Low socioeconomic status adversely influences outcomes after coronary artery bypass grafting

Abstract

Objectives:

Although socioeconomic status (SES) is believed to affect patient outcomes after coronary artery bypass grafting (CABG), readmission data are sparse. In a national cohort, we analyzed the influence of SES on readmission, resource utilization, and mortality after CABG.

Methods:

We queried the Nationwide Readmissions Database to identify patients who underwent isolated CABG from January 2016 through December 2018. We derived low, middle, and high SES from ICD-10-CM codes, patient demographics, and neighborhood-level factors. The effect of SES on risk-adjusted outcomes was assessed with multivariable analysis.

Results:

Of 523,042 patients who underwent CABG, the 134,039 (25.6%) with low SES were more likely than patients with middle (n=305,572; 58.4%) or high SES (n=83,431; 16%) to be female, younger, from rural areas, and admitted urgently. Patients with low SES were also less likely to be treated at teaching hospitals and had higher Elixhauser comorbidity scores (P<.001 for all). After risk adjustment, patients with low SES had 46% greater odds of in-hospital mortality at the index operation (odds ratio [OR] 1.464 [1.299–1.650]) than patients with high SES. Patients with low SES had the longest index hospital lengths of stay (P<.001). Low SES was associated with greater odds of readmission at 30 days (OR 1.229 [1.170–1.292]), 90 days (OR 1.281 [1.223–1.341]), and within a calendar year (hazard ratio 1.234 [1.193–1.278]) than high SES.

Conclusions:

Patients with low SES have greater adjusted odds of mortality and readmission after CABG than patients with high SES.

CENTRAL PICTURE LEGEND

Risk-adjusted readmission outcomes after CABG according to SES.

CENTRAL MESSAGE

Socioeconomic disparities exist in outcomes after CABG.

INTRODUCTION

Among patients who undergo coronary artery bypass grafting (CABG), readmission rates are 10% at 30 days and 20% at 90 days.^1–3 These readmissions are costly and associated with high rates of morbidity and mortality.⁴ Identifying populations at risk for readmission and implementing prevention measures may reduce costs and improve outcomes after CABG.

Socioeconomic status (SES) is believed to influence patient outcomes after CABG.^4–8 However, the association between an individual’s SES and readmission after CABG has not been studied using a nationally representative database. Additionally, studies often have used geographically derived measures or single-variable surrogates for SES.^5,7,9 These formulations provide only a partial picture, because both neighborhood- and individual-level factors contribute to existent socioeconomic disparities in cardiovascular outcomes.⁸

In this study, we used a composite, tiered SES metric comprising patient- and neighborhood-level factors that closely represent a patient’s individual SES¹⁰ to analyze the association between SES and outcomes after CABG in a national cohort. We hypothesized that low SES is associated with higher readmission and mortality rates.

PATIENTS AND METHODS

Data Collection and Patient Cohort

The Nationwide Readmissions Database (NRD) was developed by the Healthcare Cost and Utilization Project. Approximately 35 million patient discharges are weighted by using a complex survey design that allows national estimates of annual outcomes.¹¹ In this retrospective cohort study, the NRD was searched for adult patients who underwent isolated CABG in the United States between January 1, 2016, and December 31, 2018 by using the International Classification of Diseases, Tenth Revision, Clinical Modification/Procedure Coding System (ICD-10-CM/PCS, Table E1). Patients who underwent concomitant procedures were excluded (Figure 1). Because aggregated, deidentified admission-level data were used, this study was classified as exempt by Baylor College of Medicine’s Institutional Review Board.

FIGURE 1.

Flowchart showing patient identification, inclusion, and stratification. CABG, coronary artery bypass graft; SES, socioeconomic status.

Socioeconomic Status

As previously described, patients were stratified with a composite tiered SES metric comprising patient- and neighborhood-level factors.¹⁰ Median household income was reported as quartiles by the NRD according to patients’ zip codes.¹¹ Admission payors included Medicare, Medicaid, private payors, self-pay, and other, per NRD definitions.¹¹ SES-related ICD-10-CM codes identified social aspects contributing to SES including education, literacy, housing, economic circumstances, physical environment, and employment status (Table E1).^12,13 Low SES was defined as 1) having Medicaid or self-pay as primary payor, 2) being a Medicare beneficiary living in the lowest neighborhood income quartile, or 3) having an SES-related ICD-10-CM code and living in the lowest neighborhood income quartile. High SES was defined as living in the highest neighborhood income quartile and having private insurance or Medicare as the primary payor. Patients who did not qualify as low or high SES were categorized as middle SES.

