Eli Y. Adashi, MD, MS, CPE: Hello. I am Eli Adashi, Professor of Medical Science at Brown University and host of Medscape One-on-One. Joining me today is Dr. John Wennberg, Peggy Y. Thompson Professor in Evaluative Clinical Sciences at Dartmouth Medical School. A widely revered health policy expert, Dr. Wennberg is the founder and director emeritus of the Dartmouth Institute for Health Policy and Clinical Practice. Welcome to One-on-One.
John E. Wennberg, MD, MPH: Thank you.
Delving Into Healthcare Variability
Dr. Adashi: Back in 1973, you and Dr. Alan Gittelsohn published a seminal paper in Science wherein you laid out, perhaps for the first time, the notion of unwarranted variability across our nation in terms of healthcare cost utilization and quality. What was it that led you back in 1973, or even prior to that, to delve into this arena?
Dr. Wennberg: It's kind of a long story, but I can make it short. Basically, when I finished my residency program at Johns Hopkins, I also took a master of public health program, so I was trained in epidemiology as well as clinical medicine. I also spent some time at the Homewood campus at Johns Hopkins in the department of sociology, trying to figure out how health systems work and how individual people are motivated and their lives work. Then, I got a job in Vermont as a director of a planning program. It was called the regional medical program, or RMP, and the RMPs represented the last time the National Institutes of Health ever got involved in trying to fix healthcare. They gave very large block grants to medical schools, mostly under the assumption that they could regionalize health services for major conditions: heart disease, cancer, and stroke. It was a pretty good amount of money -- I think it was over $350,000 that the University of Vermont received -- and the school, in its wisdom, decided that I would be the director of the program. 'That's how I got the leverage and the opportunity to begin to use routine databases to analyze what was going on in clinical practice or what was going on in the regional healthcare system, and that's when Alan Gittelsohn and I put our heads together and came up with this idea of small-area analysis.
It was ideal to do so in Vermont, because Vermont is divided into 15 different communities, each with their own hospital, and people tended to use the services locally. Since we were interested in regionalization, the whole idea was, "Let's find out what's going on." We used large databases from hospitals, we got hold of the Medicare data, and we went into hospitals that were not in an organized system and actually hand-scribed the data. We also went into nursing homes, and by the time we finished we had profiled the entire system in terms of its variability from one place to the other. We also began to look at its productivity -- questions of whether more was better.
Healthcare Variations by Town
Dr. Adashi: It might be worthwhile for the benefit of our viewers to say a few words about the conclusions of the paper at that time.
Dr. Wennberg: The paper was not really a hypothesis; we basically were doing an empirical description of the system, and in that we were able to measure the location of local healthcare markets or service areas. We were able to measure the amount of resources that were used, how many doctors' inputs were made, how many nonphysician personnel were used by hospitals, how much money was spent, and how many beds were invested in the care of these populations -- and we did it all on a per capita population basis, which was probably the most unique thing about what we were doing. Instead of looking just at institutional behavior, such as occupancy rate or throughput of patients, we got down to the question how much care is actually delivered to these populations and the utilization rates. We found incredible variations in common surgical procedures, such as tonsillectomies.
We found that, for example, in Morrisville, Vermont, which was just 6 miles from where I was living at the time, the kids were having their tonsils out at a rate of about 65% by the time they reached age 15 years, whereas in my own hometown of Waterbury, the rate was about 20%. Those huge variations just didn't make sense in terms of the common assumptions about healthcare. The conclusion that we came to was that the system behaved much differently than most people intended when we set up the idea of healthcare regionalization. More fundamental issues, such as what is the value of healthcare and what is the best way of practicing healthcare, emerged, rather than the questions about underservice and how we can make sure that the advantages of the scientific innovations that have gone on through the NIH were available to all people. That was basically the theory behind regional medical programs.
What's Driving Variability?
Dr. Adashi: Having established, almost 40 years ago, the notion of variability in healthcare delivery consumption and quality, do we have a sound understanding, or at least an improved understanding, as to what might be driving this geographic variability?
