Abstract
Background
Social determinants of health, including neighborhood deprivation measured by the Area Deprivation Index, are key drivers of disparities in health outcomes, including survival after cardiac surgery. We evaluated whether the association between neighborhood deprivation and survival differs across racial and ethnic groups undergoing coronary artery bypass grafting.
Methods
We conducted a retrospective study of 739 335 Medicare beneficiaries who underwent isolated coronary artery bypass grafting between 2008 and 2019. A flexible parametric survival model with time‐dependent effects was implemented to estimate standardized (over demographics, clinical, and procedural characteristics) survival probabilities. Primary end points were differences in 30‐day and 5‐year standardized survival probabilities between beneficiaries from the lowest and highest quintiles of neighborhood deprivation (LDNs and HDNs), stratified by race and ethnicity.
Results
Unadjusted median survival was substantially higher among beneficiaries in LDNs compared with HDNs (10.65 [95% CI, 10.55–10.76] versus 9.20 [95% CI, 9.14–9.27] years). In our risk‐adjusted model, the magnitude of this difference varied significantly by race and ethnicity. At 30 days, standardized survival differences between LDNs and HDNs were 1.49% [95% CI, 0.45–2.53] among Asian American or Native Hawaiian/Pacific Islander, 1.06% [95% CI, 0.94–1.19] among White, 0.94% [95% CI,0.583–1.31] among Black, and 0.56% [95% CI, 0.25–0.91] among Hispanic beneficiaries. At 5 years, Asian American or Native Hawaiian/Pacific Islander and White beneficiaries showed the largest standardized survival differences between LDNs and HDNs (8.07% [95% CI, 5.37–10.77] and 5.01% [95% CI, 4.62 –5.39], respectively), whereas Black and Hispanic beneficiaries had smaller differences (2.00% [95% CI, 0.71–3.28] and 1.15% [95% CI, 0.05–2.26], respectively).
Conclusions
Although LDN (compared with HDN) residence was associated with improved survival after coronary artery bypass grafting, these survival differences were not equally distributed across race and ethnicity. Policies aiming to reduce socioeconomic disadvantage may yield uneven outcomes unless tailored to the specific challenges faced by different racial and ethnic populations.
Keywords: area deprivation index (ADI), coronary artery bypass grafting (CABG), ethnicity, race, social determinants of health (SDOH)
Subject Categories: Disparities, Quality and Outcomes, Cardiovascular Surgery
Nonstandard Abbreviations and Acronyms
- AANHPI
non‐Hispanic Asian American or Native Hawaiian/Pacific Islander
- ADI
area deprivation index
- HDN
highest deprivation neighborhood (ADI quintile 80%–100%)
- LDN
lowest deprivation neighborhood (ADI quintile 1%–20%)
- SDOH
social determinants of health
- SSD
standardized survival difference
- SSP
standardized survival probability
Clinical Perspective.
What Is New?
Greater neighborhood deprivation correlates with worse long‐term survival after coronary artery bypass grafting, though the impact of deprivation differs markedly by race and ethnicity.
What Are the Clinical Implications?
The survival benefit of living in less deprived neighborhoods was most evident among White and Asian American or Native Hawaiian/Pacific Islander patients, but less pronounced in Black and Hispanic patients, suggesting that structural inequities may erode the health advantages of neighborhood affluence.
Addressing socioeconomic deprivation without parallel efforts to mitigate structural and procedural inequities risks widening racial and ethnic disparities in cardiac surgical outcomes; equitable perioperative care models and community‐level interventions are needed to ensure all populations benefit equally from advances in cardiac surgery.
Race and ethnicity have long been associated with disparities in health outcomes and have been integrated into clinical risk assessment tools, including the Society of Thoracic Surgeons risk calculator. 1 , 2 However, the use of race or ethnicity alone as a proxy for genetic, cultural, or socioeconomic factors has been increasingly questioned, as it does not account for the nuanced interplay between race and social determinants of health (SDOH), potentially dissuading clinicians from recommending surgery due to perceived increased risk. 2 , 3 Indeed, socioeconomic disparities are increasingly recognized as critical determinants of health outcomes, particularly in patients undergoing complex medical and surgical interventions. 3 , 4 , 5
Despite this growing body of evidence, the interaction between race and ethnicity and SDOH and their respective associations with health outcomes has been inadequately explored, especially among patients undergoing cardiac surgery. Limited data exist on how survival disparities vary across racial and ethnic groups when factors related to SDOH are considered simultaneously. 4 , 5 Additionally, improvements in SDOH metrics may yield differential benefits depending on a patient’s race and ethnicity, further emphasizing the need for tailored approaches to reducing disparities in health care outcomes. Understanding the potential interactions between race and ethnicity and SDOH is crucial for designing equitable health care policies and improving clinical decision‐making tools.
Differences in access to health care resources, structural and institutional racism, and procedural disparities such as frequency of arterial grafting may be associated with both beneficiary race and ethnicity and neighborhood disadvantage. 5 , 6 , 7 , 8 , 9 Given this, we hypothesized that a quantitative measure of neighborhood deprivation, the area deprivation index (ADI), may have variable effects on survival across racial and ethnic groups. Therefore, we designed this study to evaluate whether differences in post–coronary artery bypass grafting (CABG) survival among Medicare beneficiaries residing in neighborhoods of low versus high deprivation vary by beneficiary race and ethnicity.
METHODS
Study Design
Data from the Centers for Medicare and Medicaid Services administrative claims database from 2008 to 2019 were retrospectively reviewed. Data used in this analysis are not publicly available due to privacy regulations but can be obtained from the Centers for Medicare and Medicaid Services through a formal data use agreement and purchase process. Analytic code and materials specific to this analysis, beyond those already included in Data S1, are available from the corresponding author upon reasonable request. The institutional review board of Baylor Scott & White Research Institute approved this study (approval #019‐270). Individual consent was waived due to the retrospective nature of the study.
Study Population
A 100% sample of Medicare beneficiary hospitalizations during which CABG was performed was identified from the Medicare Provider Analysis and Review file using International Classification of Diseases, Ninth Revision (ICD‐9) and Tenth Revision (ICD‐10) codes (Table S1). 10 Surgeon‐billed Current Procedural Terminology codes were matched from the Carrier (Fee‐for‐Service) file to doubly‐adjudicate procedural details as previously described (Table S2). 11 Beneficiaries enrolled in a Medicare Advantage plan within 1 year of the index procedure were excluded, as its inclusion may result in incomplete documentation of comorbid conditions. 12 Beneficiaries who underwent concomitant cardiovascular procedures other than atrial fibrillation ablation or coronary endarterectomy were excluded (Tables S3 and S4).
