PBP snf (cleaned view)

What it is

1. 1A normalized, analyst-ready view of Medicare Advantage (Part C) Skilled Nursing Facility (SNF) cost-sharing and benefit rules, derived from Section B2 of the CMS Plan Benefit Package (PBP) Public Use Files.
2. 2One row tells you, for a given MA plan in a given benefit year/snapshot: whether SNF is offered, the day-interval copay structure (the signature SNF cost ladder), prior-hospitalization requirements, and auth/referral rules.
3. 3It is the SNF slice of the medicare_semantic PBP family of benefit views; use it to compare SNF out-of-pocket exposure across MA plans/years.

Grain

How to read it

1. 1Compare in-network SNF out-of-pocket exposure across MA plans for a given year.
2. 2Example: SELECT contract_id, plan_id, segment_id, tier1_copay_interval1_amount AS days_1_20_copay, tier1_copay_interval2_amount AS days_21_100_copay, prior_hospitalization_required FROM medicare_semantic.vw_pbp_snf WHERE report_year=2026 AND quarter=1 AND snf_offered=true ORDER BY days_21_100_copay DESC — returns each 2026 plan's full SNF day-interval copay ladder (showing that the headline copay_amount of $0-$20 hides a ~$218/day charge after day 20), so you can rank plans by true SNF cost rather than the misleadingly cheap first-interval copay.

Traps

1. 1copay_amount and cost_share_string are the INTERVAL-1 (days 1-20) copay ONLY — they systematically understate SNF cost. The expensive days-21-100 copay (often ~$218/day) lives in tier1_copay_interval2_amount. copay_amount is provably identical to tier1_copay_interval1_amount (97,606 equal / 0 differ).
2. 2~18 columns are HARD-CODED NULL in the view definition (CAST(null AS ...)): coinsurance_pct, all limit_* and all benefit_amount_* columns, normalization_status/basis, aggregation_policy. They are schema-parity placeholders to match sibling PBP benefit views — verified 0 non-null. has_coinsurance can be true while coinsurance_pct is null, so you can never read the SNF coinsurance percentage from this view (go to medicare_silver.pbp_section_b2_snf).
3. 3The offered/observability flags (offered_status, snf_offered, numeric_amount_observability, null_amount_reason) are about the (always-null) generic benefit-amount path and the pbp_b2_bendesc_yn flag — NOT about copay availability. 92,714 rows with offered_status='not_offered'/snf_offered=false STILL carry a real copay_amount. Do not filter on snf_offered to find plans-with-a-copay.
4. 4This is a Presto VIRTUAL_VIEW with NO is_current column. Versioning is via the (report_year, quarter) snapshot itself. tier1_copay_interval_count can be '3'/'4' but the view exposes only interval 1 and interval 2 columns, so higher counts mean additional cost-ladder intervals exist in the silver source that are not surfaced here.
5. 5Grain traps: the profiler reported a 7-column superset and flagged limit_period_code/benefit_amount_period_code as candidate keys — both are 100% null and irrelevant. The true minimal key is contract_id+plan_id+segment_id+report_year+quarter (201,488=201,488). Also segment_id has empty-string values (5,074 rows) that are distinct from '0' and break CAST-to-int; plan_id is only unique within a contract; quarter's 1-4 is the PBP release quarter, not a calendar date.
6. 6prior_hospitalization_days_required and tier1_copay_interval_count are STRINGS (e.g. '1','3'), not numbers — cast before arithmetic. tier1_has_copay and has_copay are two different derivations that frequently disagree; prefer has_copay for a plan-level copay signal.

Columns (121)

Query it

describe_data_source(source_fqn="medicare_semantic.vw_pbp_snf")     # the card, live
query_substrate(sql="SELECT * FROM medicare_semantic.vw_pbp_snf WHERE contract_id = '<value>' AND plan_id = '<value>' AND segment_id = '<value>' AND report_year = '<value>' AND quarter = '<value>' LIMIT 100")