Dr. Wennberg: I think we do. We've been pursuing this thing for 40 years, and at least here within the Dartmouth circle, we think we have a pretty good understanding of the drivers of practice variation.
The first thing that's important to distinguish is whether illness is driving variability, and it turns out there's very few conditions for which we can be quite certain that illness is driving it. For example, hospitalizations for hip fractures are pretty much driven by illness, and I think the clinicians in your audience will see why. It's because people who break a hip know it. They seek care. It's not hard to diagnose, and everybody would agree they should be hospitalized. Therefore, the hospitalization rate is a pretty good indicator of the actual incidence rate -- the hip fracture rate -- but that's one of the very few things that vary that way. Surgical procedures, admissions for chronic illness, and visits to physicians for chronic illness all show much, much more variation than can be explained by illness or patient access.
Dr. Adashi: In the absence of sufficient or credible accounting for the variability, are there other hypotheses or reflections on what might be driving this variability?
Dr. Wennberg: For what we call "unwarranted variation" -- that's variation that cannot be explained by illness or by patient preferences or by access to care -- there seem to be 2 basic streams of activity that we have identified. One is for procedures and interventions for which there is more than one option: For example, hysterectomy is commonly done for uterine bleeding, but there are ways of treating that condition other than hysterectomy, and the same is true for most common procedures, particularly those that show high variation. There are other treatment options. The decision process is driven by very strong ideology and theory. Surgeons don't operate on patients without a pretty good reason for doing it. The question is, where's the evidence that the assumptions made by surgeons are true, that we really can predict what's going to happen -- but even more important, particularly as we've seen over the past 20 years, is the patient's preferences. Treatments have different outcomes; treatment options commonly have different outcomes, and even in cases where the outcomes are well known, we see plenty of evidence that patients aren't getting involved actively in the decision process.
There is the simple example of breast cancer, early-stage breast cancer. Do we treat that with lumpectomy, or do we treat it with mastectomy? Randomized controlled trials show that the effect on survival is pretty much the same, but all the other outcomes are quite different. And it would be pretty hard to argue that those other outcomes should be decided by the physician rather than the patient. Yet, there's strong evidence, particularly in the way we've always conceptualized the role of the doctor-patient relationship, that patients do delegate decision-making to physicians, and generally the preferences of the physician become more important in the decision on surgical procedures than they probably should be.
Now, the managing of chronically ill patients is where we see a much different set of issues. With chronic illness, instead of having disagreements about the value of the procedure vs alternatives, what we're mostly seeing is just variation in the intensity with which patients are treated. For example, how frequently a patient with chronic kidney disease is seen by a physician. Is it 6 months? Three months? Six weeks? How often does somebody with congestive heart failure go back for revisits? How often are they hospitalized? How often are they put in intensive care units? It turns out that this kind of behavior at the clinical level is pretty subliminal, in the sense that there isn't a lot of theory driving that issue. In fact, there's almost no evidence to support these types of decisions.
In the absence of evidence and the general assumption that more care is better, the capacity of the system tends to drive the frequency of use of medical resources in managing chronic illness. This turns out to be a really important problem in Medicare, because that's the explanation behind the 2-fold variation in Medicare spending between such places as Los Angeles and Portland, Oregon. It's nothing to do with surgical procedure rates. It's all to do with how much care each region provides chronically ill patients, including (but not exclusively) end-of-life care.
Dartmouth Atlas of Health Care
Dr. Adashi: In 1996, the first edition of the Dartmouth Atlas of Health Care saw press. One might say this was an outgrowth of your 1973 science paper and your life's work. Would you describe the scope, content, and overall universe that the atlas encompasses?
