Enhancing Care Transitions for Older Patients: A Big Data-Driven Readmission Prediction Model for Personalized Discharge Nursing Care

This study developed a machine learning-based model to predict 30-day unplanned readmissions in elderly inpatients using nursing data from a tertiary hospital in South Korea. A retrospective analysis of 58,231 patients aged 65 and older was conducted, extracting variables from nursing assessments, discharge plans, and clinical documentation. LASSO regression was used for feature selection, followed by logistic regression for model development. The final model included 23 predictors, such as age, sex, admission route, infection status, number of medications, adverse drug reactions, nursing acuity (KPCS), comorbidities (CCI), dialysis, and chemotherapy. The model showed strong performance: AUC = 0.7484, accuracy = 0.8007, F1-score = 0.8784. This tool enables early identification of high-risk patients and supports personalized discharge planning. Future research should focus on multi-site validation and incorporating social determinants of health to improve generalizability.