THURSDAY, MARCH 3, 2005

Session 3: Aging and Care-Giving: Assessing Outcomes in Medical Care

John E. Wennberg, M.D., Professor of Community and Family Medicine, Dartmouth Medical School

CHAIRMAN KASS:  Would Council members please take their seats so that we can get started?  Could I ask people in the back to join the table?

Our third session today entitled "Aging and Care-Giving:  Assessing Outcomes in Medical Care" introduces us to an aspect of the subject that we haven't considered before.  We're going to step back and look at certain features of the current health care system, its utilization, its priorities, and its goals.

Joanne Lynn's presentation in the morning already raised for us the question not only about how much we are spending but what we are getting for it and what we are spending it on.

The Council has been exhorted to look at the question of overall issues of health care.  When these things are generally discussed, both publicly and in the bioethics literature, they focus mostly on questions of justice and economy, equal access, insurance, wastage of resources at the end of life better used elsewhere in medicine or elsewhere altogether, but little attention is being paid, really, to the soundness of some of the underlying assumptions:  First, that more resources for more of the same kind of care, more available for more people, will produce better health outcomes; and, second, that our overall medical goals are clear enough and reasonable and that our priorities in medical expenditure are sound.

The two sessions this afternoon are designed to help us examine these assumptions.  They bear on the issues of aging and caregiving, perhaps only indirectly, and the implications for that topic will be left for us to draw.

But since attention to the health care and human care needs of the elderly are clearly affected by the functioning of the health care system in general, consideration of these practices based on these and other assumptions are surely relevant to our overall project in the area of aging and care.

In this first session, we are very fortunate to have with us Dr. Jack Wennberg, who is Professor of Community and Family Medicine, Professor of Medicine and Director of the Center for the Evaluative Clinical Sciences at Dartmouth Medical School.

He has been one of the leaders in the development of the field of evidence-based medicine and outcome studies, looking at various ways in which practices vary region by region and trying to raise questions about whether or not current care is, in fact, effective care.

He is also the principal investigator and series editor of the Dartmouth Atlas of Health Care, which examines these patterns of medical resource intensity and utilization in the United States.  This project has done some reporting on the patterns of end-of-life care, inequities in the Medicare reimbursement system, and the underuse of preventive care.

We are very fortunate to have Dr. Wennberg here.  We look forward to your presentation.  Welcome.

DR. WENNBERG:   Thank you very much.

 I think what I would like to do today is break the talk into two parts with a period of questions in between.  I think, Leon, you are going to do the moderating of that.  The reason is because the topics are different and they are also complex.  And I don't want to have to wait until the end to get some of the concerns and interest out on the table.

I trust that you have all had a chance to look at the manuscript that I sent down.  So I'll try to be brief in terms of my presentation and concentrate initially on what I call effective care and preference-sensitive care.  I'll come to define that a little bit further down here.

Just to give you a little background, the basic perspective that I bring to this debate about health care is that of an epidemiologist.  And that was my training, and that is what I have been doing most of my life.

Instead of looking at the incidence of illness and trying to associate environmental factors with the incidence of illness, my particular interest was in associating or studying the health care system itself as a topic of investigation.

This goes way back, but I am going to bring you up to the last decade or so of this research.  Beginning in the early '90s, we obtained access to the Medicare claims data, the full files for the whole program, at least for the traditional Medicare.

Using those files, we were able to divide the country up into regions based on the simple distribution of where people went for care, for hospital care, as a function of the zip code they lived in and the hospital they used.  And this enabled us to divide the country up into some 306 referral regions.  The variation between these regions will be the target of my initial comments at least.

More recently, we have also been able to look at individual hospitals.  And since that is where the money is, I do want to share some of those results with you because I think if we really get into the problems here, you will see this as a multifaceted problem dealing with — depending on the category that we're talking about, the remedies are different.  And the opportunities for reform are different.

But fundamentally the underlying problem in all of this work, with the exception of the effective care category, is scientific uncertainty and lack of clarity about the relationship between resource inputs, medical theories, and outcomes.  So we'll come back to that.

Okay.  Well, I want to talk about what I call unwarranted variation.  And that is variation that is not explained by illness, patient preferences, or evidence-based medicine.

Most of us, I think, assume that most of medicine is warranted in the sense that it does correspond to these definitions and explanations, but, in fact, it doesn't.  Almost all the things we see in the Medicare population demonstrate this extraordinary variation.

Effective care is the first category.  This basically is a fairly restricted amount of care.  Even though there has been an awful lot of emphasis on evidence-based medicine, so far we haven't found very many things that you can actually measure that are evidence-based.

Now, that's partly because the claims data don't allow us to, but also even when you go way behind the clinical trial definition of effectiveness, as Beth McGlynn has at RAND Corporation, you quickly run out of things to look for, even when you can do chart review.

So everybody seems to find underuse of effective care, no matter where it is.  And this was the emphasis of an awful lot of the initial IOM work.  And certainly when you hear about pay-for-performance programs now, they're going to pay doctors to do things they ought to be doing.  It turns out that that is not a bad idea, that they should do what they should do, but whether you should pay them for it is another question.  You already pay them quite a bit.

The interesting fact is that — and I'll give you sort of a — just so you can carry this thread through the conversation here, when you look at what correlates with spending in Medicare and what correlates with large amounts of resource use, it's not effective care measures.  In fact, the effective care measures are better in regions that have less intensive care, which is fascinating.  Fewer physicians leads to better effective care, maybe because there aren't so many people involved in the care of the patient.

And when it comes to costs, there is basically very little correlation between the specific items of effective care and the costs in the region or in the patient cohorts experience.

Just to make it clear — and I want to spend just a minute with these dots here in case you haven't thought about them before.  There's a dot up there for every one of the 306 regions.  People ask me, why do I bother to put these dots up like that in that funny distribution.  It's because it's easy to put a label on them so I can explain to you which dot is which.  That's the major reason.

This slide tells you quite clearly that when it comes to a simple thing about a diabetic getting an eye exam, where the criteria should be close to 100 percent of people, we're not doing it enough in most regions.  And, in fact, in some regions less than 50 percent of the population in Medicare are actually getting that eye exam, just taken a couple of years ago.

I'm not going to talk any more about this other than to say that in sort of a schematic way, we can say that in terms of the benefit-utilization curve, we're under-using effective care.  And that's what this is meant to depict here.

So the U.S. is basically in the zone where more care should be done, but it isn't.  And the irony is this is cheap to do, but it just doesn't get done.  So that is kind of the bottom line on that category, I think.

Now, preference-sensitive care is of a different nature.  It basically involves treatments for which there are options or conditions for which there are treatment options.  A classic example would be a woman with early stage breast cancer who is faced with either a lumpectomy or a mastectomy decision.

In this particular case, the clinical evidence is pretty strong.  And we do know that the mortality implications or the survival implications are about the same, but everything else is different.  The lumpectomy involves radiation, possibly chemotherapy, and the possibility of local recurrence, which then would require further surgery.  The mastectomy avoids most of those problems but clearly involves problems of reconstructive surgery and so forth.

I don't think that anybody here would try to make the case that it's the surgeon's job to prescribe for the patient what he thinks she should have, but, in fact, that is the dominant pattern of practice for most surgical decisions of this type and is the source of an awful lot of the variation that we see, namely medical opinion that has not been modified by an engaging conversation with the patient about that patient's own preferences.

So I've given you in a nutshell what my interpretation is of the causal factors and the practice variation that are most significant here.  And I will try to amplify on this a little bit as we go along here.

