Social determinants of health and hospitalization risk in heart failure: A Retrospective analysis

Abstract

Background

Heart failure affects over 6.5 million Americans and accounts for substantial healthcare utilization, with social determinants of health (SDOH) increasingly recognized as critical factors influencing patient outcomes. The present study aims to examine the relationships among specific SDOH (food insecurity risk, financial resource strain, and medical transportation needs) with subsequent hospitalization risk among patients with heart failure.

Methods

This retrospective cross-sectional study analyzed electronic health record data from the Arkansas Clinical Data Repository for 2494 patients with heart failure hospitalized between January 2014 and December 2023 at the University of Arkansas for Medical Sciences. Quantile regression on the median was used to analyze the data.

Results

Study population descriptive results demonstrated moderate hospitalization risk (average score 40.3). The population was predominantly older adults (mean age ∼ 65), female (55.8%), non-Hispanic (98.3%) and White (54.3%). In fully adjusted models, all three social determinants remained significant predictors of increased subsequent hospitalization risk: food insecurity was associated with a 9.37-point increase (95% CI: 4.83–13.90, p < 0.01), high financial resource strain with a 13.13-point increase (95% CI: 8.50–17.76, p < 0.01), and medical transportation needs with a 13.23-point increase (95% CI: 7.55–18.89, p < 0.01).

Conclusions

Food insecurity risk, financial resource strain, and medical transportation needs are associated with increased future hospitalization risk among patients with heart failure. These findings support the integration of social determinant screening into clinical risk stratification and the development of targeted interventions addressing these modifiable social factors to potentially reduce healthcare utilization and improve outcomes for vulnerable populations with heart failure.

Highlights

• •Food insecurity, financial strain, and transportation needs were linked to hospitalization risk.
• •Financial strain and transportation barriers showed the largest effects.
• •Findings support social screening in risk stratification tools for HF patients.

1. Introduction

Heart failure (HF) represents a growing public health crisis, affecting more than 6.5 million Americans with an estimated 550,000 new cases diagnosed annually [1]. Despite advances in medical therapy and quality improvement initiatives, HF continues to impose a substantial burden on healthcare systems, accounting for over 1.9 million hospitalizations nationwide and costing the United States more than $31 billion annually [1], [2]. This burden is particularly pronounced in states like Arkansas (AR), which ranks 48th nationally in overall health In AR nearly half of the adult population has high blood pressure (with 29% having uncontrolled hypertension compared to 16% nationally) and cardiovascular disease mortality rates exceed national averages (231.0 per 100,000 compared to the national rate of 173.8 per 100,000) [3], [4]. Recognizing this mounting health crisis, the current executive administration has emphasized the critical need to redirect our focus toward understanding and drastically lowering chronic disease rates, through fresh thinking on healthy lifestyles and environmental impacts. The complexity of HF management extends beyond traditional clinical parameters, as evidence demonstrates that social determinants of health (SDOH) significantly influence patient outcomes and healthcare utilization patterns, particularly in rural and underserved populations [4]. Rethinking of the role of SDOH in HF management may illuminate unforeseen pathways in clinical care to improve morality and reduce chronic disease burden.

Social determinants of health, broadly defined as the conditions in which people are born, live, work, develop, and age, have emerged as critical factors affecting HF prognosis and care delivery [3]. Research indicates that SDOH accounts for 80–90% of modifiable contributors to health outcomes in chronic diseases, including heart failure [4]. Furthermore, patients experiencing adverse SDOH are less able to access care and more likely to experience poor HF outcomes over time [1].

Among the various social determinants, food insecurity has received increasing attention as an important risk factor for hospital utilization among patients with HF. Food insecurity, defined as limited or uncertain access to nutritionally adequate food due to insufficient financial resources, affects one in eight households in the United States each year. Arkansas residents experience particularly high rates in the rural Delta counties (i.e., a region in eastern Arkansas that extends to the Mississippi Delta) where poverty and limited access to nutritious food sources compound cardiovascular risks [5], [6]. Recent studies demonstrate that food insecurity impacts residents in two ways: 1) healthcare utilization patterns (e.g., emergency department visits and hospitalizations); and 2) increased hospital service usage [7], [8]. This impact extends beyond initial hospitalizations to readmission patterns, with food-insecure patients having significantly higher rates of inpatient readmissions, suggesting that food insecurity may contribute to healthcare service demand through mechanisms such as delayed preventive care, medication non-adherence, and nutritional deficiencies that exacerbate existing health conditions [8].

Financial resource strain represents another critical social determinant that significantly impacts patients with HF and healthcare utilization. Patients with cardiovascular-related chronic diseases (like HF) who report financial strain have increased rates of hospitalization, mortality, and higher hospital costs compared to those without financial strains [9]. This is best explained by the trade-off phenomenon, where patients must choose between purchasing medications and meeting basic needs such as food and housing. As a result, this financial strain of the trade-off phenomenon can lead to medication non-adherence and subsequent clinical deterioration [10]. This financial burden is particularly pronounced among patients with chronic conditions (like HF) who face ongoing expenses for medications, medical appointments, and disease management. Given that Arkansas has the fourth-highest poverty rate in the United States at 17% [4], [11] financial resource strain in this population may disproportionately affect cardiovascular outcomes.

