Social Vulnerability and Long‐Term Cardiovascular Outcomes After COVID‐19 Hospitalization: An Analysis of the American Heart Association COVID‐19 Registry Linked With Medicare Claims Data

Abstract

Background

Patients hospitalized with COVID‐19 from socioeconomically vulnerable communities are at risk for in‐hospital cardiovascular events. However, the association of socioeconomic vulnerability and outcomes after hospitalization is uncertain.

Methods and Results

American Heart Association COVID‐19 Cardiovascular Disease Registry hospitalizations between March 1, 2020, and June 30, 2022, linked with Medicare fee‐for‐service claims, were analyzed. We used Centers for Disease Control and Prevention's Social Vulnerability Index to ascertain county‐level and Medicare‐Medicaid dual eligibility to ascertain patient‐level social vulnerability. We evaluated the association between social vulnerability and a composite of myocardial infarction, stroke, heart failure, venous thromboembolism, cardiogenic shock, cardiac arrest, and death, following discharge, using Cox regression models. The study included 8565 patients (mean age 78 years, 50% female, 16% Black, 4% Hispanic, 25% dual eligible, 34% residing in the most vulnerable counties). Patients residing in the most vulnerable counties, and dual eligible patients, were more likely to be female, Black or Hispanic, and have increased comorbidities. A total of 3783 (52%) patients experienced a composite outcome. We found no association between the most vulnerable, compared with least vulnerable, counties and cardiovascular events (hazard ratio [HR], 0.97 [95% CI, 0.87–1.07]). Dual eligibility, compared with nondual eligibility, was associated with increased cardiovascular events (HR, 1.28 [95% CI, 1.19–1.37]), which was attenuated after adjusting for comorbidities (HR, 0.97 [95% CI, 0.89–1.04]).

Conclusions

Among survivors of COVID‐19 hospitalization, patient‐level social vulnerability was associated with cardiovascular events, explained by increased comorbidities. County‐level social vulnerability was not observed to be a risk for postdischarge events. Findings suggest targeting public health efforts toward dual eligible patients to mitigate poor outcomes.

Nonstandard Abbreviations and Acronyms

AHA COVID‐19 CVD Registry

American Heart Association COVID‐19 Cardiovascular Disease Registry

FFS

fee‐for‐service

MBSF

Medicare Beneficiary Summary Files

SMD

standardized mean differences

SVI

Social Vulnerability Index

Clinical Perspective.

What Is New?

• Although patients residing in more vulnerable communities have been previously shown to have worse in‐hospital outcomes with COVID‐19 hospitalization, this study found no significant disparities in long‐term cardiovascular risk after discharge based on county‐level social deprivation.

• Patients dual eligible for Medicare‐Medicaid services continue to be at an elevated risk for worse outcomes compared with nondual eligible patients after discharge for COVID‐19 hospitalization and it is likely driven by comorbidity burden.

What Are the Clinical Implications?

• To improve patient outcomes, greater public health efforts should be focused on minimizing the burden of medical comorbidities on dual eligible patients.

As of June 2024, there have been more than 100 million confirmed cases of COVID‐19 infections with a total of over 1 million deaths in the United States. ¹ , ² Early in the pandemic, in‐hospital mortality for COVID‐19 ranged from 10% to 20%, to upwards of 50% for patients in the intensive care unit requiring mechanical ventilation. ³ , ⁴ , ⁵ More recently, overall in‐hospital mortality is approximately 5% and up to 40% for those who require intubation. ⁶ Patients that survive their COVID‐19 hospitalization have an elevated risk of adverse health outcomes, hospital readmission, and death, ⁵ , ⁷ , ⁸ and many continue to suffer long‐term sequelae from their infection. ⁹ , ¹⁰ COVID‐19 disproportionately affects socioeconomically disadvantaged populations with higher rates of hospitalization and mortality among low‐income, as well as Black, Hispanic, and other underrepresented racial and ethnic groups. ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵

Although prior studies have demonstrated that patients hospitalized with COVID‐19 who reside in more socially vulnerable communities have higher rates of in‐hospital mortality and adverse cardiovascular events, ¹⁶ whether this increased risk persists after discharge is uncertain. Additionally, minimal data reports on the effects of patient‐level socioeconomic vulnerability by low‐income status on COVID‐19 outcomes, and they focus primarily on in‐hospital outcomes. ¹³ , ¹⁷ , ¹⁸ Understanding whether these individuals are also at increased risk for long‐term adverse sequelae could inform strategies to mitigate poor outcomes.

Thus, in this study, we used a linked data set to determine the incidence of postdischarge cardiovascular complications and death among survivors of a COVID‐19 hospitalization and examine the association of social vulnerability, both at a county level and patient level, with post‐discharge outcomes.

METHODS

Data Sources

For this study we linked 3 separate data sets: the AHA COVID‐19 CVD Registry (American Heart Association COVID‐19 Cardiovascular Disease Registry), Medicare fee‐for‐service (FFS) claims, and data from the Centers for Disease Control and Prevention/ATSDR (Agency for Toxic Substances and Disease Registry) 2020 Social Vulnerability Index (SVI) as previously described. ¹⁹

The AHA COVID‐19 CVD Registry is a voluntary registry that was previously open for data collection to hospitals within the United States treating adult patients with COVID‐19, and it includes data from 136 hospitals in 34 states across the United States until summer of 2022. ²⁰ It was created with the goals of better understanding the effect of COVID‐19 on in‐hospital cardiovascular outcomes, accelerating the development of effective therapeutic strategies, facilitating quality improvement efforts, and informing preparations for future waves of the pandemic. The registry contains granular information on patients admitted with COVID‐19 infections and it has supported research on short‐term cardiovascular outcomes. The registry, however, does not capture some important historical patient information and lacks postdischarge follow‐up, limiting its utility in examining the long‐term impact of COVID‐19.

