Abstract
Background
African–Americans have worse outcomes than Caucasians in many clinical conditions studied, including trauma. We sought to analyze if mortality is different in these groups through analysis of a national data set.
Methods
Recent data from the national Trauma Quality Improvement Program were assessed with analysis, including all African–American or Caucasian patients who were brought to level I or level II trauma centers for care. Propensity scores were calculated for each African–American patient using age, sex, Injury Severity Score (ISS), Glasgow Coma Scale (GCS), injury type, insurance information and American College of Surgeons trauma level. The primary outcome of this study was in-hospital mortality, and the secondary outcomes were hospital length of stay and discharge disposition.
Results
A total of 82 150 (13.65%) out of 601 768 patients who qualified for the inclusion in the study were African–American. The remaining 519 618 (86.35%) were Caucasian. The median age (IQR) of the patients was 54 (33 to 72) years old, and approximately two-thirds of the patients were male. The median ISS and GCS score were 12 (9 to 17) and 15 (15 to 15), respectively. More than 90% of patients sustained blunt injuries. Overall, there was no significant difference found in overall in-hospital mortality between Caucasians and African–American patients (3% vs. 2.9%, p=0.2); however, the median (95% CI) hospital length of stay was 1 day longer in African–American patients compared with Caucasian patients (5 (5.5) vs. 4 (4.4), p<0.001). When the discharged destinations between the two groups were compared, a higher proportion of Caucasians were discharged to home without services (66% vs. 33%).
Conclusion
Our study showed that trauma mortalites among African–American and Caucasians are the same. Efforts to mitigate the ethnic and racial biases in the delivery of healthcare should continue, and these results (no differences in mortality) should be validated in other clinical settings.
Level of evidence
Level II.
Introduction
The disparity in healthcare delivery and outcomes among ethnic minorities in the USA is not a new topic of discussion. Efforts have been implemented to mitigate ethnic and racial bias, including education and training, improvement of equity in access to care, enhancement of regulatory vigilance, monitoring of managed care and improvement of data systems to monitor the quality of medical care.1 These efforts have met with variable success, and recently, an analysis of hospital data accountability found that the algorithm used for the assigned risk score for the patients resulted in less access to specialized care for African–American patients.2 Disparities continue to exist.
Many studies have evaluated the impact of race on clinical outcomes. African–American race represents a significant risk factor for adverse outcome in many clinical diseases.3–7 When insurance coverage is taken into account, disparities in outcome between races remain.8 Acute trauma care represents a unique situation where delivery of healthcare is expected to be provided immediately without the patient’s race, ethnicity or insurance coverage. Compared with Caucasian patients, a recent study showed that there is increased mortality among African–American trauma patients.9 However, another recent population-based study found no significant difference in mortality between Caucasian and African–American patients after a traumatic injury.10
The reasons behind the variability in mortality in different studies are not well understood. One explanation is that it may be the existence of implicit bias at the point of care towards certain races that affect outcome11 or the use of different variables and methodologies in the NTDB (National Trauma Data Bank) adopted in these studies to draw conclusions.12 This study was designed to answer the same question, whether African–American race has adversely impacted outcomes when compared with Caucasian counterparts after an injury using recent data from national Trauma Quality Improvement Program (TQIP) data set, and exact matching methodology was adopted to balance the characteristics.
Methods
Data sources and data extraction
The adult TQIP database from 2011 to 2016 was accessed for the study. The TQIP is considered a quality improvement program by the American College of Surgeons (ACS). More than 825 trauma centers all across the USA currently participate in the TQIP program (https://www.facs.org/quality-programs/trauma/tqp/center-programs/tqip; accessed date December 3, 2019). The TQIP also provides feedback two times a year to participating institutions on certain quality indicators. All African–Americans and Caucasians who sustained injury and were treated at level I and level II trauma centers were included in the study. Other patient characteristics included age (between 16 and 89 years), sex, InjurySeverity Score (ISS), injury type, Abbreviated Injury Scale (AIS) of different body regions, Glasgow Coma Scale (GCS), initial vital signs (heart rate and initial systolic blood pressure (iSBP), presence of hypotension on initial presentation (defined as iSBP <90 mm Hg) and whether insurance coverage was present. Patients aged <16 years old and identified as members of other races (eg, Hispanic or Asian) were excluded from the study. Patients who presented with no signs of life were also excluded from the study (please see table 1).
