Friday, October 2, 2015

Alternative payment models and precision medicine

You may recall that the State of the Union in which President Obama called for greater focus on "precision medicine," a term referring the concept of individual medical treatments designed and tailored to a specific patient rather than--as most treatments are today--for general populations. Most commentators seem to have taken the president's call to mean more genetics research. They have in mind a situation where the doctor takes a swab to sequence your DNA, and picks a treatment based on genetic factors. Unfortunately, a lot of what affects health actually isn't genetic, and at any rate DNA sequencing isn't the only approach to individualizing medical treatments.

Another approach is to try to target treatments to the specific condition the patient has to see what works. For example, when a cancer patient doesn't respond to the normal course of treatment, one thing that has been done in a few cases has been to take samples of that cancer and give it to lab mice, thus creating a mouse model of that specific patient's cancer on which they can test a variety of different drugs to see which, if any, the cancer will respond to. In some cases, we get a hit; one of the drugs that is not normally standard for the type of cancer does work in the mouse model, and with a little luck, also works on the patient. This wouldn't work without the animal model--you can't just administer a whole bunch of highly toxic drugs to a patient until something works. Not only would the patient be harmed, but lack of proper randomization and controls means you can't really be sure what if anything worked.

So there's a promising future in using precision medicine to help patients that don't respond to standard treatments, and potentially to help reduce the risks to patients as well. However, creating an animal model is pretty expensive and may not even yield any useful results. Studying the animal model doesn't exactly count as rendering treatment to the patient, and is inherently experimental since it literally involves running an experiment for each specific patient. The question is, who pays? This is beyond the scope of traditional insurance reimbursement. So far all examples of this technique have been funded entirely through research grants and institutional sources. If it does turn out to be as effective as initial results suggested, how will we scale this?

Traditional insurance has no incentive to fund such a procedure. Not only is it experimental, possibly won't even yield a result that would affect the course of treatment, and expensive, but it also could extend the lives of specifically those cancer patients who are most likely to relapse. I hate to be so dismal, but financially, that actually penalizes the insurer that tries it.

It seems to me that precision medicine requires an alternative payment model, one that rewards insurers for outcomes and encourages them to take risks on funding the development of individualized treatments for each of their most challenging patients. I'm not sure ACOs fit the bill. ACO payment models are designed so that providers and insurers share the benefits of any cost savings achieved while caring for a given population--but precision medicine techniques like the one above, while they could reap huge benefits, almost certainly increase costs, and quite substantially so.

With out some fresh ideas on how to reform insurer reimbursement models, I doubt many of the best precision medicine ideas will ever survive outside the research centers that dreamt them up.

Tuesday, September 29, 2015

Bush v Trump

Here's a quick comparison of the Bush and Trump tax proposals:

You can find Bush's plan here, and Trump's plan here. Sidenote: there is a typo, I think, in Bush's plan. He labeled the $11,300 bracket shown here as "standard deduction," but I think what he meant was not a standard deduction but a 0 percent tax bracket. They are very similar concepts but work differently: the standard deduction reduces your taxable income first, then you calculate how much of your income falls in each of the tax brackets using this reduced number second, whereas a 0 percent bracket doesn't deduce anything from the taxable income figure used to calculate liabilities in the other brackets. If you take Bush's tax plan literally he's saying there would be no tax on the first $22,600 of income, but since he didn't mention the tax rate on the $0-$11,300 I think he meant his "standard deduction" to actually be a 0 percent tax bracket, not a deduction.

Here's a graph of the two:

Min Income $0
Max Income $500000
What you can see is that Trump cuts taxes far more aggressively than Bush. So it's no surprise that Trump's plan costs far more than Bush's--$10.8 trillion for Trump's compared to $3.4 trillion for Bush's over 10 years. You can also see there are a handful of wealthy but not extremely rich people that pay slightly more under Bush's plan than current law, because Bush lowers the income threshold for the 28 percent tax bracket.

Sunday, September 27, 2015

Jeb Bush explains why "broad-based" tax cuts are bad policy

In a moment of clarity Jeb Bush explains why we shouldn't cut all marginal income tax rates:
"Of course, tax cuts for everybody is going to generate more for people that are paying a lot more. I mean that's just the way it is."
I've been explaining this fact to anyone who would listen for a long time. I even made this calculator to help visualize it. Any tax cut in any tax bracket necessarily gives more to the rich than to anyone else. This is even true when we cut the lowest marginal tax rate: the rich also pay taxes on income in that bracket, so they will receive the largest reduction of anyone from a reduction in the tax on that bracket.

Ok, obviously Bush was trying to defend rather than criticize his own "broad-based" tax cut plan. But in doing so he highlights how ridiculous it is. If you really wanted an across-the-board tax cut that benefits everyone equally, the only way you can pull that off is by tying any cut with an offsetting increase in the top marginal tax rates, so that the disproportionately large reduction in taxes the rich get on their first $400,000 of their income is made up by an offsetting increase on their income over $400,000.

I mean, that's just the way it is.

Friday, September 25, 2015

Audit finds excellent cybersecurity at

AP posted this story this morning titled "Audit finds slipshod cybersecurity at," criticizing security vulnerabilities in the system used to administer the federal Obamacare exchanges. Every media outlet under the sun re-posted it word-for-word:
And still others tried to re-post it word-for-word but screwed up.
You won't find the source in the AP story or any of the copy cats, but it is about this audit by the HHS Inspector General's office published (in redacted form) on 21 September 2015. (By the way, the report is exactly 4 pages long and half of that was just the title so I don't know why it took AP 3 days to cover it.)

