If I were King of Medicare.
I am still learning how best to extract useful information from the public-use Medicare Part-D Drug Utilization and Cost files. I view these as experiments of nature worth mining for what they can tell us about the clinical and business aspects of healthcare. The last few articles I have written focused on the utilization and cost of Insulin, highlighting the seemingly unjustified increases in this life-saving drug for diabetics. In this tweak, I learn how to combine data from drugs in the same therapeutic category including all the brand and generic versions of individual drugs within categories. I am still wrestling with technical issues related to presenting multiple groups in a single TreeMap visualization, but as an example, I will show that in 2015 three groups of drugs– insulin, opioids, and drugs used to treat Hepatitis-C– cost the Medicare program 17.6% of the total cost of all Part-D drugs but comprised only 7.2% of all prescriptions. These drugs were expensive for different reasons that I will illustrate.
I particularly like the TreeMap data-visualization format because it allows hundreds if not thousands of drugs to be compared at the same time and facilitates identification of unexpected or unjustified outliers much easier. Since Medicare is now standardizing the formats of these drug files, changes in utilization and cost can be tracked over several years. I believe that placing such analyses into the service of policy makers and payers can allow savings of billions of dollars without compromise of care.
Groups of drugs examined.
For this article, I use the 2015-data from the revised set of public use files more fully described in the recent insulin article. I group all the insulins together that were available for at least part of 2015. (Note that this group does not include any of the oral hypoglycemics.) In this database, Medicare flags what it categorizes as opioids and I have grouped them using that definition. Unlike the “hard” opioid group I studied in earlier analyses (“KHPI opioids”), the Medicare-flagged opioids include drugs used for medication assisted therapy (ex. buprenorphine and methadone) and the commonly prescribed tramadol HCl (ex. Ultram).
I tried to identify from clinical references all the drugs used to treat Hepatitis-C. The treatment of this relatively common but serious disease has been revolutionized by the discovery of truly antiviral drugs and combinations like Harvoni and Solvadi. Before these lifesaving but expensive drugs, various interferons were used. These latter were problematic, expensive, and not very effective. They are sill being prescribed for Medicare patients but for diagnoses that cannot be determined from the database. I include these older drugs here for comparison. (I am open to suggestions about drugs to be added or removed from the Hepatitis group.) Details about the make-up of all three groups is presented in the interactive tables and visualizations supporting the current analysis.
As noted above and in the accompanying table [Not yet added.], the three groups of drugs make up nearly 18% of all Part-D drug costs, including out-of-pocket costs to patients. This amount is out of proportion to the actual number of claims (prescriptions and refills) written. Note that Medicare was unable to take into consideration any rebates (a.k.a. bribes) that the manufacturer may have given to pharmacy benefit managers or others in the food chain. The fact that even Medicare does not know how much its drugs cost is a measure of the structural insanity in the drug pipeline between factory and patient.
The relationships between total cost and claims varies by the nature of the groups and individual outlier drugs within them. For example, the total cost of insulin comprised an amazing 8.1% of all Part-D costs but only 1.60% of all claims. For the opioids, these figures were reversed at 3.1% of costs but 5.54% of claims. [More than one in 20 Medicare prescriptions were for opioids!] The numbers for the Hepatitis-C drugs show the impact that very expensive drugs can have, encompassing 6.4% of costs, but only 0.021% of claims. One can hardly find the Hepatitis dot in the Rx-frequency TreeMap! [Find it here.] Another way to highlight the impact of cost and frequency of prescriptions is to calculate the average cost per claim from the available data. For Hepatitis-C drugs, insulin, and opioids these are $24,071, $478, and $53 per prescription or refill respectively.
Do the big dogs rule?
It is easy to assume that expensive drugs like the new biologic or cancer drugs drive the overall cost of any prescription drug insurance program. That was my initial premise. This is a hypothesis worth exploring, but a back-of-the-envelope calculation is illuminating. For example, in the current database, there are 98 drugs that cost more than $10,000 per dose. [View the list and TreeMap here.] These 3% of all the individual drugs make comprise 7.4% of the total program cost. Recall that insulin alone draws down 8.1% of total Medicare Part-D costs and none of them are on the >$10,000 list! In medicine, it is the common things done frequently that impact cost the most. Simple office visits with doctors blow away competing billing codes for outpatient costs. This is not to say that uninformed, uncaring, or unscrupulous individual providers are not taking advantage of the profits to be made from prescribing these outpatient drugs. Not doing something about this is a failing of self-governance by my profession.
Among these heavy-hitters are all 6 of the new and effective anti- Hepatitis-C virus drugs, and 3 versions of fentanyl including the illegally promoted Subsys. [ I would look into prescribing of the others too. I can get fentanyl on the street cheaper than this.]
