Understanding the Connection
Congestive heart failure (CHF) represents one of the most pressing healthcare challenges in the United States, accounting for nearly one-fifth of all deaths and serving as the leading cause of hospitalization among older adults. While traditional clinical risk factors like hypertension and diabetes have long been recognized as contributors to CHF progression, emerging evidence highlights the critical role of social determinants of health (SDOH) in shaping disease outcomes, medication adherence, and healthcare access.
Social determinants of health encompass a broad spectrum of non-medical factors that influence health outcomes, including financial stability, housing security, transportation access, food availability, educational attainment, social connections, and mental health status. As healthcare systems increasingly integrate SDOH screening into routine electronic health record (EHR) workflows, understanding how these social needs correlate with CHF status becomes essential for effective risk stratification and population health management.
The chronic nature of CHF makes it particularly susceptible to social influences that often remain invisible to healthcare providers yet significantly impact disease management. Patients facing financial hardship may struggle to afford medications, while those lacking reliable transportation may miss critical follow-up appointments. Similarly, individuals experiencing food insecurity might find it challenging to maintain heart-healthy diets, and those dealing with depression may have reduced medication adherence.
Study Overview and Methods
This comprehensive retrospective analysis examined 30,534 inpatient electronic health records from Our Lady of the Lake Regional Medical Center in Baton Rouge, Louisiana. The dataset included 7,618 patients diagnosed with congestive heart failure and 22,916 patients without CHF who had chronic conditions predisposing them to heart failure, such as obesity, diabetes, or hypertension.
The research team employed multiple analytical approaches to understand the relationship between social determinants and heart failure status. Chi-square tests examined associations between CHF status and various SDOH variables, while multivariable logistic regression evaluated adjusted predictors after controlling for demographic factors including age, sex, and race.
To address the challenge of class imbalance in the dataset, researchers applied the Synthetic Minority Oversampling Technique (SMOTE), which generates synthetic samples of the minority class to improve model learning. A Random Forest classifier was then utilized to assess the predictive importance of SDOH and demographic features, testing various configurations ranging from 100 to 500 trees with maximum depths between 10 and 1,000.
Additionally, hierarchical clustering analysis was performed to identify patterns of co-occurring social needs among patients, revealing how different social challenges tend to cluster together in specific patient populations.
Key Findings and Predictions
The analysis revealed that all social determinants of health examined were significantly associated with CHF status, with the notable exception of tobacco use. However, despite statistical significance, the individual effect sizes for most SDOH variables were relatively small, suggesting that these factors exert their influence collectively rather than in isolation.
Age emerged as the single strongest predictor of congestive heart failure, with adjusted odds increasing nearly tenfold among individuals aged 75 years and older compared to younger adults. This finding underscores the importance of age-specific interventions and monitoring strategies for heart failure prevention and management.
Beyond age, several psychosocial and economic factors demonstrated meaningful associations with CHF status. Depression registry entry and financial insecurity both increased the odds of having heart failure, highlighting the interconnected nature of mental health, economic stability, and cardiovascular outcomes. Conversely, being married or living with a partner showed a protective effect, suggesting that social support and stable relationships may play a beneficial role in heart health.
Demographics and Social Risk Factors
Patients with congestive heart failure exhibited distinct demographic characteristics compared to the non-CHF group. They were significantly older, more frequently male, and disproportionately represented among Black patients. These demographic patterns align with existing epidemiological knowledge about heart failure prevalence and underscore the importance of considering both biological and social factors in understanding disease burden.
The study examined multiple domains of social need, including substance use and sexual activity, socioeconomic factors (demographics, financial strain, food insecurity, transportation needs), lifestyle factors (physical activity, stress), and relationship dynamics (social connections, intimate partner violence). Each domain was captured through structured screening questions embedded in the Epic EHR system’s evidence-based SDOH questionnaire.
Financial resource strain was assessed through questions about difficulty paying for basic necessities like food, housing, medical care, and heating. Food insecurity was evaluated based on whether purchased food lasted and concerns about running out of food before having money to buy more. Transportation barriers were identified when lack of transportation prevented patients from attending medical appointments or obtaining medications.
Machine Learning Insights
The Random Forest classifier achieved moderate predictive performance with an area under the receiver operating characteristic curve (AUC) of 0.67 and balanced accuracy of 62%. While these metrics indicate room for improvement, they demonstrate that SDOH variables, when combined with demographic information, provide meaningful predictive utility for CHF risk stratification.
Feature importance analysis revealed that age, transportation barriers, and physical activity ranked as the top predictors in the model. Transportation access emerged as particularly significant, potentially because it affects multiple aspects of disease management including medication pickup, appointment attendance, and access to healthy food options. Physical activity levels also showed strong predictive value, likely reflecting both behavioral risk factors and functional capacity related to existing cardiovascular compromise.
Patient Clustering Patterns
Hierarchical clustering analysis uncovered distinct subgroups of patients characterized by co-occurring social needs. One cluster included patients experiencing depression, alcohol use, and high stress levels, suggesting a psychosocial vulnerability profile. Another cluster was characterized by economic hardship, combining financial strain, food insecurity, and transportation limitations.
These clustering patterns have important implications for intervention design. Rather than addressing social needs individually, healthcare systems may achieve better outcomes by developing targeted programs that address the constellation of challenges facing specific patient subpopulations.
Clinical Implications
The findings from this study support the integration of SDOH screening into routine clinical workflows and the incorporation of social need data into risk stratification models. While individual effect sizes for most social determinants were small, their collective patterns demonstrated clinical relevance and predictive utility when analyzed using machine learning approaches.
Healthcare systems can leverage these insights to identify high-risk patients who would benefit from enhanced care coordination, social work referrals, or community resource connections. The use of advanced analytics, including machine learning and oversampling techniques like SMOTE, provides a scalable approach to improve CHF risk prediction and guide targeted intervention strategies.
Conclusion
This comprehensive analysis demonstrates that demographic characteristics and social conditions contribute meaningfully to congestive heart failure burden, with age, socioeconomic stressors, and psychosocial vulnerability emerging as key factors. Despite small individual effect sizes for most social determinants, their collective patterns revealed through clustering analysis and machine learning models showed significant clinical relevance.
The study’s exploratory design provides a foundation for future predictive and longitudinal modeling while highlighting the importance of routine SDOH screening in inpatient settings. By integrating social determinants data into electronic health records and leveraging advanced analytical techniques, healthcare systems can better identify at-risk populations and implement preventive interventions before heart failure develops or progresses.
As healthcare continues its evolution toward value-based care and population health management, understanding the social context of disease becomes increasingly critical. This research demonstrates that effective CHF prevention and management requires addressing not only clinical risk factors but also the social determinants that shape patients’ ability to maintain cardiovascular health.
