and 1988 but no comparable trend in hospitalizations. The estimated inpatient cost for all motorcycle crashes in 1988 was NZ$13,485,628. Traffic crashes represented 92 percent of the total cost. Traffic collision crashes were approximately twice as expensive as traffic noncollision crashes ($6,942 vs. $3,342), and they accounted for 61 percent of the total costs.

Conclusion. The mortality associated with motorcycles is comparable with that from a disease such as cervical cancer. Compared to cervical cancer, motorcycle crashes have received insignificant media, research, and prevention attention.

Notes
New Zealand has a compulsory helmet law covering all motorcycle drivers and passengers riding on public roads. The authors cite another study that reports 93 percent helmet use for traffic motorcyclists and 63 percent for nontraffic.

The exchange rate in 1988 was New Zealand
$1 = U.S. $0.6556.

Strengths
These articles made a point of looking at both traffic and nontraffic crashes. Most articles either ignore nontraffic crashes completely or do not separate them from traffic crashes. The articles also separated traffic collisions from noncollisions. Injuries from traffic collisions are shown to be much more costly than injuries from the other two categories.

Details on severity and body region are provided for patients hospitalized for traffic crashes.

Hospitalized survivors and fatalities are analyzed separately.

Weaknesses
There is insufficient discussion of what inpatient “costs” represent. It is unclear whether the Dunedin figures represent charges, payments, or something else. Also, according to Table 1, the estimated national cost reflects readmissions, but it is not clear whether the Dunedin mean cost per patient visit calculation also included readmissions.

The sample is not representative. It includes only fatalities and hospitalizations, thus excluding crash survivors who were not injured severely enough to be admitted to the hospital.

Some of the figures presented appear to contradict other figures in the same article. Was the head the principle body region in 23 percent (p. 167) or 19 percent (p. 168) of hospital-admitted injuries incurred in traffic crashes?

Much of the focus is on collision rates as a share of the population or sub-population. While this might be useful for demonstrating the magnitude of the problem and its demographic locus for the purpose of attracting the attention of New Zealand policymakers, it is not a useful measure for our purposes in this study.

There was no examination of the impact of helmet use.

It is not clear to what extent this New Zealand study is generalizable to the U.S. New Zealand is more rural than the U.S. Do its motorcyclists ride primarily for recreation (as in the U.S.) or for everyday transport? The demographic profile of the riders, however, looks similar to that of the U.S., with a preponderance of young males.

Interesting Finding
New Zealand, unlike most countries, uses a free-text description to complement its E-codes. The authors took advantage of this field to look more closely at cases that were E-coded with a fourth digit of 9, indicating the victim’s role in the crash as “unspecified.” ICD-9 includes fourth digits for identifying motorcycle “drivers” and “passengers,” but if the data source identifies the victim simply as a “motorcyclist,” he will often be coded as “unspecified.” The authors found 18 motorcycle fatalities (1.5 percent) and 133 motorcycle hospitalizations (5.1 percent) that had been coded as “unspecified.” This suggests that relying solely on the fourth digits of E-codes to identify motorcyclists will result in undercounts.

Estimates of Inpatient Costs

Conclusions
Despite the flood of injury rates and the sometimes unclear distinctions between crash types, this article provided valuable information on the relative severity and costs of traffic vs. nontraffic and collision vs. noncollision motorcycle crashes. It also showed great detail on the body regions affected by motorcycle crashes.

Billheimer, J.W. (1998). Evaluation of California motorcyclist safety program.Transportation Research Record 1640, 100-109.

Abstract
Purpose & Study Population. The California Motorcyclist Safety Program (CMSP) is a legislatively mandated, statewide program that has trained more than 100,000 motorcyclists in the 10 years since its implementation in July 1987. The program is mandatory for riders under 21 seeking a California motorcycle license. The current evaluation traces motorcycle crash trends before and after the formation of the CMSP, compares crash trends in California with those in the remainder of the United States, and analyzes the riding records of matched pairs of 2,351 trained and untrained Southern California riders.

Data & Methods. Crash trends: Historical data on motorcycle crashes, registrations, and licensing for 1977-1995 (10 years before and 9 years after implementation of CMSP), broken down by number, severity, and age, were used to plot simple graphs to show trends in annual numbers of crashes, fatalities per registered motorcycle, and crashes per licensed rider for California. California statistics were also compared with national figures for the same period.

Matched-pair analysis: On-site interviewers went to motorcyclist hang-outs in search of riders. They assembled profiles identifying riders by age, sex, years riding, miles ridden per year, and primary purpose of riding (commuting, recreation, etc.). Over five years, beginning in late 1989, interviews with 16,000 untrained motorcyclists were obtained, resulting in 2,351 pairs of motorcyclists matched by the five key factors listed above. Of these pairs, 1,139 included a rider taking the basic 16-hour course and the other 1,182 included a rider taking the 8-hour experienced rider course (ERC). The initial profiling was followed up by telephone surveys of the participants. Key methods distinguishing this study from similar previous studies were 1) to differentiate between pairs of riders who had previously ridden more than 500 miles (524 pairs) and those who had not (615 pairs), and 2) executing the first follow-up survey six months after the training course instead of waiting a full year. Aggregate crash counts and crash rates were then compared separately for those who had previously ridden 500 miles and those who had not, at the six-month, one-year, and two-year marks.

Results. Analysis of statewide crash trends indicate that fatal motorcycle crashes have dropped 69 percent since the introduction of the CMSP, falling from 840 fatal crashes per year in 1986 to 263 in 1995. If crash trends in California had paralleled those in the rest of the United States over this period, the state would have experienced an additional 124 fatalities per year. In a similar analysis truncated at 1991, just before the helmet law went into effect, California would have experienced an additional 76 fatalities and 1,333 injury crashes per year. Using these latter figures, along with costs from Peck and Healy, gives an estimate of annual savings of $113 million from the program.

In the case of novice riders with less than 500 miles of prior experience, a matched-pair analysis indicates that trained riders experience less than half the crash rate of their untrained counterparts for at least 6 months after training. Beyond 6 months, riding experience begins to have a leveling effect on the differences between the two groups. In the case of riders with more than 500 miles of experience prior to training or interviewing, no significant differences in crash rates were detected between the two groups, either before or after riders took the basic training course. There was no evidence that riders electing to enter a safety course voluntarily rode any more safely than their untrained counterparts before taking training.

Survey responses showed that recent CMSP trainees consistently reported higher usage of such protective equipment as helmets, boots, and jackets.

Of those who take the basic course, 5 percent quit riding in large part because of their poor performance in the course. If the course weeds out 3,900 unpromising riders (5 percent of all students), it probably prevents 120 crashes a year and saves $5.9 million (at $49,500/crash) B more than the $1.3 million cost of the CMSP.

Questions
What is the theoretical relationship between training and the crash rate? Should training introduce a one-time drop or an ongoing drop? If one-time, over what time period? To the extent that crashes are caused by inexperienced, untrained riders, the crash rate should be a function of the proportion of all motorcycle operators who are both inexperienced and untrained. The extent to which training is a substitute for experience could be represented as a parameter in the model.

On page 102: “A similar [multivariate time series] analysis limited to the years 1988-1991 (after the introduction of the training but before the introduction of the helmet law) also showed that training had a significant impact in reducing fatalities per 1,000 registrations.” How can a test limited to the years 1988-1991 demonstrate the impact of a change that happened in 1987, before the period being analyzed? The author probably means 1988-1991 is to be compared to 1978-1987, but this is not clear.

Why did the crash rate for riders under age 25 increase so much in 1983-86? Were more miles being ridden? Young people taking up riding at a higher rate?

Are both halves of a pair being observed simultaneously in time, or are people from different time periods being matched ex post? If the latter, the riders might not be well matched - e.g., six summer months for one might coincide to six winter months for the other (or is this even a relevant consideration in southern California?).

Useful Reference
Peck, R., and Healy, E.J. “Accident Cost and Benefit Analysis.” DMV Research Notes, Winter 95/96, Sacramento, CA, 1995.

