An open registry for healthcare quality measures. Publish in CQL, SQL, or SQL‑on‑FHIR, get it validated, and get review from people who have solved the same problem. Today every organization rebuilds this logic on its own.
A PhD candidate writes a CQL library to screen patients for a Hospital at Home program. It runs at 30 to 40 seconds per patient. He needs it to run across 300,000. He has no idea whether the problem is his logic, his engine, or his data.
He gets lucky. His email reaches someone who has spent years optimizing CQL, and that person writes back a detailed review: functions are not cached the way defines are, a retrieve iterates every resource in the bundle every time, this correlated query is really a lateral join running dozens of semi‑joins. Rewrite it this way.
That review answered the question completely. Four people read it. It is not searchable, it is not attached to any measure, and the next person with the same problem will not find it.
This is a great example of why we need an open community for quality where experts can share their expertise, rather than relying on an introduction to the one person who happened to know someone by pure luck. — from the thread that started this
Every payer, health system, and vendor doing quality measurement staffs its own team to implement the same national measures against its own data, then pays auditors to confirm the implementation is right. The same specification ambiguities get rediscovered independently, over and over. The cost is enormous and none of it reaches a patient.
Open Quality is where that answer should live.
A valid manifest, an open license, and at least one artifact. Any language. The bar is low on purpose, so that unfinished work can still be shared.
Machine-readable measure identity, a declared data model, every value set resolvable, and a README that states intent, known limitations, and provenance.
Deep validation passes for every artifact. CQL translates to ELM. FHIR validates against the CRMI and ECR profiles. SQL parses against its declared dialect and ships its schema.
Most organizations that do quality measurement write SQL against a warehouse, not CQL. A level system that required FHIR to reach the top would exclude them.
| Piece | State | What it is |
|---|---|---|
| Package format | shipped | Manifest schema, conformance levels, license policy, content scanning. 91 tests. |
| Validation core | shipped | Validates a package directory and computes its level. Runs on real CMS eCQM content. |
oq CLI | shipped | oq validate and oq pack. Deterministic, content-addressed tarballs. |
| Deep validators | next | CQL‑to‑ELM, FHIR profile validation, SQL parsing, VSAC resolution. Level 2 needs these. |
| Registry | planned | Publish, search, install. Git-authored, registry-hosted, versions immutable. |
| Typed feedback | planned | Questions, interpretation issues, defect reports, implementation notes — attached to a measure, searchable. |
Open licenses only. Every package declares an SPDX license from a short allowlist. Value sets are referenced by OID or canonical URL, never embedded as expansions, which makes the licensing rule enforceable by machine rather than by review.
No HEDIS logic. NCQA holds copyright on those specifications. You can publish your own implementation written against a public specification. You can also publish test cases and review comments. Many employers approve releasing that even when they will not release the SQL.