Does Healthcare Do Any Good At All?

The RAND experiment showed that people with more generous health insurance consumed a lot more health care than those with less generous insurance, but didn’t have much (or maybe anything) in the way of better health outcomes.  The natural interpretation of this is that everyone, including those with less generous insurance, chooses to get all of the high-value treatments, and that the extra treatments consumed only by those with more generous insurance aren’t worth much.*  If this was true, then Robin’s suggestion to radically cut health care would follow directly; it would be the low-value marginal treatments that would get cut while the high-value infra-marginal treatments would remain.  This would also be consistent with the evidence that it is damaging to one’s health to have no insurance at all (everyone in the RAND experiment had insurance of some kind), as people with no insurance would be missing out on (at least some of) the high-value infra-marginal treatments along with the low-value marginal ones.

The problem is that some of the other evidence from the RAND study is not really consistent with this story.  It seems that the marginal care consumed only by people with more generous insurance is not just low-value stuff.  The marginal treatments consumed only by those with more generous insurance, in the opinion of expert doctors, looks a lot like the infra-marginal treatments consumed by everybody.  But if that’s true, doesn’t it have to mean that all health care is of little value?  If the marginal care looks just like the infra-marginal care, and the marginal care is of little value, then doesn’t the infra-marginal care have to be of little value too?  I don’t think anybody seriously believes that, which makes me think that there is something wrong with the studies that say that the marginal care is just like the infra-marginal care.  Does anyone have any other ideas?

*I ignore here the possibility that these marginal treatments provide little improvement in health as measured by the study, but provide substantial quality-of-life benefits.  I also ignore the possibility that things have changed in a fundamental way since the RAND study was done many years ago; a possibility that Robin has recognized and responded to with a call that the RAND experiment be repeated.

GD Star Rating
loading...
Tagged as:
Trackback URL:
  • http://hanson.gmu.edu Robin Hanson

    David, you seem to be making the common mistake of assuming that each medical treatment must have either a positive or zero health effect. But some treatments have a negative health effect.

  • janne

    Did I get you correctly that the study is about comparing marginal and infra-marginal health regimes, and ends up with the result that the marginal benefit of a marginal regime is close to zero?

    If it then appears that the marginal and infra-marginal regimes consist basically of the same treatments, isn’t the similarity in effects self-explanatory?

    That’s the way I got it, though I find it quite probable that I got messed up with your terminology.

  • Tomhs

    Is it not possible that there is diminishing returns? Even if the marginal care looks like the infra marginal care, there could be large returns to the infra marginal care.

    Have I missed something that rules this out?

  • Michael

    Compare life expectancies through the 20th century with medical treatments available in those years. Out of something like 28 years of total increase for men, a third of the increase occurs by 1940, without access to even antibiotics. Another third of the increase is by 1970, without access to anything we would call “high-tech” care now. No transplants, no MRI’s, etc.

    So my theory is that it’s the cheap stuff that makes all the difference. Public health measures, routine checkups. The high tech, expensive care makes little difference, which explains why life expectancy is so insensitive to insurance, or to different national health care systems.

    See here for a longer writeup: http://www.free-the-memes.net/writings/healthcare.html

  • Kevin Nowell

    Maybe people who live unhealthy lifestyles self-select for the more generous health care? Or maybe once they have it they then make less healthy decisions knowing that they have the generous health care benefits?

  • Caledonian

    As far as I can determine, the effective healthcare interventions are: better nutrition (although this is an ‘intervention’ only in the most technical sense), vaccination against major illnesses, basic hygiene, and first aid.

    The most dramatic interventions that save people’s lives look impressive, but they carry hidden costs – consider the statistics for how many people die due to medical errors alone. Quite a lot of resources are directed at chronically unhealthy people that only gain a little extra time, such as the elderly.

    On balance, we’d probably be better off if we abolished doctors completely. Unfortunately, what we’d lose would be very visible – occasional accidents and emergencies having a higher rate of death – and what we’d gain would be invisible – most people aren’t aware of the harm done by health services right now, so eliminating those harms would make no difference to their perceptions.

