What I thought I would try to do today was not to give you any answers, but to try to identify, from the perspective of a social scientist who had a chance to serve in government and now has had a chance to have a broad perspective on a great university, what seem to me to be some of the largest broad areas in which our research and understanding can be enhanced, and which seem to me to be particularly important for future research, and where it seems to me that analytical work can add a great deal.
I say this from the perspective of someone who believes very deeply that one of the unremarked but important intellectual revolutions that has taken place in the last 30 or 40 years is the rise of social science as an actual form of science. C.P. Snow wrote a famous essay in the late 1950s about that there were two cultures: a culture of the humanities that appreciated aesthetic things, and a culture of science that understood physical and biological things and developed weapons and new production techniques and the like. It seems to me there is an important and increasingly ascendant culture of the social sciences which is finding more and more widespread application.
The recent example that perhaps highlights this best is the book, Moneyball, which I recommend to any of you who have not read it. Moneyball is about the rather implausible subject for a social science book, baseball. What Moneyball demonstrates is that the Oakland A’s won more games than any team in the American League over a five-year period, while having a payroll that was one-third of the payroll of other teams. And the question is: how did they do it? It does not appear that they did it by being lucky. It does not appear that they did it because their athletes strived harder or were a better team than the other athletes. It happened primarily because their athletes were better and more wisely chosen than the athletes on the other teams, and played with shrewder and better strategies than the athletes on the other teams did. And how did they come to be better chosen? Through econometrics.
A very wise baseball general manager hired a Ph.D. in econometrics to study the question of what were the predictions of who would be a good baseball player. And it turned out to be the case that almost everything that the baseball industry knew was wrong. The baseball industry was not in doubt that it was very important to be and to look like a terrific athlete. The baseball industry was not in doubt that the ability to swing powerfully and hit the ball a long way was a crucial predictor of success. The baseball industry was not in doubt that one’s time in the 40-yard dash was highly predictive for how effective a baseball player would be. And the baseball industry was not in doubt that really the best athletes were the guys who would go straight from high school to the minor leagues and to the major leagues, rather than playing college ball.
There’s one very important aspect of those four propositions: they all turned out to be wrong. The appearance of athleticism, beloved of baseball scouts, turned out to be almost entirely non-predictive. How many walks a player got on his college baseball team turned out to be much more predictive of his success than how many homeruns he hit. Fielding averages turned out to have very little to do with actual effectiveness in fielding. And very systematically, teams that drafted players out of college did better than teams that drafted players out of high school. Confronted with this evidence, the reaction of the baseball industry, generally for several years, was to simply say that was a lot of nonsense that had come out of a computer, and you had to talk to people who had really played baseball and “knew the game.” The Oakland A’s have won long enough that we now have a general manager here in Boston who plays baseball that way. But what’s true of baseball is actually true of a much wider range of human activity than has been the case before.
In the last 30 years, the field of investment banking has been transformed from a field that was dominated by people who were good at meeting clients at the 19th hole, to people who were good at solving very difficult mathematical problems that were involved in pricing derivative securities. The field of environmental regulation has substantially given way in its actual application from people who were committed activists and attorneys to people who were skilled in performing cost-benefit analyses. The presidential campaigns that at one time put out the call for a group of bright lawyers to staff them, now put out the call for bright economists and bright MBAs to staff them. And I could go on and on with these examples, suggesting that the kind of analytical techniques that come out of social science are finding more and more widespread application. And so it seems to me to be very appropriate to think about health care in these ways. And certainly over the 30 years that I’ve followed these issues, since I first studied with Marty Feldstein in the early 1970s, when the idea of being an economist working in health care was a novel one, the range of economic and other social science thought in health care has increased very substantially.
I want to highlight, though, five areas where it seems to me there is a lot more that we could know than we do know.
The first is the art of diagnosis and treatment for patients presenting with given symptoms. Let me say something slightly outrageous. Whenever it is proudly asserted by practitioners of an activity that that activity is an art rather than a science, they are describing an activity that is still in relatively primitive form, and where great progress will follow the application of more scientific techniques.
