Interpretable generalized neural additive models for mortality prediction of COVID-19 hospitalized patients in Hamadan, Iran.

BACKGROUND: The high number of COVID-19 deaths is a serious threat to the world. Demographic and clinical biomarkers are significantly associated with the mortality risk of this disease. This study aimed to implement Generalized Neural Additive Model (GNAM) as an interpretable machine learning method to predict the COVID-19 mortality of patients. METHODS: This cohort study included 2181 COVID-19 patients admitted from February 2020 to July 2021 in Sina and Besat hospitals in Hamadan, west of Iran. A total of 22 baseline features including patients' demographic information and clinical biomarkers were collected. Four strategies including removing missing values, mean, K-Nearest Neighbor (KNN), and Multivariate Imputation by Chained Equations (MICE) imputation methods were used to deal with missing data. Firstly, the important features for predicting binary outcome (1: death, 0: recovery) were selected using the Random Forest (RF) method. Also, synthetic minority over-sampling technique (SMOTE) method was used for handling imbalanced data. Next, considering the selected features, the predictive performance of GNAM for predicting mortality outcome was compared with logistic regression, RF, generalized additive model (GAMs), gradient boosting decision tree (GBDT), and deep neural networks (DNNs) classification models. Each model trained on fifty different subsets of a train-test dataset to ensure a model performance. The average accuracy, F1-score and area under the curve (AUC) evaluation indices were used for comparison of the predictive performance of the models. RESULTS: Out of the 2181 COVID-19 patients, 624 died during hospitalization and 1557 recovered. The missing rate was 3 percent for each patient. The mean age of dead patients (71.17 ± 14.44 years) was statistically significant higher than recovered patients (58.25 ± 16.52 years). Based on RF, 10 features with the highest relative importance were selected as the best influential features; including blood urea nitrogen (BUN), lymphocytes (Lym), age, blood sugar (BS), serum glutamic-oxaloacetic transaminase (SGOT), monocytes (Mono), blood creatinine (CR), neutrophils (NUT), alkaline phosphatase (ALP) and hematocrit (HCT). The results of predictive performance comparisons showed GNAM with the mean accuracy, F1-score, and mean AUC in the test dataset of 0.847, 0.691, and 0.774, respectively, had the best performance. The smooth function graphs learned from the GNAM were descending for the Lym and ascending for the other important features. CONCLUSIONS: Interpretable GNAM can perform well in predicting the mortality of COVID-19 patients. Therefore, the use of such a reliable model can help physicians to prioritize some important demographic and clinical biomarkers by identifying the effective features and the type of predictive trend in disease progression.

Application of machine learning models based on decision trees in classifying the factors affecting mortality of COVID-19 patients in Hamadan, Iran.

BACKGROUND: Due to the high mortality of COVID-19 patients, the use of a high-precision classification model of patient's mortality that is also interpretable, could help reduce mortality and take appropriate action urgently. In this study, the random forest method was used to select the effective features in COVID-19 mortality and the classification was performed using logistic model tree (LMT), classification and regression tree (CART), C4.5, and C5.0 tree based on important features. METHODS: In this retrospective study, the data of 2470 COVID-19 patients admitted to hospitals in Hamadan, west Iran, were used, of which 75.02% recovered and 24.98% died. To classify, at first among the 25 demographic, clinical, and laboratory findings, features with a relative importance more than 6% were selected by random forest. Then LMT, C4.5, C5.0, and CART trees were developed and the accuracy of classification performance was evaluated with recall, accuracy, and F1-score criteria for training, test, and total datasets. At last, the best tree was developed and the receiver operating characteristic curve and area under the curve (AUC) value were reported. RESULTS: The results of this study showed that among demographic and clinical features gender and age, and among laboratory findings blood urea nitrogen, partial thromboplastin time, serum glutamic-oxaloacetic transaminase, and erythrocyte sedimentation rate had more than 6% relative importance. Developing the trees using the above features revealed that the CART with the values of F1-score, Accuracy, and Recall, 0.8681, 0.7824, and 0.955, respectively, for the test dataset and 0.8667, 0.7834, and 0.9385, respectively, for the total dataset had the best performance. The AUC value obtained for the CART was 79.5%. CONCLUSIONS: Finding a highly accurate and qualified model for interpreting the classification of a response that is considered clinically consequential is critical at all stages, including treatment and immediate decision making. In this study, the CART with its high accuracy for diagnosing and classifying mortality of COVID-19 patients as well as prioritizing important demographic, clinical, and laboratory findings in an interpretable format, risk factors for prognosis of COVID-19 patients mortality identify and enable immediate and appropriate decisions for health professionals and physicians.