We linked patient‐level data from the AHA COVID‐19 CVD Registry with Medicare FFS claims to ascertain postdischarge outcomes. We used Medicare Beneficiary Summary Files (MBSF) for patient demographics, including date of birth, sex, race and ethnicity, zip code, and dual eligibility status. ²¹ Race and ethnicity in our study were ascertained for each patient based on what was included in the administrative data set on the MBSF and categories reported include Non‐Hispanic White, Black, Asian, Hispanic, North American Native, Other, and Unknown. For our analysis, we used the racial and ethnic categories of Non‐Hispanic White, Black, and Other (which included patients categorized as Asian, Hispanic, North American Native, Other, and Unknown race or ethnicity). Patient comorbidities were identified by MBSF 30 Chronic Conditions Data Warehouse and Other Chronic or Potentially Disabling Conditions segments. ²² , ²³ Finally, hospitalization data for outcome events were ascertained from Inpatient FFS Claim Files. ²⁴ International Classification of Diseases, Tenth Revision (ICD‐10) codes were used to identify inpatient diagnoses for COVID‐19 (U07.1, B97.29, J12.82).

County‐level socioeconomic vulnerability was assessed using the Centers for Disease Control and Prevention/ATSDR's 2020 county‐level SVI. First published in 2011, the SVI was created to indicate the relative vulnerability of US Census tracts for emergency management. ²⁵ , ²⁶ SVI combines 16 census variables which are grouped into 4 themes with each variable ranked from highest to lowest vulnerability across US census tracts. A county‐level percentile rank is calculated for each variable and theme, and an overall percentile rank for each tract is also calculated. Higher ranks indicate greater social vulnerability. We used MBSF to identify the 5‐digit county level federal information processing system code for each beneficiary at the month of COVID‐19 admission.

Data Management

The Richard A. and Susan F. Smith Center for Outcomes Research in Cardiology at Beth Israel Deaconess Medical Center and Centers for Medicare and Medicaid Services had established data use agreements. Centers for Medicare and Medicaid Services Virtual Research Data Center ²⁷ was used for data management and analyses to ensure AHA COVID‐19 CVD Registry data were securely handled, and all analyses and linkage was conducted using SAS version 9.4 (SAS Institute, Cary, NC, USA). Data are not publicly available, but more extensive analytical methods regarding linkage have been previously made available for the purposes of reproduction. ¹⁹ The study was approved by the Institutional Review Board at Beth Israel Deaconess Medical Center with a waiver of informed consent.

Linkage

By linking the AHA COVID‐19 CVD Registry data with Medicare FFS claims, we created a data resource that enriches the data set and allows us to obtain claims data to assess long‐term outcomes after the initial hospitalization. We used 7 linkage variables (date of birth, sex, hospital identification, zip code, race or ethnicity, admission date, and discharge date) and performed a stepwise linkage, similar to what has been previously been described. ¹⁹ In brief, we followed a 13‐step process, with the first step requiring complete data and an exact match on all 7 of the linkage variables, with subsequent steps relaxing the criteria to link additional hospitalizations, first allowing for zip code or race and ethnicity to be missing, then allowing a nonexact match for admission and discharge dates within 1, 2, or 7 days, and finally allowing for 1 of the 3 elements for birthday (ie, day, month, and year) to not match. SVI was linked to each beneficiary by their federal information processing system code (Table S1).

Study Population

Our study included individuals ≥65 years of age who were discharged alive following a hospitalization with COVID‐19 between March 1, 2020, and June 30, 2022. We excluded individuals who were not discharged alive, beneficiaries from Veterans Affairs hospitals, and patients enrolled in Medicare FFS for <12 months to allow for appropriate assessment of chronic disease from the MBSF. We also excluded individuals when multiple AHA COVID‐19 CVD Registry hospitalizations matched to a single Medicare beneficiary or when a single registry hospitalization matched to multiple beneficiaries to avoid duplication and unreliable linkage. Inpatient claims from multiple hospitals separated by 1 or fewer days were assumed to represent a hospital‐to‐hospital transfer and were treated as a single episode of care, and if a patient had multiple successfully linked COVID‐19 hospitalizations, we included only the first hospitalization in our analysis.

Exposures

We used the overall SVI rank, which ranges from 0 to 1, to ascertain county‐level socioeconomic vulnerability. We grouped individuals based on their national SVI rank where 0 to 0.25 is group 1 (least vulnerable), 0.2501 to 0.50 is group 2, 0.5001 to 0.75 is group 3, and 0.7501 to 1.0 is group 4 (most vulnerable).

We used Medicare‐Medicaid dual eligibility status as identified in FFS claims data to ascertain patient‐level socioeconomic vulnerability and grouped individuals based on their dual eligibility status. To be dual eligible, individuals must receive Medicare (based on age or disability) and Medicaid, which additionally requires individuals be low income by meeting asset and income requirements set federally and by individual states. ²⁸ , ²⁹ Beneficiaries enrolled for full dual benefits at the time of discharge from their COVID‐19 hospitalization were classified as dual eligible.

