|
|
| |||||||||||||||||||||||||||||||||||||
Background Hospitals have difficulty justifying the expense of maintaining trauma centers without strong evidence of their effectiveness. To address this gap, we examined differences in mortality between level 1 trauma centers and hospitals without a trauma center (non–trauma centers).
Methods Mortality outcomes were compared among patients treated in 18 hospitals with a level 1 trauma center and 51 hospitals non–trauma centers located in 14 states. Patients 18 to 84 years old with a moderate-to-severe injury were eligible. Complete data were obtained for 1104 patients who died in the hospital and 4087 patients who were discharged alive. We used propensity-score weighting to adjust for observable differences between patients treated at trauma centers and those treated at non–trauma centers.
Results After adjustment for differences in the case mix, the in-hospital mortality rate was significantly lower at trauma centers than at non–trauma centers (7.6 percent vs. 9.5 percent; relative risk, 0.80; 95 percent confidence interval, 0.66 to 0.98), as was the one-year mortality rate (10.4 percent vs. 13.8 percent; relative risk, 0.75; 95 percent confidence interval, 0.60 to 0.95). The effects of treatment at a trauma center varied according to the severity of injury, with evidence to suggest that differences in mortality rates were primarily confined to patients with more severe injuries.
Conclusions Our findings show that the risk of death is significantly lower when care is provided in a trauma center than in a non–trauma center and argue for continued efforts at regionalization.
Methods
Setting
The NSCOT was conducted in 15 regions defined according to contiguous Metropolitan Statistical Areas in 14 states (Table 1). The Metropolitan Statistical Areas were selected from among the 25 largest such areas in 19 states (Arizona, California, Colorado, Florida, Illinois, Indiana, Iowa, Maryland, Massachusetts, Michigan, New Jersey, New York, North Carolina, Oregon, Pennsylvania, South Carolina, Virginia, Washington, and Wisconsin) for which routinely collected hospital-discharge data were available in 1999. We excluded Metropolitan Statistical Areas in which large non–trauma centers collectively treated fewer than 75 patients with major trauma annually, as defined according to an Injury Severity Score of more than 15, on the basis of the diagnostic codes of the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM).12,13
|
Non–trauma centers were, on average, smaller than trauma centers, were less likely to be members of the Council of Teaching Hospitals, and treated fewer patients with major trauma (Table 2). However, 17 such centers had a designated trauma team, and 8 of these had a trauma director. As compared with the universe of level 1 trauma centers and non–trauma centers located in Metropolitan Statistical Areas, the NSCOT sample consisted of larger hospitals that were more likely to be members of the Council of Teaching Hospitals.2 During the study, one of the non–trauma centers was designated a level 1 trauma center and one level 1 trauma center lost its verification. For the analysis, these hospitals were categorized according to their status at enrollment.
|
Patients were eligible for the study if they were 18 to 84 years of age, arrived alive at a participating hospital, and were treated for a moderate-to-severe injury (defined by at least one injury with a score of at least 3 on the Abbreviated Injury Scale) between July 2001 and November 2002.14 Patients who presented with no vital signs and were pronounced dead within 30 minutes after arrival were excluded, as were patients who delayed seeking treatment for more than 24 hours, patients 65 years of age or older with a first listed diagnosis of hip fracture, patients with major burns, patients who spoke neither English nor Spanish, non–U.S. residents, and patients who were incarcerated or homeless at the time of injury. The patients were selected and eligibility was determined in two stages (Figure 1). First, administrative discharge records and emergency department logs were prospectively reviewed to identify patients with a principal ICD-9-CM diagnosis code of 800 to 959 (excluding those due to late effects, foreign bodies, complications, burns, and [among patients 65 years of age or older] hip fractures). We then used a computer program to map ICD-9-CM diagnoses to Abbreviated Injury Scale scores13 to select patients with at least one diagnosis involving a score of at least 3 on the Abbreviated Injury Scale. A total of 18,198 patients met these initial eligibility criteria.
|
In stage 2, we reviewed patients' complete medical records to determine their final eligibility. Medical records were obtained for 1391 (96.7 percent) of the patients who died in the hospital. Of these, 287 were excluded, leaving 1104 eligible patients for whom medical-record data were abstracted. The most common reasons for exclusion in the second stage were death within 30 minutes after arrival and no vital signs (50.8 percent), lack of evidence of trauma (19.6 percent), and treatment sought more than 24 hours after injury (21.5 percent).
