Informative Missingness to Generate Irregular Clinical Time Series

In three linesDiffusion-based approach to generate irregular clinical time series by jointly modeling laboratory values and observation patterns. Uses DACMI benchmark from MIMIC-III, extends TimeDiff framework to capture dependencies between patient physiology and clinician testing behavior under MNAR-like missingness.

Summary generated by Claude — human-verified