Five synthetic patients, five rare diseases, one shared probabilistic architecture. The demonstrator shows how observed physiology updates a posterior estimate of current latent disease state, which is then propagated into a posterior predictive trajectory and clinically defined event risk.
Observed latent-state estimate to the vertical divider; posterior predictive median and 50%, 80% and 95% credible intervals to the right. Higher state values represent greater disease burden for this demonstrator.
The posterior narrows as patient-specific observations accumulate. In an implementation, the full joint posterior over latent state, patient-specific parameters and disease parameters would be propagated forward rather than reduced to these display summaries.