Abstract
Objective:
To determine if Black race is associated with worse short-term postoperative morbidity and mortality when compared to White race in a contemporary, cross-specialty-matched cohort.
Background:
Growing evidence suggests poorer outcomes for Black patients undergoing surgery.
Methods:
A retrospective analysis was conducted comprising of all patients undergoing surgery in the National Surgical Quality Improvement Program dataset between 2012 and 2018. One-to-one coarsened exact matching was conducted between Black and White patients. Primary outcome was rate of 30-day morbidity and mortality.
Results:
After 1:1 matching, 615,118 patients were identified. Black race was associated with increased rate of all-cause morbidity (odds ratio [OR] = 1.10, 95% confidence interval [CI] 1.08–1.13, P < 0.001) and mortality (OR = 1.15, 95% CI 1.01–1.31, P = 0.039). Black race was associated with increased risk of re-intubation (OR = 1.33, 95% CI 1.21–1.48, P < 0.001), pulmonary embolism (OR = 1.55, 95% CI 1.40–1.71, P < 0.001), failure to wean from ventilator for >48 hours (OR = 1.14, 95% CI 1.02–1.29, P < 0.001), progressive renal insufficiency (OR = 1.63, 95% CI 1.43–1.86, P < 0.001), acute renal failure (OR = 1.39, 95% CI 1.16–1.66, P < 0.001), cardiac arrest (OR = 1.47, 95% CI 1.24–1.76 P < 0.001), bleeding requiring transfusion (OR = 1.39, 95% CI 1.34–1.43, P < 0.001), DVT/thrombophlebitis (OR = 1.24, 95% CI 1.14–1.35, P < 0.001), and sepsis/septic shock (OR = 1.09, 95% CI 1.03–1.15, P < 0.001). Black patients were also more likely to have a readmission (OR = 1.12, 95% CI 1.10–1.16, P < 0.001), discharge to a rehabilitation center (OR = 1.73, 95% CI 1.66–1.80, P < 0.001) or facility other than home (OR = 1.20, 95% CI 1.16–1.23, P < 0.001).
Conclusion and Relevance:
This contemporary matched analysis demonstrates an association with increased morbidity, mortality, and readmissions for Black patients across surgical procedures and specialties.
Keywords: inequality, race, surgery, ethnicity, morbidity, mortality, outcomes
Evidence suggests that Black patients may have poorer health outcomes compared to their White counterparts. This retrospective study compared short-term outcomes between a matched cohort of Black and White patients undergoing surgery across all surgical specialties. We demonstrated increased 30-day morbidity, mortality, readmissions, and reoperations for Black patients.
Supplemental Digital Content is available in the text.
INTRODUCTION
Disparities in health exist for patients across a wide spectrum of factors, including race, ethnicity, and sexual orientation. Justice—along with beneficence, non-maleficence, and autonomy—is considered a foundational ethical pillar in the delivery of healthcare.1 Despite this, numerous studies have demonstrated health disparities analogous to racial injustices highlighted by recent societal events. Racial health disparities are multifactorial and stem from a complex interplay of patient, provider, and system-level factors known to disproportionately affect Black people—the second largest ethnic minority group (after Hispanic and Latino Americans) in the United States and North America.2 The sizeable gap in health outcomes that exists between non-Hispanic White and non-Hispanic Black populations is well characterized in the medical literature.3 These disparities are also visible in the surgical domain where a patient’s race can impact their degree of health care utilization including access to timely surgery, as well as short and long-term surgical outcomes.2,4
Black race has been associated with 20%–50% higher crude mortality rates after surgery when compared to non-Hispanic White race. Black patients also experience higher rates of postoperative morbidity, including disease recurrence.5,6 A majority of these comparisons, however, have utilized institutional and national cohorts, or are limited to specific procedures or surgical domains such as Neurosurgery, Cardiac Surgery, Orthopedic Surgery, and General Surgery.6–8 In this study, we aim to supplement and update our current understanding of racial disparities in surgery by undertaking a broader, more systematic examination of the impact of race on postoperative outcomes.
Our objective was to conduct a contemporary, matched analysis to examine the association of Black race with 30-day clinical outcomes after all surgical procedures across participating National Surgical Quality Program (NSQIP) hospitals. We hypothesized that Black race would be an independent predictor of adverse events after surgery, including 30-day morbidity and mortality, compared to non-Hispanic White race.
METHODS
Population and Data
In this retrospective matched cohort study, we identified all adult patients (18 years or older) of Black/African American or White/Caucasian ethnicity undergoing surgery between 2012 and 2018 utilizing NSQIP Participant Use Data File. As per NSQIP, a White patient was defined as a person having origins in any of the original peoples of Europe, the Middle East, or North Africa. A Black patient was defined as a person having origins in any of the Black racial groups of Africa. These categories were consistent with the US Census Bureau and reported based on the indicated race in the patient’s medical record. For patients with multiple listed races, the one listed first was selected. Patients with the following race labels were excluded from analysis: American Indian or Alaska Native, Native Hawaiian or other Pacific Islander, Asian, Hispanic, or unknown/not reported (n = 1,085,909). Patients undergoing an interventional radiology procedure were excluded (n = 909). Patients with incomplete demographic and perioperative data were excluded to allow for coarsened exact matching (n = 61,511). The final population before coarsened exact matching comprised of 4733,552 patients, including 4153,265 White patients and 5802,087 Black patients.
