June 25, 2009

(Update 9-22) -- Lowering Health Care Costs and Improving Outcomes via Incentives

(Update 9-22-09: How to implement pay-for-outcome-over-time incrementally, with low stakes, and no sudden changes. See update at the end of this post by searching on "9-22")

July Updates: Drug Costs and Setting Prices in a Public or private Plan

This post on exactly how to Pay-for-Outcomes-Over-Time combined with three other simple reforms linked here would constitute cost-effective Health Care Reform.

Also note: this method of payment (below) is a way to encourage and spread coordinated-care systems like the Mayo Clinic, Cleveland Clinic, and Bassett Healthcare by rewarding their effectiveness. Those wanting to go directly to the details can skip down past the 2nd excerpted quote below the video.

When you look across the spectrum from Grand Junction to McAllen—and the almost threefold difference in the costs of care—you come to realize that we are witnessing a battle for the soul of American medicine. Somewhere in the United States at this moment, a patient with chest pain, or a tumor, or a cough is seeing a doctor. And the damning question we have to ask is whether the doctor is set up to meet the needs of the patient, first and foremost, or to maximize revenue.
There is no insurance system that will make the two aims match perfectly.... -- Dr. Atul Gawande

There might be!....

I got a sudden idea today (6/25, below), which I'm mulling over. While I refine it, perhaps others can too. (Updates -- 7/9, 7/23 multiple refinements at this point. I think it's a good outline of how to set up incentives for both everyday and complex situations.)

First, for those wanting a more extensive, deeper look into costs, I encourage those who haven't seen the updates to the Health Care Cost Reduction post below to search (ctrl-f) on the words "effective" and "incentive" for the updates (or read the whole post if you really want a big picture look with issues like technology, psychology, end of life care, etc.).

My initial thinking on how to control costs was to limit a Public (or private) Plan to approved treatments (as existing private plans in fact do!) -- and specifically the most "effective" treatments (definition in this post), in sequence, according to actuarial data on what works best for specific, narrow diagnoses. This seems like a good idea in any case, but there is a more powerful, general way to get excellent care at lower costs and encourage innovation.

Not by specifying what kinds of treatments are used in the Plan. But by incentives alone!

Pay for outcome. Modest pay for trying. Better pay for succeeding. Effectiveness and innovation are rewarded.

I'm not proposing the previous idea of some extra pay (incentive) for a good result. More than that. Instead of a blend of fee-for-service and a little more for outcome, how about payment substantially for outcome.

Before I continue and lay out a plan that answers all the objections I could find to pay-for-outcome, let me illustrate how well-liked the basic idea is by many doctors. Here are only two examples of many:

Dr. Jack Lewin's 2nd point is the current incentives are for volume of care provided, not effectiveness.

Here's a good NPR piece on the same broad issue.

"The high-performing systems of this country are much more able to help manage patients as outpatients without having to be hospitalized, without having to be referred to the emergency room," he says. Doctors in lower-spending areas "are willing to see a patient in the afternoon, start some initial medications, follow them up later in the day to see how they're doing, and if they're doing fine, talk to them in the morning and keep them away from that unnecessary hospital stay."
But the way the government and most insurers pay doctors and hospitals works against the kinds of systems that are more efficient, Fisher says. Doctors don't get paid for things like making phone calls or sending e-mails. And hospitals lose money if patients don't come through their doors.
He and other experts want Congress to implement new payment systems that would encourage doctors and hospitals to work together and give them bonuses for keeping patients healthy, and thus using fewer expensive services.
"And I think that if we're thoughtful about creating incentives for organized systems to form — that would allow them to really practice in the ways that physicians came to medicine to do, and the hospital administrators were trained to do — we could get the kind of performance that we want," he says.

This is the subject of this post -- how to structure incentives that will work in the complexity of the real world.

Let me start by pointing out that paying for outcomes instead of designating approved treatments would actually increase "choice" -- doctors and patients could choose whatever treatment or innovation they like, even within a Public Insurance Plan.

As you will see below, the structure needed is a lot less than we might have guessed, as it is self-forming to a large degree.

