Abstract
This study uses the Healthcare Cost and Utilization Project Nationwide Emergency Department Sample to examine the number of preventable trauma deaths in US hospitals to provide a target for the National Academies of Sciences, Engineering, and Medicine mandate of zero preventable deaths after injury.
Trauma remains the leading cause of death in Americans younger than 46 years, leading to nearly 200 000 deaths per year.1 To address this issue, major quality improvement initiatives have been undertaken at the national level to improve care at trauma centers. Although results of these efforts have been generally positive, major variations in outcomes between trauma centers continue to be documented.2 The fact that some hospitals perform better than others suggests that some patients are not receiving the best possible care, which may result in preventable deaths. Recognizing this issue, in its 2016 report, the National Academies of Sciences, Engineering, and Medicine (NASEM) recommended several measures to improve the quality of trauma care, and called for achieving zero preventable deaths after injury.3 Our objective is to determine the number of preventable trauma deaths (PTDs) in US hospitals to provide a specific target for this NASEM mandate.
Methods
We used the Healthcare Cost and Utilization Project Nationwide Emergency Department Sample from 2006 to 2014 to determine weighted national estimates of PTDs.4 All individuals aged 16 years or older with blunt or penetrating injuries and an International Classification of Diseases, Ninth Revision, Clinical Modification primary diagnostic code of 800 to 905 were included. Records documenting late effects, superficial injuries, foreign bodies, interhospital transfers, and missing data on variables of interest were excluded. This work was reviewed and approved by the Johns Hopkins institutional review board. This study used deidentified, publicly available administrative data from the Agency for Healthcare Research and Quality; therefore, no individual participant informed consent was required.
National estimates of PTDs were determined using previously described methods.5 Hospitals were first classified as a high-performing hospital (HPH), average-performing hospital (APH), or low-performing hospital (LPH) based on risk-adjusted observed-to-expected in-hospital mortality ratios using standardized benchmarking methods.6 Generalized linear modeling with Poisson distributed mortality was then performed (adjusting for known demographic and injury variables) to estimate the relative risk (RR) of mortality at HPHs and APHs vs LPHs. Weighted national estimates of PTDs were calculated for the following 3 hypothetical quality improvement scenarios:
The conservative model was used if LPHs improved to an average performance: PTDs = OLPH – (OLPH × RRAPH), where O is the number of observed deaths.
The intermediate model was used if LPHs improved to an average performance and APHs improved to a high performance: PTDs = [OLPH – (OLPH × RRAPH)] + [OAPH – (OAPH × RRHPH)].
The best-case model was used if all hospitals improved to a high performance: PTDs = OLPH + APH – (OLPH + APH × RRHPH).
All analyses were performed using Stata, version 12/MP (StataCorp). P < .05 (2-sided) was considered statistically significant.
Results
A total of 18 082 170 patients from 2198 hospitals were analyzed. Performance benchmarking classified 261 facilities (11.9%) as HPHs, 1755 facilities (79.9%) as APHs, and 182 facilities (8.3%) as LPHs (Table 1). More than one-third of patients were treated at LPHs (2 689 177 [14.9%]) and HPHs (3 792 368 [21.0%]). The overall unadjusted mortality rate was 0.4% (n = 64 415); among patients with an Injury Severity Score of 9 or more the mortality rate was 3.8% (n = 52 401 of 1 366 789). Patients at APHs (RR, 0.65; 95% CI, 0.63-0.66) and HPHs (RR, 0.38; 95% CI, 0.37-0.39) had a substantially lower RR of mortality compared with those at LPHs. In addition, LPHs had significantly higher mortality rates across all demographic and injury profiles compared with APHs and HPHs (Table 1).
Table 1. Distribution of Patients and Hospitals by In-Hospital Mortality–Based Hospital Performance Status.