Study Definitions

Hospital characteristics examined included teaching status, hospital bed size, and hospital urban-rural classification per NRD definitions.¹¹ Patient characteristics and comorbidities included age, sex, patient urban-rural classification, urgency of admission, Elixhauser comorbidity index (a composite score that categorizes patients’ comorbidities according to ICD-10-CM diagnosis codes¹⁴), and additional cardiovascular risk factors such as smoking and hyperlipidemia.¹⁴

Outcomes

The primary outcome was readmission at 30 days. Secondary outcomes included in-hospital mortality, index-hospitalization length of stay (LOS) and cost, discharge disposition, and readmission at 90 days and within a calendar year. Cost-to-charge ratios (ie, the ratio of the amount spent on the patient’s treatment to the amount the patient is billed for) provided by the NRD were used to estimate the cost of each hospital admission. Index-hospitalization cost was calculated by multiplying hospital charges for each admission by these cost-to-charge ratios.¹⁵

We further characterized readmissions by mortality, LOS, cost, and reason for readmission. Because the NRD uses unique patient identifiers limited to one calendar year, 30-day readmission analysis excluded December admissions, and 90-day readmission analysis excluded October, November, and December admissions.³ Readmissions within a calendar year were determined using Kaplan-Meier analysis, with index readmission as the event being measured.¹⁶ Reasons for readmission were derived from primary readmission diagnosis codes as previously described, independently reviewed, and grouped into clinical categories. The seven most common categories are reported and include cardiovascular complications (eg, conduction disorders, heart failure, ischemic heart disease), infectious complications (eg, pneumonia, sepsis), neurologic complications (eg, stroke, TIA), pulmonary complications (eg, pneumothorax, acute noninfectious respiratory failure), and postoperative complications (eg, bleeding, pleural/pericardial effusion).

Statistical Analyses

Statistical calculations were performed in R version 4.2 (R Foundation for Statistical Computing). Figures were created in GraphPad Prism version 9.3.1 (GraphPad Software Inc). The NRD’s complex survey design was adjusted for in all calculations and statistical tests using the R package survey.¹⁷ Fewer than 1% of data were missing from any category; missing data were imputed using a multivariate imputation by chained equations algorithm. Continuous data are represented as median (interquartile range), and differences were determined with a complex survey-adjusted Kruskal-Wallis rank-sum test. Categorical data are represented as percentage (number) and were analyzed by chi-square test with the Rao and Scott adjustment. A survey-adjusted log-rank test was used to examine differences in readmission rates within a calendar year.

Regression results are represented as odds ratio (OR) or hazard ratio (both with 95% confidence intervals) for logistic or Cox proportional hazard, respectively. P values were derived from a survey-adjusted Wald test. Proportionality of the covariates was examined using Schoenfeld residuals. Gamma regression with a logarithmic link function was used to estimate the effects of covariates on admission cost. Independent variables included patient and hospital characteristics that significantly differed among groups on univariate analysis, as well as comorbidities with >5% prevalence. Variance inflation factor was used to identify and remove colinear variables. High SES was used as the reference group for regression results.

RESULTS

Patient and Hospital Characteristics by SES

Between January 1, 2016, and December 31, 2018, 523,042 US adult patients underwent isolated CABG. These patients were characterized as having low SES (25.6%; n=134,039), middle SES (58.4%; n=305,572), or high SES (16%; n=83,431) (Figure 1). Patients with low SES were more likely to be younger, female, from rural areas, and admitted urgently, and less likely to be treated at teaching hospitals (P<.001 for all, Table 1).

TABLE 1.

