Abstract
Objective
To understand hospital-level variation in triage practices for patients with moderate-to-severe injuries presenting initially to nontrauma centers.
Background
Many patients with moderate-to-severe traumatic injuries receive care at nontrauma hospitals, despite evidence of a survival benefit from treatment at trauma centers.
Methods
We used claims from the Centers for Medicare and Medicaid Services to identify patients with moderate-to-severe injuries who presented initially to nontrauma centers. We determined whether or not they were transferred to a level I or II trauma center within 24 hours of presentation, and used multivariate regression to assess the influence of hospital-level factors on triage practices, after adjusting for differences in case mix.
Results
Transfer of patients with moderate-to-severe injuries to trauma centers occurred infrequently, with significant variation among hospitals (median 2%; interquartile range 1%–6%). Greater resource availability at nontrauma centers was associated with lower rates of successful triage, including the presence of neurosurgeons (relative reduction in transfer rate: 76%, P < 0.01), more than 20 intensive care unit beds (relative reduction 30%, P < 0.01) and a high resident-to-bed ratio (relative reduction 23%, P < 0.01). However, patients were more likely to survive if they presented to hospitals with higher triage rates (odds of death for patients cared for at hospitals with the highest tercile of triage rates, compared with lowest tercile: 0.92; 95% confidence interval: 0.85–0.99, P = 0.02).
Conclusions
Injured Medicare beneficiaries presenting to nontrauma centers experience high rates of undertriage, determined in part by increasing availability of resources. Care at hospitals with low rates of successful triage is associated with worse outcomes.
Keywords: clinical practice guidelines, triage, trauma, variation
Treatment at designated trauma centers improves outcomes for patients with moderate-to-severe injuries.1–3 Accordingly, clinical practice guidelines endorse the systematic triage and transfer of these patients to trauma centers either directly from the field or after evaluation at a nontrauma center.4 Yet, between 30% to 40% of patients with moderate-to-severe injuries still receive definitive treatment at nontrauma centers, so-called “undertriage.”5,6 Much of undertriage occurs at hospitals rather than in the field, with 70% of patients initially presenting to a nontrauma center failing to reach a trauma center in a timely fashion.7,8
To address undertriage, governmental agencies and professional organizations have focused on establishing accreditation standards for referral centers and disseminating clinical practice guidelines to individual providers.9,10 In contrast, institutional characteristics that might influence triage practices have received little attention and represent a major opportunity to improve the quality of trauma care.
The objective of this study was to understand hospital-level variation in triage practices for patients with moderate-to-severe injuries presenting initially to nontrauma centers. Specifically, we sought to quantify variation in triage rates across nontrauma centers, to identify the hospital-level factors associated with failure to successfully triage patients, and to examine the relationship between successful triage and patient outcomes. We performed this analysis using Medicare claims records, both because the elderly experience high rates of undertriage and because Medicare data offer the ability to follow trauma patients across their entire episode of care, enabling a comprehensive examination of triage patterns in the United States.
METHODS
Study Design and Data
We performed a retrospective cohort study of hospital-level triage rates for patients with moderate-to-severe traumatic injuries using 2008 inpatient and outpatient data from the Centers for Medicare and Medicaid Services (CMS). We studied Medicare beneficiaries in 14 states, thought to be representative of the US population: California, Delaware, Florida, Georgia, Iowa, Massachusetts, Missouri, New Hampshire, New Jersey, New York, North Carolina, Pennsylvania, Texas, and Washington. These states contain both rural and urban populations, include a representative mix of trauma injury severity and mechanism, and implement formal regionalization of trauma care though inclusive trauma systems (which specify practice standards for all acute care hospitals) and exclusive systems (which specify standards for accredited trauma hospitals).11 Collectively, they include 51% of all Medicare beneficiaries.
We obtained hospital-level data from the 2008 CMS Healthcare Cost Report Information System (HCRIS). HCRIS contains facility-level characteristics of all nonfederal hospitals, including bed counts, ownership, and teaching status. Because HCRIS does not contain data on the trauma status of hospitals, we linked HCRIS to the Trauma Information Exchange Program to identify the trauma center designation for each hospital in 2008.12 We obtained patient-level data from the 2008 CMS Medicare Beneficiary Summary File, Medicare Provider Analysis and Review (MedPAR), Outpatient, and Carrier files. These files contain the clinical and demographic characteristics of all fee-for-service Medicare beneficiaries.