    There’s a very important bias you need to address: policy decisions are made based on the perceptions of people who know very little if anything about the topic, not the actual realities.

  • albatross

    Maybe I’m missing something, but a confounding sort of factor here is that when you’ve got generous insurance, you get somewhat different care recommendations. Some of that may be pure bottom-line-padding, and thus pure negative value to the voter–the doctor who needs to pay off that expensive machine (or that expensive mistress) and knows that you’ve got a really nice insurance plan recommends the nuclear stress test, simply to raise money. The same doctor, treating someone with much less generous insurance, doesn’t bother trying to convince him to spend $2K of his own money on a test with marginal value.

  • Cassio

    The RAND experiment showed that people with more generous health insurance consumed a lot more health care than those with less generous insurance, but didn’t have much (or maybe anything) in the way of better health outcomes.
    Not actually. Free health care improved hypertension (about 10 percent reduction in mortality), gave oportunity for the poorest to receive dental care, and reduced the incidence of serious symptoms for poorer people on
    the free plan.
    I defend cost share for health insurances, but I do not think and the study does not prove that free health care does not have effect of health, specially for the ones who cannot pay for adequate care.

  • http://stuartbuck.blogspot.com Stuart Buck

    What’s wrong with saying that some (much?) of both the infra-marginal and marginal health care is either of little value or is harmful? Just in the past month, there have been studies showing that knee operations can be more harmful than plain old rehab, that 20 million people a year are unnecessarily exposed to radiation from CT scans (potentially causing the future cancer rate to rise), and that honey works better than over-the-counter cough medicines (which don’t work at all).

    Less recently, Nortin Hadler found that “the number of people whose lives are saved by bypass surgery, angiograms, and cholesterol-lowering drugs is statistically insignificant.” Other studies have apparently shown that many of the billions Americans spend on back surgery or back pain treatment may be wasted; that antibiotics are ineffective at treating many conditions for which they are prescribed (e.g., sinus infections, ear infections, bronchitis, conjunctivitis).

    One of the objections to universal health care is that it might involve rationing. Well, my gut instinct here is that if you could ration (or even ban!) the right things — perhaps a big “if” — people would be better off.

  • Mason

    “everyone in the RAND experiment had insurance of some kind”

    True, but only in a very nominal sense, the group with the lowest insurance had something like 5% of their medical spending paid for, this was done only to provide them with an incentive to stay in the study and allow for tracking of medical spending. I don’t think 5% savings really made much difference to their healtcare choices, (how much more often would you go to the doctor if it was 5% cheaper?)

    “we’d probably be better off if we abolished doctors completely”

    I doubt it (I still go to the doctor), but we would certainly be better off if we conducted basic surveys to measure the effectiveness of procedures, and then used that information to improve them. (This would eliminate/reduce the negative treatments Robin mentioned).

  • Tiiba
  • Nominull

    Stuart Buck, the article you reference states that cough medicine is ineffective in young children, not ineffective in general. Please do not mislead people.

  • http://stuartbuck.blogspot.com Stuart Buck

    Nominull — readers aren’t limited to the links that I provided; if you use Google, you’ll find plenty of studies that over-the-counter cough medicine doesn’t work in adults either.

  • http://stuartbuck.blogspot.com Stuart Buck

    studies showing . . .

  • David J. Balan

    Robin, there is no doubt that some treatments are harmful, and that medicine (both marginal and infra-marginal) would come out looking better if people were better at identifying these. But I think the question in the post remains even after recognizing that fact.

    Janne, I was using the term “marginal” to denote the treatments received by the people with good insurance but not by the people with bad insurance, and the term “infra-marginal” to denote the treatments received by everybody.

    Tomhs, I think this is the answer to my puzzlement. There is nothing to rule out the possibility that the marginal and infra-marginal care are identical, and for diminishing returns to mean that the infra-marginal care has value and the marginal care does not. I should have thought of it but I didn’t. Nice!

    Albatross, The claim is that the marginal and the infra-marginal care are the same.

