The Hospital Decision Twin: stochastic programming on real operational data

This foundational paper introduces the Hospital Decision Twin as stochastic programming applied to real operational data, rather than a dashboard of point estimates.

It defines the atom of the model, draws the device line that keeps the twin a decision-support asset, and argues that the guardrail — recommender-mode, evidence-based promotion, kill switch — is itself an asset.

The thesis: a twin that outputs distributions and scenarios, calibrated on the hospital's own data, makes operational and strategic decisions defensible.