Generative Model Integrates Social Determinants of Health for Disease Reasoning
A new generative model proposes integrating ICD-coded proxies of social determinants of health (SDoH) into disease reasoning. The model uses a conditioned latent diffusion framework to connect multi-organ sensor data with tokenized healthcare events, aiming to improve personalized disease modeling and clinical decision support. It introduces a geometric diffusion model to characterize the temporal evolution of co-morbidities.
Key facts
- The model uses ICD-coded proxies of SDoH from chapters Z and V–Y in ICD-10.
- It is a conditioned latent diffusion framework.
- It connects multi-organ sensor data with tokenized healthcare events.
- A novel geometric diffusion model characterizes temporal evolution of co-morbidities.
- The approach addresses the lack of explicit SDoH modeling in existing generative models.
- The model is designed for in silico disease reasoning.
- It aims to improve personalized disease modeling and clinical decision support.
- The paper is published on arXiv with ID 2605.09771.
Entities
Institutions
- arXiv