Patient and hospital characteristics by socioeconomic status

Variable	Low SES (n=134,039)	Middle SES (n=305,572)	High SES (n=83,431)	P value*
Age, years	66 [58–73]	66 [59–73]	68 [61–74]	<.001
Female sex	29.8% (39,962)	23.3% (71,177)	19.6% (16,385)	<.001
Rural patient	16.5% (22,183)	9.7% (29,699)	0.4% (339)	<.001
Urgent intervention	59.1% (79,164)	51.6% (157,772)	50.6% (42,200)	<.001
Teaching hospital	76.7% (102,742)	78.7% (240,553)	87.6% (73,098)	<.001
Urban hospital	95.3% (127,700)	94.6% (289,001)	99.9% (83,389)	<.001
Hospital bed size				<.001
Large	68.4% (91,646)	65.9% (201,471)	63.6% (53,084)
Medium	22.8% (30,567)	23.5% (71,859)	28.6% (23,898)
Small	8.8% (11,826)	10.6% (32,242)	7.7% (6448)
Neighborhood income quartile†				N/A
1 (Lowest)	75.8% (100,888)	14.7% (43,851)	<1.1% (<11)
2	11.5% (15,241)	44.8% (133,829)	<1.1% (<11)‡
3	8.4% (11,119)	39.8% (119,015)	<1.1% (<11)‡
4 (Highest)	4.4% (5842)	0.7% (1965)	100.0% (83,431)
Primary payor				N/A
Medicaid	28.7% (38,434)	<1.1% (<11)‡	<1.1% (<11)‡
Medicare	61.6% (82,508)	53.2% (162,491)	60.2% (50,248)
Private insurance	0.3% (364)	41.1% (125,538)	39.8% (33,183)
Self-pay	9.3% (12,519)	<1.1% (<11)c	<1.1% (<11)‡
Other	0.1% (214)	5.7% (17,543)	<1.1% (<11)‡
Elixhauser score	8 [0–19]	7 [−1–17]	8 [0–17]	<.001
Congestive heart failure	40.1% (53,775)	33.3% (101,741)	31.7% (26,468)	<.001
Arrhythmia	44.8% (60,051)	46.5% (142,068)	50.4% (42,022)	<.001
Valve disease	16.5% (22,104)	15.6% (47,778)	17.5% (14,604)	<.001
Peripheral artery disease	16.5% (22,147)	14.4% (44,044)	14.7% (12,280)	<.001
Hypertension	88.9% (119,203)	87.7% (268,069)	87.3% (72,841)	<.001
Chronic pulmonary disease	28.4% (38,013)	21.7% (66,409)	17.3% (14,427)	<.001
Diabetes mellitus	52.4% (70,260)	47.8% (146,073)	44.2% (36,909)	<.001
Renal failure	22.8% (30,520)	20.4% (62,326)	21.0% (17,534)	<.001
Obesity	29.2% (39,121)	30.4% (92,743)	26.2% (21,831)	<.001
Pulmonary circulation disorder	6.2% (8353)	5.1% (15,610)	5.2% (4353)	<.001
Hypothyroidism	10.6% (14,153)	11.2% (34,162)	11.3% (9465)	<.001
Alcohol abuse	4.9% (6574)	3.4% (10,473)	3.1% (2622)	<.001
Drug abuse	5.0% (6703)	2.0% (5976)	1.5% (1259)	<.001
Depression	10.8% (14,532)	9.5% (29,069)	8.8% (7360)	<.001
Hyperlipidemia	77.4% (103,700)	81.1% (247,825)	84.7% (70,696)	<.001
Smoking history	56.7% (75,944)	51.3% (156,715)	46.1% (38,431)	<.001

Values presented as median (interquartile range) for continuous variables and percentage (n) for categorical variables. N/A, not applicable. SES, socioeconomic status.

^*SES levels compared with Kruskal-Wallis rank-sum test for complex survey samples or chi-square test with Rao and Scott second-order correction.

^†Based on patient zip code.

^‡Observations with cell count <11 reported as <11, per Healthcare Cost and Utilization Project regulations.

Comorbidity rates differed by SES level (Table 1). Patients with low SES had a greater prevalence of congestive heart failure, hypertension, diabetes mellitus, and smoking history. Patients with high SES more often had hyperlipidemia and arrhythmias.