This diagram here simply shows you how much variation there is among the 306 regions when you compare, on the left, hip fracture, for which the market actually clears of, quote, "need," because everybody with hip fracture gets hospitalized in this country.  And so the incidence of hip fracture is actually the major determinant of the distribution of care.

You can't see it because the dots are too small, but there is one little place that is a real outlier on the low side.  That turns out to be Hawaii.  Hawaii has a very low rate of hip fractures.  The rest of the country is very homogeneous in that.

Knee replacements, hip replacements, and back surgery are increasingly more variable.  This reflects basically some of the underlying dynamics of the market for orthopedic surgery across the country.

Here is a snapshot of what is going on in four communities in south Florida.  Fort Myers on the left does excess amounts of knee replacement, about 48 percent more knee replacement than the U.S. average, 45 percent more hip replacements, and 67 percent more back surgery.  On the other hand, Miami is low in all of them.

And if you look at Fort Lauderdale, Fort Lauderdale is below the national average for knees, above it for hips, and at it for back.  This is exactly the kind of pattern that you see:  idiosyncratic patterns of surgical practice delineated by sharp boundaries as you move from one community to the other.

Fort Myers is contiguous to Miami.  You have to go through a bit of the Everglades to get there, but it's pretty close.  And Fort Lauderdale is right next to it.  So these patterns of practice are very idiosyncratic to the community, and that occurs around the country.

I was chatting with Leon before the talk here.  The economic implications are considerable because over a decade of time, Fort Myers generates about $125 million more revenue from these 3 procedures just for hospitalization than would be the case if they were practicing at the pattern of Miami, so big money is involved in this.

The interesting thing — and this is part of the general phenomenon — is that you might expect that the relationship between knee replacement and the supply of orthopedic surgeons would be positive because, after all, orthopedic surgeons do these kinds of things.

Well, it turns out that the correlation between supply of surgeons and the rates of specific procedures which they do is quite low consistently for all of the discretionary surgical procedures.

And the reason in my opinion that this is occurring — and I think there is more than just my opinion behind this — is that orthopedic surgeons, for example, have choices.  They could be doing sports medicine.  They could be doing carpal tunnel syndrome.  They could be doing other things.  They could be doing knees and hips.  And there is a comfort level that surgeons get with specific procedures, and they look for opportunities to do it.

Just to give you a specific study here, in the case of knee replacements, the Canadians became actually quite interested in what these rates were all about.  And so they actually did a morbidity survey in which they did a questionnaire to find all people with arthritis that was sufficiently severe to engage the patient about whether they should have surgery.  They did X-rays.  And they basically found that the number of people eligible for surgery exceeded by a factor of ten the actual rates that they were performing.

Now, the interesting thing is that after that determination of need— that was, clinical need — they actually interviewed the patients and asked their preferences.  Only 14 percent of that group actually wanted surgery.  So you see that the excess morbidity that is available countered against the actual question about preferences is where these kinds of — it's in that space that these kinds of problems are happening.

This is just to show you quickly that for all the procedures we look at, there is virtually no relationship between supply of surgeons — we're talking about knee replacements, hip replacements, back surgery, CABG, PCIs, prostate surgery for cancer, for BPH, gallbladders, carotid arthrectomies, and, finally, external lower leg bypass surgery.  That makes up basically 45 percent of the Medicare outlays for surgery.

None of these is related, but what they are related to is the rates over time.  And this shows you for the 306 regions along the horizontal axis the numbers of knee replacements that were performed in 1992 and '93 along the vertical axis, the number that were performed in the year 2000 and 2001.

Now, mind you, the line of equality says there's more surgery done everywhere, but there is very little regression to the mean.   In other words, whatever is going on in the community ten years ago predicts the risk of surgery today.  And that is a five-fold variation in risk.

So these are essentially embedded attributes of the local core of physicians.  In these regions, we're not talking about a lot of people doing these things.  We're talking five or six, sometimes seven or eight in these size communities, people who are actually doing these procedures.

So we find again for all these procedures consistently that what happened ten years ago predicts what's happening today, very stable patterns of practice.

Now, what do we know about the causal issues from the clinical trial level?  We know that if decision aids that are designed to help patients understand what is at stake in these operations and which clearly delineate the problems of values and preferences, that the use of these decision aids consistently leads to different outcomes than to control groups, which are getting usual practice.  That represents essentially a clinical test of the theory that doctor opinion is different than patient opinion and that patients decide differently when they're informed than they do when they delegate decision-making to the doctors.

A second very interesting aspect of that is that in almost all the clinical trials of decision aids — and there have been about 15 of them now — the rate in the control group is higher than in the intervention group.  In other words, there's a net drop in the surgical incidence associated with the introduction of information to patients.

That is exactly what we saw in one study we did in Group Health in Seattle and the Kaiser HMO in Denver.  We basically were able to measure the rates of surgery in these regions before and after the introduction of what we call shared decision-making or the decision aids, in a structured way.

The rates dropped 40 percent.  And when we benchmark these rates against the United States distribution, which is the purple on this particular slide, we saw that the rate of the benchmark under shared decision-making was at the very bottom of the distribution of surgery rates in the United States at that time.

The clear implication is that the amount of surgery informed patients want may be much less than the amount now provided in the country as a whole.

So that brings us up to the final question then:  How do we normatively understand and assess the implications of these variations?  And at the aggregate level, what we can say is because of the way the decision process is now structured based on the old model of delegating decision-making to physicians, we simply don't know what the right rate of surgery is because we don't have models of where patient in-true-demand or patient-informed decision-making has actually led to an empirical result except for that one Kaiser example.

We do know that, on average, in all of the clinical trials, patients who are actively involved in the decision-making with informed patient choice choose less than the control group.  So very likely in my opinion, we are on the downward side of this curve.  We're providing more care than our population would demand if they were informed, but I can't say that with any certainty.

Now, that's a quick run-through, Leon, of the story on the preference-sensitive.  And I think if you'd like to open it up for —

CHAIRMAN KASS:  Why don't we pause and have discussion of this before Jack goes on with the other things?

I have a question.  The benchmark here seems to be what the patients would choose if the patients had full information, full data presented to them. But if some skeptic were to say, "Patients talked to at some greater length will become leery of certain kinds of procedures and will opt out of having them, getting what they chose but going against their best interests," what would one say?

I mean, the true measure should be benefit to patients, which is not necessarily matched by giving them what they wanted.  What you want to give them is what would be effective for improving their health.

DR. WENNBERG:   Right.

CHAIRMAN KASS:  Are there data on this with the follow-ups?

DR. WENNBERG:   Yes, there is.  Let me make one point clear that there are clinical criteria of eligibility.  So this is not somebody who just happens to have Munchausen's or something and wants to come in and get involved.

So these are people who based on — well, let me give a specific example.  In the case of prostate, the choice of surgery versus watchful waiting for a benign prosthetic hyperplasia— that's an enlarged prostate, noncancerous — it turned out that what really was at stake was a trade-off between sexual function and urinary tract function.

The urinary tract function was better managed with surgery than drugs or at the time we did this study initially, there weren't very many drugs at all; whereas, the patients who had surgery almost to a person would have retrograde ejaculation, a problem which incidentally the surgeons thought was a normal outcome, was not a normal outcome at all for a number of men.  And when they were informed about that plus the probability, the relatively small probability, of incontinence and impotence, that became the big divide in terms of the decision process.

So it is, in fact, a more — if I had time to go into this, this is a very interesting set of questions about what are these trade-offs, how do we measure them, and how do we know that the decisions that aren't being made under this model of decision-making are better decisions than those that are made under the old model?

Our group has worked very hard at that, developing what we call decision quality measures.  In this particular case, it was to learn through external questions about how much their symptoms bothered them, how severe they were by a scale, and their concern about sexual function and then to use those measures essentially to predict choice.

