If you’re thinking of having knee replacement surgery and live in DC, MD, or VA, you may want to download and have a look at this Excel (.xlsx) file:
The workbook linked above contains a list of orthopedic surgeons in DC, MD, or VA who performed knee replacements on traditional Medicare enrollees in 2018. It shows the volume of each surgeon’s Medicare-paid total knee replacements, partial replacements, and repairs (revisions) of replacements. A second sheet in the workbook provides counts of fee-for-service Medicare inpatient knee and hip replacements by hospital for the same three states.
Why should you care about how many knee replacements a surgeon has done? In a phrase, practice makes perfect. This is true of most complex surgical procedures such as joint replacements, and is particularly true of partial knee replacements. To quote one carefully-done large scale study out of Great Britain:
Caseload had a profound effect on implant survival. Low-volume surgeons had a high revision rate ... and therefore should consider either stopping or doing more UKR procedures. High-volume surgeons ... demonstrated a 10-year survival rate of 97.5%, which was similar to that reported in registries for the best-performing TKRs.
(N.B., UKR = partial knee replacement, TKR = total knee replacement, revision = repair or replacement).
Why care about Medicare-paid surgeries? Mainly, that’s the only data you can get that shows the number of procedures performed by individual surgeons. The data and methods are public information (accessible at this link). That raw information is impossible for most people to use, so I created the workbook above. As importantly, Medicare is a big piece of this market. In these three states, in 2018, the traditional (fee-for-service) Medicare program paid for roughly half of all knee replacement surgeries (documented below). And so, while this Medicare-based workbook only shows part of each physician’s practice, in most cases it shows a large part of it.
Think of this as a place to start as you decide upon a surgeon and hospital for your knee replacement. Volume of surgery alone is not a direct measure of a physician’s quality or competence. But if you’re going to have a knee joint replaced, you probably want a surgeon who replaces a lot of them. You’ll obviously want to look at more than just surgical volume before choosing a surgeon. But surgical volume is one reasonable criterion. This is one of the few places were you can find orthopedic surgeons known to do a high volume of knee replacement surgery.
For now, you only need to know a few things to use the data.
- There are no warranties or guaranties about the accuracy of this information. Use it at your own risk.
- This is a snapshot of 2018. Things change, people move around.
- These are based on claims (bills) submitted by these surgeons and paid by Medicare. There may be occasional errors, such as the inclusion of a mix of surgery and assistant-at-surgery services for an individual surgeon.
- This is restricted to self-designated orthopedic surgeons. This may omit the occasional legitimate knee surgeon who (e.g.) self-designates as a sports medicine physician or other sub-specialty.
- An orthopedic surgeon had to be paid for at least eleven knee surgeries of a given type in order to be listed. (A “type” in this case is an AMA Current Procedural Terminology (r) code.) A blank entry in this file should best be interpreted as “fewer than eleven surgeries”.
- An orthopedic surgeon practicing in this geographic area might have been correctly omitted from this file because a) they moved here after 2018, or b) the address of their main practice is listed as being in some other state, or c) they did fewer than eleven of every type of knee surgery on traditional Medicare enrollees in 2018.
This workbook also contains a crude ZIP-code based distance measure. You can use a standard Excel method (“filtering”) to find orthopedic surgeons near you. For example, you can easily reduce the list to orthopedic surgeons within 30 miles of a given ZIP code, who did at least 100 Medicare-paid knee replacements in 2018. The README sheet in the workbook briefly explains how to filter the data.
Absolutely nothing about this file is perfect. The counts are incomplete, the distance measure is crude, and so on. But for most users, it’s probably good enough to be useful. If you’ve ever tried to find a specialist, used an on-line website, and been faced with a list of hundreds of names, you’ll understand the utility of having some systematic approach to whittling down the choices. At the very least, if you’ve gotten recommendations for a surgeon, you can now look them up and see whether or not they appear to do a lot of knee replacements. And, as discussed below, you can often use the Medicare Compare website to identify hospitals to which that surgeon admits patients.
The rest of this post talks a bit more about the underlying data source, documents the estimate of traditional Medicare’s share of the market in these three states, and rambles a bit about about why I put this together.
The data, and fee-for-service Medicare’s share of the local knee replacement market.
