Hidden Markov Models to Detect SMS Survey Compliance Patterns of Youth at High Risk for Acquiring HIV Enrolled in an MHealth Intervention

• Activity Number: 540 - Statistical Methods for Adolescent HIV Trials
• Type: Topic Contributed
• Date/Time: Thursday, August 6, 2020 : 1:00 PM to 2:50 PM
• Sponsor: Biometrics Section
• Author(s): Warren Comulada* and Heather J Gunn and Dallas Swendeman
• Companies: UCLA and UCLA and UCLA
• Keywords: Hidden Markov Model; mHealth; SMS; text message; HIV; self-monitoring

Abstract

mHealth interventions for youth increasingly capitalize on the ubiquity and high uptake of social media and digital communication tools. Despite the promise of mHealth interventions, barriers to intervention delivery remain. A motivating example is provided by a study that aimed to prevent HIV among youth at high risk for acquiring HIV, including men having sex with men, transgender women, homeless and substance-using youth. The sample includes 914 youth, ages 14 to 24 who received weekly short message service (SMS) surveys on their mobile phones to help them self-monitor risk behaviors and stay engaged in the study. About half of the participants responded to surveys. It is important to understand the characteristics of non-responders for future intervention development. In this vein, we applied a hidden Markov model to the data to understand response patterns. Three clusters emerged: active responders who filled out most weekly surveys, intermittent responders, and inactive participants who rarely filled out surveys. A key predictor of cluster membership was whether the participant had ever been homeless.