Setting
The NSCOT was conducted in fifteen regions defined by one or more contiguous Metropolitan Statistical Areas located in fourteen states (California, Florida, Illinois, Indiana, Maryland, Massachusetts, Michigan, New Jersey, New York, North Carolina, Oregon, Pennsylvania, Washington, and Wisconsin) for which computerized hospital discharge data were available1. Across these regions, eighteen level-I trauma centers and fifty-one large non-trauma-center hospitals were invited to participate in the study and received approval from their institutional review boards to do so. Hospitals were identified as level-I trauma centers if they were designated as such by a state or regional authority or the designation was verified by the American College of Surgeons Committee on Trauma. Non-trauma-center hospitals were neither designated nor verified by the American College of Surgeons as a trauma center at any level but treated at least twenty-five patients with a major traumatic injury (defined by an International Classification of Diseases [ICD]-derived Injury Severity Score [ISS] of >15 points annually)2,3. Restricting the sample to larger non-trauma centers was necessary for efficiency of the study design. Non-trauma-center hospitals were, on the average, smaller than trauma centers and treated fewer patients with major trauma per year1. It is also important to note that, of the fifty-one non-trauma-center hospitals, seventeen had a designated trauma team and eight of these had a trauma director. Institutional review board approval was obtained from all sixty-nine hospitals.
Patient Population and Sample Selection
Patients were eligible for inclusion in the parent NSCOT study if they were between the ages of eighteen and eighty-four years and were treated at one of the participating hospitals for a moderately severe to severe injury (defined by an Abbreviated Injury Scale [AIS] score of =3 points)4. Patients were excluded if they presented with no vital signs and were pronounced dead within thirty minutes after arrival at the hospital, did not seek treatment within twenty-four hours after the injury, were sixty-five years of age or older and had a first-listed diagnosis of a hip fracture, had burns, were non-English or non-Spanish speaking, were not a resident of the United States, or were incarcerated or homeless at the time of injury. Patients recruited into the study were hospitalized during an eighteen-month period from July 2001 to November 2002.
The procedures used for sampling patients and determining eligibility for the NSCOT study have been previously described1. We initially selected all 1438 hospital deaths and a sample of 8021 patients who had been discharged alive, stratified within hospitals according to age, ICD-derived ISS, and principal body region injured (Fig. 1). To sample the patients who were discharged alive, a quota sampling strategy was used with the goal of enrolling sufficient numbers of patients within the strata to detect a difference of 7 to 10 points between the SF-36 subscores of the patients treated at a trauma center (approximately 250 patients within each stratum) and the SF-36 subscores of those treated at a non-trauma center (approximately 250). A 7 to 10-point difference was chosen as it is the minimal difference that is likely to be meaningful to patients or clinicians5,6.
We obtained the medical records on 1391 of the 1438 hospital deaths screened for inclusion, and 1104 of these met the eligibility criteria for the overall study (Fig. 1). Patients discharged alive and selected for the study were contacted by telephone at three months to obtain their consent for us to access their medical record and to interview them by telephone at three and twelve months. Of the 8021 patients who had been discharged alive and were initially screened for the study, 4866 (61%) were enrolled; 1635 could not be located, 1177 refused to participate, and 343 did not provide permission for us to access their medical record. Of those enrolled, 4087 were determined to be eligible for inclusion in the overall NSCOT study.
In the present analysis, we included only those patients with at least one injury to a lower limb (including the pelvis and acetabulum) with an AIS score of =3 points. These injuries generally included all types of femoral fractures; open, displaced, and comminuted fractures of the tibia; open, displaced, and comminuted fractures of the pelvis or acetabulum; all types of amputations (excluding toes); massive crush injuries and degloving injuries; penetrating injuries with substantial blood loss; and injuries to major blood vessels and nerves. We excluded forty-eight patients who had been transferred to an NSCOT hospital more than twenty-four hours after the injury as well as six who had stayed in an NSCOT hospital for less than twenty-four hours before being transferred. After excluding those patients, the final study group included 1389 patients (205 who died in the emergency department or hospital and 1184 individuals who survived to hospital discharge) (Fig. 1); 823 were treated in a trauma center, and 566 were treated in a non-trauma center. Of the 1184 patients who had been discharged alive, sixty-two died within one year after the injury (as identified by proxy or through a match with the National Death Index7). Of those surviving for one year, 925 (82.4%) consented to a follow-up interview. This final number (524 treated in a trauma center and 401 treated in a non-trauma center) meets our sample-size expectations and is sufficient for detecting meaningful differences in SF-36 scores.