  • http://twintecinc.com Steve Burrows

    Tiiba, I would venture that simple access to sanitary conditions and food supply, along with effective vaccines, has contributed the most to the increase in life expectancy shown in the graph. A sheltered childhood could also be a factor in improving child mortality. I doubt that any expensive intervention has contributed to the trend in any measurable way.

  • http://yudkowsky.net/ Eliezer Yudkowsky

    Diminishing returns requires nonindependence of medical events. So the people with more insurance would get a smaller number of treatments for the same event, fewer medical checkups but still checkups, etc. Not sure how plausible this is relative to the study.

    Maybe the judges just aren’t any good at distinguishing “necessary” from “unnecessary” care because they get sued if they ever call something “unnecessary”, but the people who have to pay for it know the difference?

  • Alan Yeung

    It’s not terribly useful to ask whether “healthcare” does anything at all, because healthcare is too broad and amorphous an idea, hence you’re inevitably ignoring confounding factors just by your choice of terms.

    A better starting point would be to divide healthcare into public health, primary healthcare, emergency healthcare (accidents, injuries, etc.), care of chronic diseases (into which one should include old age), and care of major non-chronic conditions (heart disease, cancer, etc.). Then ask whether investing in each one of those does any good at all. I think the evidence points to only the latter two categories being unsatisfactory in terms of return on investment. Coincidentally or not, those two categories also seem to involve the largest monetary expenditures.

  • michael vassar

    Tiiba: To me that chart looks a lot like the sort of linear measure that I get from almost any attempt to measure “progress” broadly by looking at outputs rather than inputs (inputs increase exponentially, outputs rose linearly from 1800 to 1920 then continued flat for a strait line increase when area under integral is measured).

  • Shane

    I’ve seen these kinds of stories many times, and never see anyone actually saying what a “health outcome” is. If “health outcomes” are how many people drop dead or not, then the results are surprising, but not intolerably so, although even here they’re conflated with a number of factors, the most prominent in my mind being the utterly wretched nutrition that the US and many Western countries have increasingly adopted. To take just one example, low-glycemic carb consumption has been steadily increasing for the last hundred years, has been shown in myriad ways to be incredibly problematic, which means that every year the people being treated are fundamentally in poorer shape. Any comparisons across eras will be confused for this reason. Gary Taube’s book is the best single source to catch up on this whole issue.

    But the more immediately obvious objection is this: if I have strep throat, for instance, failure to see a doctor won’t kill me. Seeing a doctor, in this case, will be a zero-improvement outcome, if outcomes are defined digitally (dead or not dead) But the intervention I can get from the clinic will save me maybe ten days of suffering, which is absolutely worth something. I suspect most medical care is of this sort, so if outcomes are defined [in]appropriately it’s not surprising at all that it shows up as useless.

  • Nick Tarleton

    I wonder what Tiiba’s graph would look like if there were any data points between 1920 and 2000.

  • Cássio

    if I have strep throat, for instance, failure to see a doctor won’t kill me.

    Dead wrong. There is an worldwide annual incidence of 30.000.000 cases of post-streptococcal endocarditis, a complication of untreated strep throat that causes congestive heart failure in 20% of the pacients in a 20-year follow-up. On the other hand, it can be acutely lethal in som individuals.

  • Cássio

    I meant prevalence, of course, not incidence.

  • HL

    I’m sure that Shane meant to say that “high-glycemic carb consumption” was a problem, not “low-glycemic carb consumption.” And I’ll second his recommendation for “Good Calories, Bad Calories” by Gary Taubes. It’s a superbly researched description of how the unsubstantiated opinions of a few ambitious medical researchers and politicians can become government-endorsed “facts” that guide medical practice for decades.

  • http://www.gpacharter.org Sharper

    It’s also difficult to divide treatments into marginal categories, because for some individuals in some situations (an MRI for someone with a minor concussion) a treatment is marginal and unlikely to result in a health benefit, while the same treatment in another situation (an MRI for someone with a brain tumor) is much more likely to result in a benefit.

    It’s easy to show that in some specific cases MRI is of only slight benefit, but you can’t generalize that to a statement that MRI use should be forbidden since it’s only of slight benefit.

