Survival Benefit of Treatment at or Transfer to a Tertiary Trauma Center among Injured Older Adults

Abstract

Objective:

It is well established that seriously injured older adults are under-triaged to tertiary trauma centers. However, the survival benefit of tertiary trauma centers (TC) compared to a non-tertiary trauma centers (Non-TCs) remains unclear for this patient population. Using improved methodology and a larger sample, we hypothesized that there was a difference in hospital mortality between injured older adults treated at TCs and those treated at Non-TCs.

Methods:

This was a retrospective cohort study of injured older adults (>=55 years) reported to the Oklahoma statewide trauma registry between 2005 and 2014. The outcome of interest was 30-day in-hospital mortality and the exposure variable of interest was level of definitive trauma care (TC vs Non-TC). Overall survival benefit of treatment at a TC as well as the survival benefit of transferring injured older adults to a TC were evaluated using multivariable survival analyses as well as propensity score-adjusted analyses.

Results:

Of the 25,288 patients eligible for analysis, 43% (10927) were treated at TCs. Multivariable Cox regression analyses revealed effect modification by age group and time. After adjusting for potential confounders within the age strata, overall, patients treated at TCs were significantly less likely to die within 7 days of admission and this effect was stronger for patients aged 55 – 64 years (HR 0.41, 95%CI 0.31–0.52) compared to those >=65 years (HR 0.62, 95% CI 0.55–0.70). Overall survival benefit of TCs beyond 7 days was also observed (HR 0.68, 95% CI 0.56–0.83). Similarly, for the survival benefit of transferring injured older adults, after adjusting for the propensity to be transferred and other confounders, transfer to a TC was associated with lower 30-day mortality both for patients less than 65 years old (HR 0.36, 95%CI: 0.27–0.49) and those 65 years and older (HR 0.55, 95% CI: 0.48–0.64).

Conclusions:

Our results suggest a survival benefit for injured older adults treated at TCs. This benefit was also observed for patients transferred from non-tertiary trauma centers. Further research should focus on identifying specific subgroups of patients who would especially benefit from this level of care to minimize trauma triage inefficiencies.

Introduction

The public health significance of traumatic injury in older adults cannot be overemphasized. Older adult trauma patients account for 25% of injury fatalities per year and consume 33% of health care resources spent on trauma.^1,2 A recent study also showed that, of over 6 million EMS responses in US older adults in 2014, trauma accounted for 11% of those transports, second only to cardiovascular emergencies.³ Older adults respond differently to traumatic injury in part due age-related physiologic changes, preexisting comorbidity and medication use.⁴ Age-specific mortality in patients with severe traumatic injury begins to increase as early as age of forty years.⁴ The risk of mortality following trauma is on average, four times greater in older adult patients compared to their younger counterparts.^1,5,6 Level I and II trauma centers (tertiary trauma centers) offer the most comprehensive trauma care and have the resources to manage seriously injured patients. Despite the observed increased risk of mortality following trauma, a number of studies have shown that compared to their younger counterparts, older adults 55 years and older are under-triaged to designated tertiary trauma centers (TCs) from the scene of injury.^7–14 More recent evidence also suggests injured older patients initially transported to Level III or IV (non-tertiary) trauma centers in a predominantly rural region are less likely to be transferred to TCs.⁷

Research on the benefit of TCs among seriously injured older adults has been mixed. Findings from prior studies have ranged from reporting a survival benefit^15–17, to no survival benefit^18,19 and increased mortality.²⁰ Whether injured older adults should be cared for in TCs remains an open question. Generalizability of previous studies is limited by regional differences, different inclusion criteria, variation in mortality endpoints and variation in the definition of ‘older adult’ and ‘tertiary trauma centers’. Previous studies have also been limited by modest sample sizes and number of adverse events,^18,19, plus residual confounding and lack of comorbidity data.^18,20 In addition, no prior study in this population accounted for the timing of death in their analysis.

In this study, we sought to determine the difference in 30-day in-hospital mortality between moderate to seriously injured older adults treated at TCs and those treated at non-tertiary trauma centers (Non-TCs). Secondarily, we evaluated whether inter-facility transfer of injured older adults to tertiary trauma centers conferred a survival advantage compared to similar patients who remained at non-tertiary trauma centers. We hypothesized that severely injured older adult patients treated at tertiary trauma centers (both direct from the scene and via inter-hospital transfer) would have better survival compared to patients treated at non-tertiary trauma centers.

