Pharma Payments and Medicare Prescribing: What CMS Data Can Reveal
CMS Open Payments and Medicare Part D Prescribers by Provider and Drug can be compared to study whether reported financial relationships and Part D prescribing measures occur together. They measure different things: Open Payments records reportable transfers of value and certain ownership interests; the Part D file summarizes prescription claims and utilization. A comparison can describe patterns, but cannot show that a payment caused a prescription or that prescribing was inappropriate.
QOPE’s Provider Drugs dataset includes NPI, brand and generic drug names, claims, standardized 30-day fills, days supply, drug cost, and beneficiaries. It is not a ready-made link to Open Payments. For any comparison, establish the reporting year from the source release: QOPE's selected Provider Drugs profile does not retain that year as a field. Treat suppressed values as unavailable rather than zero, and remember that standardized 30-day fills are not raw dispensing-event counts.
Table of contents
- What the two CMS datasets measure
- Which years are comparable
- What a provider match requires
- A practical workflow for descriptive analysis
- Questions these data can answer and questions they cannot
- Interpretation limits and missing data
- Frequently asked questions
- Sources
What the two CMS datasets measure
| Source | What it measures | Key interpretation |
|---|---|---|
| CMS Open Payments | Reported payments and transfers of value, plus physician ownership or investment interests | A record is a reported financial relationship, not a prescription claim. |
| Medicare Part D Prescribers by Provider and Drug | Aggregated Part D claims and utilization by prescriber and drug | It describes Part D activity, not a prescriber’s full prescribing across all patients and payers. |
Open Payments is broader than pharmaceutical-company payments. Reporting entities include applicable manufacturers and group purchasing organizations, and covered products include certain drugs and biologics as well as devices and medical supplies. Covered recipients include physicians, certain non-physician practitioners, and teaching hospitals. A study focused on pharmaceutical payments to physicians should define and filter its reporting entities, product scope, and recipient type accordingly. CMS's physician coverage excludes medical residents; its employee exclusion concerns a bona fide employee of the applicable manufacturer reporting the payment, rather than employment by any industry company. CMS describes the program’s participants and scope.
Keep the three Open Payments categories distinct. General payments are not connected to a formal research agreement or protocol; research payments are connected to one; and ownership or investment interests are a separate category, including interests held by a physician’s immediate family. The latter should not be treated as an ordinary payment. A reported transfer may also be indirect or designated to a third party, so the listed recipient did not necessarily receive the money personally. For Program Year 2024, the inflation-adjusted reporting thresholds were $13.07 for an individual transfer and $130.66 in annual aggregate; a transfer below the individual threshold could still be reportable if the annual aggregate threshold was exceeded. Publication delays for some research-related records also affect what appears in the data. See CMS’s Open Payments reporting rules.
The Part D file measures utilization among people enrolled in Medicare Part D. Claims include original prescriptions and refills, but a claim count is not the same as a count of unique patients or days of therapy. Standardized 30-day fills are calculated from days supply and adjusted using CMS coding rules; they are not raw counts of pharmacy dispensing events. Beneficiaries are unique Part D beneficiaries with at least one claim for the drug. Drug cost is a reported aggregate that can include ingredient cost, dispensing fees, taxes, applicable administration fees, and amounts paid by plans, beneficiaries, subsidies, or other third parties. It is not simply the amount Medicare paid and does not account for manufacturer rebates. CMS’s Part D dataset guidance describes these measures and their limits.
Which years are comparable
As of November 4, 2026, the latest periods listed by CMS differ: Open Payments Program Year 2025 and Part D Data Year 2024. Open Payments published its 2025 annual data in June 2026. Publication and webpage-update dates describe the release, not the year of prescribing or payment activity.
For a calendar-year comparison, 2024 is a common period: compare 2024 Open Payments records with 2024 Part D claims. Do not compare the newest Open Payments year directly with the newest Part D year and call them contemporaneous. Open Payments records can also be refreshed, so document the publication snapshot or revision used for the 2024 file.
What a provider match requires
Start by separating two identifiers: an Open Payments covered-recipient profile ID identifies a profile within that system; an NPI identifies a healthcare provider. They are not interchangeable, even when both are present.
Check the payment files and recipient-profile information for the exact year being analyzed. CMS's current public-use documentation includes a separately identified recipient NPI, but that does not prove a usable value exists for every recipient or every older release. A 2024 comparison needs checks against the selected 2024 data, not assumptions based on a newer file or an older FAQ. The CMS exploration resources provide access to the datasets and accompanying documentation.
Where a usable NPI is supplied, validate its format, coverage, and consistency with individual-provider identity before linking. Repeated NPIs are expected in payment and provider-by-drug records: the goal is a valid identity key, not one raw row per person. Check uniqueness only after aggregating to the intended provider-year or provider-year-drug grain.
