Decision Tree Algorithm for Diagnosis and Severity Analysis of COVID-19 at Outpatient Clinic

This study investigates the feasibility of decision tree algorithm like CART recursive method for classifying participants into test-based positive cases and negative cases to detect COVID-19 in the outpatient and suggest admission or home isolation according to the evaluated parameters. It also evaluates the severity of the outpatients using the values of RTPCR test and Chest X-Ray imaging results. A theoretical and predicted decision tree is proposed in the study after focus group interview with a clinical physician. Primary data was collected from the survey of patients visiting a physician for treatment of COVID-19 during the first wave. CART algorithm was applied for predicting the required decision tree. According to the predicted decision tree, it was determined that the most important feature while treating a COVID-19 patient is their history of contact with the positive coronavirus patient. Based on the valuation of dataset, the predicted decision tree provided similar results to that of the conceptual tree. Thus, comparing both trees, it can be evidently said that the predicted decision tree is a subset of conceptual decision tree and can be used by physicians for diagnosis and severity analysis of COVID-19.