Constructing Episodes of Care for Trauma Patients
We obtained the claims of all patients over the age of 65 admitted to an acute care hospital in 2008 for a primary diagnosis of trauma using International Classification of Diseases, version 9, Clinical Modification (ICD-9-CM) codes. We classified patients as having experienced a trauma if they had an ICD-9-CM code between 800 and 959, excluding those seen for late effects of injuries (ICD-9-CM codes 905–909), foreign bodies (ICD-9-CM codes 930–940), burns (ICD-9-CM codes 940–950), or minor injuries, including isolated strains/sprains (ICD-9-CM codes 840–849), superficial injuries (ICD-9-CM codes 910–919), and contusions (ICD-9-CM 920–924).13 We did not obtain records for patients who were not admitted, a group that would have included both those with minor injuries and those with severe injuries who died in the emergency department (ED). We wanted to understand determinants of transfer practices for severely injured patients, and believed that transfer would not have been possible for these patients.
We identified admissions in MedPAR, which contains final action claim records for inpatient hospitalizations. We found ED visits that occurred within 1 day of each hospitalization by linking admissions by beneficiary and date to the Outpatient and Carrier files, using validated place of service, revenue center, and procedure codes.14 We identified the location of each visit in the MedPAR and Outpatient files by linking the hospital identifier to HCRIS. For the Carrier file, which does not include institutional information, we mapped physician billing zip codes to the closest hospital, ranked by trauma volume, using linear arc distances. We then constructed episodes of care for each patient, which began at the initial ED evaluation and ended when the patient spent at least 1 day in an acute care hospital. We excluded patients whose initial hospital presentation was at a high-level trauma center (Trauma Information Exchange Program I–II) and patients initially evaluated at hospitals that could not be linked to HCRIS.
Variables
We abstracted patient demographics [age, sex, and race (white, black, Hispanic, and other)] and vital status 90-days after admission from the Medicare Beneficiary Summary file. We abstracted comorbid conditions using the Elixhauser methodology and injury characteristics using ICD-9-CM diagnosis codes from the claims.15 We used ICD-9-CM codes to identify life-threatening/critical injuries, on the basis of the American College of Surgeons Committee on Trauma interfacility transfer guidelines (see Appendix), and used a validated algorithm to translate ICD-9-CM diagnostic codes into abbreviated injury scores.4,13,16
We used HCRIS and MedPAR to define hospital characteristics that might influence trauma triage practices. From HCRIS, we identified each hospital's teaching status (defined using the resident-to-bed ratio), ownership (nonprofit, for profit, and government run), number of intensive care unit (ICU) beds, rural and urban status (defined using the size of the hospital's metropolitan statistical area), the number of trauma centers in the hospital's Dartmouth Atlas Hospital Referral Region (HRR), and the linear arc distance from the hospital to the nearest level I/II trauma center.
From MedPAR, we identified 9 radiological and subspecialty surgical services that might influence the triage of trauma patients, including computed tomographic (CT) scans, magnetic resonance imaging, neurosurgery, spine surgery, orthopedic surgery, cardiothoracic surgery, vascular surgery, urologic surgery, and maxillofacial surgery. We considered hospitals to have the resource if they issued 10 or more claims in 2008.
Analysis
We classified patients as having moderate-to-severe injuries as specified by the American College of Surgeons Committee on Trauma definition, which includes those with either a life-threatening/critical injury or an injury severity score (ISS) greater than 15.4 On the basis of the episodes of trauma care we had constructed, we defined successful triage as the transfer of a patient with moderate-to-severe injuries to a level I or II trauma center either directly from the ED or within 1 day of admission. We defined unsuccessful triage as admission to a nontrauma center or a level III/IV center for more than 1 day, or transfer to a non–level I or non–level II trauma center. We compared the characteristics of successfully and unsuccessfully triaged patients using the Student t test and χ2 text.
Hospital-Level Variation in Successful Triage Rates
Recognizing that triage rates would reflect differences in case mix and volume at hospitals, we used adjusted triage rates to study variation. For each nontrauma center, we calculated hospital-specific triage rates, defined as the proportion of evaluated patients who were triaged successfully. We then used a patient-level multivariable hierarchical logistic regression model to create hospital-specific adjusted triage rates that accounted for differences in case-mix and low reliability at small volume hospitals.17 Variables in this model included age, sex, selected comorbidities known to influence mortality, and the maximum abbreviated injury scores by body region.18 After calculating the adjusted rates, we quantified variation using standard summary statistics. We graphically examined variation by plotting observed-to-expected rates of transfer against hospital volume.
Hospital Factors Associated With Successful Triage Rates
To determine the relationship between hospital characteristics and triage rates, we categorized hospitals as having a low (<1%), moderate (1%–4%), or high (>4%) adjusted successful triage rate based on terciles of patients. We performed bivariate analyses using χ2 or analysis of variance tests. We also performed a hospital-level multivariate analysis using random-effects negative binomial regression, in which the dependent variable was a count of each hospital's successfully triaged patients, accounting for the total number of patients with moderate-to-severe injuries. We included all hospital and region-level factors as fixed effects and specified an HRR-level random effect. The exponentiated coefficients in this model were interpreted as hospital-specific incidence rate ratios, controlling for the other covariates in the model.