Patients and Methods

Study Design and Setting

Statewide trauma registry data reported between 2005 and 2014 were utilized for the study using a retrospective cohort study design. We designed the study to account for several of the limitations of previous research including consideration of additional confounders and integrating propensity score analyses to reduce bias, conducting time-to-event analyses, and utilization of a large number of diverse hospitals and a larger sample size than previous studies. As of December 2014 in Oklahoma, there was one (1) Level I hospital, 2 Level II hospitals, 26 Level III hospitals and 80 Level IV hospitals (primarily located in rural areas). The only Level I TC is located centrally (Oklahoma City) and is verified by the American College of Surgeons Committee on Trauma (ACS-COT), while the two Level II TCs (of which one is ACS verified) are located in the northeastern (Tulsa) part of the state. We considered Level I and II hospitals to be major, tertiary trauma centers, consistent with the ACSCOT definition.²¹ During the study period, six (6) Non-TCs were re-designated from Level III to IV and 1 Non-TC was redesignated from Level IV to III; there were no changes in the number of tertiary trauma centers. Oklahoma has a mandatory, inclusive trauma system, where all state-licensed hospitals designated as trauma centers are required to take part in trauma system development and report data to the state trauma registry. Patients meeting the Oklahoma major trauma definition (Table 1) and all inter-facility transfers are reported on a monthly basis. Triage guidelines play a critical role in ensuring that seriously injured patients are transported to the appropriate level of trauma care. Oklahoma’s prehospital triage guidelines are based on the CDC’s Field Triage guidelines²² and are periodically revised in line with the CDC’s recommendations. No significant changes to the guidelines were initiated during the study period. Both the Oklahoma State Department of Health (OSDH) and the University of Oklahoma Health Sciences Center Institutional Review Boards approved this study with waiver of informed consent.

Table 1.

Oklahoma’s Major Trauma Definition

Inclusion Criteria (Must meet at least one criteria in each category)

Exclusion Criteria

International Classification of Disease 9th Revision Clinical Modification code of 800.00–959.9 Abbreviated Injury Scale value of 3 or higher; or Injury Severity Score (ISS) of 9 or higher; or TRISS Survival Probability less than *.50; or Dead on arrival or died while in hospital Length of hospital stay > 48 hours; or Transferred from a lower level to a higher level trauma center with major trauma; or Dead on arrival or died while in hospital Admitted to intensive care unit; or Admitted to operating room for major surgery (head, chest, abdomen, vascular).

Persons who died at the scene, or Any of the following as the sole type of injury Overexertion injuries; Submersions; Poisonings; Asphyxiation; Injuries caused by a preexisting condition (e.g., osteoporosis, etc).

Study Population

The study included patients aged 55 years and older with at least an injury severity score (ISS) of 9. There is variability in the age threshold used to define older adults. The age cut-point of 55 years was chosen primarily because 1) age-specific mortality in major trauma patients begins to increase as early as age of 40 years;⁴ 2) studies have shown that the problem of under-triage start as early as 50 years¹⁰ and; 3) the current national (CDC) triage guidelines also use an age cut-off of 55 years and older to define older adults.²² To minimize measurement error and improve data completeness and accuracy, records for patients transferred between hospitals were linked using SAS (Cary, NC) based LinkPro 3.01 (InfoSoft Inc., Winnipeg, MB Canada). Multiple iterations of deterministic and probabilistic linkage attempts were made using last name, first name, date of birth, arrival date, gender, age, and Soundex coded last name as linking variables. A majority (88%) of transfer records were successfully linked. Between 2005 and 2014, 120683 unique patients were reported to the Oklahoma State Trauma Registry (OSTR). Of these, 25288 met study eligibility (Figure 1). Patients were excluded from the study if they were 54 years or younger (n=66769), had burn related injuries (n=1880), were transferred to or from an out-of-state hospital (n=15581), had an injury severity score (ISS) of less than 9 and were discharged alive (n=9784), or were discharged within two-hours of emergency department (ED) arrival from scene of injury (n=587). Burns were excluded because these injuries generally havea different prognosis compared to blunt/penetrating injuries and a majority are preferentially treated at specialized designated ‘Burn Centers’ not trauma centers. Deaths occurring within 2 hours were excluded because the study included transfers and the probability of transfer was lowest during the first hour of arrival, thus these patients were likely not at ‘at risk’ of being transferred even if the initial facility had planned to transfer them for higher level of trauma care.

Figure 1.

Study Population

Variables

The outcome of interest was 30-day in-hospital mortality as reported in the statewide trauma registry. The exposure variable of interest was whether a patient received definitive trauma care at a designated TC or Non-TC. We defined definitive trauma care as the final but not necessarily optimal trauma care received by the patient without the need to transfer the patient for further care. In addition to the primary exposure variable (TC vs TC Non-TC), we considered multiple confounders, predictors and effect modifiers for all multivariable outcome models: injury year, gender, race, injury etiology (traffic-related, stabbing/gunshot wound, falls, other) and blunt vs penetrating mechanism, mode of arrival to initial hospital (emergency medical service [EMS] vs non-EMS), inter-facility transfer status. Clinical factors considered included initial ED vital signs (GCS: Glasgow Coma Scale, SBP: systolic blood pressure, HR: heart rate and RR: respiratory rate), EMS/ED intubation, Injury Severity Score (ISS), serious injury by body region (Abbreviated Injury Scale score [AIS] >=3), specific injuries based on all available International Classification of Diseases, Ninth Revision (ICD-9) codes (flail chest, long bone fracture, skull fracture, pelvic fracture, wrist/ankle amputation and solid organ injuries), and comorbid conditions (pulmonary, cardiac, diabetes, coagulopathy, cancer, and liver/kidney disease). We considered the presence of shock using two definitions: SBP < 90 mm Hg and SBP < 110 mm Hg.²³

