Relation of Neighborhood Disadvantage to Heart Failure Symptoms and Hospitalizations

Abstract

Residence in socioeconomically deprived neighborhoods may influence patient’s health-related behaviors and overall health. We evaluated the association of neighborhood disadvantage on heart failure (HF) symptom burden and hospitalization rates. We characterized neighborhood deprivation in 359 HF subjects (age 56 ± 13 years, 52% black) in metropolitan Atlanta using the Area Deprivation Index (ADI). ANOVA was used to compare HF symptoms measured using the Kansas City Cardiomyopathy Questionnaire (KCCQ), and HF Self-Care Index across ADI tertiles. Zero-inflated Poisson regression was used to compare rates of recurrent HF hospitalization (HFH) across ADI tertiles. Subjects living in more deprived neighborhoods were more likely to be black, have Medicare or Medicaid insurance, and have a lower ejection fraction than those living in less deprived neighborhoods (all P≤0.005). Subjects in more deprived neighborhoods had more severe HF symptoms (P< 0.001), but there was no difference in HF Self-Care Index scores across ADI tertiles. Subjects living in more deprived neighborhoods had a higher odds of being hospitalized for HF than subjects in less deprived neighborhoods. Once subjects had experienced a HFH, however, the association between ADI and the risk of recurrent HFH varied by racial group. Among whites, increasing ADI was associated with a marginally decreased risk of recurrent HFH, while there was no association between ADI and recurrent HFH among blacks. In conclusion, individuals with HF living in more deprived neighborhoods have greater symptom burden and are more likely to experience a HFH than those living in less deprived neighborhoods.

INTRODUCTION

Heart failure (HF) is the leading cause of cardiovascular (CV) hospitalization and the fifth leading cause of hospitalization in the US overall. ^{1, 2} Patients who have been hospitalized are at particularly high risk for poor outcomes; ~25% of are readmitted within 30 days of an index HF hospitalization, with a one-year mortality rate as high as 30%.³ Current models that attempt to predict patients’ risk for death and rehospitalizations primarily employ available clinical and administrative data, but are less likely to incorporate aspects of patients’ post-discharge environment and neighborhood which may also influence the risk for HF outcomes.^{3, 4} Residential neighborhood is increasingly recognized as an independent predictor of various aspects of an individual’s health.⁵ Persons living in more impoverished and racially-segregated neighborhoods have an increased risk of HF hospitalization and readmissions.^{6, 7} While many studies have examined the association of quantitative, census-driven neighborhood deprivation with HF risk ^{8, 9}, few studies to date have examined the mechanism by which neighborhood deprivation impacts HF risk. We hypothesize that subjects who live in more deprived neighborhoods may struggle with aspects of HF self-care due to fewer neighborhood resources. We evaluated the association of neighborhood disadvantage on HF symptom burden, self-care, and frequency of hospitalization.

METHODS

We pooled individual-level data from the Atlanta Cardiomyopathy Consortium and the Metabolomics, Oxidative Stress and Vascular Function study, two prospective cohort studies which recruited outpatients with prevalent HF. The Atlanta Cardiomyopathy Consortium (TACC) enrolled 336 patients from the HF clinics at three Emory University-affiliated hospitals from 2007 to 2011, according to the inclusion and exclusion criteria previously described.¹⁰ The Metabolomics, Oxidative Stress, and Vascular Function study (MOV) is a follow-up study to the Atlanta Cardiomyopathy Consortium and enrolled 205 patients from 2015 to 2019 according to the same inclusion and exclusion criteria.⁶ All patients provided informed consent, and both studies were approved by the Emory Institutional Review Board.

All patients were followed for clinical outcomes from the baseline visit to the last date of follow-up for the study (May 2012 for TACC, October 2019 for MOV). Data on mortality was collected through review of electronic medical records, consultation with family members, and social security death index query. Data on HF-specific hospitalization was collected through review of electronic medical records, including outpatient notes from any admission to an Emory-affiliated or outside hospital, and direct interview at follow-up appointments.

