Open Payments recipient matching, with the evidence behind every match
Every payment to a clinician or teaching hospital matched to the right registry record, with the evidence behind each match.
See one case, screen by screen ↓Every payment or transfer of value has to reach the federal report against the right covered recipient: the right NPI, the right licence state and the right type — physician, physician assistant, nurse practitioner, CRNA, clinical nurse specialist, nurse-midwife or teaching hospital. The data rarely arrives that way. A rep writes “Dr Jon Ramirez” on a speaker-dinner sign-in; the registry knows him as Jonathan A. Ramirez-Ortega, PA-C, at the practice he moved to in August.
So the matching falls to people. Someone searches the NPI registry, checks the state board, compares the payment record with the HCP master and writes down why, one recipient at a time — for name variants, people who moved practice or state, initials that fit two people, grants paid to a hospital’s foundation and credentials typed wrong. When a recipient disputes a payment in April, that reason has to still be on file.
Estimated split for one transparency analyst working the exception queue, by hand and with the solution.
Six specialist agents clean, match, classify and cite every payment recipient; a transparency analyst decides each uncertain match.
A speaker-dinner sign-in reads “Dr Jon Ramirez”. Thirty-seven payments in three systems point at him, and none matched a registry record on its own. Here is what happens next, screen by screen, in the working solution.
Lena opens the solution for program year 2026: 97.0 % of records matched on their own, 14 exceptions waiting for her — 129 records, $292,474 — and 117 days until the submission window opens on Feb 1, 2027. Her weekly sweep closes Friday, Oct 9.
Agents today: “Nightly batch matched 4,671 of 4,812 records on their own.”
Each exception shows the recipient as captured, the best match, why it needs her, the systems it came from, the records and value at stake and the agents’ confidence. Name variant, two people fit, moved practice, credential unclear, NPI deactivated, paid to a foundation — every one with a due date.
M-2041: 37 records, $18,420, from speaker programs, expenses & meals and consulting contracts. The agents recommend Jonathan A. Ramirez-Ortega, PA-C, NPI 1902448317, at a 93 % match score — below the 0.95 line, so a person decides. Beside it, the source: the Veltrimab speaker-dinner sign-in of Sep 17, 2026 at Sunrise Grill, Round Rock, credential written as “Dr”, signed “J. Ramirez-O.”.
“Read from the source system as captured. Nothing changes there until you confirm.”
Candidate A agrees on specialty and is a variant on name and address; candidate B, John Ramirez, MD, a cardiologist in Houston, scores 0.41. Click any cell and the registry record opens at that field: other name “Jon Ramirez”, credential PA-C, provider type code 363A00000X, practice location 2150 Sunrise Rd, Round Rock, last updated Aug 21, 2026.
“The agents read the registry; they never edit it.”
The reason ties it together with numbered citations: the signature on the sign-in, the registry’s other name and PA-C credential, the move to Round Rock where the dinner was held, the HCP master’s licence PA-08812, and the Texas board showing that licence active. A master-data fix is proposed — credential MD → PA-C and the new address — for the steward.
“Report him as a physician assistant — a non-physician practitioner — not as a physician.”
Last night five source systems sent 4,812 payments. The agents cleaned, matched, classified and cited them in about six minutes: 4,671 matched on their own, 141 went to reviewers. M-2041 keeps its own trace of that run: the normaliser set “Dr” aside, the registry matcher found 6 candidates and hit the other name, the licence check confirmed PA-08812 active to Feb 2028, the classifier read the provider type code as physician assistant.
Lena confirms the match. The recipient is written onto all 37 records, the master-data fix goes to Dana Okafor, the profile is pre-validated against the federal rules with 0 errors, and the records join the program year 2026 submission file, batch 41.
“37 records across 3 systems now carry NPI 1902448317 · physician assistant. Decided by Lena Ortiz.”
The recipient profile shows $18,420 paid this year by nature of payment — compensation for speaking, consulting fee, travel and lodging, food and beverage — by month, every record resolved to the one NPI and the match history. The evidence pack holds the registry, licence and source records with the decision trail.
A preview of how his payments appear when the federal data is published on Jun 30, 2027: Covered Recipient Non-Physician Practitioner, physician assistant, licence state TX, $18,420 across 37 payments. The recipient sees it first, in the review period.
Not every case ends with one click. M-2047 is a $85,000 fellowship grant paid to Lakeview University Medical Center Foundation, a separate charity. The agents recommend reporting it to the hospital, which is on the program-year 2026 teaching-hospital list — and the decision goes to Omar for sign-off.
Teaching hospitals are never auto-matched — a person always decides.
The same recipients, seen by the four people who carry them — what their year looked like, and what it looks like now.
Six specialist agents clean, match, classify and cite every payment recipient; a transparency analyst decides each uncertain match.
Pulls the nightly payment batch from the expense, events, accounts-payable, grants and clinical-payment systems, and labels each line’s source system, nature of payment, product and amount.
Splits names, sets credentials aside, records nicknames and former names as variants, and standardises addresses (suites, ZIP+4).
Scores every candidate from the NPI registry, state licence lookups, the HCP master and the teaching-hospital list, field by field — name, credential, provider type code, address, licence and NPI.