Geospatial Analysis

Several measures of distance were calculated for each patient and these included distance from the scene of injury to a TC, distance to the closest Level III trauma center, and distance to the closest facility regardless of trauma level. Data sources for geocoding services included 2010 Census ZIP Code Tabulation Areas (ZCTAs), 2010 Census Counties, 2010 Census Places and 2014 ESRI U.S. ZIP Code Points. Records with Post Office Box (POB) zip codes were assigned randomly to zip code areas within 5 miles of the delivery post office address. Of all patient records, 93.3% were successfully geocoded including 75.7% geocoded based on injury zip code, 10.8% based on injury county and 6.8% based injury city. Patient injury points were located to the population weighted centroid of injury zip codes, counties or cities. ArcGIS Network Analyst extension which uses actual road networks, was used to determine distance variables of interest based on the USA Contiguous Equidistant Conic coordinate projection system (minimizes distortions while calculating distance between two points).

Statistical Analyses

Handling of missing data

While inter-facility data linkage improved completeness of our data, about 10% of eligible patients still had missing covariate information on initial ED vital signs with 5.6% (1461) missing initial ED GCS, 3.5% missing RR or HR, and 0.9% (237) missing initial ED SBP. To avoid potential bias resulting from complete case analysis, missing data on initial ED vital signs were multiply imputed using a conditional multivariable sequential regression approach (PROC MI, FCS option).^24,25

Overall survival benefit

Unadjusted comparisons between patients treated at TCs and those treated at Non-TCs were performed using the independent Student’s t test or Mann-Whitney Wilcoxon test for continuous variables, and for categorical variables, the chi-square and Fisher’s Exact tests were used. Survival analyses were performed using multivariable Cox proportional hazards regression models to compare 30-day mortality outcome differences between TC and Non-TC patients. Survival time was defined as the time elapsed between time of initial ED arrival and time of death or discharge. The survival time was censored at the date of hospital discharge or at 30 days for patients who were still hospitalized and alive at 30 days (follow-up period). The proportional hazards assumption was evaluated graphically and by including an interaction variable between survival time (in days) and level of definitive trauma care. For the overall survival benefit, the inclusion of transfers in our analysis precluded us from considering propensity score adjustment since transferred patients would have additional clinical information not available for other patients, an imbalance that would affect our propensity score model and potentially bias our results. For multivariable modeling, key predictors/confounders were included a priori, regardless of statistical significance. First-order statistical interactions were considered significant if they met a significance level of p < 0.05. We determined a priori to evaluate statistical interaction between age group (<65, >=65 years) and trauma center level (TC vs Non-TC). We considered variables to be important confounders/predictors if their inclusion or exclusion resulted in at least a 20% change in the estimate of trauma center effect. Since we included in our patient selection those patients with an ISS of 9 or greater (Table 1) and ISS >= 16 is the threshold used by ACSCOT to define under- and overtriage,²¹ we conducted sensitivity analyses to evaluate whether our estimates would be qualitatively attenuated or strengthened if we restricted our analyses to this subset of patients.

Survival benefit of transfer

Because patients are not randomly selected for transfer a propensity-adjusted survival analysis was used to evaluate the survival benefit of transfer to a TC. The propensity score was used to adjust for potential indication (selection) bias. The final propensity score model (Area Under the Curve [AUC] 0.78, 95%CI: 0.77–0.79) included the following variables: gender; age group (< 65, >=65); race; mechanism of injury; prehospital EMS transport; skull fracture/intracranial injury; internal organ injury (liver, spleen, kidney); rib fracture; pelvic and long bone fractures; EMS/ED intubation; initial ED SBP; primary payer; and injury year. Propensity-adjusted regression analysis of each covariate on transfer status was used to assess the ability of the propensity scores to create comparable groups (i.e., admitted vs. transferred). Adjusting for the propensity to be transferred removed all differences in covariates with the exception of age, presence of a serious head injury and overall injury severity (ISS). These covariates were considered for further adjustment in the subsequent outcome model. A Cox regression survival analysis evaluating 30-day mortality differences between the exposure groups was performed; the model included propensity score quintile and additional potential confounders (serious head injury, final ISS, age and presence of a penetrating injury). A graphical evaluation of propensity score distribution indicated adequate overlap across the range of propensity scores quintiles except in the first quintile where there was a predominance of patients treated at Non-TCs (25% vs 3.2%). We therefore explored “holding out” patients in this non-overlapping stratum to determine the effect on the adjusted measure of association between trauma center status and mortality.