Patient Outcomes and Resource Utilization by SES

In-hospital mortality associated with the index operation was inversely correlated with SES level, being 1.4% in patients with high SES and 2.3% in patients with low SES (P<.001, Table 2). After risk adjustment for comorbidities and hospital characteristics (Table E2), patients with low SES had 46% greater odds of in-hospital mortality (OR 1.464 [1.299–1.650], Figure 2) than patients with high SES.

TABLE 2.

Patient outcomes and resource utilization by socioeconomic status

Variable	Low SES (n=130,981)	Middle SES (n=300,050)	High SES (n=82,273)	P value*
Index hospitalization
Mortality	2.3% (3058/134,039)	1.8% (5522/305,572)	1.4% (1158/83,431)	<.001
Length of stay, d	9 [6–13]	8 [6–11]	7 [5–11]	<.001
Cost, USD	40,838 [31,046–55,777]	38,797 [29,809–52,256]	42,654 [32,015–59,386]	<.001
Disposition				<.001
Home healthcare	38.8% (50,785)	41.4% (124,058)	50.5% (41,538)
Routine	40.5% (52,988)	40.8% (122,511)	32.2% (26,525)
SNF or ICF	20.0% (26,192)	17.2% (51,650)	16.6% (13,690)
Other	0.8% (1017)	0.6% (1803)	0.6% (518)
Readmissions
30-day	13.7% (16,422/120,031)	10.8% (29,773/274,621)	10.1% (7576/74,879)	<.001
90-day	21.4% (21,057/98,466)	17.0% (38,290/225,221)	15.7% (9635/61,291)	.03
Length of stay, d	4 [2–6]	3 [2–6]	3 [2–6]	<.001
Cost, USD	8462 [4917–16,238]	8993 [5175–16,894]	10,213 [5834–19,100]	<.001
Mortality	2.3% (748/32,525)	2.2% (1335/59,667)	1.8% (271/15,136)	.04

Values presented as median (interquartile range) for continuous variables and as percentage (n) for categorical variables.

Socioeconomic status (SES) levels compared with the Kruskal-Wallis rank-sum test for complex survey samples or chi-square test with Rao and Scott second-order correction. SNF, skilled nursing facility; ICF, intermediate care facility.

FIGURE 2.

Characteristics associated with in-hospital mortality. ^aIndexed to high socioeconomic status (SES). Full regression model provided in Table E2.

Patients with low SES had a lower index-admission cost than patients with high SES (P<.001), despite having a longer index-hospitalization LOS (P<.001, Table 2). After risk adjustment for LOS, urgency of admission, teaching hospital status, and other variables, low SES was significantly associated with lower index-admission cost (exp(beta) 0.886 [0.881–0.891], Table E3). Additionally, patients with low SES were less likely to receive home healthcare (P<.001, Table 2).

Patients with low SES had higher readmission rates at 30 days (P<.001), 90 days (P=.03) (Table 2), and within a calendar year (P<.001, Figure 3). After risk adjustment, patients with low SES had greater odds of readmission than patients with high SES by 23% at 30 days (OR 1.229 [1.170–1.292], Figure 4, Table E4), 28% at 90 days (OR 1.281 [1.223–1.341], Table E5), and 23% within the calendar year (hazard ratio 1.234 [1.193–1.278], Table E6). Patients with low SES had longer readmission LOS (P<.001) and a higher mortality rate on readmission (P=.04) but lower readmission costs (P<.001) (Table 2). For all SES tiers, the most common reasons for readmission were cardiovascular, infectious, and postoperative complications (Figure 5).

FIGURE 3.

Freedom from calendar-year readmission. SES, socioeconomic status.

FIGURE 4.

Characteristics associated with readmission at 30 days (left), 90 days (middle), and within the calendar year (right). ^aIndexed to high socioeconomic status (SES). Full regression models provided in Tables E4-E6.

FIGURE 5.

Reasons for readmission after isolated coronary artery bypass grafting by socioeconomic status (SES). postop, postoperative.

Subgroup analyses of scheduled and urgent admissions were performed separately. Outcomes remained qualitatively unchanged across these analyses (Tables E7-E8).

