Overconfidence Erases Doc Advantage

Medical expenses now eat 16% of U.S. G.D.P.  That percentage has doubled every thirty years; it was 8% thirty years ago, and 4% thirty years before that.   It will probably double again, to 32%, in the next thirty years. We don’t have many good prospects for reducing that growth, but one of our best is to replace doctors with cheaper alternatives.   Primary care doctors eat a big chunk of our medical budget (median salary 155K$) , yet (confirming previous findings) a randomized trial published in JAMA in 2000 found docs no better than nurse practitioners (median salary 77K$): 

1316 patients who had no regular source of care and kept their initial primary care appointment were enrolled and randomized with either a nurse practitioner (n = 806) or physician (n = 510). … No significant differences were found in patients’ health status … at 6 months … hypertension … was statistically significantly lower for nurse practitioner patients (82 vs 85 mm Hg; P = .04). No significant differences were found in health services utilization after either 6 months or 1 year. There were no differences in satisfaction ratings following the initial appointment (P = .88 for overall satisfaction). Satisfaction ratings at 6 months differed for 1 of 4 dimensions measured (provider attributes), with physicians rated higher (4.2 vs 4.1 on a scale where 5 = excellent; P = .05).

But docs are taught more medicine than nurses; why are they no better at primary care?  Probably because docs are famously overconfident.  For example, one study found that on average when docs were 88% confident that their patient had pneumonia, in fact only 20% of such patients had pneumonia.   And overconfidence is fatal in primary care.

Imagine you are a parent with a sick child, wondering whether to take that child to the hospital.  The key to doing well at this task is to know when you don’t know.  If you see nothing unusual, you should stay home, but if you see something unusual or extreme that you do not understand, you should ask for professional advice.   

It is the same in primary care; most patients are simple and boring: sniffles, rashes, and so on.  Doctors, nurses, or paramedics can all do primary care well if they know when they do not know, i.e., if they can recognize signs that a patient is unusual, and should be referred to a specialist.   And this is where overconfidence is fatal.   Someone who knows less medicine, but admits when they do not know, can do as well as someone who knows more, but is overconfident. 

GD Star Rating
Tagged as: ,
Trackback URL:
  • michael vassar

    Can you cite that rate of growth in medical expenses? I have seen more rapid growth claimed many times but have never seen the claim that the growth is this slow.

  • Michael, I added a link to data going back to 1960.

  • spencer

    Your median salary data is dated. It is now close to $200,000
    — about the same that the manager of your local Walmart store makes .

    but even if you take the larger salary, doctors income absorbs well
    under 10% of total spending on health care.

    Since the role of the nurse practitioner is to treat normal, uncomplicated,
    easy to deal with cases and to screen and pass on complicated cases to the doctor
    I am not sure what it proves that their patients do just as well as doctors.

    Since they treat different sets of patients, shouldn’t the null hypothesis be that since nurse practitioners pass complicated cases on to the doctor nurse practitioners should have a better record, not an equal record?

  • Spencer, I referred to salaries for general practitioners, not all docs, and referred to a randomized experiment where both nurses and docs treated the same patients for the same conditions.

  • djd

    You might want to argue that the airlines might want to hire pilots with only minimal qualifications.The skills and knowledge of an experienced, high-salaried pilot is, I’m sure,needed only in a miniscule fraction of flights and only under unusual circumstances.

  • djd, if you know of any randomized experiments comparing outcomes of experienced versus inexperienced pilots, do share them.

  • Robin, djd is likely making the point that if doctors are better than nurses at relatively rare but serious cases, it might not be possible to determine this with a sample size of only 1316 despite it being important to the patients.

  • djd

    “any randomized experiments comparing outcomes of experienced versus inexperienced pilots”

    That would be a bold airlines to conduct an experiment like that!

  • Robin,

    I doubt it. Naive projections are beneath you.

    The fact is that the US is not only #1 in absolute, real medical expenditures, beating out its nearest competitors such as Norway, Switzerland, and Luxembourg by about 40-50%, but the percentage paid is also about as high as any country in the world, not able to name a single one that definitely beats the US.

    So, you might say, big deal, what happens in the rest of the world does not matter. But I think it does, and I think it will. There will be a political backlash that will contain costs. It may do so in a more restrictive manner a la national health insurance, or it may do so in a more free market way, with the simplest method being to simply open the US to immigration by doctors and other medical personnel. Dean Baker at CEPR has done some serious estimates that medical salaries could be brought down quite considerably by a fairly modest program along such lines. Doctors are just way overpaid in this country, and I think that long before we get to 32% of GDP being spent on medicine, the population will revolt and some combination of the above cost-cutting measures will be implemented.

