Scarborough K, Slone DS, Uribe P, Craun M, Bar-Or R, Bar-Or D. Reduced Mortality at a Community Hospital Trauma CenterThe Impact of Changing Trauma Level Designation From II to I. Arch Surg. 2008;143(1):22-27. doi:10.1001/archsurg.2007.2-b
Copyright 2008 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2008
To determine if a change in trauma designation from level II (L2) to level I (L1) in the same institution reduces mortality.
Design, Setting, and Patients
A retrospective cohort study of all patients consecutively admitted to a community hospital trauma center.
The upgrade to trauma L1 designation (January 1, 2003-March 31, 2007) (n = 7902) from trauma L2 designation (January 1, 1998-December 31, 2002) (n = 9511).
Main Outcome Measures
Adjusted overall mortality and adjusted mortality for severely injured patients, patients with complications, and patients with severe sites of injury.
After adjusting for age, sex, Injury Severity Score, mechanism of injury, hypotension on admission, respirations, and comorbidities, there was a significant decrease in overall mortality during L1 designation compared with L2 designation (2.50% vs 3.48%; P = .001). Severely injured patients (Injury Severity Score of ≥ 15) admitted during an L1 trauma designation had a significant reduction in mortality compared with patients admitted during an L2 designation (8.99% vs 14.11%; P < .001). Patients admitted during an L1 designation with a severe head, chest, or abdominal or pelvic injury diagnosis had a significant decrease in mortality (9.96% vs 14.51% [P = .005], 7.14% vs 11.27% [P = .01], and 6.76% vs 17.05% [P = .002], respectively), as did patients who developed acute respiratory distress syndrome during their hospital stay (9.51% vs 26.87%; P = .02).
The significant reduction in mortality of trauma patients with severe or specific injuries after the change to a higher trauma level designation may justify direct triage of these patients to L1 facilities, when available.
Proper field triage is a critical feature of a good trauma system, and transport to appropriate facilities should optimize outcomes and use of resources.1 However, there is some disagreement about whether there is improved survival for patients admitted to higher-level trauma centers. Most studies reported improvements in survival andfunctional outcomes for level I (L1) trauma centers compared with lower-level centers2- 5 and nontrauma centers.6,7 Others8- 10 contend that there is no effect on mortality for patients triaged to an L1 center compared with a level II (L2) or lower trauma center.
The accreditation process by the American College of Surgeons (ACS) has been associated with improved outcomes.11- 14 The ACS verification for L1 and L2 status is associated with improved survival, presumably through the commitment to meet national guidelines and the establishment of a trauma program.
To our knowledge, there are no studies addressing the effect on mortality after changing ACS verification to a higher level within the same institution. A unique opportunity presented at Swedish Medical Center (SMC) to compare the effect of trauma designation and verification change from L2 to L1 on mortality. This study could help resolve the uncertainty about whether survival outcomes for comparable injuries are affected by trauma level. We hypothesized that a change to a higher trauma level designation would reduce mortality and that this potential improvement in survival would be putatively associated with staffing, protocol, and policy changes made during the L1 trauma designation.
The SMC is a privately held community hospital serving the south side of the Denver, Colorado, metropolitan area, with a historical commitment to trauma care. From January 1, 1998, to December 31, 2002, SMC was designated as an L2 trauma center and was upgraded to an L1 designation by the state of Colorado and to L1 verification by the ACS in 2002.
This study was a retrospective review of all consecutively admitted trauma patients at SMC between January 1, 1998, and March 31, 2007. Patients were considered traumatized if they were admitted by, or had a consultation with, a primary trauma surgeon (trauma L1 designation) or an on-call general surgeon (trauma L2 designation). Trauma patient records were prospectively collected and entered into a database (SMC Trauma Base) by a dedicated registrant. All patients who were entered into the database were included in the study; there were no exclusion criteria.
For all analyses, the period from January 1, 2003, through March 31, 2007 (trauma L1 designation), was compared with the preceding period of January 1, 1998, through December 31, 2002 (trauma L2 designation). The outcome measure for the study was mortality, which was defined as alive at discharge. In addition to analyzing overall mortality, we examined mortality of severely injured patients, mortality for severe injury diagnoses by anatomical site, and mortality for specific complications.
