Article, Traumatology

Comparative study of ED mortality risk of US trauma patients treated at level I and level II vs nontrauma centers

a b s t r a c t

Background: Prior studies of undertriage have not made comparisons across multiple trauma levels.

Methods: Emergency department data was extracted from the Nationwide Emergency Department Sample for major trauma patients. We considered patients with moderate injuries (Injury Severity Score, ISS=16-24) and severe injuries (ISS=25-75) separately. Conditional logistic regression modeling was used to compare the odds of ED mortality for level I trauma centers (TC I) vs. nontrauma centers (NTC) and level II trauma centers (TC II) vs. NTC. An innovative 1:1:1 optimal matching (an extension of the traditional pair matching) was used to balance patient characteristics in three groups. To facilitate matching of all NTC patients, 3 subgroups were developed for ISS=16-24 and 2 subgroups for ISS=25-75. Sensitivity analyses were performed to assess the strength of the association between trauma center designation and ED mortality.

Results: For ISS=16-24, 2 of 3 subgroups had marginally significant reduced odds of ED mortality when properly triaged (TC I vs. NTC [T1:OR=0.63; 95%CI: 0.45 – 0.89, T2:OR=0.71;95%CI:0.51-0.99]). For ISS=25-75, both

subgroups had significantly reduced odds of emergency department mortality when properly triaged (H1: TC I vs. NTC [OR=0.61; 95%CI: 0.50-0.74]; TC II vs. NTC [OR=0.49; 95%CI: 0.38 – 0.63]; H2: TC I vs. NTC [OR=0.50;

95%CI: 0.41 – 0.60]; TC II vs. NTC [OR=0.42; 95%CI: 0.33 – 0.53]). Conclusions for ISS 25-75 were robust to a hy- pothesized unobserved confounding variable as shown in sensitivity analysis.

Conclusions: Trauma patients with ISS>=25 received most benefit from proper triage. Efforts to reduce undertriage should focus on this population.

(C) 2015

  1. Introduction

Trauma centers (TCs) in the United States are designated accord- ing to the resources and expertise available to handle traumatic inju- ries. Previous work has shown that better outcomes are achieved through the establishment of inclusive, regionalized systems of

? Funding source: This work was supported by a research grant (PI: Dr Huiyun Xiang, grant no.:R03-HS22277) from the Agency for Healthcare Research and Quality (Rockville, MD) and in part by The Ohio State University College of Medicine Bennett (Columbus, OH) research scholarship (BPV). The views and conclusions are solely the responsibility of the authors and do not represent official views of the funding agencies.

?? Conflicts of interest: The authors have no conflicts of interest to report.

* Corresponding author at: Center for Pediatric Trauma Research, The Research Institute at Nationwide Children’s Hospital, 700 Children’s Drive, Columbus, OH 43205. Tel.: +1 614 355 5893; fax: +1 614 355 5897.

E-mail address: [email protected] (H. Xiang).

trauma care [1,2]. In this system, all Acute care facilities in a region have the capacity to stabilize trauma patients and assess for the need to transfer major trauma patients to a level I or level II TC [3]. Ideally, the most severe injuries receive definitive treatment at level I or level II TCs, and the least severe injuries receive definitive treatment at level III or nontrauma centers (NTCs) [4]. Undertriage results when major trauma patients receive definitive treatment in a facility without the resources or expertise to care for the injury. Studies have shown that undertriage of trauma patients results in in- creased morbidity and mortality [5-7]. Thus, reducing undertriage is a target for improving outcomes of injured patients. This area of study is vital, as violence and injury are leading causes of mortality in the United States for persons between the ages of 1 and 44 years, and they caused more years of potential life lost than any other cause of mortality in individuals younger than the age of 65 years in 2011 [8].

Mann et al [9] systematically reviewed the published evidence regarding Trauma system effectiveness, noted the weak evidence, and

http://dx.doi.org/10.1016/j.ajem.2015.05.010

0735-6757/(C) 2015

called for stronger evidence from well-designed retrospective or pro- spective controlled cohort studies or well-matched case control studies. Subsequently, Mackenzie et al [10] demonstrated a 25% decrease in 1-year mortality associated with proper triage of major trauma patients to a level I TC compared with an NTC; however, to date, no study has evaluated emergency department (ED) mortality outcomes of undertriage at NTCs using a nationally representative data set. We chose to study ED mortality using the Nationwide ED Sample (NEDS) to evaluate the differences in emergency care between level I and NTCs and between level II and NTCs. Previous work on the national level has relied on data from the National Trauma Data Bank [11,12]. National Trauma Data Bank data are reported voluntarily by TCs; consequently, it is subject to convenience sample bias. In addition, few states report information from NTCs.

The Agency for Healthcare Research and Quality created and maintains NEDS through a federal, state, and industry partnership. The NEDS provides data on approximately 30 million ED records per year, which when weighted is an estimated 130 million yearly ED visits, making it the largest all-payer ED database in the United States [13]. The large sample sizes in NEDS provided the opportuni- ty to produce the strongest evidence to date of the impact of undertriage at NTCs. We used NEDS data to explore our hypothesis that severely injured Trauma victims properly triaged to a level I or level II TC have significantly lower odds of death than those undertriaged to an NTC. Contributions to this area of study are expected to better characterize the impact of undertriage and to provide evidence for improving regionalized trauma care in the United States.