DISCUSSION

In a retrospective analysis of a national cohort, we found that after undergoing isolated CABG, patients with low SES had greater risk-adjusted odds of readmission at 30 days, 90 days, and within a calendar year than patients with high SES (see Figure 6 for a graphical abstract of the study). Patients with low SES also had 46% greater risk-adjusted odds of in-hospital mortality. The higher readmission rates in patients with low SES supported our hypothesis. Surprisingly, low SES was associated with less resource utilization during index and readmission hospitalizations, despite longer stays.

FIGURE 6.

Graphical abstract.

Although SES contributes to cardiovascular disease and treatment outcomes, socioeconomic inequities remain difficult to report because SES lacks a uniform definition and is difficult to quantify.^8,18 SES is widely considered a measure of education, employment, and income; thus, single-variable proxies for SES have limited utility.^19,20 Studies have used geographically derived composite measures of SES such as the area deprivation index, distressed community index (DCI), and similar international indices.^5,9,21,22 Geographically derived SES measures effectively capture the socioeconomic environment of a geographic unit and highlight potential societal or community disparities that affect health outcomes; however, these measures may not align with an individual’s SES level.^23–25 The composite, tiered SES metric in this study uses ICD-10-CM coding for social determinants of health, combined with geographic and insurance data.^10,12,13 The Centers for Disease Control has recognized the value of these additional ICD-10 codes for enhancing understanding of social determinants of health (https://www.cdc.gov/nchs/data/icd/social-determinants-of-health.pdf), and guides for determining them are available (https://www.cms.gov/files/document/cms-2023-omh-z-code-resource.pdf; https://www.cms.gov/files/document/zcodes-infographic.pdf). These codes identify patients who experience barriers to health care. Our composite metric, which includes patient-level factors, may better represent individual-level SES than older metrics that rely primarily on geographic data. Using a more comprehensive metric could allow targeted risk assessment, empowering surgeons to address and mitigate inequities among patients who undergo CABG.

Several other reports support our finding of an inverse relationship between SES level and mortality, which is sustained across different time periods and with little variability.^5–7,22,26 Recently, Mehaffey and colleagues⁹ stratified CABG outcomes in a national cohort by using the DCI, modeling it as a continuous variable and finding that, for each 10-unit increase in DCI, the relative risk of operative mortality increased by 3%. In addition, we found a higher prevalence of urgent surgery in patients with low SES, reflecting greater acuity at presentation, which has been reported in other surgical populations.^7,27

National data surrounding readmission, resource utilization, and SES are limited. Here, we found that SES predicted readmission at all analyzed time intervals. The most common readmission reasons for all SES tiers were cardiovascular, infectious, and postoperative complications. Similarly, in a study from Denmark, Butt and colleagues¹⁹ found that being of the lowest income quartile (their proxy for low SES) predicts 1-year readmission after CABG. Fliegner and colleagues,²¹ studying Medicare beneficiaries from Michigan, found a higher risk of 90-day readmission in patients with low SES, as measured by the Area Deprivation Index. These authors also linked low SES to greater index-hospitalization cost, total 90-day cost, and LOS.

Surprisingly, we found that patients with lower SES had lower index-hospitalization and readmission costs despite longer hospital stays. This seemingly paradoxical finding is probably multifactorial. Our data suggest that for patients with low SES, higher mortality rates at index hospitalization and readmission contribute to less resource utilization per admission. An additional contributing factor may be longer intensive care unit stays in patients with high SES. Additionally, barriers to the placement of patients with low SES who are not suitable for routine discharge may prolong hospital stays without substantially increasing cost. Additionally, as proposed for other cardiovascular patients, patients with low SES who undergo CABG experience inequities in care stemming from intrinsic biases and differences in practice based on sociodemographic characteristics; these inequities may also lead to less resource utilization during index hospitalization.⁸ Studies in which various single-variable proxies are used for SES have associated these variables with readmission.^28–30 Data review has shown that academic centers and safety-net hospitals caring for populations with low SES are most likely to be penalized for readmission metrics.³¹ Thus, investing more resources up front to prevent readmission among patients with low SES may reduce overall expenditures.

