Adaptive HIV Classification in Malawi Using Particle Swarm Optimization

• Activity Number: 246
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 2:00 PM to 5:50 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Sitaram Vangala* and Chi-hong Tseng and Weng Kee Wong and Risa Hoffman and Alan Schooley
• Companies: UCLA and UCLA and UCLA and UCLA and UCLA
• Keywords: HIV ; Malawi ; Particle Swarm Optimization ; Minimax ; Interval Estimation

Abstract

Initiation of antiretroviral therapy (ART) in HIV patients is recommended by the World Health Organization if a patient's CD4 count falls below 500. We hypothesized that variability in CD4 counts results in frequent misclassification of patients close to this threshold. CD4 counts for 50 patients in Malawi were repeatedly measured to estimate the rate of error. We propose to improve practice by averaging multiple measurements. Limited health system capacity in Malawi entails that follow-up measurements be reserved for patients at increased risk of misclassification. To this end, we compute optimal intervals within which further measurements should be taken. We pose the problem as one of minimizing the maximum misclassification risk, subject to a capacity constraint. High dimensional minimax problems of this kind present computational difficulties, since the gradient is generally undefined. Particle swarm optimization (PSO) is a nature-inspired metaheuristic, which does not presuppose differentiability. We use PSO to solve the decision problem, yielding an adaptive classification scheme for ART.