Definition of Outcomes and Data Collection
The outcomes of principal interest were derived from the Medical Outcomes Study Short Form Health Survey (SF-36)8. The SF-36 consists of multi-item scales that measure eight health domains: physical functioning (PF); role limitations due to physical health (RP); bodily pain (BP); social functioning (SF); vitality, energy, or fatigue (VT); general health perceptions (GH); role limitations due to emotional problems (RE); and general mental health (MH). Scores ranging from 0 to 100 points are derived for each domain, with lower scores indicating poorer function. Because there is evidence that the SF-36 is not sensitive to the effect of trauma on mental health consequences9, we supplemented the SF-36 with the Revised Center for Epidemiologic Studies Depression Scale (CESD-R), a well-validated instrument developed to detect depressive symptoms10. In accordance with published guidelines, we classified respondents with scores of =16 points as having clinically relevant symptoms of depression.
Given our hypothesis regarding the specific effect of trauma-center care on patients with lower-limb injuries, we supplemented the SF-36 with the mobility subscale of the Musculoskeletal Function Assessment (MFA)11. The MFA mobility subscale consists of twenty items and is part of a larger instrument that taps into ten domains of everyday functioning. To be consistent with how the SF-36 score is calculated, items in the MFA mobility subscale were combined to calculate a score of 0 to 100 points, with lower scores indicating poorer function. Since the MFA was specifically designed to assess the consequences of musculoskeletal disorders, it has fewer ceiling and floor effects, higher content validity, and greater responsiveness than the SF-36 for patients with musculoskeletal trauma12.
Correlates of Outcome
Preinjury characteristics of the patient and details of the injury were obtained from the medical record and the three-month interview and were used to adjust for differences between patients treated at trauma centers and those treated at non-trauma centers.
Injuries were characterized by their mechanism, anatomic severity, and physiologic impact. The mechanism was initially categorized, according to ICD E-Codes, as a motor-vehicle accident (including pedestrian-motor vehicle accidents), low fall, high fall, other blunt injury (e.g., an injury resulting from use of machinery or being struck by or against a blunt object), or penetrating injury—firearm or other (e.g., an injury resulting from a stab or cut from a sharp or piercing instrument). The anatomic severity of individual injuries was assessed with use of the 1998 version of the AIS4. The New Injury Severity Score (NISS)13 was used to summarize overall severity across body regions. Physiological derangement was measured on the basis of the first assessment of blood pressure and pupillary response in the emergency department and the first assessment of the motor score of the Glasgow Coma Scale (GCS) in the emergency department14. Because the 1998 AIS does not adequately reflect the fracture grade or severity, we further classified these injuries according to the presence of an open fracture15, the Orthopaedic Trauma Association (OTA) fracture type (A, B, or C)16, the presence of bilateral injuries, and whether the injury resulted in an amputation. Head injuries were further classified according to the presence or absence of a midline shift of the brain.
Age, gender, race/ethnicity, insurance status, and preexisting medical comorbidities were abstracted from the medical record. According to recommendations by Groll et al.17, a score was derived on the basis of the number of comorbidities (out of twenty-one). Education, poverty status, preinjury self-reported health status (rated as excellent, very good, good, fair, or poor), and preinjury difficulty performing one or more activities of daily living or instrumental activities of daily living were determined at the three-month interview.
Statistical Analysis
In an investigation of the effect of trauma-center care on functional outcomes, it is important to take into account the competing risk of death. Examining the effect on functional outcome only among those who are known to have survived could lead to biased results. Such bias could occur if, for example, trauma centers saved the lives of patients with poor functional outcomes who would have died in a non-trauma center. Because of higher mortality in non-trauma-centers, one might then observe, even after adjusting for baseline confounders, that survivors of those centers have better functional outcomes.
