« TED Talks: Search, Translate, Subtitle | Main | Should we put a carbon tax on China? »
Medical care and comparative effectiveness
The idea is to have a commission examine which procedures should not receive full Medicare reimbursement. I favor spending cuts for Medicare so for me it's a go. Megan McArdle considers some basic issues. I'll add or second a few points:
1. When it comes to health care, it's very hard to tell what works. That's one reason why we don't pay doctors for results in the first place but rather we pay them for procedures. Having a commission look at statistics only partially remedies this problem. Sometimes looking at outcome statistics from the broader population pool makes the estimate of treatment efficacy clearer and other times it makes the estimate of treatment efficacy fuzzier (you have more data points, but not everyone responds to a treatment in the same way).
2. Where will the burden of proof be put? Will the common procedures be the only ones to receive the funding axe (they're expected to prove themselves in the statistical court, so if they can't the funds dry up?) Will "small numbers" medicine receive the benefit of the doubt or be required to prove itself? The answer to this question will make a big difference.
3. Let's say a treatment for 1000 people helps only 20 of them and so the aggregate statistics for that treatment are not so impressive. If you take those same results and define the population pool ex post as the 20 people who respond positively, suddenly the same treatment has a success rate of one hundred percent. Again, framing will matter a great deal for the results.
4. This commission, if it sticks to its statistical mandate, will be able to recommend many more possible cuts than any vote-maximizing administration will be likely to make. Some other principle will be used to determine cuts. Many defenders of the Obama administration are overestimating how scientific this process will be.
5. What does the public choice equilibrium look like? Should Medicare "strand" some chronic ailments, with large numbers of people suffering only moderately, or should the occasional person be allowed to "die in the street"? Any spending cuts policy will generate news stories of one kind or another; which will have greater political resonance?
6. The fairly arbitrary cuts we get will in some ways resemble means-testing. The discretionary procedures are mostly enjoyed by higher-income and higher-education groups.
7. Imagine an analogy from broader life. Imagine a government that would cut (some) subsidies for any input which could not be shown, statisically, to causally produce better outcomes in life. You can see how open-ended this would be. What if you applied this same metric for your personal spending? Would there be much left to spend your money on?
Addendum: Read the highly intelligent Arnold Kling.
Posted by Tyler Cowen on May 15, 2009 at 07:50 AM in Web/Tech | Permalink
Comments
One problem with state intervention in medical care is that people are really troubled by the idea of the government deciding who lives and who dies. Of course the alternative is a large corporation deciding who lives and who dies but that's somehow more palatable because a person can choose a corporation, so if a corporation doesn't provide a treatment that a person needs not to die we can tell ourselves that it was the person's lack of foresight in determining which disease they would get that's responsible (this is sort of a joke but not completely a joke).
With regard to 4. I think it would be possible to create an ordered list of prospective cuts based on a combination of absolute medical effectiveness, cost effectiveness, or a combination of the two (or more realistically different packages of procedures targeted for different levels of spending). This wouldn't completely eliminate the problem but would substantially mitigate it.
Posted by: Michael Foody at May 15, 2009 8:18:40 AM
The biggest problem with this is the cost of the research. The most expensive part of bringing a new drug to market is the phase 3 clinical trials ( http://en.wikipedia.org/wiki/Clinical_trial#Phase_III )- these determine the effectiveness of the drug. They require large populations over a long period time. Alex Tabarrok could probably elaborate on this.
It is not cheap or easy to statistically determine the "best" treatment. This also ignores individual differences. I would bet this will cost more than it saves.
Finally, if you believe the hype, medicine is moving to "personalized medicine" with drugs tailored to your genotype. Where does comparative effectiveness fit in with a sample population of 1?
Posted by: Marc at May 15, 2009 8:33:04 AM
re. 7. on personal spending, do you mean "better outcomes in life" to be higher utility or do you mean something like longer life or higher income? I think that for medical procedures, using longer (quality adjusted for pain etc.) life is a good metric because improved quality and length of life is pretty much the only reason for medical care and thus is a decent proxy for greater utility. For personal expenditures, utility is the clear metric to use, and that does seem to be a decent thing to consider when making personal expenditures. That said, the absolute drop off in marginal utility must be quite high, or members of our society would be a lot happier than 50 years ago. Because we live much longer, though, flat point in time happiness measures suggests that the hard-to-measure total utility of life has increased for people, because we get more years of utility each.
Posted by: liberalarts at May 15, 2009 8:36:32 AM
While it is expensive to meaningfully measure the effectiveness of any given treatment, is it really the long tail that breaks the health care budget?
Posted by: Cyrus at May 15, 2009 9:02:13 AM
A few years ago I had a middle-ear problem that, if left untreated, could lead to meningitis and potentially to death. This condition has been correctable for some time, but traditional surgical intervention left the patient completely deaf in the affected ear basically, it amounted to cleaning everything out of the middle ear.
