Calculating the Risk of Admission to Intensive Care Units in COVID-19 Patients Using Machine Learning

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Epidemiology

Background:

• The COVID-19 pandemic strained global healthcare systems, necessitating efficient resource allocation.
• Prioritizing patients at risk for intensive care unit (ICU) admission is crucial for managing vulnerable populations and optimizing resource use.

Purpose of the Study:

• To develop and evaluate a machine learning model for predicting COVID-19 patient risk of ICU admission during hospitalization.
• To assess the model's performance in identifying patients requiring critical care.

Main Methods:

• Retrospective cohort study involving adult COVID-19 patients from three Spanish hospitals (n=1272 for development, n=550 for validation).
• Development and validation of a classification model using machine learning techniques.
• Evaluation of model performance using metrics such as sensitivity, specificity, and area under the curve (AUC).

Main Results:

• A LightGBM model incorporating 40 variables demonstrated high predictive performance.
• The model achieved an AUC of 1.00 (0.99-1.0), specificity of 0.99 (0.97-1.00), and sensitivity of 0.92 (0.86-0.98) on the test dataset.

Conclusions:

• A highly accurate model for predicting ICU admission in hospitalized COVID-19 patients was successfully developed.
• Early identification and prioritization of high-risk patients facilitate appropriate care allocation, especially during periods of healthcare system pressure.