The NSQIP database contains perioperative measures across surgical specialties from 722 participating hospitals in North America and internationally. Data are obtained by highly trained Surgical Clinical Reviewers on over 150 variables, including demographics, perioperative risk factors, intraoperative variables, and 30-day measures of morbidity and mortality. The validity and accuracy of the measured data points are confirmed by uniformed inter-institutional certification based training and regular audits. Inter-rater reliability audits have estimated that the overall disagreement rate is 2% for the measured variables. The complete dataset comprised of 5881,881 patients. This study was exempt from ethics board review at our institution.
Exposure and Outcomes
The primary exposure was race, dividing the cohort into 2 groups: (1) Black patients and (2) White patients. The primary outcomes included 30-day morbidity and 30-day mortality. Thirty-day morbidity was defined as an aggregate measure comprising of superficial surgical site infection (SSI), deep SSI, organ space SSI, wound dehiscence, pneumonia, intraoperative or postoperative unplanned intubation, pulmonary embolism, greater than 48 hours postoperative ventilator dependence, progressive renal insufficiency, acute renal failure, urinary tract infection, cerebrovascular accident, myocardial infarction, cardiac arrest, bleeding requiring transfusion within 72 hours of operative start, deep vein thrombosis/thrombophlebitis, sepsis, and septic shock. Secondary outcomes included readmission, reoperation, discharge destination, length of stay, and failure to rescue. Failure to rescue was defined as mortality after an index case of morbidity.
Coarsened Exact Matching
To create our matched cohort, we used coarsened exact matching, a novel and intuitive non-parametric method of matching to control for confounding and improve causal inferences. Coarsened exact matching has been shown to potentially outperform other matching methodologies such as propensity score matching, by reducing covariate imbalance and thus improving the negative influence of confounding bias.9,10
We created matched groups one-to-one based on race matching on 28 covariates, including, demographic variables (age, gender, and body mass index), comorbidities, and clinical/operative variables (Table 1). Clinical and operative variables included year of operation, surgical specialty, surgical setting, surgical procedure, American Anesthesia Association class (ASA), functional status, elective designation, and emergency surgery designation. Procedural matching was conducted based on Current Procedural Terminology codes (Supplemental Table 1a, see http://links.lww.com/AOSO/A4). Age and BMI were coarsened into pre-determined clinically relevant bins (Table 1). The final matched cohort included 615,118 patients (307,559 in each cohort).
TABLE 1.
Patient Demographics and Characteristics
| White Patient | Black Patient | |
|---|---|---|
| N | 307,559 | 307,559 |
| Age | ||
| <30 | 32,232 (10.5) | 32,232 (10.5) |
| ≥30 and <40 | 44,956 (14.6) | 44,956 (14.6) |
| ≥40 and <50 | 64,433 (20.9) | 64,433 (20.9) |
| ≥50 and <60 | 69,775 (22.7) | 69,775 (22.7) |
| ≥60 and <70 | 60,011 (19.5) | 60,011 (19.5) |
| ≥70 and <80 | 28,439 (9.2) | 28,439 (9.2) |
| ≥80 and <90 | 7216 (2.3) | 7216 (2.3) |
| ≥90 | 497 (0.2) | 497 (0.2) |
| BMI | ||
| <20 | 5996 (1.9) | 5996 (1.9) |
| ≥20 and <25 | 41,848 (13.6) | 41,848 (13.6) |
| ≥25 and <30 | 86,511 (28.1) | 86,511 (28.1) |
| ≥30 and <35 | 75,673 (24.6) | 75,673 (24.6) |
| ≥35 and <40 | 46,229 (15.0) | 46,229 (15.0) |
| ≥40 and <45 | 25,808 (8.4) | 25,808 (8.4) |
| ≥45 | 25,494 (8.3) | 25,494 (8.3) |
| Sex | ||
| Male | 98,930 (32.3) | 98,930 (32.3) |
| Female | 208,629 (67.8) | 208,629 (67.8) |
| Year of operation | ||
| 2012 | 23,456 (7.6) | 23,456 (7.6) |
| 2013 | 30,430 (9.9) | 30,430 (9.9) |
| 2014 | 38,303 (12.5) | 38,303 (12.5) |
| 2015 | 46,381 (15.1) | 46,381 (15.1) |
| 2016 | 54,644 (17.8) | 54,644 (17.8) |
| 2017 | 58,047 (18.9) | 58,047 (18.9) |
| 2018 | 56,298 (18.3) | 56,298 (18.3) |
| Surgical specialty | ||
| Cardiac surgery | 128 (0.0) | 128 (0.0) |
| General surgery | 156,051 (50.7) | 156,051 (50.7) |
| Gynecology | 41,495 (13.5) | 41,495 (13.5) |
| Neurosurgery | 10,484 (3.4) | 10,484 (3.4) |
| Orthopedics | 61,793 (20.1) | 61,793 (20.1) |
| Otolaryngology | 5996 (1.9) | 5996 (1.9) |
| Plastic surgery | 10,283 (3.3) | 10,283 (3.3) |
| Thoracic surgery | 949 (0.3) | 949 (0.3) |
| Urology | 13,912 (4.5) | 13,912 (4.5) |
| Vascular surgery | 6468 (2.1) | 6468 (2.1) |
| Surgical setting | ||
| Inpatient | 160,499 (52.2) | 160,499 (52.2) |
| Outpatient | 147,060 (47.8) | 147,060 (47.8) |
| Comorbidities | ||
| Smoker | 45,412 (14.8) | 45,412 (14.8) |
| Ventilator | 5 (0.5) | 5 (0.5) |