Medical transportation barriers (i.e., being unable to drive to treatment or seek preventive care) constitute an additional but often overlooked social determinant that can significantly influence hospital utilization. Studies indicate that transportation challenges affect approximately 3.6 million Americans annually, preventing them from accessing necessary medical appointments and treatments [12]. This barrier disproportionately impacts vulnerable populations, including elderly patients, individuals with disabilities, and those living in rural or underserved communities where public transportation options are limited or nonexistent [13]. The relationship between transportation access and hospital utilization patterns demonstrates that patients with reliable transportation are more likely to engage in preventive care and routine medical management, ultimately reducing emergency department visits and costly inpatient admissions [11]. Conversely, transportation barriers often force patients to delay care until conditions become severe enough to warrant emergency intervention, leading to increased healthcare costs and poorer clinical outcomes [11]. Transportation barriers are especially problematic for patients with HF given the frequent medication monitoring, routine clinical assessments, and recurring physical tests (i.e., blood draws, chest X-rays, stress tests). These barriers can result in missed or delayed appointments, ultimately leading to poorer management of HF and possibly mortality.

The relationship between SDOH and HF outcomes is complex and multifaceted, specifically for populations in states with high rates of rural residence and socioeconomic disadvantage. Studies examining post-hospitalization outcomes have demonstrated that patients with HF who experience adverse SDOH have significantly increased risk of 90-day mortality, with hazard ratios ranging from 2.57 to 2.89 depending on the burden of SDOH [14]. Furthermore, SDOH influences not only clinical outcomes but also healthcare utilization patterns. Evidence shows significant associations between SDOH and hospital length of stay, readmission rates, and overall care quality [15], [16], [17]. These relationships may be particularly pronounced in regions like the rural Arkansas Delta, where counties such as Phillips County rank among the lowest in the nation for health outcomes and face an entanglement of factors including poverty, limited healthcare access, and social isolation that disproportionately impact cardiovascular health like HF [18].

Despite the growing recognition of SDOH's importance in HF outcomes, significant knowledge gaps remain regarding how specific individual social determinants independently influence future hospitalization risk, particularly in underserved populations. While previous research has demonstrated associations between individual social factors and HF outcomes, there is limited understanding of how food insecurity risk, financial resource strain, and medical transportation needs specifically impact the risk of hospitalization in the subsequent year among patients with established HF (a vital maker for mortality). Furthermore, most existing studies have focused on broad social determinant categories or composite measures, leaving uncertainty about the relative contribution of specific, modifiable social factors to future healthcare utilization. This limitation has significant implications for healthcare delivery and restricts our ability to develop sophisticated risk stratification models that integrate social determinants with clinical variables.

The present study addresses these knowledge gaps by examining the relationship between food insecurity risk, financial resource strain, and medical transportation needs with the outcome of hospitalization risk in the following year among patients with heart failure. These three social determinants were selected based on their established associations with cardiovascular outcomes, their potential modifiability through healthcare system interventions, and their particular relevance to the Arkansas population. Under this research question, we hypothesized that each of these social determinants would be independently associated with increased hospitalization risk, even after controlling for relevant clinical and demographic factors. By quantifying these relationships, this research seeks to inform evidence-based approaches to incorporating social determinants into clinical risk stratification and targeted interventions for heart failure patients.

2. Methods and materials

2.1. Study design and setting

The University of Arkansas for Medical Sciences is a 535-bed hospital that includes 431 adult beds, 64 newborn bassinets, and 40 psychiatry beds [19]. The University of Arkansas for Medical Sciences, as the state's only academic medical center, serves as a vital healthcare provider and provides an important opportunity to study SDOH relationships in a population that may be at particularly high risk for adverse social determinant impacts.

This retrospective, cross-sectional study analyzed deidentified electronic health record data extracted from the Arkansas Clinical Data Repository [20]. Since 2011, the AR-CDR research data warehouse has served as a unified and secure data source for clinical and translational research. This repository contains data pulled from both the current EPIC electronic health record system and older legacy EHR systems [21]. The study included all hospitalizations from the University of Arkansas for Medical Sciences. A hospitalization was defined as an inpatient admission for a continuous stay. The study data frame only included hospitalizations who had heart failure hospitalization between January 1, 2014, and December 31, 2023. This 10-year period was selected because of data availability. All patients in the dataset had a diagnosis of congestive heart failure, as characterized by left ventricular ejection fraction via transthoracic echocardiogram. Patient data included pseudonym identifiers, demographics, clinical diagnoses, medical histories, and comorbidities based on each patient's International Classification of Disease Codes Version 10.

2.2. Study variables

2.2.1. Independent variables: SDOH questionnaire

Independent variables were captured via SDOH questionnaire administered to patients during the admission process. This included 24 questions with answers on a Likert scale. The three main variables for our study are defined as follows:

2.2.1.1. Food insecurity risk

Food Insecurity Risk was defined by the survey question “Within the past 12 months, you worried that your food would run out before you got the money to buy more.” Response options included: never true, sometimes true, often true, or patient refused. Responses of sometimes true and often true were collapsed to yes, while never true was identified as no.

