Factors Associated With Nontransfer in Trauma Patients Meeting American College of Surgeons’ Criteria for Transfer at Nontertiary Centers

Key Points

Question  What are factors associated with nontransfer among trauma patients taken to nontertiary centers who meet guideline criteria to consider transfer to facilities providing higher-level care?

Findings  In this cohort study of 96 528 patients, few meeting guideline criteria to consider transfer to higher-level care were transferred from nontertiary centers. Patient-level factors associated with nontransfer included older age, chest injury, and commercial insurance, accounting for approximately one-third of the transfer variation; center-level factors included greater center resources and university affiliation, accounting for approximately half of transfer variation.

Meaning  These factors associated with nontransfer may be useful for trauma system stakeholders to target education and outreach to guide development of more inclusive trauma systems.

Importance  Secondary triage from nontertiary centers is vital to trauma system success. It remains unclear what factors are associated with nontransfer among patients who should be considered for transfer to facilities providing higher-level care.

Objective  To identify factors associated with nontransfer among patients meeting American College of Surgeons (ACS) guideline criteria for transfer from nontertiary centers.

Design, Setting, and Participants  A retrospective cohort study was performed using multilevel logistic regression to ascertain factors associated with nontransfer from nontertiary centers, including demographics, injury characteristics, and center resources. With information obtained from the National Trauma Data Bank (January 1, 2007, to December 31, 2012), relative proportion of variance in outcome across centers was determined for patient-level and center-level attributes. In all, 96 528 patients taken to nontertiary centers (levels III, IV, V, and nontrauma centers) that met ACS guideline transfer criteria were eligible for inclusion. Data analysis was performed from March 17, 2016, to May 20, 2016.

Main Outcomes and Measures  The primary outcome was nontransfer from a nontertiary center.

Results  Among 96 528 patients meeting ACS guideline criteria for transfer taken initially to nontertiary centers, 55 611 (57.6%) were male and the median age was 52 years (interquartile range, 28-77 years). Only 19 396 patients (20.1%) underwent transfer. Patient-level factors associated with nontransfer included age older than 65 years (adjusted odds ratio [AOR], 1.70; 95% CI, 1.46-1.98; P < .001), severe chest injury (AOR, 1.63; 95% CI, 1.42-1.89; P < .001), and commercial insurance (vs self-pay: AOR, 1.39; 95% CI, 1.15-1.67; P < .001). Center-level factors associated with nontransfer included larger bed size (>600 vs <200 beds: AOR, 9.22; 95% CI, 7.70-11.05; P < .001), nontrauma center (vs level III centers: AOR, 2.71; 95% CI, 2.44-3.01; P < .001), university affiliation (vs community: AOR, 9.68; 95% CI, 8.03-11.66; P < .001), more trauma surgeons (per surgeon: AOR, 1.08; 95% CI, 1.06-1.09; P < .001), and more neurosurgeons (per surgeon: AOR, 1.25; 95% CI, 1.23-1.28; P < .001). For-profit status was associated with nontransfer at nontrauma centers (AOR, 1.55; 95% CI, 1.39-1.74; P < .001), but not at level III, IV, and V trauma centers. Overall, patient-level factors accounted for 36% and center-level factors accounted for 58% of the variation in transfer practices. Patient-level factors accounted for more variation at level III, IV, and V trauma centers (44%), but less variation at nontrauma centers (13%).

Conclusions and Relevance  Only 1 in 5 patients meeting ACS transfer criteria underwent transfer. Factors associated with nontransfer may be useful for trauma system stakeholders to target education and outreach to guide development of more inclusive trauma systems. Further study is necessary to critically evaluate whether these ACS criteria identify patients who require transfer.

Organized trauma systems lead to improvements in outcome.¹ Some injured patients are initially transported to nontertiary centers. Ideally, severely injured patients should be transferred promptly to a facility that can provide a higher level of care.² However, a significant proportion of severely injured patients evaluated at nontrauma centers (NTCs) or lower-level trauma centers are never transferred.^3,4

Prior studies show that a variety of factors affect decision making in the triage of patients.⁵ The American College of Surgeons (ACS) Committee on Trauma published a set of guideline criteria to assist clinicians in identifying patients to consider for transfer to level I or II trauma centers from nontertiary centers.⁶ Some studies demonstrate that adherence to these guidelines for transfer is poor.^7,8 Few studies have investigated patients meeting ACS guideline criteria for transfer to a level I or II trauma center from a nontertiary center, and, to our knowledge, none has examined specific factors associated with nontransfer in this population.⁸ Understanding factors associated with nontransfer among patients who should be considered for transfer to a facility that can give a higher level of care may provide an opportunity for system progress, informing quality improvement strategies within trauma systems.⁹

Our objective was to identify patient-level and center-level factors associated with nontransfer among patients meeting ACS guideline criteria for consideration of transfer from nontertiary centers in a national population. We hypothesized that a combination of patient-level and center-level factors would be associated with nontransfer from nontertiary centers.