| COPD | 2192 (0.7) | 2192 (0.7) |
| Hypertension | 146,449 (47.6) | 146,449 (47.6) |
| Disseminated cancer | 1194 (0.4) | 1194 (0.4) |
| Wound infection | 681 (0.2) | 681 (0.2) |
| Steroid use | 2936 (1.0) | 2936 (1.0) |
| Pre-op weight loss | 311 (0.1) | 311 (0.1) |
| Bleeding disorder | 1883 (0.6) | 1883 (0.6) |
| Transfusion | 137 (0.0) | 137 (0.0) |
| Diabetes | 42,533 (13.8) | 42,533 (13.8) |
| Dyspnea | 6250 (2.0) | 6250 (2.0) |
| Ascites | 16 (0.0) | 16 (0.0) |
| Renal failure | 11 (0.0) | 11 (0.0) |
| Dialysis | 631 (0.2) | 631 (0.2) |
| CHF | 54 (0.0) | 54 (0.0) |
| ASA | ||
| 1 | 26,761 (8.7) | 26,761 (8.7) |
| 2 | 160,382 (52.1) | 160,382 (52.1) |
| 3 | 117,854 (38.3) | 117,854 (38.3) |
| 4 | 2560 (0.8) | 2560 (0.8) |
| 5 | 2 (0.0) | 2 (0.0) |
| Functional status | ||
| Independent | 306,833 (99.8) | 306,833 (99.8) |
| Partially dependent | 673 (0.2) | 673 (0.2) |
| Totally dependent | 53 (0.0) | 53 (0.0) |
| Pre-operative sepsis | ||
| No | 302,264 (98.3) | 302,264 (98.3) |
| Yes | 5295 (1.7) | 5295 (1.7) |
| Elective surgery | ||
| No | 31,344 (10.2) | 31,344 (10.2) |
| Yes | 276,215 (89.8) | 276,215 (89.8) |
| Emergency case | ||
| No | 292,778 (95.2) | 292,778 (95.2) |
| Yes | 14,781 (4.8) | 14,781 (4.8) |
ASA, American Anesthesia Association class.
Statistical Analysis
We assessed suitability of coarsened exact matching by pre- and post-match imbalance assessment. We assessed overall imbalance with the L1 statistic as described by Iacus et al.11. A chi-square difference and difference in empirical quantiles of the distribution of the 2 cohorts for each variable were reviewed. There were no missing values with respect to the primary outcome measures. Secondary outcome measures had missing values with respect to 30-day readmission and discharge destination. Values were reported and missing datapoints were censored from univariate analysis. Absolute values and cases per thousand patients were reported for categorical variables. Number needed to harm was calculated for both primary and secondary outcome measures. We compared Black and White groups with univariable binary and multinomial logistic regression and reported odds ratios (ORs) with 95% confidence intervals (CIs). Multinomial regression was utilized for the following outcome measures: discharge destination and total number of episodes of morbidity in 30-days. OR’s and CIls for the remaining outcome measures were compared with binary logistic regression. We did not use multivariable analyses for adjusted analysis given the use of matching for confounding control. We visualized ORs using forest plots for morbidity outcomes. A P-value of <0.05 was considered significant for all comparisons. Coarsened exact matching was conducted using the CEM package (v 1.1.20) of R (v 4.0.02). Subsequent statistical analysis was conducted using STATA v 13.0 and SPSS v 23.0.
RESULTS
Demographics and Clinical Characteristics
We identified a total of 615,118 patients after 1:1 coarsened exact matching, with 307,559 patients in each cohort. Patient and operative clinical characteristics are provided in Table 1. A total of 1099 unique operative procedures were conducted in the matched cohort (Supplemental Table 1, see http://links.lww.com/AOSO/A4). Review of the imbalance measures of the matched cohorts demonstrated a satisfactory match. The 2 cohorts demonstrated an identical match with respect to the identified characteristics. Patient and operative clinical characteristics of the pre-match cohorts, comprising of 2902 unique procedures, are demonstrated in Supplemental Table 2, see http://links.lww.com/AOSO/A5.
Morbidity
Comparison of morbidity outcomes are demonstrated in Table 2 and Figure 1. Black race compared to White race was associated with increased odds of all-cause 30-day morbidity (623 vs 684 episodes per 10,000 patients; OR = 1.10, 95% CI 1.08–1.13). This corresponded to a number needed to harm of 164 patients. Black race was also associated with increased odds of re-intubation (28 vs 21 episodes per 10,000 patients; OR = 1.33, 95% CI 1.21–1.48), pulmonary embolism (32 vs 21 episodes per 10,000 patients, OR = 1.55, 95% CI 1.40–1.71), failure to wean from ventilator for >48 hours (19 vs 17 episodes per 10,000 patients; OR = 1.14, 95% CI 1.02–1.29), progressive renal insufficiency (19 vs 11 episodes per 10,000 patients; OR = 1.63, 95% CI 1.43–1.86), acute renal failure (9 vs 7 episodes per 10,000 patients; OR = 1.39, 95% CI 1.16–1.66), cardiac arrest (10 vs 7 episodes per 10,000 patients; OR = 1.47, 95% CI 1.24–1.76), bleeding requiring transfusion (297 vs 216 episodes per 10,000 patients; OR = 1.39, 95% CI 1.34–1.43), DVT/thrombophlebitis (39 vs 31 episodes per 10,000 patients; OR = 1.24, 95% CI 1.14–1.35), and sepsis/septic shock (93 vs 86 episodes per 10,000 patients; OR = 1.09, 95% CI 1.03–1.15).