2.2.1.2. Financial resource strain

Financial Resource Strain was defined by the survey question “How hard is it for you to pay for the very basics like food, housing, medical care, and heating?” Response options included: very hard, hard, somewhat hard, not very hard, not hard at all, or patient refused. Responses were categorized as high risk (very hard), medium risk (hard and somewhat hard), and low risk (not very hard and not hard at all). This categorization was implemented for three reasons: (1) to ensure adequate cell sizes for statistical analysis, (2) to enhance clinical interpretability by grouping responses that represent similar functional limitations and would likely trigger comparable clinical interventions; and (3) to align with established clinical practice, where difficulty levels are often conceptualized on a three-tier scale (severe, moderate, minimal) for care planning and risk stratification purposes [22].

2.2.1.3. Medical transportation needs

Medical Transportation Needs was defined by the survey question “In the past 12 months, has lack of transportation kept you from medical appointments or from getting medications?” Response options included: yes, no, or patient refused.

2.2.2. Dependent variable

The dependent variable was subsequent one-year hospitalization risk. We examined prospective risk rather than concurrent events for three reasons. First, prediction enables proactive identification of high-risk patients before hospitalization occurs, creating opportunities for preventive intervention. Second, this forward-looking approach aligns with contemporary healthcare delivery models that prioritize risk stratification and preventive care management. Finally, it utilizes a validated prediction tool already integrated into clinical workflows for real-world decision-making.

Subsequent Hospitalization Risk was derived using Epic's “Risk of Hospital Admission or ED Visit Model,” a proprietary algorithm that predicts a patient's risk of needing either a hospital admission or an emergency department (ED) visit within the following year [23]. This model utilizes various factors including Medicare and Medicaid status, high-risk diagnosis, past ED and hospital visits, and twelve high-risk diagnoses [24]. The score ranges from 1 to 99, with a greater score indicating great risk of hospitalization in the following year. The score is integrated into Epic's EHR system to help clinical staff identify patients who might benefit from targeted interventions [23], [24].

2.2.3. Covariates

Confounding variables included number of registered nurses (continuous), Van Walraven Elixhauser Comorbidity Score (continuous), age (continuous, in years), gender (male, female), race (White, Black, Asian, Native Hawaiian/Pacific Islander, American Indian/Alaska Native, Other), number of care managers on the team (zero care managers present, one care manager present, two care managers present, three care managers present), ethnicity (Hispanic, Non-Hispanic), rurality (rural, non-rural), insurance type (Medicare, Medicaid, Private, Self-Pay, Other), and presence of a social worker (present, absent). The Van Walraven Elixhauser Comorbidity Score ranges from negative twelve to positive eighty-nine, with the lowest quartile (−12 to 7) representing patients with the best overall health, and patients with scores ranging from eight to eighty-nine indicating patients with poor overall health [25], [26]. This scoring system has been previously validated through international studies in patients with chronic conditions including congestive HF [25], [27], [28], [29].

2.3. Statistical Analysis

A complete case analysis approach was employed. Patients with unknown or declined responses to predictor variables were coded as missing and removed from the analysis (n = 478). Additionally, patients with missing data on confounders (n = 20) were excluded from the analyses. The total number of removed cases was 498 observations, resulting in approximately 19% of the original dataset being excluded from the final analysis.

Prior to statistical analysis, the distribution of subsequent hospitalization risk scores was assessed through visual and statistical methods. Histograms and quantile-quantile plots revealed a right-skewed distribution pattern. Formal normality testing using the Shapiro-Wilk and Anderson-Darling tests confirmed significant departures from normality (p < 0.05). To address this non-normal distribution, we applied several transformation techniques, including logarithmic, square root, and reciprocal transformations, to achieve approximate normality for subsequent parametric analyses. However, these transformations failed to adequately normalize the data distribution. Due to the failure to meet normal distribution assumptions and the robustness of quantile regression to distributional assumptions, quantile regression on the median (50th percentile) was employed as the analytical approach [30]. Quantile regression is particularly well-suited for analyzing data with skewed distributions and provides estimates that are less sensitive to outliers compared to traditional ordinary least squares regression [31], [32]. Furthermore, the use of quantile regression accommodates the characteristic non-normal, right-skewed distributions of key clinical variables (i.e., ejection fraction, length of stay, costs) and provides robust estimates unaffected by the extreme values commonly observed in patients with acute illness that would otherwise bias traditional mean-based regression analyses. The Markov Chain Marginal Bootstrap resampling method was used to construct 95% confidence intervals for quantile regression coefficients and provide a robust method to handle non-normal errors [33]. All analyses were conducted using SAS 9.4.

The analyses were conducted in three stages: an unadjusted model, a model adjusted for non-modifiable patient factors like age and race, and a fully adjusted model including both non-modifiable and modifiable factors. This staged approach allows for the assessment of how much of the observed associations can be attributed to inherent patient characteristics versus potentially modifiable clinical and behavioral factors, thereby informing targeted intervention strategies.

2.4. Ethical considerations

The study protocol (#276211) was reviewed and approved by the Ethics Committee of the University of Arkansas for Medical Sciences' Institutional Review Board. This study was conducted in accordance with all applicable government regulations and institutional research policies and procedures at the University of Arkansas for Medical Sciences. The study team only had access to de-identified data on patients and clinicians. This study presented only minimal risk for loss of confidentiality, and a waiver of documentation of the informed consent process was granted by the Ethics Committee, which is the standard for retrospective analyses involving de-identified data.

