Social determinants of health affect unplanned readmissions following acute myocardial infarction

A total of 1,055,162 adult patients were admitted with AMI between 2007 and 2014 in the aforementioned four states and met inclusion criteria, as depicted in Figure 1.

The results of this study indicate that both patient- and hospital-level social determinants of health – Medicaid insurance, Black race, median income and HBH status – are independent risk factors for 30- and 90-day readmission following hospitalization for AMI. However, uninsured status was not associated with increased odds of readmission. These findings indicate that patient- and hospital-level factors contribute to increased readmission risk in the AMI patient population; our analysis did not reveal any statistically significant two level interaction terms.

The results of this study were partially consistent with the hypothesis that uninsured and Medicaid status, as proxies for low SES, would have higher rates and risk-adjusted odds of readmission compared with patients with private insurance. This hypothesis was based on prior literature demonstrating the importance of insurance status, among other socioeconomic factors, in predicting clinical outcomes for patients with cardiovascular disease [5,26]. In a retrospective analysis of more than 210,000 STEMI patients under age 65, Patel and colleagues reported lower rates of revascularization, and higher rates of in-patient mortality, among Medicaid patients compared with those with private insurance [27]. Furthermore, a retrospective analysis of 198,420 non-elderly AMI patients in California found that, although Medicaid patients were more likely than private payer patients to be treated at facilities capable of performing cardiac catheterization and other percutaneous coronary interventions, they were less likely to receive percutaneous coronary interventions then were private insurance patients [16]. Finally, a review of 10,461 MI admissions by Foraker and colleagues [28] found that Medicaid patients were significantly less likely to be discharged with optimal medical therapy, including aspirin and β-blockers, than were patients with private insurance.

Nonetheless, the results of the current did not fully confirm our hypothesis that uninsured or Medicaid status, as a marker of lower SES, would predict higher readmission odds. Uninsured/self-pay status was instead associated with decreased risk-adjusted odds of unplanned readmission. Several factors may help explain this finding. For instance, the self-pay cohort may include relatively wealthy patients, such as those who are too affluent for Medicaid but who cannot afford to purchase private insurance, or highly affluent patients who opt to pay out-of-pocket for care at institutions of their choice. Perhaps more plausibly, uninsured patients may be more likely to be indigent or otherwise lost to follow-up; this explanation is consistent with prior research demonstrating poor continuity of care for patients without insurance coverage [20,29].

Superior care metrics among uninsured patients, relative to Medicaid patients, are not unprecedented in the medical literature. In an analysis of 17,600 patients from the National Registry of Myocardial Infarction, Sada and colleagues [5] found that uninsured patients were more likely to undergo both discretionary (for ‘low risk’ patients) and non-discretionary (for ‘high risk’ patients) angiography than were Medicaid patients. In addition, in a recent retrospective analysis of postAMI outcomes, Wadhera and colleagues [30] reported that Medicaid expansion under the Affordable Care Act was associated with lower percentages of uninsured patients but not with improved clinical outcomes, suggesting that insurance status was a less meaningful driver of postAMI complications than hypothesized [30]. Furthermore, in an analysis of Massachusetts patients before and after insurance reform and expansion, Albert et al. [31] found that increased insurance coverage did not significantly reduce disparities in rates of coronary revascularization according to non-White ethnicity and neighborhood education level, among other socioeconomic factors.

Indeed, the results of this study support the significance of non-insurance-based factors in predicting readmissions following AMI; these include hospital-level factors, such as high-burden status, and patient-level socioeconomic factors, including relative income and race/ethnicity. In the current study, high-burden status was associated with a 9.5 and 9.6% increase in risk-adjusted odds of readmission at 30 and 90 days, independent of patient insurance status. These findings are consistent with prior analyses demonstrating inferior outcomes for cardiovascular patients at HBH [32], including a review of over 2 million cases from the National Inpatient Databases, in which HBH status was associated with ‘failure to rescue’ after elective coronary artery bypass graft (CABG) and/or cardiac valve operations (OR: 1.12; 95% CI: 1.01–1.23) [33]. Similarly, an analysis of 116,857 Medicare admissions for AMI found higher risk-standardized mortality rates among patients treated at urban SNHs compared with non-SNHs (21.3 ± 0.2 vs 20.0 ± 0.1; p < 0.001) in addition to inferior reperfusion and non-reperfusion quality of care metrics [23]. HBHs have been associated with increased mortality following ST-elevation MI [32] and following percutaneous coronary intervention for AMI and non-AMI indications [21]. Hospital burden status may contribute to inferior outcomes due to financial strain from decreased revenue and limited flexibility to determining strategic care priorities in the setting of a universal care mandate [34]. Financial problems have been shown to be associated with lapses in quality of care [35]. By contrast, financial profitability is associated with increased patient satisfaction, potentially suggestive of higher quality care [36].

Our study also demonstrated a small but significant association decreased patient income by zip code, as an indicator of lower SES, and increased risk-adjusted readmission odds. Markers of community-level SES are known to predict clinical outcomes in medical and surgical fields [37]. Residence within a severely disadvantaged neighborhood [38] and exposure to higher rates of income inequality [39] have been shown to predict increased readmission rates following hospitalization for MI, pneumonia or heart failure in the Medicare population. Similarly, patients from lower income quartiles have been shown to have higher rates of 30-day readmission following AMI [10].