TABLE 2.
30-Day Morbidity
| White, N (Per 10,000 Patients) | Black, N (per 10,000 Patients) | P | |
|---|---|---|---|
| All morbidity | 19,168 (623) | 21,040 (684) | <0.001 |
| Superficial SSI | 3494 (113) | 2765 (90) | <0.001 |
| Deep SSI | 871 (28) | 829 (27) | 0.308 |
| Organ space SSI | 2508 (81) | 2577 (84) | 0.331 |
| Wound dehiscence | 713 (23) | 768 (25) | 0.152 |
| Pneumonia | 1027 (33) | 1097 (36) | 0.128 |
| Re-intubation | 651 (21) | 868 (28) | <0.001 |
| Pulmonary embolism | 646 (21) | 999 (32) | <0.001 |
| Failure to wean ventilator (>48 h) | 514 (17) | 588 (19) | 0.026 |
| Progressive renal insufficiency | 353 (11) | 575 (19) | <0.001 |
| Acute renal failure | 210 (7) | 292 (9) | <0.001 |
| Urinary tract infection | 2702 (88) | 2351 (76) | <0.001 |
| Cerebrovascular accident | 223 (7) | 270 (9) | 0.034 |
| Cardiac arrest | 211 (7) | 311 (10) | <0.001 |
| Myocardial infarction | 401 (13) | 302 (10) | <0.001 |
| Bleeding requiring transfusion | 6644 (216) | 9138 (297) | <0.001 |
| DVT/thrombophlebitis | 965 (31) | 1193 (39) | <0.001 |
| Sepsis/septic shock | 2644 (86) | 2874 (93) | 0.002 |
| Morbidity events | <0.001 | ||
| 0 (ref) | 288,391 (9377) | 286,519 (9315) | |
| 1 | 15,413 (501) | 16,657 (541) | <0.001 |
| 2 | 2551 (83) | 2902 (94) | <0.001 |
| 3 | 701 (23) | 855 (28) | <0.001 |
| ≥4 | 293 (9) | 322 (10) | <0.001 |
FIGURE 1.
Forrest plot of morbidity.
Black race was associated with progressively increasing odds of having multiple complications in a 30-day period compared to White race (Table 2). Compared to having no complications, Black patients versus White patients had increased odds of having 1 complication (OR = 1.09, 95% CI 1.06–1.13, P < 0.001), 2 complications (OR = 1.15, 95% CI 1.09–1.21, P < 0.001), 3 complications (OR = 1.23, 95% CI 1.11–1.36, P < 0.001), or 4 or more complications (OR = 1.25, 95% CI 1.11–1.41, P < 0.001).
Mortality and Failure to Rescue
Black race was associated with increased odds of mortality compared to White race (15 vs 13 episodes per 10,000 patients, OR = 1.15, 95% CI 1.01–1.31, P = 0.039). This corresponded to a number needed to harm of 5042 patients. There was no difference in odds of failure to rescue between races (173 vs 158 episodes per 10,000 patients; OR = 1.09, 95% CI 0.94–1.27, P = 0.259). Assessment of failure to rescue after index complications is highlighted in Table 3. The only complication associated with a higher odds of mortality in the Black race was cardiac arrest (65.0% vs 52.1%; OR = 1.70, 95% CI 1.19–2.43, P = 0.003).
TABLE 3.
Mortality, Failure to Rescue, Readmission, Reoperation, and Disposition
| White, N (Per 10,000 Patients) | Black, N (Per 10,000 Patients) | OR (95% CI) (Ref = White) | P | |
|---|---|---|---|---|
| Mortality | 409 (13) | 470 (15) | 1.15 (1.01–1.31) | 0.039 |
| Failure to rescue (mortality after complication) | 304 (158) | 364 (173) | 1.09 (0.94–1.27) | 0.259 |
| Wound complication* | 56 (74) | 45 (65) | 0.88 (0.60–1.31) | 0.531 |
| Pneumonia | 70 (681) | 73 (665) | 0.98 (0.69–1.37) | 0.882 |
| Re-intubation | 136 (2089) | 201 (2315) | 1.14 (0.89–1.46) | 0.293 |
| Pulmonary embolism | 17 (263) | 24 (240) | 0.91 (0.49–1.71) | 0.771 |
| Failure to wean ventilator (>48h) | 87 (1692) | 94 (1598) | 0.93 (0.68–1.29) | 0.674 |
| Progressive renal insufficiency | 16 (453) | 25 (435) | 0.96 (0.50–1.82) | 0.894 |
| Acute renal failure | 53 (2524) | 58 (1986) | 0.73 (0.48–1.12) | 0.152 |
| Urinary tract infection | 9 (33) | 10 (42) | 1.28 (0.52–3.15) | 0.593 |
| Cerebrovascular accident | 36 (1614) | 28 (1037) | 0.60 (0.35–1.02) | 0.058 |
| Cardiac arrest | 110 (5213) | 202 (6495) | 1.70 (1.19–2.43) | 0.003 |
| Myocardial infarction | 38 (948) | 43 (1424) | 1.59 (0.99–2.52) | 0.050 |
| Bleeding requiring transfusion | 104 (156) | 122 (133) | 0.85 (0.65–1.11) | 0.229 |
| DVT/thrombophlebitis | 17 (176) | 24 (201) | 1.15 (0.61–2.14) | 0.672 |
| Sepsis/septic shock | 110 (416) | 119 (414) | 0.99 (0.76–1.30) | 0.971 |
| Any readmission† | 10,789 (428) | 12,031 (479) | 1.12 (1.10–1.16) | <0.001 |
| Unplanned readmission | 10,289 (408) | 11,586 (461) | 1.26 (1.11–1.44) | <0.001 |
| Unplanned reoperation | 5324 (173) | 5549 (180) | 1.04 (1.00–1.08) | 0.029 |
| Still in hospital at 30 days | 260 (8) | 366 (12) | 1.41 (1.20–1.65) | <0.001 |
| Discharge destination‡ | <0.001 | |||
| Home (ref) | 294761 (9584) | 290303 (9446) | 1 | |
| Rehab | 3982 (129) | 6779 (220) | 1.73 (1.66–1.80) | <0.001 |
| Other facility§ | 8454 (275) | 9974 (325) | 1.20 (1.16–1.23) | <0.001 |
| Expired | 210 (7) | 273 (9) | 1.32 (1.10–1.58) | <0.001 |
*Wound complication = superficial SSI, deep SSI, organ space SSI, wound dehiscence.