Outcomes

The primary outcome was a composite of hospitalization for acute myocardial infarction, ischemic and hemorrhagic stroke, acute heart failure, deep vein thrombosis, pulmonary embolism, cardiogenic shock, cardiac arrest, and all‐cause death following discharge from index hospitalization. Secondary outcomes were death from any cause, and a composite of all‐cause death and any cause rehospitalization. Cardiovascular events were defined based on ICD‐10 codes, as previously described (Table S2) based on face validity and prior research. ³⁰ , ³¹

Statistical Analysis

Baseline demographic information, clinical comorbidities, as well as relative severity of initial COVID‐19 infection (ie, length of stay, intensive care unit admission, and ventilation requirement) were compared across SVI groups and between dual eligible and nondual eligible patients using standardized mean differences (SMD) where an absolute difference ≥10% (0.1) is conventionally deemed notable. ³²

We censored patients who did not experience the event of interest on the last day of Medicare FFS coverage or December 31, 2022, whichever came first. Cumulative incidence functions were estimated using the Kaplan–Meier method and compared using the log‐rank test. We evaluated the association between SVI or dual eligibility with each outcome using Cox proportional hazard regression models. To account for statistical dependence among patients treated at the same hospital, we used hospital‐level cluster robust sandwich variance estimators to calculate 95% CIs. We compared hazard ratios (HR) over the entire follow‐up in an unadjusted analysis (with dual eligibility status or SVI as the sole regressor), and in 2 multivariable analyses, the first adjusting for demographic variables (age, sex, race or ethnicity), and the second additionally adjusting for preexisting clinical comorbidities, diagnosed before the index hospitalization, listed in Table S3. We evaluated the association of both county‐level and patient‐level social vulnerability with the primary composite end point as well as death and rehospitalization, using SVI national ranks and Medicare‐Medicaid dual eligibility, respectively. Given the correlation between social vulnerability at the county and patient level, we also compared the primary composite end point and death across subgroups based on SVI and dual eligibility status. Additionally, we censored follow‐up data at 6 months, 1 year, and 2 years to evaluate whether the association varied over time.

RESULTS

Baseline Characteristics

The AHA COVID‐19 CVD Registry includes a total of 55 926 patients hospitalized with COVID‐19 between March 1, 2020, and June 30, 2022, of whom 25 900 were eligible for linkage. There were 1 575 051 Medicare inpatient FFS claims for COVID‐19 during the same period. A total of 12 148 records were successfully linked between the registry and FFS claims, and a total of 8565 patients met all inclusion criteria and were included in the analysis. Of those included, the mean age is 78 years, 50% are female, 75% are Non‐Hispanic White, 16% are Black, 4% are Hispanic, and 25% are dual eligible (Figure 1).

Figure 1. Flow chart demonstrating linkage of American Heart Association COVID‐19 registry and fee‐for‐service Medicare COVID‐19 admissions.

AHA indicates American Heart Association; FFS, fee for service; ICD‐10, International Classification of Diseases, Tenth Revision; SVI, Social Vulnerability Index; and VA, Veteran's Administration.

There were 1225 individuals in the least vulnerable counties (group 1) compared with 2907 in the most vulnerable counties (group 4). Compared with the least vulnerable counties, patients from the most vulnerable counties were less likely to be male (48% versus 53%; SMD −0.11), more likely to be Black (29% versus 4%; SMD 0.69), and more likely to be dual eligible (35% versus 13%; SMD 0.53). Patients in the most vulnerable counties had a higher burden of medical comorbidities and a longer length of stay (9 days versus 8 days; SMD 0.17) during their index hospitalization for COVID‐19 but had similar rates of intensive care unit admission (22% versus 18%; SMD 0.09) and mechanical ventilation requirement (7% versus 6%; SMD 0.06). Remaining characteristics can be found in Table 1.

Table 1.

Baseline Characteristics of Patients by County‐Level Social Vulnerability Index^*

