Where Doctor Data on T-MedInfo Come From and Their Limits

The datasets behind a profile

Our methodology page lists every dataset we use and the date of our latest load for each. In short:

NPPES NPI Registry

The NPPES downloadable file gives names, NPI numbers, addresses, phone numbers, taxonomy (specialty) codes, license numbers as the provider reported them, and deactivation dates. CMS publishes a full file every month and weekly updates. Providers enter their own records, and the rule at 45 CFR 162.410 requires them to report changes within 30 days, but a record can still be out of date.

NUCC taxonomy

The Health Care Provider Taxonomy code set supplies the specialty names behind the codes in NPPES. NUCC publishes new versions twice a year. A code is the specialty a provider selected, not a board certification.

Doctors and Clinicians file

CMS's Doctors and Clinicians National Downloadable File gives medical school, graduation year, Medicare assignment, the telehealth indicator and group practice links for clinicians enrolled in Medicare. The CMS data dictionary notes that it has no board certification or residency data.

Medicare Part B services

The Medicare Physician & Other Practitioners datasets list the services each clinician billed to Original Medicare: service codes, counts of services and patients, and average charges and payments for each calendar year.

Medicare Part D prescribing

The Medicare Part D Prescribers dataset lists drugs prescribed to people with Medicare drug coverage, with claims, days' supply, drug cost and patient counts for each calendar year.

Open Payments

Open Payments records are payments that drug and device companies report to CMS. CMS's data page says it publishes the previous program year on or before June 30 and refreshes the data each January.

Medicare enrollment

The Medicare Fee-For-Service Public Provider Enrollment dataset shows whether a provider has an approved Medicare enrollment, and the Order and Referring dataset shows who is eligible to order or refer certain Medicare services. Enrollment is a billing status, not a quality rating.

How current each dataset is

Two kinds of timing matter:

• Registry data (NPPES, enrollment) describe the current state of a record as of the publisher's last release. A provider who moved last week may not show the new address yet.
• Claims data (Part B, Part D) describe a calendar year that has ended. CMS's methodologies describe each release as covering one year of claims, published well after the year closes, so the latest year on a profile is usually one or two years old.

We load a new release after the publisher makes it available, and our loads are not on a fixed schedule. The methodology page shows, for each dataset, which release we have loaded, so you can see how old a figure is.

Why most Medicare data cover only Original Medicare

CMS's Part B methodology says the Part B data are based on fee-for-service claims and do not include people in any Medicare Advantage plan. The Part D data are different: CMS's Part D prescriber methodology says they come from prescription drug event records for people in both stand-alone drug plans and Medicare Advantage drug plans.

Neither includes Medicaid, private insurance or patients who paid out of pocket. A clinician who mostly sees younger, privately insured patients can have little or no Medicare data even with a busy practice.

Small-count suppression

To protect patient privacy, CMS leaves out small numbers:

• Part B: the methodology says any aggregated record derived from 10 or fewer beneficiaries is excluded.
• Part D: the methodology says records based on fewer than 11 claims are left out and other counts between 1 and 10 are suppressed.

So a missing service or drug does not mean the clinician never provides it, and adding up the rows on a profile gives less than the true total. Where a count is suppressed, we show a dash, not an estimate.

Why our profile may differ from Medicare.gov today

Medicare.gov's Care Compare tool and T-MedInfo both draw on CMS data, but differences are normal:

• Different releases: Care Compare may already show a newer release than the one we have loaded, or the other way round for a short time.
• Different sources for the same field: the CMS data dictionary describes practice locations in the Doctors and Clinicians file by Medicare enrollment record, while addresses on our profiles come from NPPES, and the two can disagree.
• Different scope: Care Compare covers clinicians enrolled in Medicare; we also list providers who have an NPI but no Medicare enrollment.

When the two disagree on something that matters to you, such as whether a clinician accepts assignment, call the office.

What the data cannot tell you

None of these datasets measures quality of care, outcomes, communication or availability. The CMS methodologies say so directly, and we repeat it next to each section of a profile. The data also do not show board certification, hospital privileges, malpractice history, insurance networks or whether a clinician is accepting new patients. Our guide on researching a doctor lists where to look for those. Our guides to Part D prescriber data and Open Payments records explain two of the datasets in detail.

How to request a correction

Because we show the data as published, most corrections must be made where the data come from:

• Name, address, phone or specialty codes: the provider updates the NPPES record, and our next NPPES load picks up the change.
• Medicare enrollment, assignment or group practice: the provider updates their Medicare enrollment with CMS.
• Open Payments records: the clinician disputes the record through the CMS review and dispute process, and the company corrects it.
• Medicare billing or prescribing data: these come from CMS's yearly releases as published, and we do not edit them; questions about the figures go to CMS.

Providers and authorized representatives can also use our data update request to ask us to update, correct or suppress their published contact information. Patients who spot an error can let the provider's office know, since the office is the one that can fix the source record.