Strengths
The article uses two different and complementary analytic approaches: 1) analysis of aggregate time series and 2) an experimental matched-pair survey that compared the traffic records of two groups of subjects.

The matched-pair study introduced two small innovations that allowed it to focus on the inexperienced riders who are most likely to benefit from training: 1) distinguishing between pairs of riders according to prior riding experience, and 2) executing the first follow-up survey after just six months instead of waiting a full year.

The study also used a relatively large sample of records from true novices (a subgroup other studies fail to tap into), which is significant because the road-experience factor may, within a very short exposure time, render potential real short-term benefits of CMSP indiscernible.

The study gives a very interesting survey finding, regarding the 16 percent of course takers who quit riding within a year, with 5 percent citing the course as the reason for no longer riding (“program as effective sieve”). If this holds, and the cost formulas used are valid and applicable, then this provides a significant benefit of the CMSP frequently overlooked by other studies.

Weaknesses
The crash trends analysis used registrations and licensed riders, rather than exposure (VMT) as the denominator in calculating rates.

The author is often vague regarding his methods. This makes it difficult to critique some of the findings. What, for example, was the method for the “multivariate time series analysis” mentioned on page 102? Was it contrasted via regressor series, explicitly via a composite ratio series, or analyzed separately, and then the two intervention coefficients tested against each other for a net difference? Or does this imply that other unmentioned series were employed as regressors, such as economic data, demographic shifts, car crash rates, etc.?

Note that, while the paper admits “no guarantee that CMSP is totally responsible for these imputed savings,” it is also possible that the effects could even be understated, if other states within the comparison series were also passing CMSP training programs during this time. In sum, it is impossible to reliably judge the time series analysis “results” other than simply giving the author the benefit of numerous possible doubts and assuming the best for the many unanswered questions.

The matched-pair study suffered from a small sample size. Even though the sample was much larger than previous comparable studies, it still included just 63 crashes. Dividing these into eight cells (by training, prior experience, and length of time after training/interview) made it almost impossible to achieve significant results. In the one instance where a significant result was achieved, one wonders if it might have been a random success.

The finding of the marginal difference in crash rates within the first six months, for the <500-mile subgroup, is based on a very small count to begin with, which by itself would be nowhere near significant (5 crashes vs. 7 crashes, out of 615 pairs). But this difference is adjusted even further apart by dividing these rare event counts by mileage exposure, which itself is imputed for a majority (63 percent) of the cases. Given that, it seems a real stretch to ascribe any significance to the p=.065 finding. (We don’t know if this is a one-tailed probability or two-tailed. With the mixed results that the other results have shown, it should use the more conservative two-tailed test.) It is even more of a stretch to make much of this questionable p=.065 result when there were 4 tests performed, with 3 of those markedly non-significant, and without even a general congruence in the directional pattern among the tests.

The “one year after” category appears to subsume the “six months after” category. This cumulative treatment is misleading, essentially double-counting the first six months. If a “second six months” category were used, it would show that untrained riders had fewer crashes in the second six months.

There is not enough information from which to judge the violation comparisons, but to his credit the author mentions that among the ERC pairs, the trained group has a lower incidence of violations pre-training, again suggesting the possibility of self-selection bias among the trained group.

It is difficult to justify a matched-pair design unless there is an a priori reason to expect that pairs of cases are likely to be statistically dependent (with correlated errors), rather than just exogenously similar on a multitude of important factors.

The study matches pairs on the basis of give of the most important factors - age, sex, years riding, miles/year, and purpose of riding - but it does not match on any measure of past driving record (e.g., moving violations) which is the factor that usually best predicts crashes.

If the state requires training for all riders under age 21, where did all the untrained riders in the study coming from? Some sort of self-selection might be embodied in this group that is willing to flout the law.

The survey follow-up found that the riding frequencies of trained and untrained riders diverged after training, with trained riders riding more. This might represent self-selection, in which case it could substantially compromise group comparability. Was the response rate (37 percent) similar between the trained and untrained groups? If one group had a higher response rate than the other, then the estimates of exposure for the groups could be biased differentially - which could be crucial (or disastrous) when the author uses those estimates to adjust crash rates for group exposures, because one group is imputed to have a 28 percent higher exposure (5,500 miles vs. 4,300 miles). If the response rates are too different, and the imputed exposures biased, then the subsequent crash rates that were analyzed would be highly questionable.

The article tells us that 16 percent of trained riders quit within one year, 5 percent because of the training course. But it does not tell us the equivalent discontinuation rates for untrained riders. This makes it hard to be sure that the 5 percent would not have discontinued anyway, finding some other convenient reason to cite. This, in turn, calls into question the estimated cost savings from the 5 percent who quit riding.

See Questions for other possible weaknesses.

Conclusions
Billheimer takes on a difficult research question. The results are instructive, but they cannot be taken as definitive. This article demonstrates the difficulty of good research on driver training.

Braddock, M., Schwartz, R., Lapidus, G., Banco, L., & Jacobs, L. (1992). A population-based study of motorcycle injury and costs. Annals of Emergency Medicine, 21(3), 273-278.

Abstract
Objective. To provide a population-based injury and cost profile for motorcycle injury in Connecticut.

Study Population. Victims of motorcycle injuries resulting in death or hospital admission in Connecticut, 1985-89.

Data & Methods. Population-based retrospective epidemiologic review of Connecticut death certificates (1985-87), hospital discharge data (fiscal 1987-89), and police crash reports (1985-89). (The three data sources were not linked.) In the first two datasets, motorcycle cases were identified by E-code. Hospitals that E-code less than 70 percent of injury discharges were dropped. The resulting E-coded subsample included only 36 percent of Connecticut’s injury discharges, so counts were divided by 0.36 to create statewide estimates.

Hospital charges were used as a cost proxy. Estimates of payer reimbursement were assigned to cases by DRG, based on the weights, rates, and outlier definitions for one large acute care hospital in the state.

Results. Connecticut death certificates identified 112 deaths from motorcycle injuries for an annual death rate of 1.2 per 100,000 persons. Death rates were highest among 20- to 24-year-old men. Nonhelmeted motorcyclists were 3.4-fold more likely to die than were helmeted riders (P<.05). An estimated 2,833 motorcycle-related hospital discharges resulted in an annual hospitalization rate of 29.6 per 100,000 persons. Head, neck, and spinal injuries accounted for 22 percent of all injuries. Total hospital charges exceeded $29.3 million. Of this total, 12 percent was generated by operating room charges, 2 percent was computed tomography charges, and 6 percent was ICU charges. 60 percent of hospitalized patients had commercial insurance, and 29 percent were uninsured. 42 percent of charges ($12.3 million) was not reimbursed by payers to the hospitals.

Conclusion. Motorcycle injuries contribute significantly to Connecticut’s mortality, morbidity, and medical costs. This study suggests that a uniform helmet law would save an estimated 10 lives and prevent more than 90 nonfatal injuries in Connecticut each year at a cost savings to the state of $5.1 million. These data are crucial in advocating re-enactment of motorcycle helmet laws.

Note
During most of the study period, Connecticut had no helmet law. In mid-1989, a law took effect requiring riders under 18 to wear helmets.

Questions
Were costs adjusted to a common year’s dollars? If so, what year?

Strengths
This article presents estimates of both hospital charges and payer reimbursement for comparison. While the estimates are based on a number of assumptions, they are nonetheless very useful for Connecticut policymakers.

Weaknesses
The reported cost estimates appear to be based on hospital charges, rather than payments. This would bias the cost estimates upwards. On the other hand, the charges do not appear to include physician charges, which would cause the cost estimates to be understated.

Hospital reimbursement rates were based on a single hospital, whose rates were slightly higher than those of other hospitals in the state. As the authors note, this would tend to overstate the amount reimbursed to hospitals.

Crash victims with injuries too minor to require hospitalization were not included in this study.

The authors had no way of determining whether hospitalized survivors were wearing helmets at the time of their crash.