It turned out that they did very well.  So people who were quite concerned about their symptoms and were not concerned about sexual function were five times more likely to choose surgery than those who were the reverse.  So that these were very — validation I think is at the level.

The other thing that these decision quality measures allow us to do is to quantify whether people in the control arm knew what they were doing.  And it turned out they didn't.  In other words, they did not have the natural history right, and they were not making decisions that were reflective of this underlying preference.

It is this sense of subjective trade-off that we're talking about here, which is very different than the classic clinical model, that there is an objective need and it has to be treated as binary.  Either you've got it or you don't.

The interesting thing is that only 20 percent of the patients who were severely symptomatic from urinary tract function under this model actually chose surgery.  It was quite similar to the result that we saw in the Canadian study, which I mentioned a few minutes ago, where only 14 percent of patients who were clinically qualified did it.

One of the other interesting clinical trials here was in Canada, where you hear a lot of argument that they are under-served in Canada for bypass surgery.  Well, it turned out that a study conducted on the waiting list for people with stable angina, not the severe blockage but the stable angina part, we saw a 22 percent reduction on the waiting list of people who were already lined up by their doctors for surgery.

So this preference issue is extremely important in terms of sorting out the variation in practices that we see.

CHAIRMAN KASS:  Thank you.

Paul McHugh?

DR. MCHUGH:   Thank you.

I think you have answered my question because I was just asking about the TURP (Transurethral Prostatectomy) for BPH (Benign Prostatic Hyperplasia) and how its association was down between 1992 and 2000.  Of course, that was the time when Flomax and Proscar and all of those drugs began to appear.

DR. WENNBERG:   Right.

DR. MCHUGH:   That could explain some of this—

DR. WENNBERG:   I think it definitely does.

DR. MCHUGH:   — the fact that there were these alternatives with BPH.  After all, sometimes you get to the point where, the patient gets to the point where whether he wants it or not, he's got to have it.  And those drugs have at least avoided that kind of problem for patients.  And isn't that —

DR. WENNBERG:   I think that's the right interpretation of that.  Thanks.

CHAIRMAN KASS:  Ben Carson?

DR.CARSON:  Thank you for that.

In terms of decision-making when more information is given and people are opting out of surgery, doesn't a lot of that depend on who is providing that information?  And also do you think it's important to distinguish the various types of surgery?

For instance, people with trigeminal neuralgia, an extremely painful condition of the face, very frequently do not get into the hands of a neurosurgeon.  And they suffer for years, sometimes for decades, when we have very simple ways of taking care of that problem.  And seizure surgery, of course, is the other big area.

DR. WENNBERG:   Clearly, the problem, from a systems perspective is to make sure that the patients who are in the primary care or other specialty care are not kept from referral on the basis of the doctor's opinion, as opposed to the patient's opinion, about what should be done.

So in the case that we are examining here of BPH, we're not sure that we didn't miss people who would have wanted it because their primary physicians never referred them.  So it's a two-way street.  And it basically says that if you're going to change the practice, you have to change it systematically for the referral process but also for the process of people who come directly to the specialist.  So it's an issue.

It sure matters who gives the data, I mean, the information.  That's the problem.  It's not that these patients don't get information.  Everybody gets informed consent.  But the problem is it's possible — and I did not go into this, but these decision aids are the result of, in the case of prostate work, a decade-long research project into what's at stake.

And that was part of the Agency for Health Care Policy and Research's original mandate, to basically study these decisions and to create rationality that made sense to patients from that work.

And the decision aids are carefully examined for bias.  Bias is something that is everywhere.  But the objectivity of the decision aids is basically a vetted process that goes through both patient-focused groups, objective questionnaires, and physician groups.

I can tell you one thing is that while patients tend to believe it's biased from the direction for which they have made the decision, the physicians believe it's biased in favor of the other specialty.  That's just a truism in this.  You figure it.  The nice thing is that you do this long enough, and everybody finally gets tired of the process and says, "Okay. It's okay."  So the long-term goal in my mind here is to improve upon the decision aids by iterative research, which learns better how to actually understand preferences of patients, their stability, and how to measure the fact that the decision that they ultimately made was coincident with their preferences.

DR. FOSTER:  Let me ask you this question.  I mean, I think it's very interesting to do all of these comparisons, but do you think that there are adequate controls?

Let's say that an informed patient decides against prostate surgery, as opposed to implanting seeds or whatever.  What is the evidence that the patient decision was the right decision?

I mean, they may say, "Well, I don't want this," but, in fact, most physicians will talk to other physicians.

DR. WENNBERG:   Right, right.

DR. FOSTER:  I mean, I can tell you very clearly if I have an illness, which I don't, I will go to a physician.  And I don't make the decision myself, no matter what he tells me.  I end up saying, "What do you think is the best thing for me?"

So my question is, the mere fact that a group of patients — I know you have thought about this as much as I have — say, "Well, we're going to opt not to do this," —

DR. WENNBERG:   Right.

DR. FOSTER:  — do we have strong evidence that their decision in terms of outcome was the proper one?  I mean, we talked about prostate.  John McConnell published a paper that got him into the Institute of Medicine about predicting when you were going to need intervention and so forth and so on.

I mean, we have those data, but my question is, do you have strong evidence that the fact that the patient decided this way after information and knowing about the complications was the right decision in terms of outcome for the future?

DR. WENNBERG:   Well, in the case that you raised, namely prostate cancer treatment, nobody has any good evidence on it.  So we wouldn't have any either.

It's important because you had said you're the patient.  You're going to go to the urologist, and he's going to decide what you get while if you went to the radiotherapist, you can guess what you would have gotten.  So simply by choosing the specialty, you have fated yourself for a decision.

I would beg to argue that the model that you have just put on the table is the old model.  That is how patients have behaved.  What I am trying to say now is that patients when they are in a care system that bothers to let them understand their options and uses validated decision aids, you get a very different decision.

Let me give you another example that might help here.  We have used the decision aids in clinical trials, of surgery, for back surgery versus watchful waiting, in which the objective of the video is to explain the treatment options but to also explain the scientific uncertainty, which is not usually explained.

You wouldn't go to a urologist probably and tell you, "Well, the clinical trials for this condition have never been completed, but I think this is what is going to happen."  You get a very different story than if you asked him "What should I do?" and he'll say, "Of course, you should have that procedure."

So it's, in fact, this cultural shift that is almost exhibited, quite interestingly, between our points of view on this is what we're about right now, learning how to communicate uncertainty, communicate values, and so forth.

And it does mean that patients have to begin to understand that.  Medicine is not written in stone.  We don't know the answers very often, as opposed to the lumpectomy situation, where the clinical trials have actually clarified that these two procedures have equivalent outcomes.

And I can tell you that you wouldn't want the gynecologist deciding whether your wife should have a mastectomy or a lumpectomy based on the fact that he always recommends mastectomies, which some of them do.

Do you follow me?

DR. FOSTER:  I think I do follow you.  Yes, I do.

DR. WENNBERG:   It's an exciting kind of concept, but it's not intuitive.  The one thing I can be quite clear about is that the evidence from the use of decision aids is that:  (a) patients made better decisions according to preferences; and (b) satisfaction is fine.  So people don't mind getting involved in uncertainty.

DR. FOSTER:  But I don't think you've answered my question.  I presume that most good physicians will change with the evidence.  You know, we gave estrogens for a long time, and now we don't do that or we understand that probably nobody should have back then.  You know, the evidence comes in.  I presume that that would be the case.

But the question I'm asking — and maybe you've answered this and I didn't get it — it's one thing to say that if you have a system — you call it the old system.  I mean, sometimes old systems are better than new systems, as you know.  I mean, many times they are.