Association Between Successful Triage Rates and Patient Outcomes
Finally, we examined the relationship between hospital-level triage rates and patient outcomes. We hypothesized that patients with moderate-to-severe injuries presenting to hospitals with higher successful triage rates would experience lower mortality, because they would ultimately be more likely to receive care at high-level trauma centers.1 To test this hypothesis, we fit patient-level logistic regression models in which the dependent variables were 30-day, 90-day, and 365-day mortality from admission and the independent variables were the adjusted successful triage rate of the presenting hospital, the size of the hospital's metropolitan statistical area, distance to the nearest trauma center, and the number of trauma centers in the hospital's HRR. We used generalized estimating equations with robust variance estimators to account for clustering by hospital.
Data management and statistical analyses were performed using STATA 12.0 (College Station, TX), with statistical significance set at P < 0.05. The University of Pittsburgh institutional review board approved this project.
RESULTS
In the 14-state sample, 1757 level III, level IV, or nondesignated trauma centers evaluated at least 1 trauma patient with moderate-to-severe injuries. We could not match 52 (3%) hospitals in HCRIS, leaving 1705 hospitals in the final analysis. A total of 35,621 trauma patients with moderate-to-severe injuries presented initially to these 1705 hospitals. Of these, 4123 (12%) were transferred to a level I or II trauma center (“successful triage”). A total of 28,828 (92%) were admitted to the initial hospital for more than 1 day, and 2670 (8%) were transferred to another nontrauma center, a level III or a level IV trauma center (“unsuccessful triage”) (Fig. 1). Among the 28,828 patients admitted to the initial hospital, 104 (0.4%) were transferred after 24 hours: 47 (45%) to a level I/II trauma center, and 57 (55%) to a non–level I/II trauma center. Patients triaged successfully were older, less likely to have comorbidities, and more likely to have higher ISSs (Table 1).
FIGURE 1.
Flowchart of patients and episodes of care in the study.
TABLE 1.
Characteristics of Patients by Triage Status*
| Variable | Successfully Triaged (n = 4123) | Unsuccessfully Triaged (n = 31,498) | P |
|---|---|---|---|
| Age, mean (SD), yr | 80 (8.4) | 78(11.6) | <0.01 |
| Female, n (%) | 2173 (53) | 17,572 (56) | <0.01 |
| Race, n (%) | <0.01 | ||
| White | 3791 (92) | 27,680 (88) | |
| Black | 155 (4) | 1538 (5) | |
| Hispanic | 55 (1) | 980 (3) | |
| Other | 122 (3) | 1300 (4) | |
| Select comorbidities, n (%) | <0.01 | ||
| Cirrhosis | 18 (0.4) | 360 (1) | |
| Congenital coagulopathy | 165 (4) | 997 (3) | |
| Congestive heart failure | 346 (8) | 3497 (11) | |
| Chronic obstructive pulmonary disease | 337 (8) | 4442 (14) | |
| Diabetes | 675 (16) | 6316 (20) | |
| 30-d mortality | 616 (15) | 3738 (12) | <0.01 |
| 90-d mortality | 843 (20) | 5492 (17) | <0.01 |
| 365-d mortality | 1134 (28) | 8291 (26) | 0.1 |
| Types of life-threatening injuries, n (%) | <0.01 | ||
| Traumatic brain injury | 826 (20) | 2546 (8) | |
| Multiple rib fractures | 584 (14) | 6094 (19) | |
| Cardiac injury | 210 (5) | 1179 (4) | |
| Open long bone fracture | 180 (4) | 3869 (12) | |
| Severe torso injury with comorbidity | 262 (6) | 2883 (9) | |
| ISS, mean (SD) | 18.0 (5.9) | 14.6 (6.0) |
Patients who presented to acute care nontrauma centers in California, Delaware, Florida, Georgia, Iowa, Massachusetts, Missouri, New Hampshire, New Jersey, New York, North Carolina, Pennsylvania, Texas, and Washington in 2008 after a traumatic injury.
Among the 1705 nontrauma centers, we found variation in successful triage rates among hospitals (Fig. 2). The median adjusted triage rate was 2%, with an interquartile range of 1% to 6% and a total range of 0% to 82%.
FIGURE 2.
Funnel plot of adjusted triage rates for patients with moderate to severe injuries in 1705 nontrauma centers. The mean triage rate was 4%, with 151 hospitals performing more than 2 standard deviations less than the mean rate and 281 hospitals performing more than 2 standard deviations more than that (shown in gray).