For these reasons, we conducted the analysis in two ways. First, we used multiple regression and averaging techniques to examine the effect of trauma-center care on only those who were observed to have survived for one year, adjusting for differences between patients treated at trauma centers and those treated at non-trauma centers. Second, to account for the potential "healthy survivor" bias described above, we estimated the effect of trauma-center care on a cohort of patients who would have survived for one year regardless of whether they were treated in a trauma center or a non-trauma center. This effect has been termed the survivors' average causal effect18-23, defined in terms of "potential outcomes."
Estimation of effects with the two approaches differs only in the way that averages of individual-specific regression estimates of the conditional means of functional outcomes (based on observed survivors who reported their outcomes) are weighted. The weights depend on individual-specific regression estimates of the conditional probability of survival given trauma center status and other covariates.
A logistic regression analysis was used to model survival status at one year. The covariates in this model included age, gender, race, preinjury insurance status, number of preinjury comorbidities, and injury characteristics. All subjects were included to estimate these regression parameters. For the outcome regressions, generalized linear models (linear for continuous outcomes and logistic for binary outcomes) were used to model the mean SF-36 score, the mean MFA mobility subscale score, and the probability of a CESD-R score of =16 points as a function of these same variables together with measures of education, poverty status, preinjury limitations in the performance of activities of daily living and instrumental activities of daily living, and self-reported health status. Estimates of the trauma-center effects were derived for all patients with at least one lower-limb injury with an AIS score of =3 points and separately for patients with at least one lower-limb injury with an AIS score of =3 points but no associated injuries of the head or spinal cord with an AIS score of =3 points.
All analyses were performed with use of data weighted back to the total population of patients who met the study inclusion criteria. This weighting was necessary because (1) the sampling protocol resulted in selection of all patients who had died in the hospital but only a sample of patients who had been discharged alive and (2) 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 for the study and the probability of being enrolled and having a medical record abstract, given that the patient was selected1. Applying the weights to the study sample of 1389 patients yields a study population of 4619 patients.
Multiple imputation techniques developed by Lepkowski et al.24 were used to impute missing covariates. Ten imputed data sets were created. For each data set, robust standard errors of the estimates were calculated to account for clustering within centers. Across data sets, estimates and standard errors were calculated with use of Rubin's combining rules25.
Despite the substantial improvements made over the last two decades in the delivery of acute trauma care and rehabilitation, moderate-to-severe lower-limb injuries often result in long-term physical and psychosocial disability28-36. The data presented here confirm these findings and underscore the need to address the post-acute-care needs of trauma patients across multiple domains of functioning.
To our knowledge, this is the first study that provides appropriate estimates of the causal effect of trauma-center care on functional outcomes. In estimating these effects, we used a relatively new statistical approach for addressing the potential "healthy survivor" bias that is introduced when differences in outcomes are examined only among patients who survived for at least one year. Given that only 7.6% of all patients with lower-limb injuries in the study died within one year, the effect of this bias was minimal.
Our results show that treatment at a level-I trauma center significantly improves the physical functioning of patients with a lower-limb injury with an AIS score of =3 points, but only when the injury was sustained from a high-energy force such as a motor-vehicle collision, a high fall, or a firearm injury. The average differences of 7.82 points in the SF-36 PF score and 6.31 points in the MFA mobility score are modest but consistent with differences shown to be meaningful to the patient and clinician5,6. A 5-point difference in the SF-36 PF score translates into a difference in the ability to perform one of the ten items included in the PF scale (e.g., the difference between being limited a lot in the ability to perform vigorous activities and being limited a little, or the difference between being limited a little in the ability to climb several flights of stairs and not being limited at all). A difference of 10 points translates into a difference in the ability to perform two of the ten items.
The effect of trauma-center care on psychosocial functioning is less consistent. While care at a trauma center may lead to improved lower-limb function per se (due to improved fracture-healing and more effective soft-tissue management), this difference does not always translate into improved role functions, social functioning, or overall well-being. This is likely due to the fact that both patients treated at trauma centers and those treated at non-trauma centers experience similar challenges after an injury, including posttraumatic stress and depressive symptoms that might overwhelm actual limitations in lower-limb function33,37-39. Only more recently has attention been drawn to the substantial impact that major limb trauma alone can have on psychosocial as well as physical outcomes.
