Unbiased Analysis of Today's Healthcare Issues

Part B drug reimbursement

Written By: Jason Shafrin - Apr• 18•16

In the past Medicare has reimbursed physicians that administer Part B drugs–typically injectable medications administered in a physician’s office–at 6% of the drug’s cost.  The 6% aims to cover the cost of purchasing and storing the drug as well as administering it.  Because physician reimbursement is proportional to the cost of the drug, physicians have an incentive to select more expensive treatments.

A recent CMS proposed rule aims to change physician reimbursement for administering Part B drugs.  CMS states:

Medicare Part B generally pays physicians and hospital outpatient departments the average sales price of a drug, plus a 6 percent add-on. The proposed model would test whether changing the add-on payment to 2.5 percent plus a flat fee payment of $16.80 per drug per day changes prescribing incentives and leads to improved quality and value. The proposed change to the add-on payment is budget neutral.

Oncologists are not happy with the change. MedPageToday reports the Bruce Gould, MD, president of the Community Oncology Alliance wrote in a letter to CMS that:

The proposed payment model “is an inappropriate, dangerous, and perverse mandatory experiment on the cancer care of seniors who are covered by Medicare.”

The question is, what is the right number?  Physician costs to administer drugs have both a fixed and variable component.  The fixed component is likely the time and labor cost to administer a drug likely does not depend on drug price.  The cost of capital tied up in purchasing a drug and the risk that the drug is not used, however, are certainly proportional to price.  Having a fixed and variable component to pricing does make sense.  However, a $16.80 price to administer a drug seems very low.  The reasonableness of the 2.5%.

MedPAC analyzed the benefits and cost of moving entirely to a flat fee system for physician reimbursement of administering Part B drugs and found:

It might increase the likelihood that a provider would choose the least expensive drug in situations where differently priced therapeutic alternatives exist, potentially generating savings for Medicare and its beneficiaries. At the same time, a flat-fee add-on might create other incentives that could increase spending. For example, questions have been raised about whether increased payment rates for very inexpensive drugs might create incentives among some providers to overuse these drugs or spur manufacturers of low-priced drugs to raise their prices.

The “appropriate” reimbursement rate may even vary across drugs.  Further research is needed to better understand the implications of changing physician reimbursement for administering Part B drugs to ensure that both cost savings and patient access to high quality medications are achieved to the greatest degree possible.

Is P4P doomed to fail?

Written By: Jason Shafrin - Apr• 17•16

There have been many pay-for-performance (P4P) programs that have been implemented to attempt to improve quality and reduce cost. The vast majority of these programs have not been able to demonstrate large or even any improvement in quality or cost. Some researchers claim that these programs have not worked due to the size of the bonus, the specific metrics measured, a learning curve in P4P design and a host of other factors.

An interesting article in Health EconomicsSherry (2016)–claims that the the lack of success may be due to a fundamental flaw in P4P design.

…the intuition that P4P increases the output of a rewarded service is guaranteed only in the most simple (and unrealistic) cases of P4P plans where a single quality metric is rewarded. When more than one service is rewarded under P4P, the change in the level of each rewarded service is ambiguous due to multitasking between rewarded services. Multiple rewarded services are more likely to increase when unrewarded services earn lower marginal revenue. The change in unrewarded services is also generally ambiguous, even in settings of joint production. A given unrewarded service is more likely to decrease due to multitasking if other competing unrewarded services are more profitable; conversely, it is less likely to decrease if it is jointly produced with other rewarded services.

To give a simple example, consider the case where a patient should receive 3 types of preventive services but only service #1 and #2 are rewarded. In this case, it is unclear if they would increase the quantity of preventive services of #1 and #2. If service #1 and #2 are rewarded equally but service #2 takes less time to accomplish, physicians may increase provision of preventive service #2, but decrease provision of preventive service #1. Further, if physician time is limited, they may be less likely to provide preventive service #3 since since this services does not receive any reimbursement. On the other hand, if preventive service #2 and #3 typically occur together, then physicians could also perform more preventive service #3.

If this seems confusing, most P4P programs have tens or hundreds of measures and due to the joint production and multitasking interactions, the net effect on quality will be unclear.


Based on some comparative statics the authors conduct on a simple 3 service model, the authors make 3 recommendations for improving P4P programs.