3. Results

Table 1 demonstrates the distribution of the study population. The study cohort included 2494 patients, with a mean hospitalization risk score of 40.3 (SD = 31.5) and a median score of 33 (IQR = 59.0). The average patient age was approximately 65 years. Demographically, the population was predominantly female (55.8%), with the vast majority identifying as not Hispanic or Latino (98.3%). The racial composition included 54.3% White, 41.9% Black, 2.3% Other Race, 0.6% American Indian/Alaska Native, 0.5% Asian, and 0.08% Native Hawaiian/Pacific Islander. Regarding social needs, 10.9% of patients reported having medical transportation needs, and 21.6% presented with a risk for food insecurity. Financial resource strain was stratified into high (18.81%), medium (38.2%), and low (42.9%). The majority of patients resided in non-rural areas (71.2%). Medicare was the most common insurance type (64.1%), followed by Private insurance (22.2%) and Medicaid (11.5%). The mean comorbidity score (i.e., the average number of patient comorbidities) was approximately 10. Most patients did not have an assigned care manager (98.48%) or a social worker (95.9%).

Table 1.

Patients with Heart Failure Demographic Characteristics.

Variables	Overall (N = 2494)
Independent Variables
Medical Transportation Needs
Medical Transport Needs Present	274 (10.99)
Medical Transport Needs Absent	2220 (89.01)
Food Insecurity
Food Insecurity Risk Present	539 (21.61)
Food Insecurity Risk Absent	1955 (78.39)
Financial resource strain
Low Risk of Financial resource strain	1071 (42.94)
Medium Risk of Financial resource strain	954 (38.25)
High Risk of Financial resource strain	469 (18.81)
Patient Characteristics
Hospitalization Risk Score
Mean (SD)	40.37 (31.58)
Median (IQR)	33.00 (59.00)
(Min., Max.)	(1,99)
Age
Mean (SD)	64.90 (14.48)
Median (IQR)	66.00 (20.00)
(Min., Max.)	(18,89)
Number of Registered Nurses
Mean (SD)	7.76 (15.83)
Median (IQR)	2.00 (10.00)
(Min., Max.)	(0,347)
Comorbidity Score
Mean (SD)	10.08 (4.26)
Median (IQR)	10.00 (6.00)
(Min., Max.)	(0,25)
Care Manager
Zero Care Manager(s) Present	2456 (98.48)
One Care Manager(s) Present	35 (1.40)
Two Care Manager(s) Present	2 (0.08)
Three Care Manager(s) Present	1 (0.04)
Gender N (%)
Female	1394 (55.89)
Male	1100 (44.11)
Ethnicity
Hispanic or Latino	42 (1.68)
Not Hispanic or Latino	2452 (98.32)
Rurality
Rural	718 (28.79)
Not Rural	1776 (71.21)
Presence of Social Worker
Social Worker Present	100 (4.01)
Social Worker Absent	2394 (95.99)
Insurance Type
Medicare	1600 (64.15)
Medicaid	287 (11.51)
Private	556 (22.29)
Self-Pay	8 (0.32)
Other	43 (1.72)
Race
White	1356 (54.37)
Black	1047 (41.98)
Asian	14 (0.56)
Native Hawaiian/ Pacific Islander	2 (0.08)
American Indian/ Alaska Native	16 (0.64)
Other Race	59 (2.37)

3.1. Food insecurity

In the unadjusted analysis, the presence of food insecurity was associated with an 18.00-point increase in the subsequent hospitalization risk (p < 0.01) (Table 2). After adjusting for non-modifiable factors, this association remained significant, predicting a 12.29-point increase in subsequent hospitalization risk (p < 0.01). In the final, fully adjusted model, food insecurity was still associated with a significant 9.37-point increase in the subsequent hospitalization risk (p < 0.01). Other predictors were also significant. For each one-year increase in age, the risk score decreased by 0.34 points (p < 0.01), while each one-point increase in the comorbidity score was associated with a 2.65-point increase in risk (p < 0.01). Compared to their respective reference groups, being male was associated with a 6.94-point decrease in subsequent hospitalization risk (p < 0.01), and being Black was associated with an 11.45-point increase in subsequent hospitalization risk (p < 0.01). Patients with Medicaid were linked to a 9.07-point increase in subsequent hospitalization risk (p = 0.02), whereas Private insurance and Self-Pay status were associated with decreases of 14.64 points (p < 0.01) and 21.79 points (p = 0.03), respectively, compared to Medicare.

Table 2.

Quantile Regression Results for Food Insecurity and Hospitalization Risk.

