10 Things You Can Do With CMS Data

farzad_mostashariFive years ago, my mother needed an orthopedic surgeon for a knee replacement. Unable to find any data, we went with an academic doctor that was recommended to us (she suffered surgical complications). Last month, we were again looking for an orthopedic surgeon- this time hoping that a steroid injection in her spine might allay the need for invasive back surgery.

This time, thanks to a recent data dump from CMS, I was able to analyze some information about Medicare providers in her area and determine the most experienced doctor for the job.  Of 453 orthopedic surgeons in Maryland, only a handful had been paid by Medicare for the procedure more than 10 times.  The leading surgeon had done 263- as many as the next 10 combined. We figured he might be the best person to go to, and we were right- the procedure went like clockwork.

Had it been a month prior to the CMS data release, I wouldn’t have had the data at my fingertips. And I certainly wouldn’t have found the most experienced hand in less than 10 minutes.

It’s been a couple of months since the release of Medicare data by the Centers for Medicare and Medicaid (CMS) on the volume and cost of services billed by healthcare providers, and despite the whiff of scandal surrounding the highest paid providers (including the now-famous Florida ophthalmologist that received $21 million) the analyses so far have been somewhat unsurprising. This week, coinciding with the fifth Health DataPalooza, is a good time to take stock of the utility of this data, its limitations, and what the future may hold.

The millions of lines of data was exactly as advertised: charges and paid services under traditional Medicare “fee-for-service,” including the billing provider’s ID and the costs to Medicare. The initial headlines touting “Medicare Millionaires” relied on some basic arithmetic and some sorting.  And the cautions piled up: the data could reflect multiple providers billing under a single ID; payments are not the same as a provider’s actual take home income; it’s not complete information as it doesn’t contain information about other insurers, or even Medicare Advantage, and so on.

But perhaps most damning was how little insight the data seemed to provide on the quality or value of care provided, as opposed to volume of services.  As Lisa Rosenbaum wrote in the New Yorker, “So much of that good isn’t captured by these numbers. You don’t bill for talking to a patient about how he wants to die. There’s no code for providing reassurance rather than ordering a test.”


Where is the value in the data?

Data bear witness to the fundamental flaw of the payment system that generates them.The absence of information on quality, safety, appropriateness, or outcomes appears to have been a genuine revelation to many, but it is in fact exactly the type of output that we should expect from this volume-based system that we have built. This is not a critique of the data release. It is an indictment of our payment system.

Data is revealing important trends in how we pay doctors differently. Not all physician payments are created equal, and the data certainly shows the disparities across specialties, primary care, and others. For example, the average total annual Medicare payment to geriatricians was less than $100,000, while dermatologists and radiation oncologists (who presumably also see non-elderly patients) received on average $200,000 and $360,000 respectively. The important question will be why and should it continue?

Figure 1: Distribution of Total Medicare Pay by Provider Type, 2012


Source: Author’s calculations based on Medicare data released in April 2014

Data is revealing important indicators of cost and pricing – a major contributor to rising health care costs. Why is it that a brief visit with a geriatrician is worth $13; a 45-minute visit with a geriatrician sorting through medications, educating family members, and developing a quality of life plan with a terminal cancer patient is worth $79; and a dermatologist treating suspected skin cancer can earn upwards of $600 for a procedure that takes them minutes?

Data sheds light on practice patterns. The data is also revealing important variances in utilization of drugs and treatments. For example, a block apart on Park Avenue, two ophalmologists differ significantly in their use of treatments for macular degeneration. One uses expensive injectable drugs and gets paid over $10,000 per injection, while the other receives less than $500 for the lower-cost equivalent.

CBS News report looked at spinal fusion surgeries—a procedure where there is almost no evidence demonstrating a net benefit to patients compared to other conservative therapies. They observed that “while the average spine surgeon performed them on 7 percent of patients they saw, some did so on 35 percent.”