The Excel workbook above is based on Medicare claims (medical bill) data. In this section, I describe the relevant parts of the Medicare program, and show that for the three states in question (DC, MD, VA), in total, traditional fee-for-service Medicare paid for about half of all knee replacements in 2018. Throughout, I ignore Medicare Part D (drug) coverage.
Roughly 60 million Americans are insured via Medicare, our federal health insurance for the aged and disabled. You can find a concise summary of enrollment statistics at this link.
Of those, just over one-third are actually enrolled in private health care plans, termed “Medicare Advantage” or “Medicare Part C” plans. For those beneficiaries, Medicare simply pays monthly premiums to those plans, just like any other health insurance coverage. With limited exceptions, the Medicare program itself never sees bills (claims) for services provided to those individuals. They are (almost by definition) excluded from the counts in the Excel workbook provided here.
Roughly two-thirds remain in the traditional “fee-for-service” Medicare program. For those individuals, Medicare hires contractors to process and pay their covered health care bills. For those individuals, the Medicare program gets an electronic copy of every bill that was paid. One way or the other, each bill shows what service was provided, and how much Medicare paid for it. Those are the counts that make up the data for the workbook above.
There is a small factor that also must be considered, which is that Medicare covers inpatient facility care (Part A) separately from payment for professional and outpatient services, such as payment to a surgeon (Part B). An increasingly large share of enrollees in traditional Medicare have Part A coverage (which is free) but not Part B coverage (which is merely heavily subsidized, not completely free). (This does not happen in Medicare Advantage, because (with rare exception) you must have both Part A and Part B coverage to be able to enroll in a Medicare Advantage plan.) Working from these enrollment statistics, about five million Medicare beneficiaries have Part A but not Part B. Because these are by definition concentrated solely in the fee-for-service portion of Medicare, this means that currently about 12.5 percent of Medicare fee-for-service beneficiaries have Part A (hospital inpatient bill) coverage, but not Part B (surgeon’s bills) coverage. These individuals tend to be relatively low users of care. But to the extent that such an individual would get an inpatient knee replacement, Medicare would see a bill from the hospital, but not from the surgeon.
For the U.S. as a whole, we can use a reference inpatient database (the AHRQ HCUP database) to estimate the fraction of all inpatient knee replacements paid by any part of Medicare. This will include both traditional fee-for-service Medicare and Medicare Advantage plans. As of 2018, Medicare paid for 57% of all inpatient knee replacements.
But that would include not only traditional Medicare, but Medicare Advantage as well. To parse those apart, I relied on an old analysis that I had done for clients back when I worked in this area. (That analysis required separating out Medicare fee-for-service discharges from records for Medicare Advantage discharges.) Based on that analysis, for knee surgery, for DC+MD+VA combined, traditional Medicare accounted for 90 percent of inpatient knee replacements in 2017. (For the U.S. as a whole, it was more like 70%).
So fee-for-service (traditional) Medicare’s share of inpatient knee replacements, for DC+MD+VA, would amount to 90% of 57%, or 51%.
Finally, I have to guess what fraction of those with Medicare fee-for-service inpatient knee replacement had Part A but not Part B. This is a lot harder, because those without Part B tend to have a much lower rate of service use for elective surgery such as knee replacement. As a reasonable guess, based on years of looking at this question for other procedures, I’d guess that 6 percent of those knee replacement patients had Part A but not Part B.
So fee-for-service (traditional) Medicare’s share of surgeon’s bills, for knee replacements, for DC+MD+VA, would amount to 94% of 51%, or 48%.
The upshot of all of that is that, if I’m looking at fee-for-service knee replacement claims paid by Medicare Part B, in DC+MD+VA, in 2018, I’m looking at just under half of all knee replacements done in this geographic area.
The actual public-use data file is Medicare’s summary of the individual claims (bills). Medicare summarized the file by physician identifier, place of service (inpatient or outpatient), and procedure (AMA Common Procedural Terminology (r) Code). As a privacy protection measure, Medicare blanks the data any time that leads to a count of fewer than 11 total services.
Because of that summary-and-redaction process, I will lose some counts of knee replacements every time a surgeon does fewer than 11 of any one particular type of knee surgery. But that factor is more-or-less irrelevant if the task is to find high-volume surgeons. It might might drop a lot of volume out of the file in total, but it should leave the counts for high-volume surgeons more-or-less unaffected.