While the benefits of trauma-center care on the physical functioning of patients with high-energy lower-limb injury are evident, the reason for this observation remains speculative. Trauma centers rely on an organized team approach that includes an orthopaedic surgeon with added qualifications, interest, and expertise in the management of complex injuries. Trauma centers also have higher case volumes, and many studies have demonstrated a direct relationship between complex surgical care and improved outcomes in high-volume practices40-43. Finally, because of the volume of cases and the level of surgical expertise and commitment, trauma centers might be supplied with a wider range of implants with which to better provide alignment, stabilization, and fixation of unusual fracture patterns. It is also possible that patients treated at trauma centers are referred to rehabilitation services that are more appropriate for their injury than are the services to which patients treated at non-trauma centers are referred. Future studies are needed to better identify the characteristics of a trauma center that lead to improved outcomes.
The results of this study must be interpreted in light of its limitations. Follow-up rates at one year were high (82%), although 38% of the patients who were potentially eligible for the study were not enrolled at three months after their injury. Limited discharge abstract data indicated that, when compared with the enrolled patients, the patients who were not enrolled were older and of lower socioeconomic status. This suggests that the results regarding poor overall recovery may underestimate the extent of the problem.
The generalizability of the results of this study is limited by several design features. First, the study was of the effectiveness of treatment at trauma centers in urban-suburban America. The results cannot be readily extrapolated to rural areas of the country. In addition, this study did not address the relative effectiveness of care at level-II and III trauma centers. Although resources available at level-I and level-II trauma centers are comparable in most states, centers categorized at level III are typically designed to provide initial assessment of injured patients with transfer of the more severely injured to a higher level of care. Such centers have traditionally played a critical role in more rural areas where no level-I or II centers are immediately accessible. Since the early 1990s, however, the numbers of level-III, IV, and V centers have proliferated in both urban and rural areas in an attempt to develop more inclusive systems of care44. Additional work is needed to assess the relative effectiveness of these different levels of trauma-center care. Finally, the study did not address important issues regarding treatment of children who have sustained traumatic injuries.
We believe that our estimates of the effect of trauma-center care on patients with lower-limb injuries are conservative since the non-trauma centers included in the study all treated at least twenty-five patients with major traumatic injuries each year. It is likely that the quality of trauma care is higher in these hospitals than in smaller hospitals, which were excluded from the study. In addition, seventeen of the non-trauma centers had a designated trauma team, and eight had a trauma director. Including these hospitals as non-trauma centers may also have biased the results toward a more conservative estimate of the overall trauma-center effect.
This study provides evidence that patients who sustain high-energy lower-limb trauma benefit from treatment at a level-I trauma center. Although the effect on one-year functional outcomes was modest, it was clinically meaningful and large enough to warrant further investigation of the underlying explanations for these differences. At the same time, the relatively poor outcomes for all patients with a lower-limb injury (regardless of where they were treated) argues for continued efforts to modify all aspects of care, including post-acute rehabilitation, pain control, and psychosocial interventions.
Note: The authors thank 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. They include A. Brent Eastman, MD (Chair), John W. Ashworth III, MHA, Robert R. Bass, MD, Gloria J. Bazzoli, PhD, Michael J. Bosse, MD, Nathan Cope, MD, Maurine Goehring, RN, MSN, David B. Hoyt, MD, Frank R. Lewis Jr., MD, James P. LoGerfo, MD, MPH, Ronald F. Maio, DO, MS, Donald W. Marion, MD, MSc, Colleen A. McHorney, PhD, J. Wayne Meredith, MD, Jeffrey Michael, EdD, John A. Morris Jr., MD, Richard J. Mullins, MD, Louis A. Quatrano, PhD, John C. Sacra, MD, Donald M. Steinwachs, PhD, Marc F. Swiontkowski, MD, Roger S. Taylor, MD, MPA, Harry Teter, JD, and John A. Weigelt, MD.