Because several surrogate measures of injury acuity were considered for multivariable analysis, we evaluated the correlation between coefficients of variables retained in final models. In addition, a crude evaluation of multicollinearity among covariates retained in final models was performed using the multiple linear regression PROC REG SAS procedure (Eigen-value analysis). Univariate, survival, and propensity-adjusted analyses were performed using SAS Version 9.4 (SAS 9.4, SAS Institute, Cary, NC)

Results

Of the 25288 patients eligible for analysis, 57% (14361) were treated at non-tertiary trauma centers while 43% (10927) were treated at tertiary trauma centers. Table 2 summarizes patient demographics, injury etiology and distance measures by level of definitive trauma care. Compared to patients treated at Non-TCs, patients treated at TCs were notably (p<0.05) younger, predominantly male, had disproportionately higher number of traffic-related and penetrating injuries, were slightly more likely to be transported by EMS from scene of injury, and were injured closer to a tertiary or Level III trauma center. Notably for patients treated at Non-TCs, there was a disproportionately higher number of patients aged 65 years and older (82% vs 64%) and a majority of the injuries (82%) were fall-related.

Table 2.

Demographic and Injury Characteristics by Level of Definitive Trauma Treatment, Injured Older Patients (>= 55 Years), Oklahoma State Trauma Registry, 2005–2014.

Variable	TC (I/II) (n=10927)	NTC (III/IV) (n=14361)	p-value
Injury Year, n (%)			<.0001
2005–2006	1537 (14.7)	2165 (15.1)
2007–2008	1913 (17.6)	2820 (19.7)
2009–2010	1918(17.6)	3014 (21.0)
2011–2012	2447(22.5)	2977(20.8)
2013–2014	3084(28.3)	3365 (23.5)
Mean Age in years (±SD)	71.2 (11.5)	76.8 (11.2)	<.0001
Age >=65 years, n (%)	6977 (63.9)	11747 (81.8)	<.0001
Male, n (%)	5990(54.8)	5687(39.6)	<.0001
Race, n (%)
White	9736 (89.1)	13102 (91.2)	<.0001
Black	506 (4.6)	354 (2.5)
AI/AN/PI	342 (3.1)	415 (2.9)
Asian	66 (0.6)	64 (0.5)
Other	277 (2.5)	426 (3.0)
Penetrating Injury, n (%)	297 (2.7)	141 (1.0)	<.0001
MOI, n (%)
Traffic-related	3726 (34.1)	1444 (10.1)	<0.0001
GSW/Stabbing	261 (2.4)	100 (0.7)
Falls	5915 (54)	11802 (82.2)
Other	1025 (9.4)	1015(7.1)
Pre-hospital EMS, n (%)	8810(80.6)	10397(72.4)	<.0001
Prehospital Mode of Transport, n (%)			<.0001
POV/Other	2117 (19.4)	3964 (27.6)
Land Ambulance	7882 (72.1)	10290 (71.7)
Air Ambulance	928 (8.5)	107 (0.8)
Transferred in, n (%)	4263 (39)	4630 (32)	<0.0001
Mean distance to closest (±SD), miles
Any Level trauma facility	6.6 (5.4)	6.5 (5.9)	0.2205
Level III trauma facility	17.2 (19.9)	19.6 (23.5)	<.0001
Level I/II trauma facility	34.0 (38.3)	53.1 (44.6)	<.0001
Mean Distance to Definitive Care Facility (±SD)	35.9 (39.5)	26 (31.9)	<.0001

% = column % unless otherwise specified; SD = Standard Deviation; AI=American Indian; AN=Alaskan Native; PI=Pacific Islander; MOI=Mechanism of Injury; GSW=Gunshot Wound; POV=Privately Owned Vehicle

Table 3 summarizes patients’ clinical characteristics and crude outcomes by level of definitive trauma care. Compared to patients treated at Non-TCs, patients treated at TCs were more likely (p<0.05) to present with hypovolemic shock, had lower GCS scores, were likely to be intubated, and had higher overall ISS; the proportion of patients with an ISS >=16 was notably higher for patients treated at TCs (43% vs 18.6%, p<0001). As for specific injuries, there was a higher incidence of serious (AIS >=3) head, chest and abdominal injuries among patients treated at TCs while a higher incidence of serious extremity injuries was noted for patients treated at Non-TCs; patients with long bone fractures (humerus or femur) were more likely to be treated at Non-TCs while patients with pelvic fractures were more likely to be treated at TCs. Albeit statistically significant, patients treated at Non-TCs had only a slightly higher prevalence of pre-existing conditions (44.7% vs 42.3%) and a notably, somewhat higher prevalence of pre-existing cardiac disease (29.1% vs 25.6%). On average, patients treated at TCs had a longer hospital length of stay and a higher proportion of unadjusted overall mortality compared to those treated at Non-TCs (10.2 vs 7.5%, p<0.0001). For both groups, a majority (>70%) of the deaths occurred within 7 days of admission.