	Overall N=8565	Group 1 (least vulnerable) N=1225	Group 2 N=2102	Group 3 N=2331	Group 4 (most vulnerable) N=2907	SMD†
Demographics
Age, y, mean±SD	78 (8)	78 (8)	77 (8)	78 (8)	77 (8)	−0.13
Male sex, no. (%)	4274 (50)	655 (53)	1078 (51)	1142 (49)	1399 (48)	−0.11
Race or ethnicity, no. (%)‡
Non‐Hispanic White	6448 (75)	1106 (90)	1712 (81)	1878 (81)	1752 (60)	−0.74
Black	1357 (16)	54 (4)	203 (10)	263 (11)	837 (29)	0.69
Other	760 (9)	65 (5)	187 (9)	190 (8)	318 (11)	0.21
Asian	160 (2)	16 (1)	49 (2)	39 (2)	56 (2)	0.05
Hispanic	329 (4)	16 (1)	70 (3)	67 (3)	176 (6)	0.25
North American Native	32 (<1)	<11 (<1)§	12 (1)	<11 (<1)§	12 (<1)	0.05
Other	118 (1)	14 (1)	24 (1)	37 (2)	43 (1)	0.03
Unknown	121 (1)	17 (1)	32 (2)	41 (2)	31 (1)	−0.03
Medicare‐Medicaid dual eligible, no. (%)	2142 (25)	157 (13)	412 (20)	568 (24)	1005 (35)	0.53
Chronic comorbidities, no. (%)
Cardiovascular comorbidities
Hypertension	7476 (87)	1045 (85)	1807 (86)	2037 (87)	2587 (89)	0.11
Hyperlipidemia	7134 (83)	984 (80)	1739 (83)	1959 (84)	2452 (84)	0.11
Ischemic heart disease	5386 (63)	747 (61)	1274 (61)	1446 (62)	1919 (66)	0.11
Diabetes	4633 (54)	579 (47)	1077 (51)	1230 (53)	1747 (60)	0.26
Heart failure	3970 (46)	517 (42)	916 (44)	1100 (47)	1437 (49)	0.15
Peripheral vascular disease	3279 (38)	412 (34)	804 (38)	920 (40)	1143 (39)	0.12
Stroke or transient ischemic attack	2114 (25)	268 (22)	506 (24)	576 (25)	764 (26)	0.10
Atrial fibrillation	2089 (24)	633 (22)	583 (25)	555 (26)	318 (26)	−0.10
Prior acute myocardial infarction	797 (9)	125 (10)	189 (9)	221 (9)	262 (9)	−0.04
Oncologic comorbidities
Prostate cancer	619 (7)	87 (7)	145 (7)	167 (7)	220 (8)	0.02
Breast cancer	483 (6)	80 (7)	111 (5)	158 (7)	134 (5)	−0.08
Colorectal cancer	336 (4)	54 (4)	60 (3)	89 (4)	133 (5)	0.01
Lung cancer	232 (3)	38 (3)	55 (3)	59 (3)	80 (3)	0.02
Endometrial cancer	111 (1)	15 (1)	22 (1)	25 (1)	49 (2)	0.04
Other comorbidities
Anemia	6033 (70)	805 (66)	1432 (68)	1667 (72)	2129 (73)	0.16
Chronic kidney disease	5165 (60)	672 (55)	1247 (59)	1383 (59)	1863 (64)	0.19
Chronic liver disease	1382 (16)	176 (15)	308 (15)	388 (17)	510 (18)	0.09
Depression	4.175 (49)	540 (44)	1014 (48)	1196 (51)	1425 (49)	0.10
Chronic obstructive pulmonary disease	3269 (38)	433 (35)	756 (36)	912 (39)	1168 (40)	0.10
Acquired hypothyroidism	2788 (33)	377 (31)	701 (33)	805 (35)	905 (31)	0.01
Alzheimer's disease	3000 (35)	398 (33)	732 (35)	829 (36)	1041 (36)	0.07
Benign prostatic hyperplasia	2268 (27)	331 (27)	563 (27)	605 (26)	769 (27)	−0.01
Glaucoma	2166 (25)	281 (23)	513 (24)	614 (26)	758 (26)	0.07
Cataract	5705 (67)	817 (67)	1384 (66)	1571 (67)	1933 (67)	−0.004
Osteoporosis	2062 (24)	296 (24)	485 (23)	550 (24)	731 (25)	0.02
Asthma	1667 (20)	222 (18)	401 (19)	471 (20)	573 (20)	0.04
Hip/pelvic fracture	652 (8)	103 (8)	154 (7)	192 (8)	203 (7)	−0.05
Rheumatoid arthritis or osteoarthritis	5906 (69)	837 (68)	1391 (66)	1620 (70)	2058 (71)	0.05
Tobacco use disorders	1327 (16)	130 (11)	329 (16)	348 (15)	520 (18)	0.21
Alcohol use disorder	530 (6)	54 (4)	120 (6)	154 (7)	202 (7)	0.11
COVID‐19 severity, No. (%)
Length of stay, d, mean±SD	9 (10)	8 (8)	9 (10)	9 (10)	9 (10)	0.17
Intensive care unit admission	1754 (21)	233 (18)	477 (23)	416 (18)	638 (22)	0.09
Ventilator Required	558 (7)	70 (6)	154 (7)	129 (6)	205 (7)	0.06

SMD indicates standardized mean difference.

^*Social Vulnerability Index from the Centers for Disease Control and Prevention national percentile where a higher percentile indicates greater vulnerability and a number from 0 to 0.25 is group 1 (least vulnerable), 0.2501–0.50 is group 2, 0.5001–0.75 is group 3, and 0.7501–1.0 is group 4 (most vulnerable).

^† SMD is difference in means from Social Vulnerability Index group 1 and group 4 divided by combined SD.

^‡ Race or ethnicity were categorized as Non‐Hispanic White, Black, and Other (Other includes the combination Asian, Hispanic, North American Native, Other, and Unknown). Race and ethnicity were collected as reported by the Medicare Beneficiary Summary File.

^§ Due to Medicare reporting requirements, values between 1 and 10 must be suppressed.

A total of 2142 dual eligible individuals were included in the study cohort. Patients who were dual eligible were less likely to be male (42% versus 53%; SMD −0.23) and more likely to be Black (27% versus 12%; SMD 0.39). Dual eligible patients had a higher burden of comorbidities and a longer total length of stay (10 versus 9 days; SMD 0.13) during their index hospitalization for COVID‐19 but had similar rates of intensive care unit admission (21% versus 20%; SMD 0.03) and mechanical ventilation (7% versus 6%; SMD 0.03). Remaining characteristics can be found in Table 2.

