Generating Synthetic Multi-national Longitudinal Cohorts for Clinically Grounded HIV Research.

Area of Science:

• Medical informatics
• Computational epidemiology
• Artificial intelligence in healthcare

Background:

• Longitudinal data for people with HIV (PWH) are crucial for research but difficult to share due to privacy regulations.
• Existing synthetic data methods struggle with the complexity of HIV clinical trajectories, including temporal dynamics and missing data.

Purpose of the Study:

• To introduce Medical Longitudinal latent Diffusion (MeLD), a novel generative model for synthesizing realistic, variable-length, longitudinal HIV cohort data.
• To address challenges in creating privacy-preserving synthetic HIV data, including complex temporal dynamics, mixed data types, and missingness.

Main Methods:

• Developed MeLD, a generative model utilizing latent diffusion to synthesize longitudinal clinical trajectories.
• Applied MeLD to the Caribbean, Central, and South America Network for HIV Epidemiology (CCASAne) cohort, a large international HIV dataset with over 30 years of follow-up.
• Evaluated MeLD against state-of-the-art methods on data utility, fidelity, and privacy.

Main Results:

• MeLD successfully synthesized variable-length, decades-spanning clinical trajectories with missingness, outperforming existing methods.
• The model accurately reproduced longitudinal inference, including time-to-death estimates and risk factor effects, while ensuring strong privacy protection.
• Demonstrated superior performance across data utility, fidelity, and privacy metrics.

Conclusions:

• MeLD provides the first large-scale, openly accessible synthetic longitudinal cohort of people with HIV.
• This resource faithfully preserves real-world data patterns and clinical associations, enabling hypothesis generation and reproducible research.
• MeLD offers an immediately deployable tool for advancing open science and data-driven innovation in HIV research.