  1. Rewarding fewer quality measures may be preferred.  “Rewarding a smaller set of quality measures limits the scope of P4P programs, but may stimulate greater performance improvement if multitasking between rewarded services is mitigated.”  Further, rewarding fewer services decreases the administrative complexity and cost of the program.
  2. Second, attention should be paid to unrewarded services as well.  The point has been made previously in the multitasking literature, most notably by Holmstrom and Milgrom 1991 (summary post here).
  3. The interactions between the production of different types of health services should be considered when designing a P4P program.  “When rewarding a single quality measure, selecting a measure that is jointly produced with other desirable health services and quality measures can contribute to broader quality improvements beyond the rewarded measure. Similarly, avoiding measures that induce higher amounts of multitasking can help avoid disruptions in care.”


Mid-week Links

Written By: Jason Shafrin - Apr• 12•16

The future of mental health?

Written By: Jason Shafrin - Apr• 11•16

Nature has an interesting article on how mobile technology is being used to treat mental illness.

Estimates suggest that about 29% of people will experience a mental disorder in their lifetime1. Data from the World Health Organization (WHO) show that many of those people — up to 55% in developed countries and 85% in developing ones — are not getting the treatment they need. Mobile health apps could help to fill the gap (see ‘Mobilizing mental health’). Given the ubiquity of smartphones, apps might serve as a digital lifeline — particularly in rural and low-income regions — putting a portable therapist in every pocket. “We can now reach people that up until recently were completely unreachable to us,” says Dror Ben-Zeev, who directs the mHealth for Mental Health Program at the Dartmouth Psychiatric Research Center in Lebanon, New Hampshire.

How do these apps work?

Several times a day, the app prompts users to answer questions such as “How well did you sleep last night?” or “How has your mood been today?” If users report that they slept poorly, or have been feeling anxious, the app will suggest strategies for tackling that problem, such as limiting caffeine intake or doing some deep-breathing exercises.

Other apps help patients connect and/or communicate with health care professionals, or even alert providers of a potential worrying change in physical activity, mood or behavior. Unfortunately, the vast majority of these apps have not been tested through rigorous studies.  There, is however, some progress.

Between 2013 and 2015, the number of mobile-health trials registered on ClinicalTrials.gov more than doubled, from135 to 300. And the number of trials specifically focused on mental and behavioural health increased by 32%.

Exciting new technology is on the horizon.  Now, we just need to prove that it works.


Meaningful Use is dead! Long live Meaningful Use!

Written By: Jason Shafrin - Apr• 10•16

In a nutshell those are the comments made in January by Acting Administrator Andy Slavitt at the J.P. Morgan Annual Health Care Conference.

The Meaningful Use program as it has existed, will now be effectively over…

You can now hear physicians groups burdened by the inflexible EMR mandate rejoicing.  But then Slavitt continues.

and replaced with something better…We will be putting out the details on this next stage over the next few months…


So what is the new meaningful use going to look like? Slavitt mentions four points:

  • Not rewarding physicians for simply using EMRs
  • “Providers will be able to customize their goals so tech companies can build around the individual practice needs, not the needs of the government.”  I’m not sure what this means since if the government wants to regulate, there needs to be some standards.  If physicians can do what they want, then there is really no point of implementing a Meaningful Use replacement.
  • Requiring open APIs.  This will allow new entrants into this space, but security could be an issue.
  • Interoperability is key.  “Technology companies that look for ways to practice “data blocking” in opposition to new regulations will find that it won’t be tolerated.” How exactly the government will enforce this rule, however, is unclear.

A report last month noted that the share of physicians participating in meaningful use has decreased.  However, EHR incentives will still be in place for clinicians treating Medicare patients.

While MACRA provides an opportunity to adjust payment incentives associated with EHR incentives in concert with the principles we outlined here, it does not eliminate it, nor will it instantly eliminate all the tensions of the current system. But we will continue to listen and learn and make improvements based on what happens on the front line.

The more things change, the more they stay the same.

HWR is up

Written By: Jason Shafrin - Apr• 07•16

Jaan Sidorov has posted A Presidential Politics-Free Health Wonk Review at The Population Health Blog.  Check it out!

How to improve measures of “value”

Written By: Jason Shafrin - Apr• 06•16

Organizations such as the Institute for Clinical and Economic Review (ICER), Memorial Sloan Kettering and the American Society of Clinical Oncology all have begun producing value frameworks to determine if the value of new drugs are worth the cost.  An interesting Forbes Op-Ed by Anupam Jena and Tomas Philipson in Forbes examines whether these organizations are looking at value in the right way. 

Few would disagree that rewarding the full social value of a drug is a good idea, but the failure of current frameworks to accurately and transparently assess value poses a threat to the future pipeline of new drugs. Put simply, our current measures of value are not only too narrow but are shortsighted. Pricing and coverage decisions made on the basis of incomplete value frameworks run the risk of reducing future innovation by penalizing drugs that actually offer value to society.

