• Physicians care about patients and money

    In case there is any doubt, a new study using experimental methods shows that how physicians get paid and by how much does affect patient care, but it is not the only factor. The health of patients matter too. I find this relatively unsurprising. Yet I think the work is worthy of note because not everyone may think as I do. Plus, the methodology is interesting and leads to very clean, clear results.

    The paper, by Heike Hennig-Schmidt, Reinhard Selten, and Daniel Wiesen, appeared in the Journal of Health Economics and is titled How payment systems affect physicians’ provision behaviour—An experimental investigation. As the title states, this was an experiment, not an observational study.

    In a controlled setting, the researchers asked medical students to choose the quantity of medical care to provide to hypothetical patients enrolled in either fee-for-service (FFS) or capitated insurance plans. Under the former, physicians are paid for each additional unit of care. Under the latter, they receive a lump-sum payment independent of units of care provided. In the experiment, quantity of care is an amount indexed by the integers 0 (no services) to 10 (the most services). In advance of making the quantity selection, the physician has full information about how the quantity selected will affect payment, costs, and profit (or income) and how it will benefit the patient. Here’s how it looks in the experimental setting (click to enlarge):

    As shown in the figure above, revenue varies by quantity for patients assigned to FFS type payment but not for patients assigned to capitated CAP type payment. What’s shown above are the payment, cost, profit, and patient benefit schedules for patient type 1 with illness type “E”. There are other types and other illnesses though, with different schedules. Patients vary by illness (there are five, labeled “A”, “B”, “C”, “D”, “E”) and by three levels of need for services (level 1 needs 5 units of service for optimal health, level 2 needs only 3 units, and level 3 needs 7 units). I know this is a lot of detail, but it is necessary to understand the results, which is where it gets interesting.

    Oh, before I get to the results, one more cool thing about the experiment. The physician (student) participants actually earn the money they generate from treatments prescribed in the experiment. Since there are no actual patients, the researchers included an incentive for the physicians to take patient concerns seriously by converting patient benefits into contributions to a charity that cares for real patients. Given the constraints of experimenting on actual people, I think this is a very cool design. Still, one might wonder how things might be different in the presence of real, flesh and blood patients.

    OK, about the findings:

    1. Payment systems matter. More services are provided under FFS than CAP. On average, patients receive more services than are optimal under the former and fewer than optimal under the latter.
    2. Patient health matters. That is, physicians do respond to how much treatment benefits patients. Still, under FFS, patients in good and intermediate health are overserved. Under CAP, patients in poor and intermediate health are underserved.
    3. Payment systems affect health (or patient benefit). Patients in good and intermediate health suffer losses under FFS due to overprovision. Patients in intermediate and poor health suffer losses under CAP due to underprovision.

    It should be perfectly clear from these results why patients in the real world might self-sort according to health, even aspects of health that are unobservable to the researcher. A patient in poor health attempting to optimize his benefit would do better under FFS. A patient in good health gets better results under CAP.

    Even though I already stated the results, the following chart conveys them so beautifully you really must take a look. (Come on, you’ve made it this far!)

    Remember, A-E are the (abstract) illness types and 1-3 index how much medical care would be optimal. The solid, black dots show the optimal level of care. Patients of type 1 (1A, 1B, etc.) would do best with 5 units of care, etc. Notice that under both payment systems, actual quantity provided is correlated with what would be optimal, highest for type 3 patients, lowest for type 2. So, patient needs matter. Still, patients needs don’t tell the whole story. Provision of care under both payment types differs from optimality, most strongly for types 1 and 2 under FFS and type 3 under CAP. Finally, CAP levels of care are systematically lower than under FFS.

    Conclusion: money matters, though so do patients. Lesson: your doctors, or medical providers generally, are not just taking care of you. They’re taking care of business (themselves) too. Don’t even try to convince me you’d behave (qualitatively) differently in their shoes either. We all like to get paid! Keep in mind, though, your doctors probably don’t know the exact optimal level of care you need, as those in this experiment did. That probably makes it more likely their decisions will be influenced by other factors, including financial. On the other hand, in the presence of an actual patient, maybe they make different choices.

    Still, I believe this study reveals a fundamental truth: money matters (more evidence here).