Now, however, microsurgery allows reconstruction of the middle ear and eardrum to restore close to normal hearing.
So the question is whether the microsurgery makes the cut. Either technique prevents the medical complications, and one could make the case that deafness in one ear is just a lifestyle adjustment (don't laugh - I've seen bloggers make just that case for delaying hip replacements), so the government plan shouldn't pay for the more-expensive surgery.
Posted by: Don K at May 15, 2009 9:04:43 AM
I hate to be the cranky Misesian, but HELLO Socialist Calculation Problem. It's not just that the statistics about which procedures "help" are often fuzzy. Even if you know this you also have to know how much they help. If 20 people out of 1000 are cured of cancer that may be a damned big result. But if 20 people out of 1000 only get an extra year of life that may not be so big. And we need to know if all of these results are different for men v. women, blacks v. whites, old v. young, etc. etc. etc. AND even if we know all that, we have to know how to value these health care gains. Um how much is a healthy year worth for a 41 year old white male? Bottom line: no government commission can know this stuff. It can't be done. It. Can't. Be. Done.
Why don't they just be honest and say, "We've socialized medicine and we're going to decide what we want to pay for an what we don't want to pay for. Deal with it," instead of all this psuedo-scientific cost-benefit nonsense.
That was a rhetorical question. I know why they can't be honest. The looters (to invoke Rand) need our permission to rob us so they first need to convince us that they're doing this all for our benefit.
Posted by: Bob Lawson at May 15, 2009 9:06:43 AM
Tyler,
Re point #3, framing effectiveness in a post-hoc way is completely dishonest. It's not so much a different framing as it is simply incorrect. Any study that attempted to do so would be attacked, and rightly so.
Also, I suggest not believing the hype about personalized medicine. If such a thing will ever exist, it's far enough away that we don't need to worry about it.
Finally, I believe that most analyses showing a benefit for a particular population but not overall are fallacious. The multiple comparisons problem figures heavily here.
Posted by: mark at May 15, 2009 9:24:52 AM
Arnold Kling has several further thoughts on this. His most important point (he says and I agree) is his point 7.
It gives the means to restrain health care costs, but not the proper motive or incentives. Indeed, without changing the incentives, more effectiveness studies can more more spending, not less. Suppose that all this effectiveness research is done with government imprimatur. Treatments are not going to be divided into just "what works and what doesn't." There's going to be a fair amount of "treatment X seems to work somewhat better than treatment Y, but is far more expensive and isn't worth it" and "treatment A sometimes works, and works for some people for whom treatment B doesn't work, but we can't predict whom and it's far more expensive and isn't worth it."
Without changes in incentives, when someone else is paying for it, people are going to see that research and say, "Hey, this sometimes works when the other treatments fail, I want to do everything possible." If they're ignoring the cost, then the cost-effectiveness studies become just effectiveness studies, and anything with nonzero effectiveness people will agitate for.
Of course, it is possible that government-as-health-insurer will promote bureaucratic restrictions on treatment of exactly the sort that government-as-regulator often (in response to political pressure) forbids insurance companies from doing. I'm not sure that the Medicare experience bears this out, though.
Posted by: John Thacker at May 15, 2009 9:26:58 AM
http://econlog.econlib.org/archives/2009/05/thoughts_on_com.html
One of his other points has to do with marginal cost effectiveness. Suppose that treatment X is cost effective when compared to doing nothing, but that it's not cost effective in terms of marginal effect compared to treatment Y. Now suppose that treatment Y is invented later, and that treatment Y is somewhat less effective but much cheaper. Do we suppose that anyone will come along and say, "Well, we used to pay for treatment X, but now that there's treatment Y we don't pay for treatment X, even though some people that used to be cured now no longer will be?" In the case of certain advanced diseases, performing treatment Y then treatment X if Y doesn't work is not an option, as the disease will have progressed too much by then if Y doesn't work.
Posted by: John Thacker at May 15, 2009 9:30:59 AM
Tyler Cowen: "When it comes to health care, it's very hard to tell what works."
That's a very broad statement, Tyler. While your statement may be true for a large number of procedures, I doubt it is true for the overwhelming majority of surgeries preformed in the U.S. My wife has been an operating room nurse for 32 years. Her brother had been an emergency room physician for even longer. Both of them have seen:
- dying patients revived;
- deaf patients able to hear for the first time in their lives;
- the faces and limbs of disfigured patients restored to "normalness";
- the severe pain of appendicitus alleviated;
- life threatening arterial blockages cleared so that patients can resume normal functions;
- gangrene halted through removal of limbs;
- patients walking freely again after degenerated hips were replaced.
Would you reconsider your statement, Tyler? As written, it seems to grossly insult the medical profession.