  1. Materials and methods
    1. Data source

Data were extracted from 2006 to 2010 NEDS. Adult, major trauma patients between the ages of 18 and 64 years with a prima- ry International Classification of Diseases, Ninth Revision (ICD-9- CM), diagnosis code of 800 to 959.9 (excluding injuries from late effects [905-909.9], superficial injuries [910-924.9], and injuries due to foreign bodies [930-939.9]) were selected for our study. This ICD-9-CM definition of traumatic injury is consistent with the National Trauma Data Standard Patient Inclusion Criteria [14]. Major trauma patients were defined by an injury severity score (ISS) greater than or equal to 16. Patients missing any study variables and patients transferred to another hospital were excluded from the study. A total of 92693 patients met criteria for inclusion in the study.

Study variables

Important variables extracted from NEDS were ICD-9-CM diagnosis, ISS, age, sex, presence of a chronic comorbid condition, presence of multiple injuries, median Household income, primary expected payer, patient location, and TC designation of the treating hospital. The outcome of interest was ED mortality, which was defined as those who died in the ED. Patients who were discharged alive from the ED and patients admitted to the hospital, regardless of whether they were discharged alive or died in the hospital were classified as “did not die in the ED.”

Statistical analysis

The study population was split into 2 groups, ISS 16 to 24 and ISS 25 to 75. We designated patients treated at an NTC to the treatment group (in statistical terms), with level I TC and level II TC serving as the 2 control groups. We were interested in understanding the ED outcome of patients treated in NTCs when compared with patients

treated at level I and level II centers. A carefully chosen second con- trol group in 3-group simultaneous analysis helps alleviate concern about potential unobserved bias between the first control group and the treatment group [15]. Our study used an innovative 1:1:1 3-way data matching technique to obtain similar baseline covariate distributions between treatment and control groups. Because of the nature of the matching technique, only the group with the least number of patients will have all patients matched with patients from control groups. To ensure all patients in the treatment group were included in the final analysis, patients treated at an NTC with ISS 16 to 24 were randomly split into 3 subgroups designated T1, T2, and T3; patients treated at an NTC with ISS 25 to 75 were randomly split into 2 subgroups designated H1 and H2. In other words, because there were limited numbers of control patients, patients within these different nontrauma subgroups could have the same matched control.

Patients were matched on ISS, age, sex, presence of chronic comorbid condition, presence of multiple injuries, Median household income, pri- mary expected payer, and patient location. We used the Mahalanobis distance matching measure to assess the similarity between 2 indi- viduals and optimal pair matching to produce matched sets that minimized the total distance between each pair of treatment and con- trol [16,17]. We chose this matching method over propensity score matching because it has advantages when there is a limited number of matching covariates. The “Pairmatch” function of Hansen [18] in his “Optmatch” R package was used in our study. Optimally matched pairs were constructed in 2 sets–the first between NTC and level I TC groups and the second between NTC and level II TC groups. Each 1:1 optimally matched pair was linked with another optimally matched pair from the other set through common NTC patients to develop 1:1:1 matched triplets.

Absolute standardized differences were calculated for all co- variates to assess for balance in postmatching data. An acceptable threshold of less than 0.25 is suggested [17]. Postmatching residual covariate imbalances were corrected through placement of the co- variate in the outcomes regression model on the recommendation of Rosenbaum [19]. Outcome analyses of ED mortality among the 3 matched groups were performed by fitting a conditional logistic regression model using SAS 9.3 (SAS, Cary, NC) software. Trauma center designation was the independent variable, and each triplet was the stratum. Imputation-based sensitivity analyses were per- formed on all significant results to assess the robustness of the asso- ciation between TC designation and ED mortality according to previously described methods [20]. Sensitivity analyses occurred in 2 steps: (1) data were created for an unobserved confounding variable based on hypothesized associations between treatment and outcome; and (2) the unobserved confounding variable was in- cluded in the conditional logistic regression model, and the average effect of the unobserved confounding variable was tested over a range of magnitudes using a multiple imputation technique to deter- mine the level of confounding required to overturn the conclusion. The associations between the confounding factor and treatment and the confounding factor and outcome required to overturn the qualitative conclusion were plotted in 2 dimensions.

  1. Results

Of the 92693 patients included in our study, 70838 patients had an ISS 16 to 24, and the remaining 21 855 patients had an ISS 25 to

75. Crude ED mortality as a function of Trauma level designation and ISS category is shown in Table 1. For ISS 16 to 24, balance was achieved in all postmatching variables in all 3 subgroups except for ISS and presence of multiple injuries in the NTC vs level II TC matched sample. For ISS 25 to 75, balance was achieved in all postmatching variables in both subgroups except for presence of chronic condition and presence of multiple injuries in the NTC vs level II TC

Table 1 Emergency department mortality before matching on patient characteristics among adult (age 18-64 years) major trauma (ISS >=16) patients, NEDS 2006 to 2010

Table 3

Emergency department mortality comparison, matched sample, NEDS 2006 to 2010

TC designation Did not die in

the ED Died in the

ED Total

ED mortality (%)

ISS 16-24, T1a

ISS 16-24

TC I vs NTC

0.63

0.45-0.89

NTC

25763

258

26021

0.99

TC II vs NTC

0.83

0.60-1.14

TC I

34043

229

34272

0.67

ISS 16-24, T2a

TC II

10460

85

10545

0.81

TC I vs NTC

0.71

0.51-0.99

Subtotal

ISS 25-75

70266

572

70838

TC II vs NTC

ISS 16-24, T3a

0.95

0.69-1.30

NTC

4554

760

5314

14.30

TC I vs NTC

0.80

0.57-1.11

TC I

12880

503

13383

3.76

TC II vs NTC

1.04

0.75-1.43

TC II

3024

134

3158

4.24

ISS 25-75, H1b

Subtotal

20458

1397

21855

TC I vs NTC

0.61

0.50-0.74

OR 95% CI

Abbreviations: TC I, level I TC; TC II, level II TC.

matched sample. All matched variables achieved balance in the NTC vs level I TC matched sample (Appendix). Original and matched ED mortality as a function of trauma level designation is shown for each subgroup in Table 2.

