Performance Analysis of Selected Decision Tree Algorithms for Predicting Drug Adverse Reaction among COVID-19 Hospitalized Patients

Nopour, Raoof and Mashoufi, Mehrnaz and Amraei, Morteza and Saki, Mojgan (2022) Performance Analysis of Selected Decision Tree Algorithms for Predicting Drug Adverse Reaction among COVID-19 Hospitalized Patients. ournal of Medicinal and Chemical Sciences.

Preview

Text 59f.pdf Download (1MB) | Preview

Abstract

Increase in drug allergies and unpleasant adverse effects caused by COVID-19 medication therapies has doubled the need for computing technologies and intelligent systems for predicting poor medication outcomes. This study aimed to construct machine learning (ML) based prediction models to better predict adverse drug effects among COVID-19 hospitalized patients. In this retrospective and single-center study, 482 hospitalized COVID-19 patients were used for analysis. First, the Chi-square test was employed to determine the most critical factors predicting adverse drug effects at P<0.05. Second, the four selected decision tree (DT) algorithms were applied to implement the model. Finally, the best DT model was acquired for predicting adverse drug effects using various performance criteria. This study showed that the 18 variables gained the Chi-square at P<0.05 as the most important factors predicting adverse drug reactions. Besides, comparing the performance of selected algorithms demonstrated that generally, the J-48 algorithm with F-Score=94.6% and AUC=0.957 was the best classifier predicting adverse drug reactions among hospitalized COVID-19 patients. Finally, it found that the J-48 algorithm enables a reasonable level of accuracy in predicting the risk of harmful drug effects among COVID-19 hospitalized patients. It potentially facilitates identifying high-risk patients and informing proper interventions by the clinicians.

Actions (login required)

View Item