After categorizing hospitals into terciles, we found that greater resources available at the nontrauma centers, including more ICU beds, a higher resident-to-bed ratio, and the presence of specialty surgical services, were associated with lower rates of successful triage in bivariable analyses (Table 2). Location in the northeast region of the country, in a larger metropolitan statistical area, and proximity to a level I/II trauma center were also associated with lower rates of successful triage. In the multivariable model, the presence of trauma-related resources at the nontrauma center, ownership of the hospital, and size of the metropolitan statistical area remained associated with lower successful triage rates (Table 3).
TABLE 2.
Hospital Characteristics by Terciles of Adjusted Successful Triage Rates at Nontrauma Centers*
| Variable | Low Success Rates; <1% (n = 276) | Moderate Success Rates; 1–4% (n = 605) | High Success Rates; >4% (n = 824) | P |
|---|---|---|---|---|
| No. moderate-to-severe trauma admissions | 33 (13–59) | 11 (3–26) | 9 (4–19) | <0.01 |
| ICU bed number | 22 (12–38) | 10 (0–21) | 7 (0–12) | <0.01 |
| Ownership | <0.01 | |||
| Government | 34 (11) | 135 (22) | 205 (27) | |
| Nonprofit | 172 (55) | 303 (49) | 445 (58) | |
| For profit | 105 (34) | 186 (30) | 120 (16) | |
| Teaching status† | <0.01 | |||
| Nonteaching | 197 (63) | 519 (83) | 654 (85) | |
| Small teaching | 87 (28) | 68 (11) | 85 (11) | |
| Large teaching | 27 (9) | 37 (6) | 31 (4) | |
| Resources available | <0.01 | |||
| Neurosurgery | 191 (61) | 160 (26) | 48 (6) | |
| Orthopedic surgery | 293 (94) | 473 (76) | 535 (69) | |
| Spine surgery | 266 (86) | 358 (57) | 347 (45) | |
| Cardiothoracic surgery | 273 (88) | 382 (61) | 391 (51) | |
| Facial surgery | 25 (8) | 32 (5) | 20 (3) | |
| Vascular surgery | 258 (83) | 341 (55) | 302 (39) | |
| Urology | 285 (92) | 433 (69) | 467 (61) | |
| CT | 300 (96) | 582 (93) | 753 (98) | |
| Magnetic resonance imaging | 85 (27) | 110 (18) | 74 (10) | |
| MSA size | <0.01 | |||
| <100,000 | 31 (10) | 162 (26) | 331 (43) | |
| 100,000 to 1 million | 110 (35) | 147 (24) | 192 (25) | |
| >1 million | 170 (55) | 315 (50) | 247 (32) | |
| Distance to nearest level I/II trauma center, miles | <0.01 | |||
| <2 | 32 (10) | 59 (9) | 48 (6) | |
| 2–7 | 99 (32) | 123 (20) | 99 (13) | |
| 8–20 | 58 (19) | 127 (20) | 129 (17) | |
| 21–49 | 58 (19) | 144 (23) | 321 (42) | |
| ≥50 | 64 (21) | 171 (27) | 173 (23) | |
| No. level I/II trauma centers in region | 1 (1–3) | 2 (1–4) | 2 (1–3) | 0.02 |
| Region | <0.01 | |||
| Northeast | 55 (18) | 150 (24) | 232 (30) | |
| South | 161 (52) | 296 (47) | 267 (35) | |
| Midwest | 12 (4) | 36 (6) | 146 (19) | |
| West | 83 (27) | 142 (23) | 125 (16) |
MSA indicates metropolitan statistical area.
Successful adjusted triage rates for patients presenting to nontrauma centers with moderate-to-severe injuries.
Teaching status categorized by resident-to-bed ratio (nonteaching = 0, small teaching = >0 and <0.25, large teaching = ≥0.25).
TABLE 3.
Association Between Hospital and Regional Characteristics and Successful Adjusted Triage Rates*
| Variable | Incident Rate Ratio | P |
|---|---|---|
| ICU bed number | ||
| <10 | Referent | |
| 10–20 | 0.86 (0.76–0.99) | 0.03 |
| >20 | 0.70 (0.60–0.81) | <0.01 |
| Ownership | ||
| Government | Referent | |
| Nonprofit | 0.88 (0.79–0.98) | 0.02 |
| For profit | 0.77 (0.67–0.88) | <0.01 |
| Teaching status | ||
| Nonteaching | Referent | |
| Small teaching | 0.86 (0.77–0.96) | <0.01 |
| Large teaching | 0.77 (0.63–0.93) | 0.26 |
| Resources | ||
| Neurosurgery | 0.34 (0.31–0.39) | <0.01 |
| Spine surgery | 0.84 (0.74–0.95) | <0.01 |
| Orthopedic surgery | 0.87 (0.75–1.01) | 0.07 |
| Cardiothoracic surgery | 0.95 (0.81–1.10) | 0.48 |
| Vascular surgery | 0.86 (0.75–0.99) | 0.03 |
| Urology | 0.87 (0.75–1.00) | 0.07 |
| MSA | ||
| <100,000 | Referent | |
| 100,000–1 million | 0.87 (0.77–0.98) | 0.02 |
| >1 million | 0.82 (0.71–0.95) | <0.01 |
| Distance to nearest TC, miles | ||
| <2 | Referent | |
| 2–7 | 0.93 (0.79–1.09) | 0.36 |
| 8–20 | 0.87 (0.73–1.03) | 0.10 |
| 21–49 | 0.89 (0.75–1.05) | 0.17 |
| ≥50 | 0.67 (0.55–0.82) | <0.01 |
| No. trauma centers in HRR | ||
| 0 | Referent | |
| 1–3 | 1.43 (1.16–1.75) | <0.01 |
| ≥4 | 1.27 (0.97–1.65) | 0.08 |
| Region | ||
| Northeast | Referent | |
| South | 1.00 (0.99–1.20) | 0.93 |
| Midwest | 1.25 (0.72–1.57) | 0.06 |
| West | 0.88 (0.72–1.06) | 0.18 |
MSA indicates metropolitan statistical area; TC, trauma center.