Calculated odds ratios (ORs) of ED mortality (Table 3) were signifi- cantly reduced in subgroup T1 for level I TC vs NTC (OR, 0.63; 95% confidence interval [CI], 0.45-0.89), in subgroup T2 for level I TC vs NTC (OR, 0.71; 95% CI, 0.51-0.99), subgroup H1 for level I TC vs NTC (OR, 0.61; 95% CI, 0.50-0.74) and level II TC vs NTC (OR, 0.49; 95% CI,

0.38-0.63), and subgroup H2 for level I TC vs NTC (OR, 0.50; 95% CI, 0.41-0.60) and level II TC vs NTC (OR, 0.42; 95% CI, 0.33-0.53). Sensi-

tivity analyses of significant associations for level I TC vs NTC indicated that the results for those with moderate injuries, subgroup T1 and subgroup T2, could be overturned with low levels of association between an unobserved confounding variable, TC designation, and ED mortality; conversely, the results for those with severe injuries, subgroup H1, were robust and were overturned only with extreme levels of association between an unobserved confounding variable, TC designation, and ED mortality (Figure). Sensitivity analyses of sig- nificant associations for level II TC vs NTC indicated that the results of those with severe injuries, subgroups H1 and H2, were robust and

Table 2

Emergency department mortality after 3-way triplet matching; adult, major trauma patients, NEDS 2006 to 2010

Original sample Matched sample

TC II vs NTC 0.49 0.38-0.63

ISS 25-75, H2b

TC I vs NTC 0.50 0.41-0.60

TC II vs NTC 0.42 0.33-0.53

a After controlling for chronic condition, multiple injury, and ISS score.

b After controlling for chronic condition and multiple injury.

were not overturned at any level of association tested between an unobserved confounding variable, TC designation, and ED mortality. The sensitivity analysis is not done for T3 because the logistic regres- sion results were not significant.

  1. Discussion

This study assesses the impact of undertriage at NTCs on major trauma victims using a nationally representative data source. Mackenzie et al [10] produced the first evidence of the Negative outcomes associated with undertriage at NTCs. Our study has a similar aim but differed in many important aspects. The NEDS provides a national characterization of the trauma system and allows us to avoid sampling data on the regional level. The study of Mackenzie et al included only NTCs that treated at least 25 major trauma patients per year, and a third of the 51 NTCs they included had designated trauma teams [10]. We provide a more complete picture of the impact of undertriage at NTCs by including all NTCs in NEDS. In addition, the large sample sizes in NEDS allowed us to compare subgroups while maintaining statistical power. We ensured balance between observed covariates in patients treated at level I, level II, and NTCs with a novel triplet matching technique. The sensitivity analysis is an important strength of our study because administrative data sets limit the amount of variables that can be balanced. This leaves potential sources of unobserved bias. By generating a confounding variable to represent unobserved bias and testing over a range of magnitudes, our sensitivity analyses provide

ISS 16-24, T1

NTC

8309

92

1.11

8309

92

1.11

TC I

34272

229

0.67

8309

56

0.67

TC II

10545

85

0.81

8309

70

0.84

ISS 16-24, T2 NTC

8963

85

0.95

8963

85

0.95

TC I

34272

229

0.67

8963

59

0.66

TC II

10545

85

0.81

8963

78

0.87

ISS 16-24, T3 NTC

8749

81

0.93

8749

81

0.93

TC I

34272

229

0.67

8749

62

0.71

TC II

10545

85

0.81

8749

80

0.91

ISS 25-75, H1

NTC

2655

362

13.63

2655

362

13.63

TC I

13383

503

3.76

2655

207

7.80

TC II

3158

134

4.24

2655

122

4.60

ISS 25-75, H2

NTC

2659

398

14.97

2659

398

14.97

TC I

13383

503

3.76

2659

197

7.41

TC II

3158

134

4.24

2659

123

4.63

n Died in ED

Mortality in ED (%)

n Died in ED

Mortality in ED (%)

evidence of the strength of the association between TC designation and ED mortality.

This study shows that the detrimental consequences associated with undertriage at NTCs are more severe than previously shown [10]. The significant reduction in odds of mortality by approximately half in severely injured trauma patients (ISS, 25-75) properly triaged to a level I or level II TC vs an NTC is an important finding. Equally important is the finding that 1 of the 3 subgroups of those with moderate injuries (ISS, 16-24) failed to achieve statistically significant reductions in mortality when properly triaged to a level I or level II TC over an NTC. However, there was a trend toward improved outcomes in ISS 16 to 24 treated at a level I TC vs an NTC. These results suggested a potential threshold point of injury severity to optimize benefit from proper triage of trauma victims.

