CROI 2025 Abstract eBook

Abstract eBook

Poster Abstracts

1313 Enhanced Language Models for Predicting and Understanding HIV Care Disengagement Junzhe Shao 1 , Waverly Wei 2 , Qiuran Lyu 1 , Rebecca Hemono 1 , Xinwei Ma 3 , Joseph Giorgio 1 , Zeyu Zheng 1 , Emmanuel Katabaro 4 , Matilda Mlowe 4 , Amon Sabasaba 4 , Caroline Lister 5 , Siraji Shabani 6 , Prosper Njau 7 , Sandra I. McCoy 1 , Jingshen Wang 1 1 University of California Berkeley, Berkeley, CA, USA, 2 University of Southern California, Los Angeles, CA, USA, 3 University of California San Diego, La Jolla, CA, USA, 4 Health for a Prosperous Nation, Dar es Salaam, United Republic of Tanzania, 5 Dodoma Referral Hospital, Dodoma, United Republic of Tanzania, 6 Tanzania Ministry of Health, Community Development, Gender, Elderly, and Children, Dodoma, United Republic of Tanzania, 7 National AIDS, STIs and Hepatitis Control Program (NASHCoP), Dodoma, Tanzania Background: Despite significant progress in HIV treatment, maintaining patient engagement in care remains a critical challenge. While UNAIDS aims to achieve the "95-95-95" targets by 2030, many people living with HIV (PLHIV), particularly in sub-Saharan Africa, tend to discontinue care within two years. Current one-size-fits-all interventions often lack efficient targeting, and resources could be better allocated to those at the highest risk. The increasing availability of electronic medical records (EMRs) in sub-Saharan Africa, combined with advancements in AI and ML techniques, offers promising solutions for predicting which PLHIV may require additional support. Methods: We developed a novel AI model by enhancing a pre-trained LLM (LLaMA 3.1, an open-source pre-trained LLM from Meta) using routinely collected EMRs from Tanzania's National HIV Care and Treatment Program from January 1, 2018, to June 30, 2023 (4,809,765 records for 261,192 people) to identify people at risk of disengaging from HIV care or developing adverse outcomes. Outcomes included risk of ART non-adherence, non-suppressed viral load, and loss to follow-up. Models were evaluated both internally (Kagera region) and externally (Geita region), with performance compared against state-of-the-art ML models and zero-shot LLMs. Additionally, a team of HIV physicians in Tanzania assessed the LLM's predictions, and LLM provided justifications for a subset of patient records to evaluate their clinical relevance and reasoning. Results: The enhanced LLMs achieved higher AUC than the supervised ML model and zero-shot LLMs in internal and external validation. When focusing on the 25% of PLHIV predicted as most likely to LTFU, the model correctly identified 78% (2,515 of 3,224) of PLHIV genuinely at risk in internal validation and 73% (7,105 of 9,733) in external validation [Figure 1]. Attention score analysis indicated that the enhanced LLM focused on keywords such as gaps in follow-up care and ART adherence. The human expert evaluation showed alignment between clinician assessments and the LLM's predictions in 65% of cases, with experts finding the model's justifications reasonable and clinically relevant in 92.3% of aligned cases. Conclusions: The results underscore the potential of LLMs in providing actionable public health insights that can guide targeted interventions to enhance patient retention. This study represents the first application of fine tuned LLMs in predicting disengagement from HIV care using large-scale EMR data.

1314 Hair Salons as a Promising Space to Offer HIV Prevention Services for Young Women in Lesotho Alain Amstutz 1 , Malena Chiaborelli 1 , Pontšo Sekhesa 2 , Madeleine Sehrt 3 , Tsepang Mohloanyane 4 , Meri Hyöky 2 , Donaldson Conserve 5 , Karen Hampanda 3 , Mamaswatsi Kopeka 5 , for the Hair SALON Collaborative 1 University Hospital Basel, Basel, Switzerland, 2 The HUB, Morija, Lesotho, 3 University of Colorado Anschutz Medical Campus, Aurora, CO, USA, 4 Central University of Technology, Bloemfontein, South Africa, 5 George Washington University, Washington, DC, USA Background: Adolescent girls and young women in southern Africa are disproportionately affected by HIV and sexual reproductive health (SRH) challenges. There is a need for more accessible and de-medicalized community spaces to offer person-centred HIV/SRH services for this key population. We aimed to assess the acceptability and feasibility of offering HIV/SRH services at hair salons in Lesotho. Methods: We used an innovative citizen scientist approach, whereby hair stylists were recruited through social media (https://hairsalonproject.com/), completed their questionnaires, and subsequently recruited their clients aged 15-35 years as respondents. All questionnaires were co-developed with stylists/ clients and a stepwise verification process using GPS, pictures and a local mobile payment system ensured data quality. In addition, we conducted individual in-depth interviews among 14 stylists/clients, following the rapid thematic analysis framework, supported by natural language processing. The protocol is registered (osf.io/VP8WZ). Results: During 4 months (Apr - Jul 2024), we recruited 157 hair stylists (median age 29; [interquartile range 25-33]; 157 salons across all ten districts) and 307 female clients (median age 26 [22-30]). Among stylists, 93.6% were comfortable offering oral HIV self-testing (HIVST), 92.4% pre-exposure prophylaxis (PrEP), and 91.7% post-exposure prophylaxis (PEP). Among clients, the corresponding proportions were 93.5%, 88.3%, 86.4%, respectively. However, immediate demand was 30.8% for HIVST, 22.1% for PrEP and 14.9% for PEP. Acceptability and demand were higher for family planning methods and menstrual health products (Figure 1). Only 10.8% of stylists thought that offering HIV/SRH services would negatively impact their business; 81.5% have a toilet on-site, 21.0% an additional confidential space. The majority of clients visit their salon 1-2 times monthly. Hair salons were more accessible than the nearest healthcare centre in terms of cost and time. The qualitative analysis confirmed the high acceptability of the hair salon as a low-threshold access space and less judgemental than clinics, but raised concerns regarding confidentiality. Conclusions: These findings suggest high acceptability and feasibility of HIV/SRH services at hair salons in Lesotho, providing the basis for a pilot intervention. Further research needs to examine the discrepancy between the hypothetical acceptability and immediate demand, the role of stylists, and individual components of the service package.

Poster Abstracts

1315 Capacity for PrEP Implementation at Pharmacies in Baltimore and San Diego: A Mixed-Methods Study Lipin Lukose 1 , Iheanyichukwu Samuel Onwubiko 1 , Shanaya Sidhu 2 , Terence Hendrix 2 , Amanda Rosecrans 3 , Gabriel A. Wagner 2 , Sheree Schwartz 1 , Christopher Kemp 1 1 The Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA, 2 University of California San Diego, La Jolla, CA, USA, 3 The Johns Hopkins University School of Medicine, Baltimore, MD, USA Background: HIV pre-exposure prophylaxis (PrEP) uptake remains disproportionately low among communities with the highest HIV incidence in the US due to stigma, structural inequities, limited access, and long wait times. Pharmacy-initiated PrEP could dramatically expand access, but pharmacy capacity for PrEP delivery has been understudied. Current legislation allows

CROI 2025 437