At the extremes, outlier “practice pattern” begins to raise questions of potential improper billing or outright fraud and abuse. For example, simply looking at the frequency and volume of services provided to individual beneficiaries can identify concerning outliers. This laboratory company billed for 28,954 blood glucose reagent strips in 2012- for 88 patients. And yes, that’s highly unusual.

Figure 2: “Outlier” Medicare Billing for Blood Glucose Reagent Strips, 2012

Source: Author’s calculations based on Medicare data released in April 2014

One clinical social worker billed for 1,697 separate days of service on 28 patients (the size of the bubble is proportional to the total amount of reimbursement by Medicare in 2012).

Figure 3: “Outlier” Medicare Billing for Days of Service, 2012

Source: Author’s calculations based on Medicare data released in April 2014

The most extreme outlier, Dr. Gary Ordog, was named by NPR and ProPublica in their examination of providers who are outliers on their pattern of coding for the highest intensity office. Ordog had previously lost the right to bill California’s state Medicaid program, and yet continued to charge Medicare for over $500,000 in billing in 2012. It’s important to caution however, that even in these extreme outliers, statistics alone cannot provide definitive evidence of abuse. There is a need for formal investigation.

Medicare and law enforcement officials will need to create new processes for dealing with a potential flood of outlier reports from amateur sleuths like me.

What’s Next for Medicare Data?

Data can be trended. Updates of data releases can begin to show us not just snapshots, but moving pictures of our healthcare system as it undergoes rapid changes. The New York Times reported on the increase in charges for certain frequent causes of hospitalization between 2011 and 2012. It will be interesting to see whether the data release itself, and the Steven Brill landmark Time article on hospital charges, have an impact on reversing these trends.  

Data can be “mashed up”.  The value of open data is hugely greater than the sum of its parts. As more and more data becomes available, the files can be cross-linked and “mashed up” to be able to answer questions no one database could have.  ProPublica linked together cobbled together data on state actions and sanctions on physicians with the Medicare data release to ask why these physicians are still being paid by Medicare.

What does the future hold? Correlations with drug prescribing data, meaningful use, and referral patterns are possible today, Sunshine Act disclosures and quality reporting, and much more is soon to come.

As we get comfortable with the data, analysts can move past the basics of arithmetic and sorting, we have an opportunity to make more ‘meaningful use’ of this data. We can begin to identify practice patterns, overuse, variations in geography or demographics, and potentially even fraud and abuse. As more and more data becomes available, the files can be cross-linked and “mashed up” to be able to answer questions no one database could have addressed. What will determine the value of the Medicare data release will be the creativity of those data scientists, epidemiologists, and health services researchers (amateur as well as professional) who can ask the challenging questions that must be answered.

17 replies »

  1. It is perfect tim to make some plans for the future and it’s time to be happy.
    I have read this post and if I could I desire to
    ssuggest you few interesting things orr advice.

    Perhaps you could write next articles referring to this article.
    I desire to read more things about it!

    Also visit myy web-site cartoon Animated Video Booba

  2. @farzad

    Many of us are tired of your anecdotes.

    The data is meaningfully useless except for outliers.

    Perhaps the ortho who did so many operated on cases that were not medically necessay at the time?

  3. One cannot leave out the caveat that the value of the data is often dependent upon a correct understanding of the CPT coding system. There are often more than one, sometimes many more, codes that can be used depending on nuances, provider choice, or practice setting. Also, searching for the specialist who you think may be the best may not be correct. Spinal injections are performed by ortho, neurosurgery, interventional radiologists, sometimes PMR, pain management docs, etc.
    It is interesting the phrasing: “We were looking….” Most people don’t know which specialist to look for. I can see the scenario where you use the data to choose a doc based on what you think the diagnosis is, get a recommendation for a procedure that that doctor likes to do, and then you again use the data to pick the best doc. Trouble is, the diagnosis you thought was it was not it, and the doc, although good at what he does, doesn’t do the best procedure for what you actually have and so doesn’t recommend it, and then you go to the best doc that does the procedure except it isn’t what you need.
    Not saying the data isn’t useful, but just saying…

  4. Average per patient of 329 test strips per is a little less than once per day. Just sayin’.

  5. See what data reveals…? Important factors of cost and pricing: visit with geriatrician is worth $13; a 45-minute visit with a geriatrician sorting meds, educating family members, and developing a life plan with a terminal cancer patient is worth $79; while a dermatologist treating suspected skin cancer can earn upwards of $600 for a procedure that takes them minutes –

    Talk about transparency.