Preoperative comorbidity data were obtained from the Chronic Conditions segment of the Master Beneficiary Summary File; diagnoses were confirmed as preexisting if the earliest date of documentation occurred before the date of surgery. 13 Diagnoses derived from Chronic Conditions‐Master Beneficiary Summary File data have been shown to remain stable over the ICD‐9/10 transition and avoid the issue of diagnosis nonpersistence associated with other flag‐based methodologies of determining comorbidities. 14 , 15 , 16
Race and Neighborhood Deprivation
Beneficiary demographics, including race and ethnicity, were ascertained from the Master Beneficiary Summary File. The index race and ethnicity variables are coded based on self‐reported responses to the Social Security Administration SS‐5 application, and since 1994 include “non‐Hispanic White” (White), “non‐Hispanic Black/African American” (Black), “Hispanic/Latino” (Hispanic), “non‐Hispanic Asian American or Native Hawaiian/Pacific Islander” (AANHPI), “non‐Hispanic American Indian or Alaska Native,” Other, and Unknown. 17 For analytic purposes, race and ethnicity were determined using the Centers for Medicare and Medicaid Services imputation algorithm rather than self‐reported values. 18 Although this approach may introduce imputation‐related bias, prior reports have demonstrated substantially improved sensitivity for identifying Hispanic and AANHPI beneficiaries using imputation‐based (as opposed to self‐reported) values. 19 Self‐reported race and ethnicity were included for descriptive purposes, and a sensitivity analysis using the self‐reported (as opposed to imputed) race and ethnicity values was conducted.
The ADI is a composite measure of neighborhood‐level socioeconomic disadvantage, derived from 17 US Census indicators reflecting education, employment, housing quality, and poverty levels. 6 ADI data are calculated at the census block‐group level (typical population of 600–3000 residents) and range from 1 (least deprived neighborhoods) to 100 (most deprived neighborhoods). These values have been mapped to >69 million 9‐digit zip codes, enabling the geographically precise assignment of census data. 6
Neighborhood deprivation was assessed using national 2015 and 2019 ADI values based on the 9‐digit zip code of the beneficiary’s residence at the time of surgery. Given that ADI data before 2015 are not available, 2015 ADI values were assigned to beneficiaries undergoing CABG from 2008 to 2015, and 2019 ADI values were assigned to beneficiaries undergoing CABG from 2016 to 2019. A sensitivity analysis comparing 2015 and 2019 ADI values demonstrated a high degree of correlation (Figure S1), supporting the use of 2015 ADI data to beneficiaries undergoing surgery in years prior. Beneficiaries missing 9‐digit zip code data or without a corresponding ADI were excluded. Neighborhoods were rank‐ordered by ADI and categorized into quintiles.
Hospital and Surgeon Volume
Annual hospital and surgeon CABG volumes in Medicare beneficiaries were determined using the Medicare provider number and National Provider Identifier, respectively. CABG volumes were calculated among Medicare beneficiaries only; thus, Medicare CABG volumes were used as surrogates for total CABG volumes.
Statistical Analysis
To guide covariate selection for risk adjustment, a causal directed acyclic graph was developed incorporating beneficiary demographics (eg, age, sex, race and ethnicity, ADI, state of residence, insurance status), preexisting comorbidities (eg, diabetes, prior myocardial infarction, chronic kidney disease, heart failure), procedural characteristics (eg, number of arterial/venous grafts, off‐pump approach, urgency of admission), and hospital/provider‐related characteristics (eg, hospital/surgeon CABG volume) (Figure S2a,b). 20 The directed acyclic graph was constructed using subject‐matter knowledge, with input from all study authors. Using covariates identified through the directed acyclic graph–informed adjustment set, a flexible parametric survival model on the log‐cumulative hazard scale was fitted (Table S5). Restricted cubic splines with varying spline knots were used for age, hospital volume, surgeon volume, and the baseline hazard function (Figures S3–S5) to capture their nonlinear associations with survival. 21 Spline terms were tested with an increasing number of degrees of freedom (up to 5), in a factorial fashion; we selected the final model according to the Akaike and Bayesian information criteria. Spline knots were placed at their default position, determined by percentiles of the observed data distribution. Models were constructed with and without an interaction term between the 2 exposure variables (ADI quintile and race and ethnicity). There was a significant interaction between race and ethnicity and ADI quintiles (likelihood‐ratio test P<0.001), and thus this interaction term was included in our final model. We also considered time‐dependent effects (nonproportional hazards) for the 2 key exposure variables (ADI and race and ethnicity) and their interaction, which we modeled using restricted cubic splines with 3 degrees of freedom.
To risk‐adjust for measured confounding variables, regression standardization while varying the 2 exposure variables of interest (race and ethnicity and ADI) was performed using the previously described flexible parametric model to obtain standardized (marginal) survival probability (SSP) and difference (SSD) curves (Methods S1). 22 Specifically, SSPs were calculated for each quintile of neighborhood deprivation (standardizing across all variables in our model, including race) and then recalculated for beneficiaries from lowest deprivation neighborhoods (LDNs; ADI quintile 1%–20%) and highest deprivation neighborhoods (HDNs; ADI quintile 80%–100%) by fixing race and ethnicity and standardizing over the remaining covariates; SSDs between LDN and HDN residents for each race and ethnicity were calculated as well. We report these measures because differences in survival probability can be interpreted as risk differences, avoiding the troublesome interpretation of hazard ratios. 23 All survival estimates are reported with 95% CIs.
Study End Point
The study end point was all‐cause mortality. Vital status and date of death were determined from the Master Beneficiary Summary File. Patients who were alive as of the data set’s last available follow‐up date (December 31, 2019) were censored. Median duration of follow‐up was calculated using the reverse Kaplan–Meier method. Unadjusted survival was evaluated using the Kaplan–Meier method, with the index date defined as the date of CABG. Thirty‐day and 5‐year SSPs and SSDs were calculated for White, Black, Hispanic, and AANHPI beneficiaries from the LDNs and HDNs.
Data processing and statistical analyses were conducted using SAS (version 9.4) and Stata software, version Stata/MP 18.0 (StataCorp, College Station, TX).
RESULTS
Between January 1, 2008, and December 31, 2019, a total of 1 520 299 Medicare hospitalizations for CABG were identified, of which 739 335 met inclusion criteria as isolated CABG procedures with adequate data for analysis (Figure 1). Among excluded cases, 9084 lacked 9‐digit zip code data, and 20 588 had 9‐digit zip codes that could not be matched to an ADI value. Our cohort included 102 023 beneficiaries from the LDNs and 152 847 from the HDNs.
Figure 1. Patient selection flow diagram for the derivation of the study cohort.
ADI indicates area deprivation index; CABG, coronary artery bypass graft; CPT, Current Procedural Terminology; and ICD‐9/10, International Classification of Diseases, Ninth Revision and Tenth Revision.
Imputed (self‐reported) race and ethnicity distributions were as follows: 83.9% (87.4%) White, 6.4% (6.5%) Black, 5.7% (1.5%) Hispanic, 1.9% (1.4%) AANHPI, 0.6% (0.6%) American Indian or Alaska Native, and 1.5% (2.5%) categorized as other or unavailable (Table 1).
Table 1.