Posted by: John Dewey at May 15, 2009 9:44:42 AM
Tyler, the commission's name is CMS. I used to date a girl who does this reimbursement dollar determination for a living.
First of all, it's extraordinarily political -- only nominally related to the actual dollars spent vs dollars compensated. The RVU metric *originally* was an accurate track of actual labor and supplies expenses used in the performance and recovery from procedures.. but then was corrupted by morons who wanted their care for free.
Tyler, hospitals have a ONE PERCENT margin - they're not growth businesses. Spouting out gratuitous rhetoric doesn't help the situation - it gets nurses fired and inpatient wings shut permanently. The ONLY deep pockets in health care are health insurance companies.
Posted by: Unsympathetic at May 15, 2009 9:54:18 AM
1. This will put an additional cost on developing new medical procedures.
2. If certain procedures are effective in different groups, I foresee that minority and women's activists will have a field day if they are not covered for a procedure that is ineffective on said groups.
Posted by: athelas at May 15, 2009 9:56:49 AM
Arnold Kling, from his blog: "Doing research on medical effectiveness gives people the means to restrain their use of medical services, but it does not give them the motive."
I disagree. If Arnold is meaning that they have no purely financial motive to reject surgeries with a low likelihood of success, then perhaps that is true for some patients. But consider that:
- even routine surgery disrupts lives and adds enormous stress to patients and families;
- though patients are often compensated for days away from work, their families often are not.
IMO, if a patient learns that a procedure is effective only 5% or 10% of the time, the patient is unlikely to take the risks and accept the life disruptions which accompany the surgery.
Posted by: John Dewey at May 15, 2009 10:11:40 AM
Isn't the purpose of the commission to scale down Medicare coverage gradually, based on the estimate of cost-effectiveness?
In other words, people are free to get whatever procedure they want, but (a) we will tell them how likely it really is to work and (b) they will have to pay a bigger co-pay/deductible the lower the estimated cost-effectiveness.
At least that's how it should work.
In terms of solving the "population cohort selection" problem, the task should be defined as follows:
A) A subject is defined as a vector of characteristics (age/race/gender/medical history/etc.)
B) A disease/condition is also defined as a vector of characteristics (major disease type/test result 1/test result 2/stage of the condition/ etc.) [Any of these vectors can be sparsely filled in]
C) A treatment is also defined as a vector (major treatment type, dosage, frequency, etc.)
D) An outcome is finally also defined as a vector of characteristics (survival rate after 1,2,3,4,5 years, disease status, side effects, etc.)
Then the task of the predictor is to take vectors (A), (B), and (C) and predict the outcome vector (D). There is still a ton of wiggle room in the task definition, however.
I agree this will be a particularly devilish problem.
Posted by: mk at May 15, 2009 10:41:08 AM
1) To John Dewey: I agree with your statement that many things (especially in emergency medicine) are causally well-known; however, when it comes to internal medicine, what works is not always well known. Remember the adage, "You don't look for a zebra in Texas." But failing to find the right horse, you start looking for the zebra. The search occurs through tests and their interpretation using statistical findings summarized in little books that every doctor keeps handy and plain intuition (the human element--why robots do not make good doctors).
2) Has anyone looked at the veterinary industry, especially small-animal (pets) predominant practices?
A) The insurance industry for veterinary care is 1970's vintage human insurance; it is geared toward catastrophic insurance, with a reasonable rider available for wellness care. Less than 10% of clients have pet insurance. The rest have to dig into their own pockets and ration care according to their means.
B) Many, many euthanasias take place because the pet will die without care, but the owner cannot afford that care.
C) Most veterinarians provide full hospital services, including dentristry, surgery, pharmacy, behavior counseling, and more. Medicare is not there to tell them that the doctor will not be reimbursed for referring to themselves (this is a big problem for the docs I know well). There is some economy of scale in a one-doctor practice, but equipment is better utilized (and the practice is more profitable) when more than one doctor shares the assets.
D) Veterinarians are considered to be trustworthy and compassionate, and perform well medically, but generally are considered to be poor business people. I suspect the same could be said of most doctors in human medicine.
E) I suspect that people take pet obesity much more seriously than they do their own. It is easier to change a dog's diet than it is your own. This is called "compliance". Non-compliant patients (a majority, in fact) get to receive the same lifestyle and medical advice from their doctor repeatedly, without any improvement in outcome, thereby taking up time that is needed by compliant patients, and transferring more dollars from the patient's (or insurer's or government's) coffer to the doctor's. When deciding what works and what should be paid for, compliance should be part of the equation.
3) Medicare is the big fish in the pond, controlling about half of all healthcare dollars in the U.S. As Medicare reimburses, so follow the insurance companies. We have had a government-dominated healthcare system for many years now, so expecting the healthcare industry to innovate under these circumstances is like expecting a politician to vote other than the party line. Doctor's are very good at medicine, and at innovating to make procedures and outcomes better.