Since we want to pay substantially more for good outcomes, we must carefully set or structure the market to set (as you will see) the initial treatment payments so that they defray most of the per-treatment marginal costs, but aren't high enough to be profitable without some success.

For example, for a certain specific condition (specific, narrowly-defined conditions could be listed in a criteria table; see below) the initial treatment payment for that condition might be set for instance at 62% of the full success fee for that condition. The initial payment is made upon treatment, and before the outcome has been determined.

After one or more time periods (listed in the table of criteria, for each condition) have passed and the outcome or intermediate outcome is determined, there would be a failure, partial success or full success, according to the table of criteria. If there is full success, the successful-outcome payment would be the remaining part of the full fee (another 38% payment in our example). For a partial success, the additional payment would be the intermediate-success portion of the fee as defined in the table. For "difficult" cases the overall full fee is increased -- see below. Well-defined levels of partial successes would have defined payments levels.

(update 7-23: establishing outcome criteria)
The tables of outcome criteria -- including time periods for partial to full outcome -- should ideally be formed by large panels of researchers and practitioners in their own specialities via participation in a polling process at the secure internet site where the criteria tables are maintained. This process can easily be formed to take advantage of the increased reliability of larger groups, and should be open to all researchers and practitioners in that speciality, so that participation could be widespread and include even hundreds for some specialties. First, criteria sets, including time-periods, for what would constitute full success and partial success(es) could be created by any participant during the pre-voting period. Participants could also vote on how many distinct levels of success to use for this condition -- 1, 2, 3, 5, etc. -- thus creating the number of partial-success levels. Then a voting period would allow all participants to choose which criteria (including time periods) are most reasonable for full-success, 2nd level success, etc. to the lowest level. There could even be a "run-off" for levels where no criteria set gains a majority on the first vote. After the sets of voting are complete, the results -- for instance the most selected (voted for) of competing criteria sets for a full success would be the result for "full success" criteria -- could then be displayed along with the voting sets (answers) of each participant (to encourage discussion). For narrowly defined conditions, a refined process of this sort could be accomplished in a cumulative 20 minutes or less from most participants as it involves simply listing one's own criteria, and then voting -- notification by email would move each step of the process along. Further, to encourage high standards and discourage vote-fixing these results could be fully visible, transparent, to the public. The resulting criteria tables should be renewed periodically by periodic re-votes and by adding more participants. In this way the criteria tables for most conditions will be an excellent, evolving guide for what constitutes a good outcome and what constitutes one or more levels of partial success. The process itself would be a way to spread knowledge and evolve standards.

Under this system the incentive for a provider -- hospital, doctor, clinic -- is to cure/heal/improve the patient as effectively and cost-efficiently as possible.

They get a initial payment for a treatment, but a nice payoff for a cure or reasonable success. (also see 7-10 update on how to set prices below)


A crucial part of the Plan must be recurring payments over time for successful preventative care, such as controlling high blood pressure or diabetes. Both the patient and the provider should gain from effective preventative care. We want to strongly incentivize cost-efficient care of all kinds, such as accurate emergency diagnoses for instance.


So, we have a basic principle to organize how to pay for medical services:

Pay for success.

(And partial payments for partial successes, according to tables).

The public or private plan could be structured in this way instead of by approved treatments, leaving doctors/providers free to innovate.

The idea is to accelerate treatments, efficiency, and innovation. There is little profit in care that doesn't accomplish much. Instead of repeated, largely-ineffective treatments, the incentives are to find real solutions quickly.

Let's work it out. Is there any reason this would not work well? Let's imagine some difficult situations.

For instance, if a doctor/clinic/provider believes it cannot meet the criteria for success or partial success for one diagnosis for a particular patient -- achieving a standard outcome for a specific condition: curing the cause of the symptoms, relieving some symptoms, or for certain conditions, palliative care -- they may turn down a patient, who is then motivated to search out a provider that believes they can succeed. The original provider that thinks they cannot succeed is not motivated to just try a few things for fees/profit. They are motivated to quickly admit to the patient they cannot succeed, and stop without excessive, expensive, wasted tests/treatments.