Demographics, Comorbidities, and Operative Characteristics, for Entire Cohort (Shaded Gray), and Stratified by ADI Quintiles
| All CABG Recipients N=739 335 | Low ADI 1%–20% N=102 023 | Low‐Mod ADI 20%–40% N=148 673 | Mod ADI 40%–60% N=167 568 | Mod‐High ADI 60%–80% N=168 224 | High ADI 80%–100% N=152 847 | |
|---|---|---|---|---|---|---|
| Age, y | 71.5 ± 8.1 | 73.2 ± 7.3 | 72.5 ± 7.4 | 71.8 ± 7.8 | 70.9 ± 8.3 | 69.7 ± 8.9 |
| <65 | 90 219 (12.2) | 5880 (5.8) | 11 399 (7.7) | 17 736 (10.6) | 24 126 (14.3) | 31 078 (20.3) |
| 65–69 | 198 098 (26.8) | 25 951 (25.4) | 39 662 (26.7) | 45 430 (27.1) | 45 752 (27.2) | 41 303 (27.0) |
| 70–74 | 186 760 (25.3) | 27 057 (26.5) | 39 543 (26.6) | 43 294 (25.8) | 41 827 (24.9) | 35 039 (22.9) |
| 75–79 | 148 397 (20.1) | 22 786 (22.3) | 32 084 (21.6) | 34 476 (20.6) | 32 556 (19.4) | 26 495 (17.3) |
| 80–84 | 86 413 (11.7) | 14 416 (14.1) | 19 218 (12.9) | 20 047 (12.0) | 18 219 (10.8) | 14 513 (9.5) |
| ≥85 | 29 448 (4.0) | 5933 (5.8) | 6767 (4.6) | 6585 (3.9) | 5744 (3.4) | 4419 (2.9) |
| Male sex | 524 006 (70.9) | 77 820 (76.3) | 109 396 (73.6) | 119 358 (71.2) | 116 108 (69.0) | 101 324 (66.3) |
| Imputed race | ||||||
| White | 620 014 (83.9) | 83 127 (81.5) | 127 636 (85.9) | 147 078 (87.8) | 145 105 (86.3) | 117 068 (76.6) |
| Black | 47 627 (6.4) | 3478 (3.4) | 6257 (4.2) | 7990 (4.8) | 10 397 (6.2) | 19 505 (12.8) |
| Hispanic | 41 771 (5.7) | 5885 (5.8) | 7819 (5.3) | 7399 (4.4) | 8398 (5.0) | 12 270 (8.0) |
| AANHPI | 14 284 (1.9) | 6459 (6.3) | 3679 (2.5) | 1966 (1.2) | 1338 (0.8) | 842 (0.6) |
| AIAN | 4565 (0.6) | 134 (0.1) | 428 (0.3) | 808 (0.5) | 1209 (0.7) | 1986 (1.3) |
| Other/Unknown | 11 074 (1.5) | 2940 (2.9) | 2854 (1.9) | 2327 (1.4) | 1777 (1.1) | 1176 (0.8) |
| Self‐reported race | ||||||
| White | 646 459 (87.4) | 86 665 (84.9) | 132 521 (89.1) | 151 712 (90.5) | 150 458 (89.4) | 125 103 (81.8) |
| Black | 48 267 (6.5) | 3595 (3.5) | 6384 (4.3) | 8123 (4.8) | 10 497 (6.2) | 19 668 (12.9) |
| Hispanic | 11 304 (1.5) | 1502 (1.5) | 2088 (1.4) | 1984 (1.2) | 2369 (1.4) | 3361 (2.2) |
| AANHPI | 10 089 (1.4) | 4348 (4.3) | 2649 (1.8) | 1478 (0.9) | 1000 (0.6) | 614 (0.4) |
| AIAN | 4493 (0.6) | 138 (0.1) | 432 (0.3) | 788 (0.5) | 1184 (0.7) | 1951 (1.3) |
| Other/Unknown | 18 723 (2.5) | 5775 (5.7) | 4599 (3.1) | 3483 (2.1) | 2716 (1.6) | 2150 (1.4) |
| Insurance—dual eligible (Medicare/Medicaid) | 119 127 (16.1) | 12 195 (12.0) | 16 038 (10.8) | 20 678 (12.3) | 29 160 (17.3) | 41 056 (26.9) |
| US Department of Health and Human Services region | ||||||
| 1 (HQ: Boston) | 32 301 (4.4) | 9447 (9.3) | 12 093 (8.1) | 6723 (4.0) | 2991 (1.8) | 1047 (0.7) |
| 2 (HQ: New York) | 54 393 (7.4) | 23 227 (22.8) | 13 915 (9.4) | 7369 (4.4) | 5653 (3.4) | 4229 (2.8) |
| 3 (HQ: Philadelphia) | 75 629 (10.2) | 12 586 (12.3) | 19 959 (13.4) | 18 257 (10.9) | 13 808 (8.2) | 11 019 (7.2) |
| 4 (HQ: Atlanta) | 186 775 (25.3) | 11 529 (11.3) | 28 780 (19.4) | 43 083 (25.7) | 51 720 (30.7) | 51 663 (33.8) |
| 5 (HQ: Chicago) | 131 642 (17.8) | 7653 (7.5) | 23 114 (15.5) | 35 285 (21.1) | 36 241 (21.5) | 29 349 (19.2) |
| 6 (HQ: Dallas) | 107 788 (14.6) | 5258 (5.2) | 14 315 (9.6) | 23 098 (13.8) | 29 754 (17.7) | 35 363 (23.1) |
| 7 (HQ: Kanas City) | 44 787 (6.1) | 1350 (1.3) | 6108 (4.1) | 11 141 (6.6) | 13 921 (8.3) | 12 267 (8.0) |
| 8 (HQ: Denver) | 17 843 (2.4) | 2254 (2.2) | 5376 (3.6) | 5031 (3.0) | 3349 (2.0) | 1833 (1.2) |
| 9 (HQ: San Francisco) | 64 479 (8.7) | 24 922 (24.4) | 16 554 (11.1) | 10 546 (6.3) | 7444 (4.4) | 5013 (3.3) |
| 10 (HQ: Seattle) | 23 571 (3.2) | 3754 (3.7) | 8422 (5.7) | 7010 (4.2) | 3333 (2.0) | 1052 (0.7) |
| Area deprivation index | 53.9 ± 26.8 | 11.4 ± 5.6 | 30.9 ± 5.7 | 50.5 ± 5.7 | 70.5 ± 5.8 | 89.8 ± 5.6 |
| Dialysis dependent | 34 185 (4.6) | 4339 (4.3) | 6041 (4.1) | 6916 (4.1) | 7652 (4.5) | 9237 (6.0) |
| Chronic Condition‐Master Beneficiary Summary File comorbidities | ||||||
| Hypertension | 661 202 (89.4) | 90 241 (88.5) | 131 928 (88.7) | 149 271 (89.1) | 150 816 (89.7) | 138 946 (90.9) |
| Hyperlipidemia | 634 594 (85.8) | 90 127 (88.3) | 129 477 (87.1) | 143 992 (85.9) | 142 405 (84.7) | 128 593 (84.1) |
| Diabetes | 379 011 (51.3) | 50 208 (49.2) | 72 874 (49.0) | 83 880 (50.1) | 87 322 (51.9) | 84 727 (55.4) |
| Ischemic heart disease | 723 586 (97.9) | 100 095 (98.1) | 145 460 (97.8) | 163 768 (97.7) | 164 446 (97.8) | 149 817 (98.0) |
| Prior acute myocardial infarction | 121 405 (16.4) | 16 060 (15.7) | 23 877 (16.1) | 26 556 (15.8) | 28 078 (16.7) | 26 834 (17.6) |