| Characteristic | Hospital Performance Statusa | P Valueb | ||
|---|---|---|---|---|
| Low Performing | Average Performing | High Performing | ||
| Hospitals, No./total No. (%) | 182/2198 (8.3) | 1755/2198 (79.9) | 261/2198 (11.9) | NA |
| Patients, No./total No. (%) | 2 689 177/18 082 170 (14.9) | 11 600 625/18 082 170 (64.2) | 3 792 368/18 082 170 (21.0) | NA |
| Weighted sample, No./total No. (%) | 12 739 850/81 304 301 (15.7) | 52 345 320/81 304 301 (64.4) | 16 219 131/81 304 301 (20.0) | NA |
| Risk-adjusted mortality rate, % | 0.7 | 0.4 | 0.2 | NA |
| Relative risk of mortality (95% CI) | 1 [Reference] | 0.65 (0.63-0.66) | 0.38 (0.37-0.39) | NA |
| Deaths, No./total No. (%) | 23 475/2 689 177 (0.9) | 34 203/11 600 625 (0.3) | 6737/3 792 368 (0.2) | |
| Age, y | ||||
| 16-25 | 3691/691 487 (0.5) | 3591/2 731 102 (0.1) | 533/815 590 (0.1) | <.001 |
| 26-35 | 2563/503 657 (0.5) | 2510/2 045 569 (0.1) | 377/646 972 (0.1) | |
| 36-45 | 2098/412 932 (0.5) | 2105/1 732 032 (0.1) | 298/549 582 (0.1) | |
| 46-55 | 2472/387 468 (0.6) | 2685/1 633 580 (0.2) | 395/534 204 (0.1) | |
| 56-64 | 2148/234 531 (0.9) | 2500/1 068 129 (0.2) | 455/359 706 (0.1) | |
| 65-75 | 2612/183 984 (1.4) | 3941/933 506 (0.4) | 807/326 324 (0.3) | |
| 76-85 | 4088/164 246 (2.5) | 7798/857 532 (0.9) | 1674/321 038 (0.5) | |
| 86-100 | 3633/109 690 (3.3) | 8941/595 438 (1.5) | 2181/237 717 (0.9) | |
| >100 | 170/1182 (14.4) | 132/3737 (3.5) | 17/1235 (1.4) | |
| Sex | ||||
| Male | 15 944/1 508 069 (1.1) | 20 491/6 039 801 (0.3) | 3894/1 893 749 (0.2) | <.001 |
| Female | 7531/1 181 108 (0.6) | 13 712/5 560 824 (0.3) | 2843/1 898 619 (0.2) | |
| Insurance status | ||||
| Insured | 18 802/2 112 418 (0.9) | 29 094/9 377 262 (0.3) | 5984/3 061 872 (0.2) | <.001 |
| Uninsured | 4673/576 759 (0.8) | 5109/2 223 363 (0.2) | 753/730 496 (0.1) | |
| Intent of injury | ||||
| Unintentional | 19 005/2 461 786 (0.8) | 29 403/11 024 389 (0.3) | 5974/3 617 696 (0.2) | <.001 |
| Self-inflicted | 1215/18 030 (6.7) | 1642/48 241 (3.4) | 230/11 715 (2.0) | |
| Assault | 2627/195 226 (1.4) | 2522/499 842 (0.5) | 438/154 848 (0.3) | |
| Undetermined | 526/6277 (8.4) | 528/9670 (5.5) | 83/2721 (3.1) | |
| Other | 102/7858 (1.3) | 108/18 483 (0.6) | 12/5388 (0.2) | |
| Type of injury | ||||
| Penetrating | 5025/462 773 (1.1) | 5448/2 116 323 (0.3) | 896/645 509 (0.1) | <.001 |
| Blunt | 18 450/2 226 404 (0.8) | 28 755/9 484 302 (0.3) | 5841/3 146 859 (0.2) | |
| Mechanism of injuryc | ||||
| Stab | 660/420 435 (0.2) | 664/2 059 478 (0.03) | 102/630 040 (0.02) | <.001 |
| Fall | 9429/1 021 530 (0.9) | 19 868/4 905 406 (0.4) | 4642/1 673 656 (0.3) | |
| Firearm | 4365/42 338 (10.3) | 4784/56 845 (8.4) | 794/15 469 (5.1) | |
| MVC | 8302/688 439 (1.2) | 8073/2 327 648 (0.4) | 1062/747 850 (0.1) | |
| Injury Severity Score | ||||
| <9 | 4192/2 358 283 (0.2) | 6585/10 813 562 (0.1) | 1237/3 543 536 (0.03) | <.001 |
| 9-15 | 7366/233 416 (3.2) | 13 836/645 526 (2.1) | 2944/208 771 (1.4) | |
| 16-24 | 6133/74 799 (8.2) | 7665/116 335 (6.6) | 1598/34 934 (4.6) | |
| 25-75 | 5784/22 679 (25.5) | 6117/25 202 (24.3) | 958/5127 (18.7) | |
| Severe head injuryd | 12 268/110 547 (11.1) | 14 645/181 357 (8.1) | 2894/55 628 (5.2) | <.001 |
| Charlson Comorbidity Index | ||||
| 0 | 15 435/2 360 716 (0.7) | 17 720/10 165 319 (0.2) | 2863/3 246 696 (0.1) | <.001 |
| 1 | 3542/229 541 (1.5) | 6092/1 020 257 (0.6) | 1358/382 084 (0.4) | |
| 2 | 1955/57 213 (3.4) | 4245/245 690 (1.7) | 985/94 479 (1.0) | |
| ≥3 | 2543/41 707 (6.1) | 6146/169 359 (3.6) | 1531/69 109 (2.2) | |
Abbreviations: MVC, motor vehicle collision; NA, not applicable.