The Self-care of HF Index¹¹ was assessed in both cohorts at their baseline visit. The score evaluates metrics such as daily weights, adherence to a low-salt diet, regular physical activity, general health maintenance, and ability to recognize and self-treat symptoms of HF. HF self-care subscales and cumulative scores for the combined dataset were calculated using the appropriate transformation factors as previously described.¹¹ Imputation to the mean was used for participants with randomly distributed missing datapoints. Of the TACC data, 54 of 275 (16.4%) participants were missing at least one datapoint, for a total of 231 (4.4%) of all values.

The Kansas City Cardiomyopathy Questionnaire (KCCQ) was self-administered in both cohorts at their baseline visit. The 23-item instrument quantifies physical function, symptoms (frequency, severity and recent change), social function, self-efficacy and knowledge, and quality of life into an overall summary score (KCCQ-OS).¹² Scores are transformed to a range of 0–100, in which higher scores reflect better HF-specific health status, with a 5-point difference considered clinically meaningful.¹³

The Area Deprivation Index (ADI) is comprised of seventeen components collected from the 5-year American Community Survey of the United States Census, that reflect the education, employment, income, and housing quality of the defined geographic area to define overall neighborhood deprivation.¹⁴ Scores range from 0–100, with a higher score indicating greater deprivation of that geographic area. ADI for all block groups in the state of Georgia were downloaded using the most updated data available from the 2015 American Community surveys. To assign each participant an ADI score, participant addresses were geocoded to street level accuracy using the US Census Bureau’s geocoder. Of the 529 total participants, 140 were excluded (91 had missing address data, and 49 failed to geocode after attempting manual correction by entering their address into a publicly available mapping software to confirm correct spelling, punctuation, and address modifiers. Block groups were chosen as proxies for neighborhoods. Block groups represent smaller geographic units within census tracts, and therefore reflect the homogenous population characteristics, socioeconomic status and living conditions of the larger census tract they are encompassed within.¹⁵ Block groups are also defined using major roads, waterways, and transit lines, creating borders defined by pedestrian limitations that make the block group an adequate proxy for neighborhood.¹⁵

Data are presented as mean (standard deviation) for normally distributed continuous variables, median (interquartile range) for non-normally distributed continuous variables, or N (%) for categorical variables as appropriate. ADI was examined as a continuous variable, and as a categorical variable based on tertiles of the national distribution of ADI (least deprived [0–33], middle [34–66], most deprived [67–100]). Baseline characteristics of the study cohort were compared across ADI groups using ANOVA for normally distributed continuous variables, the Kruskall-Wallis test for non-normally distributed continuous variables, and chi-square analysis for categorical variables. As 63% of participants in our dataset did not experience a hospitalization during follow-up, zero-inflated Poisson (ZIP) regression models were estimated to examine the association between ADI and the risk of HF hospitalizations. The ZIP model (R packages ggplot2 and pscl) accounts for the population of patients with zero hospitalizations separately, by including both a Poisson regression model for modeling counts (number of HF hospitalizations) and a logistic regression model within the Poisson model for modeling the probability of zero HF hospitalizations. Multivariable ZIP models were adjusted for the following covariates: race, age, gender, insurance status, history of hypertension, body mass index, estimated glomerular filtration rate, ejection fraction, B-type natriuretic peptide, chronic kidney disease, and KCCQ score, with an offset for follow-up time. Analyses were conducted in the total cohort, and also stratified by race (whites and blacks) conditional on experiencing at least one HF hospitalization with appropriate testing for a statistical interaction of race with ADI. All analyses were carried out in SPSS version 26.0.0.0 and R version 1.2.5001. All p values are two-tailed with a significance threshold of <0.05.

RESULTS

The baseline characteristics of the study cohort are shown in Table 1. Compared to participants in the least deprived tertile, participants living in the most deprived ADI group were more likely to be black (Figure 1). Additionally, participants living in the most deprived ADI group were more likely to have Medicare or Medicaid insurance, were less likely to have completed a college or graduate education, had a higher BMI, and a lower EF. Differences in baseline characteristics of the cohort according to racial group are shown in Supplemental Table 1.

Table 1.