I was very struck, and I have to confess somewhat worried, by a conversation I had with an advanced medical student and a junior resident a year and a half or so ago as I accompanied them on rounds in the Brigham. They went in and they saw a patient and they had a variety of tests and they discussed what to do and then they said, well we’re going to do X. And I said, gee this must be one of thousands of patients the world has seen who had this combination of symptoms. Isn’t there any database where you could plug these symptoms in and learn what the possibilities were and just in case you’ve possibly forgotten what the right thing to do was, and test your intuition against the systematic experience of thousands of cases like this? The young resident explained, no, no, no, no, medicine is an art not a science and to reduce it to something mechanistic would just be wrong. That’s done, but not at really good institutions like Harvard. And at other institutions sometimes they use computerized diagnosis, but that is not where it’s at. And I said, well is there some possibility you could perhaps check what you were doing, just in case you might have made a mistake or that there was a resident who wasn’t as good as you or it was an area you weren’t so familiar with? Well no, no, we really need to understand the science of this particular patient with this particular disease.
Maybe it’s so, but that’s what lawyers used to say before they developed forms for handling particular problems. That’s what people in 1900 said about the making of automobiles. It seems to me that we have much less in the way of standard tools for verifying judgments in the face of symptoms than one would like. I have read enough of the stories from the New England Journal over the years, and more recently the pieces in the front of the New York Times Magazine, describing the work of clever diagnosticians to understand that this process cannot be completely mechanized. But everything we know about human judgments is that they are highly fallible. That every one of us, even after we have been warned to the contrary, is, in judging any situation, much more a prisoner of our recent experience rather than our general experience than we should be. And every one of us is prevalent, every one of us is subject to, the phenomenon psychologists refer to as excessive subjective certainty, the phenomenon of fully recognizing the uncertainties that are present in any given situation.
Let me illustrate that proposition. I want everybody to think in their minds what they think the annual expenditures of Harvard University are. And I want you to not think about what you think they are, but I want you to think of an interval that you are 90 percent certain includes the total expenditures of Harvard University. Everybody have an interval in their mind? How many people’s interval includes $2.3 billion? How many people’s interval does not include $2.3 billion? Remember you were all 90 percent certain that you were going to do that. I don’t have the time to do it here, but this demonstration can be done. You can do it. It typically happens just like it did here, that about 20 percent of the people give answers that fall within the 90 percent interval. And then you say, well you probably should widen your interval everybody, and you do another question. With most groups, it takes eight or 10 questions before people can manage to give intervals that are wide enough to reflect the actual uncertainty in a situation. Now it’s possible that none of those biases ever affect any clinician as they look at a patient and form a view. But it seems very, very unlikely.
So the first area where I know, in the areas of evidence-based medicine, decision analysis for medicine, there has been important work done, but where in my own experiences with the health-care sector I am struck by how little of this data is brought to bear, where it seems to me it is very important that there be more progress, is in the bringing to bear of this kind of evidence and data to guide judgments about prescription and treatment.
A second broad area where it seems to me that we still know less and are less able than we should be, and where contemporary social science can make a big difference, is in the reimbursement of health care. What we know is that people do respond very powerfully to incentives. If you pay them to do procedures, they will do more procedures than if you do not pay them to do procedures. If you charge them money to do procedures or to make referrals, they will do fewer procedures and they will make fewer referrals. We also know something else, which is that people are far more responsive to incentives than they say they are. You take two groups, you change the incentives facing them, the groups are otherwise identical, make a big difference what the incentives are, but if you ask the members of both groups, were you affected by the incentives, they will absolutely deny that that is the case. So the incentives that we create make an enormous difference for the practices that actually result. And the problem, therefore, and the problem that we have not solved, is in classic market settings the world has a good solution to it. It turns out that when General Motors maximizes its profits, it is also maximizing the difference between the value it successfully creates for consumers and the costs that it incurs of society’s scarce resources. And so simply letting General Motors maximize profits will get us as a society reasonably close to an efficient outcome. The problem is to find something similar in health care. We have seen over the years the consequences of both the schemes that I refer to. If we pay people for doing procedures, traditional fee-for-service medicine, we see rather more procedures performed than most of us would judge to be efficient, and we’ve been through a period where I think most people would agree that we have seen some evidence that without great care in monitoring, if we simply pay people fixed amounts to provide health care for people, they don’t take as good care of those people on some important dimensions as perhaps they should. And the problem is to find some happy mean.
