Disparities in Care and Mortality Among Homeless Adults Hospitalized for Cardiovascular Conditions

Key Points

Question  Do patterns of care and mortality differ between homeless and nonhomeless adults hospitalized for acute cardiovascular conditions?

Findings  In this cross-sectional study of 24 890 hospitalizations for cardiovascular conditions among homeless adults and 1 827 900 hospitalizations for cardiovascular conditions among nonhomeless adults in New York, Massachusetts, and Florida between 2010 and 2015, homeless individuals hospitalized with acute myocardial infarction were significantly less likely to undergo coronary angiography, percutaneous coronary intervention, and coronary artery bypass graft surgery compared with nonhomeless adults and had higher mortality rates for ST-elevation myocardial infarction. Homeless persons hospitalized for cardiac arrest or stroke also received less procedural care and had higher mortality rates.

Meaning  There is a need for public health and policy efforts to support hospitals that care for homeless persons, to reduce disparities in hospital-based care, and to improve health outcomes for this population.

Importance  Cardiovascular disease is a major cause of death among homeless adults, with mortality rates that are substantially higher than in the general population. It is unknown whether differences in hospitalization-related care contribute to these disparities in cardiovascular outcomes.

Objective  To evaluate differences in intensity of care and mortality between homeless and nonhomeless individuals hospitalized for cardiovascular conditions (ie, acute myocardial infarction, stroke, cardiac arrest, or heart failure).

Design, Setting, and Participants  This retrospective cross-sectional study included all hospitalizations for cardiovascular conditions among homeless adults (n = 24 890) and nonhomeless adults (n = 1 827 900) 18 years or older in New York, Massachusetts, and Florida from January 1, 2010, to September 30, 2015. Statistical analysis was performed from February 6 to July 16, 2019.

Main Outcomes and Measures  Risk-standardized diagnostic and therapeutic procedure rates and in-hospital mortality rates.

Results  Of the 1 852 790 total hospitalizations for cardiovascular conditions across 525 hospitals, 24 890 occurred among patients who were homeless (11 452 women and 13 438 men; mean [SD] age, 65.1 [14.8] years) and 1 827 900 occurred among patients who were not homeless (850 660 women and 977 240 men; mean [SD] age, 72.1 [14.6] years). Most hospitalizations among homeless individuals were primarily concentrated among 11 hospitals. Homeless adults were more likely than nonhomeless adults to be black (38.6% vs 15.6%) and insured by Medicaid (49.3% vs 8.5%). After accounting for differences in demographics (age, sex, and race/ethnicity), insurance payer, and clinical comorbidities, homeless adults hospitalized for acute myocardial infarction were less likely to undergo coronary angiography compared with nonhomeless adults (39.5% vs 70.9%; P < .001), percutaneous coronary intervention (24.8% vs 47.4%; P < .001), and coronary artery bypass graft (2.5% vs 7.0%; P < .001). Among adults hospitalized with stroke, those who were homeless were less likely than nonhomeless individuals to undergo cerebral angiography (2.9% vs 9.5%; P < .001) but were as likely to receive thrombolytic therapy (4.8% vs 5.2%; P = .28). In the cardiac arrest cohort, homeless adults were less likely than nonhomeless adults to undergo coronary angiography (10.1% vs 17.6%; P < .001) and percutaneous coronary intervention (0.0% vs 4.7%; P < .001). Risk-standardized mortality was higher for homeless persons with ST-elevation myocardial infarction compared with nonhomeless persons (8.3% vs 6.2%; P = .04). Mortality rates were also higher for homeless persons than for nonhomeless persons hospitalized with stroke (8.9% vs 6.3%; P < .001) or cardiac arrest (76.1% vs 57.4%; P < .001) but did not differ for heart failure (1.6% vs 1.6%; P = .83).

Conclusions and Relevance  There are significant disparities in in-hospital care and mortality between homeless and nonhomeless adults with cardiovascular conditions. There is a need for public health and policy efforts to support hospitals that care for homeless persons to reduce disparities in hospital-based care and improve health outcomes for this population.

