Ballot down to Grahame's sign-off, FHIR-to-OMOP full-schema coverage, and Termbox meets Pathling on relational terminology — Aug 11, 2026
Ballot submission — one checkbox from Grahame away
The submission checklist is nearly complete. Two items are still open: the FMG IG-proposal link and the product-director version approval from Grahame Grieve. Gino's read on the FMG item is that the August 5th minutes — where the NIB and ballot were approved and the group noted the IG proposal was missing — already count as approval, and he'd link those minutes as evidence.
I looked through the minutes and they said on the 5th, when we approved this, that we didn't have an IG proposal and we would create one. And so that's why we get to count that vote as FMG approving the IG proposal.
Gino had already pinged Grahame on Zulip asking for product-director approval and hadn't heard back. John's plan is to hit Grahame at every known address — healthintersections, HL7, and Gmail — copy Lynn, and ping him on Zulip separately. Gino's PR #401 (UK → US English spellings, materialised → materialized) is his final QA pass and is ready to sign off once John reviews. Beyond that, John plans to email Lynn with the completed checklist status — everything ticked off except the one waiting on Grahame — and Lynn will move the ballot into publication when the last box is checked.
Post-ballot publishing and branching
Lynn publishes directly from the repo — the CI build and branch structure are what get published, not a zip. The ballot will get a versioned URL under HL7's directory (roughly hl7.org/fhir/uv/sql-on-fhir/3.0-ballot1), and because this is the first publication, Grahame creates the history page. Gino confirmed that once history exists, subsequent versions are auto-generated. The team's plan is to freeze main until the ballot is cut, then split ongoing work into branches so V3 final reconciliation stays clean.
Ballot reconciliation and next-version work
During the ballot period the group won't resolve Jira tickets one by one — they'll accumulate proposed dispositions and then do a block vote, which requires a one-week review window but can be handled on a regular call. Work on features beyond the current ballot goes into a separate branch so it doesn't get folded into V3 reconciliation. For pre-ballot incubation, the plan is to do the work in GitHub and file one or more Jira tickets at the end summarising the changes, rather than tracking every increment in Jira from the start.
FHIR-to-OMOP: full schema coverage and Vulcan interest
John has been building out full OMOP schema coverage from FHIR, reproduced one paper end-to-end, and has a pipeline running to reproduce five more. The remaining gap is OMOP tables that MIMIC-4 doesn't cover — cost tables and the like.
I've been basically getting full coverage over the entire OMOP schema, and I've done a successful reproduction of one paper, and I have this pipeline that's going on to do five more papers.
John plans to bring in Nikolai's test suite as a submodule with a harness wrapper and open-source the whole thing. Nikolai had presented the test suite to the Vulcan FHIR-to-OMOP working group the previous Friday and got positive interest; he now plans to split the repo into three parts — test suite, profiles, and transformation — so external contributors can drop tests into the first without pulling in the rest. The output of John's work is a set of ViewDefinitions and SQL queries; whether those could eventually live inside the Vulcan IG is a longer-term goal, complicated by the fact that John's scope is broader than what Vulcan currently implements.
Terminology as SQL tables — Termbox and Pathling converge
Nikolai wants to represent value sets and concept maps as relational tables — each concept or translation as a row — and is thinking about the shape: per-value-set table, or a generic table with value_set_url and version columns. The bigger open question is whether the tabular view should be materialised or kept virtual, since some value sets can't be expanded and have to stay dynamic.
John flagged that Pathling recently shipped a local terminology feature with exactly this shape — concepts, descriptions, properties, relationships, value-set expansions, and a closure table — enough surface to implement value-set membership, subsumption, and concept translation over ordinary SQL. He offered to pull the schemas and send them to Nikolai. Nikolai's reaction: Termbox's logical core is nearly the same shape, both inspired by Athena, and the convergence isn't coincidence.
If you're thinking Athena way about FHIR, you're getting almost to the same schema.
The follow-through Nikolai wants is bigger than a schema alignment: open-source Health Samurai's terminology loading pipelines, serve the flattened content — open terminologies plus value sets grabbed from IGs — via S3, and let implementers pick it up without running an expansion server themselves. Working name from three years ago: Babylon. The framing is an Athena-like ecosystem for FHIR terminology.
John's guardrail is that any tabular representation will be a subset — no full SNOMED post-coordination, no UCUM grammar, no BCP 47 open-ended values, no MIME-parameter parsing. The goal is roughly 80% of practical use cases, with the top-N common codes for the grammar-based systems shipped as static tables.
We kind of draw a line around what we think is the 80% and say this is going to be an alternate format. It has these features, it doesn't have these features.
Nikolai created a private Zulip chat to continue the terminology-in-database thread and told the group they're free to invite others. Follow-up actions: John sends the Pathling local-terminology schemas to Nikolai for a side-by-side, and Nikolai will connect John with Orlande to discuss publishing HS's terminology content in a standardised tabular format.
Action items
- Gino — email Grahame and ping on Zulip for product-director approval; copy Gmail as well as HL7 addresses
- John — email Lynn with checklist status; follow up on ballot submission
- Gino — get PR #401 (US English spellings) reviewed and merged; John is on review
- John — pull Pathling local-terminology table schemas and send to Nikolai for the Termbox comparison
- Nikolai — connect John with Orlande to discuss standardised tabular terminology, potentially via S3