Table 2.

Baseline Characteristics of Patients by Dual Eligibility Status

	Overall N=8565	Dual eligible N=2142	Nondual eligible N=6423	SMD*
Demographics
Age, y, mean±SD	78 (8)	77 (8)	78 (8)	−0.08
Male sex, no. (%)	4274 (50)	889 (42)	3385 (53)	−0.23
Race or ethnicity, no. (%)†
Non‐Hispanic White	6448 (75)	1189 (56)	5259 (82)	−0.59
Black	1357 (16)	585 (27)	772 (12)	0.39
Other	760 (9)	368 (17)	392 (6)	0.35
Asian	160 (2)	79 (4)	81 (1)	0.16
Hispanic	329 (4)	204 (10)	125 (2)	0.33
North American Native	32 (<1)	<11 (<1)‡	23 (<1)	0.01
Other	118 (1)	42 (2)	76 (1)	0.06
Unknown	121 (1)	34 (2)	87 (1)	0.02
SVI national percentile, mean±SD§	0.58 (0.26)	0.66 (0.24)	0.56 (0.26)	0.40
Chronic comorbidities, no. (%)
Cardiovascular comorbidities
Hypertension	7476 (87)	2003 (94)	5473 (85)	0.27
Hyperlipidemia	7134 (83)	1858 (87)	5276 (82)	0.13
Ischemic heart disease	5386 (63)	1522 (71)	3864 (60)	0.23
Diabetes	4633 (54)	1486 (69)	3147 (49)	0.42
Heart failure	3970 (46)	1255 (59)	2715 (42)	0.33
Peripheral vascular disease	3279 (38)	1105 (52)	2174 (34)	0.37
Stroke or transient ischemic attack	2114 (25)	701 (33)	1413 (22)	0.24
Atrial fibrillation	2089 (24)	509 (24)	1580 (25)	−0.02
Prior acute myocardial infarction	797 (9)	226 (11)	571 (9)	0.06
Oncologic comorbidities
Prostate cancer	619 (7)	114 (5)	505 (8)	−0.10
Breast cancer	483 (6)	102 (5)	381 (6)	−0.05
Colorectal cancer	336 (4)	82 (4)	254 (4)	−0.01
Lung cancer	232 (3)	51 (2)	181 (3)	−0.03
Endometrial cancer	111 (1)	29 (1)	82 (1)	0.01
Other comorbidities
Anemia	6033 (70)	1768 (83)	4265 (66)	0.38
Chronic kidney disease	5165 (60)	1532 (72)	3633 (57)	0.32
Chronic liver disease	1382 (16)	411 (19)	971 (15)	0.11
Depression	4.175 (49)	1326 (62)	2849 (44)	0.36
Chronic obstructive pulmonary disease	3269 (38)	1062 (50)	2207 (34)	0.31
Acquired hypothyroidism	2788 (33)	760 (36)	2028 (32)	0.08
Alzheimer's disease	3000 (35)	1080 (50)	1920 (30)	0.43
Benign prostatic hyperplasia	2268 (27)	524 (25)	1744 (27)	−0.06
Glaucoma	2166 (25)	607 (28)	1559 (24)	0.09
Cataract	5705 (67)	1497 (70)	4208 (66)	0.09
Osteoporosis	2062 (24)	598 (28)	1464 (23)	0.12
Asthma	1667 (20)	578 (27)	1089 (17)	0.24
Hip/pelvic fracture	652 (8)	176 (8)	476 (7)	0.03
Rheumatoid arthritis or osteoarthritis	5906 (69)	1585 (74)	4321 (67)	0.15
Tobacco use disorders	1327 (16)	506 (24)	821 (13)	0.28
Alcohol use disorder	530 (6)	204 (10)	326 (5)	0.17
COVID‐19 severity, No. (%)
Length of stay, d, mean±SD	9 (10)	10 (11)	9 (10)	0.13
Intensive care unit admission	1754 (21)	459 (21)	1295 (20)	0.03
Ventilator required	558 (7)	150 (7)	408 (6)	0.03

SMD indicates standardized mean difference; and SVI, social vulnerability index (national percentile).

^*SMD is difference in means from SVI group 1 and group 4 divided by combined SD.

^† Race and ethnicity were categorized as Non‐Hispanic White, Black and Other (Other includes the combination Asian, Hispanic, North American Native, Other and Unknown). Race and ethnicity were collected as reported by the Medicare Beneficiary Summary File.

^‡ Due to Medicare reporting requirements, values between 1 to 10 must be suppressed.

^§ SVI from the Centers for Disease Control and Prevention national percentile where a higher percentile indicates greater vulnerability and a number from 0 to 0.25 is group 1 (least vulnerable), 0.2501–0.50 is group 2, 0.5001–0.75 is group 3, and 0.7501–1.0 is group 4 (most vulnerable).

Outcome Events

Over a median (interquartile range) follow‐up of 620 (25th to 75th percentile: 210 to 793) days, 3783 patients experienced a primary outcome event, of whom, 2869 had died by the end of the follow‐up period leading to cumulative incidence proportions (risk) of 52.0% (95% CI, 50.6%–53.5%) and 40.7% (95% CI, 39.2%–42.2%) respectively (Figure 2).

