Prior authorization feed: datasets, templates and conventions

[Semalexa]

Summary

Adds the prior authorization feed to the Data Integration Reference: what a payer delivers so that prior authorization history can be served through Patient Access, Provider Access and Payer-to-Payer, in the shape of the Claims and Encounters pages.

• data-integration/prior-authorization/README.md — one family page built to the Da Vinci PDex STU 2.1.0 PriorAuthorization profile: Datasets, Data conventions, and three datasets with column tables, notes and "Set by Payerbox": prior_auths (one row per authorization), prior_auth_items (one row per requested service), prior_auth_documents (one row per document submitted with the request, the file as an attachment).
• Three CSV templates in assets/data-integration/ with example rows: an authorization approved with changes whose lines were decided three ways, a denial with its reason, a pending request, a retraction; their keys resolve against the patients, coverage, practitioners and organizations templates.
• Column names and vocabulary follow the family: record_id, patient_identifier, coverage_id, payer_org_npi, *_npi, *_code with *_code_system, last_updated, is_deleted; Required values are Yes, Recommended, If available, If <condition>, If retracting.
• SUMMARY.md and the Feeds table on data-integration/README.md gain the new family.

The decision model follows the profile: review_action (X12 306) on every line and, for a denial of the whole request, on the authorization; a denial carries a CARC and the reason in words; the line's adjudication entry is chosen by the decision (allowedunits, denialreason, consumedunits), which is what the profile's closed slicing admits.

Test plan

• bun lint passes (only pre-existing, unrelated warnings)
• bun assets:hash run for the three new templates; references updated
• Column set and example rows checked against a working prior-auth pipeline: thirteen scenario deliveries, the negatives failing on their own rule

🤖 Generated with Claude Code