Except, the audit did not find "slipshod" security at

There was something screwy about all this coverage because the AP published a piece that, while not verbatim, was nearly identical to this one back in September 2014. By an accident involving a local news affiliate, Charles Gaba from the esteemed pointed me to the previous story here. That prompted me to chase down the HHS Inspector General's audit, no thanks to AP who failed to properly source their article.1 It turns out that the audit actually had the opposite to say about security there is great.

The report released on 21 September 2015 was the formal writeup of an audit that the HHS Inspector General's office had conducted from August through December in 2014. In other words, this is the same audit as the AP covered in it's previous story in September 2014, and that's why the two AP stories are identical--they are literally talking about the exact same audit. The first AP story was based on a preliminary report produced by the HHS Inspector General's office about the preliminary findings in their security audit of MIDAS, a database system that and insurers use to store users' information to allow them to buy insurance through the interface. The new story is based on the formal write up of the audit (which is short because the actual technical details of the vulnerabilities were redacted to avoid giving hackers any ideas) conducted a year ago.

But here's the thing. The main reason for this new report is in fact to say that all of the security vulnerabilities have been fixed to the satisfaction of the HHS Inspector General's security team. The final line of the report:
"We have since reviewed the supporting documentation and verified CMS's remediation."
In otherwords, the point of the new report is to say that cybersecurity at is now excellent. That's the only news here. But none of the News is covering it that way.

1. See what I did there, AP? It's called a hyperlink. It turns out that when you are writing about a thing on the internet, you can "link" to that thing and then users can click it and be redirected to that thing. You do this by typing what's called an "anchor tag" into the HTML code, which looks like this <a href="[insert url here]">[insert display text here]</a>.You should try it sometime.

Wednesday, September 23, 2015

Why generics are being monopolized

You may have heard of Martin Shkreli, the CEO who tried to jack the price of a medication to treat AIDS patients up to 50 times what it is currently. Shkreli's comically evil plot aside, the part that makes this most surprising is that the drug, Daraprim, is a 62 year old generic that in principle, any company could compete with him to produce. As Dan Diamond reports, this is an increasingly common phenomenon where a firm buys up all of the makers of an old generic drug, and then jacks up the price to make enormous profits. Why doesn't someone enter the market and undercut the price?

Both Diamond and Kliff allude to a basic economic theory. Theory predicts that when we're on the increasing-returns to scale portion of the production function, the market tends towards monopoly (you may have heard of increasing-returns referred to as "economies of scale"). That's because at increasing returns, which ever firm is producing more faces a lower marginal cost of production, meaning they can undercut price and drive a competitor out.

Generally speaking, production functions all have the same basic shape. There are increasing returns to scale at low levels of output--where increasing production would lower marginal cost--and decreasing returns to scale at high levels of output--at some point, producing more will actually drive up marginal costs rather than lower them. If you have a bunch of factories operating at the increasing-returns portion of the production function, then you could cut costs by closing some of the factories and increasing production at the remaining factories. If those factories are owned by different firms, then you'd expect an increasing-returns industry start consolidating into fewer firms. Conversely, at the high end of output, if each of these factories is operating at the decreasing returns portion of production, then presumably the firm could cut costs by building more identical factories and reducing output at each of the existing factories. In other words, we'd expect fragmentation and new entrants into decreasing returns industries. For the most part, equilibrium should adjust until we're right at the linear-returns sweet spot in every factory.

But there is one exception. What if there's not enough demand? It is possible that the number of goods you'd have to produce to reach the linear-returns bliss-point exceeds the quantity that consumers will buy at that price. Such an industry will consolidate until there's only one firm or factory producing the good, and even then will still be on the increasing-returns portion of the production function. And once it has a monopoly, that firm can charge the profit-maximizing, monopoly price, because the threat of droping the price all the way down to marginal cost will keep out any potential entrants (since the monopolist would be producing more than the entrant and thus have lower marginal costs, the entrant cannot match this threat price).

It turns out that highly-effective generic drugs for rare conditions are uniquely profitable in this way. Daraprim, for example, only has about 10,000 users. Moreover, the number of users doesn't change much at all with price, meaning not only is the profit-maximizing price sky-high, way above costs, but that demand will never exceed the increasing-returns portion of production so there's zero potential for an entrant to compete on price even though the current price is way above marginal cost.

I got the impression in undergrad that most people think of economies of scale as a good thing. We like the feeling of saving money by buying more! But in fact it has some very very bad economic properties, as Martin Shkreli has illustrated perfectly.

What is the solution? There are a couple proposals floating around. The government could force these producers to sell at cost--there's a legal argument to be made that the federal government has this power now under the Bayh-Dole Act (Bayh and Dole do not agree). These are generic drugs, so there's nothing stopping the government (except maybe congress) from simply producing the drugs themselves, or perhaps contracting with a firm to produce them at cost. Basically, all of the proposals I've heard involve strong government action. Anyone have other ideas?

Side note: the theory is basically the same whether it's about marginal cost or average cost. Even if we are on the rising portion of the marginal cost curve, if average cost is falling (eg high fixed costs), market tends towards concentration in the same way. The largest firm can undercut the others until the competitors are unable to cover fixed costs and are forced to shutdown. Only major difference is that average cost operates on the extensive rather than intensive margin (see generally: firm shutdown conditions).