†Missing = 111,861.
‡Missing = 382.
§Other facility = multi-level senior community, separate acute care, skilled care- not home, unskilled facility—not home.
Readmission, Reoperation, and Disposition
Black patients had a higher likelihood of having readmission of any cause (4.8% vs 4.3%; OR = 1.12, 95% CI 1.10–1.16, P < 0.001). There was an association with increased odds of unplanned reoperations for Black patients (180 vs 173 reoperations per 10,000 patients, OR = 1.04, 95% CI 1.00–1.08, P = 0.029); however, the clinical relevance was likely minimal. Black patients had a higher odds of being discharged to a rehabilitation facility (OR = 1.73, 95% CI 1.66–1.80, P < 0.001), a facility other than home (OR = 1.20, 95% CI = 1.16–1.23, P < 0.001), or in-hospital mortality (OR = 1.32, 95% CI = 1.10–1.58, P < 0.001). The median length of stay for both Black and White patients was 1 day (interquartile range: 2, P = <0.001).
DISCUSSION
This report quantifies the magnitude of short-term adverse events following surgery in Black patients compared with a cohort of White patients matched across surgical procedure and baseline demographic and clinical characteristics. We demonstrate that Black race, after matching, is associated with an increased odds of adverse events; including morbidity, mortality, readmissions, and reoperations. This report adds to our evolving understanding of the impact of racial disparities in surgery and is significant for 3 important reasons. First, the use of the novel coarsened exact matching allowed us to report the magnitude of these outcomes and associations while minimizing the influence of critical confounders and heterogeneity. Second, we employed a well-validated large contemporary dataset that spanned across a range of surgical specialties and procedures, which broadens the applicability of the measured outcomes. Finally, this study provides specific morbidity outcome data that can be utilized at an institutional level to inform policy and strategies that will bridge the disparities in care between Black and White patients undergoing surgery.
Multiple prior studies have demonstrated a potential association with increased morbidity following surgery in Black patients.6,12 Our study builds upon this literature and is strengthened by utilizing a broad cross-specialty assessment using a matched cohort. In this analysis, we confirmed that all-cause morbidity is significantly higher in Black patients. Furthermore, we demonstrate that Black patients are particularly vulnerable to an increased risk of serious complications such as re-intubation, difficulty weaning from ventilator, venous thromboembolism, pulmonary embolism, renal failure, postoperative bleeding requiring transfusion, and cardiac arrest. Importantly, these associations held true despite matching for important patient comorbidity confounders relevant to the aforementioned outcome measures. Although the absolute difference across outcome measures was small, we demonstrated a persistent difference across the majority of outcome measures. Furthermore, a large proportion of cases were in low acuity outpatient settings, where serious complications are exceedingly rare. In total, these small differences are not only statistically significant but also cumulatively clinically relevant. In 2014, it was estimated that over 17 million hospital visits included invasive/therapeutic surgeries in the United States.13 With a population comprising of approximately 13% Black people, the number needed to harm of 163 in our analysis, could potentially account for over 13,500 extra complications annually due to race alone.14
Multiple hypotheses have been explored in the literature to explain the differential morbidity outcome between Black and White patients. The interplay of patient, healthcare provider, and health system-level factors all likely influence these outcome disparities. Studies have suggested that Black patients, for certain ailments, receive operative intervention later in the disease course than White patients.15,16 This leads to more urgent/emergent operations and may contribute to increased morbidity and mortality. The influence of such a factor was controlled to a degree in our study by matching across surgical acuity, including urgent/emergent surgical indication. However, matching did reduce the acuity of the matched Black cohort compared to the unmatched sample, indicating that a potentially healthier cohort was selected for comparison. This may have potentially underestimated the true disparity between groups. Such differences in the presentation of disease acuity speak to the potential role of structural racism in fostering an environment that results in such inequities. More advanced presentations of disease have also been linked to other major determinants of health including income and educational level. Studies have shown that controlling for these and other health determinants diminish the influence of race on advanced disease presentations.17 Comorbidity burden is a key predictor of postoperative morbidity. Multiple studies have demonstrated that Black patients have a higher comorbidity burden, including higher rates of cardiovascular disease, renal disease, and hypertension.18–20 Matching for these major comorbidities allowed us to reduce the influence this may have played on our measured outcomes. Furthermore, other consequential patient-level factors such as socioeconomic status and nutrition, which have been shown to influence operative outcomes, were not measurable in this study.21–23 These important patient-level factors have been shown to influence access to surgery, a key healthcare system-level factor.