In the United States, more than 560 000 people are homeless on any given night,¹ and 1 in 25 will experience homelessness in their lifetime.² Homeless individuals have a higher burden of medical comorbidities and also face numerous social challenges, including poor access to food, shelter, and safety. These complex medical and social issues are associated with low life expectancy and high mortality rates among people experiencing homelessness.^3-5

Cardiovascular disease is a major cause of death among homeless individuals. In addition to a high prevalence of traditional cardiovascular risk factors, such as diabetes, the burden of nontraditional cardiovascular risk factors, such as depression and substance use disorder, is greater in this population.⁶ Furthermore, financial and structural barriers impede homeless persons’ access to health care services, resulting in delayed and fragmented care, and suboptimal treatment of these risk factors.^6-8 Estimates suggest that cardiovascular mortality rates among homeless adults are up to 3-fold higher than those of nonhomeless individuals.^3,9

Despite high cardiovascular mortality rates among homeless persons,⁶ little is known about whether these broader patterns reflect differences in hospitalization-related care for cardiovascular conditions, such as acute myocardial infarction (AMI), stroke, cardiac arrest, and heart failure. One possibility is that homeless and nonhomeless persons receive similar care during a hospitalization and that worse outcomes in the homeless population are related primarily to access to outpatient care and to factors beyond hospital walls. Alternatively, homeless individuals may also be less likely to receive indicated care during a hospitalization, owing to either within-hospital or between-hospital differences in how they are treated compared with nonhomeless adults, which may be associated with their worse outcomes. Understanding whether homeless adults experience disparities in hospitalization-related care is critically important from a public health perspective, particularly given recent evidence that the homeless population is growing in size and that their rates of acute hospitalization are increasing.^10,11 Such insights could inform targeted interventions to improve health outcomes in this population.

Therefore, in this study, we aimed to answer 3 questions. First, how did demographic and clinical characteristics differ between homeless and nonhomeless adults hospitalized for acute cardiovascular conditions (AMI, cardiac arrest, stroke, or heart failure) in 3 US states from 2010 to 2015? Second, were homeless adults less likely to receive diagnostic and therapeutic procedures for acute cardiovascular conditions compared with nonhomeless adults, after accounting for differences in demographics and clinical comorbidities? If so, was this disparity due to differences in how homeless and nonhomeless adults were treated within the same hospital or, instead, to between-hospital differences in care delivery? And third, did in-hospital mortality differ between homeless and nonhomeless persons?

We used the State Inpatient Databases of the Healthcare Cost and Utilization Project,^12,13 which are created by the Agency for Healthcare Research and Quality. Each State Inpatient Database includes all inpatient discharges from nonfederal acute-care hospitals and provides information on patient demographics, clinical comorbidities, diagnoses, and expected payer. Data from Massachusetts, New York, and Florida were used because these are populous states with large homeless populations—more than one-quarter of homeless individuals in the United States reside in these 3 states.¹⁴ All inpatient discharges from January 1, 2010, through September 30, 2015, were included in the analysis. We included up to the third quarter of 2015 in our analysis because this was the last year for which data were available. Because the analysis used publicly available deidentified data, institutional review board review at Beth Israel Deaconess Medical Center was waived.

We included all hospitalizations for all adults aged 18 years or older with a principal discharge diagnosis of AMI, stroke, cardiac arrest, or heart failure. Primary discharge diagnoses were identified using International Classification of Diseases, Ninth Revision, Clinical Modification codes, which are shown in eTable 1 in the Supplement.^15,16 We identified hospitalizations among homeless individuals based on whether the patient was documented to be homeless at time of death or discharge.¹⁷ Hospitals are responsible for determining whether a patient is homeless, and this information is then coded and reported directly to the state and retained in the Healthcare Cost and Utilization Project State Inpatient Databases. This approach is used by the Agency for Healthcare Research and Quality and has also been used in prior studies.^11,18-20 All other hospitalized adults were classified as nonhomeless.