Patients discharged alive and selected for the study were contacted at 3 months by mail and then by telephone, and consent was obtained to access their medical records and interview them at 3 and 12 months. Of the 8021 such patients who were selected for the study, 4866 (60.7 percent) were enrolled, 1635 could not be located, 1177 declined to participate, and 343 completed the interview but never provided written permission for a review of their medical records. Of the 4866 who were enrolled, 779 (16.0 percent) were determined to be ineligible on review of their medical records, leaving 4087 eligible live patients for whom complete medical-record data were abstracted. The most common reasons for exclusion in stage 2 were treatment sought more than 24 hours after injury (70.8 percent) and a lack of evidence of trauma (25.4 percent).
For two reasons it was necessary to weight data on the 5191 eligible participants with complete medical-record data (1104 of whom died in the hospital and 4087 of whom were discharged alive) to the population of eligible patients. First, the sampling protocol selected all patients who died in the hospital but only a proportion of patients discharged alive. Second, not all patients selected for inclusion in the study were enrolled. The resulting "sampling" weights consist of the reciprocal product of two probabilities: the conditional probability of being selected and the probability of being enrolled and having data abstracted from the medical record, given that the patient was selected. The reference population to which inferences are made for the NSCOT consists of 15,440 patients who met or were projected to meet the inclusion criteria.
Definition of Outcomes and Data Collection
Outcomes of interest included death in the hospital and death within 30, 90, and 365 days after injury. We identified deaths that occurred after discharge either by interviewing a proxy or through a match with the National Death Index.15 To maximize the ascertainment of patients who died after being discharged, we searched the National Death Index 24 months after the last patient had been enrolled.
Characteristics of the patients and their injuries that were related to the risk of death were obtained from medical records and used in the analysis to adjust for differences between those treated at trauma centers and those treated at non–trauma centers. Nurses, trained specifically for the NSCOT and certified in scoring of the Abbreviated Injury Scale by the Association for the Advancement of Automotive Medicine, abstracted data from the patients' medical records.
Patients were characterized on the basis of their sociodemographic characteristics and preexisting medical conditions. Preexisting conditions were identified from a patient's medical record, and a score for the Charlson comorbidity index was derived.16 The index is based on 17 indicators of coexisting conditions, which are weighted and then totaled to give a single value. A value of 0 indicates that there are no serious coexisting conditions. Since the Charlson comorbidity index does not include either obesity or coagulopathy, both of which correlate with the risk of death after trauma,17,18 these conditions were included in the analysis as individual covariates. The use of alternative models in which the Charlson score was replaced with individual indicators of preexisting conditions yielded similar results.
Injuries were characterized on the basis of their mechanism, anatomical severity, and physiological effect. The anatomical severity of individual injuries was assessed with the use of the Abbreviated Injury Scale.14 Scores derived manually from a review of the medical record were used in all analyses. A total of 381 patients (7.3 percent) who were selected on the basis of having a maximal score of at least 3 were reclassified as having a maximal score of less than 3 after a review of their medical records. These patients were kept in the analysis. Several summary measures of the overall severity of injury were derived from injury-specific Abbreviated Injury Scales, including the Injury Severity Score,12 the New Injury Severity Score,19 the Anatomic Profile Score,20 and the worst survival risk ratio, as defined by Meredith and colleagues.21
We used the first assessment of blood pressure and pupillary response in the emergency department and the first assessments of the motor score of the Glasgow Coma Scale22 in the field and the emergency department to measure the degree of physiological derangement. In categorizing patients according to the motor score of the Glasgow Coma Scale, we separated patients who were pharmacologically paralyzed from those with a score of 1 who were not pharmacologically paralyzed.
Statistical Analysis
Excluded from the present analysis were 137 patients who were transferred to a participating hospital 24 hours or more after injury as well as 11 patients whose length of stay before transfer from a participating center was less than 24 hours. We included 1107 patients who were transferred to a participating hospital from another hospital within 24 hours after injury (880 within 6 hours). When the analysis was repeated excluding these 1107 patients, similar results were obtained.
