MOAE0303 | EXPLORING THE ROLE OF MACHINE LEARNING IN ENHANCING HIV CARE CONTINUITY: A STUDY OF THE “PAMOJA KUNDINI” INTERVENTION IN GEITA

Ambrose Ibihya | Health for A Prosperous Nation (HPON), Dar Es Salaam, Tanzania

The integration of machine learning (ML) in HIV service delivery offers promising avenues to improve the continuity of care for people living with HIV (PLHIV), especially in resource-constrained settings. The “Pamoja Kundini” intervention is a multi-component strategy implemented in Geita Region, Tanzania, that uses ML-based predictive models to identify PLHIV at risk of treatment disengagement to participate in tailored HIV counseling paired with conditional financial incentives in an ongoing randomized controlled trial. This sub-study used qualitative methods to explore the experiences, perceived impacts, and contextual challenges of ML usage in routine HIV care settings. Using a purposive sampling approach, in-depth interviews (IDIs) were conducted with key informants including PLHIV, healthcare providers across four HIV care facilities, and government stakeholders. Data were transcribed and analyzed using content analysis in Dedoose software. The study focused on participants’ experiences with ML, its influence on client engagement, and perceptions of fairness, data privacy, and integration into routine care practices. Overall, 42 participants participated in the qualitative study. Participants acknowledged the utility of ML in reducing bias in determining who is selected for supportive care interventions and accelerating the identification of people at risk of disengagement. However, several challenges emerged. Some healthcare providers expressed fear of job displacement due to automation from ML approaches, while PLHIV voiced concerns over privacy and lack of transparency in data usage. Moreover, all participants voiced concern about the algorithm’s ability to capture contextual barriers that may make PLHIV at risk for disengagement, such as seasonal mobility and socio-economic hardships. Despite these concerns, there was broad agreement that ML could complement, rather than replace, routine care provided by healthcare workers, provided that ethical safeguards and human oversight were implemented. The use of ML as a part of the Pamoja Kundini intervention shows both the benefits and challenges of using digital technology in HIV care. However, its success depends on building trust and ensuring clear communication among healthcare providers and patients to address concerns about data usage and privacy, while also aligning with the local cultural context.