Table 1.
Characteristics of patients before matching
| Variable | Unmatched analysis | Exact matched analysis | P value | ||||
| Values | Not African–American (n=519 618) |
African–American (n=82 150) | P values | Not African–American (n=56 161) |
African–American (n=56 161) |
||
| Age | Median (Q1 to Q3) | 57 (37 to 74) | 37 (25 to 55) | <0.001 | 42 (27 to 59) | 42(27 to 59) | 1 |
| Sex | Female | 197 616 (38) | 20 014 (24.4) | <0.001 | 14 808 (26.4) | 14 808 (26.4) | 1 |
| Male | 322 002 (62) | 62 136 (75.6) | 41 353 (73.6) | 41 353 (73.6) | |||
| Hypotension | 0 | 503 073 (96.8) | 77 937 (94.9) | <0.001 | 55 546 (98.9) | 55 546 (98.9) | 1 |
| 1 | 16 545 (3.2) | 4213 (5.1) | 615 (1.1) | 615 (1.1) | |||
| ISS | Median (Q1 to Q3) | 12 (9 to 17) | 12 (9 to 18) | <0.001 | 10 (9 to 17) | 10 (9 to 17) | 1 |
| GCS score | Median (Q1 to Q3) | 15 (15 to 15) | 15 (14 to 15) | <0.001 | 15 (15 to 15) | 15 (15 to 15) | 1 |
| Injury type | Blunt | 495 856 (95.4) | 56 859 (69.2) | <0.001 | 46 300 (82.4) | 46 300 (82.4) | 1 |
| Penetrating | 23 762 (4.6) | 25 291 (30.8) | 9861 (17.6) | 9861 (17.6) | |||
| Insurance | Blue Cross/Blue Shield | 23 915 (4.6) | 2181 (2.7) | <0.001 | 1433 (2.6) | 1433 (2.6) | 1 |
| Medicaid | 49 081 (9.4) | 19 469 (23.7) | 11 558 (20.6) | 11 558 (20.6) | |||
| Medicare | 175 971 (33.9) | 13 101 (15.9) | 11 175 (19.9) | 11 175 (19.9) | |||
| No fault automobile | 25 867 (5) | 2937 (3.6) | 2086 (3.7) | 2086 (3.7) | |||
| Not billed (for any reason) | 1245 (0.2) | 492 (0.6) | 126 (0.2) | 126 (0.2) | |||
| Other | 15 181 (2.9) | 2721 (3.3) | 1420 (2.5) | 1420 (2.5) | |||
| Other government | 11 893 (2.3) | 2573 (3.1) | 1422 (2.5) | 1422 (2.5) | |||
| Private/commercial insurance | 152 989 (29.4) | 18 005 (21.9) | 14 039 (25) | 14 039 (25) | |||
| Self-pay | 54 497 (10.5) | 19 887 (24.2) | 12 353 (22) | 12 353 (22) | |||
| Workers’ compensation | 8979 (1.7) | 784 (1) | 549 (1) | 549 (1) | |||
| ACS trauma level | <0.001 | 1 | |||||
| I | 349 881 (67.3) | 64 041 (78) | 44 347 (79) | 44 347 (79) | |||
| II | 169 737 (32.7) | 18 109 (22) | 11 814 (21) | 11 814 (21) | |||
Hypotension denotes a systolic blood pressure of <90 mm Hg.
ACS, American College of Surgeons; GCS, Glasgow Coma Scale; ISS, Injury Severity Score.
Outcomes
The primary outcome of this study was in-hospital mortality and the secondary outcomes were hospital length of stay and discharge disposition.