Finally, the present study also found increased odds of readmission among patients of black race/ethnicity, as a marker of social disadvantage, relative to White patients. These findings are consistent with prior literature describing racial disparities in cardiovascular disease outcomes, including lower rates of revascularization for Black and non-Hispanic White patients with coronary artery disease [31], and higher rates of unplanned readmissions [12]. Clinical variables are often posited as explaining the association between non-White race/ethnicity and inferior clinical outcomes among MI patients [11]. An analysis of 6402 patients admitted with AMI found that Black patients had significantly different clinical baseline characteristics, including cardiac history, comorbidities and overall health status, relative to White patients, and that these traits were independently associated increased mortality risk at 5 years [40]. However, in the current study, Black and Hispanic race/ethnicity were independently associated with increased risk-adjusted odds of readmission, even after adjusting for present-on-admission comorbidities. Of note, patients self-reported as ‘other’ demonstrated lower risk-adjusted odds, perhaps suggesting decreased readmission risk associated with racial and ethnic groups included in this cohort (e.g., Asian–American patients), or among patients who declined to specify their racial or ethnic category.

Multiple mechanisms may contribute to the SES disparities described above and observed in the current study. Less affluent patients may tend to have a greater number and severity of presenting comorbidities and modifiable risk factors, including increased rates of smoking, drug and alcohol abuse [41,42]. Among patients of low SES, reduced healthcare literacy or the ability to navigate the healthcare system and actively participate in one’s care, may adversely affect outcomes; reduced healthcare literacy has been shown to be an independent risk factor for 30-day readmission following AMI [43]. Finally, for low-SES patients who are able to access the healthcare system, quality of care may be inferior [17] due to facility- or provider-related factors [15,34–36].

Our study is limited by its retrospective nature, which prevents us from determining causality of the readmissions disparities we observed. In addition, administrative data are inherently imperfect; coding errors and misclassification may have affected our results. Also, although we excluded patients coded as readmissions for AMI, we included patients whose admission status was ‘unknown’. It is therefore possible that our data includes unrecognized readmissions.

In addition, this study likely oversimplifies SES. While SES proxies such as income, insurance status and race/ethnicity are well established in the medical literature [6], the current study was unable to account for other variables, such as education level and financial insecurity, which may also affect the relationship between SES and healthcare outcomes [40]. Similarly, even patients with private payer insurance may qualify as ‘underinsured’, due to high deductibles and elevated out-of-pocket healthcare expenses relative to income [44]. The HCUP datasets preclude identification of ‘underinsured’ patients by these definitions.

Most significantly, the HCUP databases lack certain granular clinical information, such as the severity of patient comorbidities, which are known to affect readmission and mortality outcomes following major illness [1]. Relevant to MI, these databases do not capture procedural variables (e.g., procedure duration, estimated blood loss and number of vessels treated) for revascularized patients [26], hemodynamic or laboratory parameters (e.g., blood pressure on presentation and troponin levels), or standard-of-care discharge medications (e.g., beta-blockade, aspirin, etc.), among other potentially-significant data points [11,28].

Finally, our study included a heterogeneous patient population, including adult patients of all ages and AMIs of all subtypes and clinical severity. Exploratory analyses of patient subgroups by age and AMI subtype tended to confirm primary findings regarding readmission odds and risk factors, but also raised important questions that were beyond the scope of the present inquiry, For instance, exploratory analyses stratified by age suggested higher risk-adjusted readmission odds for Medicare and Medicaid patients <65 years, compared with older patients. Although prior research has identified a similar pattern among patients with congestive heart failure [45], and indicate that readmission is a major source of morbidity and healthcare spending for AMI patients across the age spectrum [10], these findings may warrant further study. Similarly, although exploratory analyses of NSTEMI and STEMI patient groups were reassuring regarding significant differences in readmission risk factors (Supplementary Table 2), a full analysis of readmission risk associated with AMI subtypes was beyond the scope of the current inquiry. This is an area ripe for further research.

Strengths of the current study include generalizability due to large sample size, and inclusion of patients from four geographically and demographically heterogeneous states over an 8-year time period. Readmission rates following AMI are known to vary widely across the nation [46], supporting the relevance and importance of a large and diverse sample, as included in the current study.

The findings of this study indicate that hospital high-burden safety-net status and patient Medicaid status, among multiple proxies of low SES, are associated with increased risk of readmission following AMI. More research is needed to help understand and ultimately mitigate the mechanisms that underlie these disparities. In addition, the current study raises important questions regarding lower readmissions odds among self-pay and uninsured patients, relative to those with Medicaid. If low admission rates are due to decreased follow-up and poor continuity of clinical care, further research may help inform interventions to reduce care disparities and ensure optimal outcomes for patients regardless of insurance status.

We anticipate that healthcare disparities, such as unplanned readmission following AMI, will be increasingly imported as matters of health policy and social justice in coming years. More research is needed to understand the mechanisms linking social determinants of health, such as insurance status and hospital safety-net status, to readmission rates and other metrics of inferior care. We anticipate that this research will become increasingly urgent in the coming decade due to ongoing implementation of Centers for Medicare and Medicaid Services Hospital Readmissions Reduction Program and due to increased awareness of socioeconomic, racial and health-related inequalities across the USA.

MM Hoyler, X Ma and RS White contributed to the conception and design of this project. X Ma and RS White conducted statistical analyses. MM Hoyler, MD Abramovitz, X Ma and RS White interpreted the data. MM Hoyler and MD Abramovitz drafted the manuscript. MM Hoyler, MD Abramovitz, D Khatib, JD Samuels, CW Tam, R Thalappillil and RS White critically revised the manuscript.

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.