  • Rachel Soloveichik

    One possible explanation is that paying for medical care is good for you. There’s a lot of research on placebos in general, and recent work shows that expensive placebos work better than cheap placebo: http://web.mit.edu/ariely/www/MIT/Papers/Placebo1.pdf
    It is possible that paying a $10 copay makes a patient get better faster, even if the patient gets exactly identical medical care. As a result, we can’t use price to the patient as a instrument for quantity. Price has an independent effect on health.
    Unfortunately, this idea is virtually impossible to test.

  • Shane

    Whoops, I did indeed mean “high glycemic.” Thanks HL.

    Re: strep throat: nonetheless, I don’t think strep is terminal 100% of the time. If you like, choose some other non-terminal but distinctly miserable condition. The point remains the same, and I sure would like to see it addressed. (I suppose I could do the literature review myself, but in general I think when people post on some topic they should have read the paper about which they’re making claims and suppositions, and be able to answer these sorts of questions.)

  • http://dl4.jottit.com/ Richard Hollerith

    A person who pays full list price for a car will get better gas mileage than a person who pays less for the same exact car! Also, a car can run on sugar water, but only if the occupants really believe that the sugar water will work, so delete those sceptical thoughts from your mind or you’ll be stranded at the side of the road. Just do what the car dealers tells you to do because they’re the ones who conducted all this amazing research

  • K. Larson

    I believe the judgment regarding the ‘necessity’ of the additional care received was made by a panel of physicians. Were these the same physicians that provided the care? Regardless, a simple conclusion would be that medical professionals have a strong bias towards overestimating the efficacy and importance of the services they provide. Years of medical school focuses on the possible negative outcomes of underconsuming medical and the harm avoided through medical intervention- this would seem to create an environment that is highly conducive to overestimating the necessity and value one’s field.

    Ever ask a barber how many “unnecessary” haircuts they provide?

  • Stuart Armstrong

    Ever ask a barber how many “unnecessary” haircuts they provide?

    This a very interesting point, and might be worth digging into. Seems it wouldn’t be too hard either – take one of those studies that are eternally appearing about “disease X costs the economy Y billion dollars a year through sick days”, and add level of insurance, money spent on care, and care outcomes (in terms of sick days) to the mix. Might be quite subtle to tease out the income levels (some of the working poor have less insurance but also less opportunity to be absent from work, not matter how terrible they feel – lawyers have good insurance but similarly tiny opportunity to be absent), but comparing across US states with different health systems might be helpful.

  • gutzperson

    It is known that private health institutions (hospitals) supply more expensive and sometimes unnecessary treatments to people with a more extensive health insurance. This is more because of economic concerns (greed?) than of health benefit.

  • Acheman

    What I never understand in these Overcoming Bias discussions of healthcare is that anything outside the US model of healthcare provision is ignored. Here in the UK, for example, we don’t really have the consumer model for healthcare that you’re describing, because you can’t get a prescription or even a consultation with a specialist unless a GP believes it’s necessary, and the GP has to follow strict guidelines set by NICE, the National Institute for Clinical Excellence, which makes national decisions about which treatments will and will not be available based on strict evidence-based marginal cost/benifit analysis. We don’t have advertisements for prescription drugs on television. We don’t have advertisements for hospitals at bus stops. What the studies you’re quoting actually show is that the consumer model for healthcare is flawed, because consumers aren’t good at predicting how effective treatments will be.
    The only analogious situation of healthcare as a consumer good, and marker of conspicuous consumption, in the UK is the ‘alternative’ healthcare sector, where middle-class people purchase expensive placebo treatments for themselves and their families and then talk about them at length at dinner parties.

  • Joseph Delaney

    I wonder if part of the effect is from the use of health care on those with no possibility to benefit? I noticed that this showed up as a reason that t-PA (an anti-stroke treatment) actually appeared to kill people in actual clinical practice. It could be that patients and their families insist on “doing something” when the situation is hopeless.

    As these treatments may be very expensive, this could introduce a large enough error into the estimates to hide a true beneficial effect.