  • rcriii

    The participants had no regular source of care and were referred after emergency room or urgent care visits. Thus they are not a representative sample of the population. My first guess would be that this was a group likely to be less healthy than the general poulation, who could have expected to benefit from a doctor even more than most.

    I also note that in my case (in a PPO), going to a nurse practitioner would not save me any money, as I pay a flat fee for doctor visits. Even if my PPO has any nurses in the network, I’d have to go out of my way to find one and save them the money. Talk about skewed incentives…

  • Ok, looks like I misinterpreted what djd said.

    Anyway, this might not be the best post to discuss this, but I have some concerns about Robin’s and Eliezer’s interpretations of overconfidence statistics.

    1. When a human says they are x% confident, they may mean that the word “confident” is x% appropriate to describe their feelings, rather than that they estimate the probability of something as x%.

    2. People, including doctors, tend not to be all that numerate, so stating a figure of 88% may be dependent on a subjective feeling of the degree of bigness of 88%, which may not be the same thing as estimating a probability as 88%. Asking for betting odds may be better for encouraging people to think about numbers as numbers rather than vague signifiers of degrees of bigness.

    3. Experts tend to have an incentive to appear confident in order to enhance their reputations, so their apparent overconfidence may be greater than their real overconfidence or the overconfidence of people who do not have their reputations dependent on appearing confident.
    (this also applies to situations where an appearance of confidence may help one win arguments, or any other situation where one has an incentive to appear confident.)

  • Douglas Knight

    Barkley Rosser,
    zero doctor salaries can’t get us down to France level expenditures, let alone Singapore, where individual doctors are paid the same percent of GDP as in the US (although there are half as many, so the aggregate is the same as in France, which has the same doctors per capita as the US).

  • djd and simon, how many patients over how many years do you think it would take to see a difference between docs and nurses in primary care?

  • Barkley, for over six decades people have been declaring medical costs were reaching a “crisis” point where they wouldn’t be allowed to rise much further. And yet they have kept rising.

  • michael crichton

    I think what this post is really telling you is that an individual’s sense of clinical judgement is overrated to the point of being dangerous. A similar circumstance applies to psychologists, who are most accurate in making diagnoses when they are young, and tend to rely on checklists. Later, as experienced practicioners, they rely on clinical judgement and misdiagnose. This means that psychologists become demonstrably less skilled as they become more experienced. A sort of inversion of expertise. See Robin Dawes, House of Cards.

  • Robin: Quite possibly never, since I do think it is possible that there is no difference. (I was trying to clear up what seemed to me to be a misinterpretation of djd’s comment, when in fact it looks like I was the one who was mistaken). However, the real question is how many patients over how many years are needed to show that there isn’t an effect large enough to be worth caring about? I’m not sure exactly, but if doctors are better at, say, diagnosing rare severe conditions early enough to save patients, then you would need to have enough patient-years to have a fairly large number of deaths in both groups, in order to get a statistically significant difference in death rates. The study you linked to, as far as I can tell, didn’t measure death rates at all.

  • Rick Davidson

    Interesting and surprisingly short-sighted. As a primary care doc who has researched the roles of nurse practiioners since the late ’70s (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=pubmed&cmd=Retrieve&dopt=AbstractPlus&list_uids=7328200)I can tell you that while NPs can provide much of the care you discuss, they cannot and should not replace physicians as primary care providers. It’s precisely the lack of “background information” that determines the primary difference in education between NPs and MDs. All primary care patients do not have colds and need simple health maintenance and hypertension. Many, especially in internal medicine practices, have multiple chronic disease states such as diabetes, chronic lung disease or heart failure. It should not be surprising that NPs are better than physicians in caring for some basic ailments, but it’s the lack of training in slightly more complex situations that necessitates having medical consultation an important aspect of outpatient care. Additionally, if there ever was a misdirection in terms of overspending of medical resources, this is it. Blaming the astronomical rise in medical costs on primary care is absurd. Try looking at the outrageous cost of procedures, advanced technology…and if you want to look at salaries and claim that doctors are overpaid, you might want to look at some of the variation. As a senior physician in an academic health center, I am paid 25% of what a newly recruited cardiothoracic surgeon received as a salary. Luckily, money is not an issue for many primary care providers, although their debts through student loans are the same as those to their classmates who went into much more lucrative specialties such as ophthalmology or dermatology. Our average graduating student has a debt of over $100,000 starting residency, for which they are not paid much. There are certainly excesses in health care spending, but blaming those on primary care salaries is ludicrous.

  • Michael, I will read that book you recommend.

    Rick, how large (in people-years) a randomized study do you think it would take to see the difference you hypothesize?