I did considerable post-processing of the Medicare file, to achieve two things. First, I edited out aberrant-looking lines, mostly trying to get rid of claims for assistance-at-surgery. (Assistance-at-surgery is exactly what it sounds like — it’s the service of assisting the main surgeon who performs the surgery. Unfortunately, assistance-at-surgery is billed using the same codes as the surgery itself, with a separate “modifier” indicating assistance. Medicare just summarizes all the bills, regardless of modifier.) Second, I added the counts for the range of surgical codes to generate the categories you see labeled on the spreadsheet.
As a validation, I found that after all my edits, and all the CMS redactions of cells with fewer than 11 surgeries, I ended up with 93% of the “benchmark” count of U.S. total knee replacements. (The benchmark is based on the Medicare Part B National Summary file, excluding assistance-at-surgery (80s) claims.)
Why did I put this together.
I worked more than 30 years as a health economist, and spent most of that time analyzing Medicare claims data. After three decades, I was both good and quick at doing that sort of analysis.
During my career, I was repeated floored by how hard it was for the lay person to get an answer to even the simplest questions about health care, based on Medicare’s experience. If somebody in the Medicare program had not tabulated exactly what you wanted, just by chance, then you were out of luck. In most cases, for most questions, my sole option was to work up an analysis, from scratch, directly from the large public-use or limited versions of the Medicare claims and enrollment data files.
Absent that data access and analytical firepower, I could not get answers to obvious and simple questions. And it wasn’t so much that Medicare didn’t occasionally try to provide summary data that could answer some questions. It’s more that most questions you’d like to see answered require just a little bit of analysis that is specific to that question.
And so it is with these counts of surgeries. Medicare makes the raw summary file available, and even has an interactive on-line query system. And even with that, I’d bet that the average American would find it impossible to use that system to produce any sort of meaningful information. You have to know just a little bit — about the codes used to represent the surgeries, about assistance-at-surgery claims — to convert the raw data to a useful listing.
And, since I spent 30 years developing those skills, I figured I should put them to some use.
This is actually the second time I’ve done this analysis. The first time was years ago, when a friend’s wife was facing a replacement of a failing partial (unicoldylar) knee. Her knee was not done correctly the first time, it had worn prematurely, and replacements are much harder than initial implants. She really wanted to find the best of the best for her surgery. And so I did essentially this analysis, and in fact came across the surgeon who literally wrote the textbook on revisions of uni knees.
That story then had a happy ending. But without access to information like this, it’s hard to pick a surgeon. You are reduced to asking friends for suggestions, or soliciting suggestions from social media. Both of which can work, but neither of which is very systematic.
The second time I did this analysis was two weeks ago, when a question about surgery for a uni knee revision was posted on NextDoor. I figured, I might as well dust off the old analysis and see what it said now. And it actually turned up the same national expert on uni knees, at the top of the listing.
Otherwise, finding and choosing a surgery is just a hard task. You scrape together whatever information you can find from various sources. Maybe you solicit recommendations from your family physician, friends, or via social media. And you try to synthesize all of that.
What I have found is that on-line physician finder sites give you an overwhelming number of choices. And when it comes to surgeons, it seems like all of them get top ratings. They are great for showing you all of your options. They are not much good when it comes to narrowing your options.
And that’s why I like practice-makes-perfect as a guide. It’s not merely that you should avoid surgeons with low volumes of the particular surgery you need. It’s that you will find that, particularly with anything out-of-the-ordinary, surgeons with high volumes got that way because they had a reputation for being the local expert. A surgeon doesn’t get to do (say) 80 partial knee revisions per year unless that surgeon has the reputation as the go-to person for that particular surgery.
So there you have it. It’s not perfect, slick, or pretty, but it makes that Medicare data available to people who might want to use it. Now that I’ve worked out the kinks for doing this from public-use data, there’s no barrier to doing a national version of this file, and there are only small barriers to doing this for other common surgeries where the practice-makes-perfect effect is known to be important. As a final note, Medicare updates the underlying data annually in November, so if I’m still in this business at that time, I should be able to update this using newer data then.