MA pbp section b7 health prof

What it is

1. 1A CMS Medicare Advantage Plan Benefit Package (PBP) extract — Section B7 "Health Care Professional Services" — giving the member cost-sharing design (copay, coinsurance, deductible, prior-auth, referral, and benefit limits) that each MA / MA-PD plan charges for visits to doctors and other professionals.
2. 2It is the official source-of-truth for "what does an MA plan charge a member to see a primary-care doctor, specialist, therapist, chiropractor, podiatrist, mental-health provider, etc." across every contract year 2001-2026.

Grain

How to read it

1. 1Pin a single snapshot per plan-year before aggregating.
2. 2Example — median primary-care vs specialist copay for 2026 HMO plans: SELECT approx_percentile(pbp_b7a_copay_amt_mc_max,0.5) AS pcp_copay, approx_percentile(pbp_b7d_copay_amt_mc_max,0.5) AS specialist_copay FROM medicare_silver.pbp_section_b7_health_prof WHERE report_year=2026 AND quarter=1 AND pbp_a_plan_type='01' AND pbp_b7a_copay_yn='1'.
3. 3This returns PCP ~$0 and specialist ~$35, i.e. what a 2026 MA HMO member typically pays out-of-pocket to see a doctor.
4. 4Join to other pbp_section_* tables on (contract_id, plan_id, segment_id, report_year, quarter) for full benefit design.

Traps

1. 1GRAIN TRAP — quarter is part of the key, not a refresh. The SAME benefit year is re-published across multiple quarterly snapshots (e.g. 2023 appears as Q1,Q2,Q3,Q4; row counts drift 7752->7747->7751->7755). If you query a plan's PCP copay without pinning a quarter you will multi-count it ~2-4x. For 'the plan year' pick one quarter (e.g. Q1 = initial/as-bid, or MAX(quarter) for the latest correction). There is NO is_current / latest flag in this table.
2. 2NO row-level hash and NO is_current: sha256_content_hash and source_url each have only 38 distinct values (one per release ZIP), and asserted_at is just the ingest date (2 values). None of these identify or dedupe a row — the only true row identity is contract_id x plan_id x segment_id x report_year x quarter.
3. 3The profiler's reported grain is a bloated SUPERSET (24 columns incl. several *_maxplan_amt and observed_period). The real minimal key is 5 columns; observed_period (=effective_period_start always), bid_id (=concat of the 3 IDs, NULL pre-2010), and version (a bid-software number, not the snapshot id) are all functionally dependent and must NOT be added to the key.
4. 4NULL vs 0.0 carries meaning: a 0.0 copay/coinsurance is a real $0 cost-share (PCP is frequently $0); a NULL means the benefit/sub-section was not offered or not reported (older years null whole sub-sections that didn't exist yet). _yn flags are '1'=Yes / '2'=No / NULL=not-applicable — never assume 0/1 boolean.
5. 5Time coverage is gappy and quarter coverage is uneven: benefit years 2010 and 2011 are entirely missing; quarterly snapshots only start ~2018 (earlier years are Q1-only); and within recent years some quarters are absent (2020 has no Q3, 2024 has no Q2). Do not assume a continuous 4-quarter-per-year panel.
6. 6Sub-sections are columns, not rows: B7a..B7k (primary care, chiropractic, podiatry, specialist, PT/OT, psychiatric, mental-health, substance-abuse, individual-psych, combined, additional) are encoded as ~430 wide columns per row, each prefixed pbp_b7<letter>_. This differs from long/normalized fact tables — to compare service categories you select different column families, not filter a category column. pbp_b7j_mc_bendesc_cats is a packed semicolon-delimited code list that must be string-parsed.
7. 7Most amount columns come as min/max PAIRS (or single _amt with a _max ceiling for B7g/B7k) representing a cost-share RANGE within the benefit category, and many cost-shares are split by network/coverage tier suffix: _mc=Medicare-covered, _mcis=Medicare in-network/in-area, _mcgs=gap/out-of-area, _rc/_rf/_other/_cc=supplemental/routine. Pick the right suffix or you mix in-network and out-of-network dollars.

Columns (315)

Query it

describe_data_source(source_fqn="medicare_silver.pbp_section_b7_health_prof")     # the card, live
query_substrate(sql="SELECT * FROM medicare_silver.pbp_section_b7_health_prof WHERE contract_id = '<value>' AND plan_id = '<value>' AND segment_id = '<value>' AND report_year = '<value>' AND quarter = '<value>' LIMIT 100")