The authors point out that their sample of 112 fatalities from death certificates in 1985-87 is much smaller than the 214 reported in the same period by police crash reports. They were not able to explain the discrepancy.

In the discussion, head, neck, and spinal cord injuries are lumped together. The discussion would be better focused on head injuries only, for these are the injuries that helmets prevent. The proportion of spinal cord and neck injuries also seems very high.

The authors report that 42 percent of charges are not reimbursed by payers, but they do not discuss the reasons. Did the 29 percent of patients who were uninsured incur higher costs? Did payers refuse payment for insured patients?

E-coding was apparently not mandatory in Connecticut during the study period. Therefore, E-coding practice varied greatly between hospitals. The authors’ method for dealing with the problem (dividing by 0.36) assumes that motorcycle injuries are no more or less likely to be E-coded than other injuries. If records of motorcycle injury discharges are, in fact, more likely to be E-coded, then the estimates presented would be biased upwards. If, on the other hand, motorcycle crashes are often E-coded as motor vehicle crashes, without specifying that the victim was a motorcycle rider, then the estimates could be biased downwards.

Bray, T., Szabo, R., Timmerman, L., Yen, L., & Madison, M. (1985). Cost of Orthopaedics injuries sustained in motorcycle accidents. Journal of the American Medical Association, 254(17), 2452-2453.

Abstract
Purpose. To assess hospital costs, insurance profiles, and costs to taxpayers of motorcycle crashes in California.

Study Population. 51 serial admissions to the orthopedic services at the University of California, Davis, Medical Center, Sacramento, for motorcycle crash trauma with open fractures were reviewed.

Data. Motorcycle cases from a large-scale study of open fractures carried out in 1980-1983 by the Department of Orthopedic Surgery were supplemented with data on BAL at admission and disposition at discharge from hospital records, plus billing and insurance information from the patient billing and collecting department.

Results. 55 percent of those tested were alcohol intoxicated at the time of admission. 75 percent carried no insurance of any kind, and for the total group, 72 percent of hospital charges for acute hospitalization ($17,704 total per patient) were charged to the state of California, with an additional 10 percent charged to other tax-based sources.

Conclusions. Care of motorcycle trauma consumes a substantial portion of public health care funds in California. This could be reduced by legislative action concerning helmet use, licensing, and rigid enforcement of compulsory insurance.

Strengths
Detailed hospital data, including payers, on a number of motorcycle injuries.

The most interesting findings concern 1) payers and 2) intoxication. The precision of both findings is somewhat compromised by the non-representativeness of the sample, but both findings are strong enough to suggest that uninsured riders and drunk riders impose large costs on the state of California.

Weaknesses
As acknowledged by the authors, these 51 open fracture cases cannot be considered a representative sample of motorcycle injuries. Open fractures account for only about a third of all motorcycle injuries, and they tend to be significantly more serious than non-fractures or closed fractures.The article, which is based on data from 1980-83, is quite dated. The 21.2-day average length of stay is extremely high; in 1991, the average length of stay for hospital-admitted motorcycle injuries was just 7.9 days. The longer length of stay represents the combined effects of 1) the sample’s bias towards more severe injuries and 2) the less cost-conscious medical regime of the earlier period. It would be very difficult to generalize these results.

The article did not mention whether any of the 51 patients died. It is presumably to be inferred that they were all alive at the time of discharge.

The article presents a finding that 55 percent of “those tested” were legally intoxicated. The authors seem to assume that the 29 patients tested were a randomly selected representative sample. But it is likely that these patients were tested because there was reason to believe they had been drinking, whereas the 22 patients not tested had probably not been drinking at all. We are not informed whether the 13 tested patients who were not intoxicated might have tested positive at lower levels. If we assume the 22 untested patients were not drinking, this would still imply that 31 percent of the patients were legally drunk, and another 25 percent may have been drinking at lower levels.

There is no indication of whether the riders were wearing helmets when they crashed. There is also no breakdown by body region or severity, but the sample size would probably not support such analysis, anyway.

The authors’ own description of their cost proxy will serve as a sufficient criticism:

The “cost of the initial hospitalization” was defined as the total hospital charges incurred during that specific admission. Not included were charges for initial emergency service; charges incurred at an initial hospital before transfer to our institution; subsequent readmission for complications, removal of hardware, or reconstructive procedures; rehabilitation costs after discharge; transfer to skilled nursing facilities or other hospitals; and professional, legal, and administrative fees.

The use of hospital charges as a proxy for cost can cause two problems. First, it will result in an overall upward bias in the medical cost estimate, since hospitals typically overcharge. Second, it will affect the distribution of costs among payers, since some payers pay 100 percent of charges, while other payers with more clout (e.g., Medicaid) might pay less than half of charges.

On the other hand, the omission of all costs other than hospital charges will bias the cost estimate in the opposite direction - downwards. The omission of professional fees - the amounts paid directly to doctors - is particularly egregious (we estimate that these fees typically add 30 percent to the cost of a hospital stay for injury). The net effect of these opposing biases is probably an underestimate of medical costs, but there is no way to be certain.

Bried, J.M., Cordasco, F.A., & Volz, R.G. (1987). Medical and economic parameters of motorcycle-induced trauma. Clinical Orthopaedics and Related Research, 233, 252-256.

Abstract
Study Population & Methods. A retrospective study was conducted on all patients injured in a motorcycle accident who were admitted to the Arizona Health Sciences Center during a one-year period, July 1984 through June 1985. Researchers examined the paramedic report, ED report, and inpatient records.

Data & Results. The 71 hospital-admitted patients evaluated averaged 26 years of age; 79 percent were men, 75 percent were not wearing a helmet, and 24 percent were legally intoxicated. 66 percent required surgical intervention and 36 percent a second procedure. 52 percent required ICU care, (mean, 3.3 days) and 30 percent required a ventilator (mean, 6 days). There were 167 fractures, with an average of 2.4 per patient. The 27 patients requiring a blood transfusion averaged 10.5 units per patient.

Motorcyclists not wearing a helmet had an increased risk of head injury (p<.01) (see table). Those with head injuries had an increased need for intensive care (p<.0001) and a ventilator (p<.001). Patients with head injuries more commonly sustained fractures about the shoulder (p<.015) than fractures to the lower extremity (p<.005). The average hospital stay was 13 days, with average charges of $16,408 per patient. Charges were lower for motorcyclists wearing a helmet ($13,368 vs. $17,120). Charges were significantly higher in patients with a head injury ($21,945) than in patients without a head injury ($11,941). Patients sustaining a head injury were less likely to return to baseline functioning (p<.001). Of the 12 patients who became permanently impaired, none had been wearing a helmet and 10 sustained a severe head injury.

Relationship of Head Injury to Helmet Usage

Questions
Did the study really cover patients admitted to this hospital? How did it happen that helmet use was known for every patient?

Strengths
The cost proxy captured physician charges as well as hospital charges. And while, as the authors acknowledge, the cost estimates do not include post-hospital costs, a separate estimate of SCI rehabilitation costs is offered.

The article provided considerable detail on fractures.

The use of “baseline functioning” to operationally define permanent disability allows the authors to get at a question rarely examined in the motorcycle injury literature.

Weaknesses
The sample was small. It included only 18 non-helmeted riders.

The authors report a difference in charges of $17,120 vs. $13,368 for non-helmeted vs. helmeted, but they do not state whether this difference is statistically significant.

Raw hospital and physician charges are used as a proxy for costs. This is likely to bias the medical cost estimates upwards because hospital charges are greater than costs.

The study was not representative of the whole population of motorcycle injury victims. It captured only those victims who were injured severely enough to be admitted to a hospital. It would also have been useful to know more about the hospital where the study was performed - in particular, whether or not it is a trauma center.

Conclusions
Though short and simple, this article not only added more support to the usual arguments for the effectiveness of helmets, but it also shed light on the problem of permanent disability and suggested a negative correlation between fractures of the head/shoulders and fractures of the lower leg.