Statistical analyses were performed using SAS statistical software, version 9.1.3 (SAS Institute Inc, Cary, North Carolina). For physiologic vital signs, established clinical definitions were used: hypotension on admission was defined as lower than 90 mm Hg (vs ≥90 mm Hg), and respirations were categorized into normal (10-29/min), low (<10/min), and high (>29/min).1 Univariate analyses were used to determine variables associated with mortality. χ2 Tests were performed on categorical variables, and 2-tailed t tests were performed on continuous variables. Variables were stratified using receiver operating characteristic curves. Receiver operating characteristic curve analysis was inconclusive for age, and stratification was determined by graphic analysis. Dichotomous variables included comorbidities (≤1 vs ≥2), hypotension on admission (<90 vs ≥90 mm Hg), mechanism of injury (blunt vs penetrating), and sex. The Injury Severity Score (ISS) was categorized into minor injuries (<15), severe injuries (15-24), and very severe injuries (≥25). Age was modeled as a continuous variable (Table 1).
Three subanalyses examined the effect of injury severity, complications, and site of severe injury diagnoses. Severely injured patients included patients with an ISS of 15 or more. Complications were dichotomized as the presence or absence of the following: line sepsis or septicemia, acute respiratory distress syndrome (ARDS), acute respiratory failure, pneumonia (aspiration or other), cardiac complications (major arrhythmia, cardiac arrest, congestive heart failure, myocardial infarction, and pulmonary edema), urinary tract infections (developed after hospital admission), thromboembolic disorders (deep vein thrombosis, thrombosis, and pulmonary embolus), and any complication. Sites of injury were dichotomized as diagnoses that included the following injury sites with an associated abbreviated injury scale score of 3 or more: head, chest, face, neck or spine; abdomen or pelvic contents; or the extremities (arm or leg).
Multivariate linear logistic regression models were used to determine whether trauma level was an independent predictor of mortality, and the effect of trauma level on mortality, stratified by injury severity, the presence of complications, and the inclusion of severe anatomical injury diagnoses. Covariates associated with mortality in the univariate analysis (P < .20) were included in the model. Odds ratios of less than 1.0 represent a reduction in mortality during L1 compared with L2. P < .05 was considered statistically significant.
A total of 17 413 trauma patients were admitted between January 1, 1998, and March 31, 2007. Of these patients, 7902 (45.4%) were admitted and treated during L1, with statistically significant differences in patient demographics during L1 compared with L2: patients admitted during L1 were older, were more severely injured (ISS ≥ 15:
23.31% vs 19.07%; P < .001), and had fewer comorbidities; more patients were hypotensive on admission; and fewer patients had normal respirations on admission (Table 1).
Multivariate logistic regression analysis was performed on only complete data; patient records with missing response or explanatory variables were excluded from the analysis, and no attempt was made to substitute missing values for any patient records. There were 434 patient records deleted for patients without an ISS recorded, 4 patients had missing respiration and systolic blood pressure data, and 116 patients had a missing mechanism of injury. Excluded patient records were not statistically different when compared by level for mortality (P = .95). There were 16 879 patient records included for multivariate logistic regression analyses (96.9% of the total patient population).
From January 1998 through March 2007, the overall crude mortality was 3.69%, with 643 deaths of 17 413 patients. A χ2 test was computed for unadjusted mortality during L1 compared with L2 (Table 2), showing no significant decrease in mortality.
A multivariate linear logistic regression was performed to examine the effect of trauma designation on mortality, adjusting for age, sex, ISS, hypotension on admission, mechanism of injury, respirations, and the presence of comorbidities. The adjusted mortalities and the adjusted odds ratios are presented in Table 2. After adjustment, there was a statistically significant overall decrease in mortality during L1 compared with L2.
Trauma patients were considered severely injured if their ISSs were 15 or more. There were 3561 patients (20.45%) with severe injuries (n = 1775 for L1 and n = 1786 for L2). There were significantly more severely injured patients during L1 compared with L2 (23.31% vs 19.07%; P < .001). A χ2 test was computed for unadjusted mortality among severely injured patients during L1 compared with L2 (Table 2), showing no significant decrease in mortality. A multivariate linear logistic regression was performed and, after adjustment, there was a statistically significant decrease in mortality during L1 compared with L2.
The percentage of patients reporting with a site of injury diagnosis with an abbreviated injury scale score of 3 or more was statistically significantly higher during L1 compared with L2 for injury to the head, neck or spine, chest, face, and extremities (Table 1). All analyzed sites of injury were significantly associated with mortality.
