Remove 89 ICD-9 codes filed under ICD-10 fields

[KrishnaTO]

⚠️ Data change to ontologies/ari_t1d.owl and both mapping exports. 89 deletions, no additions.

The rule

Every ICD-10-CM code begins with a letter. A digit-led value in an ICD-10 field is therefore an ICD-9-CM code stored under the wrong vocabulary — 446.1, 720.0 and 390-392.99 are ICD-9; M30.3, M45 and E10 are not.

Where	Removed
ontologies/ari_t1d.owl — ARI_ICD10 values across 53 diseases	61
mappings/ari.sssom.tsv rows	14
mappings/ari.equivalencies.tsv rows	14

Letter-led values were left alone, including the range I00-I02 on Rheumatic fever — that is a separate defect (a range where a single code belongs) and is still reported by the validator.

Judgement calls worth a look

Four diseases now have no ICD-10 code at all, because everything they held was ICD-9: ARI:0001032, ARI:0001033, ARI:0001119, ARI:0001201. Those cells now read as no term recorded, which is the honest state, but it is a visible drop in apparent coverage.

Ten of the fourteen mapping rows were negative judgments, not confirmations — a curator had already reviewed the code and rejected it. Dropping them removes a record. The reasoning for doing so anyway: a rejection of a code that cannot be represented in the target vocabulary carries nothing forward, and nine of the ten already had no stored id to guard. Affected: ARI:0001012, 0001014, 0001061, 0001062, 0001063, 0001065, 0001068, 0001073, 0001074, 0001107.

Three unrelated errors fall out of this

ARI:0001012 -> icd10cm:720.0 was recorded as both confirmed and flagged wrong, and was the one id still stored on its disease after #60 despite being flagged. Both rows and the stored value are ICD-9, so all three go together. Two of the six drifts between the two exports (362.50/362.5 and 720.0/720) go with them.

Note on #57

#57 retired ICD-9 from data/3-meta-database-sources/meta-databases.csv, notebook/ari-grounding/make_match_reports.py and connecting_ontologies.drawio — but it merged into feature/metadata-manager_v2/ARI, not main. None of it is on this branch, so main still advertises ICD9/ICD9CM as sources and still groups ICD-9 into the report's ICD column. This PR covers the data only; porting #57 to main is still outstanding.

Verification

Validated with the checker from #63. Whole-repository errors 193 → 94; every icd9-under-icd10 finding cleared, along with contradiction and flagged-still-stored. The ontology diff is 61 pure deletions, all ARI_ICD10 annotation lines, zero additions, and the file still parses as XML.

Advances #23 ("Validate all diseases with database xrefs"); the other subtasks are untouched, so it is referenced rather than closed.

🤖 Generated with Claude Code