Other important healthcare system factors such as hospital quality and hospital volume also likely influence surgical morbidity and outcomes.12,24 It has been reported that Black patients tend to more often undergo operations in low-volume hospitals, and hospitals of lower quality (determined by risk-adjusted hospital mortality rates), potentially increasing risk of complications.6,12 Due to limitations in our dataset, we were not able to directly study the influence of specific hospital quality indicators; however, we hypothesize that at a baseline level, hospitals that participate in the NSQIP program are invested in surgical quality, which may have mitigated the influence of hospital-level factors on our findings.25 It has been shown that participation in NSQIP does reduce adverse events in surgery and improves quality.26 Finally, it is important to note that provider-level factors such as surgeon volume, experience, and quality are potentially associated with outcome disparities between races, as studies have demonstrated differential access between races in this setting.5,27 Moreover, the influence of provider based conscious and unconscious bias, including racial bias, is impossible to measure with such a study design but may play a potential role.28 In our study design one must question why a disparity in outcomes still exists after careful matching on key variables, and surgeons and healthcare workers must take into account potential biases that may influence the care of Black patients.
Black patients were also observed to experience significantly higher crude 30-day and inpatient mortality in relation to their White counterparts. This is a known phenomenon, which has been observed in several prior studies of race and outcomes in surgery. National Medicare data suggests that Black patients have higher operative mortality across a range of surgical procedures including coronary artery bypass, carotid endarterectomy and pancreatic resections.29 According to data from the National Inpatient Sample, Black patients are also significantly more likely to die in-hospital than White patients after undergoing appendectomy, gastric fundoplication, and gastric bypass surgeries.7 These findings may be explained by a variety of factors including differences in baseline risk and comorbidities between Black and White patients. We performed a coarsened exact matched analysis to mitigate this issue of confounding, a method that has been previously used to assess racial disparities in the trauma population, demonstrating poorer post-discharge healthcare utilization and differential mortality between Black and White patients.30–32 Ours is one of the first and largest studies to examine the association of race and failure to rescue after surgery. While we did not observe a difference in the overall odds of failure to rescue between Black and White patients, the former group was 1.7× more likely to die after a cardiac arrest. While some have argued that this is likely attributable to the poorer health of Black patients compared to their White counterparts,33 our matched analysis would suggest otherwise. Failure to rescue—or death after experiencing a complication—is an important indicator of surgical quality.34 These findings corroborate the notion that racial disparities in surgery may in part be due to the suboptimal systems in which Black patients receive care. The disproportionate comorbidity burden, higher disease acuity, decreased equitable care, and worse surgical outcomes seen in Black patients highlights the racial inequities embedded within our society, communities, and health systems. A growing body of literature has demonstrated that structural racism, known as the ingrained influence of racism across major societal systems—including education, housing, employment, healthcare—is a key health determinant.35 Efforts to abolish policies that foster structural racism, while also strengthening programs that empower such communities have the greatest likelihood of mitigating its detrimental impact on health. Structural racism must also be addressed at the health system level by fostering trust with marginalized communities through engagement, improved access, and targeted educational outreach.36
We also confirm that Black patients are at higher risk of re-admission and re-operation after surgery compared to their White counterparts.37 It is interesting that despite being identically matched on all available clinical covariates, Black patients were more likely to be discharged to rehabilitation or a facility other than home compared to White patients after surgery. This is an important patient-centered outcome, which has demonstrated greater sensitivity to patient- and surgery-level characteristics than differences in hospital characteristics.38 Our findings add meaningful substance to the current body of literature, which has yet to examine the importance of a patient’s race on discharge disposition. In addition to systemic inequalities (ie, adequate housing, access to low mortality/high-quality hospitals, socioeconomic status), Black patients may also lack meaningful access to community supports that may facilitate them remaining at home after discharge. Therefore, they may benefit from more longitudinal and multi-disciplinary disposition planning using a team of professionals including social workers, nurse navigators, and discharge coordinators.39 Furthermore, there is potential for inherent healthcare provider and health system racial biases to influence a patient’s discharge disposition. Such differences may be limited by utilization of quality initiatives that employ objective measures to ascertain a patient’s appropriate discharge disposition.