Baseline characteristics of the study cohort according to neighborhood deprivation.

	Neighborhood deprivation
Variable	Least (N=69)	Middle (N=150)	Most (N=140)	P-value
Age (years)	60.2 ± 11.8	54.2 ± 13.5	55.3 ± 12.1	0.005
Men	42 (61%)	83 (55%)	78 (56%)	0.7
White	55 (80%)	77 (51 %)	35 (25%)	<0.001
Black	11 (16%)	70 (47%)	104 (74 %)
Other	3 (4%)	3 (2%)	1 (1%)
Insurance type				0.003
• Commercial	39 (57%)	73 (49%)	38 (27%)
• Medicare	25 (36%)	57 (38%)	70 (50%)
• Medicaid	2 (3%)	8 (5%)	18 (13%)
• VA	1 (1%)	5 (3%)	3 (2%)
• None	1 (1%)	5 (3%)	8 (6%)
Education				0.002
• College or greater	59 (86%)	94 (63%)	81(58%)
• Some college	9 (13%)	52 (35%)	56 (40%)
• High school or less	1 (1%)	4 (3%)	3 (2%)
BMI (kg/m2)	29.9 ± 7.7	31.2 ± 8.1	33.4 ± 8.9	0.008
Diabetes mellitus	17 (25%)	49 (33%)	52 (37%)	0.2
Hypertension	41 (59%)	90 (60%)	96 (69%)	0.2
EF (%)	31.4 ± 13.4	25.8 ± 13.0	25.6 ± 12.7	0.005
HFpEF	9 (13%)	11 (7%)	10 (7%)	0.275
NYHA class				0.1
• 1	6 (9%)	8 (5%)	12 (9%)
• 2	38 (57%)	84 (57%)	59 (43%)
• 3–4	23 (34%)	55 (37%)	66 (48%)
BNP (pg/mL)	150 (42, 584)	211 (81, 647)	199 (78, 620)	0.6
Creatinine (mg/dL)	1.2 ± 0.7	1.2 ± 0.4	1.3 ± 0.6	0.1
CKD stage				0.9
• 1–2	45 (65%)	97 (65%)	84 (60%)
• 3	21 (30%)	48 (32%)	49 (35%)
• 4–5	3 (4%)	5 (3%)	7 (5%)
Systolic BP (mm Hg)	114.8 ± 21.4	113.6 ± 18.5	113.4 ± 18.4	0.9
Diastolic BP (mm Hg)	71.1 ± 13.3	69.2 ± 11.2	69.1 ± 12.1	0.5
Medical therapy
• ACEi/ARB	46 (79%)	117 (91%)	87 (83%)	0.07
• ARNI	3 (4%)	10 (7%)	10 (7%)	0.4
• MRA	29 (42%)	82 (55%)	80 (58%)	0.1
• Beta-blockers	64 (93%)	145 (97%)	130 (93%)	0.3
• Hydralazine	5 (8%)	31 (21%)	34 (25%)	0.012
• Oral nitrates	18 (27%)	37 (26%)	35 (26%)	0.9
• Diuretics	54 (78%)	132 (88%)	128 (91%)	0.013
ICD or CRT-D	24 (83%)	48 (72%)	63 (80%)	0.4

ACEi, angiotensin converting enzyme inhibitor; ARB, angiotensin receptor blocker; BMI, body mass index. BNP, B-type natriuretic peptide; CKD, chronic kidney disease; CRT-D, cardiac resynchronization therapy-defibrillator; EF, ejection fraction; HFpEF, heart failure with preserved ejection fraction; ICD, implantable cardioverter defibrillator; MRA, mineralocorticoid antagonist; NYHA, New York Heart Association.

Figure 1.

Figure 1 illustrates the distribution of white (red) and black (turquoise) participants within each unit of the area deprivation index. For example, <10% of Black participants resided in census tracts with ADI score 0–10 (least deprived), while ~14%% of White participants resided in census tracts with ADI score 90–100 (most deprived). Because the residential distribution of black participants is overlaid on the distribution of white participants, the third color coding (taupe) represents census tracts that have a mixed racial distribution.