Statistical methods
Univariate analysis
The patient’s demography, injury, insurance information and outcomes were compared between the African–American and Caucasian groups. Data were summarized using summary statistics (median with IQR (first quartile to third quartile) for continuous variables, and frequency and percentage for categorical variables) as described previously.13 The two groups were compared using the Wilcoxon rank-sum test for the continuous variables and the χ2 test for the categorical variables.13 Since there were significant differences identified in univariate analysis on baseline patient characteristics, therefore matching methodology was used to balance the two groups.
Exact matching and pair matched analysis
The first step in performing the exact matching was to calculate the estimated propensity score for each African–American patient using age, sex, ISS, GCS score, injury type, insurance information and ACS trauma level. Then, the one-to-one exact matching was performed using the R package ‘MatchIt’.14 Again, the data were summarized using summary statistics as described previously. The pair matched analyses were performed using the Wilcoxon signed-rank test to compare the groups on the continuous variables and McNemar’s test for categorical variables. If there was more than two categories, the Stuart-Maxwell test was performed as described previously.13 The Kaplan-Meier procedure was used to estimate the median time, and the SE was estimated using Greenwood’s formula.15 Kaplan-Meier curves were generated. The log-rank test was used to compare the time (Kaplan-Meier curves) between groups as described previously.13 All p values were two sided and values of <0.05 were considered statistically significant. All statistical analyses were performed using the R language.16
Results
A total of 82 150 (13.65%) out of 601 768 patients who qualified for the inclusion in the study were African–American. The remaining 519 618 (86.35%) were Caucasian. The median age of the patients was 54 (IQR 33 to 72) years old, and approximately two-thirds of the patients were male. The median ISS and GCS score were 12 (IQR 9 to 17) and 15 (IQR 15 to 15), respectively. More than 90% of patients sustained blunt injuries. Head injury and extremity injuries were the two most common organ-specific injuries sustained by these patients (please see tables 1 and 2).
Table 2.
AbbreviatedInjury Scale score of ≥2 different body regions
| Variable | Values | All (N=601 768) |
Not African–American (n=519 618) |
African–American (n=82 150) |
| Abdomen | 0 | 531 137 (88.3) | 463 287 (89.2) | 67 850 (82.6) |
| 1 | 70 631 (11.7) | 56 331 (10.8) | 14 300 (17.4) | |
| Brain | 0 | 386 823 (64.3) | 329 046 (63.3) | 57 777 (70.3) |
| 1 | 214 945 (35.7) | 190 572 (36.7) | 24 373 (29.7) | |
| Face | 0 | 550 195 (91.4) | 475 352 (91.5) | 74 843 (91.1) |
| 1 | 51 573 (8.6) | 44 266 (8.5) | 7307 (8.9) | |
| Neck | 0 | 593 602 (98.6) | 513 082 (98.7) | 80 520 (98) |
| 1 | 8166 (1.4) | 6536 (1.3) | 1630 (2) | |
| Pelvis | 0 | 571 990 (95.1) | 494 321 (95.1) | 77 669 (94.5) |
| 1 | 29 778 (4.9) | 25 297 (4.9) | 4481 (5.5) | |
| Spine | 0 | 586 322 (97.4) | 507 136 (97.6) | 79 186 (96.4) |
| 1 | 15 446 (2.6) | 12 482 (2.4) | 2964 (3.6) | |
| Thorax | 0 | 410 006 (68.1) | 355 054 (68.3) | 54 952 (66.9) |
| 1 | 191 762 (31.9) | 164 564 (31.7) | 27 198 (33.1) | |
| Extremities | 0 | 385 543 (64.1) | 332 477 (64) | 53 066 (64.6) |
| 1 | 216 225 (35.9) | 187 141 (36) | 29 084 (35.4) |
Unmatched (univariate) analysis
There were significant differences found between the Caucasians and African–American groups with respect to age (57 (37 to 74) vs. 37 (25 to 55)), male gender (62% vs. 75.6%), ISS (12 (9 to 17) vs. 12 (9 to 18)), penetrating trauma (4.6% vs. 30.8%) and insurance coverage, respectively. All p values were <0.001 (please see table 1). When the two groups were compared on AIS score of ≥2 on different body regions, the African–Americans sustained more torso injuries, whereas the Caucasians suffered from more head injury (please see table 2 for detailed information).