The primary outcome was in-hospital mortality after hospitalization for an acute cardiovascular condition (AMI, stroke, cardiac arrest, or heart failure). Mortality by type of AMI (non–ST-elevation myocardial infarction [non-STEMI] or ST-elevation myocardial infarction [STEMI]) and by type of stroke (ischemic vs hemorrhagic) was characterized. We also evaluated rates of diagnostic and therapeutic procedures for each respective condition, including coronary angiography, percutaneous coronary intervention (PCI), and coronary artery bypass graft surgery (CABG) for AMI, coronary angiography and PCI for cardiac arrest, cerebral angiography for stroke, and thrombolytic therapy for ischemic stroke.

Statistical analysis was performed from February 6 to July 16, 2019. Continuous variables were summarized by means and SDs and categorical variables as percentages. We first compared patient demographics (age, sex, and race/ethnicity), insurance payer, and clinical comorbidities for the homeless and nonhomeless cohorts using t tests and χ² tests.

We then categorized hospitalizations among homeless individuals and hospitalizations among nonhomeless individuals into strata based on the following demographic and clinical information: age (5-year intervals), sex (female vs male), race/ethnicity (white, black, Hispanic, or other), insurance type (Medicare, Medicaid, private, or other), and clinical comorbidities.²¹ This categorization resulted in 10 260 strata across 4 conditions, each of which contained hospitalizations among homeless and nonhomeless individuals that were defined by a unique combination of these demographic and clinical characteristics. The relative size of each stratum for the homeless cohort (eg, proportion of all homeless hospitalizations included within each stratum) was used to reweight each corresponding nonhomeless stratum. Strata in which there were either no homeless individuals or nonhomeless individuals were removed.¹¹

Next, we aggregated strata to generate an overall risk-standardized rate of each procedure for homeless and nonhomeless groups. We then calculated the difference in risk-standardized diagnostic and therapeutic procedure rates between hospitalizations among homeless individuals and hospitalizations among nonhomeless individuals. This calculation was performed separately for each acute cardiovascular condition and then repeated for subtypes of each condition (eg, non-STEMI vs STEMI). We used a similar approach to calculate differences in risk-standardized in-hospital mortality rates between homeless and nonhomeless cohorts for each condition. Overall, this approach allowed us to address potential confounding associated with the demographic and clinical variables used for stratification, to isolate the effect of homelessness, and to compare risk-standardized outcomes between homeless and nonhomeless cohorts.

We also compared procedure rates between homeless and nonhomeless patients admitted to the same hospital, using the same stratification approach for AMI and stroke. We were unable to perform this analysis for cardiac arrest owing to the low number of admissions with cardiac arrest per hospital. This comparison allowed us to understand whether there were within-hospital differences in clinical management between homeless and nonhomeless patients.

We performed 2 additional analyses. First, for the primary outcome of condition-specific in-hospital mortality, we determined E-values.²² Second, we excluded all hospitals with fewer than 50 hospitalizations of homeless individuals and then repeated our main analysis. Race and/or ethnicity was assessed in the study given the association between these patient characteristics and cardiovascular outcomes. Analyses were performed using SAS, version 9.4 64-bit (SAS Institute Inc).

We identified 1 852 790 total hospitalizations for cardiovascular conditions across the study period, of which 24 890 (1.3%) were among homeless individuals (Table). In the homeless group, 5360 hospitalizations were for AMI, 5875 for stroke, 244 for cardiac arrest, and 13 411 for heart failure. Most hospitalizations for cardiovascular conditions among homeless individuals occurred at 11 of the 525 hospitals in the study cohort (Figure 1A), while hospitalizations for cardiovascular conditions among nonhomeless individuals were distributed across all hospitals (Figure 1B).