Interesting Finding
Head injuries in motorcyclists are often accompanied by fractures in the shoulder region, but rarely by fractures of the lower leg. Only 1 patient of the 71 covered by this study suffered fractures in both regions.

Interesting Reference
“In 1982, treatment for a spinal cord injury at the Phoenix Regional Spinal Cord Injury Center averaged $75,300 for a quadriplegic requiring an average stay of 184 days.” (Young, Burns, Bowen, & McCutchen. (1982). Spinal Cord Injury Statistics. Phoenix, AZ, Good Samaritan Medical Center, p. 33.)

Hell, W. & Lob, G. (1993). Typical injury patterns of motorcyclists in different crash types -- effectiveness and improvements of countermeasures. 37th Annual Proceedings of the Association for the Advancement of Automotive Medicine, 77-86.

Abstract
Purpose & Study Population. There is a high injury rate in motorcycle crashes. In contrast to automobiles, there have been no improvements in the passive safety of motorcycles in the last few decades. The motorcyclist drives without an impact absorbing zone, and his possible safety measures are limited to a safety helmet and appropriate safety clothing. His kinematics in a crash can take many different forms. The object of this study is the investigation of 173 real-life motorcycle crashes at the scene of the accident, to get a coherent picture of the severity and injury pattern in different crash-types. Additionally, the efficiency of protective measures (147 articles of clothing and 85 crash helmets) was analyzed.

Data. 173 motorcycle crashes, involving 210 riders, in 1985-1990 in the Munich (Germany) area were documented at the scene by the Bavarian Police. This was followed up by a medical analysis in hospitals and an injury rating using AIS and ISS.

Of the 210 riders, 50 (24 percent) died. This compares to a 2 percent fatality rate for all police-reported motor vehicle crashes in Germany.

Methods. The authors did tabular analysis of their data by severity, body region, and crash type (i.e., the relative positions and directions of travel of the motorcycle and its crash opponent). They also looked at the performance of leather clothing and helmets with respect to injury severity.

Results. Of all crashes involving injuries of moderate or greater severity (AIS>=2), 43 percent involve injuries to the head, followed by lower extremities (37 percent), upper extremities (30 percent), thorax (25 percent), abdomen (16 percent), spine (12 percent), and pelvis (8 percent).

Severity and body region vary greatly depending on the crash type. Three of the eleven crash types account for 41 of the 50 fatalities: head-on collisions 1) with the front of an opposing vehicle, 2) with the side of an opposing vehicle (without fly-over), and 3) with a stationary object (e.g., tree, crash barrier). These same three crash types also have the highest rates of head injury and the highest ISS.

Two-thirds of crashes involved an opponent; the other one-third were solo. Two-thirds of crash opponents were cars, one-sixth were commercial vehicles. Most collisions with cars (90 percent) struck the front or side of the car, while a plurality of collisions with commercial vehicles (48 percent) struck the rear end. The motorcycle’s first impact point was in front in 67 percent of cases, and in the side in 29 percent.

Of 147 riders whose clothing was analyzed, 45 percent wore leather jacket and leather trousers (the article says 55 percent, but this would make the total 110 percent, and figure 4 shows that only a minority of riders wore leather trousers), 29 percent wore a leather jacket, and 26 percent wore no leather. Wearing leather safety clothing was found to be correlated with lower severity of injuries to the extremities, particularly the legs.

Of the 210 riders, 205 wore helmets. But 30 of these lost their helmets during the crash. Of 85 helmets that were investigated intensively, 16 were lost and 11 were broken in the crash. For the 58 cases where the helmets remained on riders’ heads, AIS of 0 to 2 are predominant, while more severe injuries were common when the helmet was lost. When the impact was hard enough to crack the helmet, it usually killed the victim, as well (7 of 11 cases). The most common impact regions on the helmet were the mandible (chin bar) and forehead.

Conclusions. The authors make recommendations to industry and legislators regarding the design of protective clothing, helmets, motorcycles, and other vehicles. Some of these do not derive from the data presented.

Questions
Did the data include survivors who were not medically treated? Or who were treated outside hospitals (e.g., in doctor’s offices)?

Weaknesses
The article contains no direct information on costs or payers.

Some results (e.g., body regions injured) are not separated by fatal/nonfatal (AIS 6 does not include most fatalities).

Strengths
The authors make thorough use of their data, analyzing it by 1) body region vs. severity, 2) body region vs. crash type, 3) ISS and severity vs. crash type, 4) leather clothing use vs. severity, 5) helmet performance vs. severity, and 6) helmet impact region. Several of these dimensions of motorcycle crashes (2, 3, and 4) are not commonly analyzed. Therefore, this article makes a number of unique contributions to the motorcycle crash literature.

Conclusion
With this article, Hell and Lob have made a unique and valuable contribution to the literature on motorcycle safety.

Karlson, T.A., & Quade, C.A. (1994). Head injuries associated with motorcycle use - Wisconsin, 1991. Morbidity and Mortality Weekly Report, 43(23), 423, 429-431.

Abstract
Objective. To assess 1) the risk of head injury for motorcyclists in motor-vehicle crashes, 2) the initial inpatient hospital charges for motorcyclists with head injuries resulting from crashes, and 3) the reduction in injuries and fatalities associated with universal helmet use.

Study Population. Motorcyclists hospitalized for injury after police-reported crashes in Wisconsin in 1991.

Data & Methods. Police Accident Reports and inpatient discharge records for acute-care hospitals were linked through a probabilistic method (which calculated the likelihood that a police report and a discharge record represent the same person) using the date and location of the event, the victim’s date of birth, sex, and home Zip Code, and whether the victim was transported by ambulance. For uncertain matches, additional information was reviewed manually, including diagnoses and E-codes. About 7 percent of the automated matches were determined to be incorrect.

Head injuries were separated into three categories: 1) brain injury, 2) skull fracture without intracranial injury, and 3) concussion with only brief (less than one hour) or no loss of consciousness.

Results. Of 3,184 motorcyclists involved in police-reported crashes, 2,015 (63.3 percent) were unhelmeted and 994 (31.2 percent) were helmeted. Helmet use was unknown for 175 (5.5 percent). Of those for whom helmet use was known, 545 were hospitalized and 74 died. (See Table 1, below, for more detail.)

Of the 545 hospitalized, 187 (34.3 percent) sustained head injury. (See Table 2, below, for a detailed breakdown of these 187 cases.) Of the 2,015 unhelmeted crash victims, 7.6 percent suffered head injuries, while only 3.4 percent of the 994 helmeted victims suffered head injuries. Unhelmeted head injury victims also incurred an average of 59.2 percent greater hospital charges for the initial admission.

Conclusion. Universal helmet use by all motorcyclists in Wisconsin during 1991 potentially would have prevented 60 brain injuries, 13 skull fractures with no intracranial injury, 8 concussions, and 14 deaths. It could also have saved society more than $400,000 in medical charges for initial hospital admissions.

Note
During the study period, Wisconsin required helmet use only for motorcycle riders under 19 years old. The observed helmet use rate is 42 percent of all riders.

Questions
Did the linkage technique fail to link any crash reports with hospital records? If so, how many? How do the authors know that 7 percent of the matches were incorrect?

How did Wisconsin achieve 94.5 percent reporting on helmet use? No other state without a universal helmet law has such a high rate.

Strengths
The authors began with reliable estimates of all motorcycle crash victims in the state, and used these figures as denominators in their rate calculations.

The goals, methods, and conclusions of the analysis were consistent with the limited nature of the linked dataset, which included only hospital-admitted survivors. Since the analysis focused on relatively severe head injuries, it is likely that few relevant cases were omitted by excluding non-hospitalized injuries. The explicit, if perfunctory, analysis of fatal cases separately from nonfatal provided a necessary complement (though it could, perhaps, have been made more relevant if head injuries could have been identified among the fatalities).

Crash victims whose helmet status was unknown were accounted for explicitly (although they were not included in the analysis of head injuries or the resulting cost estimates).

