Clinicians and industry payments

Open Payments' 2024 general-payment file — a manufacturer- and GPO-reported record of every transfer of value to a US physician, advanced-practice provider, dentist, or teaching hospital — joined to the clinician roster from dim_clinician, Part B utilization from fct_physician_utilization, and Part D prescribing from fct_prescriber_drug_spending. Every chart below reads a pre-aggregated source query; the 15.5M-record raw fact never leaves the warehouse. Payment payers are always rolled up on paying_manufacturer_or_gpo_id, never on raw name, so the same manufacturer under mixed-case filings collapses to one row.

How concentrated are the dollars?

Two thirds of every industry dollar reaches only the top 1% of paid clinicians, and 90% stops with the top 10%. The bottom half of the paid population combined receives a rounding-error share.

Looked at the other way — by absolute dollar bracket per clinician — about 83% of paid NPIs receive under $1,000 across all of 2024 combined, and 4,132 (0.4%) cross $100,000. Those 4,132 clinicians account for the bulk of the total.

The Food-and-Beverage paradox

Read either column at a time: Food and Beverage is by far the most common transfer type but each meal is small; Royalty or License is the smallest transaction count on the chart (about 15k records) yet tops the dollar ranking.

Who pays

The top 15 payers by 2024 dollars. Roll-up is on paying_manufacturer_or_gpo_id, so mixed-case duplicates of the same entity collapse to one row and any_value(name) chooses a single deterministic display label per id.

The list splits by business model: pharma manufacturers (AbbVie, Medtronic, Takeda, AstraZeneca) file hundreds of thousands of small meal / consulting records against a broad clinician base, while a few device and specialty firms (BioNTech, Edge Endo, DePuy Synthes) concentrate nine-figure totals on a handful of clinicians. BioNTech's number is almost entirely one royalty stream; Edge Endo's is a single $91.08M acquisition payment to one endodontist.

Who gets paid

Two rankings of the same specialty axis. First, top specialties by total dollars — dominated by high-volume Internal Medicine plus surgical fields with expensive implants. Endodontics ranks third almost entirely on the strength of Edge Endo LLC's $91.1M acquisition payment to a single clinician (see the per-clinician note below):

Second, the same table sorted by dollars per paid clinician (min 100 paid clinicians per specialty). The ranking flips: orthopaedic sub-specialties, endodontics, and other implant-centric fields average five figures per paid clinician, while primary-care specialties average low three figures each even though the summed total is large. Endodontics' top rank ($54K per clinician) is carried almost entirely by 11 Edge Endo LLC records to 9 clinicians — one $91.08M acquisition line plus a handful of small consulting and device-loan entries. Strip those 11 records out and the per-clinician average falls ~89% to ~$5.9K, dropping endodontics out of the top 15.

Payments and prescribing

The industry-payment fact and the Part D prescriber fact share NPI as a key — the intersection is 642,414 clinicians who both received industry payments in 2024 and appear in the prescriber-drug file. The two facts describe different behaviors reported by different parties, so the join surfaces an association between payment receipt and prescribing volume. It does not measure whether payments cause prescribing to change.

Grouping every prescriber by their total 2024 industry payments — a "No payments" bucket for the 497k prescribers with none, then deciles of the remaining 642k — median prescriber-billed drug cost rises steeply across the paid deciles: from ~$2.7K in decile 1 to ~$162K in decile 9 (~59×), with decile 10 slightly below decile 9 at ~$148K:

The association is real and large, but selection is heavy on both sides: manufacturers target clinicians who already prescribe a lot, and high-volume prescribers are more likely to be attend the meetings and dinners where payments are filed. Median (not mean) is the better summary here — decile 10's mean $ paid is inflated by a handful of royalty recipients receiving seven and eight figures.

Sliced by specialty rather than by decile, the paid share is high (≥ 70%) in nearly every high-volume prescribing specialty; primary care and psychiatry are the outliers with paid shares closer to half.

Do the drugs match?

The top 20 drug products by 2024 promotional spend, joined to Part D prescriber-billed spend on the same brand (upper-cased brand-name match against fct_prescriber_drug_spending). Rows with no Part D match are honest nulls, not zeros — Comirnaty and Pluvicto aren't Medicare Part D drugs, so their absence is the join working correctly, not missing data.

Ratio of promo $ to prescriber-billed $ varies by two orders of magnitude across products — Jardiance sees ~$1 of promo per $700 of prescribing, Sotyktu ~$1 per $4. Read this as a portfolio-strategy signal (mature drugs get less promotion per prescribing dollar than newly launched ones), not as an efficacy or influence signal.

