Updated: August 2016

The Centers for Medicare and Medicaid Services released data in May 2016 on the treatments performed by, and the payments made to, health care professionals who cared for beneficiaries in Medicare's Part B program in the year 2014. (The agency first released this data, covering the year 2012, in 2014.) The data covers more than 986,000 health care providers who collectively received $91 billion in Medicare payments.

Part B covers the services delivered to seniors and the disabled by doctors, optometrists, physical therapists and other practitioners. The program also covers ambulance mileage, lab tests, outpatient services and drugs administered by doctors.

We created an interactive news application called Treatment Tracker that places this newly released data in context, showing how providers compare to their peers — that is, providers in the same state and specialty — in terms of the services they perform.

Treatment Tracker includes pages for every health professional, state, medical specialty and service billed to Medicare.

It relies primarily on three files provided by CMS:

• A summary table of the total number of patients treated by each provider in Medicare's Part B program in 2014 and the total amount paid, among other things. The table also includes aggregate information about the characteristics of each provider's Medicare Part B patients.
• A detailed list of the services performed by each provider in Part B in 2014, including the number of patients treated, the number of times the service was performed, and the number of unique patient visits during which each service was performed (some services, such as drug infusions and skin tag removals, may be billed multiple times during a single visit). The table also differentiates between medical services and drugs. Medicare has published a detailed  methodology online for how it has gathered this data. The data is also available for  download online.
• Data initially requested by health data expert Fred Trotter that shows physician referral patterns. The data we are including shows the number of times Medicare patients visited one doctor (say Doctor A) within 30 days of visiting another doctor (Doctor B), during 2014. While Medicare patients do not need referrals to visit different doctors, this data may trace a doctor's medical network, because many patients seek their doctors' recommendations for specialists.

Two important caveats: Medicare only released data on services and referral patterns if they applied to at least 11 patients. If a service was performed on 10 or fewer patients, CMS redacted it and excluded it from aggregate totals.

Also, the data does not include Medicare Advantage plans, which are the health plans Medicare beneficiaries can choose in place of the traditional program. Nor does it include services delivered to patients with other coverage, such as private health insurance or Medicaid.

At the top of each provider's page, ProPublica displays summary information, including the total number of patients that provider treated in Part B in 2014, total services/procedures performed, and the total paid by Medicare. In addition, we show the average number of services delivered per patient by the provider and the average Medicare reimbursement per patient, as well as comparisons to peers in his or her specialty and state. If a provider performs more services or spends more per patient than 90 percent of peers, a note will appear in red.

Below this is a chart showing how complex the provider's patients are compared to those of peers. Providers often cite patient complexity as a factor in statistical variation. This measure, calculated by CMS, is the average Hierarchical Condition Category (HCC) risk score for beneficiaries. According to CMS, "Those scores estimate how beneficiaries' FFS spending will compare to the overall average for the entire Medicare population. ... Risk scores are based on a beneficiary's age and sex; whether the beneficiary is eligible for Medicaid, first qualified for Medicare on the basis of disability, or lives in an institution (usually a nursing home); and the beneficiary's diagnoses from the previous year." Our chart includes the 10 providers with the highest and lowest scores in each specialty and state, as well as a random selection of up to 80 other peers. The chart only appears when there are at least 20 providers in the specialty and state, and only shows providers who treated at least 20 patients in Medicare Part B in 2014.

In the middle of the page, we show how each service ranks among that provider's services and compare that to its rank among others in his or her specialty and state. We include:

• The code number for the service and a description, as well as whether it was performed in an office or a facility such as a hospital. In most cases, the office-based services are reimbursed differently than those delivered in a facility, and are broken out separately by Medicare.
• Whether the code is for a medical service or a drug. As mentioned above, much of the Medicare reimbursements for drugs is intended to cover the cost of the drugs themselves.
• Descriptions for the services, which come from two places. Services that begin with a letter come from CMS. All other codes and descriptions of the medical procedures are from the Current Procedural Terminology code set, which was created and copyrighted by the American Medical Association and is used here with the AMA's permission. Where practical, we used the AMA's consumer-friendly translation of the CPT descriptor.
• The number of times each service was performed, or units delivered, and the percentage of all of the services provided that represents. Some services, such as drug infusions (e.g., per milligram, per unit, etc.), ambulance trips (per mile) and skin tag removals are meant to be billed multiple times during a single visit.
• This service's frequency rank for this provider and also how it ranks among all those in the same specialty and state. If a service has been performed by fewer than five providers in the specialty and state, you will see an orange symbol. Though this in isolation does not necessarily indicate there's a problem, if your doctor performs a service that so few of his or her peers perform it may be worth asking for additional information about why that's the case.
• The number of the provider's patients who received the service at least once during 2014, and what percentage of all of the provider's Part B patients that number represents.
• The average number of visits that patients made to receive this service. This is calculated by using a field in the data that totals distinct visits on separate days for this service. Since a beneficiary may receive multiple services of the same type (e.g., single vs. multiple cardiac stents) on a single day, "this metric removes double-counting from the line service count to identify whether a unique service occurred," Medicare says.
• We took this figure (which is not shown) and divided it by the number of patients who received the service to arrive at the average number of unique visits per patient. For comparison, we show the average unique visits for this service for all providers in the specialty and state who delivered the service on at least 11 patients.
• The total amount billed to Medicare and the average amount paid by Medicare. These take the average payment per procedure billed and paid, provided by Medicare, and multiply it by the service count. Services performed at a facility appear to cost Medicare less than office-based services because a separate facility fee is not disclosed in this file.

Finally, if a provider had more than 100 office visits for established patients (those he or she has seen at least once before), we also have a graphic breakdown comparing those visits to peers. Such visits are coded on a scale of 1 to 5 (actually 99211 to 99215, with 99215 being the most complicated, using procedure codes created by the American Medical Association).

Nationally, a very small percentage of visits are labeled 99215, but Medicare and its inspector general  have raised concerns about upcoding, or billing for higher-level services than are actually delivered, by some providers. See our story and accompanying methodology on this topic.

Just because a provider has a higher proportion of complex visits than peers is not necessarily an indication of a problem, but it may be worth asking about.

More broadly, we do not include any comparative information when there are fewer than 11 of that kind of specialist in the state. We also do not compare providers to their peers if they recently moved to a different state or if they have not designated a primary specialty in the National Provider Identifier system.

We removed 5,732 providers from our tool because their NPI numbers had been deactivated since 2014.

Finally, when available, for each provider, we show the five providers who saw the most overlapping beneficiaries 30 days before this provider and 30 days after, in 2014. We also show the two institutions that saw the most overlapping beneficiaries during that time frame.

Below are the limitations on the data CMS has listed on its website:

1. It only describes care delivered to Medicare beneficiaries in the fee-for-service program and, as a result, may not represent a provider's entire practice. Providers may also see patients enrolled in Medicare Advantage, those with Medicaid coverage or private health insurance, or those who are uninsured.
2. It does not provide any information on the quality of care delivered by providers.
3. It does not account for differences in the sickness of patients treated by different providers.
4. Medicare pays differently when services are provided in a physician's office versus a facility (e.g., a hospital outpatient department). For services furnished in an office setting, the full payment is included. However, if a service or procedure was furnished in a facility setting, in most cases, this data only includes the payment to the provider and not the payment to the facility.
5. In general, when a provider administers drugs to a patient, the provider purchases the drug and Medicare pays the provider 106 percent of the average sales price (ASP) for the drug.
6. The way CMS counts services may differ by billing code. For example, if the number of services is 2 this may reflect two separate procedures, two 15 minute increments of a service (e.g., a 30 minute office visit), or the delivery of two units of a drug.

Do you see an error in this data?

If you are a provider and you believe your address is wrong, you can change it on the website of the National Plan & Provider Enumeration System. If you believe the procedure data is incorrect or have other questions, you can email CMS at [email protected]. You can also contact us at [email protected].