Patients transported to nontertiary trauma centers, defined as level III, IV, and V trauma centers and NTCs, between 2007 and 2012 were identified in the National Trauma Data Bank (NTDB). Centers were excluded if they did not transfer any patient from the emergency department (ED) owing to data reliability concerns.¹⁰ Patients were excluded if they were transferred to the index nontertiary center, died in the ED, or had discordant ED and center disposition. This study was exempted, with need for informed consent waived, by the University of Pittsburgh Institutional Review Board.

Patient-level characteristics were obtained, including demographics, vital signs, mechanism of injury, International Classification of Diseases, Ninth Revision (ICD-9) codes, injury severity, ED disposition, and center disposition. Center-level characteristics were also abstracted, including teaching status, profit status, bed size, and surgeon resources.

Patients were classified as meeting ACS guideline criteria for transfer from nontertiary centers if they had any 1 of the following factors: (1) carotid or vertebral arterial injury, (2) thoracic vascular injury, (3) cardiac injury, (4) major abdominal vascular injury, (5) grade IV or V liver injury, (6) unstable pelvic fracture, (7) penetrating head injury or open fracture of the skull, (8) Glasgow Coma Scale score less than 14 (scores range from 3 to 15, with higher scores representing normal mental status),¹¹ (9) spinal fracture or spinal cord deficit, (10) multiple rib fractures or flail chest, and (11) severe torso injury (Abbreviated Injury Scale score ≥3; scores range from 1 to 6, with higher scores representing more severe anatomic injury)¹² with advanced comorbidities.⁶ These criteria were identified using a combination of ICD-9 diagnosis codes, mechanism of injury, and vital sign data.

For missing data, multiple imputation using a fully conditional specification model was performed to develop 5 imputed data sets. Imputed variables included Glasgow Coma Scale score, Injury Severity Score (scores range from 1 to 75, with higher scores representing more severe anatomic injury),¹³ and admission systolic blood pressure, heart rate, and respiratory rate. Outcome models were performed using estimation techniques that combine model coefficients and SEs from each imputed data set while adjusting for the variability between data sets.¹⁴ Missing data for imputed variables ranged from 1% (Injury Severity Score) to 9% (Glasgow Coma Scale). Sensitivity analysis was performed using complete cases and similar results were obtained. Thus, imputed results are presented herein.

Patients transferred from the ED were defined as the transfer group. Patients discharged from the ED alive or admitted to the center were defined as the nontransfer group. Data analysis was performed from March 17, 2016, to May 20, 2016.

A multilevel logistic regression model was developed to identify factors associated with nontransfer from nontertiary centers. Factors included in analysis were age (children, <18 years; adults, 18-65 years; and elderly individuals, >65 years); sex; race; insurance status; mechanism of injury; admission Glasgow Coma Scale score (≤8, 9-12, and 13-15), systolic blood pressure, heart rate, and respiratory rate; severe head, chest, or abdominal injury (Abbreviated Injury Scale score ≥3); trauma center level; center bed size; profit status; teaching status; US geographic census region; number of trauma intensive care unit (ICU) beds; and number of available trauma, neurosurgical, and orthopedic surgeons. Random effects for centers were used to account for clustering. Furthermore, variance from the multilevel models and a null model with no covariates were used to calculate the proportion of variance in nontransfer among centers owing to patient-level and center-level attributes.

To evaluate center-level characteristics, the proportion of patients transferred out from the total number of patients seen at each center during the study period was determined. Centers were then divided into quartiles based on the proportion of total patients transferred and compared across profit status, teaching status, center bed size, trauma center level, trauma ICU beds, and surgeon resources available.

An exploratory analysis was performed comparing in-hospital mortality in nontransferred patients meeting ACS criteria at nontertiary centers with patients meeting the ACS criteria who were taken directly to a level I or II trauma center. Similar exclusion criteria were applied to patients from level I or II centers and those from nontertiary centers to minimize bias. A multilevel logistic regression model evaluated the association of nontertiary compared with level I or II center presentation with mortality, controlling for age, sex, race, comorbidities, mechanism of injury, insurance, Injury Severity Score, admission vital signs, head Abbreviated Injury Scale score, and trauma mortality prediction model predicted mortality,¹⁵ with a random effect for centers applied to account for clustering. Patients from nontertiary centers taken directly to the operating room were excluded from this analysis to reflect the possibility of unknown injuries until operation and patients being too unstable for transfer.