Estimates obtained from a negative binomial regression model. The incident rate ratio can be interpreted so that for a 1-imit change in the predictor variable, the rate ratio of successful triage is expected to change by the respective regression coefficient, presented in exponentiated form.
After adjustment for hospital and regional characteristics, the odds of death at 30 and 90 days after a moderate-to-severe injury were not associated with rates of successful triage. However, the odds of death at 365 days was inversely associated with rates of successful triage (odds radio = 0.97, confidence interval: 0.91–1.04, P = 0.43 if moderate success compared with low; odds radio = 0.92, confidence interval: 0.85–0.99, P = 0.02 if high success compared with low) (Fig. 3).
FIGURE 3.
Relationship between rates of successful triage and 365-day mortality. The adjusted odds of death are presented relative to patients treated at hospitals with the lowest tercile of successful triage rates.
DISCUSSION
In a retrospective analysis of triage patterns at nontrauma centers, we found that Medicare beneficiaries with moderate or severe trauma were typically not sent to trauma centers, despite long-standing quality improvement efforts by stakeholders. Even after adjusting for differences in patient characteristics, we found wide variation in how successfully these hospitals triaged patients. Nontrauma hospitals most likely to keep trauma patients tended to be better resourced, with many of the features of trauma centers, including more ICU beds and the presence of surgical subspecialties, but without the trauma center designation. Nonetheless, patients admitted to these hospitals experienced higher mortality than patients taken to worse resourced hospitals with greater success at triage.
There are several potential explanations for the association between hospital resources and triage success. First, the presence of resources for acute care may influence physician decision making. Physicians may substitute the question of whether their hospital has the resources necessary to manage the patient's case for the more complex one of whether the patient meets the criteria of the American College of Surgery for transfer. Robust evidence indicates that people faced with complex tasks routinely overestimate their abilities, resulting in predictable errors (biases).20
A second possibility is that nonclinical factors may factor into the transfer decisions, such as the social networks of physicians or institutional norms. As in other contexts, physicians may prefer to make referrals to local colleagues with whom they have a relationship, instead of sending patients to an anonymous trauma surgeon at a tertiary care center.21 Institutional norms could influence the treatment of trauma patients if hospitals belong to health systems that influence transfer rates independent of guidelines, or if hospitals exert implicit or explicit pressure on emergency physicians to retain patients in-house. Finally, patient preferences may also exert an influence on transfer rates. In other clinical contexts, patients describe a willingness to trade the convenience of receiving care close to home for the increased risk of worsened morbidity (the so-called distance-quality tradeoff).22,23
Our study corroborates and extends prior work demonstrating the importance of institutional characteristics on trauma triage practices.8 Gomez et al examined the influence of 5 physical and human resources on transfer practices in Ontario, Canada, and concluded that the presence of a CT scanner and a general surgeon reduced the likelihood of transfer to a trauma center. In contrast to their study, we found that the availability of other subspecialists also played an important role in triage practices. This observation may reflect the differences between Canadian and American hospitals; for example, only 21% of Canadian hospitals had both a CT scanner and orthopedic surgeon, whereas 96% of hospitals in the United States had both of these resources.
