How we count providers, and what the density figure means

Every place page here leads with a number: providers per 10,000 residents, plotted against every other place in the state. It is the most useful single figure on the site and also the easiest to over-read. This is how it is built and where it goes wrong.

How the number is built

The numerator is the count of individual providers whose federal registration lists a practice address in that place, whose primary provider type falls within our scope, and whose NPI is not deactivated. Organisations are excluded — only individual practitioners are counted.

The denominator is the Census Bureau’s population estimate for that place.

The comparison is the distribution of the same figure across every place in the state that has both a population estimate and at least one provider. The strip chart shows that distribution; the marker shows where this place sits; the dashed line is the median.

The median is used rather than the mean because the distribution has a long right tail — a small town containing a clinic can post an enormous ratio, and a mean would be dragged toward those cases and stop describing anywhere real.

Four ways the number misleads

1. Providers register one address; they serve a wider area.

A therapist in the county seat may see clients from five surrounding towns. The count lands entirely in the county seat. So a small town’s low figure and the neighbouring city’s high figure are both somewhat artefacts of where an address happens to be. Read your own place and its neighbours — the state page lists every place and is searchable.

2. Registered is not practising.

An NPI persists after someone retires, changes career or stops taking clients, because nothing forces an update. Every count here is therefore an upper bound on the providers actually available. It is a consistent upper bound — the same overstatement applies everywhere — so comparisons between places remain meaningful even though the absolute number is optimistic.

3. A dense place can still be unreachable.

Density measures supply, not access. A city with many providers can still have long waits, and most of that supply may be out of network, private-pay, or full. The figure tells you how hard the search is likely to be, not whether it will succeed.

4. Licence mix is not comparable across states.

States license under different titles, and the same underlying work appears under different codes. A state’s mix of clinical social workers to licensed counselors reflects its licensing history as much as anything about practice. Compare totals across states; be careful comparing the composition.

What we deliberately exclude

Deactivated NPIs. Formally retired registrations.

Organisations. Clinics and group practices hold their own NPIs. Counting both would double-count the same people.

Assistants and technicians. Physical therapist assistants, occupational therapy assistants, registered behavior technicians. They practise under supervision, and listing them as findable providers would mislead.

Neurology. The federal taxonomy files psychiatry and neurology under one classification. Taking it whole would put epilepsy and sleep medicine specialists into a therapy directory, so we list the psychiatry codes explicitly instead.

Removal requests. Anyone who has asked to be removed is excluded from every count, not merely hidden from listings. That means our totals are very slightly below the raw registry, and we would rather it that way.

Places with no population figure

Census designated places — unincorporated communities like Bethesda, Maryland — appear on this site with full provider counts but no per-capita figure, because the population series that covers incorporated places does not cover them.

Those pages show counts without density rather than an invented denominator, and the places are excluded from the distribution charts so they cannot distort them.

Why we lead with it anyway

Because the alternative is worse. “Austin has 6,967 providers” is a number nobody can interpret. “69.5 per 10,000, well above the state median of 14.6” tells you something you can act on — that your search here is likely to be easier than in most of Texas.

No paid directory can produce this figure, because they only know the providers who bought a listing. It requires the complete public record, which is the one thing this site has and they do not.

The full method, including the geographic matching and its failure modes, is on the methodology page.