For the complete documentation index, see llms.txt. This page is also available as Markdown.

Compliance Questions Cookbook

The questions a payer compliance, regulatory-affairs, or product team actually asks about the CMS prior-authorization rulemaking wave — each answered by one query. Questions 1–10 cover the obligations, deadlines, standards, and reporting duties; 11–13 cover the payer directory, the metrics-posting registry, and the posted values; 14–18 cover tracking changes over time. Queries use the DWV schema (Snowflake); on Databricks use the lowercase cms_prior_auth_compliance_dwv schema and lowercase table names. Several of these are the dataset's Snowflake Marketplace example queries (marked ★).

1. What is my full compliance calendar? ★

Every prior-authorization compliance deadline that applies to a given entity type, with the date semantics and how certain each date is.

SELECT
    entity_type_id,
    obligation_id,
    deadline_type,
    compliance_date_anchor,
    compliance_date_display,
    date_certainty
FROM DWV.COMPLIANCE_DEADLINES
WHERE entity_type_id = 'MA_ORG'
ORDER BY compliance_date_anchor NULLS LAST;

2. Everything that applies to me, in one row per obligation

The gold answer table: what applies to my entity type, by which anchor date under which date semantics, at what certainty and binding status.

SELECT
    obligation_id,
    obligation_category,
    applicability,
    compliance_date_anchor,
    compliance_date_display,
    date_certainty,
    binding_status,
    interpretation_confidence
FROM DWV.COMPLIANCE_GOLD
WHERE entity_type_id = 'MEDICAID_MC'
ORDER BY compliance_date_anchor, obligation_category;

3. What must I post publicly on my website, and by when? ★

The annual public prior-authorization metrics payers must post, the required content elements, and the recurring due rule and first due date.

4. What must I report to CMS, and when?

The usage metrics reported directly to CMS (as opposed to posted on the payer website), with their cadence and first due date.

5. What must a denial notice contain? ★

The required content elements of a prior-authorization denial notice and the channels each applies to.

6. Who is subject to the decision timeframes, and who is exempt? ★

Decision-timeframe applicability across entity types for the PA decision-clock obligation, surfacing the explicit exemption (QHP issuers on the FFEs) as its own row rather than a missing one.

7. Which FHIR / Da Vinci implementation guides does each API obligation require?

Each API obligation mapped to its named standards and implementation-guide versions, with the binding strength (required vs named vs proposed).

8. What is explicitly out of scope, and on what basis?

The program and plan types CMS did not include, each with the verbatim citation basis for exclusion — so you can prove a scope determination.

9. What is coming that is not yet binding? ★

The forward radar: proposed obligations flagged as not binding (for example CMS-0062-P drug prior-authorization) so roadmaps anticipate the next requirement.

10. Which implementation-guide versions are expiring?

The standard and IG versions with a status or expiry date on the horizon (for example the STU-2-era stack proposed to expire January 1, 2028) so upgrade planning starts early.

11. Which payers am I benchmarking against?

The payer parent organizations in a segment — for example the Medicare Advantage parents, ranked by enrollment — from the payer directory. Filter payer_kind and state / states_operating for the segment you care about.

12. Where does each major payer post its PA metrics? ★

The verified posting locations for the top payers' CMS-required annual public PA metrics, joined to the directory — only rows where a location was found. The registry gives the posting location; the posted values themselves are in PAYER_METRICS (next question).

13. What PA metric values did each payer post? ★

The posted values themselves — approval and denial rates, decision times, appeals-overturn, request volumes — joined to the metric definitions and payer directory. Every value is human-reviewed before it appears here, and carries the verbatim excerpt from the payer's own posting that licenses it.

To compare one metric across payers, filter on metric_id (for example PA_APPROVAL_RATE_STD) and sort by value_numeric.

