Predicting SMS Weekly Survey Compliance Among Youths at High Risk of HIV Infection

Background: HIV interventions using short message service (SMS) have proliferated due to the ubiquity of mobile phones, especially among youth. Frequent SMS communication with healthcare providers offers opportunities for both self-monitoring and for providers to intervene in a timely manner. We developed a weekly SMS 7-item health survey that is being administered to high-risk youth (HRY) as part of an HIV biobehavorial intervention. Few studies have examined predictors of SMS engagement. We fill in literature gaps by examining predictors of compliance to a weekly SMS health survey among HRY.

Methods: HRY (N=478) were recruited from 13 community-based and clinical centers in Los Angeles and New Orleans. Youth are 86% Black or Latino, 80% male, 62% homeless, 40% hospitalized for mental health problems, 24% treated for drug abuse, and 55% had unprotected anal sex in the last year. Three different delivery schedules were examined which included the day of survey delivery and the option to fill out the survey through SMS or an e-mail link. Predictors included sociodemographic characteristics (birth sex, race/ethnicity, age, income level, homelessness history, level of education, and sexual identity status), time in the study, and study design changes designed to increase SMS engagement. Random effects logistic regression was used to evaluate predictors for compliance to filling out weekly SMS surveys. Response rates were evaluated for each study participant under different delivery schedules.

Results: Among the 478 participants who have received at least one SMS, 7616 surveys were sent out over six months and 2289 surveys received responses (response rate=30%). The response rate gradually rose over time to about 50% across youth of different sociodemographic and risk characteristics. Higher response rates were predicted by female gender (adjusted odds ratio [OR]=1.12), minority race/ethnicity (OR=1.09), not having been homelessness (OR=1.16), having some college education (OR=1.11), being non-heterosexual (OR=1.09) and Tuesday versus Sunday SMS delivery (adjusted OR=1.06). The odds of SMS response were significantly lowered by 1% for each week that passed since participant’s enrollment (OR=0.99). Among the SMS responders, 119 participants reported potential HIV-like symptoms at least once via the weekly SMS survey (28%). Youth who reported HIV-like symptoms were 1.2 times more likely to concurrently report unprotected sex (p<0.05) and 1.9 times more likely to have attended a substance abuse treatment program (p<0.05). Within 2 months of implementing outreach, 28 youths who reported HIV-like symptoms were followed up; 18 were evaluated as not at risk, 9 youths were tested with Alere rapid diagnostic test (all were HIV-negative), and one was referred to another provider. However, 15 youth with acute or recent untreated HIV infection have been identified through the study’s routine 4-month follow-up testing or by community-based and clinic partners.

Conclusions: While the study’s routine HIV testing and partner sites have identified 15 acute and recent HIV infections to date, none have been in earliest stages of acute infection. While SMS engaged almost half of participants, a toolbox of various strategies is needed to engage HRY.