Table 3.

Clinical Characteristics and Outcomes by Level of Definitive Trauma Treatment, Injured Older Patients (>= 55 Years), Oklahoma State Trauma Registry, 2005–2014.

	TC (I/II) (n=10927)	NTC (III/IV) (n=14361)	p-value
Variable
Initial ED, Mean (±SD)
SBP	143.3 (32.3)	147.4(31.8)	<.0001
GCS	13.8 (3.0)	14.4 (2.1)	<.0001
HR	84.9 (19.4)	82.2(18.4)	<.0001
RR	18.3 (5.9)	18.8 (4.0)	<.0001
Shock (SBP <110 mm Hg), n (%)	1431 (13.1)	1230 (8.6)	<.0001
Shock (SBP < 90 mm Hg), n (%)	501 (4.6)	338 (2.4)	<.0001
TBI (GCS <9), n (%)	824(7.7)	427(3.3)	<.0001
EMS/ED Intubation, n (%)	1024(9.4)	315(2.2)	<.0001
Mean ISS (±SD)	16.0 (8.9)	11.5 (5.2)	<.0001
ISS>=16, n(%)	4707 (43.1)	2668 (18.6)	<.0001
Serious Injury (AIS≥3), n (%)
Head	4768(43.6)	3691(25.7)	<.0001
Chest	3138 (28.7)	1578(11.0)	<.0001
Abdomen	819(7.5)	249(1.7)	<.0001
Extremity	3535(32.4)	8737(60.8)	<.0001
Anatomic Injury (ICD-9), n (%)
Skull fracture/Intracranial	5210 (47.7)	3786 (26.4)	<.0001
Rib Fracture	3175 (29.1)	1527 (10.6)	<0.0001
Heart/Lung	1281 (11.7)	472 (3.3)	<0.0001
Long bone fracture (hum/fern)	2413 (22)	8084 (56.3)	<0.0001
Pelvic	1156 (10.6)	519 (3.6)	<.0001
Spleen	373 (3.4)	109 (0.8)	<0.0001
Liver	264 (2.4)	49 (0.3)	<.0001
Kidney	151 (1.4)	25 (0.2)	<.0001
Other Pelvic/Abdominal organs	227 (2.1)	53 (0.4)	<.0001
Pre-existing Comorbidity, n (%)	4617 (42.3)	6412 (44.7)	0.0001
Pulmonary Disease, n (%)	998 (9.1)	1553 (10.8)	<.0001
Cardiac Disease, n (%)	2798 (25.6)	4179 (29.1)	<.0001
Diabetes, n (%)	1942 (17.8)	2442 (17.0)	0.1100
Coagulopathy, n (%)	400 (3.7)	288 (2.0)	<.0001
Cancer, n (%)	96 (0.9)	168 (1.2)	0.0240
Liver/Kidney, n (%)	441 (4.0)	782 (5.5)	<.0001
Mean Hospital LOS (±SD)	7.6 (7.2)	5.8 (5.6)	<.0001
Overall In-hospital Mortality, n (%)	1118(10.2)	1081(7.5)	<.0001
Time-specific mortality (days), n (%)			0.0186
1–7	805 (72.0)	817 (75.6)
8–14	176 (15.7)	169 (15.6)
15–30	101 (9.0)	79 (7.3)
>30	36 (3.2)	16 (1.5)

% = column % unless otherwise specified; SD = Standard Deviation; TBI = Traumatic Brain Injury; LOS = Length of Stay

Multivariable Analyses

Overall Survival Benefit

An evaluation of the proportional hazards assumption graphically suggested a violation of this assumption and as such, stratified survival analyses (<= 7 and 7–30 days) were performed. Within these models, statistical interaction, determined a priori between age group (< 65 and >=65 years) and level of definitive trauma care (TC vs Non-TC) was evaluated; there was a significant interaction between age group and level of trauma care for deaths occurring within 7 days of admission (p=0.0021) but not for deaths occurring after 7 days of admission (p > 0.05).Table 4 summarizes results from the stratified survival analyses. Overall, the 7-day survival analyses shows that patients treated at TCs were significantly less likely to die after adjusting for potential confounders and that this effect was much stronger for patients aged less than 65 years (HR 0.45, 95%CI 0.36–0.56) compared to those 65 years and older (HR 0.65, 95% CI 0.58–0.73). The protective effect of treatment at a TC was also observed for patients surviving beyond 7 days (HR 0.69, 95% CI 0.56–0.83); patients treated at Non-TCs had consistently higher 30-day cumulative risk of death (Figure 2). Sensitivity analyses restricting analyses to patients with an ISS >=16 marginally strengthened the TC protective effect for all groups (results not shown).

Table 4.