Previous studies investigating outcomes of major trauma patients treated at a level I vs level II TC have arrived at different conclusions. Some have reported improved outcomes with treatment at a level I TC [11,12,21], whereas others have found no significant difference between major trauma patient mortality outcomes between level I

Figure. Sensitivity analyses contour plot of significant results for TC I vs NTC. OR1 denotes association between the confounding factor and TC designation; OR2 denotes association between the confounding factor and ED mortality. Each line indicates the significance point (P = .05) at which qualitative conclusions of significant results change. All points including and superior to each line represent magnitudes of confounding, which if present, have the strength to overturn the conclusion. For example, in ISS 16 to 24, the presence of a confounding factor with an OR of 2 between both TC designation and ED mortality has the strength to overturn the conclusion in this group. Whereas in ISS 25 to 75, H1, a confounding factor with an OR of 7 between both TC designation and ED mortality is required to overturn the conclusion in this group. Sensitivity analyses of significant results for TC II vs NTC were not plotted because conclusions were not overturned at any range of magnitude tested.

and level II TCs [22]. These differing conclusions likely result from the considerable variability shown to exist between regional trauma systems [23,24]. Although our study did not directly compare mortality outcomes between level I and level II TCs, we can say that they both produced a similarly significant reduction in odds of ED mortality vs NTCs for severely injured trauma patients.

After the American College of Surgeons Committee on Trauma (ACS-COT) call for a reduction in undertriage rates to below 5%, Mohan et al [25] looked at current practice patterns and concluded that this benchmark may not currently be feasible given the required increase in TC capacity. Mohan et al [25] used ISS greater than 15 or an injury satisfying the ACS-COT definition of life threatening or critical in their analysis of those who should be treated at a level I or level II TC. However, our results indicated that the focus in reduction of undertriage rates must be directed toward the most severely injured patients (ISS, 25-75). This is a pressing issue, as 24.3% of the 21855 patients with ISS 25 to 75 in the 2006 to 2010 NEDS data received defin- itive treatment at an NTC.

Using 2010 NEDS data, Xiang et al [26] showed that level I and level II TCs in the United States would have to increase trauma team capacity 6.2% to treat the estimated 6924 undertriaged patients per year with ISS greater than or equal to 25 vs a 51.5% capacity increase to treat the estimated 57609 undertriaged patients per year with ISS greater than or equal to 16. The feasibility of a 6.2% capacity increase requires further study.

  1. Conclusion

In conclusion, the significant reduction in odds of mortality from proper triage of severely injured trauma patients cannot be ignored.

Improving Triage protocols and accommodating additional trauma vic- tims at TCs or improving the response of NTCs to major trauma are 2 strategies for producing better outcomes.

Study limitations

The clinical picture of our patients was limited to variables pro- vided in the NEDS database, primarily ICD-9-CM Diagnosis codes and ISS. Some researchers have advocated the use of other severity scores [27], and others have used physiologic criteria as well as re- sources and interventions required during trauma patient care to identify severe injuries [28]. However, ISS is correlated with trauma patient mortality and is used by the ACS-COT as a criterion for iden- tifying severe injury [29,30].

The NEDS data are compiled on an event basis and not a patient basis; therefore, the same patient could have multiple records in the database for each time they visited an ED. To account for this in the setting of acute traumatic injury, we excluded patients who were transferred out of the ED. We were unable to examine mortality beyond the ED or the impact of undertriage on patient morbidity. Despite some of the limi- tations of NEDS, the statistical strength generated from the large sample size has produced some of the strongest evidence to date on the impact of undertriage at NTCs.

Acknowledgment

Sarah A Johnson assisted in the manuscript preparation and submission.

Appendix. Balance checking of observed covariates, matched sample, NEDS 2006 to 2010

Balance checking, ISS:16-24,T1

Original sample Matched sample

Mean/percent Standardized difference Mean/percent Standardized difference

NTC

TC I

TC II

NTC vs

NTC vs

NTC

TC I

TC II

NTC vs

NTC vs

(N = 8309)

(N = 34272)

(N = 10545)

TC I

TC II

(N = 8309)

(N = 8309)

(N = 8309)

TC I

TC II

Age (mean, yrs) Mean

43.3

39.6

40.7

0.2607

0.1779

43.3

42.4

41.5

0.0594

0.1244

SD

(14.0)

(13.9)

(14.2)

(14.0)

(13.7)

(14.1)

ISS (mean) -0.5608 -0.4523 -0.0783 -0.2901

Mean

17.4

18.6

18.4

17.4

17.5

18.0

SD

(2.0)

(2.5)

(2.4)

(2.0)

(2.0)

(2.2)

Female

28.7

25.3

26.2

0.0767

0.0549

28.7

26.8

26.7

0.0426

0.0442

With chronic condition

59.9

72.9

72.6

-0.2781

-0.2712

59.9

65.0

69.8

-0.1101

-0.2115

Multiple injury

63.1

86.5

83.2

-0.5594

-0.4651

63.1

67.4

79.2

-0.1014

-0.3726

Patient location

Large central metropolitan

21.5

25.0

15.2

-0.0823

0.1620

21.5

22.5

17.9

-0.0230

0.0925

Large fringe metropolitan

22.5

22.2

26.5

0.0070

-0.0931

22.5

22.5

24.8

0.0014

-0.0528

Medium metropolitan

19.9

22.0

28.3

-0.0499

-0.1968

19.9

20.1

24.5

-0.0029

-0.1069

Small metropolitan

9.6

8.0

8.4

0.0573

0.0395

9.6

9.5

9.5

0.0039

0.0038

Micropolitan

15.1

13.2

10.8

0.0551

0.1277

15.1

15.0

12.8

0.0040

0.0695

Not metropolitan or

9.9

8.2

9.8

0.0568

0.0013

9.9

9.9

9.6

0.0003

0.0081

micropolitan

Median household income Q1 (0-25%)

