Published Paper Details:

Chronic Kidney Disease (CKD) remains one of the most serious complications observed among people living with HIV, affecting anywhere between 4% and 28% of the population depending on geographic and demographic factors. The early detection and accurate staging of CKD in this vulnerable group are critical--not only to slow disease progression but also to tailor antiretroviral regimens that minimize nephrotoxic burden. In this study, we explore how two unsupervised machine learning algorithms--K-means and hierarchical agglomerative clustering--can stratify a real-world cohort of 400 HIV-infected patients into clinically meaningful CKD stage groups without requiring pre-labeled training data. We describe the full analytical pipeline from raw data ingestion through preprocessing, feature engineering, dimensionality reduction, and clustering validation. Using a dominant-class mapping strategy, K-means achieved an accuracy of 78.25% and hierarchical clustering reached 77.50% when evaluated against clinical classification labels. While neither algorithm replaces traditional GFR-based staging, both reveal clinically coherent patient subgroups and offer a promising foundation for decision-support tools in resource-limited healthcare settings. We also present graphical outputs including confusion matrices, performance comparisons, and cluster distribution analyses to make the results accessible to both clinical and computational readers.