Dr. Wennberg: Let me start with a little bit of the history of the atlas. You're right; in fact, it's pretty much the same methods that we originally developed in Vermont, at least in the early editions of the atlas. The atlas was conceived in the context of the Clinton Administration health plan, and the efforts at that time to reform healthcare. We were working in some degree as advisors to Hillary Clinton and her colleagues, and it was clear that they were trying to set up this competition between managed care, and it was going to happen in regions, and we knew from all the work we'd done that geographic regions vary enormously in the amount of money being spent, the resources available, and the way healthcare is delivered. If one is going to set up competition in a region, one has to know what's going on in the region, and that was our goal -- to basically provide an analysis of the instruments that would be developed under that plan and put in place.
Of course, when the Clinton health plan tanked, here we were with a nice grant, which we received from the Robert Wood Johnson Foundation, to do this: We developed a geographic analysis for the whole country using Medicare data. The work that we'd done up to that point was pretty much restricted to a few states, so now we had a national database. What do you do when you have data and no audience? Well, you create an atlas, and that's the way it happened. That was always part of our premise: that the feedback of information on performance to broad audiences was one of the most important remedies that one could bring about in terms of the practice variations, because if you don't know variations are there and you don't pay attention to them, you're not going to be able to do anything about it.
Influence of the Dartmouth Atlas on Policymaking
Dr. Adashi: I cannot help but believe that since its inception and to this day, the atlas almost certainly has influenced health policymaking in the United States. I would be very interested in your thoughts on that.
Dr. Wennberg: The important step of the atlas was to make transparent what previously had been hidden -- namely, very striking differences between the amount of care provided for similar illnesses in different local places. It wasn't just that we could show that remote Vermont was off the mark, but this now was a message that could be brought anywhere in the country. Initially, we saw a lot of interest from the American Hospital Association, which basically was the publisher of the first Dartmouth Atlas. As things moved on, it became clear that there was going to be controversy, there was going to be dissidence, there was going to be debate about what this all meant, and we got the sponsorship to go back to the Robert Wood Johnson Foundation and have since then been independently funded to do this. We haven't depended on any one of the actors in the system.
Once you get into the trenches and start trying to figure out what to do about practice variation, you're always led to the fundamental problem that science is poor: Simply, our clinical science has not been nearly as developed as our biomedical science. We really don't understand what we're doing in terms of outcomes in many, many examples, and we see big differences in spending levels and evidence that if you look more like the Mayo Clinic or if you look more like some of the better-organized practice systems, you would spend a lot less money on a per capita basis than what is now being spent on healthcare in most parts of the country. It became clear that there were some best-practice models we had not seen earlier because we weren't looking deeply enough into the problem. You could see that group practices tended to be more efficient, not just in terms of their cost per case or per service, but rather in terms of the per capita costs that actually went on in the regions in which they were located. This became part of the issue.
At the same time, we were able to see extraordinary differences in which some hospitals because we were able to get hospital-specific as well as region-specific information. We could show, for example, that while Los Angeles was quite costly compared with many other parts of the country, there was also a 2-fold difference among the individual hospitals within Los Angeles in terms of costs in managing chronically ill patients.
In terms of political impact, I think by being able to profile major academic medical centers across the country and seeing 2-fold variation and greater among our leading institutions, that the original comparisons that we'd made between Boston and New Haven -- which caused quite a stir -- now became the rule rather than the exception. That again brought great evidence toward the need to do something about the scientific basis in medicine.
The Influence of the Dartmouth Atlas on the Affordable Care Act
Dr. Adashi: To what degree, if any, do you think the atlas influenced the Affordable Care Act (ACA) or the process that led to its formulization?
Dr. Wennberg: We'd said many times that if you use benchmarks of efficient practice, we can indicate large opportunities for saving money by becoming more efficient. When the concept came out that we were going to increase entitlement, our argument was that if we keep the system steered the way it is, if we don't increase capacity, our analysis says you basically won't increase spending as much as you think you would. Certainly Peter Orszag (former Office of Management and Budget director) and the group that drafted the ACA picked up on that and moved ahead on it. Certainly more interesting is the work we did in medical effectiveness and the days of the prostate studies.
