Predictive Modelling for Hospital Readmission Using Socioeconomic and Clinical Data

Hospital readmissions pose a significant burden on healthcare systems, impacting patient outcomes and inflating medical costs. This study explores the development and implementation of predictive modeling techniques to identify patients at high risk of hospital readmission using an integrated dataset comprising socioeconomic and clinical variables. Drawing from de-identified records of over 50,000 patient discharges across multiple hospitals in the United States, we employed a hybrid machine learning framework incorporating logistic regression, random forest, and gradient boosting algorithms to assess and predict 30-day readmission likelihood. Socioeconomic indicators such as income level, insurance status, employment, education, and neighborhood deprivation index were combined with clinical factors including comorbidities, length of stay, discharge disposition, primary diagnosis, and prior hospitalization history. Feature selection techniques and recursive elimination were applied to identify the most influential predictors. The models were validated using stratified 10-fold cross-validation and evaluated using precision, recall, F1-score, and area under the receiver operating characteristic (ROC) curve. The gradient boosting model demonstrated the highest performance, achieving an AUC of 0.84, indicating strong discriminatory power. Among the top predictors were chronic conditions such as congestive heart failure and diabetes, low-income status, limited education, and discharge to non-home settings. Furthermore, the study underscores the synergistic effect of integrating socioeconomic determinants with clinical data, highlighting that models using only clinical data underperformed compared to those incorporating both domains. This integrative approach provides healthcare providers with a more nuanced understanding of the factors influencing readmission risk and supports proactive intervention strategies, such as post-discharge care planning and targeted resource allocation for vulnerable populations. The study emphasizes the importance of incorporating social determinants of health into predictive analytics to advance equitable and efficient healthcare delivery. Future work will focus on the deployment of these predictive tools within electronic health record systems and the development of real-time dashboards for care teams. By leveraging predictive modeling rooted in comprehensive patient profiles, healthcare systems can enhance care continuity, reduce preventable readmissions, and support policy frameworks aimed at value-based care.