If a reliable identifier route is unavailable, names and addresses can be ambiguous. Report match rates, keep uncertain and unmatched records separate, and use other available attributes to assess matches. An unmatched record is not evidence of zero payments or zero prescribing; do not force a match to make the table complete.
Company and drug fields also require care. Open Payments may identify a reporting company and reported related products; Part D identifies dispensed drugs using brand and generic names. These are not interchangeable identifiers. A company payment cannot automatically be attributed to one brand, generic, product, or indication. If a payment record names multiple related products, duplicating the payment amount across each product would inflate totals unless there is a defensible allocation method.
A practical workflow for descriptive analysis
- Choose a common period. Use Program Year 2024 Open Payments and 2024 Part D data for a calendar-year comparison, and record the Open Payments snapshot used.
- Define the population and measures. Select the relevant reporting entities and individual covered recipients. Keep general payments, research payments, and ownership or investment interests separate.
- Validate identities and grouping keys. Check whether the selected Open Payments release supplies a usable NPI and how complete and consistent it is. Exclude organization records when studying individual physicians. After aggregation, check that each output grouping key occurs once.
- Aggregate before joining. For a provider-year analysis, aggregate payments once to provider, year, and category (and company if needed). Sum additive Part D measures such as claims, standardized fills, and drug cost across drug rows only when the question calls for those totals. Do not sum drug-level beneficiary counts to estimate provider-year beneficiaries: the same person can appear under more than one drug. Keep beneficiary counts at provider-year-drug grain or use a separately deduplicated provider-level measure if available. For a drug-level question, retain Part D at provider, year, and drug, and do not repeat a provider-year payment total on every drug row.
- Preserve missingness. Keep suppressed Part D measures blank, not zero. Report unmatched providers and recognize that reporting thresholds and delayed publication affect Open Payments coverage.
- Describe, then adjust. Compare patterns by relevant characteristics such as specialty and practice size, and account for patient mix where possible. Interpret the results as descriptive associations, not effects of payments.
Hypothetical example: Suppose a provider-year general-payment total is $100 and is repeated onto two drug rows; summing the joined rows would incorrectly make it $200. If the two rows show 20 and 15 beneficiaries, and 10 people appear in both, the unique total is 25, not 35. The provider-by-drug aggregates do not reveal that overlap, so do not calculate provider-wide beneficiaries by summing them. Keep a suppressed measure unavailable and a provider with no reliable match marked unmatched—not as zero.
Questions these data can answer and questions they cannot
With a validated match and a compatible time period, researchers can ask: What reportable financial relationships were disclosed for covered providers? How do Part D utilization measures vary among providers with different payment categories or reported company relationships? Do descriptive patterns differ by specialty, or across brand and generic drug measures?
These data alone cannot establish whether a payment caused a prescription, what diagnosis or indication motivated it, or whether a prescription was appropriate. They do not show that a payment affected a particular beneficiary, represent all of a clinician’s prescribing across payers, or provide a drug’s net cost after rebates. A company’s payment to a provider also does not prove that the provider prescribed that company’s product.
Interpretation limits and missing data
Open Payments is limited to relationships that fall within the program’s reporting scope and rules. A missing record may reflect thresholds, recipient or product coverage, reporting practices, or delayed publication—not necessarily the absence of a financial relationship. Part D counts can also be suppressed: provider–drug combinations with fewer than 11 claims are not included, and beneficiary counts below 11 are suppressed. Missing, suppressed, and unmatched values should remain distinct from zero.
The Part D population is limited to beneficiaries enrolled in Part D. Its measures are not a complete record of a provider’s prescribing, and CMS cautions that the file is not intended to indicate quality of care. Specialty, practice size, patient mix, and other factors may influence both payments and prescribing. Reverse causation is also plausible: higher use of a drug may attract industry engagement. A descriptive association does not establish causation, inappropriate prescribing, or misconduct.
For a practical starting point, inspect the QOPE Provider Drugs dataset, confirm its source period, and choose one prescribing measure. Keep Open Payments as a separate source until both provider identities and the intended comparison grain have been validated. This guide describes a method; it does not report results from a completed linked analysis.
Frequently asked questions
Can I compare 2025 Open Payments with 2024 Part D data?
You can describe those two releases side by side, but the periods do not match. For a common calendar-year comparison, use Program Year 2024 Open Payments and Data Year 2024 Part D claims.
Does a 30-day fill mean one prescription was dispensed?
No. It is a standardized utilization measure based on days supply, not a raw count of dispensing events. It should not be interpreted as the number of prescriptions or pharmacy transactions.
Does a missing payment or prescribing record mean zero?
No. Reporting thresholds, coverage limits, suppression, delayed publication, missing identifiers, and matching uncertainty can all result in absent or unavailable records.
Does a provider-level association show that a payment influenced prescribing?
No. These sources can support descriptive comparisons, but do not establish a causal relationship or explain the clinical context of a prescription.