MA pbp section b3 cardiac rehab

What it is

1. 1A flattened CMS Medicare Advantage Plan Benefit Package (PBP) Section B-3 extract describing each MA plan's member cost-sharing and coverage rules for cardiac and pulmonary rehabilitation services (the four sub-services: cardiac rehab / intensive cardiac rehab / pulmonary rehab / supervised exercise therapy for PAD).
2. 2It is the benefit-design side of healthcare cost data — what a plan charges its members (copay/coinsurance) and what utilization/authorization limits apply — not provider-negotiated rates.

Grain

How to read it

1. 1Latest-snapshot member cost share for cardiac rehab across MA plans: SELECT bid_id, contract_id, plan_id, pbp_a_plan_type, pbp_b3_auth_yn, pbp_b3_copay_amt_mc_max_crs, pbp_b3_coins_pct_mc_max_crs FROM (SELECT *, row_number() OVER (PARTITION BY bid_id, report_year ORDER BY quarter DESC) rn FROM medicare_silver.pbp_section_b3_cardiac_rehab WHERE report_year=2025) WHERE rn=1.
2. 2Answers 'for plan year 2025, what copay/coinsurance and prior-auth does each MA plan impose on cardiac rehabilitation, using the freshest quarterly PBP snapshot' (~7,570 plan-years, deduped from the multi-quarter raw rows).

Traps

1. 1NOT one row per plan. Multiple quarterly snapshots of the same plan-year coexist with NO is_current flag. A naive COUNT(*) (113,962) over-counts plans ~2x. For one-row-per-plan-year (55,872), take the latest snapshot: ROW_NUMBER/max(quarter) per (bid_id, report_year). bid_id alone, or (bid_id, observed_period), or (contract_id,plan_id,segment_id) are all NON-unique.
2. 2bid_id is fully derived: it equals contract_id_plan_id_segment_id with zero exceptions, and pbp_a_plan_type/orgtype are stable per bid_id-year. The profiler's 8-column 'verified key' is a bloated superset; the true minimal grain is just (bid_id, report_year, quarter).
3. 3Snapshots are NOT strictly nested across quarters: between Q1 and Q4 of the same year a few plans appear or disappear (2023: Q1=7752, Q4=7755, union=7766). 'Latest quarter' is the best single snapshot but does not perfectly superset earlier quarters; union across quarters if you need the full plan universe.
4. 4Many cost columns are structurally null and meaningful: pbp_b3_ded_amt and pbp_b3_maxenr_amt/maxenr_per are ALWAYS null (cardiac rehab has no benefit-level deductible or enrollee max — vestigial template columns). The real cost share lives in the 'mc_' (Medicare-covered) copay/coinsurance columns, split across 4 sub-service suffixes crs/icrs/prs/setpad. The non-'mc_' copay/coins columns (supplemental portion) are largely null.
5. 5YN flags use 1=Yes, 2=No (not boolean). bendesc_yn='2' means 'no EXTRA benefit description', NOT 'no benefit' — copay/coins/auth are still populated when bendesc_yn='2'. Also _d decode columns (e.g. bendesc_per_*_d) are mostly null even when the base code is set, so never use a non-null _d to detect a limit. The value 20.0 appears as both a coinsurance percent and a copay dollar amount — disambiguate by column family, and pair each *_ehc bitmask (position = sub-service) with its copay/coins columns.
6. 6Sibling-table caveat: this is one of many pbp_section_* tables sharing the same key shape and provenance columns; it covers ONLY cardiac/pulmonary rehab (B-3). It is plan benefit DESIGN (what the member pays), not provider-negotiated commercial/Medicare fee-schedule rates — join to plan/enrollment or fee-schedule tables for rate-vs-benefit analysis. cms_vintage_sequence usually mirrors quarter but diverges for ~14K re-released 2020/2021 Q1 rows.

Columns (111)

Query it

describe_data_source(source_fqn="medicare_silver.pbp_section_b3_cardiac_rehab")     # the card, live
query_substrate(sql="SELECT * FROM medicare_silver.pbp_section_b3_cardiac_rehab WHERE bid_id = '<value>' AND report_year = '<value>' AND quarter = '<value>' LIMIT 100")