Prescribers

Who drives spending on the top Part D drugs? This page reads three pre-aggregated marts. Prescriber-level totals come from fct_prescriber_profile (one row per NPI, no drug-level suppression). The prescriber × drug detail lives in fct_prescriber_drug_spending (~28M rows), which CMS strips of any row with fewer than 11 claims — so its aggregate totals sit about 13.5% below the profile file's on claims and 21% below on dollars. National and state-level rankings come from fct_drug_geography. All dollar figures are prescriber- billed drug cost: what pharmacies billed for the prescriptions a prescriber wrote, before rebates — not net Part D program spending.

Top drugs by prescriber-billed cost

The list splits into two regimes: mass-market drugs written by hundreds of thousands of prescribers at a few hundred dollars per claim (anticoagulants, GLP-1s), and specialty drugs written by a few thousand prescribers at five figures per claim.

Claim leaders vs cost leaders

The two rankings above hide the fact that no drug is both a top-10 claim leader and a top-10 cost leader. fct_drug_geography National rows make this concrete: the ten drugs written for the most claims — generic statins, antihypertensives, thyroid hormone — cost only a few dollars per claim. The ten drugs that drive the most spending — anticoagulants, GLP-1s, immunomodulators, an oncology brand — average hundreds to tens of thousands of dollars per claim.

The top 10 by claim volume account for

How concentrated is each top drug?

For each of the top 25 drugs, the share of its billed cost written by its top 1% of prescribers (by cost; the 1% headcount is rounded up, so every drug keeps at least one). Spending on the biggest drugs is not dominated by a handful of prescribers:

For Eliquis — the biggest drug at $19.9B — the top 1% of its

Which specialties drive the spending?

Primary care (nurse practitioners, internal medicine, family practice, physician assistants) bills the most in aggregate through sheer claim volume, while oncology and rheumatology reach the top ten with a tiny fraction of the claims — their cost per claim runs roughly 5–30× higher. Both cost and claim totals here come from the sub-11-claim- suppressed detail file, so they undercount every specialty (13.5% of claims and 21% of cost are missing at the file level); the profile- file view in the next section aligns to the full $288B and adds a brand / opioid / antibiotic / antipsychotic breakdown.

Prescriber-type prescribing patterns

Same specialties, but from fct_prescriber_profile — one row per NPI, no drug-level suppression — so both the totals and the mix columns cover every prescribing NPI, not just those with 11+ claims on a given drug. Restricted to types with at least 5,000 prescribers so the rate columns are stable. Rates are computed only over rows where CMS didn't blank the numerator (opioid, antibiotic, and antipsychotic- GE65 measures are individually suppressed when their counts are 1–10). prescriber_type here comes directly from the profile file — the FK to dim_clinician is intentionally not used, since only ~66% of NPIs match the clinician roster.

The single highest brand share in the chart belongs to Pharmacists (~69%) — an unusual category whose Part D claims skew toward specialty dispensing rather than office practice. Among physician specialties, Pulmonary Disease (~50%), Endocrinology (~48%), Ophthalmology (~36%, with Optometry slightly higher at ~39%), and Infectious Disease (~26%) prescribe from brand-heavy therapeutic areas (inhalers, GLP-1s / insulins, ophthalmic biologics, and HIV regimens respectively). Primary-care specialties sit near 10–13% brand by claim.

Antibiotic prescribing is dominated by three specialties that make sense clinically: Infectious Disease (~31% of claims), Urology (~16%), and Dermatology (~10%). Primary-care antibiotic rates run around 3–7% (Physician Assistants highest at ~7%, Family Practice lowest at ~3%).

Psychiatry accounts for ~8% of its claim volume in antipsychotics prescribed to beneficiaries 65 and older — an order of magnitude above every other specialty. Neurology (~2%) and Nurse Practitioners (~2%) sit next; the rest are below 1%.

Where is the spending written?

State totals track population; the state column also carries territories and military/unknown codes, which is why there are more than 51 values in the source.

State cost-per-beneficiary variation

Aggregate state spending mostly tracks population, but per-beneficiary cost varies more than the aggregate totals suggest — especially on specialty drugs. This chart walks the biggest drug (Eliquis) across the 50 states + DC, with the national cost-per-beneficiary marked as a reference line. fct_drug_geography carries the national benchmark inline on every state row, so the reference value doesn't require a separate roll-up.

Even the largest drug — where per-patient dosing is nearly identical by protocol — shows a

State-level total_beneficiaries is suppressed on about 19% of state rows (CMS blanks counts of 1–10), so tiny (drug × state) cells drop out of the per-beneficiary column — this filter alone excludes the territories and armed-forces pseudo-codes that would otherwise skew the min/max. The four missing drug-state cells in the table — Montana and Wyoming for Vyndamax, Wyoming for Pomalyst and Ofev — are the ones where CMS suppressed the state's beneficiary count for that drug.

Top prescriber-drug combinations

The largest single prescriber×drug lines among the top 25 drugs. Even the biggest individual line is a small sliver of its drug's total — the share_of_drug_cost column makes the concentration story above concrete.

Prescriber-billed vs gross Part D spending

The top 25 drugs joined to fct_part_d_drug_spending's manufacturer-roll-up gross spending for the same drug, via drug_key (exact upper-cased brand + generic match; staging strips the Part D spending file's trailing-* aggregate marker, so the names line up).

Caveats

• Single snapshot, no trend. All three marts are single snapshots (calendar-2024 claims, vintage in each mart's as_of column) with no year column, so nothing on this page is a time series.
• Small rows are suppressed entirely. CMS removes prescriber-drug rows with fewer than 11 claims from the detail file, so every row in the detail-file sections has ≥ 11 claims, totals undercount true spending (by ~13.5% of claims and ~21% of dollars vs the profile mart), and low-volume prescribers are invisible on the detail file (which also inflates the top-1% shares slightly). The profile mart itself has no such row filter — CMS only drops NPIs with 10 or fewer total claims from that file — so the aggregate totals in the headline BigValues and the prescriber-type-patterns section include those small prescriber-drug lines.
• Measure-level suppression on the profile file. total_beneficiaries is NULL (suppressed at 1–10) on 10.3% of profile NPIs; the opioid, antibiotic, and antipsychotic-GE65 measures are individually NULL when their own counts fall in that band. The prescriber-type-patterns section computes each rate only over rows where the numerator is populated, so denominators shrink from row to row.
• Geography state cells drop out too. In fct_drug_geography, total_beneficiaries is NULL on ~19% of State rows, which is why the state cost-per-beneficiary charts show 51 states for Eliquis but only 49–50 for Vyndamax / Pomalyst / Ofev.
• Star-marked spending rows aggregate brand and generic versions. The Part D spending file star-marks drug names whose estimates aggregate brand and generic versions; staging strips the marker so every row here conforms to dim_drug ( 0.0% of detail-file rows orphan, enforced by a dbt test) and the gross- spending comparison covers all 25 top drugs — but an aggregated gross figure can overshoot its brand-only prescriber-billed counterpart, which is why Stelara's billed/gross ratio runs low.
• Billed ≠ net. Prescriber-billed drug cost ignores manufacturer rebates and DIR; actual net Part D spending is materially lower, especially for high-rebate brand drugs.
• Specialty is self-reported. prescriber_type comes from the provider's Medicare enrollment/claims specialty. The detail file carries 182 distinct values; the profile file carries 214 (with 2 NPIs missing a specialty). Both files' organizational NPIs (2 rows in the profile) are omitted from the dim_clinician relationships test rather than reported as orphans.