28.1

34.1

25.1

-0.1301

0.0670

28.1

29.6

26.7

-0.0316

0.0324

Q2 (25%-50%)

27.4

26.7

31.5

0.0142

-0.0914

27.4

27.5

30.5

-0.0020

-0.0694

Q3 (50%-75%)

22.7

20.9

21.4

0.0451

0.0323

22.7

22.5

22.3

0.0053

0.0097

Q4 (75%-100%)

17.9

14.6

19.7

0.0902

-0.0471

17.9

17.6

18.0

0.0084

-0.0031

Primary expected payer Medicare

10.0

6.2

5.8

0.1405

0.1583

10.0

10.0

7.3

0.0000

0.1011

Medicaid

11.9

14.9

12.0

-0.0867

-0.0028

11.9

11.9

11.7

0.0009

0.0071

Private including HMO

43.2

46.4

53.4

-0.0642

-0.2058

43.2

43.4

49.8

-0.0042

-0.1332

Self-pay

23.1

19.3

18.3

0.0916

0.1190

23.1

22.7

20.6

0.0098

0.0601

No charge

1.5

2.1

0.7

-0.0448

0.0848

1.5

1.5

0.8

0.0000

0.0675

Other

9.6

10.1

9.7

-0.0165

-0.0037

9.6

9.6

9.5

0.0003

0.0017

Balance checking, ISS: 16-24, T2

Original sample Matched sample

Mean/percent Standardized difference Mean/percent Standardized difference

NTC

TC I

TC II

NTC vs

NTC vs

NTC

TC I

TC II

NTC vs

NTC vs

(N = 8963)

(N = 34272)

(N = 10545)

TC I

TC II

(N = 8963)

(N = 8963)

(N = 8963)

TC I

TC II

Age (mean, yrs) Mean

43.6

39.6

40.7

0.2858

0.2022

43.6

42.7

41.3

0.0668

0.1604

SD

(13.9)

(13.9)

(14.2)

(13.9)

(13.7)

(14.1)

ISS (mean) -0.5800 -0.4710 -0.0903 -0.3300

Mean

17.3

18.6

18.4

17.3

17.5

18.1

SD

(1.9)

(2.5)

(2.4)

(1.9)

(2.0)

(2.2)

Female

28.6

25.3

26.2

0.0756

0.0538

28.6

26.8

26.8

0.0406

0.0413

With chronic condition

60.1

72.9

72.6

-0.2733

-0.2665

60.1

65.4

70.4

-0.1125

-0.2201

Multiple injury

62.6

86.5

83.2

-0.5723

-0.4778

62.6

67.5

80.5

-0.1172

-0.4142

Patient location

Large central metropolitan

20.6

25.0

15.2

-0.1031

0.1411

20.6

21.6

17.2

-0.0229

0.0904

Large fringe metropolitan

21.8

22.2

26.5

-0.0104

-0.1105

21.8

21.7

24.6

0.0014

-0.0660

Medium metropolitan

20.1

22.0

28.3

-0.0464

-0.1933

20.1

20.1

25.2

-0.0015

-0.1208

Small metropolitan

9.6

8.0

8.4

0.0583

0.0405

9.6

9.5

9.3

0.0028

0.0115

Micropolitan

15.2

13.2

10.8

0.0571

0.1297

15.2

15.1

12.3

0.0026

0.0844

Not metropolitan or micropolitan

11.1

8.2

9.8

0.0975

0.0421

11.1

11.1

10.5

0.0003

0.0206

Median household income Q1 (0-25%)

28.6

34.1

25.1

-0.1194

0.0777

28.6

30.8

26.5

-0.0486

0.0479

Q2 (25%-50%)

26.5

26.7

31.5

-0.0052

-0.1108

26.5

26.4

30.4

0.0027

-0.0861

Q3 (50%-75%)

22.6

20.9

21.4

0.0434

0.0307

22.6

22.2

21.9

0.0119

0.0179

Q4 (75%-100%)

18.0

14.6

19.7

0.0941

-0.0432

18.0

17.9

18.8

0.0041

-0.0205

Primary expected payer Medicare

10.1

6.2

5.8

0.1432

0.1610

10.1

10.1

6.7

0.0000

0.1257

Medicaid

12.8

14.9

12.0

-0.0591

0.0249

12.8

12.8

12.2

0.0012

0.0179

Private including HMO

43.9

46.4

53.4

-0.0486

-0.1901

43.9

44.1

51.2

-0.0028

-0.1465

Self-pay

21.9

19.3

18.3

0.0643

0.0917

21.9

21.6

19.8

0.0084

0.0530

No charge

1.2

2.1

0.7

-0.0727

0.0575

1.2

1.2

0.7

0.0000

0.0491

Other

9.4

10.1

9.7

-0.0247

-0.0119

9.4

9.3

9.1

0.0007

0.0089

Balance checking, ISS: 16-24, T3

Original sample Matched sample

Mean/percent Standardized difference Mean/percent Standardized difference

NTC

TC I

TC II

NTC vs

NTC vs

NTC

TC I

TC II

NTC vs

NTC vs

(N = 8749)

(N = 34272)

(N = 10545)

TC I

TC II

(N = 8749)

(N = 8749)

(N = 8749)

TC I

TC II

Age (mean, yrs) Mean

43.6

39.6

40.7

0.2817

0.1987

43.6

42.6

41.4

0.0667

0.1527

SD

(14.1)

(13.9)

(14.2)

(14.1)

(13.7)

(14.1)