Weaknesses
The presentation of the results carefully avoided showing the helmet-status breakdown of the 545 hospitalized motorcyclists whose helmet status was known. (Perhaps the space and format constraints of the MMWR prevented the raising of a question that would require a digression on the prevalence of limb injuries among helmeted motorcyclists who are hospitalized.)

Charges for initial hospital visits were used as a proxy for health care costs. Professional fees and other costs were not accounted for, which would cause costs to be underestimated. (The absence of accounting for long-term costs is particularly serious for brain injuries, which can have long-term consequences.) On the other hand, charges were not adjusted to reflect the fact that hospitals overcharge.

Table 1.
Victims of Police-Reported Motorcycle Crashes, Wisconsin 1991:
Distribution by Injury Status and Helmet Use

Table 2.
Motorcycle Crash Victims Hospitalized for Head Injury, Wisconsin 1991:
Distribution, Injury Rates, and Average Charges by Injury Type and Helmet Use

The estimation of savings from universal helmet use assumed that the observed distribution of outcomes for helmeted riders would be replicated in the unhelmeted riders if they were to wear helmets. This might not be the case if helmet use is correlated with other riding behaviors (e.g., speed, alcohol consumption) or rider characteristics (e.g., experience).

Conclusions
This article achieves its modest goal, clearly demonstrating that helmet use is associated with lower rates of head injury and lower hospital charges for injured motorcyclists.

Kelly, P., Sanson, T., Strange, G., & Orsay, E. (1991). A prospective study of the impact of helmet usage on motorcycle trauma. Annals of Emergency Medicine, 20(8), 852-856.

Abstract
Objective. To determine the effect of the use of a motorcycle helmet on reducing the mortality, morbidity, and health care costs resulting from motorcycle crashes.

Study Population & Data. All motorcycle crash victims presenting less than 24 hours after injury for whom helmet information was known. Data were collected during April-October 1988 (the “motorcycle season,” seven months long) at the emergency departments of eight medical centers across Illinois, including representatives from urban, rural, teaching, and community facilities.

Method. A prospective, multicenter study of all eligible motorcycle crash victims. The examining physician filled out the questionnaire for each patient and a member of the research team recorded follow-up information from the treating facility and from the patient, the family, or paramedic or police reports. County coroners were contacted for information on fatalities taken directly to the morgue.

Results. Of 398 patients, 58 (14.6 percent) were helmeted, and 340 (85.4 percent) were not. The nonhelmeted patients had higher Injury Severity Scores (11.9 vs. 7.0), sustained head/neck injuries more frequently (41.7 percent vs. 24.1 percent), and had lower Glasgow Coma Scores (13.73 vs. 14.51). 25 of the 26 fatalities were nonhelmeted patients. By logistic regression, the lack of helmet use was found to be a major risk factor for increased severity of injury, along with speed and mechanism of injury.

Health care charges were found to be 23 percent higher for nonhelmeted patients (average charges $7,208 vs. $5,852). Nonhelmeted patients were also transported by ambulance more frequently (63.4 percent vs. 46.4 percent) and more likely to be uninsured (54.6 percent vs. 44.2 percent), though none of these differences was significant at the 5 percent level.

Conclusions. Helmet use may reduce the overall severity of injury and the incidence of head injuries resulting from motorcycle crashes. A tendency toward higher health care costs was demonstrated in the nonhelmeted patients.

Questions
Mechanism of injury is said to be a predictor of injury severity. Which mechanisms are associated with more severe injuries -- vehicle-vehicle, vehicle-fixed object, overturn, or other? How did the logistic regression account for mechanism?

Data were collected only for patients whose helmet status was known. Does this mean that the questionnaire was not distributed to patients whose helmet status was unknown to the researchers a priori? Or does it mean that they dropped those patients from their sample who did not report their helmet status on the questionnaire? For how many patients was helmet status unknown?

Strengths
The study collected data on interesting and relevant factors, including a few that are often overlooked or not adequately measured: time of day, type of roadway, estimated speed, license possession, ambulance transport, and blood alcohol.

Despite data limitations, the study found four factors that are good predictors of injury severity -- speed, mechanism of injury, helmet status, and age -- all of which are highly significant.

The analysis by severity, body region, and helmet use was appropriate, though perhaps more detail could have been reported.

For at least some of the analysis, fatalities were examined separately from survivors. For other analysis, this distinction was taken into account in the severity measure.

Weaknesses
Physician charges for inpatients were not included in costs. This would tend to lead to an underestimate of medical costs.

The small sample of nonhelmeted riders (58, or 14.6 percent) made it very difficult to obtain significant results. Likewise, the sample included only 26 fatalities, and only 1 of these was nonhelmeted.

The study was not representative of the whole population of motorcycle injury victims. It captured only those victims who died or received medical treatment in an ED. Those who were injured less severely were missed.

The payers reported were apparently expected, rather than ultimate, payers. Also, more payer detail could have been reported.

No explanation was offered for the significance of age in predicting injury severity. Did it proxy for riding experience? Or for risk aversion?Interesting Finding

Figure 2 shows that helmeted riders had higher rates of limb injuries, but lower rates of external injuries (e.g., lacerations, abrasions, contusions). The former is a common finding, often resulting from the selection bias inherent in analyzing only those crash victims who were injured. But the latter result is contrary to this bias. (Does it suggest that riders who wear helmets are also more likely to wear protective clothing, which can prevent minor external injuries?)

Conclusions
This article employed good methods in collection and analysis, but the small sample of helmeted riders constrained the reporting and the conclusions. The authors were still able to obtain significant results regarding the importance of various factors in motorcycle injury severity.

The medical cost estimates show little detail. Moreover, they embody the usual problems of state-level hospital studies -- reliance on raw hospital charges and exclusion of physician charges. Nonetheless, despite the small sample, the study arrives at results similar to those from other states, pointing to the efficacy of helmet use.

Max, W., Stark, B., & Root, S. (1998). Putting a lid on injury costs: The economic impact of the California motorcycle helmet law. Journal of Trauma, Injury, Infection, and Critical Care, 45(3), 550-556.

Abstract
Objective. To analyze the effect of California’s motorcycle helmet law, which took effect in 1992, on injury costs.

Study Population. All victims of motorcycle injuries in California, 1991-93.

Data & Methods. California state hospital discharge data, 1991-93. Motorcycle injuries (11,163) were selected by ICD-9 E-code. Cases were classified by admission date. Costs were calculated from charges using hospital-specific cost/charge ratios. Professional fees were estimated as 25 percent of hospital costs. Transfers were identified by discharge and admission dates, discharge disposition, and admission source. Unique record-linkage number were used to link readmissions, rehabilitation, and (hospital-based) nursing home care.

San Diego County cost data, 1991-93. Data on costs of ambulance transport, emergency air transport, ED visits, and follow-up ambulatory care visits were obtained from billing records. These costs were adjusted downwards by the ratio of California/San Diego per diem hospital costs.

Statewide Integrated Traffic Records System. This dataset contains all motor vehicle crash reports submitted by local and state police. Of four severity categories - killed, severe wound, other visible injuries, complaint of pain - it was assumed that the middle two were all treated in a hospital or ED. The count of hospital admissions was subtracted from this total, and the remainder were assumed to be treated in the ED and released.

Head and spinal injuries, identified by ICD-9 diagnosis, were considered separately.

Hospital costs, all direct costs, and lost productivity were analyzed separately. Total and per person costs were compared by year and by head and spinal injury status. Each outcome was compared for 1 year before and 2 years after the implementation of the helmet law.

Productivity losses from fatalities were valued at average lifetime earnings (including an imputed value for household production) by age and sex, discounted at 3 percent. Years of potential life lost by motorycycle fatalities also were estimated.

Results. Total medical care costs for motorcycle injuries were $35 million less in 1993 than in 1991, a reduction of 35 percent. Costs decreased for all payer categories, and 73 percent of the reduced hospitalization costs were attributable to reduced costs for patients with head injuries. Head injuries also accounted for a disproportionate share of the reduction in the rate (60 percent) of hospitalized injuries per 100,000 motorcyclists. Initial hospital costs for patients with head injuries averaged $18,527, compared with $10,350 for patients without head injuries. (See table 1, below, for more results.)

