Since my initial exploration of Medicare’s Physician Payment Database, I have not done much with it. The expectation that the information would be of great interest to many has been validated, and the utility and shortcomings of the data better understood. The potential is great that these data can be used to improve the quality, affordability, and availability of medical care. As might have been predicted however, a great deal of attention has been focused on identifying medical malfeasance and fraud.

I have always been of the opinion that examination of outliers in big data sets like this one is extremely valuable in health policy research. This is especially true in American medicine where there is such great variation both in the frequency in which various medical services are provided, and the amount of money charged. Looking at outliers does not automatically assume that something inappropriate is going on. A place or provider where a large number of things are being done may be a center of recognized excellence. On the other hand, and as we have seen in this series of articles, such “hotspots” of activity may represent inappropriate, abusive, or even illegal medical practice. I believe that large utilization data sets like this one beg us to ask questions about how to use our healthcare resources fairly, efficiently, and most of all effectively. A strategy I recommend is to start by looking more closely at the top 10 and the bottom 10 on any such list. Is that entry there for the best of reasons to be emulated, or for reasons of less value to be corrected? When you are done with the top ten, go on to the next, and so on.

Data is non judgmental, but is it easier to find the bad stuff?
With this as my working background, a series of reports about convictions and settlements related to two oncology practices caught my eye. One practice in Somerset was inappropriately buying non-FDA approved chemotherapy drugs manufactured in Third World countries and selling them here at inflated American prices. In a second practice in Elizabethtown, two physicians were markedly prolonging the duration of chemotherapy infusions in order to charge Medicare at a higher rate. The defensive rationale offered for this unique approach was in my opinion indefensible and I have not yet seen any other medical authority venture a different opinion. Presumably, these practices had been going on for some time, wasting money and exposing patients to harm. I wondered how the named providers in these settlements would have appeared in the 2012 Medicare Provider Payment Database. In fact were at or near the top end of all the charge and payment categories.

From the database of Kentucky physicians, I extracted all the Oncologists (22), Hematologists/Oncologists (93), and Hematologists (12) and prepared an Excel spreadsheet containing the aggregate Medicare data. [I am aware that there are physicians practicing oncology who are listed as “Internal Medicine.” Such individuals are not captured in this analysis.] The Medicare payment data for the designated 128 physicians is presented in a table form that allows me, and you, to sort and rank the information based on any desired element.

Results.

Total Medicare Payment.
The range here was stupendous, from $2000 to over $2 million. (Still less than some rheumatologists in Kentucky!)  To a large extent, the range of payments reflects the number of patients seen. As was true for rheumatologists, it also reflects whether or not chemotherapy drugs are given in the office, or in a hospital. In this ranking, the physicians of interest ranked 7th, 10th, and 15th. The top 9 physicians received more than 2, and the top 4 physicians received more than 3 standard deviations above the mean. These extremes can be considered statistical outliers.

In an effort to correct for patient volume, I divided total Medicare payment by what Medicare defines as the number of unique patients. Although the three physicians of interest were not seeing among the very highest volumes of unique patients (25th, 32d, 38th), the payments per unique patient were ranked at 3d, 6th, and 8th.  The top 7 physicians received more than 2, and the top 2 physicians more than 3 standard deviations above the mean.

Total charges per patient.
Medicare pays providers according to defined fee schedules, but providers are free to charge whatever they want. Correcting total charges for the number of unique patients, the three physicians at issue ranked 2d, 4th, and 20th in highest charges per unique patient.  The difference between the amount charged to the patient and the amount paid by Medicare is often substantial. One (different) physician charged his patients 10 times more than Medicare paid! One is left to wonder how much of the residuals ended up having to be paid by somebody? Total charges do count! Medicare does provide patients protection from excessive charges but there are always copays, deductibles, and non-covered services. Presumably similar charges are being made to non-Medicare patients who are also liable for some or all of the total amount, but may not have government or private insurance protecting their interests.

The top of the lists.
The three physicians who entered settlements with the US attorney were hardly alone at the top of these rankings. Some of these others were clearly statistical outliers.  A health policy scientist, a hospital medical officer, a medical licensing official, a quality evaluating agency, a US attorney, or even a potential patient looking for a way to provide or obtain quality and affordable medical care should want to understand these rankings more fully.

The bottom of the lists.
I do not have any immediate insights into why some providers ranked lower on the list. Patient volume is a factor. Whether or not chemotherapy, laboratory tests, or imaging were provided in the office probably is also important. The fact that providers at the bottom of the Charges/Payments lists were billing for many fewer different HCPPS codes/services supports that billing for laboratory and diagnostic testing as opposed to cognitive services explains (but does not justify) some of the differences. This hypothesis can be tested by looking at the companion database that breaks down each physician’s services by category but which is more cumbersome to use. I will try. Clearly some physicians are more efficient or at least less expensive than others.

Enough for now. I am continuing to learn how to understand and manipulate both the physician and hospital Medicare payment databases. Click on the Medicare button below to see earlier articles in this series.  Many of our readers have downloaded my Kentucky or Louisville subsets. I hope you will share your experience and understanding with the rest of us. How should we be using this information, if at all?

As always, if I have made an error of fact, help me to correct it.

Peter Hasselbacher, MD
President, KHPI
Emeritus Professor of Medicine, UofL
18 July 2014

• Table of Medicare Payments to Kentucky Oncologists (Excel .xlsx 83 KB)