  • Douglas Knight,

    France is reputed to have the best medical system in the world. But I do not think we need to get down to their level of expenditures in either absolute or percentage terms. I would be fine with a cessation of further increases in the percent of GDP going to medical care. Capping, better yet, cutting (not to zero, please, let us not be ridiculous) doctors’ overblown salaries, would help with this. To give a ballpark number, Baker (one can argue with his model of course) suggests that a fairly modest program of allowing doctors to immigrate could bring average doctors’ salaries down from around $193,000 (almost certainly an underestimate of the actual, he says) to around $135,000, a pretty respectable income I would say, but enough to make a dent in those expenditures.


    It is not just doctors’ salaries. There are a lot of other factors that are contributing to costs, including to pick two that are on the radars of the opposite sides of the political spectrum, insurance costs for overblown malpractice suits and extra administrative costs associated with our private insurance system and its efforts not to pay for peoples’ expenditures. Reforms of one sort or another are possible, if not politically achievable, in a number of areas that could slow the rise of costs. If costs do keep rising astronomically, at some point presumably the political will will come to be there.

  • 1) Anyone have data on veterinarians? My understanding is that they make substantially less than MDs, though the time they spend in vet school is the same. Apparently to date malpractice awards are limited to the cost of the animal. I.e., no pain and suffering by the owner. As a result, less need to over-test. Also, the owner’s price elasticity of demand for the life of the patient is more elastic than presumably ours for our own or our family’s lives.
    2) Robin: my experience with Army medics is consistent with your hypothesis. They were good at the simple stuff, and knowing how little they knew, were quick to refer to an MD if something looked hinky. They had been taught humility; my sense of medical school is that one is taught, for some good reasons including placebo effects, to exude confidence, which is easier to do if one actually has it, merited or not.

  • theCoach

    Predicting is hard, especially about the future.

    “That percentage has doubled every thirty years”
    “We don’t have many good prospects for reducing that growth.”

    I do feel confident in predicting that this growth will not continue indefinitely. I would think our best prospect for reducing this prospect is simple mathematics — 32, 64, 128 — my understanding of GDP is that 128% going to health care is unlikely.

  • Shakespeare’s Fool

    But, “theCoach,” to get to 128% of GDP, do we need do anything more than keep borrowing from China?

    Robin Hanson,
    Excellent topic. I, too, am headed to the book Michael Crichton recommended.

  • Douglas Knight

    Barkley Rosser,
    All I was saying was that a backlash that simply contained doctors’ salaries would not contain costs, unless you have a model of how they cause all the other costs (which is unlikely to be consistent with Singapore; but malpractice costs would probably go down). A backlash in the form of nationalization also seems unlikely to contain costs, given how small the gap between costs in the US and France is. Costs are rising exponentially everywhere (including Singapore).

    You can point to a lot of sources of rising costs in health care, but even if you can get the accounting to work out, I don’t think tackling them individually will have much effect. There’s probably some central reason and the individual line items are just leaking out (eg, rent-seeking, as in both doctors’ salaries and malpractice).

    My impression of the claim that France has the best health care in the world is that it means that the outcomes are at the uniform level of Europe (and the US!) without problems that suggest rationing. I think Singapore passes this test as well, but that people who make these reputations are simply ignorant of it.

  • More Reason To Support Walk-in Clinics

    On the overconfidence of doctors: 1316 patients who had no regular source of care and kept their initial primary care appointment were enrolled and randomized with either a nurse practitioner (n = 806) or physician (n = 510). … No

  • Stuart Armstrong

    If costs do keep rising astronomically, at some point presumably the political will will come to be there.

    How can you be sure? Someone dying on a hospital bed would presumably pay any amount of money he could afford to survive (with a good quality and length of life afterwards).

    If we let the market decide (either the monetary or the political market) I don’t see any reason why medical expenditure wouldn’t reach 80-90% of GDP.

    There are people on this blog (myself included) who have ambitions of becoming immortal. How much of your income would you be willing to pay out to achieve that goal? Because that’s the market upper bound on the percentage of GDP devoted to healthcare.

  • Stuart Armstrong

    Blaming the astronomical rise in medical costs on primary care is absurd. Try looking at the outrageous cost of procedures, advanced technology…

    Indeed. There are well established industries, like water companies and agriculture – they are characterised by low profits, large amounts of consumers who pay little, stable business models and a declining proportion of GDP. New industries behave differently – high profits, high innovation, little stability, relatively high prices for consumers, and a rising share of GDP.

    Healthcare is an established industry that behaves much more like a new industry. Also patients have the propensity to want a “life and quality of life”-saving treatment, if it exists, at whatever the cost. This means that medical innovations push up costs more than in most industries.

    So if we really want to reduce medical spending while maintaining acceptable standards of care, the most rational thing to do is to reduce the pace of medical innovation. Removing patent protections from medical drugs would be the easiest way to do this, and would bring down medical costs in the short term as well.