Exact matching and pair matched analysis
After one-to-one exact matching, each group contained 56 161 patients. No difference existed on the baseline patient’s characteristics after matching (please see table 1). The median age, ISS and GCS score in both groups were 42 (27 to 59) years, 10 (9 to 17) and 15 (15 to 15). However, after matching, the proportion of most of the body region injuries between the groups was balanced except extremity injury. A higher proportion of African–American patients sustained extremity injury compared with their Caucasians counterpart (37.7% vs. 34.2%; see table 3).
Table 3.
AbbreviatedInjury Scale score >2 of different body regions
| Variable | Values | All (N=112 322) |
Not African–American (n=56 161) |
African–American (n=56 161) |
| Abdomen | 0 | 98 339 (87.6) | 49 157 (87.5) | 49 182 (87.6) |
| 1 | 13 983 (12.4) | 7004 (12.5) | 6979 (12.4) | |
| Brain | 0 | 78 805 (70.2) | 39 422 (70.2) | 39 383 (70.1) |
| 1 | 33 517 (29.8) | 16 739 (29.8) | 16 778 (29.9) | |
| Face | 0 | 103 794 (92.4) | 52 133 (92.8) | 51 661 (92) |
| 1 | 8528 (7.6) | 4028 (7.2) | 4500 (8) | |
| Neck | 0 | 110 753 (98.6) | 55 339 (98.5) | 55 414 (98.7) |
| 1 | 1569 (1.4) | 822 (1.5) | 747 (1.3) | |
| Pelvis | 0 | 106 940 (95.2) | 53 737 (95.7) | 53 203 (94.7) |
| 1 | 5382 (4.8) | 2424 (4.3) | 2958 (5.3) | |
| Spine | 0 | 109 326 (97.3) | 54 882 (97.7) | 54 444 (96.9) |
| 1 | 2996 (2.7) | 1279 (2.3) | 1717 (3.1) | |
| Thorax | 0 | 78 062 (69.5) | 38 545 (68.6) | 39 517 (70.4) |
| 1 | 34 260 (30.5) | 17 616 (31.4) | 16 644 (29.6) | |
| Extremities | 0 | 71 937 (64) | 36 952 (65.8) | 34 985 (62.3) |
| 1 | 40 385 (36) | 19 209 (34.2) | 21 176 (37.7) |
Outcomes
There were no significant difference in overall in-hospital mortality between Caucasians and African–American patients (3% vs. 2.9%, p=0.2); however, the median (95% CI) hospital length of stay was 1 day longer in African–American patients compared with Caucasians (5 (5.5) vs. 4 (4.4), p<0.001) (please see table 4). When the discharged destinations between the two groups were compared, a higher proportion of Caucasians were discharged to home without services (66% vs. 33%; please see table 5).
Table 4.
Outcomes
| Variable | Before matching | All (N=601 768) |
Not African–American (n=519 618) |
African–American (n=82 150) |
P value | OR (95% CI) | Absolute risk difference (95% CI) |
| Total death | 0 | 569 649 (94.7) | 491 872 (94.7) | 77 777 (94.7) | 0.852 | 0.997 (0.964 to 1.030) |
−0.00017 (−0.0018 to 0.0015) |
| 1 | 32 119 (5.3) | 27 746 (5.3) | 4373 (5.3) | ||||
| Exact matched |
All (N=112 322) |
Not African–American (n=56 161) |
African–American (n=56 161) |
||||
| Total death | 0 | 109 014 (97.1) | 54 477 (97) | 54 537 (97.1) | 0.2 | 0.95 (0.87 to 1.03) | −0.001 (−0.003 to 0.001) |
| 1 | 3308 (2.9) | 1684 (3) | 1624 (2.9) |
Table 5.
