Chronic Kidney Disease Stage Identification In HIV Infected Patients Using Machine Learning

Abstract

Chronic Kidney Disease (CKD) is a chronic and progressive medical condition that is often not associated until it is in its advanced stages, which consequently leads to extreme morbidity and increased mortality. The condition aggravates this risk in individuals who have contracted Human Immunodeficiency Virus (HIV) because of the long-term use of antiretroviral drugs, persistent systemic inflammation, and co morbid metabolic disturbance. In turn, timely diagnosis of CKD and accurate staging of the disease is of utmost importance in order to support the process of clinical intervention. The current study suggests a machine-learning-based decision support system in identifying CKD stages in HIV-infected patients using routine clinical and laboratory parameters. The system uses the supervised learning algorithm such as Support Vector Machine, Random Forest, Decision Tree, and XGBoost to classify the renal status as either normal or abnormal. Meanwhile, the estimated Glomerular Filtration Rate (eGFR) is determined by CKD-EPI equation and it is further matched with standard CKD levels, that is, Stage 1 to Stage 5. Experimental evaluation shows that models built in the form of an ensemble achieve better predictive accuracy with the Random Forest algorithm having the highest accuracy, precision and recall rates compared to the other analyzed models. The proposed approach will increase the reliability and interpretability of diagnostics by integrating data-driven machine -learning predictions with clinically validated eGFR-based staging. The resulting system is likely to provide healthcare professionals with an opportunity to screen CKD in its early stages, to reduce the diagnostic time, and to facilitate evidence-based decisions, especially in relation to the high-risk group of HIV-positive patients.