This study must be interpreted in the context of several limitations, many of which are inherent to the use of large datasets such as NSQIP. First, this study is at risk of influence from unmeasured confounders such as surgeon experience, surgeon race, hospital quality, hospital location, socioeconomic status, insurance status, genetics, and nutrition. We attempted to limit the impact of confounding by utilizing CEM across all clinically relevant measured variables. CEM is an intuitive matching methodology that has been shown to potentially outperform other methods of confounding control such as propensity score matching and multivariable regression analysis.9,10 Furthermore, matching across a large set of covariates allowed us to reduce heterogeneity and likely improved balance over unmeasured confounders, as such reducing the influence of biased effect estimates. Second, despite matching improving confounding bias, it reduces heterogeneity and as such the applicability and validity of our outcomes are limited to the patient and clinical characteristics of the matched cohort. Our pre-matched population comprised of 516,070 Black patients, after matching approximately 60% of all Black patients matched and were included in the final analysis. This significant proportional match demonstrates the persistent broad applicability of our results to the general NSQIP population. However, it is important to note that approximately 40% of Black patients were excluded from the analysis, and assessment of the pre-match baseline characteristics demonstrates that they tended to have greater comorbidities, higher ASA scores, and were more likely to have undergone emergency surgery. Thus, the matched cohort of Black patients was a healthier subset of the overall NSQIP Black patient population, and as such our findings may underestimate the true difference in adverse event rates between cohorts. Ideally, a separate multivariate analysis would be done on the complete cohort; however, we felt this was not feasible given that such a model would need to account for over 2900 unique operative procedures. As such, we felt that coarsened exact matching would provide the most reliable comparison while accounting for a broad range of operations. Furthermore, although we did observe a difference in morbidity across race, it is important to note that NSQIP does not report on severity of outcomes and as such this was not measured. We did attempt to quantify this by measuring failure to rescue, or mortality after complication, which is reviewed above. Similarly, NSQIP does not report on disease severity/acuity, a factor which may bias our findings. Matching was performed across comorbidities and emergency/elective surgical designation to reduce the influence of disease severity on outcome measures. Nonetheless, while acuity of a case type was matched for, it cannot fully approximate where a patient is in their disease course. Finally, this study due to its design is at risk of selection bias. Participation in the NSQIP program is completely voluntary at the hospital level, a factor that may affect patient selection and subsequent generalizability of outcome measures. NSQIP hospitals account for approximately 12% of U.S. Hospitals, and have been shown to be larger, more well-resourced, and academic-affiliated compared to nonparticipating hospitals.40 Given that Black patients are known to have reduced access to high-quality care, our results likely underestimate the true difference in measures across the general population.
CONCLUSIONS
This is among the largest contemporary surgical studies, and first to use the novel coarsened exact matching methodology, to directly compare short-term complications and mortality between Black and White patients. The findings of this study demonstrate an association with higher morbidity and mortality for Black patients undergoing surgery than a comparable cohort of White patients undergoing the same procedures. Further research is needed to identify etiologic factors that result in such disparities, including the potential influence of racial bias, with an aim to abolish this gap and optimize outcomes for Black patients.
Supplementary Material
Footnotes
Disclosure: The authors declare that they have nothing to disclose.
Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s Web site (www.annalsofsurgery.com).
REFERENCES
- 1.Page K. The four principles: can they be measured and do they predict ethical decision making? BMC Med Ethics. 2012; 13:10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Haider AH, Scott VK, Rehman KA, et al. Racial disparities in surgical care and outcomes in the United States: a comprehensive review of patient, provider, and systemic factors. J Am Coll Surg. 2013; 216:482–92.e12 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Smedley BD, Stith AY, Nelson AR, eds. Institute of Medicine (US) Committee on Understanding and Eliminating Racial and Ethnic Disparities in Health Care. Unequal Treatment: Confronting Racial and Ethnic Disparities in Health Care. 2003, Washington (DC): National Academies Press (US) [PubMed] [Google Scholar]
- 4.Shavers VL, Fagan P, Jones D, et al. The state of research on racial/ethnic discrimination in the receipt of health care. Am J Public Health. 2012; 102:953–966 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Morris AM, Wei Y, Birkmeyer NJ, et al. Racial disparities in late survival after rectal cancer surgery. J Am Coll Surg. 2006; 203:787–794 [DOI] [PubMed] [Google Scholar]
- 6.Rangrass G, Ghaferi AA, Dimick JB. Explaining racial disparities in outcomes after cardiac surgery: the role of hospital quality. JAMA Surg. 2014; 149:223–227 [DOI] [PubMed] [Google Scholar]
- 7.Ricciardi R, Selker HP, Baxter NN, et al. Disparate use of minimally invasive surgery in benign surgical conditions. Surg Endosc. 2008; 22:1977–1986 [DOI] [PubMed] [Google Scholar]
- 8.Trivedi AN, Sequist TD, Ayanian JZ. Impact of hospital volume on racial disparities in cardiovascular procedure mortality. J Am Coll Cardiol. 2006; 47:417–424 [DOI] [PubMed] [Google Scholar]
- 9.King G, Nielson R. Why propensity score should not be used for matching. Political Analysis. 2019; 27:435–454 [Google Scholar]
- 10.King G, Nielson R, Coberley C, et al. Comparative effectiveness of matching methods for causal inference. http://gking.harvard.edu/publications/comparative-effectiveness-matching-methods-causal-inference. Accessed July 3, 2020