Conclusions. During the first 2 years of implementation of California’s helmet law, there were reduced costs for injuries and fatalities and large dollar savings to the state and other payers compared with the previous year.

Strengths
Comprehensive datasets. The California hospital discharge data capture all hospitalizations in the state, and mandatory E-coding enables selection of motorcycle riders. The Statewide Integrated Traffic Records System captures all motor vehicle crashes in the state.

Hospital-specific cost-to-charge ratios were used to convert charges to costs.

The helmet law took effect on 1 January 1992, making for a clear cut-off between pre-law and post-law periods. Data were collected for one year before and two years after law’s implementation.

The study attempts to account for all medical costs from ambulance through rehabilitation, for both hospital-admitted and non-admitted patients.

The study accounts separately for fatalities and survivors.

Some critics of California’s helmet law have attributed the fall in injury rates after passage of the law to the decline in motorcycle registrations. But Max et al. show calculations demonstrating that injury rates fell much faster than registrations between 1991 and 1992.

Weaknesses
Rates were calculated using the number of registered motorcycles, rather than VMT, as the denominator.

Professional fees were calculated as a flat 25 percent of hospital costs.

Medical expenditures were found not to be normally distributed, but the actual distribution shape was not discussed, nor were the consequences.

Why were non-fracture spinal cord injuries (ICD 952) not counted as spinal injuries?

There was no mention of the productivity losses of non-fatal injury victims. Productivity savings would therefore be underestimated.

Conclusion
The stark contrast between 1991 (before the law) and 1992-93 (after the law) in the number and cost of head injuries leaves little room for quibbling with the article’s conclusion that the helmet law saves lives, health, and money

Table 1
Results from Putting a Lid on Injury Costs:
The Economic Impact of the California Motorcycle Helmet Law

McSwain, N.E., & Belles, A. (1990). Motorcycle helmets, medical costs and the law. Journal of Trauma 30(10), 1189-1199.

Abstract
Objective. Since 1975, 26 states have repealed or modified their motorcycle helmet laws. Louisiana (LA) reinstated the helmet law in 1982. The medical and financial impact of repeal in Kansas (KS), reinstatement in LA (accident, fatality, and critical injury rates) have been studied through 1987. Current FARS data and studies from KS, LA, 10 states, and 5 countries are compared and reported.

Study Populations, Data, & Methods. Texas. The authors selected 99 patients transported from the scenes of motorcycle crashes by the Bexar County EMS from September 1986 to December 1987 (16 months). Eight other subjects were dropped because of unknown helmet use, or because they were transferred to or from the Medical Center Hospital, from which some data were not available. A medical student assigned AIS scores to each patient. To look at rates of fatality, hospitalization, injury, and helmet use, 6,600 Texas motorcycle collisions from the same time frame were examined.

Louisiana. 1) A detailed study examined 616 motorcycle crashes (of 704 total; including 555 injuries with 16 fatalities) in three major population centers - Lake Charles, Baton Rouge, and New Orleans - in June-September of 1981 and 1982 (two periods of four months). 2) The fatality study included all 15,741 motorcycle fatalities in Louisiana in 1981-1987 (seven years). Data sources included summary reports from the Highway Safety Commission, crash reports, hospital discharge files, coroner’s reports, and helmet observation studies.

Kansas. The methods were referenced to other articles. One of these, published by NHTSA, was “Impact of the repeal of the Kansas mandatory motorcycle helmet law: 1975-1978,” (DOT HS-805 773, and executive summary DOT HS-803 959). It outlines the following methods: Data were collected retrospectively for the months July-September of the years 1975-1978 (four periods of three months). During this period, the number of riders involved in crashes rose from 410 to 508, while the number for which helmet data were available fell from 353 to 278. Identification of motorcycle crashes and victims was carried out with the aid of various state agencies, notably the Kansas Department of Transportation, whose data included time and location of crash, type and number of vehicles involved, road conditions, number of riders, and a rough determination of the extent of injuries suffered. Hospitals in the study area helped to confirm the identities of injury victims and assign AIS severity scores to the injuries. In addition, an observational survey of 2,000 motorcycles was carried out during the July-September 1977 at representative locations in the study area. Observations were distributed among the hours of the day and days of the week.

Results. After repeal of the helmet law, the LA user rate dropped from 97 percent to 50 percent. With reinstatement, the user rate rose to 95 percent. Average hospital stay for helmeted (H) riders is 5.8, non-helmeted (NH) 11.8. Fatality rate per 1,000 motorcycle registrations is 6.2 NH, 1.6 H. Changes effected through LA helmet legislation: fatality rate was 1.17 (1981), falling to 0.44 (1987) with legislation (62 percent decrease); fatality rate per 1,000 crashes fell 28 percent, from 42.68 NH to 30.81 H; injury rate dropped from 84 percent of crashes to 73 percent; and the number of critical injuries was reduced by 44 percent (1981 to 1987). Risk of head injury: NH 2.07>H. Risk of fatality: NH 1.44>H. Crash rate is less with helmet legislation than without (19 percent KS, 48 percent LA).

The medical costs (LA 1981-1987) decreased 48.8 percent. Length of stay decreased 37 percent. The major impact hospital stay >20 days: 80 percent decline. Cost of long-term disability >30 days: 81.2 percent decrease (LA). Average disability was 26.7 days H vs. 51.1 days NH (KS); 25.5 percent H required hospitalization vs. 41.6 percent NH. Medical costs: NH 306 percent>H (KS). In Bexar Co., only 18 percent of NH riders’ charges had been paid, compared with 59 percent of H riders’. In LA, short-term disability costs per case fell from $18,314 in 1981 to $17,301 in 1982, while average long-term (>30 days) disability costs fell from $29,800 to $5,600 in the same period. In 1989 dollars, $120.8 million of additional medical care and rehabilitation expenses per year were due directly to non-use of helmets (U.S.). $4.9 billion was absorbed by the public in the form of increased taxation, higher insurance costs, and lost taxes.

Conclusion. Motorcycle helmet legislation decreases medical costs. In this era of spiraling health care costs, legislation mandating the use of protective helmets should be considered as a viable alternative to raising taxes.

Strengths
This article tapped a wide array of data from a number of original studies using data from Louisiana, Texas, and Kansas. It also summarized a wide scope of literature in the field.

The Bexar County study included potentially valuable information on ICU days and percent of charges covered. The Louisiana study included information on both short-term and long-term disability. (The helmet law appears to make a huge difference on long-term disability costs; unfortunately, there is no indication of sample size or statistical significance, so it is difficult to evaluate the result. A few high-cost patients can skew the means, especially if the sample is small, as is likely for motorcyclists in long-term care for their injuries.)

Weaknesses
The number of different studies rolled into this one paper makes for some confusion. It is not always clear which data are the source of a given result.

For Louisiana, the authors often compared 1981 with 1987, rather than with 1982. Any changes resulting from the helmet law should have occurred relatively soon after the law took effect. Stretching out the comparison period increases the likelihood of contamination by other influences. Still, the authors sometimes used 1982, giving the appearance of choosing the comparison year that gave the strongest result.

The authors attributed the falling collision rate in Louisiana to the helmet law. They offer no theoretical reason why this should be so. Moreover, there was no reduction in the first year after the law took effect; the entire reduction took place in subsequent years. It appears that other factors were at work in the state after 1982, reducing the collision rate.

Many of the numbers in the abstract are not contained in the text. Some appear contradictory, others represent different computations.

Rate comparisons use motorcycle registrations as the denominator. There is no discussion of how mileage per motorcyclist might have been changing.

Details on data and methods are often lacking, as are sample sizes and indications of significance, making it difficult to put some of the reported results in context and evaluate them.

There is no discussion of the generalizability of the results from LA, KS, and TX. Without knowing more about the motorcycle-riding populations of these states, it is difficult to apply the results more broadly.

