Multivariate Modeling of Biomarkers for Cross-Sectional HIV Incidence Estimation

• Activity Number: 76
• Type: Contributed
• Date/Time: Sunday, August 9, 2015 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Jacob Konikoff* and Ron Brookmeyer
• Companies: UCLA and UCLA
• Keywords: HIV ; Incidence ; MCMC ; Multivariate modeling

Abstract

Monitoring of the HIV epidemic depends critically on the ability to measure new infections in the population. Reliable estimates of incidence, the rate at which new infections emerge from the susceptible part of the population, additionally empower researchers seeking to target and evaluate prevention efforts. In order to estimate incidence from a cross-sectional study one must model the probability that persons will maintain markers of recent infection t years after seroconversion. We present a novel approach for modeling this probability curve based on a joint multivariate model of the continuous underlying biomarker-based assays. We discuss the advantages and disadvantages of this approach and conclude that this approach has unique utility when trying to set specific cutoffs for recent infection with biomarker-based assays.