Figure 2. Cumulative primary outcome and death over time following live discharge by social vulnerability.

Cumulative incidence of (A) primary outcome and (B) death as a function of time following discharge stratified by SVI groups. Cumulative incidence of (C) primary outcome and (D) death as a function of time following discharge stratified by dual eligibility status. Primary outcome is a composite of hospitalization for acute myocardial infarction, ischemic and hemorrhagic stroke, acute heart failure, deep vein thrombosis, pulmonary embolism, cardiogenic shock, cardiac arrest, and all‐cause death following discharge from index hospitalization during entire follow‐up period. SVI from the Centers for Disease Control and Prevention national percentile where a higher percentile indicates greater vulnerability and a number from 0 to 0.25 is group 1 (least vulnerable), 0.2501–0.50 is group 2, 0.5001–0.75 is group 3, and 0.7501–1.0 is group 4 (most vulnerable). SVI indicates Social Vulnerability Index.

Outcomes by SVI

Risk of the primary composite outcome was similar between the most and least vulnerable SVI groups (approximately 51% in both group 1 and group 4, P=0.49) (Figure S1) and there were no differences between the most and least vulnerable SVI groups in unadjusted models (HR, 0.97 [95% CI, 0.87–1.07]) and models adjusting for demographics and comorbidities (HR, 0.90 [95% CI, 0.81–1.06]). Risk of death between the most and least vulnerable SVI groups were also similar (40% versus 41% in group 1 and group 4 respectively, P=0.87) (Figure S1) without differences in unadjusted (HR, 1.01 [95% CI, 0.90–1.14]) and adjusted (HR, 0.95 [95% CI, 0.84–1.08]) models (Table 3). Our findings were similar when censoring follow‐up data at 6 months (Table S4), 1 year (Table S5), and 2 years (Table S6). Additionally, when patients were grouped based on dual eligibility status, there was no significant differences in primary outcome (Table S7) or death (Table S8) when comparing across SVI groups. There also was no difference when comparing the composite of death and rehospitalization between the most and least vulnerable SVI groups in unadjusted (HR, 1.04 [95% CI, 0.94–1.14]) and adjusted (HR, 0.98 [95% CI, 0.87–1.10]) models (Table S9).

Table 3.

Primary Outcome and Death Based on Social Vulnerability

	Model 1* HR (95% CI)	Model 2† HR (95% CI)	Model 3‡ HR (95% CI)
Outcome
Primary outcome§
SVI Group 1‖	Ref	Ref	Ref
SVI Group 2‖	1.04 (0.94–1.16)	1.08 (0.97–1.20)	1.03 (0.93–1.15)
SVI Group 3‖	0.98 (0.88–1.09)	0.99 (0.89–1.09)	0.91 (0.82–1.02)
SVI Group 4‖	0.97 (0.87–1.07)	1.01 (0.90–1.12)	0.90 (0.81–1.06)
Death
SVI Group 1‖	Ref	Ref	Ref
SVI Group 2‖	1.09 (0.97–1.24)	1.14 (1.01–1.29)	1.07 (0.94–1.21)
SVI Group 3‖	0.98 (0.87–1.11)	0.99 (0.88–1.12)	0.91 (0.80–1.03)
SVI Group 4‖	1.01 (0.90–1.14)	1.06 (0.94–1.20)	0.95 (0.84–1.08)
Primary outcome§
Nondual eligible	Ref	Ref	Ref
Dual eligible	1.28 (1.19–1.37)	1.36 (1.26–1.47)	0.97 (0.89–1.04)
Death
Nondual eligible	Ref	Ref	Ref
Dual eligible	1.28 (1.19–1.39)	1.39 (1.28–1.52)	0.99 (0.91–1.08)

HR indicates hazard ratio; and SVI, social vulnerability index (national percentile).

^*Model 1 is unadjusted.

^† Model 2 is adjusted for age, sex, and race or ethnicity. Race and ethnicity for our analysis were categorized as Non‐Hispanic White, Black and Other (Other includes the combination Asian, Hispanic, North American Native, Other and Unknown). Race and ethnicity were collected as reported by the Medicare Beneficiary Summary File.

^‡ Model 3 is adjusted for age, sex, race or ethnicity, and medical comorbidities.

^§ Primary outcome is a composite of hospitalization for acute myocardial infarction, ischemic and hemorrhagic stroke, acute heart failure, deep vein thrombosis, pulmonary embolism, cardiogenic shock, cardiac arrest, and all‐cause death following discharge from index hospitalization during entire follow‐up period.

^‖ SVI from the Centers for Disease Control and Prevention national percentile where a higher percentile indicates greater vulnerability and a number from 0 to 0.25 is group 1 (least vulnerable), 0.2501–0.50 is group 2, 0.5001–0.75 is group 3, and 0.7501–1.0 is group 4 (most vulnerable).