Conclusion
The authors bit off more than they could chew in one article. The datasets assembled for this article deserve separate, more focused articles that can present the data, methods, and results in a more coherent fashion. As it stands, this article contains a lot of valuable results, but it takes a lot of work to mine them, dust them off, and evaluate them. Still, it might be worth the effort to do so if the data in this article can fill data gaps, such as disability costs.

Miller, T., Levy, D.T., Spicer, R.S., Lestina, D.C. (1998). Allocating the costs of motor vehicle crashes between vehicle types. Transportation Research Record 1635, 81-87.

Miller, T.R., Spicer, R.S., Lestina, D.C., Levy, D.T. (1999). Is it safest to travel by bicycle, car, or big truck? Journal of Crash Prevention and Injury Control, 1(1), 25-34.

Miller, T.R. (1994). Costs of safety belt and motorcycle helmet nonuse. Testimony before the Subcommittee on Surface Transportation, House Committee on Public Works and Transportation, hearing on Designating the National Highway System, as required by the Surface Transportation Efficiency Act of 1991 (ISTEA), March 3, 1994.

Abstract
The two published articles employ essentially the same methods to address similar policy questions. The testimony also uses the same methods, though they are not explicitly described. Hence, the reviews have been combined, and findings from all three sources are included here.

When different types of vehicles are involved in a crash, the harm to occupants tends to vary with the weight of the vehicles involved. In determining the appropriate level of government expenditures for traffic safety, costs in multi-vehicle crashes involving different vehicle types must be allocated between occupants and nonoccupants of a particular vehicle type.

Purpose. These studies compare highway crash incidence, injuries, and costs by vehicle type. Four methods for allocating costs among different vehicle types are considered in the first article (1998); the second (1999) includes three of these methods. The methods correspond to different perspectives, including that of occupants of a vehicle and that of society under different property right assignments. Costs based on the allocation methods for the United States as a whole and per vehicle mile are estimated and presented by vehicle type. The 1999 article also includes bicycles, railroad transportation, and air travel.

Study Population, Data, & Methods. Annual crash and injury incidence were estimated using Crashworthiness Data System (1988-1991), National Automotive Sampling System (1982-1986), General Estimates System (1992-1993), and Fatality Analysis Reporting System (1993) data. Costs were computed based on restraint use, body region, and threat-to-life severity of the injury. Costs estimated included medically related costs, emergency services, property damage, lost productivity, and quality of life. Costs were then allocated between vehicle types using the different allocation methods in order to answer comparative safety questions. The cost allocation methods are:

• Occupant cost method: The cost of each victim is assigned to the vehicle type that he or she occupied. Hence, the costs of all motorcycle crash injuries are allocated to motorcycles.

• Optimal externality method (omitted from the 1999 article): All costs are assigned to every vehicle in the crash, and then to adjust for double-counting of multi-vehicle crashes, the total costs for each vehicle type (e.g., passenger cars) is multiplied by the proportion of total costs represented by that vehicle type. For example, if passenger cars account for $100,000 dollars of the total cost of $300,000, then passenger cars are assigned a cost of $33,333 (100,000 times 100,000/300,000).

• Excess cost method: Costs are allocated to the heaviest vehicle to the extent that these costs exceed costs that would have resulted had the lighter vehicle been involved in a crash with another similar lighter vehicle.

• Heaviest vehicle method: All costs are assigned to the heaviest vehicle.

Results & Conclusions. The method of allocating costs to vehicle type was found to be an important determinant of the relative cost for each vehicle type. Motorcycle crashes accounted for 2-5 percent of total crash costs (using the willingness-to-pay methodology). However, motorcycle crashes account for the greatest cost per 1000 vehicle miles ($1,960 over three times greater than the cost for buses, the next most costly per mile) using the excess cost method the authors prefer.

Motor vehicle and bicycle crash costs total $389 billion annually (as reported in the 1999 article); 75 percent resulting from passenger vehicles. Motorcycles and bicycles have the highest costs per 1,000 vehicle and passenger miles; costs per victim are highest for pedestrians, bicyclists, and motorcyclists. Costs per vehicle mile for heavy trucks and passenger cars are comparable but exceed costs for light trucks. Passenger vehicle occupants are safest if a crash occurs. Light truck, other single truck, and bus occupants have the lowest cost per passenger mile, but higher costs than air and rail travelers. Motorcyclists face the greatest risks. Combination trucks may not impose an excess risk to other drivers, but their drivers face large risks.

Detailed findings from the two articles as well as the testimony are presented at the end of this review in several summary tables. (below)

Corrections
There are two corrections to the 1998 paper: The WTP cost for fatalities in Table 1 ($2,716) should read “$2,716,000.” On page 85, in the “Table 4” paragraph, the second-to-last sentence reads: “Other single vehicle truck injuries in multi-vehicle crashes may be more severe.” The word “vehicle” should be removed, so that it reads, “Other single truck injuries . . .”

Strengths
Total costs, including medical, productivity, property, and quality of life, are calculated using both human capital and willingness-to-pay methods. The cost calculations take account of injury severity, including fatal vs. nonfatal.

The studies provide a useful comparative perspective in terms of the relative risk and cost of crashes among motor vehicle types. They also point out the need to consider carefully the perspective of the analysis (cost to occupants vs. society) and hence the method for cost allocation. Whatever methodology is used to compare crash types, motorcycles are the most costly and dangerous mode of transportation of those studied in terms of cost per miles driven. A useful analysis is provided of external costs, i.e., costs not borne by the involved parties, that is useful for public policy purposes.

Weaknesses
The cost allocation methods are difficult to follow. There is no breakdown of results by helmet use or body region.

The analysis assume that the light truck category is heavier than the passenger car/van category, even though the latter includes mini-vans and sport utility vehicles, which are heavier than many pickups. The variation within these two categories is probably greater than the difference , them.

The methodologies assume that crashes are the fault of the vehicle, and leave no allocation to the driver.

The articles do not develop independent estimates of injury costs; rather they use previously published estimates. While this is not a weakness of the articles (the studies were done to allocate costs, not estimate them), it is a weakness for our purposes of reviewing the literature on crash costs.

Conclusions
These articles are useful for indicating the relative importance of motorcycle crash costs relative to the costs of other motor vehicle crash costs. The findings are unambiguous for our purposes, i.e., looking at motorcycle crashes. Whatever method used or rate calculated, motorcycles are more costly than other motor vehicle types per vehicle mile (except railroads), per person miles, per victim, and per survivor. However, motorcycles are among the least costly in terms of total costs because there are relatively few motorcycles on the road compared to cars, trucks, and other vehicles. While the cost data in this study are not presented in sufficient detail to be useful for any specific cost estimation, they provide a valuable comparative perspective and are quite useful for context.

Additional comments
Method 3 assigns excess costs to the heavier vehicle except in the case of motorcycles, reasoning that “. . . these costs are not due primarily to the weight of the other vehicle (i.e., they reflect the lack of protection afforded to riders) . . .” But similar reasoning could be applied to subcompact cars - their drivers are assuming the risk of a known hazard - collision with a much heavier vehicle. Just as methods 3 and 4 assign highway property rights to the drivers of lighter vehicles (except motorcycles), a fifth method could assign property rights to the drivers of heavier vehicles, on the grounds that their vehicle choice shows that they value these rights (or their safety) more highly.

The “Method 3” section concludes: “Assignment of costs in the case of motorcyclists and nonoccupants is also somewhat judgmental and depends upon the purpose of the analysis.” This statement applies to all vehicles and all four methods. It might have been helpful for the authors to spell out more explicitly what judgments and purposes are associated with each method. For example, the property right preference for lighter vehicles might be attributed to the desirability of avoiding a vehicular arms race, in which all drivers are encouraged to armor themselves in the heaviest vehicle available.

Summary Tables of 1993 Motorcycle Crash Costs
from articles by Miller et al.
(all costs in September 1995 dollars)

Costs of Highway Crashes by Method of Analysis

Breakdown of WTP Motorcycle Cost, by Excess Cost Method

Ultimate External Costs of Highway Motorcycle Crashes...