Food Insecurity	Variables	Estimate	95% CI	p-value
Unadjusted	Food Insecurity Present Food Insecurity Absent	18.00	(12.59, 23.40)	<0.01
		Ref.	Ref.	Ref.
Adjusted (unmodifiable)	Food Insecurity Present	12.29	(7.62, 16.95)	<0.01
	Food Insecurity Absent	Ref.	Ref.	Ref.
	Age	−0.14	(−0.28, −0.00)	0.04
	Gender
	Male Female	−8.86	(−12.87, −4.84)	<0.01
		Ref.	Ref.	Ref.
	Race
	Black	18.57	(14.33, 22.80)	<0.01
	Asian	−6.71	(−23.53, 10.10)	0.43
	Native Hawaiian/ Pacific Islander	25.14	(−1245.88, 1296.17)	0.97
	American Indian/Alaska Native	18.71	(−14.79, 52.22)	0.27
	Other	6.00	(−12.10, 24.10)	0.52
	White	Ref.	Ref.	Ref.
	Ethnicity
	Hispanic	−10.71	(−28.99, 7.56)	0.25
	Not Hispanic	Ref.	Ref.	Ref.
Adjusted (unmodifiable + modifiable)	Food Insecurity Present	9.37	(4.83, 13.90)	<0.01
	Food Insecurity Absent	Ref.	Ref.	Ref.
	Age	−0.34	(−0.47, −0.21)	<0.01
	Number of Registered Nurses	0.23	(0.09, 0.36)	<0.01
	Comorbidity Score	2.65	(2.24, 3.06)	<0.01
	Gender
	Male	−6.94	(−10.24, −3.64)	<0.01
	Female	Ref.	Ref.	Ref.
	Race
	Black	11.45	(8.06, 14.83)	<0.01
	Asian	2.89	(−7.29, 13.06)	0.58
	Native Hawaiian/ Pacific Islander	43.44	(−400.77, 487.66)	0.85
	American Indian/Alaska Native	9.38	(−12.79, 31.56)	0.41
	Other	9.15	(−3.96, 22.26)	0.17
	White	Ref.	Ref.	Ref.
	Ethnicity
	Hispanic	−8.23	(−19.73, 3.27)	0.16
	Not Hispanic	Ref.	Ref.	Ref.
	Number of Care Managers
	One Care Manager(s) Present	−9.25	(−26.26, 7.76)	0.29
	Two Care Manager(s) Present	36.59	(−169.64, 242.81)	0.73
	Three Care Manager(s) Present	−22.78	(−743.27, 697.73)	0.95
	Zero Care Manager(s) Present	Ref.	Ref.	Ref.
	Presence of a Social Worker
	Social Worker Present	0.76	(−7.01, 8.55)	0.85
	Social Worker Absent	Ref.	Ref.	Ref.
	Rurality
	Rural	2.24	(−1.45, 5.93)	0.23
	Not Rural	Ref.	Ref.	Ref.
	Insurance Type
	Medicaid	9.07	(1.73, 16.42)	0.02
	Private	−14.64	(−18.50, −10.78)	<0.01
	Other	−6.20	(−14.72, 2.32)	0.15
	Self-Pay	−21.79	(−41.23, −2.36	0.03
	Medicare	Ref.	Ref.	Ref

3.2. Financial resource strain

A similar pattern of association was observed for financial resource insecurity (Table 3). In the unadjusted model, high insecurity was associated with a 21.00-point increase in subsequent hospitalization risk compared to low insecurity (p < 0.01), while medium insecurity was not significant. After adjusting for non-modifiable factors, high insecurity predicted a 17.67-point increase in risk (p < 0.01), and medium insecurity became statistically significant (5.50-point increase, p = 0.01). In the final, fully adjusted model, high financial insecurity remained a significant predictor, corresponding to a 13.13-point increase in the risk score (p < 0.01). Interestingly, each additional year of age was associated with a 0.39-point decrease in the subsequent hospitalization risk (p < 0.01), while each point increase in comorbidity score corresponded to a 2.59-point increase (p < 0.01). Being male was associated with an 8.69-point lower risk score than being female (p < 0.01), and Black patients had scores 11.33 points higher than White patients (p < 0.01). Compared to those with Medicare, having Medicaid was linked to a 7.32-point increase in subsequent hospitalization risk (p = 0.03), while Private insurance and Self-Pay were associated with decreases of 14.89 points (p < 0.01) and 23.05 points (p = 0.03), respectively.

Table 3.

Quantile Regression Results for Financial Resource Strain and Hospitalization Risk.