| Congestive heart failure | 273 171 (37.0) | 37 027 (36.3) | 51 436 (34.6) | 59 067 (35.2) | 62 796 (37.3) | 62 845 (41.1) |
| Atrial fibrillation | 91 214 (12.3) | 14 112 (13.8) | 19 503 (13.1) | 20 740 (12.4) | 20 101 (11.9) | 16 758 (11.0) |
| Chronic kidney disease | 246 281 (33.3) | 32 799 (32.1) | 48 557 (32.7) | 54 895 (32.8) | 55 914 (33.2) | 54 116 (35.4) |
| Chronic obstructive pulmonary disease | 201 105 (27.2) | 20 452 (20.0) | 34 074 (22.9) | 43 807 (26.1) | 50 386 (30.0) | 52 386 (34.3) |
| Prior hip fracture | 9551 (1.3) | 1346 (1.3) | 1858 (1.2) | 2167 (1.3) | 2240 (1.3) | 1940 (1.3) |
| Depression | 182 641 (24.7) | 20 373 (20.0) | 33 084 (22.3) | 40 441 (24.1) | 44 262 (26.3) | 44 481 (29.1) |
| Stroke/transient ischemic attack | 109 831 (14.9) | 14 746 (14.5) | 21 024 (14.1) | 24 235 (14.5) | 25 323 (15.1) | 24 503 (16.0) |
| Alzheimer/dementia | 42 756 (5.8) | 6192 (6.1) | 8491 (5.7) | 9156 (5.5) | 9520 (5.7) | 9397 (6.1) |
| History of cancer | 92 921 (12.6) | 15 357 (15.1) | 20 312 (13.7) | 21 217 (12.7) | 19 621 (11.7) | 16 414 (10.7) |
| Hospital CABG volume | 199 ± 143 | 212 ± 172 | 200 ± 142 | 196 ± 136 | 197 ± 136 | 196 ± 139 |
| Surgeon CABG volume | 53 ± 31 | 46 ± 29 | 51 ± 29 | 53 ± 30 | 55 ± 32 | 56 ± 33 |
| Surgery year [interquartile interval] | 2013 [2010–2016] | 2013 [2010–2016] | 2013 [2010–2016] | 2013 [2010–2016] | 2013 [2010–2016] | 2012 [2010–2016] |
| Surgery urgency | ||||||
| Elective | 362 889 (49.1) | 49 500 (48.5) | 74 164 (49.9) | 83 530 (49.8) | 82 447 (49.0) | 73 248 (47.9) |
| Urgent | 176 215 (23.8) | 23 219 (22.8) | 34 460 (23.2) | 39 407 (23.5) | 41 416 (24.6) | 37 713 (24.7) |
| Emergent | 197 381 (26.7) | 28 865 (28.3) | 39 504 (26.6) | 44 080 (26.3) | 43 749 (26.0) | 41 183 (26.9) |
| Other/Unknown | 2850 (0.4) | 439 (0.4) | 545 (0.4) | 551 (0.3) | 612 (0.4) | 703 (0.5) |
| Redo sternotomy | 25 027 (3.4) | 2752 (2.7) | 4593 (3.1) | 5728 (3.4) | 6237 (3.7) | 5717 (3.7) |
| Atrial fibrillation ablation | 25 362 (3.4) | 3572 (3.5) | 5380 (3.6) | 5811 (3.5) | 5852 (3.5) | 4747 (3.1) |
| Coronary endarterectomy | 16 101 (2.2) | 1951 (1.9) | 3064 (2.1) | 3547 (2.1) | 3823 (2.3) | 3716 (2.4) |
| Off‐pump CABG | 124 084 (16.8) | 18 008 (17.7) | 25 000 (16.8) | 27 628 (16.5) | 27 900 (16.6) | 25 548 (16.7) |
| No. arterial graft(s) | ||||||
| 0 | 47 466 (6.4) | 4792 (4.7) | 8184 (5.5) | 10 603 (6.3) | 11 949 (7.1) | 11 938 (7.8) |
| 1 | 628 208 (85.0) | 85 629 (83.9) | 126 099 (84.8) | 142 340 (84.9) | 143 536 (85.3) | 130 604 (85.4) |
| ≥2 | 63 661 (8.6) | 11 602 (11.4) | 14 390 (9.7) | 14 625 (8.7) | 12 739 (7.6) | 10 305 (6.7) |
| No. venous graft(s) | ||||||
| 0 | 45 753 (6.2) | 7208 (7.1) | 9820 (6.6) | 10 389 (6.2) | 9847 (5.9) | 8489 (5.6) |
| 1 | 132 778 (18.0) | 18 961 (18.6) | 26 707 (18.0) | 29 756 (17.8) | 29 776 (17.7) | 27 578 (18.0) |
| 2 | 294 316 (39.8) | 41 143 (40.3) | 59 465 (40.0) | 66 349 (39.6) | 66 875 (39.8) | 60 484 (39.6) |
| 3 | 201 031 (27.2) | 26 733 (26.2) | 39 845 (26.8) | 45 895 (27.4) | 46 381 (27.6) | 42 177 (27.6) |
| ≥4 | 65 457 (8.9) | 7978 (7.8) | 12 836 (8.6) | 15 179 (9.1) | 15 345 (9.1) | 14 119 (9.2) |
Values are presented as mean ± SD or median [interquartile interval] for continuous variables and as count (percentage) for categorical variables.
AANHPI indicates Asian American, Native Hawaiian, or Pacific Islander; ADI, area deprivation index; AIAN, American Indian or Alaska Native; and HQ, headquarters.
The median follow‐up duration was 6.67 years (95% CI, 6.65–6.69), with an interquartile interval of 3.43 years (95% CI, 3.41–3.44) to 9.61 years (95% CI, 9.60–9.63).
Patient Demographics and Operative Characteristics by Area Deprivation
Significant disparities were observed in demographic and clinical characteristics when stratified by ADI (Table 1). Patients in HDNs were younger (mean age: 69.7 ± 8.9 years versus 73.2 ± 7.3 years), more frequently female (33.7% versus 23.7%), and more likely to qualify for dual Medicare and Medicaid coverage (26.9% versus 12.0%), a marker of poverty. These patients also exhibited a greater burden of comorbidities at the time of surgery, including congestive heart failure (41.1% versus 36.3%), diabetes (55.4% versus 49.2%), chronic kidney disease (35.4% versus 32.1%), dialysis dependence (6.0% versus 4.3%), chronic obstructive pulmonary disease (34.3% versus 20.0%), and depression (29.1% versus 20.0%). Conversely, patients in LDNs had higher rates of atrial fibrillation (13.8% versus 11.0%) and history of malignancy (15.1% versus 10.7%).