ISS (mean) -0.5784 -0.4695 -0.0841 -0.3208

Mean

17.3

18.6

18.4

17.3

17.5

18.0

SD

(1.9)

(2.5)

(2.4)

(1.9)

(1.9)

(2.2)

Female

28.2

25.3

26.2

0.0658

0.0441

28.2

26.1

26.6

0.0470

0.0362

With chronic condition

60.3

72.9

72.6

-0.2704

-0.2636

60.3

65.4

70.2

-0.1093

-0.2119

Multiple injury

63.0

86.5

83.2

-0.5633

-0.4690

63.0

67.7

79.9

-0.1133

-0.3930

Patient location

Large central metropolitan

21.2

25.0

15.2

-0.0884

0.1559

21.2

21.8

17.5

-0.0135

0.0982

Large fringe metropolitan

21.5

22.2

26.5

-0.0182

-0.1182

21.5

21.5

24.1

-0.0010

-0.0619

Medium metropolitan

19.9

22.0

28.3

-0.0509

-0.1978

19.9

19.9

25.3

-0.0005

-0.1259

Small metropolitan

10.1

8.0

8.4

0.0748

0.0569

10.1

10.0

9.6

0.0024

0.0162

Micropolitan

15.4

13.2

10.8

0.0649

0.1374

15.4

15.3

12.5

0.0034

0.0882

Not metropolitan or

10.4

8.2

9.8

0.0734

0.0179

10.4

10.4

10.1

0.0000

0.0090

micropolitan

Median household income Q1 (0-25%)

28.4

34.1

25.1

-0.1225

0.0746

28.4

30.4

26.8

-0.0432

0.0368

Q2 (25%-50%)

27.6

26.7

31.5

0.0187

-0.0869

27.6

27.4

30.9

0.0040

-0.0737

Q3 (50%-75%)

22.9

20.9

21.4

0.0487

0.0359

22.9

22.5

22.3

0.0080

0.0133

Q4 (75%-100%)

16.8

14.6

19.7

0.0608

-0.0765

16.8

16.4

17.5

0.0094

-0.0187

Primary expected payer Medicare

10.2

6.2

5.8

0.1445

0.1623

10.2

10.2

7.0

0.0000

0.1185

Medicaid

12.5

14.9

12.0

-0.0696

0.0143

12.5

12.4

12.1

0.0015

0.0104

Private including HMO

42.9

46.4

53.4

-0.0698

-0.2115

42.9

43.0

50.4

-0.0022

-0.1503

Self-pay

23.1

19.3

18.3

0.0928

0.1202

23.1

22.8

20.4

0.0091

0.0685

No charge

1.3

2.1

0.7

-0.0605

0.0696

1.3

1.3

0.8

0.0000

0.0554

Other

9.3

10.1

9.7

-0.0253

-0.0126

9.3

9.3

9.3

0.0000

0.0027

Balance checking, ISS: 25-75, H1

Original sample Matched sample

Mean/percent Standardized difference Mean/percent Standardized difference

NTC

TC I

TC II

NTC vs

NTC vs

NTC

TC I

TC II

NTC vs

NTC vs

(N = 2655)

(N = 13383)

(N = 3158)

TC I

TC II

(N = 2655)

(N = 2655)

(N = 2655)

TC I

TC II

Age (mean, yrs) Mean

38.5

37.2

37.8

0.0968

0.0546

38.5

38.3

38.5

0.0160

0.0060

SD

(13.7)

(13.6)

(13.9)

(13.7)

(13.5)

(13.8)

ISS (mean) 0.1233 0.1112 0.0417 0.0829

Mean

34.2

32.5

32.6

34.2

33.6

33.0

SD

(16.6)

(10.9)

(11.9)

(16.6)

(15.1)

(12.7)

Female

26.9

25.4

24.6

0.0224

0.0404

26.9

24.5

25.7

0.0548

0.0284

With chronic condition

51.9

71.3

71.3

-0.4573

-0.4569

51.9

56.6

67.9

-0.0988

-0.3368

Multiple injury

66.8

92.3

89.1

-0.6373

-0.5297

66.8

73.7

87.0

-0.1800

-0.5063

Patient location

Large central metropolitan

15.4

24.5

13.9

-0.2046

0.0665

15.4

17.7

15.2

-0.0559

0.0062

Large fringe metropolitan

20.4

22.2

24.6

-0.0912

-0.1470

20.4

19.8

21.9

0.0147

-0.0353

Medium metropolitan

21.4

20.9

28.8

-0.0188

-0.2025

21.4

21.1

25.5

0.0066

-0.0968

Small metropolitan

11.0

7.9

9.1

0.1545

0.1117

11.0

10.7

10.2

0.0089

0.0250

Micropolitan

17.5

13.6

12.0

0.1226

0.1710

17.5

17.2

14.2

0.0083

0.0936

Not metropolitan or

12.5

9.1

10.8

0.1189

0.0636

12.5

12.4

11.9

0.0035

0.0200

micropolitan

Median household income Q1 (0-25%)

29.7

35.4

26.2

-0.1289

0.0712

29.7

32.0

27.5

-0.0492

0.0486

Q2 (25%-50%)

27.8

26.3

32.4

0.0729

-0.0597

27.8

27.5

32.1

0.0068

-0.0937

Q3 (50%-75%)

21.9

20.6

21.4

0.0400

0.0199

21.9

21.1

21.5

0.0193

0.0092

Q4 (75%-100%)