The burden of participants’ HF symptoms and HF self-care are shown in Table 2. There was no difference in the cumulative HF self-care score according to ADI group. The KCCQ overall summary score demonstrated that participants living in the most deprived ADI group had more severe HF symptoms than those living in either the middle or least deprived ADI group. Specifically, the physical limitation, symptom frequency, symptom burden, total symptom scores, and social limitation sub-scale scores were more severe for participants in the most deprived ADI group compared to those living in both the middle and least deprived ADI groups.

Table 2.

Comparison of HF self-care index and KCCQ scores according to neighborhood deprivation.

	Neighborhood deprivation
Variable	Least (N=69)	Middle (N=150)	Most (N=140)	P-value
HF Self-care Index
- Self-care maintenance subscore	75 (70, 85)	75 (60, 80)	75 (61, 85)	0.1
- Self-care management subscore	73 (65, 85)	75 (60, 83)	75 (59, 85)	0.2
- Self-care self-confidence subscore	69 (56, 81)	75 (63, 88)	75 (63, 94)	0.09
- Cumulative self-care score	214 (193, 239)	223 (185, 243)	222 (190, 252)	0.4
KCCQ
- Physical limitation	83 (58, 96)	71 (50, 88)	59 (42, 79)	<0.001
- Symptom stability	50 (50, 50)	50 (50, 50)	50 (50, 75)	0.3
- Symptom frequency	82 (63, 96)	75 (54, 92)	70 (38, 88)	0.003
- Symptom burden	83 (67, 100)	83 (58, 92)	75 (50, 92)	0.028
- Total symptom score	82 (67, 96)	77 (56, 92)	70 (44, 87)	0.007
- Self-efficacy	88 (75, 100)	88 (75, 100)	100 (75, 100)	0.9
- Quality of life	67 (42, 83)	67 (38, 83)	50 (33, 75)	0.08
- Social limitation	75 (44, 96)	67 (33, 88)	50 (25, 88)	0.024
- Overall summary score	75 (54, 89)	68 (45, 83)	59 (36, 79)	0.002

KCCQ, Kansas City Cardiomyopathy Questionnaire. Scores are transformed to a range of 0–100, in which higher scores reflect better HF-specific health status, with a 5-point difference being clinically meaningful.

During a median follow-up of 1,062 (IQR 646, 1,420) days, 144 (40%) participants had ≥1 HF hospitalization, and 86 (24%) had ≥2 HF hospitalizations (Figure 2). The crude frequency of HF hospitalization was higher in participants who lived in more deprived neighborhoods (44.2 per 100 person-years for participants in the most deprived ADI group vs. 40.2 per 100 person-years for participants in the middle ADI group vs. 22.7 per 100 person-years for participants in the least deprived ADI group). After adjustment for covariates, living in the most deprived ADI group was associated with lower odds of never experiencing a HF hospitalization (middle ADI group odds ratio [OR] = 0.39, 95% confidence interval [CI] 0.17 – 0.88, P = 0.024; most deprived ADI group OR = 0.36, 95% CI 0.16 – 0.84, P = 0.018). In the count portion of the ZIP model, covariates associated with an increased risk of recurrent HF hospitalizations included black race (rate ratio [RR] 1.42, 95% CI 1.03 – 1.95; P=0.032), lower education (RR 1.75, 95% CI 1.37 – 2.25; P<0.0001), male sex (RR 1.49, 95% CI 1.13 – 1.94; P=0.004), CKD stage 3 (RR 1.49, 95% CI 1.14 – 1.94; P=0.003).

Figure 2.

Figure 2 is a stacked frequency plot demonstrating the frequency of total HF hospitalizations according to racial group.

Amongst those who experienced at least one HF hospitalization, the risk of recurrent hospitalizations varied by race (P=0.013 for race*ADI interaction). Among blacks, there was no association between ADI and recurrent HF hospitalization among blacks (Table 3). Among whites, increasing ADI was associated with a marginally decreased risk of recurrent HF hospitalization. Of the remaining risk factors for HF hospitalization, increasing KCCQ score and the presence of HF with preserved EF were associated with a reduced risk of recurrent hospitalization, while male sex, less education, impaired renal function, and higher BNP were associated with an increased risk of recurrent HF hospitalization in blacks and whites.