When one provider fails (or only gives a diagnosis and quits), the full-success payment for that particular patient would be increased by a fixed percentage (e.g.--another +15 percentage points, so that the full fee with success becomes 115% for this patient), which can be earned by other providers who have no remunerative relationship to the first provider, sweetening the pot, and making the patient more attractive as a customer in spite of their difficult condition. More failures or refusals to treat could lead to bigger jackpots for success (125%...then 135%...then 150% for success), but still only the original initial payment for failed treatments or an exam-only fee for exams without treatment. But, note that the partial-success payment levels would be increasing in proportion to the full-success fee (e.g.-- for instance, a partial-success payment of 18% of full fee amount for meeting certain criteria for a certain condition would increase in value as the full fee amount increases.))

Therefore, even a partial-success becomes more and more worth a try even for the least responsive patients.


Diagnoses themselves should be incentivized, with initial payments for tests, but substantial payments (according to table) for proven-correct diagnoses, contingent on treatment proving a diagnosis via treatment-response criteria. If the diagnosis indicates a certain condition, and the treatment results show a change that meets criteria that prove the condition was present, then the diagnosis was correct. The initial payments for diagnostic tests would be carefully structured to support the diagnostic test costs, but not make tests too profitable in the absence of correct diagnosis. The diagnosing doctor has an incentive to get the diagnosis right, and earns more per hour of work if he/she finds the correct diagnosis, or several correct diagnoses, sooner. The test costs are covered but not especially profitable, and the main payoff comes from getting correct diagnoses.

The patients themselves help police the system (along with the success incentives) -- wanting to be cured quickly and well, they will try to find the best providers available for their condition, and they have freedom to choose providers without plan restriction, since the cost control is built-in (also see pricing update below).

===Update 7/8 --Drug Costs=====

A basic aspect to consider is drug costs. Today, we need effective law against the prescribing doctor receiving any kickbacks or participation in profits from prescriptions. But this need could become obsolete under a better system.

The central issue is getting the best drug choices for the dollar.

Providers should have an incentive to choose drugs wisely, balancing cost and effectiveness. A simple idea to incentivize smart drug choices is to require 1/2 (or perhaps even all) of the drug cost be paid by the prescribing doctor. This would then be a part of the full success fee, and the fee would be set accordingly by the price of the commonly effective drug; but the fee itself would also be independent of drug choice except when specific drugs fail or are ruled out (see below). The doctor has to choose what is cost-effective and works effectively, to get the best economic outcome -- the full success fee without undue drug cost. The doctor then has a strong incentive to find cost-effective drugs for his/her patients.

Sometimes a patient will not respond to one drug or family of drugs or cannot take that type of drug, and needs a different type of drug. If effective alternative drugs are significantly more expensive, then the full success fee would be appropriately higher for the condition of non-response or non-usability of the initial best-practice type of drug.


===Update 7-10, 7-17 -- Setting Prices===

One key practical question in a Public Option (aka Public insurance Plan) or a Private Insurance Plan is how the insurer sets its prices -- the fees paid to doctors/providers.

To understand this section, be sure to first read "establishing outcome criteria" above.

Whether the public or private insurer is paying for piecemeal treatments in an old-style (and costly) fee-for-service system or instead is paying for outcomes (based on specific, narrow diagnoses -- "conditions") in a fee-for-outcome system as described above, the insurer must have a rational way to set prices.

Either way, the answer is an auction.

The exact way to run an auction and set up incentives for bidding are practical questions. Below, I will offer some price setting ideas that can be tested and refined by experience over time by any insurer, public or private.

The Plan (public or private insurer) can simply solicit bids to perform specific treatments or bids for treating specific conditions (bidding is for the "full-success" fee -- see establishing outcome criteria above), and have a community or regional target of having a certain minimum number (or percentage) of providers in that area for each kind of treatment.

Doctors and providers need do no extra work to submit bids. They could simply submit their current pricing (for a fully-successful treatment or outcome) perhaps with a volume-discount of 10%-20% in anticipation of more patients (due to the more people having insurance). Thus little or no effort to calculate prices is needed.

No complex calculations are needed.