We used multiple imputation techniques23 to account for missing covariates. Data were missing for fewer than 5 percent of patients except for the categories of prehospital intubation (6.9 percent had data missing), the first score for the Glasgow Coma Scale (13.4 percent), and the score for the Glasgow Coma Scale obtained before hospitalization (30.9 percent). Ten imputed data sets were created. For each data set, robust standard errors were computed to account for clustering within hospitals. Across data sets, estimates and standard errors were computed with the use of Rubin's combining rules.24
All analyses were performed with the use of data weighted to the population of eligible patients. To adjust for observable differences between patients treated at trauma centers and those treated at non–trauma centers, we used the inverse probability of treatment weighting approach described by Robins and colleagues.25 In this approach, data on each patient are further weighted according to the reciprocal of the conditional probability of receiving care at a trauma center given all demographic and injury characteristics listed in Table 3 together with relevant interaction terms. These "adjustment" weights, often referred to as propensity scores, serve to create an "adjusted population," which has two important characteristics: the receipt of care at a trauma center is not confounded by covariates, and the effect of care at a trauma center is the same in the adjusted population as it is in the original reference population. This method hinges on the correct specification of a model for the propensity score. To check the adequacy of this model, we evaluated the balance on covariates in the adjusted population.26 We also trimmed the adjustment weights to reduce the effect of influential observations on the overall results. The degree of trimming was chosen to minimize mean squared error.27
|
As compared with patients treated in trauma centers, those treated in non–trauma centers were older; had more coexisting conditions; were more likely to be female, non-Hispanic white, and insured; and tended to have less severe injuries (Table 3). After further weighting according to propensity scores, the two groups of patients were similar (Table 3).
The observed (unadjusted) case fatality rate in the hospital was higher among patients treated at trauma centers than among patients treated at non–trauma centers (8.0 percent vs. 5.9 percent). An additional 3.1 percent of patients died after discharge, with a smaller percentage dying after discharge from a trauma center than after discharge from a non–trauma center (1.9 percent vs. 6.3 percent).
After adjustment for differences in the case mix, the risk of death within one year after injury was significantly lower when care was provided in a trauma center than when care was provided in a non–trauma center (10.4 percent vs. 13.8 percent; relative risk, 0.75; 95 percent confidence interval, 0.60 to 0.95) (Table 4). The relative reduction in risk was similar for in-hospital, 30-day, and 90-day mortality (Table 4). We assessed whether the relative risk of death in a trauma center as compared with a non–trauma center varied according to the overall severity of injury. We observed a significant interaction between the score for the Abbreviated Injury Scale and treatment at a trauma center with regard to in-hospital mortality (two-sided P=0.02 by a global test for two-way interactions between the type of hospital and maximal scores), 30-day mortality (P=0.03), and 90-day mortality (P=0.02) but not 365-day mortality (P=0.61). As shown in Table 4, the relative risks of death among patients with a maximal score for the Abbreviated Injury Scale of 4 or a maximal score of 5 or 6 were lower than the risks among those with a maximal score of only 3. On the other hand, there were minimal differences in risk between patients with a maximal score of 4 and those with a maximal score of 5 or 6.
|
Discussion
Previous studies of the effectiveness of trauma centers have been inconclusive and hampered by limitations in study design and reliance on in-hospital mortality as a measure. Most problematic has been the difficulty in adequately adjusting for referral bias — that is, the reality that trauma centers treat a higher proportion of young, severely injured patients, whereas non–trauma centers treat a higher proportion of elderly patients with coexisting conditions. We addressed this issue by stratifying the patients according to the type and severity of injury and age, collecting detailed information on important covariates known to influence the risk of death, and by using propensity-score weighting to adjust for potential biases in the analysis.
After adjustment for differences in the case mix, the overall ris k of death was 25 percent lower when care was provided at a trauma center than when it was provided at a non–trauma center. Relative differences in risk, however, varied according to the severity of injury, with evidence to suggest that differences in the risk of death according to the type of hospital were primarily among patients with Abbreviated Injury Scale scores of 4 or higher. Although there is insufficient evidence to establish a hospital-based effect among patients with scores of less than 4, the risk of death in this group of patients, especially among the young, is low. It is possible, however, that treatment at a trauma center could benefit these patients by reducing complications and overall treatment costs or improving functional outcomes and increasing the likelihood that they will return to productivity.