Continuous data are presented as median (interquartile range) and compared using Mann-Whitney tests. Proportions were compared using χ² tests. A 2-tailed P value <.05 was considered significant. Adjusted odds ratios (AORs) with 95% CIs were obtained from regression models. Model performance was assessed using the C statistic. Data were analyzed using SPSS, version 22 (IBM Corp), and Stata, version 13MP (StataCorp).

Given the evidence of unique needs among geriatric trauma patients,^16,17 the analyses described above were performed separately for patients older than 65 years. Because there may be important differences between lower-level trauma centers and NTCs, subgroup analyses were performed separately for level III, IV, and V trauma centers and for NTCs.

A total of 475 525 patients were initially seen at a nontertiary center during the study period, with 76 143 patients (16.0%) transferred. Of those initially seen at a nontertiary center, 96 528 patients (20.3%) met ACS criteria for consideration of transfer from a nontertiary center (55 611 [57.6%] were male and the median age was 52 years [interquartile range, 28-77 years]), with 19 396 patients (20.1%) undergoing transfer (Figure). Transferred patients were more likely to be children, of nonwhite race, and self-pay and to have severe head or chest injury. They were also associated with centers with fewer beds and surgeons (Table 1).

Elderly patients had a nearly 2-fold increase in the adjusted odds of nontransfer (AOR, 1.70; 95% CI, 1.46-1.98; P < .001) (Table 2). Among injury characteristics, severe thoracic injury and penetrating injury were associated with nontransfer. Government-subsidized and commercial insurance were also associated with nontransfer compared with self-pay patients. Of specific ACS transfer criteria, factors associated with nontransfer included multiple rib fractures or flail chest (AOR, 1.47; 95%, CI, 1.26-1.72; P < .001) and severe torso injury with comorbidities (AOR, 1.76; 95% CI, 1.37-2.27; P < .001). Patient-level factors accounted for 36% of variation in transfer practices between centers.

Among patients with penetrating injuries who were not transferred, 1642 individuals (47.2%) were taken directly to the operating room at the nontertiary center. In this group, 352 patients (21.4%) presented with hypotension. When these patients were excluded from the model, penetrating injury was no longer associated with nontransfer (Table 2), indicating that the association between penetrating injury and nontransfer was due to patients taken directly to the operating room. Thirty-one patients (1.9%) of this subgroup subsequently underwent transfer and 272 patients (16.6%) died; 203 of the nonsurvivors (74.6%) died on the day of admission. Finally, similar patient-level factors were associated with nontransfer when the above analysis was repeated excluding early deaths within 24 hours at nontertiary centers.

Among center-level factors, increasing bed size had incrementally increasing adjusted odds of nontransfer (Table 2). Nontrauma centers were associated with nontransfer compared with level III centers. University-affiliated centers were also strongly associated with nontransfer compared with community centers. Compared with patients from the Northeast, those from the Midwest, South, and particularly the West had higher adjusted odds of nontransfer. More trauma surgeons and neurosurgeons at a center were associated with increased adjusted odds of nontransfer. Center-level factors accounted for 58% of the variation in transfer practices between centers.

When stratifying centers by the proportion of transfer quartile, increasing trauma care resources, including total number of beds, number of ICU beds, and number of trauma surgeons, neurosurgeons, and orthopedic surgeons, correlated with lower proportions of patients being transferred. University-affiliated centers and NTCs were significantly more represented in the lower quartiles of patients transferred (Table 3).

Among patients meeting ACS transfer criteria who were not transferred out of nontertiary centers, in-hospital mortality was 5.6%, compared with 8.7% mortality for patients meeting ACS criteria who were taken initially to level I or II centers. However, the risk-adjusted odds of mortality were significantly higher for patients not transferred from nontertiary centers compared with those treated at a level I or II center (AOR, 1.14; 95% CI, 1.03-1.28; P = .03).

When examining data on elderly patients, the only patient-level factors associated with nontransfer noted were pelvic fracture (AOR, 1.49; 95% CI, 1.09-2.03; P = .01) and severe torso injury with major comorbidity (AOR, 2.63; 95% CI, 1.58-4.40; P < .001); severe thoracic injury and penetrating injury were no longer associated with nontransfer. A strong interaction was also found between elderly age and government-subsidized insurance (P < .001).