Together, these findings have important implications for improving regionalization in trauma. Existing quality improvement efforts have focused on establishing accreditation standards for referral centers and disseminating clinical practice guidelines to individual providers.9,10 This study confirms that these efforts have had only modest success. A possible solution might involve creating financial incentives for transfer through pay-for-performance, rewarding hospitals with high successful triage rates. At the same time, hospitals that excel at triaging trauma patients have the potential to provide insights to institutions that wish to improve their performance—structural factors at these hospitals, such as decision support tools, might be exported, potentially improving transfer rates.24 Other possibilities include greater effort at regionalizing the resources most likely to influence triage patterns, such as the availability of neurosurgery, as have Veteran Affairs hospitals.23 Finally, use of basic behavioral science methods to analyze variability might result in novel approaches to quality improvement. Groups like the Transportation Security Administration have effectively used strategies like recalibrating the heuristics of baggage screeners to improve threat detection.25
We also confirmed an association between triage quality and patient outcomes. Although our primary goal was to examine variation in transfer guideline adherence, many experts question the validity of the triage guidelines in elderly patients.26 Our results support the notion that improved guideline adherence may be associated with lower mortality, at least at the hospital level. We avoided a patient-level analysis because of concerns that we could not adequately adjust for indication biases inherent in patient-level triage decisions. We acknowledge that hospital triage rates might serve as a proxy for other performance metrics, such as overall quality of care.
Our study has several limitations. First, we restricted our sample to Medicare beneficiaries, which may prevent generalizability of our findings to other trauma patients. Elderly patients are a unique population with specific needs, and they may have had prior interactions with local hospitals, making triage patterns less representative of the general population. Nonetheless, Medicare patients account for more than 25% of all deaths from injury and experience high rates of undertriage.27,28 Moreover, by concentrating on this population, we could reduce the effect of 2 patient-level characteristics known to influence triage patterns: age and insurance status.29,30 Second, we sampled only patients with moderate-to-severe injuries, which might also limit generalizability. However, we hypothesized that factors associated with undertriage would differ from those associated with overtriage (the inappropriate transfer of patients with minor injuries to trauma centers) and would bias our results if we conflated the groups. Third, we used HRRs as a surrogate for trauma referral regions. HRRs were developed to describe referral patterns for cardiothoracic surgery and neurosurgery, and little is known about the validity of these regions in trauma. If HRRs do not accurately describe trauma referral patterns, then we may have underestimated the influence of the region on hospital performance. Fourth, we limited our analysis to hospital-level variables, not including other potentially important determinants of variation, such as patient preferences and the practice style of individual physicians.
CONCLUSIONS
Improving the quality of trauma regionalization is a high priority for multiple health care stakeholders.24 Using Medicare claims, we found substantial room for improvement, with wide variation in hospital rates of triage, even after adjusting for differences in case mix. Several hospital-level factors served to explain this variation, suggesting that current clinical decision making is driven by heuristics surrounding resource availability rather than the existing guidelines. Given the association between variation and outcome, this study indicates the need for greater implementation efforts designed to improve triage quality, as well as, the need for research aimed at understanding of how resource availability influences triage decision making.
ACKNOWLEDGMENTS
The authors contributed to this study in the following ways— Dr Mohan had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis; Study concept and design: Mohan, Barnato, Rosengart, Wallace, Angus, and Kahn; Acquisition of data: Mohan; Analysis and interpretation of data: Mohan, Barnato, Rosengart, Wallace, Angus, and Kahn; Drafting of the manuscript: Mohan; Critical revision of the manuscript for important intellectual content: Barnato, Rosengart, Wallace, Angus, and Kahn; Statistical analysis: Mohan, Kahn; Obtained funding: Mohan, Barnato, Rosengart, Angus; and Administrative, technical, or material support: Mohan, Kahn, Angus.
Disclosure: This work was supported by the National Institutes of Health through grants number 1K23 GM101292-01 (Mohan), K12-HL109068 (Wallace), 3UL1RR024153-06S2, and 8UL1TR000005-07. The authors have no financial conflicts to disclose. This work was performed at the University of Pittsburgh School of Medicine.
APPENDIX.