This goes back to the steps that were taken after we finished the Vermont study. We also had the opportunity at that time to begin to look at what was going on in the state of Maine, and at that time the editor of the now-defunct Maine Medical Journal was Daniel F. Hanley, MD, who was a very influential physician in Maine. He was on the US Olympic Committee, he was a Bowdoin College physician, and he was the secretary of the Maine Medical Association. He published the series of articles showing practice variations in Maine. He said it was the responsibility of practicing physicians to look seriously at practice variations and begin to address them. Working with Dan and my colleagues here at Dartmouth and in Boston, we set up a strategy for taking feedback to physicians to show them what the practice patterns were in their communities and in other communities. We asked 3 questions: Will practice variation information itself lead to a reduction in variation? If not, can we go to the research literature and figure out what is actually going on and make some judgments about inappropriate care? And, if we can't do that, can we do studies that would actually clarify what was going on?
Asking those questions for benign prostatic hyperplasia treatment led to the first example of medical effectiveness research. It showed a consistent effort to try to take the theories that were being exercised by local physicians, subject them to critical appraisal, and go on to the data -- and, in the case of benign prostatic hyperplasia, that led to the clear identification of the role of the patient in decision-making and the necessity of changing the doctor-patient relationship from informed consent to informed patient choice. This led to a whole stream of activity that ultimately influenced the Obama plan and the concept of shared decision-making. The establishment of shared decision-making is now one of the requirements for Accountable Care Organizations.
State of Healthcare in America
Dr. Adashi: It's undeniable that we as a nation are going through a difficult patch with respect to a variety of aspects, not least of which is healthcare. If we were to liken for a moment the American healthcare system to a patient, what is the prognosis of this patient?
Dr. Wennberg: It's certainly going to have to figure out how to go from a spurt of growth to a spurt of maturity and stability.
Dr. Adashi: From puberty to adulthood.
Dr. Wennberg: We could use those words so that everybody understands that the growth curve cannot be sustained, and certainly the great difficulty is whether the adolescent can enter into a new phase in which there is cooperation and an understanding of the requirements for building systems of care, and doing things that physicians were not trained to do, work in teams, and ultimately limiting capacity -- which is another problem because you get into the macroeconomic system and what drives novelty and what drives growth.
Dr. Adashi: Finally, and more on a personal note, may I take you to a time preceding 1973? Maybe you could share with our viewers what it was that led you to the area of healthcare in general, to become a physician in particular, and then one who also embraced population health, having obtained a degree in public health and having become a leading proponent of the same?
Dr. Wennberg: It's hard to answer a question that probably has many different threads coming into it. I know that my undergraduate training was entirely in the social sciences or in humanities, so I was not the classic premed student. I came to medicine with a different viewpoint -- or, let's put it this way, different sets of experience -- and I had one experience when I was at Hopkins that I might mention. At that time, I was on the nephrology service, I was basically trained as a nephrologist, and we had a patient who died after kidney failure apparently caused by a reaction to a drug that was given to visualize the gallbladder. I began to look at that and uncovered the fact that there had been some 25 deaths in the Washington and Baltimore area from this drug. I also began to give the drug to cats as an experiment to see how much they could tolerate and noticed that the cats weren't doing so well, and some of them died. That became sort of an eye-opener to me about the consequences and problems of failure to organize and understand the outcomes of what you're doing, and it also was an interesting example of where there wasn't any process in place that could really deal with this information.
The administration at Johns Hopkins took the drug off their market, but that was the limit of their responsibility as they saw it, and other places continued to use it, until eventually Congress got involved. This was a time when the FDA didn't really exist as it does now -- it was about the time it was being formulated -- so this particular example was used as one of the reasons why we needed to do something. Learning the relationship between what goes on at the macrosystem level and the regulatory system level, and what we as physicians have to deal with, really got me interested in public policy issues.
Dr. Adashi: Thank you for sharing.
Dr. Wennberg: You're welcome.
Dr. Adashi: On this note, sincere thanks to Dr. Wennberg and to you, our viewers, for joining Medscape One-on-One.