Artificial Intelligence Techniques in Prediction of COVID-19 Mortality and Its Related Factors: A Multi-Center Study

This study utilized machine learning (ML) algorithms to predict mortality risk among 7115 COVID-19 patients from 35 hospitals in Khorasan-Razavi, Iran. The random forests (RF) model outperformed others, achieving an AUC of 88%, accuracy of 96%, precision of 96%, sensitivity of 99%, and specificity of 80%. Key predictors of mortality included intubation, SpO2, age, time from symptoms, loss of consciousness, reception season, and diabetes. The RF model shows promise for aiding clinical decisions and resource allocation but requires external validation in larger, diverse populations.