    • Very interesting article, thanks for posting. It is important to note that over-provision of healthcare is not without real health costs to the patient. More tests and more procedures increase risks of adverse events (whether do to excess care itself or excess care driven by excess diagnositics – the WSJ healthblog had an article this week on how myriad blood draws for inpatients causes material amounts of anemia!). Adding this lense, it becomes clear that the CAP method delivers close to optimal care, on average, whereas the FFS model has large risks for patients in good and intermediate health.

    • This is great stuff, and the kind of thing I would never come across. Thanks for posting.

      Question though, as I dont have a copy of the study. What was unique about illness 3, if investigators specified, that caused underprovision of care in CAP scenario. Its differentiation is distinguised (and stark) from #1 and 2.

      Also, if I understand correctly, participants actually got the virtual cash to donate to a real charity, yes? The amount also may (or may not) influence outcomes. I think of a poker game with nickel/dime stakes vs dollar/five–you get more “serious” as amount increases, even if money does not enter your pocket.

      I agree though, rather ingenious, and I wonder if that style of testing has been validated and reflects behavior as it would unfold real world.


      • The first two-panel chart in the post shows you exactly the type of thing the participants saw when making Tx decisions. Type 3 just got maximal benefit from a higher quantity than 2 or 1, as I stated and as is shown in the second chart (black dots).

        Participants got paid an amount proportional to the profit associated with their Tx choices (as per the first chart). They were converted to Euros by some factor. Benefits were also converted to Euros and given to a charity that cares for patients. Participants did not get to select the charity, nor get to handle the benefit Euros.

    • This conforms with what I have witnesses in private practice. I think that everyone is at least partially affected by money,whether they realize it or not. I have also observed over the years that there is a smallish subset of docs for whom money is the primary motivation in their practice. Before retiring, my wife worked for a guy in Florida who had sheets in the office outlining what care patients should be given based upon their insurance.


    • Often these decisions are posed as though physicians/providers make decisions all on their own, as in a traditional small private practice model that it seems we all carry around in our heads. In such a practice, the question of who is making the decisions on behalf of whom is relatively simple. There are advantages to that simplicity in the sense that everyone at least has a chance of understanding what process is going on and who benefits from what.

      What so many of us do now though is work for large practices, hospitals, clinics, or other similar systems. In such systems, administrators with administrative concerns have a lot of influence on providers’ decisions. This is true even when administrators carefully draw a line and refrain from directly telling their physicians etc. what they should and should not do in terms of medical decisions. There are more subtle forms of influence that are very powerful.

      If I am a physician and am expected to achieve X amount of billable hours by my employer, that impacts how I practice. Chart reviews, for example, are a casualty of this, since they don’t usually contribute to the all holy billable. Is there a good study on, say, correlations between time spent charting and reviewing charts vs. errors and #s of tests? I’ve often wondered. Since action is paid and thinking is not, it seems to me we get a great deal of action and not a whole lot of thinking sometimes.

      Getting paid to think is a bit of a joke, but if we are going to persist in these models in which providers are made sensitive to revenues, it may be exactly what we need.

      Similarly, if I’m told my organization makes more for seeing a new patient than for doing a good job maintaining the health of an existing patient, that may sway my choices too, even in subtle things, like who gets the last appointment of the week. I am after all subject to performance reviews by my employer and perhaps I’d like to be promoted someday within the organization. Or, perhaps I’m less venal and instead motivated by the desire to be seen as a helpful team member within my organization. Perhaps I want to make sure that my organization can continue to provide well-paid jobs with good benefits for the people who work with me and I don’t want to see the receptionist laid off if I fail to bring in the billable.

      These things matter and it is only the most insensitive physician (in both social and economic terms) who will ignore them. Increasingly I suspect that such physicians may find it more difficult to obtain/retain employment, since if you do not cooperate with admin, admin will find someone else who will. Perhaps I am too pessimistic about this…but it seems clear that this is how business operates, and that large healthcare organizations are increasingly businesses. Non-profits are not immune either, since they too need their billable #s to work out.

      I greatly enjoy your blog, thank you for providing these updates and summaries. I find them tremendously helpful in trying to understand what is going on in our healthcare system. I apologize if my comments are reinventions of the wheel; I am a new-ish reader and admit I have not reviewed past entries in the detail they undoubtedly deserve. I wanted to put in my $.02 to encourage you to continue to look at this issue of how people are actually influenced vs. how they think they are influenced.