Compared with nonhomeless adults, homeless adults were younger (mean [SD] age, 65.1 [14.8] vs 72.1 [14.6] years), more likely to be black (38.6% vs 15.6%) and insured by Medicaid (49.3% vs 8.5%), and less likely to be insured by Medicare (44.7% vs 71.4%) (Table). Comorbidities differed among homeless and nonhomeless cohorts. Homeless adults had higher rates than nonhomeless adults of alcohol use disorder (7.5% vs 3.4%), drug use disorder (8.4% vs 2.3%), and psychoses (4.4% vs 2.8%). Homeless adults were also more likely than nonhomeless adults to have diabetes (39.0% vs 32.5%) and hypertension (75.9% vs 74.8%) but were otherwise generally less likely to have other comorbidities. The mean (SD) length of stay was longer for the homeless cohort than the nonhomeless cohort (6.9 [11.1] days vs 5.7 [7.3] days), and homeless individuals were less likely to be discharged to a skilled nursing or intermediate care facility (16.7% vs 24.0%) (Table; eTable 2 in the Supplement).

Observed procedure rates for each cardiovascular condition are shown in eTable 3 in the Supplement. After risk standardization (for age, sex, race/ethnicity, insurance payer, and clinical comorbidities), homeless adults who were hospitalized for AMI were less likely to undergo coronary angiography compared with nonhomeless adults (39.5% vs 70.9%; difference in procedure rates between homeless and nonhomeless groups, −31.4% [95% CI for difference, −33.7% to −29.1%]; P < .001), PCI (24.8% vs 47.4%; difference, −22.6% [95% CI, −23.8% to −21.4%]; P < .001), and CABG (2.5% vs 7.0%; difference, −4.5% [95% CI, −5.2% to −3.8%]; P < .001) (Figure 2A). In the subgroup of patients with STEMI, risk-standardized rates of coronary angiography were lower in the homeless group compared with the nonhomeless group (55.8% vs 85.9%; difference, −30.2% [95% CI, −34.0% to −26.3%]; P < .001), as were rates of PCI (55.4% vs 75.9%; difference, −20.5% [95% CI, −23.5% to −17.4%]; P < .001) and CABG (1.7% vs 4.8%; difference, −3.0% [95% CI, −4.1% to −2.0%]). Similarly, among patients with non-STEMI, rates of coronary angiography (35.7% vs 66.1%; difference, −30.4% [95% CI, −32.4% to −28.4%]; P < .001), PCI (17.6% vs 37.6%; difference, −20.0% [95% CI, −21.1% to −18.8%]; P < .001), and CABG (2.5% vs 7.7%; difference, −5.1% [95% CI, −5.9% to −4.3%]; P < .001) were lower in the homeless group vs the nonhomeless group (Figure 3).

Among individuals hospitalized with stroke, those who were homeless were less likely than those who were nonhomeless to undergo cerebral angiography (2.9% vs 9.5%; difference, −6.6% [95% CI, −7.5% to −5.7%]; P < .001) but as likely to receive thrombolytic therapy (4.8% vs 5.2%; difference, −0.4% [95% CI, −1.0% to 0.3%]; P = .28) (Figure 2B). These patterns were similar among the subgroup of adults with ischemic stroke (cerebral angiography, 2.1% vs 6.7%; difference, −4.7% [95% CI, −5.3% to −4.0%]; P < .001; thrombolytic therapy, 5.9% vs 6.1%; difference, −0.2% [95% CI, −1.1% to 0.7%]; P = .69). Homeless adults hospitalized with hemorrhagic stroke were also substantially less likely than nonhomeless adults to undergo cerebral angiography (5.1% vs 18.1%; difference, −13.0% [95% CI, −15.5% to −10.5%]; P < .001).

In the cardiac arrest cohort, homeless adults were less likely than nonhomeless adults to undergo coronary angiography (10.1% vs 17.6%; difference, −7.5% [95% CI, −11.3% to −3.6%]; P < .001). Homeless adults were also less likely to undergo PCI (0.0% vs 4.7%; difference, −4.7% [95% CI, −6.1% to −3.3%]; P < .001) (Figure 2C).

