Statistical Analysis of Partner Concurrency Using Retrospective Sexual History Survey Data

• Activity Number: 688
• Type: Contributed
• Date/Time: Thursday, August 13, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Hilary Aralis* and Pamina Gorbach and Ron Brookmeyer
• Companies: UCLA and UCLA and UCLA
• Keywords: Multistate Modeling ; HIV/AIDS ; Concurrency ; Disease Modeling ; Incomplete Data

Abstract

The extent to which sexual partner concurrency drives the HIV epidemic within a community and the effectiveness of potential interventions can be explored using retrospective partnership information collected from a cross-sectional survey of individuals. Data are frequently analyzed at the partnership level requiring an independence assumption and resulting in issues of both censoring and truncation. We propose an alternative approach that employs a multistate modeling framework with each state being defined as the number of ongoing partnerships engaged in by an individual at a given time. At the individual level, the proposed model constitutes a continuous-time Markov process in which transition probabilities are allowed to depend on the current state and explanatory covariates. The proposed method has several advantages including the joint modeling of the concurrent partnership and duration distributions and greater flexibility in accounting for variation within and between individuals. A newly developed method for incorporating incompletely observed state data is also presented. We fit this model to data from an epidemiological study of men having sex with men in Los Angeles.