Differences in the risk of death according to the type of hospital also appeared to be greater among younger patients than older patients. Although the risk of death was lower among older patients treated at trauma centers than among those treated at non–trauma centers, the differences were not as large as those between younger patients and the relative risks of death were not significantly different from 1.0. An important limitation of our study, however, was the small number of older patients with severe injuries, resulting in wide confidence intervals for this cohort. This limitation may have contributed to our inability to detect a significant interaction between the type of hospital and age. Elderly patients with trauma represent a serious challenge, because they are at high risk for complications and death from injuries that would not necessarily prove fatal to their younger counterparts.28,29,30 Paying more aggressive attention to coexisting medical conditions during the acute and post-acute phases may improve the outcome among such patients and is worthy of further study.10,31,32,33,34
Our estimates may be conservative for two reasons. First, we included only non–trauma centers that treated at least 25 patients with major trauma per year. Most non–trauma centers are small and may have a lower quality of trauma care than larger facilities. More important, 17 of the non–trauma centers in our study had a designated trauma team, and 8 of the 17 had a trauma director. Including these hospitals as non–trauma centers may have biased the results toward a more conservative estimate of the effect.
Caution is needed in generalizing our results. Because the NSCOT is a study of the effectiveness of trauma centers in urban and suburban America, our results cannot readily be extrapolated to rural areas of the country. In addition, we did not address the relative effectiveness of intermediate levels (2, 3, or 4) of trauma care. Finally, we excluded children and adolescents; the effect of care in a trauma center in this population must be addressed in a separate study.
Our results show that the overall risk of death is significantly lower when care is provided in a trauma center than when it is provided in a non–trauma center, and they argue for continued efforts at regionalization. At the same time, they highlight the difficulty in decreasing the risk of death among elderly patients with trauma.
Funded by a grant (R49/CCR316840) from the National Center for Injury Prevention and Control of the Centers for Disease Control and Prevention and a grant (R01/AG20361) from the National Institute on Aging of the National Institutes of Health.
No potential conflict of interest relevant to this article was reported.
We are indebted to the members of the NSCOT National Advisory Committee who provided invaluable assistance in the design of the study and in the interpretation of the results, including: A. Brent Eastman, M.D. (chair), John W. Ashworth, III, M.H.A., Robert R. Bass, M.D., Gloria J. Bazzoli, Ph.D., Michael J. Bosse, M.D., Nathan Cope, M.D., Maurine Goehring, R.N. M.S.N., David B. Hoyt, M.D., Frank R. Lewis, Jr., M.D., James P. LoGerfo, M.D., M.P.H., Ronald F. Maio, D.O. M.S., Donald W. Marion, M.D., M.Sc., Colleen A. McHorney, Ph.D., J. Wayne Meredith, M.D., Jeffrey Michael, Ed.D., John A. Morris, Jr., M.D., Richard J. Mullins, M.D., Louis A. Quatrano, Ph.D., John C. Sacra, M.D., Donald M. Steinwachs, Ph.D., Marc F. Swiontkowski, M.D., Roger S. Taylor, M.D., M.P.A., Harry Teter, J.D., and John A. Weigelt, M.D.; to Anthony R. Carlini, M.S., Lele Tang, M.S. and the NSCOT nurse coordinators — Linda Agnello, R.N. Marcia Baldwin, R.N., J.D.; Sharon Blassingame, R.N. Linda Carrier, R.N., Carla Kimberlin, R.N. Elaine Kooima, R.N., Leah LeClerc, R.N. M.S. Cynthia Lemmon, R.N., Dana McDermott, R.N., M. Christine Michaelis, R.N., Yeni Quintana, R.N., Allana Richmond, M.S., R.N.C., Carleen Sparks, R.N., Eleanor Walsh, R.N., and