Enhancing care transitions for older patients: A big data-driven readmission prediction model for personalized...

Given that readmission rates and the associated costs constitute a crucial management concern, this study analyzed nursing records of patients aged 65 years and older who were admitted to a tertiary general hospital in South Korea between November 2019 and December 2023, with the aim of identifying factors associated with 30-day unplanned readmissions and developing a predictive model using machine learning. Data were extracted from electronic medical records, including general demographics, admission nursing assessments, and discharge summaries. Least absolute shrinkage and selection operator (LASSO) regression was used for variable selection, followed by logistic regression to construct the prediction model. Key predictors identified included age, gender, admission route, infection status, number of medications prescribed, adverse drug reactions, Korean Patient Classification System (KPCS) classification, Charlson Comorbidity Index (CCI), dialysis, and chemotherapy. The final model achieved an area under the curve (AUC) of 0.7484, accuracy of 0.8007, and an F1-score of 0.8784, indicating moderate discriminatory performance. The model has potential utility when it comes to supporting nurses in identifying high-risk patients and implementing targeted care interventions to prevent readmissions. Based on these findings, the study recommends the development of a hospital-integrated program to enable early identification and proactive management of patients at high risk of readmission-a measure which could have positive health implications for such patients.