Sources and methodology

Datasets and how current they are

"Last retrieved" is when our copy was rebuilt from the agency's file, and is printed again at the foot of every section that uses it. A rebuild only happens when the agency has republished, so a date that stands still means the source has not changed.

NPPES - the provider registry

Every clinician and healthcare organisation billing a US health plan must hold a National Provider Identifier, and the registry behind it is public. It supplies names, credentials, practice and mailing addresses, phone numbers, self-declared specialties (taxonomy codes) and the state licence numbers a provider entered. A further 351,911 identifiers are deactivated; they are held out of search rather than shown as live listings.

CMS Care Compare - hospital quality

The overall star rating is CMS's own composite of many measures. The patient-survey percentages are HCAHPS results as published, shown against the state and national averages CMS publishes alongside them. Where a hospital is called better, no different or worse than the nation, that is CMS's statistical comparison for that measure; we neither compute nor adjust it. Hospitals are joined to the survey and measure files by their CMS certification number (CCN).

Emergency department: from CMS's Timely and Effective Care files - the median time patients spent in the emergency department before leaving, the share who left before being seen, stroke brain-scan timing and sepsis care, each beside CMS's own state and national figures. CMS also publishes national ED times by emergency department volume, which we show so a very busy department is compared with its peers. These are medians over a year, not live wait times.

What we derive: on the hospital hub, a state's average star rating (the plain mean of CMS's ratings over the hospitals in that state that have one) and the count of hospitals CMS does not rate by hospital type; on a hospital page, the tally of measures CMS marked better or worse.

Nearby comparison: the hospitals set beside one another on a hospital page are chosen by us, by location only: the same city first, then the same county, then the same 3-digit ZIP area, closest ZIP code first. General hospitals are compared with general hospitals (acute care, critical access, VA and military), and psychiatric or children's hospitals only with their own kind. The order says nothing about quality, and every figure in the table is CMS's own.

CMS Open Payments - industry payments to teaching hospitals

Under the Sunshine Act, drug and device makers report what they pay teaching hospitals. Yearly totals come straight from CMS's per-hospital aggregate. For the latest program year we read every individual payment record for teaching hospitals and sum them by nature of payment, by paying company and by research study; the amounts are as the companies reported them, and every hospital links to its record on CMS's site. A payment is disclosure, not a finding of wrongdoing: at most teaching hospitals the bulk is research funding and royalties on licensed inventions.

CMS Medicare Physician & Other Practitioners - a doctor's Medicare practice

For each provider who billed Original Medicare Part B, CMS publishes how many Medicare patients they treated, how many services they billed, what they charged, what Medicare paid, and their patients' age mix and chronic conditions. It covers only traditional Medicare - not Medicare Advantage, Medicaid or private insurance - so it shows part of a practice, and it is published about two years after the year it describes. CMS withholds any count below 11 patients.

What we derive: the "typical" values shown beside a provider's figures are the median across all providers CMS lists under the same provider type (for example, all Internal Medicine providers), computed by us from the same file. They are context, not a benchmark of quality.

HHS-OIG exclusion list (LEIE)

The Office of Inspector General lists individuals and entities excluded from federal health care programs. We match it to a profile by NPI only - never by name, because a name match can point at the wrong person and this is not something to get wrong about a named clinician. Most LEIE entries carry no NPI, so "no exclusion under this NPI" is not proof: the OIG's own search is the official record.

What the data cannot tell you

Specialties are self-declared. A provider chooses their own taxonomy codes when registering. The registry does not verify that they practise it, and one provider often carries several.

Addresses go stale. The "record last updated" date on a profile is when the provider last edited their own entry - some have not touched it in a decade. Practices move, phone numbers get reassigned.

A licence number is not a licence check. The registry stores what was entered. Whether that licence is currently active, restricted or revoked is held by the state board, not here.

Volume is not quality. How many Medicare patients a doctor sees, or what Medicare paid them, says nothing about how good their care is. Doctor listings are ordered by relevance or alphabetically, never by merit.

Hospital measures are hospital-wide. They do not rate an individual doctor, nurse or department, and a measure CMS could not compare (too few cases) is shown as such rather than guessed.

Corrections

We cannot change a federal record - the provider or the agency corrects it at the source, and the correction reaches us at the next refresh. If a page here misrepresents what the source actually says, that is our bug and we will fix it: see how to report an error.