Table 3.

Multivariable predictors of recurrent heart failure hospitalizations (from the count portion of the zero-inflated Poisson regression model).

	Whites		Blacks
	RR (95% CI)	P-value	RR (95% CI)	P-value
ADI (per 1-pt increase)	0.98 (0.97 – 0.99)	<0.001	1.00 (0.99 – 1.01)	0.5
Age (per 1-yr increase)	1.00 (0.99 – 1.02)	0.7	0.98 (0.97 – 0.99)	<0.001
Male	2.21 (1.34 – 3.57)	0.001	1.80 (1.36 – 2.39)	<0.001
Education
• College graduate or higher	REFERENCE		REFERENCE
• Some college	3.24 (2.10 – 5.00)	<0.001	1.57 (1.19 – 2.07)	0.001
• High school or less	0.41 (0.05 – 3.16)	0.4	0.97 (0.24 – 3.93)	0.9
Hypertension	2.03 (1.24 – 3.33)	0.005	0.71 (0.52 – 0.96)	0.028
CKD stage
• Stage 1–2	REFERENCE		REFERENCE
• Stage 3	3.10 (2.03 – 4.74)	<0.001	1.57 (1.19 – 2.07)	0.001
• Stage 4–5	4.18 (1.81 – 9.68)	0.001	0.97 (0.24 – 3.93)	0.9
HFpEF	0.13 (0.05 – 0.33)	<0.001	0.07 (0.01 – 0.50)	0.008
BMI (per kg/m2)	1.01 (0.98 – 1.05)	0.4	1.03 (1.01 – 1.04)	<0.001
KCCQ-OS (per 1-pt increase)	0.98 (0.97 – 0.99)	<0.001	0.99 (0.99 – 1.00)	0.08
BNP (per SD increase)	1.40 (1.17 – 1.67)	<0.001	1.29 (1.14 – 1.45)	<0.001

The rate ratios (RR) and 95% confidence intervals (CI) are conditional on having suffered at least one hospitalization, and thus only include participants who had a hospitalization during the course of follow-up. These results are based on a Poisson regression model, and thus describe the incidence of recurrent events (first and subsequent HF hospitalizations).ADI, area deprivation index; BMI, body mass index; BNP, B-type natriuretic peptide; CKD, chronic kidney disease; HFpEF, heart failure with preserved ejection fraction; KCCQ-OS, Kansas City Cardiomyopathy Questionnaire overall summary score

DISCUSSION

In this prospective cohort analysis, we prospectively examined the association of neighborhood deprivation on hospitalization risk in HF patients living in the state of Georgia. Our main findings were that: (1) participants living in more socioeconomically deprived neighborhoods reported a greater HF symptom burden, but did not report differences in HF self-care; (2) participants living in more socioeconomically deprived neighborhoods had a higher risk of having at least one HF hospitalization compared to those living in less deprived neighborhoods; and 3) the association between neighborhood deprivation and recurrent hospitalization differed between racial groups, whereby increasing neighborhood deprivation was associated with a marginally reduced risk of recurrent HF hospitalization among whites, but we did not identify an association between neighborhood deprivation and recurrent HF hospitalization among blacks.

Prior studies have established a strong and consistent relationship between neighborhood of residence and individual health. Neighborhood deprivation is inversely related to health-promoting environmental features, including the availability of grocery stores to acquire fresh and healthy foods, as well as green space and sidewalks to encourage physical activity, are likely to impact on lifestyle behaviors that influence overall CVD risk. Patients with HF are encouraged to restrict dietary sodium to reduce symptoms of congestion, engage in regular physical activity to improve functional status, and to utilize social support networks to reduce the risk of hospitalization.¹⁶ In our cohort, self-reported HF self-care scores did not differ based on neighborhood deprivation. Despite this, there was a strong association between increasing neighborhood deprivation and HF symptom burden, as quantified by the KCCQ. Since all participants in our study were being followed in an academic HF clinic, the higher burden of symptoms can not be attributed to variations in access to healthcare or specialized HF education. Still, subjects who lived in more deprived neighborhoods may have had reduced access to cardiac rehabilitation or other facilities to participate in exercise training, which improves symptom burden in HF patients.¹⁷ Thus, future investigations should examine novel aspects of the built environment at the neighborhood level that may influence symptom burden.

