The premise underlying the development of state trauma systems was that an organized system of trauma care would ensure that critically injured patients are appropriately triaged, transferred to high-quality definitive care, or both, without delay. Despite its conceptual simplicity, implementation of the necessary infrastructure for system development is exceedingly complex. First, state trauma systems must be supported by appropriate state legislation to ensure that administering organizations (usually the state emergency medical service office) have the legal authority to develop and enforce trauma system policy. 5 Administering organizations frequently encounter opposition to proposed policies and without a legal framework have no recourse for requiring compliance. Second, there must be a process of trauma center designation such that critically ill patients are cared for by those most experienced in all phases of trauma care. 4 Finally, it has been suggested that prehospital medical direction, triage protocols, interfacility transfer agreements and mechanisms for quality assurance be implemented as part of trauma system policy. 5,22–24 In view of the infrastructure required, the costs of developing and maintaining a comprehensive statewide trauma system are significant. The combination of complexity, cost, and questionable effectiveness has limited trauma system implementation to only a small number of states. We set out to evaluate the effectiveness of trauma systems in reducing injury-related mortality in this context.
By using a population-based cross-sectional analytic approach comparing injury mortality rates in states with trauma systems to those without, we have demonstrated that the presence of a trauma system is associated with a 9% reduction in global injury mortality. The greatest effect was evident when deaths caused by unintentional injuries, specifically motor vehicle crashes, were examined separately. Furthermore, this reduction in mortality remains even when adjustment is made for a number of covariates known to affect MVC-related mortality. Finally, when evaluated by age stratum, trauma systems affected their greatest reduction on injury mortality in the young and elderly, patients typically at highest risk for adverse outcome.
There have been several other evaluations of regional trauma systems; however, these analyses are typically limited to either a single state or several counties. For example, in a report of MVC-related trauma mortality in counties in the state of Florida, Alexander reported a 24-fold reduction in mortality in counties with Level I trauma centers. 12 These data are difficult to interpret because there was no attempt to control for speed limits or population density, both of which may impact on outcome. To correct for these potential confounders, Rutledge et al. evaluated each county in North Carolina by using multivariate analytical techniques and demonstrated a 20% reduction in per capita trauma deaths in counties with trauma centers compared with counties without. 16 Kane et al. used a slightly different approach in the evaluation of the Los Angeles County trauma system. 15 In this study, MVC-related mortality was assessed before and after the designation of regional trauma centers. In contrast with the above data, regionalization of trauma centers did not improve mortality, although post hoc subset analysis suggested a mortality benefit in the most severely injured patients. Finally, Mullins et al. compared Washington and Oregon, states similar in population and geography, both before and after the implementation of the Oregon regional trauma system. 25 By using the state of Washington as a control for changes over time, there remained a 20% reduction in the odds of death after trauma system implementation in Oregon.
The present study offers several advantages to those described above. First, all currently available studies are limited in scope because they tend to evaluate smaller, defined jurisdictions, whether a city, county, or state. The effectiveness of a trauma system compared with an ad hoc system of care may be dependent on geography or referral patterns in any particular region. By analyzing the nation as a whole, these data provide a reasonable estimate of the national impact of trauma system implementation. Furthermore, the geographic distribution of the trauma system states and large sample size precludes their being any specific state or regional distinctions that may confound these data. Finally, almost all previous interregional comparisons have failed to consider other parameters that may profoundly effect MVC-related mortality, including speed limits, 26,27 restraint laws, 28 and population distribution. 21,29 The presence of certain legislative initiatives such as lower speed limits and primary enforcement of restraint laws tend to parallel trauma system legislation. Additionally, regional trauma systems tend to be organized in more urban areas where crash mortality rates are significantly lower than in less-populous regions. Taken together, these data suggest that failing to control for these factors may overestimate any mortality benefit contributed by an organized system of trauma care.
This study design carries with it three inherent limitations. First, the use of mortality data tends to underestimate the effectiveness of trauma care. For example, injuries with a high on-scene case fatality rate will not benefit from the presence of a trauma system. The data sources used in this study do not permit separate analysis or exclusion of deaths attributable to unsalvageable injuries or deaths at the scene. This effect is most apparent when evaluating the effect of trauma systems on mortality caused by intentional injuries. The increased mortality rate associated with trauma systems in patients with intentional, or penetrating injuries, may simply be a reflection of the higher homicide rates in the more urban trauma system states and probably has little to do with quality of care. 30 Because of the lower on-scene case fatality rate, MVC-related deaths provide a more reliable estimate of efficacy, as these deaths are most amenable to system intervention.