Even without submitting a bid or being in a network, any doctor can participate fully or partially in the Plan with little paperwork, or none (see below).

Once the sealed bids are in (via on-line submission), then the Plan will simply accept a bid level high enough to include roughly the target number of providers for each specific treatment or outcome. In this way, the low bidders don't lose, they gain -- they will get the accepted bid level, thus they gain the extra pay difference between their original low bid and the accepted price level.

One good bidding incentive would be to reward bidders that submit bids below the accepted final price with a bonus above that of the price difference they will already receive, perhaps another 5% of that difference. A further refinement is that once an average number of patients per doctor/specialist are treated by a provider during a time period, the low-bid bonus could be phased out for additional patients, gaining an extra price efficiency for the insurer and thus the policy holders. (Note that this is a suggestion of a method and bonus percentage that can be tested and refined over time through experience by the insurer.)

For instance if the bids for a specific treatment or outcome range from $105 to $287, and the target portion of providers is 50% in that region, then the bid at or below which 50% of the providers come in would be the new Plan price for that specific treatment or outcome. In this example, for our range of $105 to $287, this price level that includes 50% of providers might be at $172, thus the new Plan price would be $172 for all in-network providers. But this same payment would be paid to any doctor who chooses to accept a Plan patient, even without any prior paperwork or network status (more on this below).

Providers that submit bids below our example $172 would then would get both the full $172 plus a bonus percentage for submitting a bid below price. (Again, such details can be tested and refined over time by the insurer.)

This is done for all the specific treatments/conditions covered by the Public (or private) Plan in a computerized on-line pricing auction, thus effecting a complete set of Plan prices. Providers then can choose whether to become a Plan in-network provider at these prices.

A few simple rules could handle the question of all-or-nothing in-network status.

For instance, to be certified as a "Full (Public or private) Plan Provider" or a "Preferred Plan Provider", etc., a provider would accept all the Plan prices for the treatments they offer. But...a provider could choose to offer certain treatments at above-Plan pricing (and perhaps advertise as a "Partial Plan Provider" etc.) by giving clear notification to the patient at sign-in of those price premiums, and requiring the patient sign-off on a specific price premium they must pay out of pocket (or via supplemental insurance) before treatment in order for the provider to be fully reimbursed from the Plan for the portion paid by the Plan (lack of a price notification signed off by patient would result in the patient being only liable for 1/3 of the premium above Plan, this amount capped at 10% of the Plan price).

This is only an optional, pragmatic method to allow non-Plan providers to opt-in to widespread Plans with less paperwork and administration.

This allows 100% flexibility. The Plan (Public or private Plan) using this method does not exclude any provider. Any above-price provider could accept any Plan patient with simply a clear notification of their premium pricing the patient or supplemental insurance must cover.

(Another interesting possibility is to allow private insurers to use any Public Plan as a building block within their own plans, in addition to being able to offer "supplemental" insurance. This would allow individuals to purchase private insurance for less established or less reliable treatments which are not fully reimbursed by any particular Plan.)

This is 100% freedom of choice, and 0% interference in the choices of patients and doctors.

Treatments or treating conditions can be auctioned on an ongoing basis, annually or semi-annually. Providers would not need to change their existing bids unless they feel their pricing needs to change. Providers could gradually modify their bids as new ways of treating conditions emerge that are more cost-effective, thus gaining the additional small bonus percentage from lower-than-plan bids.

This promotes cost-improvement over time, as in other areas of service technology.

New Treatments -- providers could offer new unique bids for new possible outcomes. Once enough bids occur for a new type of outcome (a new successful treatment of a previously incurable condition), the Public or private Plan can then initiate an accepted bid level for this new type of treatment at its own discretion, thus including the new treatment into its benefits. Before new treatments/drugs or potential new outcomes are fully accepted as Plan-Established, they could have co-insurance requirements (patients must pay part of the cost out-of-pocket, or via private supplemental insurance).

Straightforward rules like these can handle all situations and provide complete flexibility.


In principle, this kind of pay for outcome could work quite well. It is widely used in a simple form we all are familiar with -- work guarantees.

