How the specialist finder works
Research and trial activity, not a measure of clinical quality.
What it shows, where each number comes from and what it does not claim.
For the physician at a rural or critical access hospital: you refer out more of your serious cases and have the fewest colleagues to ask. Each condition page can be filtered by state, and shows Medicare's rural-urban code where the file has the physician.
What it is
For a serious condition, the finder lists the treating physicians in the U.S. with the most research and trial activity on it: physicians who write the most-cited recent research and lead its trials and who have a federal clinician record in a fitting specialty. Each name carries the evidence behind it, and every number links to its public source.
It follows the do-it-yourself search Eric Topol describes in The Creative Destruction of Medicine (Basic Books, pages 49 to 53): find the most-cited recent papers, note the senior author, check how much that person has published on the condition, and confirm where they work now. The finder does the lookup for you and shows the checks the book recommends before a referral.
It is a starting point for a referral or a second opinion.
What it is not
- It does not say who is best. The order is research and trial activity. It is not a ranking of skill, outcomes or quality of care.
- It shows no complication or outcome rates. Individual rates are not public.
- It uses no patient reviews and no reputation surveys.
- It does not say whether an expert takes a patient's insurance, accepts outside referrals or offers a tele second opinion. Ask.
- It is not medical advice.
How a name gets on a list
- Papers. We search PubMed for research papers on the condition from the last 5 years. Reviews, meta-analyses, editorials, letters, comments, case reports, news, retraction notices and corrections are left out. The senior author is the last named author, counted when that author's own affiliation is in the U.S. A group name at the end of the author list (such as "PARTNER 3 Investigators") is not an author, so the last named person before it is the senior author, as in PubMed's own last-author search. A paper with only one named author has no senior author and is not counted.
- Citations. NIH's iCite gives each paper a Relative Citation Ratio, which adjusts for field and year. A ratio of 1.0 is a median NIH-funded paper. People are ordered by the sum of those ratios for their senior-author papers. That measures research activity. Google Scholar is not used.
- Rising authors. Two of the ten places are kept for people with at least three senior-author papers in the last 2 years, so newer experts are not missed. Those entries say so.
- Trials. ClinicalTrials.gov studies on the condition that list the person as an overall official (principal investigator, study chair or study director), with the status and whether industry is the lead sponsor.
- Clinician record. The federal NPI registry (NPPES) shows whether the person is a licensed clinician, in a specialty that fits the condition, and where they practice. People without a fitting record are not listed: the finder lists treating physicians only.
How a name is matched to a person
A name alone is never enough, because the same name can belong to several clinicians. A person is matched only when independent signals agree: the practice state in the federal registry matches the affiliation on the papers, an ORCID record lists the same employer in the same state, a ClinicalTrials.gov record lists the same name and institution, or a person on our team checked an official page of the institution.
- Confirmed: two independent signals agree.
- Likely: one signal agrees and nothing contradicts it.
- Anyone else is not shown.
A registry record in a different state from the papers is never accepted without a second signal. Very common names always go to a person for review before they are shown. Each entry lists the signals that matched it. Where a reviewer checked a page, the entry links it and puts the page's words in quotation marks only when they are its own words, word for word; otherwise it gives a summary. When the registry's practice states do not include the state on the person's most recent paper, the entry says so beside the locations, because either record may be out of date: check the location before referring. An identity can still be wrong, so every entry has a correction link.
Medicare volume and industry payments
- Medicare volume. For treatments with clear billing codes, the page shows the physician's Medicare services in the latest data year. It covers Medicare Part B fee-for-service only, not Medicare Advantage or commercial insurance. CMS leaves out any row derived from 10 or fewer patients, so a figure means at least 11 patients and a blank means unknown, not zero.
- Industry payments. From CMS Open Payments: payments from the companies that make the main drugs and devices for the condition. Each company is counted with its affiliates as reported to CMS: every reporting entity of the same company group, by the exact name CMS publishes (for example Boehringer Ingelheim International GmbH with Boehringer Ingelheim). Totals are all payments from those companies and their affiliates in the program year, not only for the condition's product. In CMS's words, "financial ties alone do not signify an inappropriate relationship" (MLN Matters SE1303). It is context, not a judgment.
- Figures from CMS are shown as CMS reports them. CMS does not guarantee their accuracy. Our groupings and summaries are ours and are not CMS data.
Who this favors, and what we do about it
Counting papers and trials favors academic centers, English-language journals, senior physicians and groups that hold more senior-author positions. A skilled community operator who does not publish will not appear on a list. Each list says so in its first line.
Where a billing code fits the treatment, the page also shows how many physicians billed Medicare for it in each state, and how many practice in small towns and rural areas, so experienced operators outside the research centers are visible.
For rural and critical access hospitals
On each condition page you can filter the treating physicians by practice state. Where the Medicare file has the physician, the page shows the rural-urban code of the practice. Before you refer, ask about outside referrals, the patient's insurance and a tele second opinion. The page cannot answer them.
Contact information
Each physician shows the practice location and the practice telephone number that the federal NPI registry (NPPES) lists: the street lines, city, state and ZIP code, as the registry gives them, with the date the registry record was last updated. A physician with more than one practice location in the registry shows each. Where our team recorded an official page about the physician on an institution's own site while checking identity, the entry links it as the institution profile.
Registry entries are filed by the physician or an office and can be out of date, so confirm before referring. We use only the registry's practice location, never its mailing address. We show no fax number and no email address, and nothing from the affiliations printed on papers.
Distance
Distance is the straight-line distance in miles from the visitor's location, ZIP code or city to the nearest practice location the physician has in the federal registry, measured to the center of that location's ZIP code (or of its city when the ZIP code is not available), rounded, and not a driving distance.
ZIP code and city centers are the internal points in the U.S. Census Bureau's 2026 Gazetteer files (public domain), and the arithmetic runs in the visitor's browser, so a location or ZIP code is never sent to us.
How often it is updated
PubMed, iCite, ClinicalTrials.gov and the NPI registry are refreshed monthly. Open Payments is refreshed each year after CMS publishes it, and the Medicare file each spring. Every condition page shows the date of each source.
Corrections
Every entry has a Request a correction link to our contact form. Tell us the physician's name, the condition and what the source says. We answer within 5 business days, and an identity correction removes the entry until it is resolved.
Money
No hospital, physician or company can pay to be listed or to be listed higher. We take no money from hospitals, physicians or companies for this tool, and none may sponsor it or advertise beside it.