Multivariable Analysis of the Impact of Treatment at a Tertiary Trauma Center on Mortality among Injured Older Adults

Variable	7-day Mortality 55–64 years HR (95% CI) n=6564	7-day Mortality >=65 years HR (95% CI) n=18724	*8–30 day Mortality HR (95% CI) n=6704
Tertiary TC	0.45 (0.36–0.56)	0.65 (0.58–0.73)	0.69 (0.63–0.76)
Non-tertiary TC	ref	ref	ref
Male Gender	1.1 (0.89–1.40)	1.19 (1.08–1.31)	⫞
Age >=65 years			2.74 (2.15–3.48)
Pre-existing condition	⫞	1.28(1.16–1.41)	⫞
Mechanism of Injury
Traffic-related	ref	ref	ref
Stabbing/GSW	3.45 (2.50–4.75)	3.43 (2.54–4.66)	1.72 (0.90–3.30)
Falls	1.02 (0.80–1.31)	1.43 (1.25–1.64)	1.81 (1.43–2.29)
Other	0.84 (0.60–1.16)	0.94 (0.74–1.20)	1.24 (0.85–1.81)
Prehospital Transport			⫞
POV/Other	ref	ref	⫞
Air EMS	1.53 (0.95–2.44)	1.20 (0.93–1.55)
Ground EMS	2.12(1.45–3.10)	1.57 (1.37–1.80)
Initial ED SBP	0.99 (0.988–0.994)	0.996(0.995–0.998)	⫞
ED/EMS Intubation	4.37 (3.48–5.49)	3.47 (3.01–3.96)	⫞
Serious Head Injury	2.20 (1.73–2.80)	1.34 (1.19–1.52)	1.36 (1.12–1.65)
Long Bone Fracture	0.73 (0.59–0.97)	0.56 (0.49–0.65)	⫞
ISS	1.04 (1.03–1.05)	1.06 (1.05–1.06)	1.03 (1.02–1.04)
Injury Period			⫞
2005–2006	Ref
2007–2008	1.03 (0.75–1.42)	1.10 (0.93–1.30)
2009–2010	0.97 (0.67–1.47	1.27 (1.08–1.49)
2011–2012	1.01 (0.73–1.41)	1.20 (1.02–1.41)
2013–2014	1.51 (1.12–2.05)	1.39 (1.18–1.64)

HR=Hazard Ratio; ˧ = dropped from model, p >0.05; GSW= Gunshot Wound; POV=Privately Owned Vehicle

^*No interaction (p=0.3323) between age group (< 64 and >=65 years) and level of definitive trauma treatment (TC vs NTC), hence no stratification by age group.

Figure 2. *Adjusted thirty-day (8–30 days) cumulative risk of death by level of definitive trauma care and survival time (days).

*Adjusted for gender, age group, pre-existing comorbid conditions, mechanism of injury, prehospital EMS transport, initial ED SBP, EMS/ED intubation, serious head injury, long bone fracture, ISS, and injury period.

Survival Benefit of Transfer

A total of 18,643 study eligible patients were initially transported to non-tertiary trauma centers. Of these, 4,305 (23.1%) were transferred to a tertiary trauma center within twenty-four hours. The crude mortality was significantly higher for transferred patients (9.2% vs 7.5%, p=0.0005). Survival analyses were performed to evaluate the adjusted association between transfer status and 30-day mortality. Overall, patients treated at Non-TCs had a consistently higher 30-day cumulative risk of death and the proportional hazards assumption was not violated (Figure 3). Statistical interaction between age group and level of definitive trauma treatment (p<0.0001) was however noted. Table 5 summarizes the adjusted stratified results by age group (< 65 and >=65 years). After adjusting for the propensity to be transferred, ISS, presence of a serious head injury, and presence of a penetrating injury, transfer to a tertiary trauma center (Level I or II) was associated with significantly lower 30-day mortality both for patients less than 65 years old (HR 0.36, 95%CI: 0.27–0.49) and those 65 years and older (HR 0.55, 95%CI: 0.48–0.64). Sensitivity analysis ‘holding out’ patients in the first propensity score quintile accentuated the protective effect of treatment at a TC on mortality (see Table 5).

Figure 3. *Adjusted thirty-day cumulative risk of death by transfer status and survival time (days).

*Adjusted for propensity score, serious head injury, penetrating injury and overall injury severity.

Propensity score model (AUC 0.78, 95%CI: 0.77–0.79) included the following variables: gender; age group (< 65, >=65); race; mechanism of injury; prehospital EMS transport; skull fracture/intracranial injury; internal organ injury (liver, spleen, kidney); rib fracture; pelvic and long bone fractures; EMS/ED intubation; initial ED SBP; primary payer; and injury period.

Table 5.