But what I want to know is it's not whether they got a mastectomy or a lumpectomy.  You can get the outcome then.  I want to know about the change of decision.

Look, I'm an endocrinologist.  Let's say that if a person comes in and has got hyperthyroidism, one thing I can say is, "Okay.  You can treat it with radioactive iodine.  You're going to get this.  You get surgery.  You're going to do this.  You know, if you want to take long-term propylyuricil. All of those things will make you well.  And there are risks and so forth with each one of them."  And so let's say that the patient then makes a decision that on the basis of the facts that they learn that they are going to choose to do surgery.

My question is, do we have firm evidence that that decision conforms to outcome in a higher percentage than if you use the old system?  I guess that's what I'm saying.

My guess is that if you got sick and you had something that was not a streptococcal sore throat or something like that, that you would probably ask around at Dartmouth and so forth.  And what you would say is, "Who is the best person to treat me for this?"

And that person may tell you, and you may decide to do one.  He may give you options.  But don't you —

DR. WENNBERG:   Let me try —

DR. FOSTER:  The question is, I guess I'm saying, is, do we have data that shows that the patient decision when informed has a better outcome than the decision that was made by a physician, what you're calling physician preference here?  I'm trying to separate out the economic reasons for doing a lot of surgery or something.

DR. WENNBERG:   Let me say that we have to keep the categories straight.  So, in other words, if there is something that works and everybody knows it works, we're not talking about a preference-sensitive problem.  What we're talking about is, the major categories that this applies to is discretionary surgery, where you could do this or you could do that.

Sometimes the outcomes are fairly well-known.  And sometimes they aren't.  The ultimate goal that we have is actually in cases, like back surgery where there were no clinical trials, to actually conduct them.

And we conducted the clinical trial by using the decision aid.  And we argued forcefully I think with our colleagues that in this case, the ethical criteria for entering the trial should be patient equipoise and not doctor equipoise because we couldn't find very many surgeons who were at equipoise about the value of back surgery.

So what this allowed us to do in 11 center sites was to actually get over 900 people enrolled in the clinical trial.  And we're also following up those who had strong preferences.  So we're able to look at the internal and external validity problem.

So when the probabilities are not known but the treatment is optional, then the long-term ethical, in my opinion the ethically correct way, is to learn by works, not just to keep doing things when we don't know the answer.

So there's this distinction between understanding what the outcomes are given the choice of treatment — sometimes we know, sometimes we don't — and then the question about the value to the patient of choosing one way or the other.

Go back to the prostate situation.  We have actually done this a number of times.  We have informed men about the clinical uncertainties associated with the treatment.  And some of them will say, "I want that surgery because I think the evidence, shaky as it is, is better than nothing."

Others will say, "That evidence is shaky, and I don't want to risk the incontinence and impotence," which we do know a lot about.  And you find people divide on that.  They really do.

So if I can try to make sure that my point isn't getting lost here, the question of preferences has to do with the choices between the options given the knowledge we have.  The question of outcomes has to do with how good are the studies that have established our information base, of which we're informing patients.

In many cases here, it's shaky.  Prostate cancer is a great example.  And we do need to do these studies.

CHAIRMAN KASS:  There are a couple of people in the queue, but just if I could intervene here and see if I can clarify for myself.  In a way, Dan reasked the question I asked also.

I take it that the point of departure is this very wide variation, say, between Miami and Fort Myers, a much wider variation for knee replacement, hip replacement, and back surgery, which are relatively optional compared to hip pinning for a fracture, where the variation is much smaller.

I suppose, I mean, one possibility is to say the people in Miami are not having their backs done sufficiently.  They're being under-served and that it might be a mistake to talk people in Fort Myers out of doing that by giving them more information so they choose like Miamians.

But I think the presumption is that somewhere that there is this tremendous variation means that there might very likely be a certain amount of excess surgery.  I think that's some of the starting assumption.

And then the question is — and you don't even have to attribute bad motives.  It might be certain sort of standard ways of proceeding in various hospitals or in various regions, where people do it the way they were trained.

Then the question is, what happens if you lay out all of the details where the cost-benefit calculus of the patient is somewhat up in the air?  I take it the point is in those cases, the argument is that the patients are probably as well able to make that kind of calculated decision where, really, you're dealing with optional surgery, rather than with, say, life-saving surgery.  Is that —

DR. WENNBERG:   Yes.  I mean, this would not apply to hip fracture.

CHAIRMAN KASS:  So I take it it was a kind of limited scenario where Jack was suggesting that this kind of information — where what you're getting is a kind of trade-off of certain kinds of benefits at certain kinds of symptoms, where that reduces the incidence of surgery, where you suspect that there is a certain kind of excess, might not be a cause of concern.

Now, I don't know that that is reassuring or whether that is a correct understanding of the problem.

DR. WENNBERG:   I think ultimately the ethical question to me has been it's not a good idea to do surgery on people who don't want it.  And we have never been very good at finding out what people want.

That's the story here.  I don't want to give a lumpectomy to a woman who wants a mastectomy just because I have been trained that lumpectomies are the wave of the future.  And that's because there is heterogeneity among people.  We all know that.  We're not all the same.

Medicine is fraught with decisions for which there are options for which information is not complete.  So sharing uncertainty with patients is part of the ethical requirement for the discourse that we're after.  An obligation on top of that is to follow up with patients to find out what happened, preferably with randomization if you can do it, but if you can't, just do it on the basis of the cohorts.

Just an example on this very study we're talking about here is that one of the big uncertainties about prostate surgery when we were engaged in this research was, what is the incidence of acute retention if you don't have surgery.

The urologists were all over the place in terms of what they thought it was.  Some thought it was about one percent.  Some thought it was ten percent in a year.  Well, by having a whole bunch of people who had significant prostate disease that we were following, we answered that question within about two years very well.  And it was quite low, actually.

So if it had been ten percent, then the whole decision process would have been very different because you would have to face that gradient.

CHAIRMAN KASS:  Let's have a couple of questions.  Ben Carson and Gil, and then we will let you do the second part.

DR.CARSON:  Just sticking with the whole prostate situation, you know, in the 1980s, Pat Walsh worked out a wonderful nerve-sparing prostatectomy procedure.  And a number of other people have adopted that.  And there by the mid '90s were a handful of people who had spectacular results with prostatectomy.  And that number has increased exponentially since that time.

Is that kind of information utilized in this new information system?  A lot of people think that a surgeon is a surgeon like a plumber is a plumber, and it's not true.

DR. WENNBERG:   Yes.  There are two questions in a sense that we are being asked there.  I think one was whether or not the specific surgeon that was about to do the procedure was giving his statistics or her statistics, rather than Pat Walsh's.  That's a big problem.

There's also a question about what Pat Walsh's statistics really are, too, because systematic follow-up of — not Pat.  He likes to do this, but he also categorically does not operate on anybody over 60, 65 if you really push him.  And that's not what is happening in most practices.  So the question about how to get surgeon-specific information into these kinds of databases is complicated and difficult.

The other question is about whether there has been innovation.  The obligation of doing decision aids is frequent updates because new information comes in all the time.  Sometimes it's positive.  Sometimes it's not.  Actually, when Vioxx went off the market, it upset one of the decision aids on osteoarthritis.  And it had to be dealt with immediately.

So I don't know if that answers your question, but it is a complicated one, particularly when one gets into the question about operative mortality variations, too, and how they become communicated to patients.  The statistics are difficult because of the incidents, even among a not-so-good surgeon, and still pretty low, thank God.

CHAIRMAN KASS:  Gil Meilaender?

