Predicting Mortality of COVID-19 Patients based on Data Mining Techniques

Abstract

If Coronavirus (COVID-19) is not predicted, managed, and controlled timely, the health systems of any country and their people will face serious problems. Predictive models can be helpful in health resource management and prevent outbreak and death caused by COVID-19. The present study aimed at predicting mortality in patients with COVID-19 based on data mining techniques. To do this study, the mortality factors of COVID-19 patients were first identified based on different studies. These factors were confirmed by specialist physicians. Based on the confirmed factors, the data of COVID-19 patients were extracted from 850 medical records. Decision tree (J48), MLP, KNN, random forest, and SVM data mining models were used for prediction. The models were evaluated based on accuracy, precision, specificity, sensitivity, and the ROC curve. According to the results, the most effective factor used to predict the death of COVID-19 patients was dyspnea. Based on ROC (1.000), accuracy (99.23%), precision (99.74%), sensitivity (98.25%) and specificity (99.84%), the random forest was the best model in predicting of mortality than other models. After the random forest, KNN5, MLP, and J48 models were ranked next, respectively. Data analysis of COVID-19 patients can be a suitable and practical tool for predicting the mortality of these patients. Given the sensitivity of medical science concerning maintaining human life and lack of specialized human resources in the health system, using the proposed models can increase the chances of successful treatment, prevent early death and reduce the costs associated with long treatments for patients, hospitals and the insurance industry.

Comparing machine learning algorithms for predicting COVID-19 mortality

“…Other study conducted [ 42 ] four ML techniques were trained based on 10,237 patients' data and, finally, SVM with the sensitivity of 90.7%, specificity of 91.4%, and ROC of 0.963% had the best performance. Moulaei et al [ 31 ] also predicted the mortality of Covid-19 patients based on data mining techniques and concluded that based on ROC (1.00), precision (99.74%), accuracy (99.23%), specificity (99.84%) and sensitivity (98.25%), RF was the best model in predicting mortality. After, the RF, KNN5, MLP, and J48 were the best models, respectively [ 31 ]…”

“…Moulaei et al [ 31 ] also predicted the mortality of Covid-19 patients based on data mining techniques and concluded that based on ROC (1.00), precision (99.74%), accuracy (99.23%), specificity (99.84%) and sensitivity (98.25%), RF was the best model in predicting mortality. After, the RF, KNN5, MLP, and J48 were the best models, respectively [ 31 ]…”

“…The selected features are used as inputs for developing ML-based models for severity, deterioration, and mortality of COVID-19 patient risk analysis. The strongest predictive features included basic data such as age (aged) [ 11 , 17 , 28 , 30 , 43 – 46 ], gender (male) [ 10 , 11 , 18 , 27 , 29 , 44 , 46 ], BMI (high) [ 15 – 17 ], type of patient encounter (inpatient vs. outpatient) [ 11 , 23 , 27 , 29 ], occupation (related to healthcare) [ 17 , 23 , 29 , 30 ], clinical symptoms include dyspnea [ 15 , 16 , 23 , 30 , 31 , 44 , 47 ], low consciousness [ 11 , 17 , 18 , 28 ], dry cough[ 15 , 17 , 18 , 23 , 27 , 28 , 44 ] fever [ 11 , 17 , 18 , 43 – 45 , 47 ], para-clinical indicators consisting of spo2 (decreased) [ 16 , 18 , 29 , 45 , 47 ], lymphocyte count (low) [ 10 , 23 , 27 – 29 ], platelet count (low) [ 16 , 27 – 29 , 47 ], leukocyte count (raised) [ 15 , 16 , 27 , 28 , 30 ,…”

“…The predictive classifier models were developed for accurately predicting COVID-19 mortality. In the modeling stage, to select the appropriate ML algorithms, the related studies in this field were reviewed [ 10 , 11 , 15 , 18 , 23 , 27 – 31 ] along with considering the type and quality of the selected dataset. To construct the mortality prediction model, we applied seven ML algorithms including the J48 decision tree, random forest (RF), k-nearest neighborhood (k-NN), multi-layer perceptron (MLP), Naïve Bayes (NB), eXtreme gradient boosting (XGBoost), and logistic regression (LR).…”

Supervised Machine Learning Approach to COVID-19 Detection Based on Clinical Data

“…The LR algorithm is less efficient in their research (F - score = 82%) than in our findings (F - score = 87.5%).Other studies have been performed in Iran to diagnosis COVID-19 using ML algorithms that have used CT images dataset ( 39 ) or routine blood tests ( 40 ). Other studies based on ML algorithms in Iran in the field of predicting mortality ( 41 , 42 ) and intubation prediction ( 43 ) have been performed. The strength of our research compared to other research is the multi-center data set and the large volume of data set.…”