Algorithm Used to Select Injuries Described by the ACS-COT as Indicating the Need for Transfer
| Injuries the ACS-COT Considers Life-Threatening or Critical | ICD-9-CM Codes | Assumptions/Notes |
|---|---|---|
| Carotid or vertebral injuries | 900 | |
| Aorta or great vessel injuries | 901 | |
| Cardiac rupture | 861.0 and 861.1 | 1. Unable to identify cardiac rupture from diagnosis codes so included all cardiac injuries. |
| Bilateral pulmonary contusions with PaO2/FiO2 ratio < 200 | — | 1. Unable to identify based on discharge codes |
| Major abdominal vascular injury | 902 | |
| Grade IV or V liver injury with >6 units pRBC | 864.04 or 864.14 | 1. Unable to identify RBC transfusion so included all Grade IV liver lacerations |
| Unstable pelvic fracture with >6 units pRBC | 808.43 or 808.53 | 1. Unable to identify unstable fractures so used the surrogate of the disrupted pelvic circle. |
| Fracture or dislocation with loss of distal pulses | 903.1–903.3 in conjunction with 812, 813, and 818 |
1. Unable to identify fractures with loss of pulses so used the surrogate of fracture with vascular injury. |
| 904.1–904.5 in conjunction with 821, 822, 823, 824, and 827 |
2. Excluded patients with amputations, which would technically fit into this category, because of the problem of misclassification.13 |
|
| Open skull fracture or penetrating injury | 800.5–800.9 | 1. Unable to identify penetrating injury so captured only the open skull fractures. |
| 801.5–801.9 | ||
| 803.5–803.9 | ||
| 804.5–804.9 | ||
| GCS < 14 or lateralizing neurologic injury | 800–804 | 1. Unable to calculate GCS from discharge diagnosis codes so used the fifth digit subclassification to identify patients who had either a moderate (1–24 h) or prolonged (>24 h) loss of consciousness (eg, 852.03). We assumed these patients would appear clinically to have a GCS < 14. |
| 850.2–850.5 | ||
| 851–854 | ||
| Spinal cord deficit or lateralizing neurologic sign | 806 | |
| 952–955.2 | ||
| 956–956.2 | ||
| Spinal column fractures | — | 1. Excluded because of the problem of misclassification.13 |
| >2 unilateral rib fractures or bilateral rib fractures | 807.03–807.1 | 1. Unable to identify patients with bilateral rib fractures so included only patients with >2 rib fractures. |
| 807.4–807.6 | ||
| Open long bone fracture | 812.1, 812.3, 812.5 | |
| 813.1, 813.3, 813.5 | ||
| 820.1, 820.3, 820.5 | ||
| 821.1, 821.3, 821.5 | ||
| 823.1, 823.3, 823.5 | ||
| 824.1, 824.3, 824.5 | ||
| Significant torso injury with advanced comorbid disease | 860.1, 860.3, 860.5 | 1. ICD-9-CM codes indicated significant torso injury. We included patients who had these ICD-9 codes as well as an Elixhauser comorbidity code. |
| 861.2–861.9 | ||
| 862–870 |
ACS-COT = American College of Surgeons - Committee on Trauma; MSA = metropolitan statistical area; TC = trauma center. ACS-COT indicates American College of Surgeons - Committee on Trauma.
REFERENCES
- 1.MacKenzie EJ, Rivara FP, Jurkovich GJ, et al. A national evaluation of the effect of trauma-center care on mortality. N Engl J Med. 2006;354:366–378. doi: 10.1056/NEJMsa052049. [DOI] [PubMed] [Google Scholar]
- 2.Celso B, Tepas J, Langland-Orban B, et al. A systematic review and meta-analysis comparing outcome of severely injured patients treated in trauma centers following the establishment of trauma systems. J Trauma. 2006;60:371–378. doi: 10.1097/01.ta.0000197916.99629.eb. [DOI] [PubMed] [Google Scholar]
- 3.Demetriades D, Martin M, Salim A, et al. The effect of trauma center designation and trauma volume on outcome in specific severe injuries. Ann Surg. 2005;242:512–517. doi: 10.1097/01.sla.0000184169.73614.09. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Committee on Trauma—American College of Surgeons . Resources for Optimal Care of the Injured Patient 2006. American College of Surgeons; Chicago, IL: 2006. [Google Scholar]
- 5.Macias CA, Rosengart MR, Puyana JC, et al. The effects of trauma center care, admission volume, and surgical volume on paralysis after traumatic spinal cord injury. Ann Surg. 2009;249:10–17. doi: 10.1097/SLA.0b013e31818a1505. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Nathens AB, Jurkovich GJ, MacKenzie EJ, et al. A resource-based assessment of trauma care in the United States. J Trauma. 2004;56:173–178. doi: 10.1097/01.TA.0000056159.65396.7C. [DOI] [PubMed] [Google Scholar]
- 7.Mohan D, Rosengart MR, Farris C, et al. Assessing the feasibility of the American College of Surgeons' benchmarks for the triage of trauma patients. Arch Surg. 2011;146:786–792. doi: 10.1001/archsurg.2011.43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Gomez D, Haas B, de Mestral C, et al. Institutional and provider factors impeding access to trauma center care: an analysis of transfer practices in a regional trauma system. J Trauma. 2012;73:1288–1293. doi: 10.1097/TA.0b013e318265cec2. [DOI] [PubMed] [Google Scholar]
- 9.US Department of Health and Human Services [Accessed February 14, 2006];Model trauma system planning and evaluation. Available at https://www.socialtext.net/acs-demo-wiki/index.cgi?regional_trauma_systems_optimal_elements_integration_and_assessment_systems_consultation_guide.