Another possible limitation is that we may not have considered all potential differences between trauma system states and nonsystem states that may impact on trauma mortality. For example, we had no way of assessing the extent of use of either advanced or basic life support used by emergency medical service in the two groups of states. Although somewhat controversial, the availability of advanced life support may impact on mortality. 31 Unfortunately, it is not possible to assess advanced life support use from state to state, because its use within a state is not uniform, tending to vary from county to county within a state.
Finally, the use of a cross-sectional study design assessing the effects of a trauma system at one point in time provides no insight into how trauma systems evolve or how trauma system effectiveness changes over time. For example, this analysis makes the assumption that less well-established trauma systems are as effective as systems that have evolved over several years. This assumption may not be valid, and if anything, will tend to underestimate the effectiveness of trauma systems as relatively new systems are classed with those that have matured over time.
Despite these limitations, these data clearly provide evidence that the presence of a state trauma system is associated with a reduction in the risk of death attributable to MVC of approximately 9%. Currently, only 61% of the United States population live in a state with a trauma system. Although the funds and personnel necessary to both develop and maintain the infrastructure of a regional trauma program are substantial, they are relatively insignificant when considering the potential productive life years gained of an additional 3,160 survivors of the 35,100 Americans in nontrauma system states dying of their injuries annually, equivalent to productivity gains of over 300 million dollars. 32
1. National Safety Council. Accident Facts: 1997. Itasca, Ill: National Safety Council; 1997.
2. National Committee of Trauma, and Committee on Shock. Accidental Death and Disability: The Neglected Disease of Modern Society. Washington, DC: National Academy of Sciences/National Research Council. 1966.
3. American College of Surgeons. Optimal hospital resources for care of the seriously injured patient. Bull Am Coll Surg. 1976; 61:15–22.
4. West JG, Williams MJ, Trunkey DD, Wolferth CC. Trauma systems: current status-future challenges. JAMA. 1988; 259:3597–3600.
5. Bazzoli GJ, Madura KJ, Cooper GF, MacKenzie EJ, Maier RV. Progress in the development of trauma systems in the United States: results of a national survey. JAMA. 1995; 273:395–401.
6. West JG, Cales RH, Gazzaniga AB. Impact of regionalization: the Orange County experience. Arch Surg. 1983; 118:740–744.
7. West JG, Trunkey DD, Lim RC. Systems of trauma care: a study of two counties. Arch Surg. 1979; 114:455–460.
8. Neuman TS, Bockman MA, Moody P, et al. An autopsy study of traumatic deaths. Am J Surg. 1979; 144:722–727.
9. Guss DA, Meyer FT, Neuman TS, Baxt WG, Dunford JV, Griffith LD. The impact of a regionalized trauma system on trauma care in San Diego County. Ann Emerg Med. 1989; 18:1141–1145.
10. Cales RH. Trauma mortality in Orange County: the effect of implementation of a regional trauma system. Ann Emerg Med. 1984; 13:1–10.
11. Shackford SR, Hollingsworth-Fridlund P, Mcardle M, Eastman AB. Assuring quality in a trauma system: the medical audit committee: composition, cost and results. J Trauma. 1987; 27:866–875.
12. Alexander RH, Pons PT, Krischer J, Hunt P. The effect of advanced life support and sophisticated hospital systems on motor vehicle mortality. J Trauma. 1984; 24:486–490.
13. Mullins RJ, Veum-Stone J, Helfand M, et al. Outcome of hospitalized injured patients after institution of a trauma system in an urban area. JAMA. 1994; 271:1919–1924.
14. Mullner R, Goldberg J. An evaluation of the Illinois trauma system. Med Care. 1978; 16:140–151.
15. Kane G, Wheeler NC, Cook S, et al. Impact of the Los Angeles County trauma system on the survival of seriously injured patients. J Trauma. 1992; 32:576–583.
16. Rutledge R, Fakhry SM, Meyer A, Sheldon GF, Baker CC. An analysis of the association of trauma centers with per capita hospitalizations and death rates from injury. Ann Surg. 1993; 218:512–524.
17. Bazzoli GJ, Madura KJ. Inventory of Trauma Systems. Chicago: Hospital Research and Educational Trust. 1993.
18. Centers for Disease Control, and Prevention. CDC WONDER Website. Available at http://wondercdcgov. Accessed September, 1998.
19. National Highway Traffic Safety Administration. Fatality Analysis Reporting System Website. Available at: http://www.fars.nhtsa.dot.gov/fars/fars.cfm. Accessed September. 1998.
20. National Safety Council. Accident Facts: 1995. Itasca, Ill: National Safety Council; 1995.
21. Muelleman RL, Walker RA, Edney JA. Motor vehicle deaths: a rural epidemic. J Trauma. 1993; 35:717–719.
22. US Department of Health and Human Services. Setting the national agenda for injury control in the 1990s. Proceedings from the Third National Injury Control Conference, Atlanta, GA: Centers for Disease Control. 1992.
