MA pbp section c

What it is

1. 1CMS Medicare Advantage Plan Benefit Package (PBP) "Section C" — the medical/hospital benefit cost-sharing detail (inpatient hospital, SNF, outpatient service groups, deductibles, plan maximum-out-of-pocket, prior-authorization/referral rules, and out-of-network / point-of-service / visitor-travel terms) for each MA plan, parsed from the public CMS PBP PUF zips.
2. 2It documents what an MA enrollee pays for Part-A/Part-B-style medical services under each plan-year.

Grain

How to read it

1. 1Latest out-of-network inpatient coinsurance for 2026 MA plans: SELECT contract_id, plan_id, segment_id, pbp_c_oon_coins_iph_pct FROM medicare_silver.pbp_section_c WHERE report_year=2026 AND quarter=(SELECT max(quarter) FROM medicare_silver.pbp_section_c WHERE report_year=2026) AND pbp_c_oon_coins_iph_pct IS NOT NULL.
2. 2This answers 'what does each MA plan charge an enrollee in coinsurance for an out-of-network inpatient hospital stay in 2026' (real values cluster at 50%).
3. 3Always pin quarter to the max for the year to avoid double-counting re-releases, and remember in-network cost-sharing columns are NULL for 2012+.

Traps

1. 1BIGGEST TRAP — in-network cost-sharing is empty for all modern years. Every pbp_c in-network benefit column (general deductible, in-network MOOP/gen_oop_med_amt, inpatient iha/iph/ihs copay & coinsurance, outpatient groups 1-5, gated_vs_open, auth/referral flags) is 100% NULL for benefit years 2012-2026. Only the OON (out-of-network), POS (point-of-service), and VT (visitor/travel) families are populated post-2012 (2023: pos 93/130 cols, oon 73/91, vt 3/3). In-network values exist ONLY in 2004-2006, and sparsely. Do NOT use this table to look up a current MA plan's in-network deductible/copay/MOOP — those columns are blank.
2. 2No is_current / no snapshot dedup flag. The same plan-year is re-published every quarter (and at multiple submission versions), all stored as separate rows. quarter and version are BOTH part of the grain. To get the 'latest' view of a plan-year you must filter to the max quarter (and max version) per contract_id+plan_id+segment_id+report_year yourself — nothing does it for you. Naively counting rows double-counts plans across quarters.
3. 3bid_id looks like the natural key but is NULL for all 2001-2006 rows (8,590 rows) and only 17,463 distinct vs 20,668 distinct contract+plan+segment. When non-null it is exactly contract_id||'_'||plan_id||'_'||segment_id. Use contract_id+plan_id+segment_id (not bid_id) for joins across the full history.
4. 4pbp_c_outpt_group_num_id is the hidden grain disambiguator but is NULL 97.9% of the time — it only splits a plan into multiple rows (outpatient groups 1-10) in 2004-2006. For 2012+ ignore it. observed_period is also a false key: it equals Jan-1-of-report_year and is identical across a year's quarterly re-releases, so it is coarser than report_year+quarter.
5. 5Time coverage has holes: benefit years 2007-2011 are entirely absent (data jumps from 2006 straight to 2012) and 2020 is missing its Q3 release. report_year is the benefit/effective year (what plan terms apply to), NOT a publication date. asserted_at (only 2 dates) is the Vlada ingest date, not a data dimension. All cost amounts are USD and all coinsurance fields are percentages 0-100; _yn flags use '1'=yes / '2'=no.

Columns (465)

Query it

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