CQL Foundry

A consent-safe, rule-first blueprint for turning clinical notes into auditable CDS logic with LLMs.

Why I Wrote This

LLMs are good at reading clinical notes, but clinical decision support has to be predictable, reviewable, and respectful of patient consent. CQL Foundry is a concept paper on how to get both: use LLMs to help draft decision logic, and make sure what actually runs is deterministic Clinical Quality Language (CQL) that clinicians and informaticians can inspect.

The Idea

  • Rules come first. The LLM proposes, and the CQL artifact is what gets executed and audited.
  • Consent and sensitivity checks sit in the pipeline from the start, building on my SHARES consent engine work.
  • Every step from note to CDS artifact leaves a trail a reviewer can follow.