16.8

13.4

17.3

0.0280

-0.0800

16.8

16.5

16.1

0.0095

0.0199

Primary expected payer Medicare

5.5

4.7

3.4

0.0140

0.0786

5.5

5.5

4.0

0.0000

0.0714

Medicaid

12.2

16.7

14.5

-0.1259

-0.0658

12.2

12.2

13.1

-0.0009

-0.0265

Private including HMO

43.4

47.6

55.2

-0.0810

-0.2338

43.4

43.3

52.5

0.0014

-0.1837

Self-pay

26.1

18.9

16.3

0.1716

0.2395

26.1

25.8

19.0

0.0062

0.1761

No charge

1.2

1.7

0.3

-0.0572

0.0971

1.2

1.2

0.3

0.0000

0.1095

Other

10.8

9.2

10.0

0.0646

0.0354

10.8

10.7

10.9

0.0050

-0.0013

(continued on next page)

Balance checking, ISS: 25-75, H2

(continued)

Appendix (continued)

Original sample Matched sample

Mean/percent Standardized difference Mean/percent Standardized difference

NTC

TC I

TC II

NTC vs

NTC vs

NTC

TC I

TC II

NTC vs

NTC vs

(N = 2659)

(N = 13383)

(N = 3158)

TC I

TC II

(N = 2659)

(N = 2659)

(N = 2659)

TC I

TC II

Age (mean, yrs) Mean

38.3

37.2

37.8

0.0795

0.0374

38.3

38.2

38.5

0.0045

-0.0140

SD

(13.7)

(13.6)

(13.9)

(13.7)

(13.6)

(13.9)

ISS (mean) 0.1184 0.1065 0.0461 0.0791

Mean

34.1

32.5

32.6

34.1

33.5

33.0

SD

(16.6)

(10.9)

(11.9)

(16.6)

(14.9)

(12.7)

Female

26.4

25.2

25.0

0.0415

0.0445

26.4

23.9

25.5

0.0566

0.0188

With chronic condition

49.5

70.9

71.0

-0.4103

-0.4133

49.5

55.6

67.4

-0.1278

-0.3755

Multiple injury

68.1

91.4

87.5

-0.7269

-0.5982

68.1

74.0

87.1

-0.1556

-0.4808

Patient location

Large central metropolitan

16.3

24.9

14.0

-0.2121

0.0672

16.3

18.4

15.3

-0.0527

0.0282

Large fringe metropolitan

18.5

22.6

25.1

-0.0593

-0.1202

18.5

18.4

21.3

0.0037

-0.0671

Medium metropolitan

20.1

20.8

29.1

-0.0094

-0.2025

20.1

19.8

26.0

0.0084

-0.1363

Small metropolitan

12.5

7.7

8.6

0.1382

0.1069

12.5

12.5

10.5

0.0010

0.0642

Micropolitan

18.1

13.5

11.4

0.1147

0.1769

18.1

17.4

14.0

0.0197

0.1162

Not metropolitan or micropolitan

12.8

8.7

10.9

0.1118

0.0374

12.8

12.6

12.0

0.0058

0.0267

Median household income Q1 (0-25%)

29.4

35.4

26.7

-0.1345

0.0533

29.4

33.1

27.1

-0.0790

0.0520

Q2 (25%-50%)

29.6

26.3

32.0

0.0665

-0.0586

29.6

28.9

33.0

0.0167

-0.0731

Q3 (50%-75%)

22.2

20.6

21.3

0.0364

0.0194

22.2

21.3

22.1

0.0229

0.0029

Q4 (75%-100%)

14.4

13.4

17.5

0.0639

-0.0495

14.4

14.2

15.0

0.0052

-0.0175

Primary expected payer Medicare

5.0

4.8

3.6

0.0242

0.0834

5.0

4.9

4.0

0.0033

0.0470

Medicaid

12.3

16.9

15.1

-0.1074

-0.0589

12.3

12.3

13.1

0.0000

-0.0244

Private including HMO

43.6

47.0

54.9

-0.0985

-0.2582

43.6

43.7

52.1

-0.0022

-0.1726

Self-pay

26.1

19.0

16.2

0.1903

0.2643

26.1

25.9

19.0

0.0036

0.1742

No charge

1.0

1.7

0.2

-0.0447

0.1137

1.0

1.0

0.3

0.0000

0.0907

Other

11.1

9.4

9.7

0.0374

0.0265

11.1

11.2

11.2

-0.0013

-0.0013

Abbreviations: ISS, injury severity score; NTC, nontrauma center; TC I, level I trauma center; TC II, level II trauma center. Each T subgroup represents a third of randomly divided patients with ISS 16 to 24 treated at a nontrauma center. Each H subgroup represents a half of randomly divided patients with ISS 25 to 75 treated at a nontrauma center.