4) What makes government any better than the individual at choosing whether to pay for a treatment? The fundamental difference between the two is that the government has (seemingly) deeper pockets.
Posted by: Free Spirit at May 15, 2009 11:01:29 AM
The first thing we need to do in the United States of America is to get people to take better care of themselves. I really want everyone to have health care, but just walking around an looking at some of the people out there, it's appalling. When you see stores like Wal Mart offering scooters for customers because they are too obese to shop, it makes you really take a step back and question whether health care in and of itself should be a human right. It's highly disturbing to my sense of moral being to think along these lines, but I think it has to be considered. We're spending millions of dollars considering how to deploy further billions and for what?
Posted by: Ian at May 15, 2009 11:05:48 AM
Imagine a government that would cut (some) subsidies for any input which could not be shown, statisically, to causally produce better outcomes in life. You can see how open-ended this would be. What if you applied this same metric for your personal spending? Would there be much left to spend your money on?
That sounds like an excellent plan for government spending on health care. I'm nervous about government involvement, but would be jump right in if the guiding principle was "do nothing which hasn't been proven."
And no, I wouldn't spend much privately on stuff if I had to know ahead of time that it would always pay off. But the government shouldn't spend like a private party.
Posted by: Dan Weber at May 15, 2009 11:20:56 AM
.. lots of people talking past each other, as far as i can tell ..
what can i read that will give me the basic framework for thinking this out, including numbers for costs?
Posted by: babar at May 15, 2009 11:28:16 AM
Ian: "The first thing we need to do in the United States of America is to get people to take better care of themselves"
I agree completely. If our schools spent as much time teaching health and nutrition - and the consequences of ignoring them - as they do teaching multiculturism, acceptance of physical condition, and environmental zealism, perhaps we would have a healthier population.
Ian: "When you see stores like Wal Mart offering scooters for customers because they are too obese to shop"
On this point I disagree. WalMart offers scooters to customers who cannot walk around the many aisles of its large stores. Some who take advantage of this benefit are obese. But there are other reasons for mobility disabilities: degenerated or injured limbs and joints; muscular dystrophy; cerebral palsy; muscle weakness due to aging; paraplegia.
Posted by: John Dewey at May 15, 2009 12:08:50 PM
IMO, if a patient learns that a procedure is effective only 5% or 10% of the time, the patient is unlikely to take the risks and accept the life disruptions which accompany the surgery.
For a life-threatening illness, John Dewey? You really believe that most people will say, "oh, this has a 5 to 10% chance of saving my/my child/my parent's life/giving whoever several more years of life, but it's too disruptive so I'll let them die?" I'm sorry, but I don't believe that.
Perhaps people will decide against, say, hip replacements for that reason, but for cancer treatments? I don't think it's all that unlikely that people will say, "If it's got a 5% chance, let's try it."
Posted by: John Thacker at May 15, 2009 12:17:50 PM
John,
I don't know what the wal-mart where you shop is like, but when I worked there 99% of the people who used the scooter were on it because they were obese. Granted this was in a rural setting where most work was related to ag, so it didn't get used much.
Posted by: nelsonal at May 15, 2009 12:19:12 PM
One problem with state intervention in medical care is that people are really troubled by the idea of the government deciding who lives and who dies.
This is a strawman argument and ignores that somebody could pay for the treatment themselves, or have purchased private insurance that will cover it. Do you really think that Medicare will be able to continue to pay for all care regardless of cost? Do you think that the US will get rid of Medicare?
Posted by: Byrk at May 15, 2009 12:21:46 PM
The UK has been doing (or trying to do) something similar for while now:
http://www.nice.org.uk/index.jsp
Posted by: Nigel at May 15, 2009 12:23:52 PM
Assuming doctors make decisions about what to recommend based on scientific studies, simply doing the studies and publishing them should have some positive effect on costs. There are costs to procedures other than money (inconvenience, risk of complications, and so on) so people would be less likely to try them if they've been shown to be ineffective.
Posted by: Brian Slesinsky at May 15, 2009 12:28:12 PM
John Thacker: "For a life-threatening illness, John Dewey? ... Perhaps people will decide against, say, hip replacements for that reason, but for cancer treatments? I don't think it's all that unlikely that people will say, "If it's got a 5% chance, let's try it."
I agree with you, Mr. Thacker. But I also believe that the overwhelming majority of surgical procedures are not directed at life-threatening illnesses or injuries. That's based on observations of my wife, a 31 year veteran of the operating room.
I'll amend my original statement:
"if a patient faced with a non-life-threatening illness or injury learns that a procedure is effective only 5% or 10% of the time, the patient is unlikely to take the risks and accept the life disruptions which accompany the surgery."
Posted by: John Dewey at May 15, 2009 12:32:07 PM
