الفهرس | Only 14 pages are availabe for public view |
Abstract Background: The COVID-19 pandemic presented a difficult challenge to physicians. Timely diagnosis, risk stratification, and proper decision on hospitalization are crucial to save the most lives possible. Despite many patients’ success in overcoming the virus, some are exposed to the disease’s poor outcomes. Aging is considered a primary risk factor for these poor outcomes. Machine learning has been increasingly used for risk classification, prediction, screening, and decision-making in medicine. The aim of this study was to identify the predictors of mortality in elderly patients hospitalized with COVID-19 and present a valid model for predicting on-admission mortality risk. Patients and methods: This is a retrospective hospital-based study that included consecutively hospitalized patients aged 50 years and above with COVID-19. Five machine-learning classification models were used; the support vector machine (SVM), random forest (RF), logistic regression (LR), k-nearest neighbor (KNN), and decision tree (DT). Results: This study included 301 patients who were hospitalized for COVID-19. Assessment of the machine learning models’ performance showed that all the trained models exhibited good performance for internal validation (AUC > 80%). The Decision Tree and Random Forest algorithms recalled the highest accuracy (100%). The strongest predictors of mortality were mechanical ventilation, quick SOFA (qSOFA) score, National Early Warning Score (NEWS), and ICU admission. Conclusion: The present work emphasizes the performance of ML methods in predicting mortality in older adult patients hospitalized with COVID-19. The assessment tools: NEWS, qSOFA, MV, and ICU admission strongly predicted mortality in this age group. |