Our methodology starts with public federal data: every number on contact.doctor comes from it. This methodology page explains which datasets we use, how we combine them, and what they can and can’t tell you.
Data as of: NPI registry November 2026 · Medicare prescribing 2024 · Medicare services 2023 (by procedure) and 2024 (per provider) · Open Payments 2025 · Census population 2020–2024.
Methodology: data sources
| Dataset | Publisher | Years | Updated |
|---|---|---|---|
| NPPES National Provider Identifier (NPI) registry | CMS | November 2026 release | Monthly |
| Medicare Part D Prescribers, by provider and drug | CMS | 2024 | Yearly |
| Medicare Physician & Other Practitioners | CMS | 2023 (by service), 2024 (by provider) | Yearly |
| Doctors and Clinicians national file (Care Compare) | CMS | Current | Monthly |
| Open Payments | CMS | Program year 2025 | Yearly (June) |
| Health Care Provider Taxonomy code set | NUCC | Version 26.1 | Twice a year |
| American Community Survey 5-year estimates | U.S. Census Bureau | 2020–2024 | Yearly |
| Cartographic boundaries, Gazetteer and ZIP–county relationship files | U.S. Census Bureau | 2024–2025 (relationships: 2020) | Yearly |
| Physician license files with first-issue dates (list of states and sources) | California, Texas, Illinois, Washington, Colorado, Connecticut, Delaware, Vermont and Alabama licensing agencies | Current | Daily or monthly (we refresh with each update). Illinois data: Open Database License; our matched Illinois file is shared under the same license. |
| Drug labels, recalls and shortages (drug pages) | U.S. Food and Drug Administration, openFDA (drug labels, NDC directory, enforcement reports, drug shortages) | Current | With each data refresh |
| Research papers and journals by field (specialty pages) | OpenAlex (open index of scholarly works, CC0); papers with at least one author at a U.S. institution | Last two years and this year | With each data refresh |
| Training length by specialty | ACGME program requirements (usual minimum years); American Board of Obesity Medicine | Current | Checked by hand |
Who is counted
- Every individual provider in the NPI registry with an active NPI (not deactivated, or reactivated) and a practice address in the United States, including Puerto Rico: about 7.5 million people. We keep organizations (about 2 million) separate and don’t count them on the density maps.
- Physicians are providers whose primary taxonomy code is in the NUCC group “Allopathic & Osteopathic Physicians” (MDs and DOs): about 1.20 million with a mapped practice county. However, providers report their own credentials, so we also check them: a provider under a physician code whose credential names another profession and no medical degree (for example “NP”, “PA-C”, “DPT”, “DDS” or a PhD alone) does not count as a physician. As a result, we exclude about 19,000 registrations. In contrast, we do count providers with a blank credential.
- We count each provider once, under their primary specialty: the taxonomy they marked as primary, or the first one listed.
How specialties are grouped
For physicians, we group by specialty regardless of the parent field. For example, “Addiction Medicine” under family medicine, internal medicine or psychiatry is one specialty. For other clinicians, however, we use the NUCC display name, so a family nurse practitioner and a family psychologist stay separate. Combined groups on the map, such as “Primary care” or “Mental health professionals”, list the specialties they include in the map’s menu.
The menus show about 85 physician specialties and 30 other clinician types. In addition, we count narrow subspecialties with their main specialty (for example, glaucoma specialists with ophthalmology, and geriatric psychiatry with psychiatry). On the other hand, we don’t offer codes that describe a role or a lab science rather than patient care, such as Independent Medical Examiner, Radiological Physics or Ph.D. Medical Genetics. Some NUCC titles appear both as physician and non-physician codes (Legal Medicine, Ph.D. Medical Genetics); we keep the two apart.
Maps and search count specialties differently. Maps and rankings count each clinician once, under the primary specialty marked in the NPI registry. Doctor search, however, also finds clinicians who list a specialty as an additional one: many endocrinologists, cardiologists and other subspecialists mark internal medicine as their primary specialty. A search for a specialty can therefore show more clinicians than the map counts for the same place. Also, results matched this way show the primary specialty as well.
How providers are placed on the map
- First, we use the 5-digit ZIP code of the practice address. Then we match each ZIP code to the county that holds most of its land, using the Census ZIP Code Tabulation Area (ZCTA) to county relationship file.
- However, where a ZIP code spans several counties, we use the county that contains the provider’s city (Census place centre), when that county is one the ZIP code touches.
- Virginia’s smaller independent cities share ZIP codes and mailing names with the counties around them, so, following the U.S. Bureau of Economic Analysis, we combine 28 of them with their surrounding county (for example, James City County & Williamsburg). Larger independent cities, such as Richmond, Norfolk and Virginia Beach, stay separate.
- Similarly, Connecticut replaced its counties with nine planning regions in 2022, so we match ZIP codes there to the nearest planning region.
- Finally, map locations are ZIP code centres, not street addresses.
What the numbers mean
- Per 100,000 residents: the number of providers in an area divided by its population (ACS 2020–2024), times 100,000.
- None: no provider of that specialty lists a practice address in the area. People there may still get care in a nearby county.
- Prescriptions per 1,000 residents aged 65+: Medicare Part D claims (each fill or refill counts once) written by prescribers in an area, divided by its residents aged 65 and over (ACS 2020–2024), times 1,000. Part D also covers some people under 65 with disabilities, so this is a rate for comparing places, not the share of older adults taking a drug. In other words, we place prescriptions where the prescriber practices, not where the patient lives.
- Payments per resident (payments map): Open Payments general payments placed in the county of each clinician’s practice address, divided by the area’s population (ACS 2020–2024). The main view and company views leave out royalties and license fees and acquisition payments, which a few very large payments would dominate. Drug views use the products each payment record names (up to five per payment), matched to our drug list by name or NDC code; a payment naming several drugs counts under each.
- Left out of payment totals: one 2025 record (a single $64.3 million speaking and faculty payment to one clinician, about 18 times the next-largest) looks like a reporting error, so we leave it out of maps and totals; it remains in CMS’s data. We also leave out payments to teaching hospitals: CMS identifies a hospital by its own number, which can coincide with a clinician’s record number, so we count only the payments CMS lists for individual clinicians.
Prescribing, colors and rankings
- None reported (prescribing maps): CMS removes any prescriber and drug pair with fewer than 11 claims, so small counties can show none even when doctors there wrote a few prescriptions.
- Map colors use quantiles: each color step holds about the same number of areas. In addition, each legend shows the break values.
- Rankings of counties, similarly, include only counties with at least 50,000 residents, because a few doctors can swing the rate in a small county.
Limits of our methodology
- The NPI registry is self-reported. For example, some clinicians who have retired or no longer see patients keep an active NPI, so counts run higher than the number actively practicing. Also, addresses can be out of date.
- Some providers list a billing or headquarters address rather than where they see patients. As a result, counts in cities can run high.
- Medicare data covers Medicare patients only and lags about two years. CMS hides counts under 11 to protect patient privacy.
- Open Payments lists payments that drug and device makers report to CMS. A payment isn’t evidence of wrongdoing, and doctors can dispute records with CMS.
- Prescribing and payment patterns describe what happened in Medicare. Therefore, they aren’t measures of a doctor’s quality.
License and attribution
CMS and Census data are public domain works of the U.S. government. The National Uniform Claim Committee (NUCC) publishes the provider taxonomy. Map boundaries come from the U.S. Census Bureau.