Improving ICU Resource Allocation through Machine Learning-Based Comorbidity-Aware Admission Prediction

The COVID-19 pandemic has significantly disrupted global healthcare systems, necessitating advanced data-driven approaches to manage patient care. The dataset utilized in this research comprises detailed information on comorbid patients, including demographic profiles, medical history categorized by disease groups, blood test results, and vital signs. The study employs a hybrid index approach for ICU admission prediction using machine learning classification techniques. Strategies are developed to optimize critical resource allocation, ensuring the effective distribution of healthcare resources and timely interventions for COVID-19 patients. The Omega index is evaluated for feature selection to minimize ICU admission prediction time and is compared against the Alpha index for performance assessment. Experimental results demonstrate that integrating the Omega index with Decision Tree models yields superior performance, achieving 97.87% accuracy, 100% recall, and an ROC of 0.99. Overall, this study advances predictive analytics in healthcare, contributing to improved optimization of ICU admissions during the COVID-19 pandemic.