Background: Despite the World Health Organization’s announcement that the coronavirus pandemic had ended in 2020, the nature of the disease remains unclear, and there are no proven treatments, including effective drugs, to treat and prevent patient death. Therefore, this study was designed to predict the risk of death in hospitalized COVID-19 patients using modern statistical modeling and machine learning techniques. Materials and Methods: This cross-sectional study was conducted on data from the COVID-19 disease registration program in 2024. Various machine learning algorithms were used to analyze the data. The performance of the models used was evaluated by determining accuracy, sensitivity, specificity, and area under the curve (AUC). The final model was used to determine the most important predictors. Results: The area under the ROC curve for the three models (LR, NN, and NB) is used to compare their performance. The LR model has the highest accuracy (AUC = 95.7). Accordingly, the most important variables predicting death in patients were WBC, BUN, age, and glucose levels. As a result, the model accurately predicted the risk of death in patients by 77.17%, with an accuracy of 95% and a sensitivity of 88.02. In addition, the results of the model test showed that the model can predict the risk of death by 94.54% and the probability of survival by 93.81%. Conclusion: Based on the results, the most important variables predicting death from COVID-19 were WBC, BUN, age, and glucose levels in patients, so that the model was able to predict the risk of death in hospitalized patients with very high accuracy.
Sayyadi et al. (Sun,) studied this question.