Fewer studies have examined the association between neighborhood deprivation and the risk for clinical events in patients with established HF. Prior analyses have noted that neighborhood socioeconomic disadvantage predicts increased risk for hospital readmissions.^{7, 18} Our study is the first to utilize the 17-component census-derived ADI to evaluate the risk of HF hospitalization among patients with prevalent HF. Although we found that residence in more deprived neighborhoods was associated with a higher odds of ever experiencing a HF hospitalization, the association between ADI and risk for recurrent HF hospitalization differed by racial group in our cohort among those who experienced a hospitalization. Blacks had a higher overall risk of recurrent hospitalization that was not associated with the ADI. However, in whites, there was a marginal but paradoxical decrease in the risk of recurrent HF hospitalization with increasing neighborhood deprivation. We hypothesize that this may be due to the fact that, within our study population, most of the black participants living in more deprived neighborhoods were clustered in urban settings, while white participants living in more deprived neighborhoods were clustered in rural settings. It is possible that participants in rural settings may have had hospitalizations outside of our healthcare system, and thus the number of hospitalizations was underestimated. It is also possible that once a patient has reached the stage of illness requiring a HF hospitalization, other clinical factors may be more relevant than neighborhood deprivation. While it has been established that place of residence influences multiple dimensions of overall health, it is likely that additional variables that were unmeasured in this study may contribute to increased morbidity among those living in deprived neighborhoods.

There are several important clinical implications to our findings. Since the introduction of the Hospital Readmissions Reduction Program (HRRP), it is even more important to note aspects of the post-discharge environment that may impact the risk for rehospitalization, and subject safety-net hospitals to higher penalties because of excess readmissions.¹⁹ As patients with HF have a high risk of repeat hospitalizations, the examination of factors that affect the overall burden of hospitalizations may be more clinically meaningful than analyses that only examine time to first event, since recurrent hospitalizations can adversely impact patients’ quality of life and identifies a subset of patients who are at higher risk for death. Prior data shows that patients who live in more impoverished neighborhoods have a higher risk for recurrent all-cause and HF-specific hospitalizations.^{6, 20} Black patients had a higher risk for recurrent hospitalizations in our analysis, but were also more likely to live in deprived neighborhoods. Future studies must comprehensively examine the clinical factors as well as the sociodemographic factors that put patients at much higher risk for repeat hospitalizations.

The current study has several limitations. First, the participants in this study were established patients at a university-based advanced HF clinic, and the results may not be generalizable to patients in other settings. Prior studies have shown that patients followed in specialty HF clinics, whether academic or community, have a reduced risk of all-cause and 30-day readmission, and decreased overall mortality compared to similar patients with HF.^{21, 22} Second, there were 140 participants in the original cohort for whom ADI classification was not possible due to either missing or partial address data, or addresses were outside the state of Georgia. Finally, the current participant address may not reflect the length of residence (and exposure to area deprivation), or childhood address and exposures on adult outcomes. However, prior studies suggest that individuals tend to move to neighborhoods of similar SES even when they change address, which may lessen the impact of this shortcoming.²³

In conclusion, we found that individuals with HF living in more deprived neighborhoods had greater symptom burden and, for those not yet hospitalized, increased HF hospitalization risk. Association can not prove causation, and these results highlight the complexity of examining an individual’s neighborhood of residence, individual health, and outcomes, particularly among patient populations as heterogeneous as those with HF.

Figure 3.

Figure 3 depicts that among those who experienced at least one HF hospitalizations, increasing ADI was associated with a decreased risk of recurrent HF hospitalization among whites, while there was no association between ADI and recurrent HF hospitalization among blacks.