Outcomes by Dual Eligibility Status

Risk of the primary composite outcome differed based on Medicare‐Medicaid dual eligibility status (62% for dual eligible versus 49% for nondual eligible, P<0.01) (Figure S2). Individuals dual eligible were at an increased risk of the primary outcome (HR, 1.28 [95% CI, 1.19–1.37]), which persisted after controlling for demographics (adjusted HR, 1.36 [95% CI, 1.26–1.47]) but was attenuated after additionally controlling for comorbidities (adjusted HR, 0.97 [95% CI, 0.89–1.04]) (Table 3). Risk of death was also different across dual status (49% for dual eligible versus 38% for nondual eligible, P<0.01), and dual eligible patients had an increased risk of death (HR, 1.28 [95% CI, 1.19–1.39]). The increased risk of death remained when controlling for demographics (adjusted HR, 1.39 [95% CI, 1.28–1.52]) but was lost after additionally controlling for comorbidities (adjusted HR, 0.99 [95% CI, 0.91–1.08]) (Table 3). We found similar differences to uncensored follow‐up data when censoring follow‐up at 6 months (Table S4), 1 year (Table S5), and 2 years (Table S6). Additionally, when comparing subgroups within the same SVI group, similar trends were seen in the primary outcome (Table S10) and death (Table S11). The composite of all‐cause death and rehospitalization was also higher among dual eligible patients (HR, 1.36 [95% CI, 1.26–1.47]), which remained when controlling for demographics (HR, 1.42 [95% CI, 1.31–1.53]) but attenuated after controlling for comorbidities (HR, 1.04 [95% CI, 0.97–1.12]) (Table S9).

DISCUSSION

This study aimed to investigate the incidence of postdischarge cardiovascular events and death among older adult survivors of COVID‐19 hospitalizations, and examine the impact of social vulnerability, both at the county and patient level, on postdischarge outcomes. We used a data set linking the AHA COVID‐19 CVD Registry with Medicare FFS claims data to obtain rich patient demographic, in‐hospital, and postdischarge clinical information. We found that almost half of the patients in our cohort experienced a primary outcome event following discharge over a median of only 620 days of follow‐up. We found no significant association with county‐level vulnerability and postdischarge events or death. We did, however, find patient‐level social vulnerability was associated with an increased risk of postdischarge events and death, with this increased risk attributable to the presence of a greater burden of clinical comorbidities.

Patients from vulnerable communities have been disproportionately affected by COVID‐19, with a greater number of cases, ³³ , ³⁴ , ³⁵ and ecological and population based studies demonstrating increased COVID‐19 mortality. ³⁶ , ³⁷ , ³⁸ Although county‐level social vulnerability has been associated with in‐hospital events and death, ¹⁶ , ³⁹ we found no association with postdischarge events or death. This may indicate that for COVID‐19, county‐level vulnerability has a temporal effect, where these communities are at highest risk during the acute phase of their infection, but that this effect is attenuated should they survive to discharge. In addition, it is possible that patients who previously were not connected to important outpatient medical or social services, might be connected to these services once hospitalized. Given that all patients are enrolled in Medicare, a public health insurance, this may also be demonstrating the success of public health insurance, that it is appropriately servicing individuals from various counties irrespective of vulnerability.

Dual eligibility is a proxy of patient‐level social vulnerability. Compared with nondual eligible patients, dual eligible beneficiaries have been found to experience worse cardiovascular outcomes, ⁴⁰ , ⁴¹ , ⁴² are more prone to COVID‐19 hospitalizations, ¹³ , ¹⁸ , ⁴³ and have an increased risk of death. ⁴⁴ , ⁴⁵ In our study, dual eligible individuals hospitalized with COVID‐19 had increased risk of postdischarge events and death that was maintained throughout the entire follow‐up period. The increased risk remained despite adjusting for demographics but after adjusting for comorbidities the association was no longer significant. This may demonstrate public health insurance succeeding among patients when appropriately risk adjusted; however, our findings indicate that comorbidities are likely mediating a large portion of the effect in our population. Chronic disease and poverty are interconnected, and although dual eligibility status might be driven by increased comorbidities and medical complexity, county‐level vulnerability was also associated with an increased number of comorbidities. However, there were no differences in events between SVI groups. Although we cannot discern this from our data, dual eligibles may have more advanced or insufficient control over chronic conditions compared with nondual eligibles, which may be acting as a risk enhancer. Prior studies have shown increased readmissions for chronic diseases in dual eligible patients which may also indicate worse control. ⁴⁶ , ⁴⁷ , ⁴⁸ Dual eligibility status is also a more granular metric compared with SVI. SVI encompasses an entire county by federal information processing system codes where significant heterogeneity in patient‐level vulnerability may exist. In our study, about one third of individuals within the highest‐risk counties were dual eligible.

A unique aspect, and strength, of our study is the use of a data set linking registry data with FFS claims data. The AHA COVID‐19 CVD Registry captures extensive information regarding patient hospitalization, allowing more rigorous adjustment for baseline clinical characteristics than analyses based on claims data alone. Prior studies have used this registry data for evaluating in‐hospital outcomes with COVID‐19, but due to an intrinsic limitation in registry data, long‐term outcome data are scant. ¹⁶ , ⁴⁹ , ⁵⁰ By linking registry data with FFS claims, as well as SVI and dual eligibility status allowed us to overcome this limitation and evaluate the association of social vulnerability with long‐term outcomes. Other strengths of our study include the large sample size with a long follow‐up period: our study cohort included more than 8000 individuals with a median follow‐up of almost 2 years, allowing for adequate statistical power and time to evaluate outcomes following discharge. Our linked cohort had previously been shown to be generally representative of general Medicare FFS beneficiaries, ¹⁹ thus our findings are generalizable to patients above the age of 65 years enrolled in Medicare FFS residing across the United States.

