Novel Algorithms for Predictive Analytics on Retention in the HIV Patient Care Continuum
Abstract – Retention of individuals living with HIV on the HIV Care Continuum is not only critical to public health initiatives addressing the HIV epidemic, but also to individual health outcomes. Despite this, less than 50% of people living with HIV are classified as retained in care. With no existing system to identify patients that fail to attain follow up treatments, machine learning algorithms are being rapidly developed to produce predictive analytics on the HIV Care Continuum. This study investigates novel algorithms that provide predictions pertaining to retention of patients in HIV care. A total of five peer-reviewed studies were selected from research search engines to provide a comprehensive review of algorithms in this field, as well as a comparison to identify ideal models that can optimize the current care system for HIV. Models analyzed and tested in this study include random forest, logistic regression, linear regression, XGBoost and Bayesian networks. Such models hold high potential for identifying risk factors to HIV retention and should be integrated into the current healthcare system to optimize domestic and international responses to the HIV epidemic.