- 10.Myers RA. Advanced trauma life support courses. J R Soc Med. 1990;83:281–282. doi: 10.1177/014107689008300501. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Utter GH, Maier RV, Rivara FP, et al. Inclusive trauma systems: do they improve triage or outcomes of the severely injured? J Trauma. 2006;60:529–537. doi: 10.1097/01.ta.0000204022.36214.9e. [DOI] [PubMed] [Google Scholar]
- 12.American Trauma Society . National Trauma Center Maps from TIEP. American Trauma Society; Falls Church, VA: [Accessed November 20, 2012]. 2010. Available at http://www.amtrauma.org/resources/trauma-center-maps/index.aspx. [Google Scholar]
- 13.Mohan D, Barnato AE, Rosengart MR, et al. Trauma triage in the emergency departments of non-trauma centers: an analysis of individual physician caseload on triage patterns. J Trauma. 2013;74:1541–1547. doi: 10.1097/TA.0b013e31828c3f75. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Hennessy S, Leonard CE, Freeman CP, et al. Validation of diagnostic codes for outpatient-originating sudden cardiac death and ventricular arrhythmia in Medicaid and Medicare claims data. Pharmacoepidemiol Drug Saf. 2010;19:555–562. doi: 10.1002/pds.1869. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Elixhauser A, Steinter C, Harris RD, et al. Comorbidity measures for use with administrative data. Med Care. 1998;36:8–27. doi: 10.1097/00005650-199801000-00004. [DOI] [PubMed] [Google Scholar]
- 16.MacKenzie EJ, Steinwachs DM, Shankar B. Classifying trauma severity based on hospital discharge diagnoses: validation of an ICD-9CM to AIS-85 conversion table. Med Care. 1989;27:412–422. doi: 10.1097/00005650-198904000-00008. [DOI] [PubMed] [Google Scholar]
- 17.Kahn JM, Werner RM, Carson SS, et al. Variation in long-term acute care hospital use after intensive care. Med Care. 2012;69:339–350. doi: 10.1177/1077558711432889. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Morris JA, MacKenzie EJ, Edelstein SL. The effect of pre-existing conditions on mortality in trauma patients. JAMA. 1990;263:1942–1946. [PubMed] [Google Scholar]
- 19.Kahneman D, Tversky A. Judgment under uncertainty: heuristics and biases. In: Kahneman D, Slovic P, Tversky A, editors. Judgment Under Uncertainty: Heuristics and Biases. Cambridge University Press; New York, NY: 2003. pp. 3–20. [Google Scholar]
- 20.Plous S. The Psychology of Judgment and Decision Making. McGraw Hill; New York, NY: 1993. [Google Scholar]
- 21.Kinchen KS, Cooper LA, Levine D, et al. Referral of patients to specialists: factors affecting choice of specialist by primary care physicians. Ann Fam Med. 2004;2:245–252. doi: 10.1370/afm.68. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Finlayson SR, Birkmeyer JD, Tosteson ANA, et al. Patient preferences for location of care: implications for regionalization. Med Care. 1999;37:204–209. doi: 10.1097/00005650-199902000-00010. [DOI] [PubMed] [Google Scholar]
- 23.Liu JH, Zingmond DS, McGory ML, et al. Disparities in the utilization of high-volume hospitals for complex surgery. JAMA. 2006;296:1973–1980. doi: 10.1001/jama.296.16.1973. [DOI] [PubMed] [Google Scholar]
- 24.Bradley EH, Curry LA, Webster TR, et al. Achieving rapid door-to-balloon times: how top hospitals improve complex clinical systems. Circulation. 2006;113:1079–1085. doi: 10.1161/CIRCULATIONAHA.105.590133. [DOI] [PubMed] [Google Scholar]
- 25.Institute of Medicine (IOM) Regionalizing Emergency Care: Workshop Summary. The National Academies Press; Washington, DC: 2010. [PubMed] [Google Scholar]
- 26.Wolfe JM, Horowitz TS, Van Wert MJ, et al. Low target prevalence is a stubborn source of errors in visual search tasks. J Exp Psychol Gen. 2007;136:623–638. doi: 10.1037/0096-3445.136.4.623. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Calland JF, Ingraham AM, Martin N, et al. Evaluation and management of geriatric trauma: an Eastern Association for the Surgery of Trauma practice management guideline. J Trauma. 2012;73:S345–S350. doi: 10.1097/TA.0b013e318270191f. [DOI] [PubMed] [Google Scholar]
- 28.Centers for Disease Control and Prevention [Accessed on March 15, 2013];Injury prevention and control: data and statistics. http://www.cdc.gov/injury/wisqars/fatal.html.
- 29.Chang DC, Bass RR, Cornwell EE, et al. Undertriage of elderly trauma patients to state-designated trauma centers. Arch Surg. 2008;143:776–781. doi: 10.1001/archsurg.143.8.776. [DOI] [PubMed] [Google Scholar]
- 30.Nathens AB, Maier RV, Copass MK, et al. Payer status: the unspoken triage criterion. J Trauma. 2001;50:776–783. doi: 10.1097/00005373-200105000-00002. [DOI] [PubMed] [Google Scholar]