Although hospital and surgeon CABG volumes and procedural urgency did not differ significantly by neighborhood deprivation, patients in HDNs more frequently required redo sternotomy (3.7% versus 2.7%) and coronary endarterectomy (2.4% versus 1.9%) at the time of surgery. Additionally, these patients were more likely to undergo all‐venous grafting (7.8% versus 4.7%) and were less likely to receive multiarterial grafting (6.7% versus 11.4%).
Patient Demographics and Operative Characteristics by Race and Ethnicity
White and AANHPI beneficiaries disproportionately resided in LDNs (81.5% and 6.3%, respectively) compared with HDNs (76.6% and 0.6%, respectively). In contrast, Black and Hispanic beneficiaries disproportionately resided in HDNs (12.8% and 8.0%, respectively) compared with LDNs (3.4% and 5.8%, respectively). Geographically, beneficiaries in the Northeast and West Coast were more likely to live in LDNs, whereas those in the Midwest and South were more likely to reside in HDNs.
Unadjusted Kaplan–Meier Survival Outcomes by Neighborhood Deprivation
Of 739 335 beneficiaries undergoing CABG, 242 062 were deceased, and 497 273 were alive at the date of last known follow‐up. Unadjusted Kaplan–Meier analysis demonstrated a median survival of 9.89 (95% CI, 9.86–9.92) years for the entire cohort, with 30‐day, 5‐year, and 10‐year survival rates of 97.0%, 75.6%, and 49.4%, respectively (Figure S6). When stratified by neighborhood deprivation (without risk‐adjustment), patients residing in LDNs had significantly longer median survival compared with HDNs (10.65 [95% CI, 10.55–10.76] versus 9.20 [95% CI, 9.14–9.27] years, respectively) (Figure 2A).
Figure 2. Survival after CABG by neighborhood deprivation.
A, Unadjusted Kaplan–Meier survival. B, Standardized survival probability (risk‐adjusted for all measured covariates). Neighborhood deprivation categorized by national ADI quintiles, with low ADI indicating least‐deprived and high ADI indicating most‐deprived neighborhoods. Shaded regions represent 95% CIs. ADI indicates area deprivation index.
Flexible Parametric Survival Model
A flexible parametric survival model was created to risk‐adjust for beneficiary demographics (eg, age, sex, race and ethnicity, ADI, state of residence, insurance status), preexisting comorbidities (eg, diabetes, prior myocardial infarction, chronic kidney disease, heart failure), procedural characteristics (eg, number of arterial/venous grafts, off‐pump approach, urgency of admission), and hospital/provider‐related characteristics (eg, hospital/surgeon CABG volume) (Table S5). SSPs standardizing across all measured covariates, including race, were generated to demonstrate the association of level of neighborhood deprivation with survival after risk‐adjustment (Figure 2B).
Standardized Survival Probabilities and Differences Comparing LDNs and HDNs
Using this model, SSPs were calculated for beneficiaries residing in LDNs and HDNs across White, Black, and AANHPI race and Hispanic ethnicity (Figures 3A through 3D). Although LDN beneficiaries had higher SSPs compared with HDN beneficiaries across all races and ethnicities, the degree of change varied by race and ethnicity (Figures 4A and 4B). For example, the 30‐day SSDs between residents of LDNs and HDNs were 1.5% (95% CI, 0.5–2.5) among AANHPI, 1.1% (95% CI, 0.9–1.2) among White, 0.9% (95% CI, 0.6–1.3) among Black, and 0.6% (95% CI, 0.3–0.9) among Hispanic beneficiaries, and 5‐year SSDs between residents of LDNs and HDNs were 8.1% (95% CI, 5.4–10.8) among AANHPI, 5.0% (95% CI, 4.6–5.4) among White, 2.0% (95% CI, 0.7–3.3) among Black, and 1.2% (95% CI, 0.1–2.3) among Hispanic beneficiaries (Table 2, Figure 5).
Figure 3. Standardized survival probabilities (standardized over all covariates aside from race and ADI listed in Data S1) after CABG.
SSPs are shown for White (A), Black (B), Hispanic (C), and AANHPI (D) beneficiaries residing in the lowest (dotted line) and highest (straight line) quintiles of neighborhood deprivation. AANHPI indicates Asian American, Native Hawaiian, or Pacific Islander; ADI, area deprivation index; and CABG, coronary artery bypass graft.
Figure 4. Thirty‐day and 5‐year standardized survival probabilities, stratified by race and ADI quintile.
A, SSP at 30 d. B, SSP at 5 y. Each race is represented by a distinct color: red/salmon for White, yellow/orange for Black, green for Hispanic, and blue for AANHPI. For each race, progressively darker color shades indicate increasing neighborhood deprivation (from lowest deprivation, ADI 1%–20%, to highest deprivation, ADI 81%–100%). AANHPI indicates Asian American, Native Hawaiian, or Pacific Islander; ADI, area deprivation index; and SSP, standardized survival probabilities.
Table 2.
Standardized Survival Probabilities and Differences at 30 Days, 1 Year, and 5 Years, for Beneficiaries in the Lowest and Highest Quintile of Neighborhood Deprivation, Stratified by Imputed Race and Ethnicity
| 30 D | 1 Y | 5 Y | ||
|---|---|---|---|---|
|
Standardized Survival Probabilities: Low Deprivation (ADI 1%–20%) |
White | 97.89% [97.80,97.98] | 93.08% [92.92,93.24] | 78.37% [78.09,78.67] |
| Black | 98.12% [97.79,98.44] | 92.86% [92.20,93.52] | 76.54% [75.36,77.73] | |
| Hispanic | 98.10% [97.84,98.36] | 93.61% [93.11,94.11] | 79.44% [78.52,80.36] | |
| AANHPI | 97.76% [97.48,98.04] | 93.73% [93.24,94.22] | 82.18% [81.32,83.05] | |
|
Standardized Survival Probabilities: High Deprivation (ADI 81%–100%) |
White | 96.83% [96.74,96.92] | 90.77% [90.62,90.92] | 73.37% [73.12,73.62] |
| Black | 97.17% [97.00,97.34] | 91.44% [91.14,91.74] | 74.54% [74.02,75.06] | |
| Hispanic | 97.52% [97.32,97.73] | 92.63% [92.28,92.99] | 78.28% [77.66,78.91] | |
| AANHPI | 96.27% [95.27,97.28] | 90.12% [88.51,91.76] | 74.11% [71.59,76.72] | |
|
Standardized Survival Differences: (SSP low deprivation)—(SSP high deprivation) |
White | 1.06% [0.94,1.19] | 2.31% [2.09,2.53] | 5.01% [4.62,5.39] |
| Black | 0.94% [0.58,1.31] | 1.42% [0.70,2.15] | 2.00% [0.71,3.28] | |
| Hispanic | 0.58% [0.25,0.91] | 0.97% [0.36,1.58] | 1.15% [0.05,2.26] | |
| AANHPI | 1.49% [0.45,2.53] | 3.61% [1.92,5.30] | 8.07% [5.37,10.77] |
ADI indicates area deprivation index; AANHPI, Asian American, Native Hawaiian, or Pacific Islander; SSD, standardized survival difference; and SSP, standardized survival probability.
