The American Medical Association (AMA) is committed to transparency and supports the release of data that can help improve quality of care. For that reason, the AMA believes that certain safeguards are needed to ensure accurate information is presented to the public. Given that CMS has once again released Medicare claims data without pre-verification by physicians to ensure accuracy and with little context, members of the media will be integral to ensuring that the public gets clear, accurate information.

As you may be aware, there were several misrepresentations in the media of the Medicare claims data released in 2014. Thus, the AMA highly recommends that reporters consider the following concerns about the limitations of this data before utilizing it in potential stories or publications. Medicare claims data is complex and can be confusing and the manner in which CMS is broadly releasing physician claims data, without context, can lead to inaccuracies, misinterpretations and false conclusions.

In order to ensure that the information reported is helpful, accurate and complete, media should take into account the following:

1. Errors: Data being released may contain errors because there is currently no mechanism for physicians and other providers to review and correct their information.  For example, it is not uncommon for Medicare’s data base to list physicians with wrong addresses, wrong specialties and wrong affiliations. In addition, inaccurate claims data may lead to errors in the reporting of what services physicians actually performed. These types of errors have the potential to distort CMS calculations and comparison of charge and payment data for the physicians.
2. Quality: The data does not include explicit information on the quality of care provided or quality measurement. It solely focuses on payment and utilization of services so it cannot be used to evaluate the care actually provided. For example, a physician who has provided more than the average number of prevention services or ordered more lab tests for certain patients may actually have better outcomes than another physician with lower utilization of these services. Also, the link between volume and quality is not well-defined so that while some observers may regard physicians who perform a particular procedure more frequently as more experienced and therefore higher quality, others might argue that the same physician is performing the procedure too often.
3. Number of Services: Residents, physician assistants, nurse practitioners and others under a physician’s supervision can all file claims under one physician’s National Provider Identifier (NPI), which can make it appear as though a single physician performed far more services than was the case. Additionally, there are several instances in which it can appear that two surgical procedures were done when in fact there was only one. For example, when there are co-surgeons or an assistant present during a surgery, the procedure should be counted as only one surgery, not two. This poses  the very real possibility that  the information will be misinterpreted— leading to fallacious accusations of fraudulent behavior based on assumptions that physicians are billing for more services than they are.
4. Patient population: The data being released is an incomplete representation of the services physicians provide, as it is not risk adjusted. Additionally, it does not include care for private insurance patients or Medicaid beneficiaries making it a limited view of the patients a physician cares for. Without adjustments for differences in patients’ health and socioeconomic status, physicians who treat the sickest and most disadvantaged patients will be wrongly labelled as outliers in public reports. Failure to include data from other payers may under-represent a physician’s true experience and/or lead to invalid calculations based on a small number of patients.   
5. Site of service: The data does not include facility fees, which could change the amount ultimately paid by Medicare. To reflect practice cost differences, Medicare pays physicians less for services provided in a hospital or hospital outpatient department than for the same service in a physician’s office. Yet for services in the hospital or its outpatient department, there is a payment to the hospital as well as the physician so that the combined payment and beneficiary cost-sharing for the service is often higher than the payment for the same service in the physician’s office. Patients seeking to lower their cost of care might therefore decide to seek care at an outpatient facility rather than in their long-time physicians’ office only to end up paying more, not less, for their care.
6. Provider comparisons: There is a lack of context and specificity in specialty descriptions and practice types in the data, which could be misleading when making comparisons between physicians and specialties. For example, average payments within specialties differ due to the mix of patients and services the specialty typically provides. Also, in some cases, physicians who appear to have the same specialty can serve very different types of patients thus impacting the mix of services provided. For example, internists, also known as hospitalists, who focus on hospitalized patients will be compared to internists who treat patients mostly outside the hospital, which could lead to the possible misrepresentation of the hospitalists as high cost outliers.
7. Missing information: The data does not account for patient mix and demographics. Additionally, it does not point out that a significant share of Medicare payments is used to cover such costs as office overhead, employee salaries, supplies, and equipment. Payment rates are designed to reflect these osts and are therefore higher when a service requires highly-trained staff or high cost supplies and/or equipment. 
8. State Comparisons: As CMS has noted, the data does not reflect adjustments that Medicare makes regarding geographic cost or practice differences. This is an important contributor to payment differences across states. The mix of specialties and patients and site of service variations within states also contributes to differences in average payments per physician.  
9. Coding and billing changes: Any analysis using the data should take into account changes in Medicare’s coding and billing rules that may be different over time and across regions of the country. For example, an analyst who was unaware of a 2010 Medicare policy changed regarding physician consultations might have viewed the uptick in high level visit codes that followed as evidence of fraud when in fact this was the expected result of a Medicare directive.  In addition to changes in national policy, Medicare payment and billing reflects “local coverage determinations” imposed by the program’s contractors. A service that is routinely covered in one part of the country may be generally denied in another, thereby potentially hampering accurate comparisons of physician practice across regions.