χ2 Tests were computed for unadjusted mortality following injury to anatomical sites (Table 3). After adjustment, patients admitted with an included diagnosis of injury to the head, chest, or abdomen or pelvic contents had a significant reduction in mortality during L1 compared with those patients admitted during L2 (Table 3). The mortality of patients with a head injury decreased from 14.51% to 9.96%; for those with a chest injury, from 11.27% to 7.14%; and for those with an abdominal or pelvic injury, from 17.05% to 6.76%. The mortality data of patients with facial injuries, neck or spine injuries, or injuries to the extremities were statistically nonsignificant during L1 compared with L2.
There were significant differences for L1 compared with L2 for patients reporting with any complication, and specifically for those with a complication of thromboembolic disorders and urinary tract infections, with fewer reported during L1 (Table 4). All analyzed complications were significantly associated with mortality, except for urinary tract infections (P = .20).
χ2 Tests were computed for unadjusted mortality following development of complications (Table 5). Patients who developed ARDS (n = 130) had a significant reduction in mortality (Table 5). The adjusted mortality was reduced from 26.87% to 9.51% for L1 compared with L2 (Table 5). There were no significant differences observed for the other complications examined.
To our knowledge, the effect on mortality after changing ACS verification or designation to a higher level has not previously been reported. This study has the unique advantage of comparing L1 with L2 trauma center outcomes within the same hospital, where the preceding period served as its own control after L1 designation. Our results indicate a significant reduction in adjusted mortality for severely injured patients (ISS of ≥ 15) and the subsets of patients presenting with severe injury diagnoses to the head, chest, and abdominal or pelvic contents treated at SMC during L1 trauma designation compared with the preceding years (L2 trauma designation).
This study suggests that modifying field triage protocols for traumatized patients for transfer to the appropriate trauma facility would improve survival. The number of patients needed to be treated at an L1 trauma center over an L2 trauma center to save 1 life is as follows: overall, 70 patients; ISS of 15 or more, 22 patients; head injury, 17 patients; chest injury, 20 patients; and abdominal or pelvic injury, 8 patients. In addition, every fourth patient who developed ARDS may have been saved had the patient been triaged to an L1 trauma center.
Similar to the findings reported by Demetriades et al,3 severely injured patients treated at L1 trauma centers have significantly reduced mortality vs those treated at L2 trauma centers. Demetriades et al found the overall mortality among severely injured patients to be 14.9% for L1 and 15.4% for L2 centers using the National Trauma Data Bank, which is not significantly different from the mortality in the present study (13.18% and 14.11%, respectively). These researchers discussed a difficulty in identifying the specific resources responsible for the reduced mortality rates at L1 trauma centers. Their analyses combined 45 L1 facilities and 39 L2 facilities whose length of ACS accreditation was not uniformly known and whose specific protocols, policies, and staffing criteria were also unknown; therefore, no meaningful conclusions could be drawn as to why L1 facilities were significantly better than L2 facilities. They concluded that attaining L1 ACS accreditation justified the financial and personnel commitments.
At SMC, several changes were implemented during L1 trauma designation that may collectively explain the reduction in mortality observed after the higher trauma designation was obtained. These changes can be divided into 3 broad categories: staffing, protocol, and policy changes.
Five changes in staffing were made after attaining L1 designation. First, the trauma service established a core group of dedicated trauma surgeons with critical care certification available in-house 24 hours a day. A dedicated core of trauma surgeons, as defined by the American Association for the Surgery of Trauma as “acute care surgeons,” committed and trained in trauma care, emergency surgery, and critical care has been demonstrated to significantly lower the mortality of traumatized patients within the same system.15 This group of trauma surgeons attended to, or consulted with, all of the admitted trauma patients. They were consecutively employed throughout the study period (January 1, 2003-March 31, 2007). Previously, the surgical staff consisted of general surgeons with critical care credentials who also cared for up to 15% of nontrauma private patients. Second, operating room nurses became available in-house 24 hours a day after ACS L1 accreditation, whereas the operating room nurses were on call during L2. Third, a unique and specialized medical trauma team that exclusively serves SMC was established. This team consults with trauma surgeons to treat less severely injured trauma patients presenting with comorbidities. Fourth, a dedicated subset of specialized intensive care unit nurses who care for critically injured trauma patients was established. This group of nurses requires additional continuing medical education in trauma. Last, there is a consistent rotation of medical students, emergency department residents, and general surgery residents during the L1 period compared with the L2 period. Their presence may be beneficial in 2 ways: teaching stimulates surgeons to remain up-to-date on current practices, and the addition of the students and residents at daily rounds allows for a larger group for education and discussion in daily care.16
Before L1 designation, there were several evidence-based trauma protocols in place; however, they were not consistently followed. The use of these protocols was at the discretion of the individual surgeon. The main protocol change that took place after attaining L1 designation was compliance to these evidence-based protocols. Validated protocols for ARDS, blunt aortic injury, chest tube management, deep vein thrombosis prophylaxis, insulin for glucose control, pelvic fracture care, and the factor VII protocol were consistently adhered to.