PROF. MEILAENDER:  Ben, a plumber is not a plumber.  There are distinct differences. (Laughter.) Is there a way to sort of factor into the thinking that you're doing about this or is it even important to factor in the possibility that it's not just what you a while back, I think answering Dan, talked about as a culture shift but there might be temperamental differences?

In other words, maybe I really want all of the information.  And maybe the person next to me experiences a greater level of contentment with their care if they just say, "What would you do, Doctor?" and the doctor says something.

CHAIRMAN KASS:  Right.

PROF. MEILAENDER:  How is that difference, which isn't strictly — I don't think that's simply changing culturally.  There still are and always will be people who react differently.  How does that make a difference in how you think about these things?

DR. WENNBERG:   That's a very good question.  Quantitatively, we find in most of these clinical trials about ten percent of people fit that category after they have looked.  They still don't know what they want.

And in this case, we have taken a fairly simple rule of thumb that if the patient doesn't know what he wants, do the less invasive thing because some day he might find out what he wants and he will come back to you.

So that it's basically very pragmatic.  If the patient doesn't have a choice for surgery, I wouldn't do it.  So it does have a simple pragmatic answer.

CHAIRMAN KASS:  Why don't we go ahead, Jack, into the second part?

DR. WENNBERG:   Sure.  Well, having gotten you through all of that difficulty, wait until you see the next part here.  This is basically now dealing with supply-sensitive care.  And the distinction between preference-sensitive care and supply-sensitive care is partly because of the role of supply itself, as I will show you, but also because in preference-sensitive care, the treatment options are fully articulated in terms of medical theory and sometimes evidence.  Surgeons don't operate without a good reason for it, even though the reason may not be ultimately correct.

But in the case of supply-sensitive care, what we are talking about is essentially the frequency with which chronically ill people visit physicians, not surgeons but medical specialists and general internists and family practitioners, how often people with chronic illness are hospitalized, how frequently they're put in intensive care units, and how often they're monitored with diagnostic testing and imaging.

For the concept of how frequently should I revisit or schedule, our patient with moderate congestive heart failure, there are basically no clinical guidelines at all about this.

It's never entered the discourse, clinical discourse.  You don't go to clinical meetings and see physicians arguing vehemently about every three months, every five months, every six months.  In fact, it turns out that interval shifts from six months to three months accommodate a doubling of the supply of physicians because it takes that much more time to see those patients.

And so it is the question here of the variation in the intensity with which chronically ill people are treated.  And the reason we are focusing on chronic illness is because that is where most of the visits, hospitalizations for medical hospitalizations go, not surgery but go for the Medicare population.

So what we see is huge variation in the number of primary physician visits, particularly medical specialist visits.  They show about a five-fold difference between the regions, five-fold differences in congestive heart failure and COPD discharges.  So that's what we're looking at here in terms of the basic problem.

When we correlate the hospitalizations for medical conditions, the green dots on the slide, we see a strong relationship between hospitalization and the number of acute care beds in a region.  And that's not surprising because acute care beds tend to be fully occupied, maybe not so much anymore.  Maybe that's why we only explain half of the variation, but that's an awful lot of it; whereas, hip fractures, which we have already discussed, are essentially where the hospitalization rate is pretty much under-determined by the incidence rate.  We don't see any relationship between bed supply and admissions or hospitalization for hip fracture.

I might remind you that under the classic concept that utilization was regulated by scientific information and effectiveness criteria, the hip fracture would be what you would expect.  But, in fact, hip fractures — there are only about three other conditions that we have looked at that even come close to the hip fracture.  One of them is colectomy for people with cancer of the colon.  Most people who have cancer of the colon will ultimately get a colectomy, even if it's late.

And it used to be acute heart attacks would follow this pattern, but recently there have been so many redefinitions of what an AMI is in terms of the enzymatic definitions that you see increasing variation in the use of hospitals for AMI also, the same kind of thing we see for the physician visit rate.

So here is the number of cardiologists per capita ranging from 2.5 to 12.5 or about 5- or 6-fold difference.  And that's the visit rate to cardiologists, not surprising because cardiologists in large part see people who are over 65.  It's not as if we were pediatricians with this.  So that is kind of to be expected.

Again, this is quite consistent with the scenario when you have more physicians you get more visits or, in fact, what is happening is that the revisit rates are being narrowed as you get more, there being a limit to the number of people with cardiological problems.

Now, the kind of obvious question here is not so much that it varies but, rather, what's the value of the additional increment in utilization?  That's the question that underlies this question about whether we are rationing care in the low-rate regions, or are we overtreating or is there excess capacity in the high-bed side?  So that's the kind of question that comes out of that.

We have been pursuing this problem now for a long time.  And consistently we have found basically no relationship between health outcomes that we could measure at the population level and the utilization of hospitals and doctors.  In other words, we see in Miami twice as many hospitalizations per capita for medical conditions as we see in Minneapolis or we see in Los Angeles.

The study that I am going to talk now briefly about essentially took the area dimension down to the cohort dimension, instead of asking the question about general survival in the region among Medicare patients.  What about people who had had a heart attack, cancer of the colon treated surgically, and patients who had had hip fractures?

These were the kind of conditions which we could have some confidence were being treated similarly between regions.  Particularly in the case of heart attacks, we were able to get some clinical evidence that allowed us to sort out these newer categories of AMIs that are enzymatically determined.

Then to follow these people over long periods of time, five years, and ask the question if they lived in regions that had high-intensity care, did they do better than the ones living in the low-intensity care regions?  So it was an aggregate study across the country.

What I can show you on this slide is that comparing the highest quintile to the lowest quintile, we saw about 60 percent more Medicare spending, 32 percent more hospital beds, lots more medical specialists and surgeons and so forth, and correspondingly high utilization rates.

But what we didn't see was any improvement in outcomes.  In fact, when we compared the relative risk of death over that five-year period, in each case, the mortality rates were higher in the more intensive regions.  Heart attacks, colon cancer, and hip fractures all had significantly higher mortality rates in the high-rate regions than they did in the low-rate regions.

In addition, when we looked at other data sources that Medicare provides in terms of functional status satisfaction, we basically saw no difference, which is coming at least in the directions about quality of life.

Finally, fascinatingly enough, access was actually worse in the high-rate regions for such things as preventive care and beta blockers.  The reason I think that may be happening is that in the high-rate regions, many more doctors are involved in the treatment of patients.  And the continuity of care question is really very questionable there.  I'll come back later and talk about that.

By our understanding of the available data, we can say that at best, we're at flat on the curve, which means we're not getting anything out of this incremental investment or, as the mortality evidence suggests, we're actually on the downward slope.  So we're losing, as opposed to gaining.

Now, you could ask me, how in the world could that ever happen?  Well, if you pay attention to the medical error literature, where the claim is that about one percent of Medicare population experiences an iatrogenic death for each hospitalization that occurs; in other words, one out of 100 hospitalizations is associated with that, if you do twice as many hospitalizations, you're going to have twice as many people on a per capita basis who are actually dying in a region from the iatrogenic illness or medical illness.  And that, in effect, is quite consistent with what we're seeing in the data.

The importance of this last category is — I can quickly illustrate here by the green, is the low-spending quartile, and the red is the high spending.

When you increase spending, you don't get more effective care.  Ironically, you don't get more preference-sensitive care.  In other words, the likelihood of having bypass surgery in Minneapolis is just as high as it is of having it in the high-cost regions.  But what you get is a lot more hospitalizations, medical specialist visits.  And this percent, seeing ten or more doctors, is our measure of continuity or lack of continuity.