Discharged disposition
| Variable | Values | All (N=109 014) |
Not African–American (n=54 477) |
African–American (n=54 537) |
P value |
| Hospital disposition | Another hospital | 3275 (3) | 1713 (3.1) | 1562 (2.9) | <0.001 |
| Home: healthcare | 7958 (7.3) | 3493 (6.4) | 4465 (8.2) | ||
| Home: no services | 70 763 (64.9) | 36 293 (66.6) | 34 470 (63.2) | ||
| Hospice care | 370 (0.3) | 207 (0.4) | 163 (0.3) | ||
| Intermediate care | 826 (0.8) | 325 (0.6) | 501 (0.9) | ||
| Left against medical advice | 1298 (1.2) | 644 (1.2) | 654 (1.2) | ||
| Long-term care | 5640 (5.2) | 2664 (4.9) | 2976 (5.5) | ||
| Skilled nursing care | 18 884 (17.3) | 9138 (16.8) | 9746 (17.9) |
Discussion
Our study analyzed the recent TQIP data set and found no significant difference in overall in-hospital mortality between African–American and Caucasians trauma victims. However, African–American patients stayed longer in the hospital and required more additional services at the time of hospital discharge.
Most prior studies reported higher mortality among ethnic minority patients after trauma.8 Haider and colleagues examined the NTDB data set almost 10 years ago and included only those patients aged 18–64 years who had an ISS of ≥9 and compared African–Americans, Hispanics and Caucasians.8 They included more than 400 000 patients in their analysis and found higher in-hospital mortality among the African–American and Hispanic populations in unadjusted and adjusted analyses. The uninsured Hispanic patients had the highest odd of mortality followed by African–Americans (2.30, 95% CI 2.13 to 2.49; 1.78, 95% CI 1.65 to 1.90), respectively. The same authors and colleagues later performed a systemic review; 7 out of initially 35 selected studies showed higher odds of mortality among African–American patients compared with Caucasian counterparts (1.19, 95% CI 1.09 to 1.31).9
Contrary to previous studies, our study used the ACS–TQIP database data. ACS sends the feedback to all participating trauma centers all across the USA on some of the quality matrix, including in-hospital mortality, using the same database. Our study clearly showed no significant difference in mortality outcome regardless of the patients being African–American or Caucasian. This result may surprise some of the readers, but the important point to note is that we used very recent national quality data. Furthermore, we also used exact matching methodology to balance the two ethnic groups (black and white patients). We matched insurance coverage as well to determine if disparity in race generated the disparity in mortality (no disparity was noted). This may reflect the positive influence of ongoing efforts to limit implicit or explicit biases at the point of care, at least among trauma providers. A recent survey study among clinicians in an acute care setting found no implicit bias in the decision-making process,17 which may represent more evidence of progress.
Limitation
Although the study was performed using a well-recognized national quality database, the retrospective nature of the study carries inherent limitations. We attempted to include all the available characteristics of the patients that can influence the mortality and did exact matching on one-to-one basis to balance the group; however, exact matching cannot account for unobserved variables.
Conclusion
Our study showed that trauma mortalities among African–Americans and Caucasians are the same. Efforts to mitigate the ethnic and racial biases in the delivery of healthcare should continue, and these results (no differences in mortality) should be validated in other clinical settings.
Footnotes
Contributors: NA performed the statistical analysis, cleaned and analyzed the data, and drafted the article. DK revised the article and is the guarantor.
Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests: None declared.
Patient consent for publication: Not required.
Provenance and peer review: Not commissioned; externally peer reviewed.