^*Propensity-adjusted survival analyses of the association between transfer status and 30-day mortality among patients initially presenting to non-tertiary trauma centers

Variable	< 65 years HR (95% CI) n=3970	>=65 years HR (95% CI) n=14673
Transferred to TC	0.39 (0.29–0.53)	0.59 (0.51–0.68)
Transferred to TCΦ	0.37 (0.27–0.48)	0.52 (0.46–0.59)
Treated at NTC	ref	ref
Propensity score quintile*	1.40 (1.18–1.67)	1.24 (1.17–1.31)
Serious head injury (AIS>=3)	1.81 (1.34–2.45)	1.40 (1.22–1.61)
Penetrating Injury	7.68 (5.12–11.54)	3.80 (2.64–5.47)
ISS	1.07 (1.06–1.08)	1.06 (1.05–1.06)

HR=Hazard Ratio

^*Propensity score model (AUC 0.78, 95%CI: 0.77–0.79) included the following variables: gender; age group (< 65, >=65); race; mechanism of injury; prehospital EMS transport; skull fracture/intracranial injury; internal organ injury (liver, spleen, kidney); rib fracture; pelvic and long bone fractures; EMS/ED intubation; initial ED SBP; primary payer; and injury period.

^ΦEffect calculated while ‘holding out’ patients in the non-overlapping propensity score quintile (first quintile).

Discussion

The concept that seriously injured older adults are under-triaged to tertiary trauma centers is now well established.^7–14,26 However, research remains mixed regarding the survival benefit conferred by treatment at tertiary trauma centers in this patient population. The findings of our study suggest that in a predominantly rural state, moderate to seriously injured older adults treated at tertiary trauma centers have a significant survival advantage over their counterparts treated at non-tertiary trauma centers after adjusting for potential confounders, with a much stronger protective effect observed in patients ages 55 to 64 years. Of note, this benefit extends to patients transferred from non-tertiary trauma centers as well. In a related study conducted in a similar patient population, we demonstrated a two-fold survival benefit associated with direct transport from the scene of injury, regardless of age.²⁷ This suggests that benefit derived from treatment at a TC is further augmented when patients are transported directly from the scene. The American College of Surgeons Committee on Trauma (ASCOT) defines under- and over-triage based on ISS >=16, as this threshold denotes the subgroup of patients most likely to benefit from treatment at a tertiary trauma center.^19,28 Restricting our analysis to patients with ISS >=16 marginally strengthened the protective effect of treatment at a tertiary trauma center, but results remained qualitatively similar to those of the full study cohort. For predominantly rural states like Oklahoma, it is impractical to transport all trauma patients directly to a tertiary trauma center. However, we have shown that, compared to their younger counterparts, moderate-seriously injured older adults are less likely to be transferred to TCs for higher level of trauma care.^7,29 Other studies have shown inter-hospital transfer also occurs more slowly for older adults.³⁰ While secondary triage of patients for inter-hospital transfer is designed to ensure that patients requiring a higher level of trauma care are rapidly identified and transferred in a timely manner, this does not always occur in practice, especially among older adults. Given the large proportion (74%) of patients initially presenting to Non-TCs in our study, trauma system effectiveness on improving outcomes may largely hinge on this process, as our findings demonstrate improved survival among transferred patients. Notably in our study, based on our propensity score model predicting transfer to a TC, insurance status independently influenced whether patients were transferred to TCs; compared to patients with commercial insurance, patients with Medicare coverage were less likely to be transferred while self-paying patients were more likely to be transferred.

While comparison of studies can be complicated by differences in study locale, case mix, trauma system structure, definitions of key variables and risk adjustment strategies, there have been previous studies that found a survival advantage among injured older adult patients treated at tertiary trauma centers. ^15–17 Pracht et al. using an instrumental variable (IV) approach, concluded that severely injured elderly patients (>=65 years) treated at designated trauma centers (DTCs) had a survival advantage over their counterparts at non-tertiary trauma centers; this study however excluded transfers and did not further adjust for mechanism of injury, anatomic injury and physiologic factors. ¹⁷ Meldon et al. also found a survival benefit in the very elderly (>=80 years) treated at TCs but similarly did not adjust for potentially confounding factors such as mechanism of injury, pre-existing conditions, presence of shock, etc.¹⁶ In terms of morbidity, Smith et al. showed that patients with a femoral shaft fracture treated at trauma centers suffered fewer complications compared to those treated at non-trauma centers, particularly in the older age group.¹⁵ In their analysis however, the authors did not adjust for potential confounders such as pre-existing conditions, severity of concomitant injuries; the study’s focus on a single injury further also limits its generalizability. Other studies have observed no survival benefit or increased mortality for older patients treated at TCs.^18–20 A study by Scheetz et al. concluded that injured older adult (>=65 years) patients admitted to non-tertiary trauma centers had more than a 4-fold increase in the likelihood of survival after adjusting for ISS and age.²⁰ This study was however limited by potential residual confounding resulting from not adjusting for potentially confounding variables such as physiologic factors, pre-existing conditions, prehospital care. In a study by Staudenmayer et al. there was no difference in 60-day mortality between injured older adults (>=55 years) treated at trauma centers and those treated at non-trauma centers. Similarly, this study did not adjust for pre-existing comorbidities nor did the authors consider specific injuries (e.g. head vs extremity injuries); and the mortality analysis was based on only 41 deaths, which likely affected the precision of the estimated effect. Using a multi-state sample, MacKenzie et al. evaluated the effect of trauma center care on mortality and found no significant difference in mortality outcomes among older adults. ¹⁹ As alluded by the authors, their analysis was limited by the small number of older patients with severe injuries resulting in imprecise estimates.