Savings accumulate to the plan in three ways -- more preventative care due to incentives, fewer tests and wasted treatments due to incentives, and innovation leading to reduced costs for many conditions which will then over time gradually lead to some fees being reduced in the compensation tables.

Combined with the three other simple and easy reforms which involve increasing information to patients and encouraging preventive care, we have complete reform.

It's encouraging that "changing incentives" was one of Obama's major points on ABC the other night.

I'll update this preliminary post as I think of further aspects.

In its simplest form, the Public (or private) Plans would pay set fees for specific narrow diagnoses/ages in fixed regions. The patient could then find a provider willing to try to earn the success (or partial success) fee, in town, or nearby, or in some cases by traveling to a specialized center.

The Public (or private) Plan, and any private insurer choosing this method, would pay its successful-outcome payment after specified time periods for various conditions when a successful outcome is verified by an independent inspector. One key part of this plan is periodic payments for continued preventative success, perhaps annually.

Your thoughts?


7-7: Just saw this great article by David Leonhardt on this whole question of costs and incentives. Highly recommended.


Update 9-22:

Reading this brief Peter Orszag interview it occurs to me to point out an easy, low-stakes way to gradually move from pure fee-for-service towards more pay-for-outcome-over-time.

The objective is to transition with small changes anyone could like and which doctors universally (or near universally) could support.

Reform without big, sudden change would be much easier to implement politically.

Start with a very short list of only a few narrowly defined conditions (such as a severely arthritic knee which needs replacement, or certain heart conditions) and implement small success payments, such as 5%-15% of the normal full cost of complete treatment. The success payment would be made after the treatment is shown successful as indicated by no need for extra treatment(s) beyond the normal follow ups needed for that specific condition within a certain specified time period for each specific condition in the list.

The initial payment, made when the main treatment is complete, would be most of the full cost of complete treatment. For example, if the success payment is 10% (before we also add a bonus percentage for outcome quality), then the initial payment would be 90%. Thus the complete payment for a successful treatment in this example would be the total of 90% + 10% + the bonus percentage.

Another example, for clarity: for a certain condition, upon treatment an 88% initial payment (88% of the full fee) could be made, and if the patient does not need further treatment (other than normal follow up and normal therapy) for that condition during the specified success time period, a 12% success-over-time fee payment would be made after that time period. The outcome quality would then be used to also generate an additional bonus percentage payment. --- For instance, in our example of an 88% initial fee, the success payments might total 14%, for a grand total of a 102% successful treatment payment. For this condition, some patients might not be treated successfully and would thus result in a payment of only 88% of the full fee, while most patients on the other hand would be treated successfully and would result in a payment of 102%.

This outcome-incentive system could be implemented initially for a short list of 5-20 specific, moderately-expensive narrow conditions which have clear enough typical outcomes such that it is easy to specify what is a good outcome-over-time in terms of a no-relapse, or success, time period. Any success bonus percentage (above the normal full fee) can vary according to what is shown to work well over time by experience.

Gradually, several new conditions/success-criteria could be added each year to the list, for instance by vote of panels of doctors, or even better by the full process I described earlier in this post.

I recommend that the establishing outcome criteria internet-polling process as laid out above be implemented from the beginning, to gain several kinds of powerful advantages, such as spreading information about best-practice and new techniques, encouraging market-driven innovation, etc.

This is a simple, limited version of pay-for-outcome-over-time which would have low stakes, and allow a gradual implementation.

Slowly, over time, a pay-for-outcome system could come into being, improving quality and value.

Note: Several ideas from comments below have been incorporated into this Pay-For-Outcome-Over-Time system, which was developed over several months. Note also that comments are forwarded to me, and I respond usually. Also, anyone can contact me via email at: halbhh45@gmail.com


  1. Several problems:
    1) The correct path to follow, known in the trade as practice standards which come from outcome studies, is not available for a lot of what doctors do, sorry. And 'Pay for Performance" does exist some places, but doesn't work as well as it might.
    2) Medicare tried 'payment for diagnosis' in hospitals, called DRGs. Hospital simply found ways to game the system to make more money from the 'simple' DRG. Co-existing conditions, different coding and other things.
    3) What is the 'correct path' for an arthritic knee? How would you define success in someone with an arthritic knee?
    It's not as easy as it sounds. Therefore, allowing doctors to decide, in their own area, what is appropriate and comparing to other areas probably stands the best chance. The Mayo Clinic doesn't rely on 'best practices' to be as good as they are. They're probably ahead of best practices in many areas.
    Doctors can do it. They need to be incentivized somehow.