  6. I’m all for transparency and the trend towards open and accountable in our midst. Lets be mindful of the perfect storm that is emerging under the weight of exposing the very tender underbelly of our seeming imploding house of cards healthcare borg.

    As we know, the best intentioned enabling law or enabling policy has a way of generating burdensome and unintended consequences.

    The regulatory reform and enabling pathways (including blended comp plans to bridge the zeitgeist shift) to get us from volume to value need appreciate the continuing (from drip, drip, drip to tsunami scale) exits from traditional bill for collections FFS practices increasingly drawn to the world of ‘direct practice’ & all it’s derivative forms from membership medicine, to retainer to concierge.

    The perfect storm metaphor is we are stoking the fires of physician exodus (at least those specialties most suitable for the new model ie, PCPs) at the precise time we’re exponentially expanding the demand for their services via the ACA et sequelae.

    Just saying….

  7. I’m sure there are some patients who may be “under-tested” with regard to glucose monitoring. I am not seeing many of them in the ED. What I am seeing is people who are simply non-compliant with their diet and medications.

    Maybe we need to be asking “why do all these patients need to be testing 4-5x per day”??

    Most of the patients on Medicare are NOT even insulin dependent, and the ones who are – they are not on a sliding scale, so what is the need to test multiple times per day? All this does is cost the taxpayer $$, enrich the companies who send Medicare enrollees hundreds of strips per month, and generate fairly useless (and unreliable) data which the doctor usually never even sees.

    I have had patients come to the ER because their “sugar was high” – amazingly, this was 45 mins AFTER eating a huge meal of pancakes, syrup, juice, toast, and coffee. Wow, what a surprise. Did your doctor want you to see how high you could drive your post-prandial blood glucose? What exactly are we supposed to do with you? Admit you to the hospital?

  8. “The leading surgeon had done 263- as many as the next 10 combined. We figured he might be the best person to go to, and we were right- the procedure went like clockwork.”

    Did you ask him how many he’d F’d up before you made the appointment?
    At what number did he become proficient – 43, 68? If everyone avoids the lower numbers how will they develop a skill set?

  9. I’ve worked with hospital data as a consultant and a member of a leadership team. I appreciate the value here, it is real and it can indeed help to solve many of the problems we face – but am concerned that we’re setting up a system that encourages people to try to game the system. Data is not by definition objective. It can be manipulated and it can be easily falsified – we only have to look at the VA scandal – which was in fact about this very problem – to get a sense of how grave the potential for problems is. This is a reality we’re only now begrudgingly acknowledging in the academic world. Go take a look at a site like retraction watch and look at the number of researchers who have fudged, faked their way and otherwise manipulated the system. The revealing thing? Most of them seem to feel they have to if they want to keep up with the competition. If we can’t expect scientists and researchers at some of the top institutions in the world to play by the rules, how reasonable is it to expect business people and doctors trying to run profitable practices to not fall into the same trap?

  10. While you are analyzing all that data, don’t forget that absolute values need to be standardized in some way. On the chart about glucose test strips, the number of test strips averages out to 329 test strips per patient. I test 3-4 times per day. I would conclude from the graphic that most providers aren’t having their diabetic patients test nearly often enough.

  11. We have fallen into the trap of literally worshiping data – and a mechanistic male model of health

  12. we are only one great app and big data dump away for perfect health in this country.