95% CI shown in brackets. SSDs are determined by the difference between the SSPs of the low deprivation and the high deprivation cohorts (Figure 5).
Figure 5. Standardized survival differences comparing low (ADI 1%–20%) versus high (ADI 81%–100%) neighborhood deprivation, stratified by race.
Standardized survival probabilities, from which standardized survival differences are derived, are shown in Figure 3A–D. Standardized survival probabilities and differences are summarized in Table 2. Shaded regions represent 95% CIs. AANHPI indicates Asian American, Native Hawaiian, or Pacific Islander; and ADI, area deprivation index.
A sensitivity analysis using self‐reported race and ethnicity, as opposed to imputed, was conducted (Figures S7a–d, S8a,b, and S9), and results were consistent across both measures of race and ethnicity.
DISCUSSION
This study examined the association between neighborhood deprivation, as measured by ADI, and survival in nearly 740 000 Medicare beneficiaries undergoing isolated CABG between 2008 and 2019 in the United States. Our report comprises 3 key findings. First, beneficiaries from HDNs had distinct demographic profiles and a higher burden of comorbid conditions compared with those from LDNs. Second, procedural disparities were evident, with beneficiaries residing in HDNs being more likely to receive all venous grafting and less likely to receive multiarterial grafting. Third, the magnitude of predicted survival differences between beneficiaries from HDNs and LDNs varied by race and ethnicity.
We noted a higher proportion of White and AANHPI beneficiaries residing in LDNs and a higher proportion of Black and Hispanic beneficiaries residing in HDNs; geographic clustering was also evident, with LDN beneficiaries concentrated in the Northeast and West Coast, whereas HDNs were more common in the South and Midwest. These findings align with prior studies demonstrating regional variability in the distribution of socioeconomic disadvantage and its disproportionate impact on minoritized populations. 5 , 6 Patients from HDNs also exhibited greater financial vulnerability (more required state assistance to pay Medicare premiums), were younger, and were more frequently female. They also carried a higher burden of comorbidities at the time of CABG, including congestive heart failure, diabetes, chronic kidney disease, dialysis dependence, and depression. These patterns are concordant with findings from Patrick et al. and Thompson et al., who similarly reported that patients residing in socioeconomically disadvantaged areas undergoing CABG were more likely to be of races and ethnicities other than White, carry a higher comorbidity burden, and experience worse outcomes, including mortality, stroke, wound infection, and reoperation. 24 , 25 Conversely, LDN beneficiaries had higher rates of atrial fibrillation and malignancy, possibly reflecting differences in beneficiary age‐associated health risks at the time of CABG. 26
Procedural disparities were evident between beneficiaries of HDNs and LDNs who underwent CABG. Beneficiaries in HDNs were more likely to undergo redo sternotomy, coronary endarterectomy, and, most concerningly, receive all‐venous grafting despite the fact that the use of at least 1 internal mammary artery had a class I indication for CABG over the duration of the study. 27 This contrasted with beneficiaries in LDNs who more frequently underwent multiarterial grafting, which may be associated with improved long‐term survival. 7 , 28 , 29 The reasons for these differences are likely multifactorial and related to unmeasured factors such as variations in lesion complexity, local resource availability, institutional and surgeon bias, patient mistrust or limited advocacy, and broader social determinants including housing instability, language barriers, and insufficient outpatient care access. 24 , 25 Regardless of cause, these disparities serve as quantifiable quality metrics that could guide interventions ensuring that minoritized and socioeconomically disadvantaged patients receive guideline‐directed CABG care.
As demonstrated in prior studies by Patrick et al. and Thompson et al., lower socioeconomic status is associated with decreased long‐term survival following cardiac surgery. 24 , 25 These disparities are likely compounded by structural bias and racism experienced by minoritized populations. 8 When accounting for both population descriptors and their interactions in our model, both short‐ and long‐term predicted survival increased as a beneficiary’s level of neighborhood deprivation decreased; however, the extent of this increase varied by race and ethnicity. Specifically, we found that the difference in survival between residing in LDN (compared with HDN) was highest for AANHPI and White beneficiaries and more modest for Black and Hispanic beneficiaries. In other words, improvements in factors related to SDOH may be less likely to improve postoperative survival among patients from minoritized groups undergoing CABG. These findings suggest that improvements in SDOH alone may be insufficient to produce equitable postoperative outcomes across racial and ethnic groups. The attenuated benefit observed among minoritized populations possibly reflects the enduring influence of structural inequities, differential access to high‐quality health care, and variation in procedural delivery. 5 , 6 , 7 , 8 , 9 Additional contributing mechanisms may include disparities in hospital quality, cumulative psychosocial stress, provider bias, and diminished trust in the health care system among historically marginalized communities.
The observed differential impact on survival of reduced deprivation by race and ethnicity has important implications for public health and health care equity. Specifically, they underscore the need for targeted interventions aimed at addressing the unique challenges faced by minoritized populations, regardless of whether they reside in socioeconomically deprived neighborhoods. These patients may experience barriers to follow‐up care, cardiac rehabilitation, medication adherence, and lifestyle modifications due to factors such as transportation insecurity, housing instability, and lack of community‐based resources. 3 , 7 For example, the barbershop‐based blood pressure reduction program for Black men noted a significant improvement in hypertension control and demonstrated the potential effectiveness of culturally tailored, community‐engaged approaches to reducing cardiovascular health disparities. 30 Similarly, telehealth‐based cardiac rehabilitation programs have demonstrated increased uptake, improved physical activity, and enhanced medication adherence among patients with coronary heart disease. 31 Other well‐considered policy interventions aimed at improving patient education, access to primary care physicians, improving transitional care, and promoting evidence‐based surgical practices such as single‐arterial, or even multiarterial, grafting, are critical steps toward reducing disparities. 7 , 24 , 28
Interestingly, Hispanic beneficiaries from neighborhoods of high deprivation exhibited improved post‐CABG survival (both 30 days and 5 years) compared with Black, White, and AANHPI beneficiaries from HDNs. This finding has been observed in numerous prior studies noting that Hispanic patients experience decreased all‐cause mortality and adverse cardiovascular events despite a higher prevalence of cardiovascular comorbidities. 24 Although the underlying mechanism for these observed differences in the Hispanic population is unknown, it may be due to their typically younger age at presentation or the possible psychosocial benefits of strong social and familial relationships. 24 , 32 This observation also highlights the fact that individuals of different racial and ethnic backgrounds may experience the same neighborhood environment in fundamentally different ways. Structural factors such as redlining, residential segregation, and variable exposure to institutional bias can result in differing lived experiences within the same geographic area.