Although several strategies have been shown to reduce readmissions, no single strategy is likely to be effective in all patients with low SES (Figure 7). A multidimensional approach that begins preoperatively may include improving health literacy, patient-provider communication, postdischarge support such as home healthcare and cardiac rehab, and SES-tailored discharge protocols.³² We found that patients with low SES less frequently received home healthcare. Access to home healthcare may increase medical compliance, which is reportedly lower in patients with low SES.³⁰ Avoidable complications may be prevented through improved compliance with postoperative medication, sternal and wound precautions, and follow-up. Additionally, professionals may identify developing complications and ensure timely intervention without readmission. Compared with routine discharge, discharge followed by home health visits reduces readmission rates, although resource utilization is greater initially.³³ Our findings are hypothesis generating, and the potential value of home health care in reducing readmission after CABG warrants further investigation.

FIGURE 7.

Proposed solutions to mitigate socioeconomic disparities among patients who undergo coronary artery bypass grafting (CABG). ICU, intensive care unit; SES, socioeconomic status.

Pursuing equity in health care is a multidisciplinary effort. Early involvement of case managers and social workers may help overcome socioeconomic barriers to care. Customized postoperative plans for patients with reduced access to care may also mitigate complications.

Improving health literacy through SES-tailored patient education provides an opportunity to recognize complications earlier.³⁴ The best-fit teaching method should be selected for each patient and adjusted to education and personal needs.³⁵ In a disparate population, Chudgar and colleagues³⁶ implemented discharge care protocols that encompassed preoperative health literacy, inpatient disease-management education, and rigorous follow-up after discharge, thereby reducing readmissions. Encouragingly, implementing these protocols did not require additional resources but only shifting of resource allocation.³⁶ Maniar and colleagues³⁰ associated patient education and prior cardiologist establishment with less 30-day readmission. At our regional safety-net hospital, case management consults are initiated immediately after surgery for all CABG patients. Additionally, follow-up with county health providers is arranged within 1 week after discharge. In some cases, hospital stays are prolonged to ensure optimal social disposition, balancing the benefits against the risks of extended hospitalization.

This study had limitations inherent to retrospective analyses of large administrative databases. Limitations of the NRD include dependence on ICD-10–based derivations of comorbidities. Furthermore, the NRD lacks granular admission data such as ejection fraction, intraoperative details, and intensive care unit LOS. The NRD does not capture every out-of-hospital death, so we could not perform an accurate competing risk analysis. The NRD does not currently include race and ethnicity data. From the data available through the NRD, we are unable to fully examine the reasons for the observed lower odds of mortality in patients with obesity or diabetes, despite these comorbidities being associated with adverse cardiac surgical outcomes. This association may result from preoperative risk stratification and risk-modification strategies, or from selection bias, a limitation commonly encountered in retrospective studies. Nonetheless, the NRD generates standardized, robust calculations reflecting national-level in-hospital outcomes and readmissions. Additionally, while we believe the tiered SES metric used in this study improves on common single-variable and geographic SES measures, an ideal approach would assess SES as a continuous variable at the individual level to enable the most precise targeted risk assessment. Using the data available in the NRD, we could only treat SES as a categorical variable; this could impair the applicability of our findings, because our SES metric may not fully capture the complexity of SES-related disparities.

CONCLUSIONS

In a national cohort, we identified SES-related disparities in patients’ outcomes after CABG. Discovering the primary drivers of greater readmission in patients with low SES may lead to the development of targeted interventions to reduce health care expenditure and improve outcomes in this vulnerable population.

PERSPECTIVE STATEMENT.

While reports have indicated that post-CABG in-hospital outcomes differ by patients’ socioeconomic status (SES), the association of SES with readmission remains underreported. SES-related differences in readmission rates suggest that health inequities exist among patients who undergo CABG. Further study is needed to find ways to mitigate these differences.

ABBREVIATIONS AND ACRONYMS

CABG

coronary artery bypass grafting

DCI

distressed community index

ICD-10-CM

International Classification of Diseases, Tenth Revision, Clinical Modification

LOS

length of stay

NRD

Nationwide Readmissions Database

OR

odds ratio

SES

socioeconomic status