Identifying Factors Associated with COVID-19 All-Cause 90-Day Readmission: Machine Learning Approaches

The COVID-19 pandemic has placed immense strain on healthcare systems. In response to this challenge, our study employs machine learning techniques to identify and analyze risk factors associated with COVID-19 all-cause 90-day readmission. Leveraging the Optum® de-identified COVID-19 Electronic Health Record data set, we developed predictive models with comparable efficacy, particularly the optimized XGBoost model in prognosticating readmission risks. Our analysis reveals several key risk factors aligned with existing research and finds specific laboratory tests that may serve as potential indicators of readmission risk. By elucidating these critical determinants, our study expands the knowledge base for clinical decision-making, offering healthcare practitioners deeper insights into the factors affecting COVID-19 patient readmission risks. These findings can potentially empower clinicians to refine interventions and care strategies, mitigating adverse outcomes and advancing healthcare delivery for individuals affected by COVID-19.