When examining nontertiary trauma centers (levels III, IV, and V), the same set of patient-level and center-level factors associated with nontransfer emerged; however, mortality for nontransferred patients at nontertiary trauma centers was not significantly different compared with the mortality for patients initially taken to a level I or II center (AOR, 1.18; 95% CI, 0.94-1.41; P = .07). Patient-level factors accounted for 44% and center-level factors accounted for 55% of variation in transfer practices between nontertiary trauma centers.

When data on NTCs were examined, again, similar patient-level factors were associated with nontransfer. At the center level, for-profit status was associated with nontransfer (AOR, 1.55; 95% CI, 1.39-1.74; P < .001). Patient-level factors accounted for only 13% and center-level factors accounted for 55% of the variation in transfer practices between NTCs. University affiliation was most strongly associated with nontransfer (AOR, 8.42; 95% CI, 7.12-9.96; P < .001). Because of this finding, NTCs without and with university affiliation were further examined separately. At NTCs without university affiliation, similar patient-level and center-level factors were again associated with nontransfer, including for-profit status. For NTCs with university affiliation, the only patient-level factors associated with nontransfer were government-subsidized and commercial insurance; however, for-profit status, larger bed size, and geographic region continued to be associated with nontransfer. The odds of mortality were significantly greater for patients not transferred from NTCs without university affiliation compared with patients at level I or II centers (AOR, 1.24; 95% CI, 1.04-1.48, P = .02); however, increased odds were not seen at NTCs with university affiliation (AOR, 0.94; 95% CI, 0.71-1.24, P = .64).

The present study demonstrates that only 1 in 5 patients meeting ACS guideline criteria for consideration of transfer from a nontertiary center is transferred. Several patient-level factors were associated with nontransfer, including age, injury mechanism and patterns, and insurance status. Increasing center resources and rurality were also associated with nontransfer in these groups. Center-level factors account for more variation in transfer practices than patient-level factors. Exploratory analysis demonstrated higher odds of mortality among patients not undergoing transfer from a nontertiary center compared with patients initially arriving at a level I or II center that meets ACS transfer criteria. Subgroup analysis suggests that this increased mortality was primarily driven by NTCs without university affiliation.

There are several possible explanations for the high rate of nontransfer among patients meeting ACS transfer criteria. The first explanation is that high-risk injuries were not recognized. This is common among elderly patients leading to undertriage.¹⁸ Lane et al¹⁸ found that one-third of severely injured elderly patients were treated in NTCs without undergoing transfer to a trauma center. Others have found a similar trend.^19-21

Penetrating injury was an unexpected factor of nontransfer. However, some injuries constituting transfer criteria may not have been recognized until laparotomy, since nearly half of the patients with penetrating injuries were taken from the ED to the operating room at the nontertiary center.

Another potential explanation for the high rate of nontransfer among patients meeting ACS transfer criteria is that, despite recognition of injuries, nontertiary centers believed that patients could be adequately cared for within their capabilities. Chest injuries were a factor associated with nontransfer and may be attributable to the infrequent need for operative intervention, leading some centers to believe that these injuries can be managed successfully without transfer. Centers with greater resources were also less likely to transfer patients, even when meeting ACS transfer criteria. In Canada, severely injured patients who received initial care in resource-rich NTCs were less likely to be transferred to a trauma center.²² Others have reported comparable associations between availability of center resources, including ICU beds and surgical subspecialists.⁴ Nontrauma center status may be associated with nontransfer since they are less integrated into trauma systems. Nontrauma centers are not reviewed by an external accrediting body and have no obligation to transfer patients; however, lower-level trauma centers undergo such review and are expected to transfer patients whose care would exceed their capabilities within the trauma system. This finding may partially explain why patient-level factors were less important at NTCs but more important at lower-level trauma centers. These results suggest that the risks of certain injury patterns may be underappreciated and that nontertiary centers with greater resources may elect to care for these patients.

More rural regions were less likely to transfer patients compared with the more densely populated Northeast. The Northeast is also more densely populated with level I or II trauma centers, while the Midwestern and Western regions rely on level III, IV, and V trauma centers for initial care.²³ The logistics and time to transfer a patient may lead some centers to conclude that patients are better served staying at the nontertiary center. Newgard et al²¹ found that rurality and distance to the closest tertiary center were inversely associated with transfer to a higher-level trauma center.

Patients with commercial insurance were also less likely to be transferred. Furthermore, for-profit status was associated with nontransfer in NTCs. Others have reported that insured patients are less likely to be transferred from NTCs.^5,24,25 This phenomenon deserves further study since it is conceivable that centers may be keeping patients with favorable insurance owing to reimbursement incentives. Although the present study cannot elucidate the underlying factors, this issue is important to trauma systems and requires investigation.

