Predicting Congestive Heart Failure Risk Factors in King Abdulaziz Medical City A Machine Learning Approach

Abstract

Congestive heart failure (CHF) is one of the diseases with a high burden on the healthcare systems. Patients visits and follow-up at the out-patient clinics are associated with high direct and indirect costs and affect the patient treatment outcomes. In this study, we have tried to test and use machine learning models to predict the risk level and class of CHF patients to confidently extend the timing for the next out-patient cardiac clinic visit. The data for 700 patients’ records were statistically analyzed with Waikato Environment Knowledge Analysis version 3.9.4 (Weka) using eight different machine learning models. Among the eight tested models, the Random Forest and Logistic regression models were found to be the best. Overall performance of the models was promising with these excellent measures (Precision, Recall, F-measure, and ROC) for the Random Forest and Logistic regression models with high accuracy around 0.89. Future work with more balanced datasets and records are needed to test such models which could save the healthcare systems a lot of direct and indirect costs and improve patients’ outcomes.