How the index works.

Every rate in the index is submitted by a clinician who actually worked the shift. We don’t scrape job boards for posted rates — those are aspirational. We capture what was actually paid.

At submission time, a clinician verifies their NPI once. The submission itself records the specialty, state, city, shift type, engagement (1099 or W2), facility type, and rate. The NPI link is hashed at write time and never re-identifiable from the stored submission.

Submissions are bucketed by specialty × state × shift type. For each bucket with at least 5 submissions, we compute the median and the 10th, 25th, 75th, and 90th percentiles.

Buckets with fewer than 5 submissions are shown but flagged as low sample size. We always publish n alongside any rate. If you see a number on locu.ms, you can see how many submissions backed it.

Outliers more than 3 standard deviations from a bucket’s mean are flagged for manual review before being included in the aggregate. We err on the side of including unusual rates if the submission is otherwise complete and the clinician’s NPI checks out.

• We never share individual submissions with anyone. Not agencies, not employers, not other clinicians. Only the aggregate (median + percentiles + n) is published.
• We never re-identify a submitter. The NPI hash is one-way; there’s no path from a row in the index back to the person who submitted it.
• We never sell the data. Reading the index is free for everyone, including agencies. The product is the transparency, not access to it.
• We never paywall coverage. If we have data for your specialty in your state, you can read it without an account.

The index started in Hampton Roads, Virginia (Norfolk, Virginia Beach, Hampton, Newport News) because that’s where the founder works and has direct access to first-hand rate data. Specialties with verified submissions today: anesthesiology and a small set of related fields, with more entering each week.

Outside Hampton Roads, the index is sparse. We show empty rows rather than fabricated medians — every gap is an honest invitation to be the first submission for that specialty in that geography.

Down the road, we plan to supplement self-reported submissions with public-record data (state-published Medicare and Medicaid rates, court filings, FOIA-released contracts where applicable). Any data ingested this way will be tagged with its source so readers can distinguish self-reported from scraped.

This is methodology v1.0, dated May 2026. Material changes — new aggregation rules, new minimum sample thresholds, changes to outlier handling — will be published with a date and a rationale. Previous versions stay archived.