So that's the story at the regional level as we look at it across the United States.  The question is — I don't know whether, Leon, you want to take questions now, but I do want to go over to look at what is actually happening in some of our academic medical centers.

CHAIRMAN KASS:  Why don't you finish up, and we'll hold the questions to the end.

DR. WENNBERG:   Okay.  Because I'm going to be dealing with the same problem, namely the supply-sensitive care, rather than the preference-sensitive care.

In the last few years now, we have been working to change or to add the population-based analysis to the experiences of specific hospitals.  And that's possible because when people become chronically ill, they do not change providers very often.  So if you follow back from death for two years and look at the providers that were used during that time and assign the patient to the hospital that they most frequently used, you end up with loyalty of 80 to 90 percent of the care being actually delivered at that place.

So we have been able through this process to quantify the amount of care provided in managing chronic illness across these hospitals.  We have concentrated on end-of-life care.  There were a couple of reasons.  One reason is that I think people do not resist our argument that when you're about to die, you're pretty sick, no matter where you are.

So measuring variation in end-of-life care is a way of implicitly controlling for illness differences, particularly if one goes to the trouble of case mix adjustment because some cancer, for example, has lower utilization rates than congestive heart failure.  But by adjusting for that, you can take care of that problem.

This is the difference that we see among what we call the 77 best U.S. hospitals.  This was slightly tongue in cheek.  This was done in that British Medical Journal article.  We took the ones from the U.S. News and World Report list of the best hospitals in the United States for geriatric care and for chronic disease management basically just to see how different they actually were.

What you see here is the average number of days that patients will spend in hospitals ranges from around 9 days at the low end of the distribution among these hospitals up to almost 28 days in the high end of the distribution.  Remember, that's the average number of days people in these cohorts with chronic disease are spending in these institutions.

If you put labels on them, you can see some fairly familiar places.  Patients who are loyal to NYU spend almost a month in the hospital, patients loyal to Stanford about ten days.  UCSF is about 11.5 and UCLA 16.1.  You actually do see these extraordinary differences within the University of California's system.  I just showed two of them here.

What you then go on to see is this interesting association here, namely that if you're high for one disease, you're high for all diseases.  This (slide) happens to compare congestive heart failure patients to cancer patients.  And if you're high for cancer patients, you're likely to be high for congestive heart failure patients.

The black line is equality.  So cancer patients get less care on average than congestive heart failure patients.  But what really matters about how much care you get is not the disease you've got but the place where you're given care.

This is also true for black-white comparisons.  Blacks get slightly more care in the last six months of life than non-blacks, but what really determines how much you get is where you're going.  And that turns out to be true for all the conditions we looked at and for all of the different measures of utilization that we used.

Here's the number of physician visits in the last six months of life.  NYU is using 76, and Stanford is using 22, almost 3 times as much — actually, more than 3 times as much.  Cedars-Sinai in San Francisco is 66 visits on average per person, UCLA 43.9, and UCSF 27.2.

Days in the hospital are related to physician visits.  If you are in the hospital, you are much easier to visit.  You get lots of referrals.  So lots of time in the hospital equals lots of physician visits.

Intensive care units, let me just go back to this.  The same study that I showed you before at the regional level for all hospitals was repeated to show the same thing happens among academic medical centers, namely academic medical centers located in regions with greater intensity of care, actually do not have better outcomes.  They have slightly worse outcomes than those in the low-rate regions.  So it's a consistent finding.

And this basically is a day spent in intensive care.  Notice the differences between the University of California in San Francisco, where you have 2.6 days on average and UCLA, more than a week in intensive care on average for patients in the last 6 months of life.

Deaths in hospital associated with admissions to ICU, 20 percent of Mt. Sinai Hospital and 36 percent at Cedars-Sinai.  These issues have a lot to say about the quality of dying, for sure.  I'm not dwelling on that at this time, but these are real people.  And 36 percent of people at Cedars who happen to use that system are going to end up on average dying in an intensive care unit, compared to 20 at Mt. Sinai.  And there are other hospitals who are even lower.  I just didn't have them named here.

Now, the final kind of thing I want to talk about here is that this gets a little bit back to what some of the options might be for actually doing something about this.  In addition to measuring utilization, which I have just gone through, we also are able to measure resource allocation.  So how many doctors were used?  What type of doctors were used in treating these populations?  And also how much money was spent in treating them?

I just want to briefly show you that this is the number of primary care plus medical specialists whose labor was allocated.  I won't have time to go into the details of this, but these are standardized full-time equivalent input.  So you can compare one place to the other in terms of the actual labor inputs.  It's based essentially on the RVUs that were allocated to the different places.

So NYU is using 24.6 physicians, whereas, Stanford is using 8.7 for treating essentially the same patients over time.  UCSF falls where it falls.  Cedars-Sinai Medical Center in Los Angeles is again quite high.

Here are the differences between medical specialists and primary care.  The Cedars system and the UCLA system are highly dependent on specialist care.  So if you divide specialist by primary care physicians, you get 2.8, a ratio of 2.8, at UCLA.  If you do the same thing for UCSF, you get 0.67.  In other words, UCSF is using more primary physicians than is UCLA by quite a bit.

Finally, costs.  Payments along the vertical line shows you how much money Medicare spent on these patients in the last six months of life, ranging from around $11,000 upwards of $35,000.

Along the horizontal axis is how much money was spent for the same patients in the 18 to 24 months prior to death.  Of course, it's a lot less.  It ranges from 2 and a half to 2 to around 7 and a half, still a 3-fold variation or a 2 and a half-fold variation, but notice that they are highly correlated.

So, in other words, what we're seeing here is not just an end-of-life phenomenon but, rather, a consistent difference in the threshold for treating chronically ill people across time, which could have been looked at in terms of visits and any of the other parameters I have shown you.

I've chosen this one because of another point that I want to make.  And that is, when you look within a community, like San Francisco or Los Angeles, there is a great deal of variation in total payments and utilization among the hospitals actually located in that region.  Some of them are spending well below the average of the region and some are spending well above the average of the region.

And that introduces some very interesting new information and debate for the so-called pay-for-performance strategies that are unfolding around the United States right now because it means that payers who are interested in locating efficient hospitals with regard to chronic illness can actually locate them through this data set.  And opportunities for selective contracting for otherwise rewarding patients or rewarding physicians who are more efficient could be devised.

I don't have a prediction about what will happen, but we are going to release this information nationally sometime over the next few months, starting with California, where we have already made connections with both the academic medical centers, some of them with the quality improvement organizations and with a business group.  We're trying to work out ways that this information could begin to filter back to see if we can affect change in California.

So I think that kind of gets it.  I just wanted to say we find widespread underuse of effective care, things that doctors believe work and do work because of clinical trials.

We call it misuse of preference-sensitive care because it should be based on informed patient choice, but generally it's not.  And then we say an overuse of supply-sensitive care, particularly in the management of chronic illness, based largely on the fact that we find no marginal value in the increased use of care.

In terms of the things that need to be done, we need to work on the underuse of effective care.  And that's actually happening fairly well around the country.  People are willing to pay people to do more.  But we need to learn what works.

These questions that were raised here about what the actual outcomes are here, we need to assure informed patient choice.  We need to achieve efficient and effective management of supply-sensitive care, namely chronic illnesses.

Finally, — this is a tough one — we have to achieve efficient allocation of resources geared to the size of the population served.  The measures that I have just discussed allow you to know where those hospitals are.  The question about how to downsize the excessive investment in acute care in this country, of course, is another story.

CHAIRMAN KASS:  Thank you very much.

The floor is open for discussion.  Robby?

PROF. GEORGE:  If no one else is jumping in, if I could take Dr. Wennberg back to the preference-sensitive care issue?  I just wanted to clarify for my own sake something in the exchange between you and Dr. Foster.