- 11.Iacus SM, King G, Porro G. Matching for causal inference without balance checking. 2008http://gking.harvard.edu/files/abs/cem-abs.shtml. Accessed July 4, 2020
- 12.Birkmeyer JD, Siewers AE, Finlayson EV, et al. Hospital volume and surgical mortality in the United States. N Engl J Med. 2002; 346:1128–1137 [DOI] [PubMed] [Google Scholar]
- 13.Steiner CA, Karaca Z, Moore BJ, Imshaug MC, Pickens G. Surgeries in Hospital-Based Ambulatory Surgery and Hospital Inpatient Settings, 2014: Statistical Brief #223. In: Healthcare Cost and Utilization Project (HCUP) Statistical Briefs. 2017, Rockville (MD): Agency for Healthcare Research and Quality (US) [PubMed] [Google Scholar]
- 14.“US Census Bureau July 1 2019 Estimates” (web). 2019, United States Census Bureau; Retrieved August 2, 2020 [Google Scholar]
- 15.Ball JK, Elixhauser A. Treatment differences between blacks and whites with colorectal cancer. Med Care. 1996; 34:970–984 [DOI] [PubMed] [Google Scholar]
- 16.Chew DK, Nguyen LL, Owens CD, et al. Comparative analysis of autogenous infrainguinal bypass grafts in African Americans and Caucasians: the association of race with graft function and limb salvage. J Vasc Surg. 2005; 42:695–701 [DOI] [PubMed] [Google Scholar]
- 17.Lee SL, Shekherdimian S, Chiu VY, Sydorak RM. Perforated appendicitis in children: equal access to care eliminates racial and socioeconomic disparities. J Pediatr Surg. 2010; 45:1203–1207 [DOI] [PubMed] [Google Scholar]
- 18.Jolly S, Vittinghoff E, Chattopadhyay A, et al. Higher cardiovascular disease prevalence and mortality among younger blacks compared to whites. Am J Med. 2010; 123:811–818 [DOI] [PubMed] [Google Scholar]
- 19.Kershaw KN, Diez Roux AV, Burgard SA, et al. Metropolitan-level racial residential segregation and black-white disparities in hypertension. Am J Epidemiol. 2011; 174:537–545 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Norris KC, Agodoa LY. Unraveling the racial disparities associated with kidney disease. Kidney Int. 2005; 68:914–924 [DOI] [PubMed] [Google Scholar]
- 21.Williams DR, Priest N, Anderson NB. Understanding associations among race, socioeconomic status, and health: Patterns and prospects. Health Psychol. 2016; 35:407–411 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Downing NS, Wang C, Gupta A, et al. Association of racial and socioeconomic disparities with outcomes among patients hospitalized with acute myocardial infarction, heart failure, and pneumonia: an analysis of within- and between-hospital variation. JAMA Netw Open. 2018; 1:e182044. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Satia JA. Diet-related disparities: understanding the problem and accelerating solutions. J Am Diet Assoc. 2009; 109:610–615 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Nathan H, Frederick W, Choti MA, Schulick RD, Pawlik TM. Racial disparity in surgical mortality after major hepatectomy [published correction appears in J Am Coll Surg. 2010;211:301]. J Am Coll Surg. 2008; 207:312–319 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Parsons HM, Habermann EB, Stain SC, et al. What happens to racial and ethnic minorities after cancer surgery at American College of Surgeons National Surgical Quality Improvement Program hospitals? J Am Coll Surg. 2012; 214:539–47discussion 547 [DOI] [PubMed] [Google Scholar]
- 26.Cohen ME, Liu Y, Ko CY, et al. Improved surgical outcomes for ACS NSQIP Hospitals over time: evaluation of hospital cohorts with up to 8 years of participation. Ann Surg. 2016; 263:267–273 [DOI] [PubMed] [Google Scholar]
- 27.Aranda MA, McGory M, Sekeris E, Maggard M, Ko C, Zingmond DS. Do racial/ethnic disparities exist in the utilization of high-volume surgeons for women with ovarian cancer? Gynecol Oncol. 2008; 111:166–172 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Santry HP, Wren SM. The role of unconscious bias in surgical safety and outcomes. Surg Clin North Am. 2012; 92:137–151 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Lucas FL, Stukel TA, Morris AM, et al. Race and surgical mortality in the United States. Ann Surg. 2006; 243:281–286 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Sharma R, Johnson A, Li J, DeBoard Z, Zikakis I, Grotts J, Kaminski S. Racial disparities and the acute management of severe blunt traumatic brain injury. Trauma Surg Acute Care Open. 2019; 4:e000358. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Hicks CW, Hashmi ZG, Velopulos C, et al. Association between race and age in survival after trauma. JAMA Surg. 2014; 149:642–647 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Chun Fat S, Herrera-Escobar JP, Seshadri AJ, et al. Racial disparities in post-discharge healthcare utilization after trauma. Am J Surg. 2019; 218:842–846 [DOI] [PubMed] [Google Scholar]
- 33.Silber JH, Rosenbaum PR, Kelz RR, et al. Examining causes of racial disparities in general surgical mortality: hospital quality versus patient risk. Med Care. 2015; 53:619–629 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Portuondo JI, Shah SR, Singh H, Massarweh NN. Failure to rescue as a surgical quality indicator: current concepts and future directions for improving surgical outcomes. Anesthesiology. 2019; 131:426–437 [DOI] [PubMed] [Google Scholar]
- 35.Bailey ZD, Krieger N, Agénor M, et al. Structural racism and health inequities in the USA: evidence and interventions. Lancet. 2017; 389:1453–1463 [DOI] [PubMed] [Google Scholar]
- 36.Egede LE, Walker RJ. Structural racism, social risk factors, and COVID-19—a dangerous convergence for Black Americans. N Engl J Med. 2020; 383:e77. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Havens JM, Olufajo OA, Cooper ZR, Haider AH, Shah AA, Salim A. Defining rates and risk factors for readmissions following emergency general surgery [published correction appears in JAMA Surg. 2017 Jul 1;152(7):708]. JAMA Surg. 2016; 151:330–336 [DOI] [PubMed] [Google Scholar]
- 38.Jerath A, Austin PC, Wijeysundera DN. Days alive and out of hospital: validation of a patient-centered outcome for perioperative medicine. Anesthesiology. 2019; 131:84–93 [DOI] [PubMed] [Google Scholar]
- 39.Ko NY, Snyder FR, Raich PC, et al. Racial and ethnic differences in patient navigation: results from the Patient Navigation Research Program. Cancer. 2016; 122:2715–2722 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Sheils CR, Dahlke AR, Kreutzer L, et al. Evaluation of hospitals participating in the American College of Surgeons National Surgical Quality Improvement Program. Surgery. 2016; 160:1182–1188 [DOI] [PubMed] [Google Scholar]