23. National Highway Traffic Safety Administration. Development of Trauma Systems (DOTS): A State and Community Guide. Washington, DC: US Department of Transportation; 1989.
24. US Department of Health and Human Services. Model Trauma Care System Plan. Washington, DC: Bureau of Health Resources Development. 1992.
25. Mullins RJ, Mann NC, Hedes JR, Worrall W, Jurkovich GJ. Preferential benefit of implementation of a statewide trauma system in one of two adjacent states. J Trauma. 1998; 44:609–617.
26. Garber S, Grahman JD. The effects of the new 65 mph speed limit on rural highway fatalities: a state-by-state analysis. Accid Anal Prev. 1990; 22:137–149.
27. Wagenaar AC, Streff FM, Schultz R. Effects of the 65 mph speed limit on injury morbidity and mortality. Accid Anal Prev. 1990; 22:571–585.
28. Rivara FP, Thompson DC, Cummings P. Effectiveness of primary and secondary enforced seat belt laws. Am J Prev Med. 1999; 16:30–39.
29. Baker SPH, Whitfield RA, O’Neill B. Geographic variations in mortality from motor vehicle crashes. N Engl J Med. 1987; 316:1384–1387.
30. Fingerhut LA, Ingram DD, Feldman JJ. Homicide rates among US teenagers and young adults: differences by mechanism, level of urbanization, race, and sex, 1987 through 1995. JAMA. 1998; 280:423–427.
31. Spaite DW, Criss EA, Valenzuela TD, Meislin HW. Prehospital advanced life support for major trauma: critical need for clinical trials. Ann Emerg Med. 1998; 32:480–489.
32. Rice DP, MacKenzie EJ. Cost of Injury in the United States: A Report to Congress. Baltimore, Md: Injury Prevention Center, School of Hygiene and Public Health, The Johns Hopkins University; 1989.
Dr. William Schwab (Philadelphia, Pennsylvania): Allow me to thank the Program Committee for the opportunity of discussing this overall excellent paper. The authors are especially commended for their survey and statutory corroboration of trauma systems for the year 1995. This study is the type used by public policy analysts to make decisions regarding the crafting of legislation. It is important! As our Firearm Injury Control Center at the University of Pennsylvania has been active in this type of public policy analysis concerning the effect of certain firearms on death rates in America, I feel comfortable trying to discuss this paper’s methodology. Unfortunately, as a trauma surgeon interested in advancing trauma systems, I emotionally support every possible conclusion, so I offer that to you as a true “bias” of this discussant. In addition, I must thank Dr. Charles Branas, a health systems analyst and senior scholar from the Department of Clinical Epidemiology and Biostatistics at the Wharton School for his help at better understanding the application of the mathematical modeling used by the authors. With that as your reference, let me offer several constructive criticisms and questions.
This study is a 1-year national cross-sectional analysis of states with trauma systems versus those without. Because it is at the state, there are a few records (n = 50). By design, it would definitely benefit from being extended to a multiyear, time-series analysis. Previous studies that examined individual states before and after trauma systems better control for factors specific to a particular state in a more robust way than the current single year cross-sectional interstate comparison. This is a limitation that needs to be better explored and, it seems to me, could be remedied by using NCHS and FARS data (available as far back as 1975), for several years in serial as opposed to just 1 year. Each individual state could be better described before and after their specific trauma system implementation dates using longitudinal data analysis techniques for a stronger conclusion. For the authors, (1) did you consider such an analysis, and if so, why did you elect the cross-sectional method for a single year? (2) Was 1995 an outlier year and did you do any year-to-year comparisons to assure ’95 trauma systems performed in similar manners in other years. Please explain.
The Poisson regression model is generally competent. The authors should, however, better explain the specifics of their multivariate modeling. In short, I would have liked to have seen the inclusion and testing of more independent variables that could have modified the effect of trauma systems on death rates. (3) Why is age included in a separate model? Why is the full model with vehicle miles driven not reported alongside the reduced model? (4) Why were variables that were initially discussed, such as median household income, driving habits, and access to prompt trauma care never included in regression models?
Possibly, this last covariable, access to the trauma system, is the most important variable not explored. Triage of major trauma into trauma systems and centers seems to me to be a key variable to this cross-sectional analysis. You yourself said time delays may effect outcome. Perhaps you attempted to analyze the effectiveness of triage in each state studied through the questionnaire. If so, please tell us. Your conclusions might be better supported if you modeled the strength of triage into the multivariate regression or handled this separately. This more than the use of advanced life support, which you discussed, is likely to greatly affect whether a system can reduce mortality, since trauma care and outcome is so time dependent. From our own work, a comparison of the states of Maryland and Pennsylvania prehospital triage performance would have shown a marked difference in prehospital personnel’s compliance of trauma triage and outcome in the two neighboring trauma system states. In addition, considerable insight to how established “systems” differ vastly when this one key covariable is studied would be apparent.