    But if we want to maintain fast medical innovations, we’re probably addicted to rising healthcare costs (not that we can’t try and slow that rise through other means – but slow is all we can do).

  • Rick Davidson

    Robin, there have been a number of studies in the past that looked at comparisons between specialists and primary care physicians in the management of various conditions; in fact, a systematic review has been done that pointed out a number of methodologic issues (http://archinte.ama-assn.org/cgi/content/abstract/167/1/10), but a comparison could be done. Because of the natural history of common ailments, a very large sample size would be expected to avoid a type 2 error–probably too large to be realistic. However, an old friend of mine, Tim Carey, published a NEJM article about the care provided for back pain by multiple specialists, including NP’s, and found no difference in the quality of care using a very simple outcome measure. Interestingly, patients of chiropracters were most satisfied with their care….and spent the most money on care, even though there were no differences in outcomes (http://content.nejm.org/cgi/content/abstract/333/14/913).

    Many years ago Dave Sackett was involved in several studies that demonstrated a decline in the knowledge of the best treatment of hypertension that was associated with the year the physician graduated from medical school. A strong argument for evidence-based medicine, as it’s the lack of reading the current literature that is primarily responsible for this deterioration.

    At our institution the tuition in vet school is less than med school, and most vets begin their practice immediately after graduating; med students have a minimum of 3 and a maximum of 8 years of residency before starting their practice. Residents’ salaries will not make much of a dent in a large debt. Don’t get me wrong, I’m not arguing that doctors are underpaid; they’re not, but the distribution of the payment is what’s wrong with salaries. Why should an ophthalmologist make $5000 for a 20 minute cataract extraction and an internist who spends an hour and half trying to diagnose a complex problem get reimbursed $60?

    If you look at the figures regarding rising medical costs, there was one time (the early ’90s) that the rise leveled off and in fact even decreased (http://hspm.sph.sc.edu/Courses/Econ/Classes/nhe00/). The reason? Significantly managed care. That was the time of the aggressive HMO, when all procedures were questioned, and there were significant limitations in provider choice. What happened? The public would not accept those kind of restrictions. If you saw “As Good As It Gets”, remember Helen Hunt’s rant about HMO’s? So health plans changed. Less than 1% of requests are denied, some plans don’t even ask for approval now, and the selection of providers is much broader with more options. And with that relaxation, costs immediately began rising again. Thing will get worse. The population is aging, and there’s new and expensive technology developed every day. I agree with the comment about the political will being the driving force. It’s one thing to have lower middle class workers lose their health insurance; when this starts happening to a broader section of the population, who wield more political power, there will be a national health plan of some kind.

  • Rick, thanks for the reference to that interesting review article comparing specialists to generalist docs.

  • I found it easy to imagine that people mistakenly believe pilot safety improves with experience, so I did a quick search and found signs that the evidence is inconclusive.
    There seems to be evidence that pilots with 100+ hours of experience in the type of aircraft they’re flying are substantially safer than less experienced pilots.
    For much larger amounts of experience, there’s evidence that people overestimate the value of experience, but we can’t conclude that increasing experience is worthless. Some quotes from http://www.faa.gov/library/reports/medical/age60/media/age60_1.pdf:
    “Outcomes from Golaszewski (1983, 1991, 1993) where certain pilot groups demonstrated an increased incidence of accidents with increased age still showed evidence that experience and recent flight time were somewhat protective.”
    “Researchers and the general public are typically of the opinion that experience will reduce the magnitude of the age differences in cognitive functioning. As Salthouse (1990) points out, despite the general acceptance of this concept, the scientific evidence is less clear cut.”

  • Peter, thanks for taking the trouble of looking that up!

  • Pingback: TheMoneyIllusion » That American entrepreneurial spirit()

  • Thomas H.

    Questionable pneumonias are nothing new. The question is who decides what represents a pneumonia and what the gold standard is for the study you site? Oftentimes two radiologist both with years of experience reading chest x rays will differ in interpretation son chest x rays. That is where experience and training and the art of medicine come in. Ad a patient you want a Doctor who is going to weigh the risk benefit of treatment correctly in the case of a questionable pneumonia. It wasn’t overconfidence but caution that leads to “overduagnosis” of pneumonia and let’s not forget that antibiotics currently have a role in the treatment of copd under current treatment guidelines. The scoring and “core measures” used to measure physicians are what is deficient. They aren’t sophisticated or even correct enough to measure the effects of superior medical education. One thing is for sure less training and knowlege are not beneficial in highly complex fields such as medicine.

  • Pingback: Overcoming Bias : Trust Govt More?()

  • Pingback: John Fox on AI safety | Machine Intelligence Research Institute()