Provider Pulse
Original research on the US healthcare workforce, computed directly from all 9,417,885 active providers in the national NPI registry. No surveys, no estimates. We publish the numbers and how we got them.
Latest study
RESEARCH
The Praxis MD State Provider Supply Index
A population-normalized benchmark of all 50 states and Washington, D.C. using active individual NPI records and 2024 Census population estimates.
Explore the index & download the data →Research library
RESEARCH
The Fastest-Growing Healthcare Provider Roles by New NPI Registrations
We compared 2019 and 2025 registrations across established provider classifications. Home Health Aide led the ranking at 639.3% growth.
Read the study →RESEARCH
The Most Common US Healthcare Provider Isn't a Doctor
We ranked every active individual provider by specialty. Counselors (644,271) lead the list, the top three behavioral-health roles (1,690,047) outnumber all 1,250,607 physicians combined, and nurse practitioners (521,998) outnumber every single physician specialty.
Read the study →Download every Provider Pulse dataset and view press resources →
What Provider Pulse covers
The NPI registry is the most complete public record of who delivers healthcare in the United States, and almost nobody reads it at full scale. Provider Pulse does. We compute workforce composition, specialty growth, and geographic patterns from the same data that powers the Praxis MD directory, then publish each finding with its methodology so you can check it.
Everything here is free to cite with attribution. If you are a journalist, researcher, or operator who needs a number about the healthcare workforce, this is the source. More studies are on the way, covering specialty distribution by state, the growth of advanced-practice roles, and how provider supply maps to population.
How the data works
Every figure comes from active National Provider Identifier (NPI) records in NPPES, the public registry maintained by CMS. We refresh from the source on a fixed schedule, count exact totals rather than samples, and exclude deactivated NPIs. Each study links to its own methodology so the numbers are reproducible.