AI Uncovers Psychosocial Pathways to Cardiometabolic Multimorbidity

The link between social disadvantage and multimorbidity is well-established, yet the specific pathways connecting social conditions to disease burden remain poorly understood. To shed light on these complex relationships, researchers developed an AI-driven multimodal mediation framework. This framework integrates a wide array of health data, including socioeconomic, psychosocial, clinical, laboratory, behavioral, and genomic information, sourced from the extensive All of Us Research Program. The methodology involved using modality-specific variational autoencoders to derive latent representations for each data domain. Subsequently, mediation analyses were performed within this latent space to evaluate indirect associations between socioeconomic disadvantage, psychosocial factors, and multimorbidity. The study cohort comprised 20,804 participants with complete multimodal data, allowing for a comprehensive analysis. The findings revealed that mediation signals were concentrated within a small number of latent dimensions across 800 exposure-mediator-outcome combinations. The most significant indirect association identified linked a socioeconomic disadvantage dimension, a psychosocial vulnerability dimension, and a cardiometabolic multimorbidity dimension. The psychosocial dimension was characterized by factors such as poorer mental health, increased loneliness, lower social well-being, and reduced health literacy, while the outcome dimension included conditions like hypertension, diabetes, and obesity. Bootstrap analyses confirmed the stability of this leading pathway, suggesting that psychosocial vulnerability plays a crucial role in mediating the relationship between socioeconomic disadvantage and cardiometabolic multimorbidity.