  2. Robert, I've been thinking on your third problem, a difficult one, in a general way for hours since I wrote the post. Conditions that are hard to solve I've seen in some people I know. A relative who lives nearby has had very serious arthritis and a worsening knee problem for a long time which in fact led finally, today, to a replacement; actually today. She is doing well after the surgery I hear.

    By chance, another relative I knew well had two replacement knees in the late 1980s. That was quite successful, and he had very good use of both legs thereafter (for more than 12 years).

    My best answer at this moment is that we need fine definitions of conditions (diagnoses) to the extent possible (not more than possible, just to the extent possible). So more or less we'd say the knee condition is exactly of ___ kind to degree ____ with ____ extra conditions. That's a lot of variation. So you'd have these very large tables of condition variations, with some moderate precision, and in the table, across the row, what constitutes a good outcome, an partial success outcome, and outcomes less than those criteria would be unsatisfactory. This would handle many conditions well, and not all.

    Now, because of the complexity of many conditions, the minimal compensation payment may itself need to be also set (not always some flat amount like "60%") for many conditions so that the minimal compensation doesn't rule out a reasonable gamble. Also, we would need a lot of "partial success" definitions and rates (perhaps more than one or two for a lot of conditions), and the full success rates may need to vary also.

    With this, a lot of conditions would be handled adequately, and a good number would not be. There may need to be a "general rate" for each significant category (e.g.--"Arthritic Knee General Rate") that would apply as a default when precise diagnosis isn't possible.

    No matter our system, we can't expect perfection. Like in a weather forecast -- exactly like a weather forecast -- we often won't know what might prevent success in an individual. There will be plenty of imprecise diagnosis, or incomplete diagnoses.

    But something can be done about the diagnoses problem itself.

    I'm not thinking this idea is perfect. Rather, it's a step in the right direction.

    It actually doesn't bother me if the diagnosis process itself has complexity and requires several tests. How can diagnosis itself be incentivized correctly? By doing it "piecemeal". By placing a diagnosis success incentive also.

    If after a treatment in response to a diagnosis, there is success in regards to certain symptoms, then this implies that this particular diagnosis for part of the patients problems was correct, thus a successful diagnosis, and thus earns the "success payment" for that particular diagnosis, regardless of other conditions concurrently present.

    In other words, piecemeal success it still a valuable success often and it's okay to have piecemeal success for a patient, and payments for pieces of success.

    I'm aware that regarding point #1 there are plenty of patients/conditions where unclear outcomes are the norm. I don't expect complete success in all patients. Some people can only be partially helped, in any given moment in time. If they have 4 problems that are evident, and doctors fix 2 successfully, and 1 partially, and the other one not at all, and another problem becomes apparent -- this is all okay, good medicine, good practice I think. We can do what we can do. This is reality.

    Have I addressed your concerns?

  3. I've updated the post with more on diagnoses incentives and tonight on drug cost incentives.

  4. I still think your biggest problem with performance incentives is reaching commonly acceptable definitions of "success" in given situations. It has to be immune to quirks in patients, too, so as to allow a doctor who has done his best in an impossible case to be paid something, even if the patient thinks the doctor is ignoring him or out to get him.

  5. The partial success criteria table would list specific types of partial success that would gain partial payments of the full success fee. For some conditions some of the criteria could be subjective, dependent on the patient's own evaluation, but some safegaurds are possible. For instance if a subsequent doctor finds that the patient isn't realistic in evaluating their own outcome, that could result in the previous doctor gaining additional payment instead of only the minimal payment or partial payment.

  6. I like the idea as a concept, but implementation will entail a huge bureaucracy that in my mind cannot be good.

  7. Sue, actually, due to the bidding system being on-line and using existing prices, etc., it wil be simpler than the existing system we now have -- especialy the outcome-based system!