Financial Resource Insecurity	Variables	Estimate	95% CI	p-value
Unadjusted	High Financial Resource Strain Medium Financial Resource Strain Low Financial Resource strain	21.00	(14.84, 27.16)	<0.01
		3.00	(−0.94, 6.93)	0.14
		Ref.	Ref.	Ref.
Adjusted (unmodifiable)	High Financial Resource Strain	17.67	(10.75, 24.58)	<0.01
	Medium Financial Resource Strain	5.50	(1.11, 9.89)	0.01
	Low Financial Resource Strain	Ref.	Ref.	Ref.
	Age	−0.17	(−0.32, −0.01)	0.04
	Gender
	Male Female	−10.17	(−14.28, −6.05)	<0.01
		Ref.	Ref.	Ref.
	Race
	Black	17.50	(12.96, 22.04)	<0.01
	Asian	−4.83	(−24.17, 14.51)	0.62
	Native Hawaiian/ Pacific Islander	21.33	(−864.77, 907.44)	0.96
	American Indian/Alaska Native	15.83	(−14.46, 46.13)	0.31
	Other	4.33	(−11.10, 19.77)	0.58
	White	Ref.	Ref.	Ref.
	Ethnicity
	Hispanic	−11.00	(−27.44, 5.44)	0.19
	Not Hispanic	Ref.	Ref.	Ref.
Adjusted (unmodifiable + modifiable)	High Financial Resource Insecurity	13.13	(8.50, 17.76)	<0.01
	Medium Financial Resource Insecurity	1.94	(−1.32, 5.20)	0.24
	Low Financial resource strain	Ref.	Ref.	Ref.
	Age	−0.39	(−0.52, −0.27)	<0.01
	Number of Registered Nurses	0.24	(0.11, 0.36)	<0.01
	Comorbidity Score	2.59	(2.21, 2.97)	<0.01
	Gender
	Male	−8.69	(−11.66, −5.74)	<0.01
	Female	Ref.	Ref.	Ref.
	Race
	Black	11.33	(7.99, 14.66)	<0.01
	Asian	3.04	(−8.89, 14.96)	0.62
	Native Hawaiian/ Pacific Islander	36.86	(−514.77, 588.49)	0.89
	American Indian/Alaska Native	10.24	(−12.13, 32.60)	0.37
	Other	10.90	(−3.49, 25.30)	0.14
	White	Ref.	Ref.	Ref.
	Ethnicity
	Hispanic	−7.10	(−18.15, 3.94)	0.21
	Not Hispanic	Ref.	Ref.	Ref.
	Number of Care Managers
	One Care Manager(s) Present	−12.78	(−28.62, 3.08)	0.11
	Two Care Manager(s) Present	21.02	(−277.01, 319.06)	0.89
	Three Care Manager(s) Present	−23.33	(−1341.87, 1295.22)	0.97
	Zero Care Manager(s) Present	Ref.	Ref.	Ref.
	Presence of a Social Worker
	Social Worker Present	2.41	(−6.99, 11.80)	0.62
	Social Worker Absent	Ref.	Ref.	Ref.
	Rurality
	Rural	1.83	(−1.56, 5.22)	0.29
	Not Rural	Ref.	Ref.	Ref.
	Insurance Type
	Medicaid	7.32	(0.90, 13.73)	0.03
	Private	−14.89	(−18.68, −11.11)	<0.01
	Other	−5.96	(−17.17, 5.27)	0.29
	Self-Pay	−23.05	(−43.19, −2.89)	0.03
	Medicare	Ref.	Ref.	Ref.

3.3. Medical transportation needs

The analysis of medical transportation needs yielded different findings (Table 4). In the unadjusted model, the presence of medical transportation needs was linked to a 26.00-point higher subsequent hospitalization risk (p < 0.01). This was attenuated to a 17.45-point increase after adjusting for non-modifiable factors (p < 0.01) and further to a 13.23-point increase in the final fully adjusted model (p < 0.01). Within the final model, several other variables were significant predictors of hospitalization risk. Age was associated with a 0.37-point decrease in subsequent hospitalization risk (p < 0.01), and the comorbidity score was associated with a 2.57-point increase (p < 0.01). Male patients had risk scores 7.63 points lower than female patients (p < 0.01). Black patients had scores 11.69 points higher than White patients (p < 0.01). Regarding insurance type and compared to Medicare, Medicaid was linked to an 8.42-point increase in subsequent hospitalization risk (p < 0.01), while Private insurance and Self-Pay were associated with decreases of 13.25 points (p < 0.01) and 24.28 points (p < 0.01), respectively. Across all final models, ethnicity, number of care managers, presence of a social worker, and rurality were not found to be significantly associated with subsequent hospitalization risk score.

Table 4.

Quantile Regression Results for Medical Transportation Needs and Hospitalization Risk.

Unadjusted	Medical Transportation Needs Present Medical Transportation Needs Absent	26.00	(18.73, 33.27)	<0.01
		Ref.	Ref.	Ref.
Adjusted (unmodifiable)	Medical Transportation Needs Present	17.45	(11.68, 23.22)	<0.01
	Medical Transportation Needs Absent	Ref.	Ref.	Ref.
	Age	−0.15	(−0.29, −0.01)	0.03
	Gender
	Male Female	−9.50	(−13.02, −5.98)	<0.01
		Ref.	Ref.	Ref.
	Race
	Black	17.25	(13.29, 21.21)	<0.01
	Asian	−7.90	(−25.52, 9.72)	0.38
	Native Hawaiian/ Pacific Islander	36.15	(−955.23, 1027.53)	0.94
	American Indian/Alaska Native	17.70	(−16.06, 51.46)	0.30
	Other	3.95	(−12.22, 20.12)	0.63
	White	Ref.	Ref.	Ref.
	Ethnicity
	Hispanic	−9.95	(−22.21, 2.31)	0.11
	Not Hispanic	Ref.	Ref.	Ref.
Adjusted (unmodifiable + modifiable)	Medical Transportation Needs Present	13.23	(7.55, 18.89)	<0.01
	Medical Transportation Needs Absent	Ref.	Ref.	Ref.
	Age	−0.37	(−0.49, −0.22)	<0.01
	Number of Registered Nurses	0.27	(0.13, 0.39)	<0.01
	Comorbidity Score	2.57	(2.22, 2.92)	<0.01
	Gender
	Male	−7.63	(−10.28, −4.99)	<0.01
	Female	Ref.	Ref.	Ref.
	Race
	Black	11.69	(8.34, 15.04)	<0.01
	Asian	0.41	(−12.59, 13.41)	0.95
	Native Hawaiian/ Pacific Islander	50.51	(−388.82, 489.83)	0.82
	American Indian/Alaska Native	9.61	(−15.30, 34.53)	0.45
	Other	8.66	(−6.02, 23.35)	0.25
	White	Ref.	Ref.	Ref.
	Ethnicity
	Hispanic	−4.96	(−18.59, 8.67)	0.48
	Not Hispanic	Ref.	Ref.	Ref.
	Number of Care Managers
	One Care Manager(s) Present	−16.02	(−34.65, 2.60)	0.09
	Two Care Manager(s) Present	20.59	(−379.14, 420.34)	0.92
	Three Care Manager(s) Present	−25.93	(−971.87, 920.01)	0.96
	Zero Care Manager(s) Present	Ref.	Ref.	Ref.
	Presence of a Social Worker
	Social Worker Present	−0.02	(−9.72, 9.69)	0.99
	Social Worker Absent	Ref.	Ref.	Ref.
	Rurality
	Rural	2.04	(−1.35, 5.43)	0.24
	Not Rural	Ref.	Ref.	Ref.
	Insurance Type
	Medicaid	8.42	(2.86, 13.99)	<0.01
	Private	−13.25	(−16.90, −9.59)	<0.01
	Other	−6.83	(−17.18, 3.53)	0.19
	Self-Pay	−24.28	(−37.78, −10.78)	<0.01
	Medicare	Ref.	Ref.	Ref.