There is an obvious solution that presents itself at first blush, and that is to pay people based on their success in keeping people healthy and improving their health. That is, after all, value-added. If you want value to be added, then you should pay for value-added. And it is frequently observed that if health-care providers are reimbursed on the basis of how healthy their patients are, then they will have enormous incentives to do all the right things. They will perform procedures when those procedures will be helpful. If it is more efficient to counsel people not to eat so much than it is to do procedures on them; they will have a strong incentive to counsel people not to eat so much. And so what is held out as a solution is outcome-based reimbursement, and it is a very powerful idea. There is one problem, and it is a problem that to date it seems to me has most been fatal with respect to schemes of this kind, which is that the outcome in terms of the health of a 65-year-old male depends upon two things. It depends upon the value-added of the health care provider during that man’s 65th year, and it depends upon how healthy he was at the end of his 64th year. And if you pay people on the basis of health outcomes, there is an enormous incentive to compete, not by being a better service provider, but rather by being a cleverer screener of patients. One can, to be sure, eliminate the more obvious ways of doing this. One can require that you can’t give people a physical and turn people away based on their physical. There are a variety of things one can do to stop this kind of creaming. But I can’t help but wonder whether it is an accident that many of the most profitable HMOs, on the one hand, have terrific sports medicine practices, terrific, responsive and prompt well-baby care practices, and gosh, they just don’t turn out to be very good at treating chronic asthma. They just somehow didn’t manage to get to be good this year at treating chronic asthma. It’s probably not an accident. There is a very powerful incentive to compete directly and to compete indirectly by altering the terms of selection of patients. This is not a phenomenon that is unique to health care. It is absolutely pervasive in the insurance industry. There is a man walking around in America who is a billionaire, and the insight that is behind his billion dollars is that if you are a married male with a stable job and a good credit rating, over the age of 40, who rides a motorcycle, you are actually very safe. And therefore, you can be sold motorcycle insurance at a much cheaper price than all the other motorcyclists. And by having that insight, he was able to produce a fortune, and it is exactly that kind of creaming that is the central risk in health care. What is the obvious answer? The obvious answer is that our success in being able to craft satisfactory reimbursement arrangements depends upon our ability to measure value-added, which depends upon our ability to measure initial risks in satisfactory ways. This is an area that is exactly right for precisely the kinds of statistical analyses that I’ve been talking about. And the success in carrying them out and being able to risk-adjust patterns of reimbursement could well be crucial in improving the efficiency of competition and hence, ultimately, the efficiency of the health-care system.
The third area where, it seems to me, we have done less research and we know less than one could imagine that we could, is what I would call getting people to do what they should. I don’t know the precise numbers. I have asked people this question a number of times, and I get a range of answers to the question based on what medical science, social medical science, now knows, if you assign an average adult male in America to take a pill once a day, and you tell him that it’s important to his health that he take that pill once a day, what is the probability that he will in fact have taken that pill once and only once on more than 28 days in a 3l-day month? My impression, from having surveyed a number of people who regard themselves as knowledgeable, is that the answer to that question is about 50 percent. And if you think about the lost access to pharmaceuticals that comes from people not taking the drugs that they should, it is an urgent social problem. So is the problem of people who become addicted to things they should not become addicted to. So is the problem of people who eat and exercise in ways that they shouldn’t. Are there ways of modifying behavior? There are techniques. There are different ways of dispensing pills that have differing efficacy. There are computer screens that remind you on your exercise and diet. There are a range of things that are out there. But I can’t help but think that the degree of intellectual talent that has been applied in our society to getting people to do what they should and what they probably like doing or not mind doing if they got used to it, has not been 1 percent of the talent that has been applied to any of 20 aspects of cell biology.
And if I think about the social payoff to successful research in that area, it seems to me that it is enormous. I don’t have particular suggestions as to how best this can be done. I’m rather skeptical that everything we know, or everything the private sector knows, about how to get people to do things that they maybe should do or maybe shouldn’t do, involving spending their money, if we were as effective in getting people to stop doing things that are bad for them as we were at getting them to buy things that they don’t need, it would represent very substantial progress. But the degree of research effort that goes into these behavioral questions seems to me to be a small fraction of what it ought to be.
The fourth question I would throw out for you is how little, and it’s related to the third, is how little it seems to me we know and we allow rational planning to guide the allocation of our research fund across all the areas in which research is possible. I just mentioned one area where I have a strong suspicion that the social payoff to increased research is pretty large just because of the paucity of research that has been done so far. I can’t help but wonder why some of the research that we’re now seeing, and is having enormous impact on human diet and on human exercise, that is actually not dependent on the successful sequencing of the genome, is actually not dependent on scanning equipment of a kind that has been developed only in the last five or 10 years, is only coming to have such large effect now. Part of it, I suspect, is the prejudice towards associating the word “science” with physical and biological processes rather than social phenomena. Some of it is the sets of cultural surrounds that are associated with public health rather than medicine and private health, even though most of the increases in life expectancy that have taken place over the last century have come more from public health than have come from increased medical care. Part of it, I suspect, is that we don’t have any accurate estimates that anybody can cite for the payoff from any research, and therefore we have a tendency to go with our intuition and to go with existing tradition. I don’t think I know what the right solutions are to this problem, but it seems to me that some more satisfactory metrics for allocating society’s research funds would be likely to have very substantial payoffs. In that same regard, it seems to me that there is a canonical problem that exists in many areas of health-care research, that’s perhaps most explicitly posed in the area of diseases that are only prevalent in less-developed countries.
