Although racial/ethnic disparities in health care and outcomes have been extensively documented in prior research, several studies have identified lower mortality rates in non-Hispanic Black than non-Hispanic White patients following hospitalization for common medical conditions such as heart failure (HF) or pneumonia.1–4 Similar comparisons of mortality for Hispanics are lacking.5,6

The explanation for increased survival for non-Hispanic Black compared with non-Hispanic White patients following hospitalization is unclear. Prior studies have used claims-based measures to adjust for differences in clinical risk. However, a previous study found that algorithms that predict risk using claims-based measures assigned non-Hispanic Black and non-Hispanic White patients to the same risk level even though non-Hispanic Black patients had worse health.7 As an example, prior receipt of a coronary artery bypass graft is a covariate in risk-adjustment models for mortality in HF. However, underuse of this procedure among non-Hispanic Black patients would lead to racial differences in measured clinical risk when these differences instead reflect unequal access to care.8,9 It is therefore possible that including objective clinical measures in risk-adjustment models may attenuate the observed survival advantage for non-Hispanic Black Americans after hospitalization.10

The Veterans Health Administration (VA) is an ideal setting to examine racial/ethnic differences in hospital mortality for several reasons. First, the VA represents a safety net health care system that serves larger proportions of racial/ethnic minorities and individuals with lower socioeconomic status than the general US population.5,11,12 Second, the VA ensures access to health care for all enrollees and reduces financial and other barriers to timely health care, with prior studies identifying minimal racial differences in ambulatory health care utilization among Veterans.6,13 Multiple prior studies have identified a survival advantage for hospitalized non-Hispanic Black Veterans, though the most recent of these studies used 2006–2011 data.11,13–15 Little is known about VA hospital mortality rates among Hispanic Veterans because prior studies often excluded Hispanic patients because of small sample sizes.1,3,16

Our aims were to compare mortality for non-Hispanic White, non-Hispanic Black, and Hispanic Veterans hospitalized for HF and pneumonia and determine whether observed mortality differences were based on risk adjustment for claims-based comorbid conditions and/or clinical variables. To achieve these aims, we examined data between January 2009 and September 2015, a time period during which the VA launched national efforts to measure and improve hospital mortality.17–19 We additionally examined whether assessments of racial/ethnic mortality differences vary depending on whether claims-based comorbid conditions or clinical variables are used for risk adjustment.7

METHODS Data Sources and Study Population

We used the VA Informatics and Computing Infrastructure (VINCI) data warehouse to extract VA hospital admissions, vital signs, laboratory values, vital status, outpatient claims, claims files, and fee-basis files for health services that occurred in non-VA settings.20 The Providence VA Medical Center’s Institutional Review Board reviewed and approved the study.

The study cohort included admission to a VA Medical Center (VAMC) between January 1, 2009, and September 30, 2015, with a principal diagnosis of HF or pneumonia, as well as admissions with a principal diagnosis of sepsis or respiratory failure (RF) and a secondary diagnosis of pneumonia. We focused on these medical conditions because they are among the most common conditions for hospitalization and are publicly reported by the VA and Medicare. This time frame was selected to avoid the impact of the transition to International Classification of Diseases, 10th Revision (ICD-10) codes that took place in October 2015. We included patients with a principal diagnosis of sepsis or RF in our pneumonia cohort because prior work identified large increases in the number of patients with a primary diagnosis of sepsis/RF and a secondary diagnosis of pneumonia (Table, Supplemental Digital Content 1, https://links.lww.com/MLR/C334).20,21 We excluded Veterans with <12 months of continuous VA enrollment in the year before admission, who left against medical advice, with hospice utilization in the year before admission, and who were discharged alive within 24 hours of their admission, consistent with Centers for Medicare and Medicaid Services (CMS)/VA measure specifications. Veterans with missing race/ethnicity or from a racial/ethnic group other than non-Hispanic White (hereafter, “White”), non-Hispanic Black (hereafter, “Black”), or Hispanic were excluded [3404 admissions (2.3%) for HF; 3543 (2.7%) for pneumonia]. The final study cohort included 143,520 admissions for HF and 127,782 admissions for pneumonia across 132 VAMC (Fig., Supplemental Digital Content 2, https://links.lww.com/MLR/C334).

Measures Primary Study Outcome and Independent Variable

Our study outcome was all-cause mortality in the 30 days following admission, determined using the VA vital status file and consistent with VA’s methodology for calculating hospital mortality.22Veterans were linked to the VAMC they were originally admitted to, regardless of postadmission transfers to other VA and non-VA medical facilities. The primary independent variable was race/ethnicity, defined as White, Black, or Hispanic.