Finally, another consideration is patient preference in transfer decisions. Finlayson and colleagues²⁶ reported that nearly half of patients would accept a doubling of mortality risk for high-risk surgery to receive care at a local hospital rather than a regional center. Older age was associated with higher risk tolerance to remain local. Trauma patients may also be willing to accept a higher risk of poor outcome to remain at a local center.

An alternative interpretation that must be considered is whether the ACS criteria for consideration of transfer accurately identify patients who should undergo transfer. The overall mortality was only 5.6% when patients were not transferred, and increased mortality compared with tertiary centers primarily appears to occur at NTCs without university affiliation. Although the ACS notes that the published criteria are not inclusive, there is little evidence supporting these criteria for identifying patients who have better outcomes when transferred. Thus, these results support further research to systematically evaluate specific criteria to identify patients who require transfer from nontertiary centers.

These findings have implications for trauma system process improvement. At the patient level, the factors identified herein may allow educational initiatives to focus on early consideration of transfer for specific high-risk groups. Within individual systems, regional trauma center leaders may be able to identify centers with greater resources in an attempt to reinforce criteria that should prompt transfer to a tertiary trauma center, refine formal transfer agreements, as well as invite them to participate as partners in the system-wide quality assurance process.⁹

This analysis has several limitations. The first are those associated with a retrospective design. The second are those of the NTDB²⁷; some data were missing; however, multiple imputation was used to mitigate this and has been validated in the NTDB previously.²⁸ The NTDB offers a large number of patients in a national sample; however, there are a limited number of variables for analysis. We were able to identify 11 of 14 ACS transfer criteria, but the remaining 3 criteria could not be reliably identified using the data set. The ACS criteria are guidelines for considering transfer and do not mandate transfer. It was not possible to determine whether transfer had been considered among patients who ultimately were not transferred. Trauma systems may alter transfer criteria based on system resources; however, we were unable to evaluate regional variation using the NTDB. We excluded ED deaths since these patients were not “at risk” for transfer; however, some may have had poor-quality ED care and died as a result, introducing bias. Despite this, patients excluded represent a small proportion, with 25% dead on arrival and a median ED stay of 27 minutes, making this bias less likely to influence our results.

Among transferred patients, we cannot determine the destination center or subsequent outcomes. Transfer timing was also unavailable. Thus, we cannot determine the effect of transfer on outcomes. Furthermore, it is unclear how accurate or complete the diagnosis of injuries was for patients undergoing transfer. Some patients may not have had all injuries diagnosed, leading to inaccurate injury severity assessment; however, the mean number of ICD-9 codes per patient was similar between the transferred and nontransferred groups (3.6 vs 3.4). Finally, withdrawal of care was not specified, which may have influenced nontransfer, particularly in elderly patients.

Only 1 of 5 of patients meeting ACS guideline criteria for consideration of transfer from nontertiary centers underwent transfer. These patients may have worse outcomes compared with patients initially taken to a level I or II center, particularly at non–university-affiliated NTCs. Several patient- and center-level factors associated with nontransfer were identified, and center-level factors appear to be the primary driver of variation in transfer practices. These factors may be useful for trauma system leaders and stakeholders to target educational outreach to guide development of more-inclusive trauma systems. Further study is necessary to critically evaluate whether these ACS criteria identify patients who have worse outcomes without transfer and to understand patient values in the location of their care.

Corresponding Author: Joshua B. Brown, MD, MSc, Division of General Surgery and Trauma, Department of Surgery, University of Pittsburgh Medical Center, 200 Lothrop St, Pittsburgh, PA 15213 (brownjb@upmc.edu).

Accepted for Publication: October 4, 2016.

Published Online: January 4, 2017. doi:10.1001/jamasurg.2016.4976

Author Contributions: Dr Brown had full access to all 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: Zhou, Rosengart, Peitzman, Sperry, Brown.

Acquisition, analysis, or interpretation of data: Zhou, Billiar, Sperry, Brown.

Drafting of the manuscript: Zhou, Rosengart, Brown.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Zhou, Rosengart, Brown.

Administrative, technical, or material support: Peitzman, Brown.

Study supervision: Rosengart, Billiar, Sperry, Brown.

Conflict of Interest Disclosures: None reported.

Funding/Support: Dr Brown receives support from institutional T-32 Ruth L. Kischstein National Research Service Award training grant 5-T32-GM-008516-20 from the National Institutes of Health.

Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: Committee on Trauma, American College of Surgeons. National Trauma Data Bank NTDB 2007-2012, Chicago, IL. The content reproduced from the NTDB remains the full and exclusive copyrighted property of the American College of Surgeons. The American College of Surgeons is not responsible for any claims arising from works based on the original data, text, tables, or figures.