Using Machine Learning to Predict Drug-Related Problems for Diabetes Patients

Drug-related problems issue is considered an important issue in healthcare systems worldwide. It appears commonly with diabetic patients, which leads to undesired healthcare complications. In this study, we aim to experiment some machine learning methods to predict the Drug-Related Problems Status (DRPS) for diabetes patients. Another goal is to compare the results of the machine learning methods with the statistical methods to figure out the best results. The dataset used in this paper was collected from outpatient clinics for cardiovascular, endocrine, and respiratory diseases from six main hospitals in different regions in Jordan. The collected data has been prepared and preprocessed and then used to build different machine learning and statistical models. Several experiments have been conducted to evaluate the models. The results of the experiments showed that the random forests machine learning algorithm achieved the highest classification accuracy for predicting the DRPs status. When comparing the machine learning algorithms with the logistic regression using the original non-preprocessed dataset, the findings showed that the random forests method outperformed the logistic regression classifier for both sensitivity and specificity.