Pointing out the outliers and sore thumbs.
The TreeMaps allow Individual drugs or their branded products to stand out from among the 3205 different brands and generics prescribed in 2015. For example, one Hepatitis drug, Harvoni, is the single most expensive drug for the Medicare Part-D program with an aggregate cost some two and a half times more than Crestor– the next most expensive drug. Harvoni alone made up 5.1% of all Medicare Part-D cost in 2015! Since then a second expensive product, Sovaldi [available for only part of 2015] will surely alter the usage profile and cost of Hepatitis-C drugs, but I am not sure how. Medicaid drug utilization and cost data is available for 2017 and should predict clinical shifts in the use of Hepatitis-C drugs.
Lantus brands– the most commonly prescribed insulin– made up 3.2% of total Medicare Part-D costs and 0.61% of total claims. At the other end of the cost/claims ratios is the combination opioid, hydrocodone with acetaminophen, which at an average of $25 per claim comprises fully 2.1% of all claims but only 0.53% of total cost. This drug is cheap in the retail market, but pulls its weight with volume! The total cost to our society is incalculable!
How should we use this information?
I believe that looking at groups of drugs in this way can help insurers, health policy makers, and the public decide how best to spend our limited financial resource of taxes, premiums, and pockets. I have already argued that if it were medically defensible to use insulins that are less expensive than Lantus, that this Medicare outpatient program alone could save as much as $7.7 billion that could be used for other purposes. I freely admit that there may be some data-driven or at least rational clinical reasons to use one type or brand of insulin over another. Who will justify the rising high costs of insulin for us? I cannot.
Although expensive at present, there is no question that Harvoni works immeasurably better than the older drugs available. Is there any reason not to consider Harvoni or its fellow-travelers as the standard of care? Is it a quality issue not to, as for example in prisons? Are these new drugs being priced fairly, or are we abusing the vulnerable.
Certainly I am not alone with the opinion that healthcare professionals are prescribing opioids with unjustifiable frequency, but there is no possible reason anyone should pay more for Lortab or Vicodin than for generic hydrocodone at $233, $68, and $25 per claim respectively. Use the savings to help pay for treatment of addiction (or prosecution of irresponsible manufacturers or prescribers.)
Possible next groupings.
Nexium, the brand name, purple pill for heartburn, was the second most expensive drug for Medicare Part-D in 2015 with a total cost of more than $2 billion. As a physician, I am pretty confident that unless you like the color or shape of the pill better, that generic omeprazole [whose brand name Prilosec ruled the roost before Nexium and is now available over-the-counter] is just as good as Nexium for 99% of all patients needing proton-pump inhibitors for heartburn or ulcers. Compared to Nexium at $368 per claim in 2015, lots of omeprazole was being prescribed at $13 per claim. If I were King, I would decree that Medicare would pay no more for Nexium than for generic omeprazole. Doing so for Medicare Part D in 2015 would have saved us collectively almost $2 billion! I suspect that when I group all the heartburn pills together, that the amount we pay as a nation for drugs in this category with little or no advantage will be staggering. (You can crown me later!)
Although I intended the focus of this article to be about the value of grouping drugs by therapeutic category, I cannot help myself. Inspection of the frequency TreeMap reveals that just below the combined total costs of the three index groups considered here, Crestor [a “statin” drug to decrease cholesterol] came in as a single drug to be number four on the cost hit-parade. We paid $2.9 billion for this brand name drug in 2015 at $331 per claim. Although there was no generic equivalent for Crestor at that time, there was certainly no lack of other effective statins. What is the matter with atorvastin at $18 per claim, or pravastatin at $29, or lovastatin at $11? These drugs were prescribed much more frequently than Crestor in 2015. How bad can they be? Perhaps their manufacturers were no longer providing free lunches anymore! I will explore the cholesterol drugs later, including usage profiles over the three-year period for which I have comparable data. For example, what happens when a generic becomes available, or one drug raises its price? How long does it take before we get wise as consumers or professionals? As a matter of public health, we health-professionals are flogging the nation to lower our collective cholesterol. We therefore have an obligation to make the effort cost-effective. If I am off-base here, let me know how far and why.
You play with the data.
Enough for now. You can manipulate the data visualizations and tables yourself on my Public Tableau website. They can be expanded to full screen within the site to display more labels, and searched or sorted at your whim. Let me know where and how you would save our resources and improve our health– preferably both at the same time. Where do you think we are being abused by the system? Where I should look next?
Peter Hasselbacher, MD
Emeritus Professor of Medicine, UofL
March 21, 2018
Let me know if you have problems with the data visualizations. I am trying to find ways to embed the visualizations directly into the blog. Let me know if things like download times or problems with smaller devices are a problem. How does it work when I put the Viz in the comment section below?