Utilization: paid vs unpaid

For the 1.24M individual-clinician rows in fct_physician_utilization that also carry a payments-vs-no-payments label, the median clinician who received any 2024 industry payment bills roughly 1.8× the Medicare beneficiaries and 1.77× the Medicare dollars of the median clinician who received none. Average HCC risk score is essentially identical between the two cohorts — this isn't about paid clinicians treating sicker patients.

Same caveat as the prescribing section: manufacturers target high-volume clinicians, so the cohorts are not exchangeable. The utilization gap is what the two selected populations look like side by side, not the effect of receiving a payment.

What Medicare pays for

Switching from the payments-to-clinicians side to the payments-from-Medicare side. fct_hcpcs_service rolls the CMS-published Physician & Other Practitioners by-geography-and-service file up to (hcpcs_code, place_of_service) — 13,463 National-grain rows in the current 2026-05-21 vintage, one per procedure code × facility/office split. Every headline below is computed from that National grain; state rows never contribute to totals here.

3.55 billion services and $120.8B of Medicare program payment — the "program payment" here is sum(total_services * avg_medicare_payment_amount) across all 13,463 National rows, which is the CMS-published all-USA dollar total for the fee-for-service physician side of Medicare (carrier + DME) as reported in this file. Every dollar is exposed at HCPCS grain, so a top-N ranking captures where the money actually goes.

The top 20 (HCPCS, place) cells by total payment:

Two shapes dominate the list. Office E&M (99214, 99213, 99215, 99204, G0439 wellness visits) rings up billions on very high volume — 99214 alone is $8.0B across 95M services. Facility inpatient E&M (99223, 99232, 99233, 99285 ED visits, 99291 critical care) is the same story on the hospital side. Punctuating both tails are a handful of per-unit-expensive items: cataract removal (66984), Part-B drugs (J0178 aflibercept, J9271 pembrolizumab, J2777 faricimab), and a few graft/wrap codes (Q4205, Q4271) where the average charge runs into four figures.

Facility vs office

Office / non-facility accounts for the bulk of both service volume (3.05B of 3.55B, 86%) and dollars ($85.6B of $120.8B, 71%). Facility services are fewer (506M) but each one costs Medicare more on average ($69.52 vs $28.09) — the site-of-service differential is worth about 2.5× per service in this vintage.

Payment vs charge

Providers submit charges (avg_submitted_charge) that are far higher than what Medicare's fee schedule actually pays. The median HCPCS × place cell has a payment_to_charge_ratio of

Bucketed distribution across all 13,463 National-grain rows:

Just under half the rows (6,579 of 13,463, 49%) land in the 10–20% band, and a further 2,783 (21%) in 20–30%. Only 588 rows (4.4%) pay back more than half of the submitted charge, and 181 (1.3%) pay 75% or more — mostly low-dollar items where the submitted charge is already close to the fee schedule.

Absolute-dollar gaps concentrate in the highest-volume services. The 15 HCPCS with the largest total submitted − paid gap (at least 100k services nationally):

The largest single gap is 99214 at $18.4B — 95M services with an average charge of $276 against a paid amount of $83.75. ED visits (99285), knee replacement (27447), and critical-care (99291) all pay back 10–20% of the submitted charge. Read this as reporting convention rather than as denial: avg_submitted_charge is what the provider population billed, which under Medicare is largely irrelevant because the program pays the fee schedule regardless of what the claim says. The submitted charge is a data-quality signal about the provider's non-Medicare billing baseline, not a negotiation counter-offer.

Geography benchmarks

fct_physician_service_geography carries a National benchmark inline on every state row (national_avg_medicare_payment_amount and friends), which makes state-vs-national comparisons a single-table lookup. Aggregating a state's own service mix at state prices versus the same mix at national prices gives a "payment index" — greater than 1 means the state gets paid above what the national fee schedule would apply to its mix, less than 1 means below. All the state rankings here are the 50 states + DC only; the territory codes (60/66/69/72/78) and Armed Forces / unknown / foreign pseudo-codes (9A–9E) are excluded because their volumes are small enough that a handful of outlier cells swing the aggregate, and the five state rows CMS emitted with NULL geography_code (see caveats) are dropped as well.

Alaska tops the ranking at 1.162, then a cluster of high-cost / high GPCI states (NY 1.083, DC 1.076, NJ 1.071, CA 1.069). Maine anchors the bottom at 0.875, alongside Alabama, West Virginia, South Dakota, and Kentucky (all ≤ 0.92). The median state runs at 0.969 — most states get slightly less per unit of service than the national average, which is exactly what happens when the population-weighted national average is pulled up by a small number of large high-cost-of-labor states.