Baseline Sociodemographic, Claims-based, and Clinical Variables

In addition to the site of hospitalization and calendar quarter based on the month and year of admission, we abstracted patient sociodemographic, claims-based, and clinical variables for risk adjustment.

The sociodemographic variables were patient age, sex, Area Deprivation Index (ADI), homelessness, admission to the hospital from a nursing home, VA enrollment priority, and number of nursing home days in the year before admission. The ADI is a composite socioeconomic disadvantage measure based on US Census data aggregated to the census block or census tract that summarizes 17 variables pertaining to poverty, education, housing, and employment.23–25 The ADI is constructed with a national mean of 100 and a standard deviation of 10. We defined homelessness by the presence of an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code of V60.0 or V.60.1 diagnostic code.26Veterans are assigned 1 of 8 enrollment priority groups by the VA based on factors like the presence and extent of a service-related disability, income, receipt of VA pension benefits. The variable for nursing home days before admission included the following categorizations: no evidence of nursing home use, short-term nursing home residence (<90 d), and long-term nursing home residence (>90 d).

We derived claims-based variables using VA administrative data from the index admission, in addition to outpatient and inpatient encounter data for the 12 months preceding admission. The individual comorbid conditions (22 for HF, 30 for pneumonia) are listed in Tables 1 and 2. We did not include comorbid conditions from Medicare claims to maintain consistency with VA’s current risk-adjustment methodology.27

TABLE 1 - Comparison of Baseline Characteristics for Veterans Hospitalized With Heart Failure Stratified by Race/Ethnicity Groups n (%) Characteristics White (N=101,226) Black (N=38,044) Hispanic (N=4250) P WA P WH Sociodemographic variables  Age (mean±SD) (y) 73.2±10.7 67.3±11.9 73.1±11.8 * 0.59  Male 99,641 (98.4) 37,299 (98.0) 4218 (99.2) * *  Area Deprivation Index (mean±SD)† 104.2±13.9 108.7±12.5 106.0±14.2 * *  Homeless 10,625 (10.5) 6826 (17.9) 445 (10.5) * 0.98  Discharged from nursing home 2033 (2.0) 516 (1.4) 64 (1.5) * *  Enrollment priority   Group 1 25,495 (25.2) 8899 (23.4) 1078 (25.4) * *   Group 2 5305 (5.2) 2311 (6.1) 179 (4.2)   Group 3 9234 (9.1) 3373 (8.9) 336 (7.9)   Group 4 5766 (5.7) 2135 (5.6) 405 (9.5)   Group 5 49,515 (48.9) 19,918 (52.4) 2136 (50.3)   Group 6 646 (0.6) 243 (0.6) 24 (0.6)   Group 7+8 5265 (5.2) 1165 (3.1) 92 (2.2)  Nursing home stay   None 88,285 (87.2) 34,549 (90.8) 3896 (91.7) * *   <90 d 11,415 (11.3) 3028 (8.0) 293 (6.9)   >90 d 1526 (1.5) 467 (1.2) 61 (1.4) Claims-based variables  Percutaneous transluminal coronary angioplasty 12,101 (12.0) 2940 (7.7) 352 (8.3) * *  Coronary artery bypass graft 25,496 (25.2) 3976 (10.5) 767 (18.0) * *  Congestive heart failure 80,515 (79.5) 31,701 (83.3) 3045 (71.6) * *  Acute myocardial infarction 8448 (8.3) 2559 (6.7) 442 (10.4) * *  Ischemic heart disease 12,726 (12.6) 4348 (11.4) 613 (14.4) * *  Chronic atherosclerosis 71,155 (70.3) 21,630 (56.9) 2560 (60.2) * *  Cardiorespiratory failure and shock 20,003 (19.8) 6265 (16.5) 568 (13.4) * *  Valvular/rheumatic heart disease 24,734 (24.4) 7177 (18.9) 832 (19.6) * *  Hypertension 89,815 (88.7) 34,746 (91.3) 3720 (87.5) * *  Stroke 9862 (9.7) 4715 (12.4) 474 (11.2) * *  Renal failure 51,156 (50.5) 22,348 (58.7) 2040 (48.0) * *  Chronic obstructive pulmonary disease 48,119 (47.5) 14,735 (38.7) 1129 (26.6) * *  Pneumonia 23,970 (23.7) 7868 (20.7) 849 (20.0) * *  Diabetes mellitus with complications 62,231 (61.5) 22,362 (58.8) 2815 (66.2) * *  Dementia/senility 11,460 (11.3) 3825 (10.1) 609 (14.3) * *  Functional disability 8577 (8.5) 3668 (9.6) 409 (9.6) * *  Peripheral vascular disease 32,558 (32.2) 9759 (25.7) 1211 (28.5) * *  Cancer 4048 (4.0) 1228 (3.2) 129 (3.0) * *  Trauma 31,475 (31.1) 8965 (23.6) 1175 (27.6) * *  Major psychiatric disease 14,402 (14.2) 5959 (15.7) 621 (14.6) * 0.50  Chronic liver disease 5142 (5.1) 4124 (10.8) 324 (7.6) * *  Surgery 4008 (4.0) 2227 (5.9) 192 (4.5) * 0.07 No. claims-based comorbidities (mean±SD) 6.4±2.9 6.0±2.8 5.9±3.0 * * Clinical variables (mean±SD)  Systolic blood pressure (mm Hg)† 127.6±29.1 131.0±31.2 130.9±30.5 * *  Diastolic blood pressure (mm Hg)† 79.6±22.6 87.8±23.4 81.4±23.1 * *  Heart rate† 80.2±18.1 82.5±17.2 80.3±17.4 * 0.55  Respiratory rate† 23.3±27.2 23.1±25.7 22.3±23.5 0.09 *  Blood urea nitrogen (mg/dL)† 30.6±16.9 28.7±17.2 30.2±17.2 * 0.11  Oxygen saturation (%)† 94.1±11.8 95.2±11.5 95.2±11.4 * *  Creatinine (mg/dL)† 1.6±0.8 1.8±1.1 1.7±1.0 * *  Potassium (mEq/L)† 4.2±0.6 4.1±0.6 4.3±0.6 * *  Hematocrit (%)† 36.5±6.1 36.6±6.2 36.2±6.3 0.17 *  B-type natriuretic peptide (pg/mL)† 1344.3±1654.8 1538.5±1591.8 1348.1±1423.3 * 0.87  Body mass index (kg/m2)† 31.5±7.9 30.7±8.0 30.0±7.5 * *  Left ventricular ejection fraction (%)† 39.8±15.2 34.5±15.6 39.1±15.6 * *  Sodium (mEq/L)† 138.0±4.3 139.0±3.9 138.5±4.3 * *