References
- 1.Williams DR, Rucker TD, Williams DR, Rucker TD. Understanding and addressing racial disparities in health care. Health Care Financ Rev 2000;21:75–90. [PMC free article] [PubMed] [Google Scholar]
- 2.Ledford H. Millions of black people affected by racial bias in health-care algorithms. Nature 2019;574:608–9. 10.1038/d41586-019-03228-6 [DOI] [PubMed] [Google Scholar]
- 3.Chen AY, Schrag NM, Halpern MT, Ward EM. The impact of health insurance status on stage at diagnosis of oropharyngeal cancer. Cancer 2007;110:395–402. 10.1002/cncr.22788 [DOI] [PubMed] [Google Scholar]
- 4.Richardson LC, Tian L, Voti L, Hartzema AG, Reis I, Fleming LE, Mackinnon J. The roles of teaching hospitals, insurance status, and race/ethnicity in receipt of adjuvant therapy for regional-stage breast cancer in Florida. Am J Public Health 2006;96:160–6. 10.2105/AJPH.2004.053579 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Kelz RR, Gimotty PA, Polsky D, Norman S, Fraker D, DeMichele A. Morbidity and mortality of colorectal carcinoma surgery differs by insurance status. Cancer 2004;101:2187–94. 10.1002/cncr.20624 [DOI] [PubMed] [Google Scholar]
- 6.Roetzheim RG, Pal N, Tennant C, Voti L, Ayanian JZ, Schwabe A, Krischer JP. Effects of health insurance and race on early detection of cancer. J Natl Cancer Inst 1999;91:1409–15. 10.1093/jnci/91.16.1409 [DOI] [PubMed] [Google Scholar]
- 7.Shafi S, de la Plata CM, Diaz-Arrastia R, Bransky A, Frankel H, Elliott AC, Parks J, Gentilello LM. Ethnic disparities exist in trauma care. J Trauma 2007;63:1138–42. 10.1097/TA.0b013e3181568cd4 [DOI] [PubMed] [Google Scholar]
- 8.Haider AH, Chang DC, Efron DT, Haut ER, Crandall M, Cornwell EE. Race and insurance status as risk factors for trauma mortality. Arch Surg 2008;143:945–9. 10.1001/archsurg.143.10.945 [DOI] [PubMed] [Google Scholar]
- 9.Haider AH, Weygandt PL, Bentley JM, Monn MF, Rehman KA, Zarzaur BL, Crandall ML, Cornwell EE, Cooper LA. Disparities in trauma care and outcomes in the United States: a systematic review and meta-analysis. J Trauma Acute Care Surg 2013;74:1195–205. 10.1097/TA.0b013e31828c331d [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Osler T, Glance LG, Li W, Buzas JS, Wetzel ML, Hosmer DW. Trauma care does not discriminate: the association of race and health insurance with mortality following traumatic injury. J Trauma Acute Care Surg 2015;78:1026–33. 10.1097/TA.0000000000000593 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Hall WJ, Chapman MV, Lee KM, Merino YM, Thomas TW, Payne BK, Eng E, Day SH, Coyne-Beasley T. Implicit racial/ethnic bias among health care professionals and its influence on health care outcomes: a systematic review. Am J Public Health 2015;105:e60–76. 10.2105/AJPH.2015.302903 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Haider AH, Saleem T, Leow JJ, Villegas CV, Kisat M, Schneider EB, Haut ER, Stevens KA, Cornwell EE, MacKenzie EJ, et al. Influence of the National trauma data bank on the study of trauma outcomes: is it time to set research best practices to further enhance its impact? J Am Coll Surg 2012;214:756–68. 10.1016/j.jamcollsurg.2011.12.013 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Ahmed N, Greenberg P. Assessing the impact of blood alcohol concentration on the rate of in-hospital mortality following traumatic motor vehicle crash injury: a matched analysis of the National trauma data bank. Injury 2019;50:33–8. 10.1016/j.injury.2018.09.022 [DOI] [PubMed] [Google Scholar]
- 14.Ho DE, Imai K, King G, Stuart EA. MatchIt : Nonparametric Preprocessing for Parametric Causal Inference. J Stat Softw 2011;42:1–28. [Google Scholar]
- 15.Pokhrel A, Dyba T, Hakulinen T. A Greenwood formula for standard error of the age-standardised relative survival ratio. Eur J Cancer 2008;44:441–7. 10.1016/j.ejca.2007.10.026 [DOI] [PubMed] [Google Scholar]
- 16.R Core Team R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. 2018. https://www.R-project.org/.
- 17.Haider AH, Schneider EB, Sriram N, Dossick DS, Scott VK, Swoboda SM, Losonczy L, Haut ER, Efron DT, Pronovost PJ, et al. Unconscious race and social class bias among acute care surgical clinicians and clinical treatment decisions. JAMA Surg 2015;150:457 10.1001/jamasurg.2014.4038 [DOI] [PubMed] [Google Scholar]