4. Discussion

The current study demonstrates that social determinants, specifically food insecurity risk, financial resource strain, and medical transportation needs, are associated with increased hospitalization risk among patients with HF, even after controlling for clinical and demographic factors. Our findings align with emerging evidence linking food insecurity to adverse hospital outcomes. A recent study by Bansal et al. (2024) found that patients with HF and food insecurity experience higher mortality rates, longer hospital stays, increased hospitalization charges, and greater complication rates, including cardiac arrest and cardiogenic shock, compared to non-food-insecure patients [34]. The mechanistic relationship between food insecurity and hospitalization likely involves multiple interconnected pathways. Nutritional inadequacy, driven by the inability to afford heart-healthy foods such as fresh produce, lean proteins, and low-sodium options, forces patients to rely on cheaper, processed foods high in sodium and preservatives, which can worsen fluid retention and cardiac function [35], [36], [37], [38]. Additionally, the association of food insecurity with increased psychological stress may worsen cardiometabolic conditions through neuroendocrine mechanisms, directly impacting future hospital utilization [39], [40].

High financial resource strain emerged as a significant predictor of subsequent hospitalization in our analysis. This finding corroborates previous research demonstrating that financial barriers significantly impact cardiovascular outcomes [9]. A systematic review by Swarup and others (2024) found that financial stress was positively associated with major cardiac outcomes [41]. Financial strain creates a cascade of barriers that compromise HF management and ultimately increase hospitalization risk. Patients facing economic hardship often reduce or discontinue prescribed medications due to cost concerns, leading to suboptimal HF management and increased risk of acute exacerbations requiring emergency care [42]. Similarly, financial constraints drive patients to delay routine medical appointments and preventive care visits to avoid copayments and other expenses. This lack of clinical monitoring may further allow the progression of HF symptoms until hospitalization becomes necessary. Beyond these direct pathways, the chronic psychological stress of financial insecurity can activate inflammatory biological systems that add stress to the cardiac system, further worsening HF conditions [43], [44]. This multi-faceted impact of economic barriers on HF outcomes highlights the critical need for healthcare systems to integrate financial support and resource assistance into comprehensive HF management programs to reduce preventable hospitalizations.

Medical transportation needs also emerged as a strong predictor of hospitalization risk, highlighting a frequently overlooked barrier to optimal heart failure management. Our findings align with research demonstrating that transportation barriers were associated with missed cardiology appointments and subsequent adverse events [45]. Beyond missed appointments, these transportation challenges create cascading effects that compromise medication adherence through difficulty accessing pharmacy refills, reduced participation in cardiac rehabilitation programs, and delayed ability to seek timely care when experiencing worsening symptoms [46], [47]. These findings demonstrate that addressing transportation barriers represents a potentially high-impact, modifiable risk factor that could significantly reduce preventable hospitalizations and improve outcomes for vulnerable populations with HF.

While our analysis examined each social determinant independently, broader literature suggests these factors often co-occur and interact synergistically. Patients experiencing food insecurity frequently also face financial strain and transportation barriers, creating compound disadvantage [48]. Tucher et al. (2025) demonstrated that patients with multiple needs of financial strain, food insecurity, housing instability, and transportation barriers had exponentially higher healthcare utilization compared to those with single needs [48]. This intersectionality is particularly evident when thinking about how transportation barriers affect food access (i.e., patients who cannot reach grocery stores face both transportation and food insecurity challenges). Similarly, financial constraints simultaneously limit food purchasing power, transportation options, and medication adherence, creating cascading effects on hospitalization outcomes. [49]

The magnitude of associations between food insecurity, financial resource strain, medical transportation needs and hospitalization risk has profound implications for healthcare delivery. Traditional risk stratification models (i.e., that focus primarily on clinical parameters may substantially underestimate risk for socially vulnerable patients [50], [51]. Research from the REGARDS study demonstrates that having any social determinant of health nearly triples the risk prediction of 90-day mortality after heart failure hospitalization, independent of other factors. Our study extends this evidence to show similar patterns for future hospitalization risk among patients with HF [52]. These findings underscore the critical need for systematic SDOH screening as a standard component of heart failure care, but assessment alone is insufficient, healthcare systems must commit to building the infrastructure and allocating resources necessary to intervene on identified social needs. For example, although Centers for Medicare & Medicaid Services initiatives incorporate social risk factors into quality metrics represent progress, implementation remains inconsistent. Furthermore, value-based care models that penalize hospitals for excess readmissions may inadvertently disadvantage facilities serving socially vulnerable populations unless social risk factors are adequately adjusted for in payment models [53].