    In general, the therapies that seem to work well are also not terribly expensive. Antibiotics to cure syphilis are cheap and actually reduce overall costs (as insane people cost money one way or the other — although you might hide the expense by putting it in other government bodies).

    In the same sense, statins for the secondary prevention (given after one heart attack has already occurred) really do seem to be worth the current cost given net benefit (and will be a very cost-effective therapy once the drugs are off patent).

    On the other hand, there is a lot less evidence for other (expensive) interventions — see the whole controversy with drug-eluting stents. It’s not clear to me that this (clever idea that was undertaken in good faith) is going to work out as a good idea from any point of view in the end.

  • David J. Balan

    A couple of the comments above call into question the expert claim that the marginal and the infra-marginal treatments are really the same. And they may well not be. In fact, when I wrote the post, I thought that this was the most likely solution to the puzzle (the puzzle being the seemingly nonsensical result that if the marginal treatments didn’t do any good, and the infra-marginal treaments are just like the marginal ones, then that must mean that the infra-marginal ones don’t do any good either). But now Tomhs’s diminishing marginal returns story seems more likely to be the correct answer. I’m not sure that the story is right, but it is convincing enough to me that I no longer regard the puzzle as a puzzle. Eliezer is right that this story only makes sense if the treatments are non-independent, but I would imagine that this condition is satisfied.

  • Cássio

    Dear Shane, the point remains the same: I do not think you have read anything on streptococal infection and its complications. Despite this, you do not consider yourself wrong. By the way, when the example for your conclusion is not valid, which seems to be the case, the conclusion itself has not been proven.

  • Shane

    Cassio, you are nit picking this to death, and I really hope you’re missing the point on purpose, since if you’re not you are astoundingly dense. Here it is again, with training wheels, which you seem to require:

    There are a huge number of things that someone might go to the doctor for, for which medical intervention is NOT required to stave off death. Since you are so passionate about strep, lets forget that, and go with a sinus infection, for instance, or blood poisoning, both of which I’ve had, without seeing a doctor, both of which didn’t kill me. Or hay fever. Or back spasms. There are an infinity of other things just like these, for which medical intervention could have significantly improved my quality of life. Defining outcomes s.t. saying that this sort of health care has no effect on them is just absurd.

    I do hope you can wrap your head around this now.

  • http://hanson.gmu.edu Robin Hanson

    Acheman, what do you think is the best evidence about the health effectiveness of medicine in the UK?

  • Cássio

    Dear Shane
    Making it clear, since you may have not realised my point: no, I do not think you have to look for a MD every time you sneeze.
    Let us not loose the point of it all, and go back straight to it: The RAND experiment does not say that having no healthcare is better than having some. Read the study.
    Be a good sport and keep your good manners.

  • Mike

    The healthcare system of a nation doesn’t have a whole lot of effect on the life expectancy of the population.
    http://www.nationalcenter.org/NPA547ComparativeHealth.html
    “More robust statistical analysis confirms that health care spending is not related to life expectancy. Studies of multiple countries using regression analysis found no significant relationship between life expectancy and the number of physicians and hospital beds per 100,000 population or health care expenditures as a percentage of GDP. Rather, life expectancy was associated with factors such as sanitation, clean water, income, and literacy rate.8”

  • J Thomas

    Many of us have a chance to test our beliefs about healthcare.

    Give up your health insurance and stop seeing doctors. You can save a significant amount of money that way just in copayments alone. And you can perhaps go to your employer and ask how much of a raise they’d give you in exchange for no health benefits. Something like 48 million americans spent some time uninsured last year, up from 46 million the year before. You could join them. If you’re sure that health care does no good, it would be the right thing to do.

    Then there’s Alan Yeung’s split:

    “A better starting point would be to divide healthcare into public health, primary healthcare, emergency healthcare (accidents, injuries, etc.), care of chronic diseases (into which one should include old age), and care of major non-chronic conditions (heart disease, cancer, etc.).”

    If you could get public health benefits and insurance just for emergency health care, and leave the rest….