Decides physician, physician assistant, nurse practitioner, CNS, CRNA, CNM, teaching hospital or not covered, from the registry’s provider type code and the licence type — not the credential the business typed.
Writes the short reason for each match, with numbered citations to the exact registry, licence and source passages, and says what the person must decide.
Writes confirmed recipients into the federal file layout and pre-validates NPI, licence and name, reporting any profile error back to the exception list.
Decides each uncertain match. The agents propose; a named person decides.
Anyone on the team can ask about a recipient, an exception or a source system — or tell it a matching rule to add. Answers cite the registry, licence and source records.
Which exceptions are worth the most?
Your five largest open exceptions by value: A. Thompson (PI, site 114), $96,400; Lakeview University Medical Center Foundation, $85,000; Northbridge Oncology Associates, $27,000; Dr. Lucas Ferreira, $22,400; Dr Jon Ramirez, $18,420. I have sorted your list by value.
Why is Jon Ramirez a physician assistant?
Because the NPI registry says so: Jonathan A. Ramirez-Ortega is listed with the credential PA-C and the physician-assistant provider type code, and his other name is “Jon Ramirez”. The rep wrote “Dr” on the sign-in, which is why the payment looked like it went to a physician. He moved to Round Rock on Aug 21 and the Texas board shows licence PA-08812 active. Confirming fixes 37 records worth $18,420.
Confirm credential fixes automatically when the state board confirms them
Done. I added the rule “Confirm a credential fix on its own when one registry record matches name and address and the state board confirms the licence” to Settings → Matching rules. It applies to two of your open exceptions — M-2052 (Angela Brooks, RN) and M-2078 (Marcus Bell, Midwife). Teaching hospitals and “not covered” calls still come to a person.
How ready are we for the February submission?
On track. 117 days until the window opens on Feb 1, 2027; files must be attested by Mar 31. 97.0 % matched on their own, 31,956 records confirmed by reviewers, 3,068 waiting for the team, 14 exceptions waiting for you, and 0 profile errors in the last pre-validation. Next: your weekly sweep closes Fri Oct 9, then year-end pre-validation on Jan 8.
12 screens from the working solution, on its sample data. Pick one to see it large.
How the agents match, who signs off and where the reference data comes from are settings — and new rules are added in plain words.
| Setting | Default | Choose from |
|---|---|---|
| Auto-match line | 0.95 | 0.90 · 0.95 · 0.98 |
| Send master-data fixes to the steward | On — Dana Okafor | On · Off |
| Licence before name | On | On · Off |
| Teaching hospitals always need a person | On | On · Off |
| Confirm credential fixes on their own | Off | On · Off |
| Sign-off: teaching hospitals and foundations | Omar Haddad | Omar Haddad · Dana Okafor · Sam Patel |
| Sign-off: not a covered recipient | Omar Haddad | Omar Haddad · Dana Okafor · Sam Patel |
| Sign-off: duplicate master records | Dana Okafor | Omar Haddad · Dana Okafor · Sam Patel |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample program-year data · “estimated” = our estimate (3–5% of records matched by hand today, at about 1.8 million records a year) · “typical” = published figures (42 CFR Part 403 Subpart I, Open Payments — civil money penalties). People, companies and products named on this page are fictional — characters and sample data in the working solution.
Resolving each payment or transfer of value to the covered recipient it must be reported against — the right NPI, licence state and covered-recipient type. The Recipient Matcher does this for every record from your spend systems, against the NPI registry, state licence lookups, your HCP master and the CMS teaching-hospital list, and keeps the evidence for each match.
The normaliser splits names, sets credentials aside and standardises addresses. The registry matcher then scores every candidate field by field — name, credential, provider type code, address, licence and NPI — and prefers the licence number over the name when both are present. At 0.95 or above with no rule hit, the match is confirmed on its own; below it, a reviewer decides.
Yes. The covered-recipient classifier decides physician, physician assistant, nurse practitioner, clinical nurse specialist, CRNA, certified nurse-midwife, teaching hospital or not covered from the registry’s provider type code and licence type, not the credential typed by the business. Registered nurses and pharmacists are not covered, and a “not covered” call needs a sign-off.
No. The agents never edit a source system or the HCP master. Credential, address and duplicate fixes found while matching are proposed and sent to your master-data steward.
The recipient’s evidence pack: every record behind the payment, the registry and licence lookups at the time of the match, who decided and why. Each decision keeps its evidence for at least five years after publication.
Your expense, events, accounts-payable, clinical-payments and grants systems; the NPI registry weekly file; the CMS teaching-hospital list; state licence lookups; your HCP master; and the federal profile file used for pre-validation.
Yes. A reviewer decides every match below the auto-match line, and teaching hospitals, foundations, “not covered” calls and duplicate master records need a named person’s sign-off. The auto-match line, the sign-off owners and the matching rules are settings your team controls.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your source systems, matching rules and a sample of past spend records — and it goes live once every quality gate has passed. We will walk you through it on your own payments first.
Every speaker program scored against the OIG's eight red flags, before and after it runs.
1-min filmThird-Party Due DiligenceEvery distributor, agent and vendor checked for bribery risk before the contract is signed.
1-min filmInvestigation BuilderEvery investigation starts with its evidence gathered, ordered and cited, and closes in a median of 18 days.
1-min filmWe’ll run the Recipient Matcher on a sample of your own spend records.