References

  1. Celso B, Tepas J, Langland-Orban B, Pracht E, Papa L, Lottenberg L, 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(2):371-8 [Epub 2006/03/02].
  2. Utter GH, Maier RV, Rivara FP, Mock CN, Jurkovich GJ, Nathens AB. Inclusive trauma systems: do they improve triage or outcomes of the severely injured? J Trauma 2006;60(3):529-35 [Epub 2006/03/15].
  3. American College of Surgeons Committee on Trauma. Regional trauma systems: op- timal elements, integration, and assessment: systems consultation guide; 2008[Chi- cago, IL].
  4. American College of Surgeons Committee on Trauma. Resources for Optimal care of the injured patient 2006; 2006[Chicago, IL].
  5. Haas B, Stukel TA, Gomez D, Zagorski B, De Mestral C, Sharma SV, et al. The Mortality benefit of direct trauma center transport in a regional trauma sys- tem: a population-based analysis. J Trauma Acute Care Surg 2012;72(6): 1510-5.
  6. Haas B, Gomez D, Zagorski B, Stukel TA, Rubenfeld GD, Nathens AB. Survival of the fittest: the hidden cost of undertriage of major trauma. J Am Coll Surg 2010; 211(6):804-11.
  7. Mackenzie EJ, Rivara FP, Jurkovich GJ, Nathens AB, Egleston BL, Salkever DS, et al. The impact of trauma-center care on functional outcomes following major lower-limb trauma. J Bone Joint Surg Am 2008;90(1):101-9.
  8. Centers for Disease Control and Prevention. National Center for Injury Prevention and Control: Web-based Injury Statistics Query and Reporting System (WISQARS). [cited 2014 June 23] http://www.cdc.gov/injury/wisqars/.
  9. Mann NC, Mullins RJ, MacKenzie EJ, Jurkovich GJ, Mock CN. Systematic review of published evidence regarding trauma system effectiveness. J Trauma 1999;47(3): S25-33.
  10. MacKenzie EJ, Rivara FP, Jurkovich GJ, Nathens AB, Frey KP, Egleston BL, et al. A na- tional evaluation of the effect of trauma-center care on mortality. N Engl J Med 2006; 354(4):366-78 [Epub 2006/01/27].
  11. Demetriades D, Martin M, Salim A, Rhee P, Brown C, Chan L. The effect of trauma center designation and trauma volume on outcome in specific severe injuries. Ann Surg 2005;242(4):512-7 [Epub 2005/09/30].
  12. Demetriades D, Martin M, Salim A, Rhee P, Brown C, Doucet J, et al. Relationship be- tween American College of Surgeons trauma center designation and mortality in pa- tients with severe trauma (injury severity score N 15). J Am Coll Surg 2006;202(2): 212-5 [Epub 2006/01/24].
  13. Healthcare Cost and Utilization project. Overview of the Nationwide Emergency De- partment Sample (NEDS). [cited 2014 June 24] http://www.hcup-us.ahrq.gov.
  14. National Trauma Data Bank. National Trauma Data Standard: data dictionary; 2014[http://www.ntdsdictionary.org/].
  15. Rosenbaum PR. The role of a second control group in an observational study. Stat Sci

1987;2(3):292-306.

  1. Rubin DB. Bias reduction using Mahalanobis-metric matching. Matched sampling for

causal effects; 2006 160-5.

  1. Stuart EA. Matching methods for causal inference: a review and a look forward. Stat Sci 2010;25(1):1-21.
  2. Hansen BB. Optmatch: flexible, optimal matching for observational studies. R News 2007;7:18-24.
  3. Rosenbaum PR. Covariance adjustment in randomized experiments and observa- tional studies. Stat Sci 2002;17(3):286-327.
  4. Lu B, Qian Z, Cunningham A, Li C-L. Estimating the effect of premarital cohabitation on timing of marital disruption: using propensity score matching in event history analysis. Sociol Method Res 2012;41(3):440-66 [Epub June 29, 2012].
  5. Cudnik MT, Newgard CD, Sayre MR, Steinberg SM. Level I versus level II trauma cen- ters: an outcomes-based assessment. J Trauma 2009;66(5):1321-6.
  6. Rogers FB, Osler T, Lee JC, Sakorafas L, Wu D, Evans T, et al. In a mature trauma sys- tem, there is no difference in outcome (survival) between level I and level II trauma centers. J Trauma 2011;70(6):1354-7.
  7. Cudnik MT, Sayre MR, Hiestand B, Steinberg SM. Are all trauma centers created equally? A statewide analysis. Acad Emerg Med 2010;17(7):701-8.
  8. Staudenmayer K, Lin F, Mackersie R, Spain D, Hsia R. Variability in California triage from 2005 to 2009: a population-based longitudinal study of severely injured pa- tients. J Trauma Acute Care Surg 2014;76(4):1041-7.
  9. Mohan D, Rosengart MR, Farris C, Cohen E, Angus DC, Barnato AE. Assessing the fea- sibility of the American College of Surgeons’ benchmarks for the triage of trauma pa- tients. Arch Surg 2011;146(7):786-92 [Epub 2011/03/23].
  10. Xiang H, Wheeler KK, Groner JI, Shi J, Haley KJ. Undertriage of major trauma patients in the US EDs. Am J Emerg Med 2014;32(9):997-1004.
  11. Cook A, Weddle J, Baker S, Hosmer D, Glance L, Friedman L, et al. A comparison of the injury severity score and the trauma Mortality prediction model. J Trauma Acute Care Surg 2014;76(1):47-52.
  12. Kohn MA, Hammel JM, Bretz SW, Stangby A. trauma team activation criteria as pre- dictors of patient disposition from the emergency department. Acad Emerg Med 2004;11(1):1-9 [Epub 2004/01/08].
  13. Baker SP, O’Neill B, Haddon Jr W, Long WB. The injury severity score: a method for describing patients with multiple injuries and evaluating emergency care. J Trauma 1974;14(3):187-96.
  14. Sasser SM, Hunt RC, Faul M, Sugerman D, Pearson WS, Dulski T, et al. Guidelines for field triage of injured patients: recommendations of the National Expert Panel on Field Triage, 2011. MMWR Recomm Rep 2012;61(RR-1):1-20.

Leave a Reply

Your email address will not be published. Required fields are marked *