SVM-enabled prognostic method for clinical decision making: The use of CD4 T-cell level and HIV-1 viral load for guiding treatment initiation and alteration

Papers with focus on biomedical data analysis have investigated the initiation of HIV treatment on the basis of several factors including the patient's clinical and immunological state, mental health and the likelihood of adherence to treatment. Here we focus on a different domain, namely the decision whether to change treatment based on sequential laboratory test indicators of the patient. In this paper, we give attention to CD4 and blood viral load-based parameters since they have been shown to be significant disease indicators. We develop an SVM-based method and show that our method provides results consistent with guidelines built on clinical knowledge of the impact of various factors on treatment response, and more importantly, indicates which factors contribute the most to the decision via bootstrap analysis. Among our findings are that CD4 T-cells and HIV-1 viral load absolute levels are strong contributors, while CD4 T-cell and blood viral load sequential changes are weaker contributors.