Abstract: Using Artificial Intelligence to Augment Network-Based, HIV Prevention for Youth Experiencing Homelessness: Results from a Large Community-Based Trial (Society for Social Work and Research 25th Annual Conference - Social Work Science for Social Change)

All live presentations are in Eastern time zone.

Using Artificial Intelligence to Augment Network-Based, HIV Prevention for Youth Experiencing Homelessness: Results from a Large Community-Based Trial

Schedule:
Wednesday, January 20, 2021
* noted as presenting author
Eric Rice, PhD, Associate Professor, University of Southern California, Los Angeles, CA
Laura Onasch-Vera, MSW, Project Specialist, University of Southern California, Los Angeles, CA
Graham DiGuiseppi, ScM, PhD Student, University of Southern California, Los Angeles, CA
Chyna Hill, MA, Student, University of Southern California, Woodland Hills, CA
Robin Petering, PhD, Founder, Senior Researcher, Lens Co, Los Angeles, CA
Nicole Wilson, MSW, Project Administrator, University of Southern California
Nicole Thompson, MSW, Research Associate, Lens Co
Darlene Woo, MSW, Clinical field faculty, University of Southern California, los angeles, CA
Milind Tambe, Professor, University of Southern California, Los Angeles, CA
Bryan Wilder, doctoral student, University of Southern California, los angeles, CA
Amulya Yadav, MS, Doctoral student, University of Southern California, los angeles, CA
Background:

Each year, there are nearly 4 million youth experiencing homelessness (YEH) in the United States with HIV prevalence ranging from 3 to 11.5%. Peer change agent (PCA) models for HIV prevention have been used successfully in many populations, but there have been notable failures. In recent years, network interventionists have suggested that these failures could be attributed PCA selection procedures. The change agents themselves who are selected to do the PCA work can often be as important as the messages they convey. To address this concern, we tested a new PCA intervention for YEH, with three arms: (1) an arm using an Artificial Intelligence (AI) planning algorithm to select PCA, (2) a popularity arm, the standard PCA approach, operationalized as highest degree centrality (DC), (3) an observation only comparison group (CG).

Methods:

714 YEH were recruited from 3 drop-in centers who provide food and crisis management services to YEH in Los Angeles, CA. Youth were consented and completed a survey which collected self-reported data on HIV testing, condom use, HIV knowledge, and social network information. Each arm (AI, DC, and CG) was deployed at each agency with at least 6 months between deployments to allow for new youth to enter the networks at these agencies. 472 youth (66.5% retention) were interviewed immediately after intervention deployment (1 month post-baseline) and 415 youth (58.5% retention) were interviewed 3 months post-baseline. Network data was used to select peer leaders for the DC arm and for the AI arm. In each intervention arm (AI and DC) 20% of youth were selected as PCA’s, given a 4 hour initial training, followed by 7 weeks of 1 hour follow up sessions. Youth disseminated messages promoting HIV testing and condom use.

Results:

Using GEE models, there was a significant change over time (p<.01) and a significant time by arm interaction (p<.05) for conodmless anal sex, with AI showing the most rapid decrease in condomless anal sex. The AI arm reported 24.0% at baseline, 16.8% at 1 month and 14.0% at three months, relative to DC which reported 38.5%, 36.9% and 29.5%, and relative to CG which reported 21.4%, 20.9% and 23.5%. There was a significant change over time (p<.01) for condomless vaginal sex, but no arm interaction. Likewise, there was a increase in new HIV testing across all three arms (p<.01) but no arm interaction. There was a significant increase in HIV knowledge over time (p<.01) and a significant arm by time interaction (p<.01) with AI having the most increase in knowledge.

Conclusions:

PCA models that promote HIV testing, HIV knowledge and condom use are efficacious for YEH. Both the AI and DC arms showed improvements over time. AI-based PC selection, however, led to better outcomes and increase the speed of intervention effects, as the changes in behavior observed in the AI arm occurred by 1 month, but not until 3 months with DC. Given the transient nature of YEH and the high risk for HIV infection, more rapid intervention effects are desirable.