Tracking changes over time

This is a maintained reference, not a one-time snapshot, and it tracks change at the row and field level rather than by versioning source files. Three layers work together: FEEDS / FEEDS_FILES tell you which release you are on, RULE_AUDIT_LOG tells you what changed and why, and the SCD2 lifecycle columns (row_version, is_current, curation_valid_from, curation_valid_to) let you reconstruct what the reference asserted on any past date. See How time works in this product for the mental model.

14. What has changed in the reference, and why? ★

The field-level audit trail of every substantive editorial change — corrections, verifications, and supersessions — each with its customer-facing reason. This is the proof the reference is maintained.

To see only what changed since your last sync, add a date predicate — AND changed_at >= '2026-06-01' (your last-checked date) — and drop the change_type filter to include the initial CREATED events as well.

15. Which release am I on, and what shipped in each one?

The full curation-release history from FEEDS_FILES — one row per published release, with is_latest = TRUE marking the release you are currently on. Every data row's feeds_files_id points at the release that shipped it.

16. What did the reference assert on a given date?

Point-in-time reconstruction. Because rows are versioned append-only (never overwritten), filtering an obligation's applicability rows to the versions whose curation-validity interval contains your as-of date returns the reference exactly as it stood then. Replace the date with the as-of date you care about; a date earlier than a row's curation_valid_from returns nothing for that row — the reference had not asserted it yet.

17. What is the full history of one obligation row?

Two queries reconstruct the complete life of a single applicability fact. First, every version of the row from OBLIGATION_APPLICABILITY — the current version plus any superseded predecessors, in order:

Then the field-level audit entries for that same row. In RULE_AUDIT_LOG, row_pk is the row's natural key: for the composite-key applicability table it is obligation_id | entity_type_id (space-pipe-space); single-key tables use the bare primary key.

18. What is currently under verification, and why?

The quarantine ledger from CURATION_STATUS. Rows flagged needs_verification are excluded from COMPLIANCE_GOLD and every derived view until confirmed; this query surfaces them with the reason they are held, how long they have been held (days_in_quarantine), and the current row count of the table they belong to (table_rows_current) so you can read quarantined items as a share of the whole.

Using Snowflake Cortex on the text columns

Much of this dataset is dense free text — verbatim Federal Register rule text (requirement_text_verbatim), payers' free-text service lists (PA_ITEMS_SERVICES_LIST values), exemption notes, and the audit trail's change reasons. Snowflake Cortex AI functions run directly on those columns inside your account — no data movement, no external API keys. Four patterns consumers use:

Cortex AI functions are billed to your account per use, and model availability varies by cloud region — swap in any model available in yours, or enable cross-region inference. AI_EXTRACT and AI_CLASSIFY need no model parameter; AI_COMPLETE takes a model name.

Ask the whole product a question with the semantic view

The share includes the COMPLIANCE_ANALYST semantic view. For fully natural-language querying, point Cortex Analyst at it (Snowsight → AI & ML → Cortex Analyst → DWV.COMPLIANCE_ANALYST). The same semantic model is queryable in SQL with the SEMANTIC_VIEW() construct — no joins to write, relationships and definitions are built in:

Ask the rule text a question

For a compliance analyst: extract a specific answer from the verbatim rule text instead of reading the Federal Register — here, the standard prior-authorization decision deadline.

Normalize payers' free-text service lists

For a benchmarking analyst: every payer phrases its "items and services requiring prior authorization" differently. Classify the free text into comparable categories to build a cross-payer view.

Summarize an obligation for a briefing

For regulatory affairs: one-paragraph summaries of the longest rule-text passages, straight from the verbatim source.

A plain-English obligations register

For consultants and readiness reviews: generate a one-sentence plain-English statement of every obligation that applies to an entity type, joined from the gold table. Uses COMPLETE with a named model — pick one available in your region.


Informational reference compiled from public Federal Register, eCFR, and CMS sources. Not legal advice; consult counsel for compliance decisions.

Last updated