Annual Cost of Motorcycle Crashes
by Cost Category (WTP, excess method)

Motorcycle Crash Victim Costs...

(numbers in parentheses identify method used, as per table at top of page)

Injury Costs per Motorcycle by Helmet Use, $1992

Annual Public Cost Savings with Universal Helmet Use, $1992

Muelleman, R.L., Mlinek, E.J., & Collicott, P.E. (1992). Motorcycle crash injuries and costs: Effect of a reenacted comprehensive helmet use law. Annals of Emergency Medicine, 21(3), 266-272.

Abstract
Objective. To document the effect of a reenacted comprehensive helmet use law on injuries and fatalities.

Study Population & Data. 671 patients reported as injured to the Nebraska Department of Roads in the period from one year before through one year after the reenactment on January 1, 1989, in two urban counties representing 40 percent of Nebraska’s population.

Methods. Retrospective before-and-after analysis.

Results. The helmet use law was temporally associated with a 26 percent decrease in the reported rate of motorcycle crashes in Nebraska compared with five other midwestern states. There were sharp declines in the number (and rates) of reported injured, hospital transports, hospital admissions, severe nonhead injuries, severe head injuries, and deaths. Serious head injuries (AIS Ž 3) decreased 22 percent. The percentage of injured motorcyclists with serious head injuries was significantly lower among the helmeted motorcyclists (5 percent) than among the unhelmeted cyclists (14 percent) for the two years combined.

The total acute medical charges decreased by $324,648 (38 percent) after implementation of the helmet use law, consistent with the overall declines in crashes, injuries, and admissions after the law went into effect. Because the study represented 46 percent of the reported injured motorcyclists in the state, the total decrease in charges for the state may have been more than $700,000. If Rivara et al.’s estimate that acute hospitalization charges for injured motorcyclists amount to 60.5 percent of the total medical costs, then the helmet use law may have decreased total medical costs in Nebraska by more than $1.1 million.

Of the patients with known insurance status, 59 percent had private insurance, 34 percent had no insurance, and 7 percent had Medicaid or Medicare. Of the nearly $1.4 million charged over both years, 48 percent was either unpaid or paid by state funds.

Conclusion. The reenactment of a helmet use law resulted in fewer crashes, fatalities, and severe head injuries.

Acute medical charges and payments
by payer and year (1989 dollars)

Questions
What is the mechanism by which the helmet law is thought to reduce the number or rate of crashes? This benefit obviously cannot be attributed to the wearing of helmets, which reduce the severity of crash injuries, not the likelihood of crashing. Does the law simply discourage certain kinds of riders -- risk takers or the inexperienced -- from riding at all?

How were the payers and payments collected determined? Did the hospitals record expected payers and apply standard payment/charge ratios ex ante to estimate payments, or did they record actual payers and collections ex post? If it is based on hospital bills, it would be misleading, since bills indicate the expected payer, which often differs from the ultimate payer.

Strengths
The authors try to perform the right kinds of analysis, despite the fact that their sample sizes are sometimes too small to support strong conclusions.

The authors look at changes in riding patterns that coincide with the helmet law, including the percentage of riders who are licensed.

The authors implicitly acknowledge the shortcoming of hospital charges as a proxy for total medical costs by trying to extrapolate from it to a more comprehensive state-wide medical cost estimate.

The authors deal explicitly with three phenomena that are sometimes cited by anti-helmet activists: basilar skull fractures, cervical fractures, and organ donations.

Weaknesses
Crash Rate Study. It is very difficult to draw conclusions from a comparison of states, even well-chosen ones, because they differ in so many other ways (though their results are rather striking).

Clinical Injury Study. Much of the analysis, particularly the breakdowns by body region and injury type (Table 4), is based on small samples, and the results cannot be considered significant. The clearest evidence of the efficacy of helmets would come from a demonstration of a reduction in the rate of serious head injuries. But the conclusion that “serious head injuries decreased 22 percent,” based on a total of 34 cases over the two years, is particularly sensitive to sample size; a single additional head injury in 1989 would have cut this decrease by half. Given the small population, the secular decline in all motorcycle riding figures, and the decrease in unlicensed driving, a one-year post-law sample consisting of 92 injuries and 38 hospitalizations limits the strength of the conclusions that can be drawn.

The sample was restricted to victims transported to a hospital for treatment by ambulance or helicopter. Those who died at the scene or who were not injured severely enough to require emergency transport were not captured.

Hospital charges were used as a proxy for health care costs. Professional fees and other costs were not accounted for, which would cause costs to be underestimated. On the other hand, charges were not adjusted to reflect the fact that hospitals overcharge. The authors try to make a rough extrapolation from hospital charges to total medical costs, but the ratio they use does not appear to be specific to Nebraska.

Even though the hospital dataset included information on payments, as well as charges, the authors do little with it. They seem to assume that charges are real, and that when payments fall short of charges this represents a social loss. In reality, this is the usual result of hospitals’ practice of overcharging. It would also have been useful to see a the payment/charge ratios for the various payers.

Payers are broken down only by patient, not by charges. They are not broken down by helmet use. And where aggregate payments are reported, self-pay is combined with insurance, and Medicaid with Medicare; these categories should be reported separately.

Fatalities do not appear to have been separated from survivors for the analysis.

Conclusions
This article overcomes data limitations by slicing the data in enough different directions to give an overall picture that suggests the benefits of helmet use. Even though the article suffers from the common problems of 1) a sample selection process that systematically excludes some segments of the injury population and 2) reliance on hospital charges as a cost proxy, it still manages to produce a rough but reasonable estimate of statewide motorcycle injury cost savings resulting from the helmet law.

Murdock, M.A., & Waxman, K. (1991). Helmet use improves outcomes after motorcycle accidents. Western Journal of Medicine 155(4), 370-372.

Abstract
Objective. To determine the effects of motorcycle helmet use on patient outcomes.

Study Population & Data. The authors studied patient outcomes and demographic and epidemiologic variables of 474 patients injured in motorcycle collisions and treated at a Level I trauma center over a 45-month period. Of those involved in a motorcycle collision, 50 percent were not wearing a helmet, 23 percent were wearing a helmet, and in 27 percent helmet use was unknown.

Methods. For analysis, patients were separated into groups according to such variables as injury severity, body region injured, and disability. Counts and percentages of helmeted and nonhelmeted riders in various groups were compared, and differences were tested using Student’s t test or Fisher’s Exact Test.

Results. Those who were wearing a helmet had fewer and less severe head and facial injuries, required fewer days on a ventilator, and sustained no serious neck injuries. Fewer patients who wore helmets were discharged with disability (5 percent vs. 9 percent); this difference was especially sharp with respect to disabling head injuries (i.e., injuries resulting in decreased cognitive functioning) (1 percent vs. 7 percent). Average hospital charges were also lower for helmeted patients ($16,154 vs. $29,905). 53 percent of helmeted patients had third-party insurance coverage, compared with only 40 percent of nonhelmeted patients.

Conclusion. These results support the need for both increased public education regarding helmet use and mandatory helmet use legislation.

Interesting Finding
Over a 45-month period, 474 of 3,941 trauma patients (12 percent) at a Level I trauma center were involved in a motorcycle collision.

Strengths
Fatality and disability cases were examined separately, in Tables 3 and 4, though they also appear to be mixed in with the other cases for the broader analysis.

The authors’ attention to the details of severity and body region -- particularly head and neck injuries -- compensates, to a large degree, for the overall non-representativeness of the sample.

Payers were recorded at the same time as charges -- presumably at discharge.

Weaknesses
The study captured only patients treated at a Level I trauma center. This is not a representative sample of all injured riders, as it misses those who die at the scene and those who avoid significant injury.

The article does not specify the years during which the 45-month study took place, nor does it mention what year’s dollars the monetary results are reported in.

Raw hospital charges are used as a proxy for costs. This is likely to bias the medical cost estima