Limitations
Several limitations of our analysis require acknowledgement. The study’s retrospective design and reliance on administrative claims data may introduce biases related to unmeasured confounding and misclassification. Our analysis was limited to Medicare beneficiaries with traditional coverage, a predominantly older population, which may limit the generalizability of our findings to younger populations or differing insurance status. ADI data were available only for 2015 and 2019. Although sensitivity analyses demonstrated temporal stability of ADI values, applying 2015 ADI to patients undergoing CABG in 2008 introduces a potential risk of misclassification. The use of imputed race and ethnicity, although improving sensitivity for Hispanic and AANHPI race, may introduce bias related to the imputation algorithm; importantly, we performed a sensitivity analysis using self‐identified race, which yielded similar findings. 18 , 19 Our use of hospital and surgeon volume as proxies for institution and provider quality may not fully capture variation in care delivery. The absence of individual‐level socioeconomic data precluded multilevel (hierarchical) modeling, which would have allowed for more precise estimation of compositional (individual‐level) as opposed to contextual (neighborhood‐level) effects. As a result, residual confounding may persist. The exclusion of participants without available zip code or matched ADI information may also introduce bias, as this subgroup could disproportionately include individuals experiencing homelessness or housing instability. In a sensitivity analysis (Table S6; Figures S10 and S11), demographic and comorbidity profiles were similar across groups. However, beneficiaries missing zip code data exhibited worse unadjusted survival than all other cohorts, including those from the highest ADI quintile. We hypothesize that this population may experience disproportionately worse outcomes due to cumulative socioeconomic disadvantage, limited access to preventive and postoperative care, a higher likelihood of loss to follow‐up after hospital discharge, and greater exposure to environmental and psychosocial stressors that compound postoperative risk. Conversely, beneficiaries with 9‐digit zip codes but without a matched ADI demonstrated higher unadjusted survival relative to all other groups, including those from the lowest ADI quintile. We hypothesize that this finding most likely represents incomplete geographic mapping of certain postal codes, such as post office boxes or newly developed areas rather than a true clinical or socioeconomic distinction.
Although our focus on ADI extremes (highest and lowest quintiles) offers insight into the impact of neighborhood deprivation, it may oversimplify the spectrum of socioeconomic disadvantages. The interpretation of the interaction between race and ethnicity and ADI was conducted on a multiplicative scale, which may limit its applicability to public health or clinical decision‐making. Although an additive approach might better inform intervention strategies, the chosen model represents a balance between interpretability and statistical complexity. Finally, although we observed differences in survival across ADI quintiles, our study design does not allow for the inference that reducing neighborhood deprivation causally improves survival. Further research is needed to explore the mechanisms underlying the observed disparities in survival across levels of neighborhood deprivation.
CONCLUSIONS
Neighborhood deprivation and race and ethnicity are SDOHs that interact in complex ways to influence survival after CABG. Predicted differences in post‐CABG survival between patients from HDNs and LDNs were substantially greater for White and AANHPI patients than for their Black and Hispanic counterparts. These findings suggest that policies aimed at reducing neighborhood deprivation may not confer equitable survival benefits across racial and ethnic groups. Rather, our results highlight the need for targeted, equity‐focused interventions to improve access to high‐quality surgical care for at‐risk populations. Promising strategies include ensuring minority patients receive guideline‐based therapies, enhancing early and frequent postdischarge follow‐up for those residing in HDNs and minoritized populations, and implementing community‐based, culturally tailored programs to promote engagement and continuity of care. Ultimately, our study supports a more nuanced and equitable approach to cardiac surgical risk stratification, care delivery, and policymaking.
Sources of Funding
Data acquisition and effort of John J. Squiers, Taylor Pickering, and Jasjit K. Banwait supported by a philanthropic gift from Satish and Yasmin Gupta to Baylor Scott & White The Heart Hospital, Plano.
Disclosures
None.
Supporting information
Data S1
This article was sent to Mahasin S. Mujahid, PhD, MS, FAHA, Associate Editor, for review by expert referees, editorial decision, and final disposition.
Supplemental Material is available at https://www.ahajournals.org/doi/suppl/10.1161/JAHA.125.043095
For Sources of Funding and Disclosures, see page 13.
References
- 1. Flanagin A, Frey T, Christiansen SL; AMA Manual of Style committee . updated guidance on the reporting of race and ethnicity in medical and science journals. JAMA. 2021;326:621–627. doi: 10.1001/jama.2021.13304 [DOI] [PubMed] [Google Scholar]
- 2. Vyas DA, Eisenstein LG, Jones DS. Hidden in plain sight—reconsidering the use of race correction in clinical algorithms. N Engl J Med. 2020;383:874–882. doi: 10.1056/NEJMms2004740 [DOI] [PubMed] [Google Scholar]
- 3. Brandt EJ, Tobb K, Cambron JC, Ferdinand K, Douglass P, Nguyen PK, Vijayaraghavan K, Islam S, Thamman R, Rahman S, et al. Assessing and addressing social determinants of cardiovascular health: JACC state‐of‐the‐art review. J Am Coll Cardiol. 2023;81:1368–1385. doi: 10.1016/j.jacc.2023.01.042 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Ioannidis JPA, Powe NR, Yancy C. Recalibrating the use of race in medical research. JAMA. 2021;325:623–624. doi: 10.1001/jama.2021.0003 [DOI] [PubMed] [Google Scholar]
- 5. Elfaki LA, Nwakoby A, Keshishi M, Vervoort D, Yanagawa B, Fremes SE. Race and ethnicity in cardiac surgery: a missed opportunity? Ann Thorac Surg. 2024;117:714–722. doi: 10.1016/j.athoracsur.2023.10.025 [DOI] [PubMed] [Google Scholar]