The rest of the article presents eight examples demonstrating that the current frameworks are not adequately capturing the value of innovative treatments.  The economics-based discussion of value is certainly an interesting perspective for the recent debate around care value.

Disclosure: Dr. Jean and Dr. Philipson are my colleagues at Precision Health Economics.

Are drug prices higher or lower in the U.S.

Written By: Jason Shafrin - Apr• 05•16

The high price of prescription drugs in the U.S. has received a lot of press in recent years.  However, are drug prices really higher in the U.S. than other countries?  Tomas Philipson makes an interesting point regarding U.S. drug prices:

It is well known that free-market prices of branded drugs still on patent are higher in the US than in other rich countries with price-controlled prices. However, it is also true and often not highlighted that prices for generic drugs tend to be lower in the US than other countries. Prices of generics in the US tend to be lower due to more free-market competition and the more streamlined process at the FDA. Given that about 88 percent of prescriptions in the US in 2014 were for generics, the value of a more competitive generics market in the US deserves more praise.

This comment was made in response to a recent FDA policy change to allow easier market access for generic drugs.  The value of low priced generic drugs in the U.S. should not be underestimated.


Value Framework Guidelines

Written By: Jason Shafrin - Apr• 04•16

In recent years, a number of organizations have developed value frameworks to assess new treatments.  These organizations include the Institute for Clinical and Economic Review (ICER), Memorial Sloan Kettering Memorial Cancer Center (Drug Abacus), American Society for Clinical Oncologists (ASCO),  National Comprehensive Cancer Network (NCCN), European Society of Medical Oncology (Magnitude of Clinical Benefit Scale).  What are the best practices to conduct these value analyses?  

The National Pharmaceutical Council has put together 28 recommendations for how best to conduct these value assessments.  The recommendations are divided into six categories: assessment process, methodology, benefits, costs, evidence, and dissemination and utilization.

The infographic below summarizes their recommendations.

How Should Value in Health Care be Assesed
Source: National Pharmaceutical Council

The Intuition behind Bayes Theorem

Written By: Jason Shafrin - Apr• 03•16

Bayes Theorem is well-known in law of probability. Mathematically, you could write it as:
P(A|B)=P(A and B)/Pr(B) = P(B|A)*P(A)/P(B).

An interesting interview in Scientific American with Decision theorist Eliezer Yudkowsky explains Bayes Theorem more intuitively.

I might answer that Bayes’s Theorem is a kind of Second Law of Thermodynamics for cognition. If you obtain a well-calibrated posterior belief that some proposition is 99% probable, whether that proposition is milk being available at the supermarket or global warming being anthropogenic, then you must have processed some combination of sufficiently good priors and sufficiently strong evidence. That’s not a normative demand, it’s a law. In the same way that a car can’t run without dissipating entropy, you simply don’t get an accurate map of the world without a process that has Bayesian structure buried somewhere inside it, even if the process doesn’t explicitly represent probabilities or likelihood ratios. You had strong-enough evidence and a good-enough prior or you wouldn’t have gotten there.

On a personal level, I think the main inspiration Bayes has to offer us is just the fact that there are rules, that there are iron laws that govern whether a mode of thinking works to map reality. Mormons are told that they’ll know the truth of the Book of Mormon through feeling a burning sensation in their hearts. Let’s conservatively set the prior probability of the Book of Mormon at one to a billion (against). We then ask about the likelihood that, assuming the Book of Mormon is false, someone would feel a burning sensation in their heart after being told to expect one. If you understand Bayes’s Rule you can see at once that the improbability of the evidence is not commensurate with the improbability of the hypothesis it’s trying to lift.

To give some other examples. Let’s say that you are a kid and you are waiting for your mom to pick you up. Let’s say that your mom has picked you up from school every day for the past 2 years between 0 and 5 minutes after the school day has ended. If your mom is 1 minute late, what is the probability your mom is not going to pick you up? The answer is probably pretty low, given that your prior your expected probability (your prior) of your mom picking you up is so strong.

On the other hand, let’s say your mom is very irresponsible and over the past 2 years will pick you up between 0 and 5 minutes after the school day has ended half the time, and the other half of the time she forgets to show up and you have to walk home. In this case, a 1 minute delay in your mom’s arrival would be much more convincing evidence that your mom will not show up because your expected probability (your prior) was much lower.

Thus, with the same current data available (i.e., your mom is 1 minute late), Bayes Theorem tells you that you should not be worried if your mom has a great track record, but a bit more worried if over the past 2 years your mom has been a bit of a deadbeat.