Related Experiment Videos
A Machine Learning-Guided Lyric Analysis Peer Support Intervention (LAPSI) for Psychological Distress in South
Lucas Banda1,2, Mathildah Mokgatle1, Olanrewaju Oladimeji3,4
1Department of Public Health, Faculty of Healthcare Sciences, Sefako Makgatho Health Sciences University, Pretoria, Gauteng, South Africa.
Background:
Psychological distress disproportionately affects women in South Africa and is shaped by complex structural, socioeconomic, and psychosocial determinants, particularly within the context of HIV. Despite policy commitments to integrate mental health services into existing health systems, current interventions are limited in their capacity to address these intersecting determinants and are frequently insufficiently targeted, contextually responsive, or inclusive. Data-driven strategies that identify subgroup-specific vulnerabilities may improve the design of contextually relevant and culturally grounded psychosocial interventions.
Objective:
This study seeks to develop and implement a machine learning (ML)-informed lyric analysis peer support intervention (LAPSI) and to evaluate its preliminary impact, feasibility, and acceptability among South African women living with or at heightened risk of HIV.
Methods:
An exploratory sequential mixed methods design is used across five phases: (1) a scoping review of mental health interventions in sub-Saharan Africa; (2) ML-based risk profiling using national survey data with Shapley additive explanations (SHAP)-based interpretability; (3) intervention development and implementation; (4) a single-arm pre-post quantitative assessment of psychological distress using the Kessler 10-Item Psychological Distress Scale (K10), without a comparison group; and (5) qualitative evaluation of feasibility, usability, and acceptability.
Results:
Secondary data analyses were conducted between October 2024 and April 2025. LAPSI was implemented between May and August 2025 in 2 provinces. Outcome assessments were completed in August 2025 (Eastern Cape) and September 2025 (Gauteng), and qualitative interviews were conducted 5 to 7 days after the intervention. A total of 30 participants were recruited and completed the intervention. Quantitative and qualitative analyses were ongoing at the time of manuscript submission, with full findings anticipated to be reported in 2026 and 2027.
Conclusions:
This protocol describes a potentially scalable framework for integrating explainable ML with culturally grounded, peer-delivered intervention design to address psychological distress among women in HIV-affected settings.
International Registered Report Identifier (Irrid):
DERR1-10.2196/90426.