Is it a kind of presupposition of the profession that there will always be a right answer to these care choice questions, although it's often difficult to know what the right answer is?  And then you have the question, well, since it's difficult to know what the right answer is, should the doctor make the call or should the patient and the patient's family make the call or is it the case that in some cases, what the patient and doctor are faced with is not a choice where there is a right answer but there is a choice where there is no right answer? There are incommensurable pros and cons here.

The case that Dr. Wennberg cited about the patient before the development of better-quality drugs facing the question in the case of benign prostate disease of a TURP outcome where there would be diminished sexual function, on the one hand, or an outcome in which there would be diminished or problematic urinary function, that sounded like to me — I may just be misunderstanding — that case sounded to me like an example of a case where there is not really a right or wrong answer.  There are incommensurable advantages and disadvantages to the two.

So that since it is not a question of trying to get the answer right in a situation where it is difficult to know what the right answer is but, rather, just with a choice that doesn't emit a right answer in a case like that, would there be any argument for the doctor making the decision, rather than the patient?

It would sound to me as though there would not be any argument for the doctor making the decision, that that is just a pure decision that is left for the patient as to which of those two kinds of outcomes he would prefer to live with, either Dr. Foster or Dr. Wennberg.  I just would like to have that clarified.

DR. WENNBERG:   What you just defined is what the preference-sensitive care category is all about, basically.  Options for which — and I think I am going to challenge you on one thing.

There is a right answer, but it's not in the doctor's head.  It's in the patient's head because the patient has to sort out his values with regard to sexual function and urinary tract function.  You only can learn these by asking questions about that.

PROF. MEILAENDER:  But that's a values-determined —

DR. WENNBERG:   Exactly.

PROF. MEILAENDER:  — conception of right outcome, as opposed to —

DR. WENNBERG:   A probability —

PROF. MEILAENDER:  — the medical.

DR. WENNBERG:   — because the probabilities in this case were pretty well understood.  You got a lot better improvement in urinary function, but you also had a high probability of retrograde ejaculation and the other problems.  So that was not a decision where there was a scientific uncertainty.  It was a question about the values.

The value side in our formulation of informed patient choice, as opposed to informed consent, belong to the patient as an active participant in the decision process.

PROF. MEILAENDER:  So we're really talking here about two separate cases, one where there is scientific uncertainty.  And there would be at least some argument in that situation for just letting the patient make the call, although others would say let the doctor make the call, and then other cases where it's not a question of scientific uncertainty.  It's sorting out the patients' preferences, strictly speaking, values or desires.  Am I right?

DR. WENNBERG:   Can I come back to you?

PROF. MEILAENDER:  Yes, please.  I'd like to —

DR. WENNBERG:   Essentially what the surgery for cancer, not for BPH —

PROF. MEILAENDER:  I understand, but the cancer, it would be a different case.

DR. WENNBERG:   Well, basically the way we frame that question in terms of our understanding of what in the subjective mind of the patient is at stake is at follows.  Here is a set of evidence about the range in which the clinical trials if they were done would show benefit.

The most you can get out of this is a 10 percent reduction or a 15 percent reduction, whatever it happens to be, improvement in life expectancy.  Against that, you've got to basically have a very strong probability of incontinence and impotence, even despite Dr. Walsh.  It's still a big problem around the country.

So the question here is essentially the patient has to make a bet between a wager that the clinical trials if they came out would show enough improvement so it was worth it to him.  And some people would just say, "Okay.  I'm going to do that because I want to.  I don't want to risk this thing.  And my dad died of prostate cancer."  That, by the way, is a big deal.  So they go that route.

And other people will say — and we see it all the time — "It's just not worth it for me given my current situation, my current age, and all of that kind of thing."

The thing I would most like to do is I would like to convert that opportunity into clinical trials so we actually can begin to resolve these problems.

Where there is such strong opinions on different sides, the urologists, the radiologists, and the general internists, particularly in the U.K., are completely against this thing.  In this country, they have become a little bit more persuaded.  But it's a real deal.

CHAIRMAN KASS:  Peter?

DR. LAWLER:  This is a scenario about which I know nothing, but this bothers me what you said.  Let's say I'm old, which is far-fetched, and I have a chronic disease that will kill me, like congestive heart failure.

I would think I should move to a research university town, where I get superior medical care, but this turns out not to be the case.  I would just be in the hospital prodded by specialists a lot, go through a certain kind of medical torture that wouldn't produce a better outcome.

And the reason for this, if I understand your paper correctly, is because in an area I am now living, there are lots of hospital beds and lots of specialists.

So I should move to a more rural area, where there aren't so many hospital beds, aren't so many specialists, where they wouldn't want me to be in the hospital much.  I would just see the same old boring doctor all along, have the continuity of care, and die about the same time.  Is that about right?

DR. WENNBERG:   You might not have to move.  Let's say you're in Los Angeles.  The variation that we saw in these slides is not quite replicated, but it's definitely in the same trend within L.A.

So if you really did not want to get involved in this business, (a) the quality measure that shows that if you go to this place, you would at least get your diabetic eye exams because we can measure those at this level.  It's good quality from that perspective.

It's all over the park with surgery, but you're not going to do that.  You're going to worry about congestive heart failure.  Then you as an informed patient would say, "Doctor, I want to go to that hospital, and I want to seek care there."  So you don't have to move.

DR. LAWLER:  You have kind of an iron law here that the beds will be filled, the doctors will be used.  And so what is the remedy to this?  Fewer beds?  Fewer doctors?

DR. WENNBERG:   Well, achieve an efficient allocation of resources is what I have said here.

DR. LAWLER:  But be a little bit more specific perhaps.

DR. WENNBERG:   Well, we definitely have in my opinion an excess capacity in the acute sector for managing chronic illness.  That includes the whole group.

Now, if you want to get into the what do you do about this, you as a patient do have some options.  Whether we as a society have any options in terms of removing excess capacity is a $64,000 question right now.  It's actually the $64 billion question.

And the immediate complexity is that it is tied, for one thing, to the labor markets.  There are lots of people who are employed and many more in Los Angeles on a per capita basis than in San Francisco in this industry, many of whom are poor and are needing jobs.

There is a question about the equity market.  A lot of the L.A. hospitals are proprietary hospitals, Tenet hospitals, for example.  And then there is the bond market, like our friends at UCLA have got huge bonds out there.  So disrupting this in any way that a market would disrupt it has some very interesting implications.

My hope would be given this — well, put it this way.  One possibility, given this, would be that there would be a gradual shift of patients towards more efficient hospitals, either through people like yourself, who make wise choices, shall we say, but also by employers who steer people to them and by this Medicare choice, which is this capitated system, which would love to have this information because now they know where to get their networks designed.

So there is a system within this pay-for-performance strategy that if they were extended to include this level of service, namely supply-sensitive service, you could begin to get some movement.

The loser is going to be Medicare, traditional Medicare, because they don't do anything about steering patients.  And patients who stay at hospitals that are now inefficient while others are leaving it will just get more care because the care will be picked up.  And their frequency of those interventions will increase.  Yes, that is the problem.

CHAIRMAN KASS:  Ben Carson?

DR.CARSON:  Yes.  In many of the major medical centers, we were looking at the data, showing that length of stay in the major medical centers and in the regional hospitals might have been approximately the same, the outcome might have been the same.

Does that take into consideration the fact that in many of those major medical centers, there are people who attract the most complex cases that there are in the country and many of the community hospitals and regional hospitals are not dealing with those kinds of patients?