- 6. Kind AJH, Buckingham WR. Making neighborhood‐disadvantage metrics accessible—the neighborhood atlas. N Engl J Med. 2018;378:2456–2458. doi: 10.1056/NEJMp1802313 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Wadhera RK, Bhatt DL, Kind AJH, Song Y, Williams KA, Maddox TM, Yeh RW, Dong L, Doros G, Turchin A, et al. Association of outpatient practice‐level socioeconomic disadvantage with quality of care and outcomes among older adults with coronary artery disease: implications for value‐based payment. Circ Cardiovasc Qual Outcomes. 2020;13:e005977. doi: 10.1161/CIRCOUTCOMES.119.005977 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Banerjee S, Cader FA, Gulati M; Quinn CapersIV . Racism and cardiology: a global call to action. CJC Open. 2021;3:S165–S173. doi: 10.1016/j.cjco.2021.09.014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Johnson AE, Zhu J, Garrard W, Thoma FW, Mulukutla S, Kershaw KN, Magnani JW. Area deprivation index and cardiac readmissions: evaluating risk‐prediction in an electronic health record. J Am Heart Assoc. 2021;10:e020466. doi: 10.1161/JAHA.120.020466 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Mehaffey JH, Schaffer JM, Hawkins RB, Mori M, Strobel R, Hameed I, Subramanyan R, Bowdish M, Badhwar V. Longitudinal datasets in cardiac surgery. Ann Thorac Surg. 2025;S0003‐4975(25)01157‐9. doi: 10.1016/j.athoracsur.2025.11.013 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Schaffer JM, Squiers JJ, Banwait JK, Hale S, Ryan WH, Mack MJ, DiMaio JM. Differences in administrative claims data for coronary artery bypass grafting between international classification of diseases, ninth revision and tenth revision coding. JAMA Cardiol. 2021;6:1094–1096. doi: 10.1001/jamacardio.2021.1595 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Meyers DJ, Johnston KJ. The growing importance of Medicare Advantage in health policy and health services research. JAMA Health Forum. 2021;2:e210235. doi: 10.1001/jamahealthforum.2021.0235 [DOI] [PubMed] [Google Scholar]
- 13. Schaffer JM, Squiers JJ, Kluis A, Banwait JK, Gaudino MFL, Mack MJ, DiMaio JM. Insurance‐related differences in Chronic Conditions Data Warehouse comorbidities of Medicare beneficiaries. Am J Manag Care. 2025;31:e336–e346. doi: 10.37765/ajmc.2025.89824 [DOI] [PubMed] [Google Scholar]
- 14. Ellis RP, Hsu HE, Song C, Kuo T‐C, Martins B, Siracuse JJ, Liu Y, Ash AS. Diagnostic category prevalence in 3 classification systems across the transition to the International Classification of Diseases, Tenth Revision, Clinical Modification. JAMA Netw Open. 2020;3:e202280. doi: 10.1001/jamanetworkopen.2020.2280 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Weeks WB, Huynh G, Cao SY, Smith J, Weinstein JN. Assessment of year‐to‐year patient‐specific comorbid conditions reported in the Medicare chronic conditions data warehouse. JAMA Netw Open. 2020;3:e2018176. doi: 10.1001/jamanetworkopen.2020.18176 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Tu N, Henderson M, Sundararajan M, Salas M. Discrepancies in ICD‐9/ICD‐10‐based codes used to identify three common diseases in cancer patients in real‐world settings and their implications for disease classification in breast cancer patients and patients without cancer: a literature review and descriptive study. Front Oncol. 2023;13:1016389. doi: 10.3389/fonc.2023.1016389 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Application for a Social Security Card. Social Security. Accessed June 10, 2025. https://www.ssa.gov/forms/ss‐5.pdf.
- 18. Filice CE, Joynt KE. Examining race and ethnicity information in Medicare administrative data. Med Care. 2017;55:e170–e176. doi: 10.1097/MLR.0000000000000608 [DOI] [PubMed] [Google Scholar]
- 19. Huang AW, Meyers DJ. Assessing the validity of race and ethnicity coding in administrative Medicare data for reporting outcomes among Medicare Advantage beneficiaries from 2015 to 2017. Health Serv Res. 2023;58:1045–1055. doi: 10.1111/1475-6773.14197 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Lipsky AM, Greenland S. Causal directed acyclic graphs. JAMA. 2022;327:1083–1084. doi: 10.1001/jama.2022.1816 [DOI] [PubMed] [Google Scholar]
- 21. Royston P, Parmar MKB. Flexible parametric proportional‐hazards and proportional‐odds models for censored survival data, with application to prognostic modelling and estimation of treatment effects. Stat Med. 2002;21:2175–2197. doi: 10.1002/sim.1203 [DOI] [PubMed] [Google Scholar]
- 22. Syriopoulou E, Wästerlid T, Lambert PC, Andersson TM‐L. Standardised survival probabilities: a useful and informative tool for reporting regression models for survival data. Br J Cancer. 2022;127:1808–1815. doi: 10.1038/s41416-022-01949-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Hernán MA. The hazards of hazard ratios. Epidemiology. 2010;21:13–15. doi: 10.1097/EDE.0b013e3181c1ea43 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Patrick WL, Bojko M, Han JJ, Kelly JJ, Iyengar A, Helmers M, Smood B, McCarthy FH, Williams ML. Neighborhood socioeconomic status is associated with differences in operative management and long‐term survival after coronary artery bypass grafting. J Thorac Cardiovasc Surg. 2022;164:92–102.e8. doi: 10.1016/j.jtcversus2020.08.024 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Thompson MP, Yaser JM, Fliegner MA, Syrjamaki JD, Nathan H, Sukul D, Theurer PF, Clark MJ, Likosky DS, Prager RL, et al. High socioeconomic deprivation and coronary artery bypass grafting outcomes: insights from Michigan. Ann Thorac Surg. 2022;113:1962–1970. doi: 10.1016/j.athoracsur.2021.07.021 [DOI] [PubMed] [Google Scholar]
- 26. Schultz WM, Kelli HM, Lisko JC, Varghese T, Shen J, Sandesara P, Quyyumi AA, Taylor HA, Gulati M, Harold JG, et al. Socioeconomic status and cardiovascular outcomes: challenges and interventions. Circulation. 2018;137:2166–2178. doi: 10.1161/CIRCULATIONAHA.117.029652 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Squiers JJ, Mack MJ. Coronary artery bypass grafting‐fifty years of quality initiatives since Favaloro. Ann Cardiothorac Surg. 2018;7:516–520. doi: 10.21037/acs.2018.05.13 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Loop FD, Lytle BW, Cosgrove DM, Stewart RW, Goormastic M, Williams GW, Golding LA, Gill CC, Taylor PC, Sheldon WC. Influence of the internal‐mammary‐artery graft on 10‐year survival and other cardiac events. N Engl J Med. 1986;314:1–6. doi: 10.1056/NEJM198601023140101 [DOI] [PubMed] [Google Scholar]
- 29. Rotar EP, Scott EJ, Hawkins RB, Mehaffey JH, Strobel RJ, Charles EJ, Quader MA, Joseph M, Teman NR, Yarboro LT, et al. Changes in controllable coronary artery bypass grafting practice for white and black Americans. Ann Thorac Surg. 2023;115:922–928. doi: 10.1016/j.athoracsur.2021.11.074 [DOI] [PubMed] [Google Scholar]
- 30. Victor RG, Lynch K, Li N, Blyler C, Muhammad E, Handler J, Brettler J, Rashid M, Hsu B, Foxx‐Drew D, et al. A cluster‐randomized trial of blood‐pressure reduction in black barbershops. N Engl J Med. 2018;378:1291–1301. doi: 10.1056/NEJMoa1717250 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Shi W, Green H, Sikhosana N, Fernandez R. Effectiveness of telehealth cardiac rehabilitation programs on health outcomes of patients with coronary heart diseases: an umbrella review. J Cardiopulm Rehabil Prev. 2024;44:15–25. doi: 10.1097/HCR.0000000000000807 [DOI] [PubMed] [Google Scholar]
- 32. Delgado JL. Beyond diversity—time for new models of health. N Engl J Med. 2022;386:503–505. doi: 10.1056/NEJMp2115149 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data S1