Three main policy changes that accompany the change in trauma designation to L1 include the following: (1) regularly scheduled trauma conferences for physicians, nurses, and basic science researchers;
(2) an active basic science and clinical research program; and (3) a dedicated operating room for trauma surgery available 24 hours a day. Previous studies report that a full-time trauma service, consisting of a dedicated admitting unit, a trauma core curriculum, and quality assurance meetings, is associated with improved outcome.17
Mortality was significantly reduced for patients who developed ARDS during L1 compared with L2. This reduction in mortality is most likely associated with improved compliance of the ARDS protocols for low tidal volume18 and weaning from mechanical ventilation. In a recent study19 examining ARDS development among patients with a head injury, the proportion with ARDS or acute lung injury increased with the higher initial tidal volume settings in a dose-response relationship. Among our study population, the unadjusted mortality for patients with a head injury who developed ARDS differed significantly for L1 (after compliance to ARDS protocols) compared with L2 (9.30% [n = 4] vs 29.73% [n = 11]; P < .001). Acute respiratory distress syndrome is associated with a 3-fold increased risk of dying after severe head injury, and patients with a head injury have high rates (20%) of ARDS development.20 This finding warrants further studies examining whether the reduction in mortality among patients with a head injury is related to an improvement in survival for patients who developed ARDS.
There are several limitations to this study. This is a retrospective study; therefore, the database used for analysis was predetermined, and not all of the covariates that could be significantly associated with outcome were collected. For example, alternative measures to ISS could be better predictors of mortality, including the New ISS, the International Classification of Diseases, Ninth Revision (ICD-9), ISS, and the revised trauma score.21- 23 However, ISS was the severity score most consistently calculated at SMC. In addition, there were variables that we were unable to use because they were not routinely collected or recorded. These include scene and emergency department Glasgow Coma Scale scores, which were more consistently recorded after 2004. We are unaware of any systematic changes made to the variables we used in our analysis. Last, we included covariates previously associated with mortality following traumatic injury, but they may not represent a comprehensive list.2,3,5- 7,24- 29
In conclusion, severely injured patients (ISS of ≥ 15) and patients with head, chest, and abdominal or pelvic content injuries treated during L1 had significantly reduced mortality compared with patients admitted and treated during L2; mortality was significantly reduced for patients who developed ARDS during L1 compared with L2. Larger multicenter studies, comparing mortality in urban trauma center settings, are necessary to conclude if the observations of this study can be generalized at a local and national level.
Correspondence: David Bar-Or, MD, Trauma Research Department, Swedish Medical Center, 501 E Hampden Ave, Room 4-454, Englewood, CO 80113 (email@example.com).
Accepted for Publication: August 25, 2007.
Author Contributions:Study concept and design: D. Bar-Or. Acquisition of data: Uribe and Craun. Analysis and interpretation of data: Scarborough, Slone, and R. Bar-Or. Drafting of the manuscript: Scarborough, R. Bar-Or, and D. Bar-Or. Critical revision of the manuscript for important intellectual content: Slone, Uribe, Craun, and D. Bar-Or. Statistical analysis: Scarborough and R. Bar-Or. Administrative, technical, and material support: Slone, Uribe, and Craun. Study supervision: D. Bar-Or.
Financial Disclosure: None reported.
Additional Contributions: Karin Whinery provided access to, and initially formatted, the database (Trauma Base); the trauma coordinator staff, especially April Settell, RN, BSN, performed data collection.