Our study suggests that there is room for greater public health interventions for dual eligible individuals. Dual eligible patients had a disproportionately increased risk of events and death despite surviving a COVID‐19 hospitalization. Unfortunately, outpatient care tends to be less available and underused among dual eligible patients, ⁵¹ who also tend to have decreased use of optimal guideline‐directed care for many chronic conditions. ⁵² , ⁵³ Availability and use of primary care decreases the probability of hospitalizations, ⁵⁴ thus efforts to increase access to outpatient primary and specialty care is essential in this population. Reducing the burden of chronic disease in this population can help mitigate poor outcomes, and it is worth not only focusing public health efforts at a county level but also at a patient level. Additionally, given the increased risk of poor cardiovascular outcomes among dual eligible patients, physicians and care teams may begin considering dual eligible status when evaluating the overall postdischarge risk of patients with COVID‐19. A potential strategy for practicing clinicians caring for these patients would be early case management involvement to maximize and target postdischarge resources and ensuring adequate follow‐up. Finally, given the elevated risks among dual eligible patients, steps can be taken by Centers for Medicare and Medicaid Services in various capacities to optimize the care of this population both in and out of the hospital. The Home Health Value‐Based Purchasing model, a value‐based payment model to home health agencies that incentivizes successful discharge to the community after acute hospitalization and prevention of readmissions, was recently expanded across the country after the pilot program demonstrated improved quality performance metrics, reductions in hospitalizations and skilled nursing facility stays, and system‐level financial savings. ⁵⁵ Specific provisions that provide incentives for improved postdischarge care for dual eligible patients could be considered, with careful attention to the potential for unintended consequences, which have worsened existing health disparities among vulnerable patients in other value‐based purchasing programs. ⁵⁶ Financial incentives should therefore also be implemented to health systems that care for dual eligible patients to help provide the adequate resources to mitigate risk in this population. Centers for Medicare and Medicaid Services has already taken steps in this direction with the planned implementation of health equity adjustments, which will likely shift resources toward health systems and organizations caring for dual eligible and other vulnerable populations. ⁵⁷ , ⁵⁸ Finally, efforts to integrate the currently fragmented Medicare and Medicaid system for dual eligible patients may also lead to decreased system costs and improved outcomes among dual eligible patients. ⁵⁹

Our study has limitations. The cohort was restricted to patients 65 years or older enrolled in traditional Medicare plans; our findings may not be generalizable to younger patients or to those enrolled in Medicare Advantage plans. However, it is well established that COVID‐19 disproportionately affects older adults, highlighting the need to focus on this population. Moreover, our population includes only insured patients and thus does not account for social vulnerability among uninsured patients, including those not eligible or who have not signed up for Medicare. Exclusion of uninsured patients, who are at the extreme of social vulnerability, is likely attenuating the association between social vulnerability and outcomes towards null. The AHA COVID‐19 CVD Registry may overrepresent patients hospitalized in medical centers with the resources to participate in the voluntary program, particularly large academic centers in well‐resourced urban areas. Additionally, of the original 55 926 records from the AHA COVID‐19 CVD Registry, only about 22% were successfully linked and 15% were ultimately included in our analysis. Our data set, however, has been shown in prior analyses to be representative of the general Medicare FFS population, and when compared with the unlinked cohort is similar with respect to demographics and burden of chronic diseases. ¹⁹ As with any clinical database analysis, there is the potential for misclassification bias. Misclassification at the level of COVID‐19 diagnosis is attenuated through the requirement of an active infection during index hospitalization and excludes individuals with prior infections who continue to test positive without any symptoms. Finally, our study evaluated COVID‐19 hospitalizations during numerous waves of COVID‐19 with various strains and included hospitalizations following the availability of different COVID‐19 directed vaccines and therapies and likely included individuals with prior uncaptured infections. Although this increases the heterogeneity of the population, it also is more reflective of spectrum of patients hospitalized with COVID‐19 who have been exposed to various strains through the numerous waves of the pandemic.

CONCLUSIONS

In this study of Medicare FFS beneficiaries 65 years or older who survived a hospitalization with COVID‐19, socioeconomic vulnerability at a patient level is associated with adverse cardiovascular outcomes and deaths; this association appears to be mediated by a greater burden of comorbidities in dual eligible individuals. County‐level vulnerability, which has previously been associated with poor outcomes during COVID‐19 hospitalization, was not associated with long‐term cardiovascular outcomes and death among those surviving to discharge appears to be less pronounced. Findings suggest the need to focus public health efforts on dual eligible individuals to help mitigate the burden of comorbidities in this population.

Sources of Funding

This project is supported by the National Institutes of Health (grant number R01HL157530). The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the article; and decision to submit the article for publication.

Disclosures

Dr Yeh receives support from the National Heart, Lung, and Blood Institute who also funded this study (R01HL157530). Rishi K. Wadhera receives research support from the National Heart, Lung, and Blood Institute, the Donaghue Foundation, and the American Heart Association, and serves as a consultant for Abbott, Chambercardio, and CVS Health (outside the submitted work). Issa J. Dahabreh reports being principal investigator of a research agreement between Harvard University and Sanofi and receiving consulting fees from Moderna, both unrelated to the content of this article. The remaining authors have no disclosures to report.

Supporting information

Tables S1–S11

Figures S1–S2