Improving retention: predicting ART retention among PLHIV using a supervised machine learning approach

Poor ART care retention due to loss to follow up have dire clinical consequence including poor treatment outcomes among HIV+ individuals and increased transmission rates in populations especially in sub-Saharan Africa (notably Nigeria) and has long beleaguered goals to achieve HIV global targets. Several studies have been conducted to predict or prevent this loss. Machine learning, a field in artificial intelligence, employs statistical, probabilistic and optimization techniques for pattern recognition applied for predicting future events from large (and often complex) data sets and has been used successfully in business, engineering and health analytics. Owing to the need to reduce loss to follow up, the objective of this study is to predict candidate HIV+ individual who would potentially be lost to follow up when they are still enrolled in ART care using a computational approach.