The integrity of the study and the quality of the data collection, management, and analysis heavily depended on the efforts of Anthony R. Carlini, MS, Lele Tang, MS, and the NSCOT nurse coordinators: Linda Agnello, RN, Marcia Baldwin, RN, JD, Sharon Blassingame, RN, Linda Carrier, RN, Carla Kimberlin, RN, Elaine Kooima, RN, Leah LeClerc, RN, MS, Cynthia Lemmon, RN, Dana McDermott, RN, M. Christine Michaelis, RN, Yeni Quintana, RN, Allana Richmond, MS, RNC, Carleen Sparks, RN, Eleanor Walsh, RN, Karen Yuhas, RN, MPH. In addition, the authors are grateful to Ciprian M. Crainiceanu, PhD, and Zhiqiang Tan, PhD, for their assistance in refining our approach to the statistical analysis of the data.
The authors also thank the sixty-nine participating hospitals: Beverly Hospital, Beverly, MA; Boston Medical Center, Boston, MA; Brockton Hospital, Brockton, MA; Cape Fear Valley Health System, Fayetteville, NC; Caritas Good Samaritan Medical Center, Brockton, MA; Carolinas Medical Center, Charlotte, NC; Citrus Valley Medical Center, Covina, CA; Cook County Hospital, Chicago, IL; Deaconess Hospital, Evansville, IN; Doctors Medical Center, Modesto, CA; Forsyth Medical Center, Winston-Salem, NC; Frederick Memorial Hospital, Frederick, MD; Froedtert Memorial Lutheran Hospital, Milwaukee, WI; Garden City Hospital, Garden City, MI; Gaston Memorial Hospital, Gastonia, NC; Greater Baltimore Medical Center, Baltimore, MD; Harborview Medical Center, Seattle, WA; Henry Ford Hospital, Detroit, MI; The Hospital of the University of Pennsylvania, Philadelphia, PA; Jackson Memorial Hospital, Miami, FL; Jacobi Medical Center, Bronx, NY; Johns Hopkins Hospital, Baltimore, MD; Kaiser Foundation Hospital, Woodland Hills, CA; Kaiser Foundation Hospital, Los Angeles, CA; Kaiser Foundation Hospital, San Diego, CA; Kendall Medical Center, Miami, FL; LAC/USC Medical Center, Los Angeles, CA; Lawrence Hospital, Bronxville, NY; Lehigh Valley Hospital, Allentown, PA; Little Company of Mary Hospital, Evergreen Park, IL; Long Island College Hospital, Brooklyn, NY; Loyola University Medical Center, Maywood, IL; Maimonides Medical Center, Brooklyn, NY; Mary Washington Hospital, Fredericksburg, VA; Memorial Medical Center, Modesto, CA; Methodist Hospital of Southern California, Arcadia, CA; Montefiore Medical Center, Bronx, NY; J. Mount Clemens General Hospital, Mount Clemens, MI; North Carolina Baptist Hospital, Winston Salem, NC; NorthEast Medical Center, Concord, NC; Oakwood Hospital and Medical Center, Dearborn, MI; PinnacleHealth Harrisburg Hospital, Harrisburg, PA; Presbyterian Intercommunity Hospital, Whittier, CA; Providence Hospital and Medical Centers, Southfield, MI; Saint Mary's Medical Center, Racine, WI; Saint Mary's Medical Center, Saginaw, MI; San Francisco General Hospital Medical Center, San Francisco, CA; San Joaquin General Hospital, French Camp, CA; Shady Grove Adventist Hospital, Rockville, MD; Sharp Grossmont Hospital, La Mesa, CA; Sinai Grace Hospital, Detroit, MI; South Jersey Hospital-Newcomb, Vineland, NJ; St. Catherine Hospital, East Chicago, IN; St. Joseph Medical Center, Towson, MD; St. Luke's Hospital of New Bedford, New Bedford, MA; St. Mary Mercy Hospital, Livonia, MI; St. Mary's Health Services, Evansville, IN; St. Mary's Hospital Medical Center, Madison, WI; St. Luke's Medical Center, Milwaukee, WI; Swedish Health Services, Seattle, WA; Swedish Medical Center, Seattle, WA; Tri-City Medical Center, Oceanside, CA; University of Maryland Medical Center, Baltimore, MD; University of California San Diego Medical Center, San Diego, CA; University of Pittsburgh Medical Center Presbyterian Hospital, Pittsburgh, PA; Virginia Mason Medical Center, Seattle, WA; Waukesha Memorial Hospital, Waukesha, WI; White Memorial Medical Center, Los Angeles, CA; and William Beaumont Hospital-Troy, Troy, MI.