Observed in-hospital mortality rates are shown in eTable 4 in the Supplement. Risk-standardized in-hospital mortality did not significantly differ between homeless (3.5%) and nonhomeless (3.7%) adults hospitalized for AMI (difference, −0.2% [95% CI, −0.8% to 0.4%]; P = .56) (Figure 4A). For subtypes of AMI, however, mortality was higher among homeless patients hospitalized for STEMI (8.3% vs 6.2%; difference, 2.1% [95% CI, 0.04%-4.2%]; P = .04) but no different for those hospitalized for non-STEMI.

Among all hospitalizations for stroke, homeless adults had higher risk-standardized in-hospital mortality rates than their nonhomeless counterparts (8.9% vs 6.3%; difference, 2.5% [95% CI, 1.4%-3.6%]; P < .001). This pattern was also observed among the subgroup of patients with ischemic stroke (4.6% vs 3.0%; difference, 1.6% [95% CI, 0.5%-2.7%]; P = .003), although mortality was similar between homeless and nonhomeless adults with hemorrhagic stroke (Figure 4B).

For cardiac arrest, in-hospital mortality rates were higher among homeless adults (76.1% vs 57.4%; difference, 18.8% [95% CI, 13.4%-24.2%]; P < .001) compared with nonhomeless adults (Figure 4C). In contrast, mortality among homeless patients hospitalized for heart failure did not significantly differ from the nonhomeless cohort (1.6% vs 1.6%; difference, −0.02% [95% CI, −0.3% to 0.2%]; P = .83) (Figure 4D).

To understand whether there were within-hospital differences in clinical management, we also compared standardized procedure rates for AMI and stroke among patients admitted to the same hospital (eTable 5 in the Supplement). Risk-standardized rates of coronary angiography (37.9% vs 46.8%; difference, −8.8% [95% CI, −13.6% to −4.0%]; P < .001) and PCI (24.7% vs 28.0%; difference, −3.3% [95% CI, −5.5% to −1.1%]; P = .001) were still significantly lower for homeless adults compared with nonhomeless adults hospitalized for AMI at the same site of care. These patterns persisted when stratified by STEMI and non-STEMI. The rates of cerebral angiography (2.8% vs 4.5%; difference, −1.7% [95% CI, −3.2% to −0.1%]; P = .02) were lower among homeless adults with stroke than among nonhomeless adults with stroke, although rates of thrombolysis did not differ.

E-values for the risk-standardized in-hospital mortality rates between homeless and nonhomeless adults were 1.27 for AMI, 2.15 for stroke, and 1.98 for cardiac arrest. Among the subgroup of patients with STEMI, the E-value for in-hospital mortality was 2.01, and among those with ischemic stroke, the E-value for in-hospital mortality was 2.47. In addition, we repeated our analysis after excluding hospitals with fewer than 50 hospitalizations among homeless individuals across the study period, the findings of which were consistent with our main analysis (eTables 6 and 7 in the Supplement).

In this study of homeless adults hospitalized for cardiovascular conditions, we observed significant disparities in the use of diagnostic and therapeutic procedures compared with nonhomeless adults, even after accounting for differences in demographics (age, sex, and race/ethnicity), insurance payer, and clinical comorbidities. These patterns of procedure use were, in part, explained by within-hospital differences in the clinical management of homeless and nonhomeless patients. More importantly, we also found that risk-standardized in-hospital mortality rates for most cardiovascular conditions were substantially higher among homeless persons.

There has been significant concern about the health outcomes of homeless individuals in the United States, particularly because this population is now growing and the number of individuals who are homeless and elderly is projected to triple by 2030.^10,23 Despite greater policy and public health focus, homeless persons have premature mortality and high mortality rates, in part due to cardiovascular disease.⁵ A study of homeless adults residing in Philadelphia, Pennsylvania, in the 1980s found that heart disease was the second leading cause of death,⁴ and more recent evidence from the Boston Health Care for the Homeless Program has shown that heart disease remains a major cause of death among homeless adults,³ unchanged compared with 15 years prior.²⁴ Our findings build on these prior studies by focusing on hospitalization-related care and outcomes in a contemporary cohort of homeless adults admitted for acute cardiovascular conditions across 3 states.