Karen Yuhas, R.N., M.P.H. for their commitment to the study; to Ciprian M. Crainiceanu, Ph.D., and Zhiqiang Tan, Ph.D., for their assistance in refining our approach to the statistical analysis of the data; and, for their participation in the study, to the following hospitals: Beverly Hospital, Beverly, Mass.; Boston Medical Center, Boston; Brockton Hospital, Brockton, Mass.; Cape Fear Valley Health System, Fayetteville, N.C.; Caritas Good Samaritan Medical Center, Brockton, Mass.; Carolinas Medical Center, Charlotte, N.C.; Citrus Valley Medical Center, Covina, Calif.; Cook County Hospital, Chicago; Deaconess Hospital, Evansville, Ind.; Doctors Medical Center, Modesto, Calif.; Forsyth Medical Center, Winston-Salem, N.C.; Frederick Memorial Hospital, Frederick, Md.; Froedtert Memorial Lutheran Hospital, Milwaukee; Garden City Hospital, Garden City, Mich.; Gaston Memorial Hospital, Gastonia, N.C.; Greater Baltimore Medical Center, Baltimore; Harborview Medical Center, Seattle; Henry Ford Hospital, Detroit; Hospital of the University of Pennsylvania, Philadelphia; Jackson Memorial Hospital, Miami; Jacobi Medical Center, Bronx N.Y.; Johns Hopkins Hospital, Baltimore; Kaiser Foundation Hospital, Woodland Hills, Calif.; Kaiser Foundation Hospital, Los Angeles; Kaiser Foundation Hospital, San Diego, Calif.; Kendall Medical Center, Miami; Los Angeles County–University of Southern California Medical Center, Los Angeles.; Lawrence Hospital, Bronxville, N.Y.; Lehigh Valley Hospital, Allentown, Pa.; Little Company of Mary Hospital, Evergreen Park, Ill.; Long Island College Hospital, Brooklyn, N.Y.; Loyola University Medical Center, Maywood, Ill.; Maimonides Medical Center, Brooklyn, N.Y.; Mary Washington Hospital, Fredericksburg, Va.; Memorial Medical Center, Modesto, Calif.; Methodist Hospital of Southern California, A rcadia, Ca.; Montefiore Medical Center, Bronx, N.Y.; J. Mount Clemens General Hospital, Mount Clemens, Mich.; North Carolina Baptist Hospital, Winston Salem, N.C.; NorthEast Medical Center, Concord, N.C.; Oakwood Hospital & Medical Center, Dearborn, Mich.; PinnacleHealth Harrisburg Hospital, Harrisburg, Pa.; Presbyterian Intercommunity Hospital, Whittier, Calif.; Providence Hospital & Medical Centers, Southfield, Mich.; Saint Mary's Medical Center, Racine, Wisc.; Saint Mary's Medical Center, Saginaw, Mich; San Francisco General Hospital Medical Center, San Francisco; San Joaquin General Hospital, French Camp, Calif.; Shady Grove Adventist Hospital, Rockville, Md.; Sharp Grossmont Hospital, La Mesa, Calif.; Sinai Grace Hospital, Detroit; South Jersey Hospital-Newcomb, Vineland, N.J.; St. Catherine Hospital, East Chicago, Ind.; St. Joseph Medical Center, Towson, Md.; St. Luke's Hospital of New Bedford, New Bedford, Mass.; St. Mary Mercy Hospital, Livonia, Mich.; St. Mary's Health Services, Evansville, Ind.; St. Mary's Hospital Medical Center, Madison, Wisc.; St. Luke's Medical Center, Milwaukee; Swedish Health Services, Seattle; Swedish Medical Center, Seattle; Tri-City Medical Center, Oceanside, Calif.; University of Maryland Medical Center, Baltimore; University of California San Diego Medical Center, San Diego; University of Pittsburgh Medical Center Presbyterian Hospital, Pittsburgh; Virginia Mason Medical Center, Seattle; Waukesha Memorial Hospital, Waukesha, Wisc.; White Memorial Medical Center, Los Angeles; William Beaumont Hospital-Troy, Troy, Mich.
Source Information
From the Johns Hopkins Bloomberg School of Public Health, Center for Injury Research and Policy, Baltimore (E.J.M., K.P.F., B.L.E., D.S.S., D.O.S.); and the University of Washington School of Medicine, Harborview Injury Prevention and Research Center, Seattle (F.P.R., G.J.J., A.B.N.).
Address reprint requests to Dr. MacKenzie at Johns Hopkins Bloomberg School of Public Health, 624 N. Broadway, Rm. 554, Baltimore, MD 21205-1996, or at emackenz{at}jhsph.edu.
References
| |||||||||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Previous