Model adjusted for age, sex, insurance status, intent of injury, mechanism of injury, trauma mortality prediction model, severe head injury (head Abbreviated Injury Scale score, ≥3), and Charlson Comorbidity Index.
Significant at P<.05, comparing patients at low-performing with average-performing and high-performing hospitals, respectively.
Some categories not shown owing to low cell counts.
Head Abbreviated Injury Scale score, ≥3.
Table 2 describes the weighted national estimates of PTDs. If all hospitals were to deliver the highest quality of care, an estimated 167 746 (95% CI, 164 534-170 861) lives could potentially be saved.
Table 2. National Estimates of Preventable Trauma Deaths.
| Estimate | Estimated No. of Preventable Trauma Deaths (95% CI) | ||
|---|---|---|---|
| Conservative Modela | Intermediate Modelb | Best-Case Modelc | |
| Overall 9-y estimate | 40 129 (38 529-41 696) | 137 310 (133 849-140 681) | 167 746 (164 534-170 861) |
| Per year estimate | 4459 (4281-4633) | 15 257 (14 872-15 361) | 18 638 (18 281-18 985) |
If low-performing hospitals improve to average-performing status.
If low-performing hospitals improve to average-performing status, and average-performing hospitals improve to high-performing status.
If all hospitals improve to high-performing status.
Discussion
This nationally representative evaluation of trauma data suggests that if all US hospitals achieved outcomes similar to those at the highest-performing centers, 100 000 lives could be saved in approximately 5 years.
The limitations of this study arise from use of retrospective administrative data, including a lack of information on severity of physiological injury and inability to identify systemic factors to distinguish HPHs and LPHs. In addition, in the absence of specifics of each death, our estimates should be interpreted as potentially preventable deaths.
Although this study focuses on preventable in-hospital trauma deaths (owing to differential quality of care), a substantial proportion of potentially preventable deaths may occur in the prehospital setting (attributable to differential access to care). As efforts are being made to standardize prehospital care and improve access, the fact that thousands of lives could be saved by implementing the existing highest standards of care at our hospitals harkens back to NASEM’s call for “the nation to improve its approach to trauma care.”3(p32) Whereas dedicated initiatives in the last few decades have substantially reduced trauma mortality, future efforts should focus on ensuring that all patients receive the best possible trauma care. The results of this study provide a tangible target for such efforts, and serve as a baseline to evaluate future gains.
References
- 1.Rhee P, Joseph B, Pandit V, et al. Increasing trauma deaths in the United States. Ann Surg. 2014;260(1):13-21. [DOI] [PubMed] [Google Scholar]
- 2.Nathens AB, Cryer HG, Fildes J. The American College of Surgeons Trauma Quality Improvement Program. Surg Clin North Am. 2012;92(2):441-454, x-xi. [DOI] [PubMed] [Google Scholar]
- 3.Committee on Military Trauma Care’s Learning Health System and Its Translation to the Civilian Sector; Board on Health Sciences Policy; Board on the Health of Select Populations; Health and Medicine Division; National Academies of Sciences, Engineering, and Medicine; Berwick D, Downey A, Cornett E, eds. A National Trauma Care System: Integrating Military and Civilian Trauma Systems to Achieve Zero Preventable Deaths After Injury. Washington, DC: National Academies of Sciences, Engineering, and Medicine; 2016. doi: 10.17226/23511 [DOI] [PubMed] [Google Scholar]
- 4.Healthcare Cost and Utilization Project, Agency for Healthcare Research and Quality NEDS overview. https://www.hcup-us.ahrq.gov/nedsoverview.jsp. Updated December 14, 2017. Accessed January 29, 2018.
- 5.Fiscella K, Franks P, Gold MR, Clancy CM. Inequality in quality: addressing socioeconomic, racial, and ethnic disparities in health care. JAMA. 2000;283(19):2579-2584. [DOI] [PubMed] [Google Scholar]
- 6.Newgard CD, Fildes JJ, Wu L, et al. Methodology and analytic rationale for the American College of Surgeons Trauma Quality Improvement Program. J Am Coll Surg. 2013;216(1):147-157. [DOI] [PubMed] [Google Scholar]