We observed important differences in care and mortality for homeless patients compared with their nonhomeless counterparts. Homeless adults with AMI, for example, underwent coronary angiography at approximately half the rate of similar nonhomeless adults. This finding suggests that clinicians were less likely to pursue a diagnostic procedure often needed to guide therapeutic decision-making, resulting in lower rates of PCI and CABG in the homeless cohort. More important, among the subgroup of patients with STEMI, a condition for which urgent revascularization is indicated, homeless individuals were less likely to undergo coronary angiography (55.8% vs 85.9%) or PCI (55.4% vs 75.9%) than nonhomeless persons. These differences may explain the high in-hospital mortality rates among homeless patients hospitalized for STEMI. We also found similar gaps in care for other cardiovascular conditions. Condition-specific procedure rates were much lower among homeless individuals hospitalized for cardiac arrest or stroke, and associated mortality rates were substantially higher. In contrast, mortality did not differ among patients with heart failure, possibly because their outcomes are less dependent on the delivery of specific procedures.

There are several potential explanations as to why homeless patients receive lower-intensity care for acute cardiovascular conditions. Most concerning is the possibility that homeless patients are treated differently for nonclinical reasons, as suggested by the within-hospital differences in care observed in our study. Prior investigations have demonstrated racial/ethnic and sex disparities in the delivery of life-saving procedural interventions for AMI,^25,26 potentially due in part to implicit bias.^27,28 It is possible that the stigma associated with homelessness also contributes to disparities in care during a hospitalization. These differences in procedural care were not primarily a problem of insurance coverage, given that only a small minority of the homeless individuals in our study were uninsured.

Alternatively, other factors not captured in our study may explain these patterns. For instance, homeless persons face structural barriers that can result in late presentation to acute care facilities for emergency conditions, potentially limiting clinicians’ ability to deliver potentially life-saving interventions. In addition, decision-making on whether to perform procedures for homeless patients, such as PCI for AMI, may be affected by concerns about lack of adherence to medications (eg, dual antiplatelet therapy) and the potential for downstream harm. Homeless persons often face financial barriers to obtaining prescription medications and experience competing priorities (eg, food and shelter), which may impede adherence.^6,29,30 Efforts to tailor decision-making regarding the delivery of cardiovascular procedures (eg, PCI) to the circumstances of each individual patient, to use case management services, and to provide close, integrated follow-up care could mitigate concerns about adherence and reduce disparities in care intensity for homeless persons with acute cardiovascular conditions.⁶

The cumulative burden of cardiovascular, psychosocial, and environmental risk factors may also be associated with worse outcomes in the homeless population. Homeless individuals often lack reliable access to a usual source of care.³¹ This lack of access impedes the delivery of primary and preventive care, as well as the timely diagnosis and treatment of disease, particularly given the unique barriers to medication access and adherence that this population faces.³⁰ As a result, compared with the general population, homeless individuals often have poorer blood pressure control,³² greater complications associated with diabetes,³³ and high rates of cigarette smoking,^34,35 all of which are associated with excess cardiovascular mortality.⁶ Furthermore, high rates of serious mental illness (eg, schizophrenia), posttraumatic stress disorders, and polysubstance use also are associated with worse cardiovascular outcomes in this population.^36-42

Our findings have important policy implications. Most hospitalizations for homeless adults were concentrated at a small number of hospitals, and when we evaluated homeless and nonhomeless persons treated at the same sites, differences in procedural care were partially attenuated. This finding indicates that hospitals caring primarily for homeless persons may have limited resources and services that may be associated with disparities in delivery of care. In addition, these hospitals are likely to be safety-net institutions that care for other vulnerable populations of people who may also face similar barriers to accessing primary or follow-up care and adhering to medications. Although the Patient Protection and Affordable Care Act expanded Medicaid eligibility and increased funding for outpatient health centers that care for homeless persons, our findings suggest that policy initiatives should also focus on supporting hospitals that primarily serve this population. Providing additional resources to safety-net hospitals—whether financial or in-kind—to implement multidisciplinary care pathways, enhance care coordination, facilitate transitions to medical respite units,⁴³ and provide intensive case management specifically for homeless individuals, coupled with ongoing efforts to enhance longitudinal access to ambulatory health care services and promote housing stability through a “housing first” approach,^44-47 could collectively help improve health outcomes in this highly vulnerable population.