Mean±SD reported for continuous variables; n (%) reported for categorical variables.

*P<0.05; PWA: P-value comparing Whites to Blacks; PWH: P-value comparing Whites to Hispanics.

†Summary statistics reported among nonmissing values.

TABLE 2 - Comparison of Baseline Characteristics for Veterans Hospitalized With Pneumonia Stratified by Race/Ethnicity Groups n (%) Characteristics White (N=101,868) Black (N=19,590) Hispanic (N=6324) P WA P WH Sociodemographic variables  Age (mean±SD) (y) 71.6±11.8 66.7±12.9 72.9±13.5 * *  Male 99,053 (97.2) 18,820 (96.1) 6225 (98.4) * *  Area Deprivation Index (mean±SD)† 105.1±13.2 108.4±13.0 106.9±13.4 * *  Homeless 11,461 (11.3) 3624 (18.5) 916 (14.5) * *  Discharged from nursing home 4439 (4.4) 988 (5.0) 172 (2.7) * *  Enrollment priority   Group 1 28,507 (28.0) 5019 (25.6) 1727 (27.3) * *   Group 2 5695 (5.6) 1194 (6.1) 301 (4.8)   Group 3 9521 (9.4) 1813 (9.3) 522 (8.3)   Group 4 6463 (6.3) 1410 (7.2) 755 (11.9)   Group 5 45,765 (44.9) 9386 (47.9) 2855 (45.2)   Group 6 847 (0.8) 157 (0.8) 43 (0.7)   Group 7+8 5070 (5.0) 611 (3.1) 121 (1.9)  Nursing home stay   None 87,893 (86.3) 16,872 (86.1) 5733 (90.7) * *   <90 d 10,521 (10.3) 1878 (9.6) 430 (6.8)   >90 d 3454 (3.4) 840 (4.3) 161 (2.6) Claims-based variables  Percutaneous transluminal coronary angioplasty 5647 (5.5) 543 (2.8) 192 (3.0) * *  Coronary artery bypass graft 9228 (9.1) 640 (3.3) 394 (6.2) * *  Congestive heart failure 30,086 (29.5) 5306 (27.1) 1388 (21.9) * *  Acute myocardial infarction 3063 (3.0) 450 (2.3) 256 (4.0) * *  Ischemic heart disease 4638 (4.6) 775 (4.0) 333 (5.3)