Based on our findings, several recommendations are needed for improving care delivery and outcomes. One, the Epic Risk of Hospital Admission model used in this study demonstrates the feasibility of integrating predictive analytics but should be enhanced with real-time social determinant data to improve risk stratification accuracy. Two, the complexity of available resources, from SNAP applications to various transportation programs, necessitates dedicated resource navigation and coordination support. Embedding community health workers or social workers within clinical teams has shown promise in addressing social needs and connecting patients with appropriate community resources [54], [55]. These personnel can help patients understand eligibility requirements, complete applications, and access services that directly address their social determinants of health. Three, healthcare systems should form strategic partnerships with local grocery stores, transportation providers, and other community stakeholders to create direct interventions on SDOH and deliver more seamless care, rather than fragmented referral-based approaches. From a policy perspective, current value-based payment models and readmission penalties inadequately account for these social risk factors, potentially penalizing hospitals serving vulnerable populations; our findings support incorporating social determinant adjustments into quality metrics and reimbursement, as well as strengthening upstream programs like SNAP, Medicaid expansion, and rural transportation services. Rural patients face especially severe challenges, with no public transit, taxi, or Metro Link services available outside metropolitan counties, forcing reliance on limited regional programs with restrictive eligibility criteria. Finally, documented cuts to programs like Care Link, which previously served individuals aged 60 and older, have created critical service voids and underscore the need for sustained policy advocacy and funding initiatives.

4.1. Study limitations

Several limitations warrant consideration. First, our cross-sectional design prevents causal inference. Second, the 19% exclusion rate due to missing data may introduce selection bias, potentially underestimating effects if patients with missing data had more severe social needs. Third, our single-center study in Arkansas may limit generalizability, though the state's high burden of cardiovascular disease and social vulnerability makes it an important setting for this research. Fourth, our sample size provided insufficient statistical power to detect meaningful differences across racial and ethnic subgroups, limiting our ability to examine potential disparities in the relationship between social determinants and hospitalization risk. Finally, the use of social determinant measures, while previously validated, relies on self-report and may underestimate true prevalence due to stigma or social desirability bias. Despite these limitations, our findings provide valuable insights into the relationship between social determinants of health (i.e., food insecurity risk, financial resource strain, and medical transportation needs) and hospitalization risk outcomes in a vulnerable population, suggesting that routine screening and intervention programs may be warranted.

4.2. Future research

Future research is needed to employ longitudinal designs to establish temporality and examine trajectories of social needs over time. Additionally, research examining the cost-effectiveness of social interventions from health system perspectives would strengthen the institutional case for addressing social determinants among patients with HF. Finally, the development and validation of integrated risk models that combine clinical and social factors could improve population health management strategies in hospital settings.

5. Conclusion

This study provides compelling evidence that social determinants of health, specifically food insecurity risk, financial resource strain, and medical transportation needs, are associated with increased hospitalization risk among patients with HF. Moving forward, healthcare providers and health systems should consider implementing systematic social determinant screening protocols, developing partnerships with community organizations to address identified social needs, and integrating social risk factors into clinical decision-making and care planning processes.

CRediT authorship contribution statement

Alisha A. Crump: Writing – review & editing, Writing – original draft, Software, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Milan Bimali: Writing – review & editing, Writing – original draft, Project administration, Methodology, Formal analysis. Sanchia McCall: Writing – review & editing, Conceptualization. Pearman Parker: Writing – review & editing. Taren Massey-Swindle: Writing – review & editing. Kevin Wayne Sexton: Writing – review & editing. Emel Seker: Writing – review & editing. Maryam Y. Garza: Writing – review & editing. Tremaine B. Williams: Writing – review & editing, Writing – original draft, Supervision, Project administration, Investigation, Funding acquisition, Conceptualization.

Ethics statement

The study protocol (#276211) was reviewed and approved by the Ethics Committee of the University of Arkansas for Medical Sciences' Institutional Review Board. This study was conducted in accordance with all applicable government regulations and institutional research policies and procedures at the University of Arkansas for Medical Sciences. The study team only had access to de-identified data on patients and clinicians. This study presented only minimal risk for loss of confidentiality, and a waiver of documentation of the informed consent process was granted by the Ethics Committee, which is the standard for retrospective analyses involving de-identified data.

Clinical trial number

Not applicable.

Funding

This work was supported by the National Institute of Health (National Institute of Nursing Research: R21NR021063).

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:

Tremaine Williams reports financial support was provided by National Institute of Health. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

The data that support the findings are available from the University of Arkansas for Medical Sciences, but restrictions apply to their public availability and so are not publicly available. Data are, however, available from the authors upon reasonable request and with permission from the University of Arkansas for Medical Sciences.