    Right now, you pay on average about 20% of your insurance premium just for the insurer's bureaucracy and net (about 80% is paid out for actual healthcare on average). This is one cost that the above system can bring down (regulation is an alternative though). If you read through the full post, including the "update" sections, you'll see what I mean about less overhead than you'd guess.

  8. So you are going to pay physicians based on outcomes? Never mind that the definition of “good outcome” is often fuzzy - or that patients are complex as individuals and not all are equally “fixable," which will mean physicians simply avoid taking on high-risk patients by saying, "there's nothing we can do for you" (cherry picking). Maybe that’s a good idea - rationing by risk of outcome, but tell that to the population at large and you'll have an outcry!

    An even bigger question is, “When will outcome be determined and payment made?” Take prostate surgery, for example - are you going to wait to pay the surgeon until 10 years later, when the PSA remains undetectable? Often, outcomes aren’t known for weeks, months, or even years. If you want to simply address "short-term outcomes," there are plenty of ways to jigger the system to improve short-term outcomes in ways that can actually make things worse in the long run! This is exceptionally complex, and I remain pessimistic that any form of "payment reform" is really going to eliminate waste. I think it fundamentally comes down to reforming physicians' mindsets and making patients bear a higher share of their health care costs, so they have a greater stake in deciding what gets done.

  9. Anon--fuzzy outcomes are addressed in a central part of the post. To see exactly how I think we should handle fuzzy outcomes (partial success or no success) in detail just search through the post (use Ctrl-f on your keyboard) for "difficult situations" and read several paragraphs. Of course, for that section to make complete sense you'd also need to have read most of the pieces before it also.

    Regarding prostate cancer, the David Leonhardt article in the NYTimes is worth anyone's time:


    Leonhardt writes (among several other interesting things):
    "Some doctors swear by one treatment [prostate], others by another. But no one really knows which is best. Rigorous research has been scant. Above all, no serious study has found that the high-technology treatments do better at keeping men healthy and alive."

    So, clearly this question needs more research.

    I would expect the panels of expert doctors setting the criteria for various specialties (cardiology, etc.) to take into account their own knowledge or lack thereof. Perhaps I should go ahead and insert what I presumed into the post: the success criteria should be set by panels of expert practitioners.

  10. Anon, I've now included a new update to lay out how to best set up and evolve the tables of outcome criteria (search "update 7-23").

    Let me know if you can see any problems. Thanks.

  11. OK, thanks ... I can see how that would work. As an aside, I also think that EMR's would greatly facilitate outcomes reporting - in fact, I'm not sure if outcomes-based payments would be possible WITHOUT an EMR (way too much paper and busywork, etc.).

    Logistically, have you thought about whether outcomes should be reported by physicians or directly by patients? If by physicians, would there be an auditing process? How to deal with situations when patients give different answers about their outcome than the physician submitted?

    Since I brought up the example of prostate cancer outcomes, you may find the following article to be of interest. Published in the latest issue of Urologic Oncology – "Quality of care indicators for prostate cancer: progress toward consensus." Urologic Oncology. 2009 Jul-Aug;27(4):427-34. Do you have online access to an academic library?

  12. Yes, EMRs (electronic medical records) would be essential. I haven't tried to specify in detail how the outcome evaluation would be done, but perhaps dedicated evaluators (agents) would be a good idea. This could be an agent of an agency that works full time in evaluating outcomes. One provision I've suggested in a previous comment for patient-reported outcomes when the patient reports little or no success and then a subsequent doctor determined that the previous treatment was reasonably successful...is for the previous doctor would then receive additional payment above only the minimum fee, etc. This is only a comment idea at the moment, it might need more pieces. But in general, I think dedicated agents could work full time to determine outcome status over the phone or via other contact.

    I don't have free time enough to read further about prostate treatment, so probably abstracts are all I'll look at on prostates. I do think a large group is better at evaluating questions like this. The voting process I suggest is my best idea so far to set standards for complex outcomes. If you have further refinements or ideas on that process, please feel free to suggest them here.