This study has some limitations. First, we focused on homeless hospitalizations in only 3 states, which may not be representative of the homeless population in the United States. Nonetheless, more than one-quarter of the homeless population in the United States resides in the 3 states examined in our study.

Second, our analysis was observational and may not have captured unmeasured differences in acuity of illness between homeless and nonhomeless patients. However, our stratification approach accounted for important observable differences between these populations, including age, sex, race/ethnicity, insurance payer, and clinical comorbidities. Third, homelessness was defined based on whether a hospital directly reported a patient to be homeless at the time of discharge or death. We suspect that this method is reasonably specific in that hospitals are unlikely to make this designation in the absence of housing problems. However, given that the spectrum of housing issues is large, this approach may not have captured individuals who experienced housing insecurity or who were transiently homeless.⁴⁸ In addition, we were unable to determine the approach used by hospitals to obtain and record information about homelessness, and it is possible that there was variation in the definition of homelessness across hospitals. Finally, we were unable to investigate postdischarge care patterns and outcomes in the homeless population, which represent important areas for future study.

We found that mortality rates were higher among homeless persons than nonhomeless persons hospitalized with acute cardiovascular conditions, even after accounting for differences in demographics (age, sex, and race/ethnicity), insurance payer, and clinical comorbidities. We also observed substantially lower use of potentially life-saving procedures for homeless adults, which may explain differences in mortality. This finding was explained, in part, by within-hospital differences in the clinical management of these populations. Further qualitative research is needed to understand how health care professional decision-making, when caring for homeless individuals, contribute to these patterns of care. Collectively, our findings suggest a need for public health and policy efforts to focus on reducing disparities in hospital-based acute care for homeless individuals to improve health outcomes in this population.

Accepted for Publication: October 14, 2019.

Corresponding Author: Rishi K. Wadhera, MD, MPP, MPhil, Richard A. and Susan F. Smith Center for Outcomes Research in Cardiology, Division of Cardiology, Beth Israel Deaconess Medical Center, Harvard Medical School, 375 Longwood Ave, Boston, MA 02215 (rwadhera@bidmc.harvard.edu).

Published Online: November 18, 2019. doi:10.1001/jamainternmed.2019.6010

Author Contributions: Drs Wadhera and Shen had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Wadhera, Shen, Joynt Maddox.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Wadhera, Choi, Joynt Maddox.

Critical revision of the manuscript for important intellectual content: Wadhera, Khatana, Jiang, Shen, Yeh, Joynt Maddox.

Statistical analysis: Wadhera, Choi, Shen.

Obtained funding: Yeh.

Supervision: Wadhera, Yeh, Joynt Maddox.

Conflict of Interest Disclosures: Dr Wadhera reported receiving research support from grant K23 HL148525-01 from the National Heart, Lung, and Blood Institute, and receiving personal fees from Regeneron outside the submitted work. Dr Yeh reported receiving research support from grant R01HL136708 from the National Heart, Lung, and Blood Institute and the Richard A. and Susan F. Smith Center for Outcomes Research in Cardiology; receiving grant funding from Abiomed; receiving personal fees from Asahi Intecc, Medtronic, and Teleflex; receiving grants from AstraZeneca; and receiving grants and personal fees from Boston Scientific outside the submitted work. Dr Joynt Maddox reported receiving research support from grant R01HL143421 from the National Heart, Lung, and Blood Institute, grant R01AG060935 from the National Institute on Aging, and the Commonwealth Fund; and previously performing contract work for the from US Department of Health and Human Services outside the submitted work. No other disclosures were reported.

Meeting Presentation: This study was presented at the Annual Meeting of the American Heart Association; November 18, 2019; Philadelphia, Pennsylvania.