MA pbp mrx tier

What it is

1. 1CMS Medicare Part D Plan Benefit Package (PBP) Section-D drug-tier cost-sharing file: one row per formulary tier per Medicare drug plan, giving the member's out-of-pocket cost (copay or coinsurance) for that tier across every pharmacy channel and days-supply.
2. 2This is the canonical public source for "what does a Part D / MA-PD plan charge the patient for a Tier-N drug." Silver layer = lightly cleaned CMS PBP extract spanning contract years 2010-2026.

Grain

How to read it

1. 1To get the current member copay for each formulary tier of a Medicare drug plan, filter to one snapshot and the initial-coverage phase, then read the retail-standard 30-day cost column.
2. 2Example: SELECT mrx_tier_id, mrx_tier_label_list, mrx_tier_rstd_copay_1m AS copay_30d, mrx_tier_rstd_coins_1m AS coins_30d_pct FROM medicare_silver.pbp_mrx_tier WHERE report_year=2024 AND quarter=4 AND mrx_tier_type_id='1' AND bid_id='H0028_014_0' ORDER BY cast(mrx_tier_id AS int).
3. 3Returns one row per tier (Preferred Generic $0, Generic $0, Preferred Brand $47, Non-Preferred $99, Specialty 33% coinsurance) - the patient's out-of-pocket drug cost ladder for that plan.
4. 4Switch mrx_tier_type_id to '3' to see the coverage-gap (donut-hole) cost-sharing for the same tiers.

Traps

1. 1mrx_tier_id (the tier number 1-7) is NOT a unique key for a plan: each tier appears up to twice, split by mrx_tier_type_id = '1' INITIAL-coverage vs '3' COVERAGE-GAP cost-sharing. You MUST include mrx_tier_type_id, or filter to one phase (mrx_tier_type_id='1' for the normal/initial-coverage copay), or you double-count. Verified: all 116,195 type-3 rows carry mrx_tier_gap_cost_share; zero type-1 rows do.
2. 2This table keeps MULTIPLE CMS quarterly releases per contract year and there is NO is_current flag. The 7-key includes both report_year AND quarter. For a single clean per-plan-per-tier view you must pick one snapshot, e.g. WHERE report_year=2024 AND quarter=4 (the latest available quarter for that year). Quarters retained per year are uneven: 2010-2017/2021/2025/2026 have only 1 quarter, 2018/2019/2023 have 4, 2020/2024 have 3, 2022 has 2 - so raw row counts per year are NOT comparable.
3. 3The cost lives in ~200 sparse columns (channel x days-supply x copay/coinsurance). Any single row has only a few non-null cost fields and copay is mutually exclusive with coinsurance on the same row (verified: retail-std-30d copay_only=165k, coins_only=63k, both=49). The everyday number most users want is mrx_tier_rstd_copay_1m (retail-standard 30-day copay) OR mrx_tier_rstd_coins_1m (coinsurance %). A NULL copay does not mean free - check the coinsurance column and the right channel family.
4. 4Identifier hygiene: plan_id zero-padding is inconsistent across vintages (unpadded in 2020, 3-digit padded elsewhere) - join/group on bid_id (= contract_id_plan_id_segment_id, canonical padded) rather than raw plan_id. effective_period_start/observed_period are just Jan-1 of report_year (annual), and asserted_at/cms_vintage_sequence are ETL provenance (2026-05 load dates), NOT data recency or a reliable quarter proxy.
5. 5mrx_tier_label_list is the friendly tier name but is NULL for all of 2010 and sometimes a bare numeric code in early years; mrx_tier_cstshr_struct_type does NOT cleanly map to copay-vs-coinsurance (observed struct='2' rows carrying copays) - infer copay vs coinsurance from which value column is populated, not from the struct code.

Columns (309)

Query it

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