DR. WENNBERG:   Of course, that's not what I'm talking about here.  You would have to say that NYU attracts five times more sick people than Stanford does or UCSF.  So, in other words, we're looking here at similar kinds of hospitals.

The interesting thing is that within a market, obviously the academic medical centers are more expensive generally speaking.  So in San Francisco, you might think UCSF was the most expensive hospital by our measure.  It's not.  It ranks about fourth.  So there are others that are more expensive.  And the ones on the other side tend to be fairly large community hospitals.  So there is a difference there.

And if we're going to subsidize academic medical centers to be different because we want them to be different, then that is not what we are doing now.  We are subsidizing them.  We are paying them because they have — UCLA has a smaller share of the L.A. market relative to its capacity than does UCSF in San Francisco.  And that's the key behind this problem.

There is no question that academic medical centers generally are more expensive, even if they have the same capacity or the same utilization rights because their prices are higher because they're getting indirect medical education and they're getting — well, their disproportionate share in Medicare is about the same, actually.  But yes.

CHAIRMAN KASS:  Paul McHugh?

DR. MCHUGH:   I very much enjoyed this because it has so many different implications, but perhaps the problem for me — and maybe I just don't have it yet, Doctor — is that we're talking about a multi-variable schema of improving the services in care to our patients.  We're talking about eliminating underuse, informing, making sure the patients are better informed, and then effectively measuring and managing our supplies.  Okay?

All three of them are in play here.  In each one of them, no one was going to have any argument that each one of them is good.  I mean, we certainly want to underuse effective care.

A condition that I know very well is cystic fibrosis, a chronic condition and amongst children a genetic condition.  And it has an ultimate shortening of life.  But the length of life that a person has with cystic fibrosis very much depends on what doctor and what center is taking care of them because the centers that do best with these cases are the ones that look after every little nuance of the breathing.  I'm not telling you anything that you don't know, but maybe the others don't know it.

But if you look after every evening's sleep for these little children and the atmosphere that they breathe and the moisture of the breath, they live for many years more than if somebody is just dealing with the acute problems that turn up with cystic fibrosis.

And so that is a matter of effective care.  And we should be teaching that effective care to everyone who ever takes care of a patient.

Informed patient choice.  I mean, that is fundamentally emphasizing the fact that the patients aren't invalids but are consumers.  And consumers have meaning, as you say.  And then, finally, there is this management.

But, look, when you have a multi-variable thing that is going to go into outcome— I think this comes back to really what Dan is saying — which one of these things should we work on first to get the biggest bang for the buck as quick as we can because, after all, we have got an expensive health care system and we really want to take better care of patients every day?

DR. WENNBERG:   Yes.  From the point of view of the way the data looked to me, if you want to save money, go after the supply-sensitive.  That is what is correlated with Medicare spending.

We have a twofold difference between Miami and Portland, Oregon and Minneapolis and places like that.  It's all supply-sensitive care.

But I say that, and I am going to have to emphasize the difficulty of doing anything about it because it is going to take a lot of societal will to do this.

And I think behind this whole facade here, I mean, behind this whole story, is the fundamental belief that Americans have that more is better.  And we're saying it's not better.  You could give it up, and you could save a lot of money and you could reinvest it in something else, like infrastructure for cystic fibrosis care in the community.  That is where we want to go on this.

DR. MCHUGH:   Well, you know, I am not all that interested necessarily in saving money on this matter, although it is good to save money.  I really want to take better care of patients so that patients do better and that patients who come to doctors ultimately, whatever happens, that they do better.  Isn't that the effective care area?  And shouldn't we be working on that perhaps first?

DR. WENNBERG:   Well, yes and no because the problem is it just depends on our concept of how much more inflation and how much more costs we can bear.  There is evidence of marginal harm.  So we would want to try to deal with that.

I think the thing that is missing in here is the active engagement of academic medicine in trying to help us understand the scientific basis of managing chronic illness.  We just do not have any research projects.  The NIH leaves this completely alone.  And you can see what it is like.

These places should be embarrassed by their differences, but they are not even aware of them.  And we should try to begin to put pressure on academia to learn how to efficiently manage chronic illness.  If we did that, they might begin to respond not only to the specific question about what this theory A does with theory B but also with the whole question about the efficient allocation of resources.

The other thing that bothers me — and I should have mentioned this — is I really don't understand how in light of this kind of information here, the number of doctors being used by these places — the academic medical centers are out there agitating for increasing the supply of physicians right now.  You know, what in the world are they thinking about?  They don't know how many they should use, and they don't know the mix between primary care and medical specialists.

So trying to get the attention of people to these kinds of statistics in itself might enliven the debate a bit, like it has around this table today, by the way.

CHAIRMAN KASS:  Let me have the last, if I might, just very briefly, to try to tie this into some of the concerns that Joanne Lynn raised for us in the last session about end-of-life care.

Really quite striking disparities in the days in hospital and physician visits in the last six months of life, as you showed us here.  Let's assume for the moment that Joanne is right and that you stretch back also since this correlates with the way these people have been treated also for 12 to 18 or 18 to 24 months prior to that.

DR. WENNBERG:   You can go further than that.

CHAIRMAN KASS:  Right.  Let's assume that she's right and that looking forward, it's the long-term care, rather than the acute medical interventions, that make the most difference here.

Is there any way to take advantage of these particular discrepancies and the knowledge of them to begin to fiddle with the incentives?  I know you have indicated what some of the difficulties are.  It's the hospital beds that's driving this.  There are bond issues, and there are city employees whose jobs are at risk if one changes this.  It's a very large obstacle to doing something.

Have you thought really about — apart from getting more data and having academic medical centers sort of take this seriously, is there something that the people who now pay for all of this can do —

DR. WENNBERG:   Right.

CHAIRMAN KASS:  — to begin to shift the incentives in such a way that the services most needful are most available, rather than the suppliers getting their supply houses filled?

DR. WENNBERG:   Yes.  We thought a lot about that.  I have been working for several years now, goes back almost eight years, trying to get CMS to design a demonstration project to allow interested, quality places to experiment with the reimbursement system so they can deal with all of these categories at once.

And we have a consortium that includes Intermountain Health in Utah, Mayo Clinic, Marchfield Clinic, the Hitchcock Clinic, and, interestingly enough, interest on the part of Mass General.

We went to the trouble of getting Senator Jeffords to actually sponsor legislation that is in the '03 MMA section 646, which asked CMS to do a demonstration project.  And we had an idea that we would be doing it, mind you.

That has not happened yet.  It's still hanging up there in CMS space but the idea that the authority is there for waivers to basically pay on partial capitation for managing chronic illness and for implementation of shared decision-making in ways that are not feasible through the fee system so that you could do this without formal enrolling of patients into staff model HMOs, which is very difficult to do.

What we had hoped to do here was starting with places that we were certain because we were already working on these problems, we had encountered severe difficulty with the reimbursement system, to buy the infrastructure, for example, of congestive heart failure and for any of these diseases, that we wanted to have the opportunity to step forward and do this, not because we thought that this was necessarily replicable but, rather, if there were a few good models out there of how accountability across those six points I mentioned to you could be met, then there would be a lot of latitude for figuring out other, more aggressive ways, like selective contracting.

CHAIRMAN KASS:  Well, thank you very, very much for an eye-opening presentation, both in the written materials.  And there is much to think about here, very complicated.

It does bear upon seeing how some of these built-in structures and practices of utilization really do bear upon trying to do something for the long-term care.

We are running, as is our usual habit, behind.  Let's take 15, but really 15.  Dan Callahan is here and ready to go.  Thank you.

(Whereupon, the foregoing matter went off the record at 3:48 p.m. and went back on the record at 4:05 p.m.)