The case and conclusions the authors make for motor vehicle crash analysis is strong to stand on its own merit. Including the “intentional” injury analysis seems to weaken or distract from the message. To equate this poorer correlation with the effect of urban homicides seems superficial, and if you choose to include it, I would recommend either to add further modeling with a larger and stronger Poisson regression or make it the subject of your next paper. Please explain your rationale behind your handling of homicides.
Finally, let me conclude by saying that the figure of 3,160 survivors based on a 9% reduction of mortality with a state trauma system is so weighty that the authors must be absolutely sure of this number. It has such potential impact on policy that it must be as correct as possible and each regression coefficient as accurate as possible. But the analysis and conclusions are very broad and beg for better explanations in the text. If substantiated by the econometricians and other analysts, this number, 9% reduction in lives lost with trauma system, will soon be the “oft-quoted punch line” of this manuscript.
I would like to thank Dr. Nathens and his coauthors for the early receipt of the manuscript and the privilege of discussing their paper.
Dr. Richard J. Mullins (Portland, Oregon): Integral to the public policy approach to trauma systems for the past 40 years has been the concept that Injury in America is a neglected disease, akin to an infectious disease, which should be approached as a public health problem. Thus there is a combined emphasis on injury prevention programs and improved acute care. Do you conclude that in states with sufficient public awareness to establish a trauma system there is also more active public demand for effective injury prevention programs? In other words, is some of the reduction in trauma-related death in these states with a trauma system caused by a concurrent program of injury prevention?
There have been over the past 10 years attempts by the federal government, through allocation of money linked to compliance with federal requirements regarding trauma systems, to impose on individual states a generic model of a trauma system. On the other hand, many trauma systems have been developed from the grass roots efforts of community leaders and coalitions, and as such these trauma systems have been individualized to the communities needs. Do you feel the 20 successful trauma systems you evaluated are homogenous and similar in organization, or more heterogenous? Do you feel that trauma system development should follow the model of the federal government imposing from the top down, or should there be an emphasis on development in individual states from the bottom up?
Dr. Avery B. Nathens (closing): I’d like to thank Dr. Schwab for his insightful comments. First, there are limitations to a cross-sectional study. We can’t possibly control for every confounder as we don’t even know what all the potential confounders are. What we used as confounders or covariates in the Poisson regression model were those parameters that we know clearly have an impact on crash mortality.
We can look at other factors like median household income, which is mentioned in the study, but a lot of those factors are actually related to the initiation of a trauma system; that is, states that are wealthy will implement a system. So they’re not really covariates or confounders in that respect.
We are left having to decide what’s reasonable and what’s appropriate to put in the model. We left variables in the model only if they were associated with a significant change in the point estimate of the incident rate ratio. For example, if the point estimate changed by greater than 10% on addition of a variable to the model, the parameter of interest was left in the model. This is referred to as a “change in estimates” approach to the use of confounders.
In terms of longitudinal time-series analysis, this graph demonstrates what happens over time in a trauma system state. You don’t really see much of a change in crash mortality until about 6 to 8 years after a trauma system is implemented. By about 13 to 15 years later, there’s actually a significant reduction in crash mortality by, again, 9%, or 9 to 10%.
I won’t go into the details of this approach. It’s very complex in terms of mathematical modeling. But it does suggest that it does take time for a trauma system to evolve in a state, and that we can’t expect the benefit to become apparent.
As far as your comment about looking at triage and times of definitive care, it’s almost impossible to look at that in an ecological study of this nature. It’d be nice to get individual records regarding time to definitive care, but it’s impossible, given the nature of this study.
Dr. Mullins’ questions, first with regard to Injury in America and the importance of prevention, it’s clear that the best prevention policies are those that are implemented as a part of a system, and every trauma system should, in fact, have a prevention component to it.
So, in fact, aside from trauma systems providing better care, they should have another goal, the reduction of global rates of injury. Systems that don’t have prevention policies and don’t make the public aware of the importance of prevention are not fulfilling their full potential.
Regarding the variation between states and whether the systems should be built from the top down or bottom up, I think that every state is different. The geography is different, the population distribution is different, and the politics are different. I think that each state has a unique set of problems and rules coming from the top probably aren’t going to be useful for every state, so things have to be flexible and most appropriate for the state in question.
I thank the Association for the privilege of the podium.