Zooming in on a single anchor code — 99214 O, the highest-payment single cell in the file at $8.0B nationally — the same pattern shows up cleanly:

Alaska pays 1.211× the national average ($101.38 vs $83.75), DC 1.158, New York 1.156, California 1.129, New Jersey 1.125. At the other end North Dakota, Maine, and Arkansas run at 0.81–0.85. The spread — top state pays about 1.50× what the bottom state does — is entirely fee-schedule geography (GPCI locality adjustments). It is not a quality or intensity signal.

Caveats

• Service-file snapshot has no year. fct_hcpcs_service and fct_physician_service_geography are single-vintage rollups of the Physician & Other Practitioners by-geography-and-service file (2026-05-21 in the current vintage, as_of column). CMS does not publish a claims year on this file, so nothing in the service and geography sections is a time series.
• National-grain is authoritative for totals. fct_hcpcs_service sources its 13,463 rows from the National rows of the upstream staging model rather than aggregating the 9.78M-row by-provider-and-service file. The two disagree because CMS drops (npi, hcpcs, pos) cells with 10 or fewer beneficiaries from the provider file — 22.4% of program-wide services (798M of 3.55B) live entirely in that sub-11 suppression tail. Summing state rows in fct_physician_service_geography also under-counts national volume for the same reason. Every total on this page is computed from the National grain.
• avg_submitted_charge is not a negotiation. Under Medicare fee-for-service, the program pays the fee schedule regardless of what the claim's submitted amount says. The submitted charge in this file is a provider-population average — a data-quality reflection of what the provider bills their non-Medicare book — not evidence of a rejected offer. All the payment-vs-charge charts above should be read as descriptive, not as denial rates.
• Geography rankings exclude territories and NULL-code rows. The 50-states-plus-DC filter drops five state rows CMS emitted without a geography_code (all E&M / psychotherapy codes: 90833/F, 99223/F, 99232/F, 99233/F, 99239/F), the five territory codes 60 (AS), 66 (GU), 69 (MP), 72 (PR), 78 (VI), and the five pseudo-codes 9A (Armed Forces Central/South America) through 9E (Foreign Country). All 10 non-state codes are populated as ordinary state rows in the mart — this is a display choice for the rankings only.
• State payment_to_charge_ratio can exceed 1.0. 13 state rows in the current vintage cross the 1.0 line — 12 for HCPCS M0010 (chelation therapy, ~$70 baseline) and one for G0442. National rows never do. Geographic-adjustment multipliers occasionally push a small-dollar service above the provider-population average submitted charge; those rows are legitimate and are included as-is.

Open-Payments caveats

• One program year. fct_industry_payments currently holds program year 2024 only (payment_year = 2024, published mid-2025). Nothing on this page is a time series; a paid clinician here means a clinician who received at least one reportable transfer during calendar 2024.
• Coverage recipients are individuals + teaching hospitals. 51,278 records (~0.33%) carry no covered_recipient_npi and drop out of every NPI-based aggregation above: 37,388 name a teaching hospital rather than an individual clinician, and the remaining ~13.9K (7,729 non-physician practitioners + 6,161 physicians) are individual-clinician records filed without an NPI. All of these rows are still in the total dollar headline.
• Roster coverage is partial. 32.4% of paid NPIs don't appear in `dim_clinician` — the roster covers only currently Medicare-enrolled clinicians while Sunshine reports on every US physician / dentist / APP a manufacturer paid, including retirees, out-of-network dentists, and clinicians who never billed Medicare. Charts that only need Open Payments fields (nature, payer, specialty, product) are unaffected; the utilization and prescribing joins silently drop those NPIs.
• Prescribing snapshot has no year. fct_prescriber_drug_spending is a single vintage of Part D prescriber-drug totals with no year column; the association charts compare 2024 industry payments to a snapshot of prescribing that covers calendar 2024 claims. Rows below 11 claims are suppressed by CMS and drop out of the join denominator.
• Association ≠ causation. Every payment-vs-prescribing or payment-vs-utilization comparison on this page is observational. Manufacturers target high-volume clinicians (payment goes toward where the demand is), so the cohorts on either side of any split are not exchangeable. Nothing here estimates the effect of receiving a payment on how much a clinician prescribes or bills.
• covered_recipient_specialty uses NUCC pipe-delimited paths. The "specialty" tables above show them verbatim rather than normalizing to a shorter label; that keeps the taxonomy honest (Allopathic & Osteopathic Physicians|Orthopaedic Surgery and …|Orthopaedic Surgery|Orthopaedic Surgery of the Spine are different specialties in the source) but forces a bit of scrolling.
• Products are position-1 only. Open Payments records can carry up to five associated products; fct_industry_payments exposes slot 1 (populated on ~94% of records) as product_name / product_category. The remaining slots stay in the staging layer.