Albeit not perfect, an inherent strength of our study is that we were able to improve upon previous methodology utilized in prior studies. In this study, we were able to link most of the patients transferred between facilities thereby increasing the accuracy of covariates of interest. We were also able to evaluate the survival benefit of transferring injured older adults to TCs using a propensity-adjusted survival analysis. Furthermore, we adjusted for additional potential confounders not adjusted for in prior studies such as physiologic factors, mechanism of injury, pre-existing comorbidity, etc. Prior studies have also been hampered by small sample sizes and/or number of outcome events analyzed.^18,19 In our study, we utilized a large sample size and had at least 1000 deaths in each exposure group (TC and Non-TC). Additionally, we are not aware of any prior study that considered the timing of death in their analysis as we did in this study. It is possible that treatment at TCs confers a survival advantage for early mortality (e.g. 2 weeks) but not for in-hospital mortality after a certain time period.

While the results of this study provide strong evidence of the survival benefit conferred by management and treatment at tertiary trauma centers among injured older adults, caution is warranted in interpreting and translating these results. In practice, realities involving such factors as scarcity of resources and patient preference/decision play an important role. Injured older patients may prefer to stay at a Non-TC and receive definitive treatment, given their established relationships with local physicians. Furthermore, triaging injured older patients to TCs based on age alone would be highly inefficient and not cost-effective. Nakamura et al. estimated a number needed to treat (NNT) ranging from 60 to 65 i.e. one (1) older adult patient with an ISS of 16 identified for every 60 to 65 additional patients transported to a TC. ¹³ In the Oklahoma trauma system, a large proportion of injured older adults are treated at Non-TCs (Level III/IV). There is heterogeneity in the capacity to care for seriously injured patients among Non-TC hospitals e.g. some high functioning Level III trauma centers are capable of caring for certain high-need patients while some Level IIIs will generally transfer such patients. As long as appropriate care is provided in a timely manner, our results should not be misconstrued to imply that transport to/treatment of seriously injured older adults at a Non-TC is inappropriate. The challenge is to identify that subgroup of patients that would benefit from TC care rather than Non-TC care, as well aspects of care at TCs that yield this benefit. A recent study by Bosson et al. concluded that no individual triage criterion of the currently recommended CDC guidelines, ²² definitively identifies patients who would benefit from transport to a TC and that these recommendations should be considered in their entirety.³¹ A few studies have also explored specific field trauma triage criteria for older adults or ways to improve triaging of these patients to tertiary trauma centers with the general finding of improved sensitivity but decreased specificity of their proposed criteria. ^32–36 While these studies are an important step, more work including additional education of prehospital providers, needs to be done to improve triaging of older adults to TCs in a cost-effective manner.

Inter-facility transfer guidelines exist in the State of Oklahoma and reinforcement of these guidelines and continued education of medical providers in Non-TCs is warranted. To give Non-TCs a sense of ownership these trauma facilities should be integrated into the planning and audit aspects of the transfer process.

Our study has limitations that warrant further discussion. The retrospective nature of the study precluded us from considering additional information that could alter our study estimates. To that point, we were not able to take into account patient preference to stay at Non-TCs or withdrawal of care particularly in the very elderly in our adjustment. It is unclear how excluding such patients would affect our results. Another limitation of our study is that deaths were not tracked after hospital discharge. Some patients may have been discharged to intermediate care facilities and died there within 30 days. The use of in-hospital mortality as an outcome measure is further limited by the fact that older adults have been shown to have significant high mortality in the first 60 days post-discharge;^37,38 it’s possible inclusion of post-discharge deaths might alter our estimates of the effect of TC on mortality. It also important to note that this study presents an average TC effect on mortality and did not identify specific subgroups of patients or patients with injuries or combinations of injuries who would benefit from TC care. Additionally, our study was conducted within a predominantly rural state and as such, may not be generalizable to regions that have an urban prominence. Aforementioned limitations notwithstanding, the major strengths of our study are that we were able to improve upon previous methodology in evaluating the effect tertiary trauma center care and were also able to assess the survival benefit of transfer to TC in this patient population.

Conclusions

In this sample, there was a survival benefit for injured older adults treated at tertiary trauma centers compared to their counterparts treated at non-tertiary trauma centers. Further research should focus on identifying specific subgroups of patients who would especially benefit from treatment/transfer to tertiary trauma centers to prevent excessive and inefficient triage to these centers. In addition, barriers to effective secondary triage to TCs by non-tertiary trauma centers should also be evaluated to improve this process.