The basic economics of pharmaceuticals and software are the same, to lapse into economics lingo for a second, to develop a new type of software or to develop a new drug, there is a big fixed cost, and once you’ve developed it, the cost of making it available to more people is zero or is very small. High fixed cost, low marginal cost. Here’s the problem. If, after you have produced the knowledge, you can only collect marginal cost for it, there is no very strong incentive to produce the knowledge. On the other hand, if, after you produce the knowledge, you charge a sufficient price to recover your fixed costs, something terrible has happened. This stuff that costs almost nothing to produce and is very valuable is being withheld from people for no reason other than to help some guy to recover a cost he has already incurred. That’s a problem with respect to software. That’s the basic problem with respect to AIDS medications. That’s the problem with respect to a new idea. Suppose I spent $1 billion and as a consequence of developing the $1 billion, I found a particular diet. I’m going to choose an implausible example. But I chose a particular diet involving a particular rotation of meat and fish and vegetables that was enormously efficacious in terms of improving people’s health. How could I profit from that invention? It would be very hard for me to keep it secret. Once I sold it, it would be out there for everyone to have, and therefore there is no incentive to invest $1 billion in the first place. And so, while we don’t know much about where there’s going to be too little research, we do know something, that it’s in general going to be situations where it will be very difficult to charge more than marginal costs and to recover your research funds. It seems to me that the suggestion that our economics colleague, Michael Kramer, has made in the tropical disease area for a so-called pull-incentive, is one that potentially could find quite widespread application here. Kramer’s suggestion is that the government announce that if a successful vaccine is found for any of a number of diseases, it will guarantee a market of a given size at a given price, and it will in effect subsidize the ultimate demanders so the successful discoverer of that vaccine is not put in the position of either having to give the vaccine away and not recover costs or deny lifesaving treatment to very poor people.
A similar list of important discoveries that would call forth a very substantial payment from the government, it seems to me, in a broader range of areas, could make a significant contribution to this problem.
Let me make a final observation, and it’s of a rather different sort. And that is, I am struck as I look on very much as an outsider by the absence of an animating idea around the domestic health-care reform debate right now. For a time, the animating idea was national health insurance. There was a powerful idea in the early 1990s. It was managed competition, managed competition at the level of individual employers, proposed unsuccessfully, managed competition at the level of the nation. There was a theory that managed competition would call forth competitive improvements in health care and lead to better performance that might or might not be coupled with reforms to improve access. At least in the popular mind, the bloom is off the rose of managed competition, in part because of the problems of undercare that I referred to, perhaps also in part because of the problems of creaming the most select patients that I referred to. And so the sense that we know how to solve the problem, we just need to introduce managed competition that people had in the 1990s, wasn’t there. And I am struck that while there are a lot of ideas and there are all kinds of experts who one can go to, how should we reform Medicare, what’s wrong with the prescription drugs in Medicare bill, what’s the plan for broadening children’s coverage? If you say at the broadest level, as a president of the United States would, there are 43 million Americans without health insurance, health-care costs are again going up at two or three times the prevailing rate of inflation, perhaps reflecting increased services, but nonetheless they are going up very fast. There is still a lot of evidence suggesting that in some cities in the Unites States people get many procedures at two or three times the rate that they get those procedures in other cities, and it is not to their evident benefit. What is the broad character of the principles approach that we should be taking to address this problem? I am completely stuck by the absence of voices who are convincing even to themselves, let alone to the nation as a whole, in answering those questions. And it seems to me that in many ways it is the largest of all questions. It seems to me that that question, like the first four that I’ve asked, is a question that can valuably be addressed in a great university like this one. It does seem to be that the lesson of history in the very longest run, not over five years, not over 10 years, but in the various longest run, is that the success of societies depends very much on how systems are organized to help people do their jobs, and how incentives are created. And it is within that that the scientific breakthroughs take place and that their application takes place, and that progress in thinking about those questions ultimately makes the most profound difference of all. And that’s why I think this University is so fortunate to have so many extraordinary scholars working in the area of health policy, because I think it is an area in which there is a very great amount to do.
I hope these reflections have provoked and challenged you, at least a little bit. And if I didn’t say anything that you find to have been foolish, or at least provocative, I will have not succeeded.
Thank you very much.