Capitalisn’t: The New Economics of Industrial Policy
Harvard’s Dani Rodrik visits the podcast to discuss changing attitudes toward globalization.
Capitalisn’t: The New Economics of Industrial PolicyJosh Stunkel
(light piano music)
Hal Weitzman: Big data is changing many industries, and health care is no different. From Fitbit to the data crunches in hospital systems, the amount of information being produced about our health is growing fast. Yet much of it is spread among organizations that don’t always communicate very well with each other—hospitals, primary-care providers, researchers, insurance companies, and state and federal governments.
So is big data making us healthier? And how are hospitals and health-care systems using the information they collect to improve patient outcomes and lower costs? Welcome to the 25th edition of The Big Question, the monthly video series from Capital Ideas at Chicago Booth. I’m Hal Weitzman, and with me to discuss the issue is an expert panel
Dan Adelman is a professor of operations management at Chicago Booth. He’s an expert in management analytics and helps companies use data and decision analysis to improve their management and strategy. He leads the Healthcare Analytics Laboratory at Chicago Booth, where PhD and MBA students work on real-world projects with health-care institutions.
David DiLoreto is chief clinical operations and innovation officer for Presence Health, the largest Catholic health system, based in Illinois. He’s trained as an ophthalmologist and an oculoplastic surgeon and he also holds an MBA from Emory University
Jonathan Silverstein is vice president and Davis Family Chair of Informatics and head of the Center for Biomedical Research Informatics at NorthShore University HealthSystem. He’s also a research associate at the University of Chicago’s Pritzker School of Medicine.
And Harold Pollack is Helen Ross Professor at the University of Chicago School of Social Service Administration. He’s also codirector of the University of Chicago Crime Lab and an executive committee member of the Center for Health Administration Studies at the University of Chicago.
Panel, welcome to The Big Question.
David DiLoreto, let me start with you. Just give us an update, for those of us who haven’t been following this, how have data analytics—analyzing these big data sets—changed health-care delivery?
David DiLoreto: Well, now when folks get services for health care, there’s information about them in many, many places: in physicians’ offices, maybe in a pharmacy, maybe in a hospital, or a lab. Increasingly, people are even wearing sensors like I am, so there’s lots of information now.
How we actually gather that information is what’s changing, and we in the past didn’t really bring that information together. It sat in individual silos. The ability now to bring that information into a central repository gives us tremendous opportunities to analyze that information and actually find out about gaps in care, opportunities for improvement, or even predictive modeling about what might happen. We’re in the early stages. It may even be in its infancy, some folks may say, but it has a very significant impact in today’s operations environment.
Hal Weitzman: Jonathan Silverstein, what have you seen about how health care has changed since big data, you know, came more into play?
Jonathan Silverstein: Well, I think, again, I would agree that it’s still pretty early, but the areas in which the impact really is coming is in population health. Changes in health-care financing that we’re all seeing really have driven health systems to need to understand the population that they serve and begin to classify groups of patients in different ways, understand what may be the best therapies for them, where they’re going, and how they’re progressing through the care system.
I think you know really the ultimate in big data gets very much down to the individual patient and understanding enough to really know what to do with the individual. We’re not there yet. I think to me the the problem that we’re really trying to solve that we’re approaching is one of mass customization on an incredible scale, where every single patient is different and context matters tremendously.
So I would say it’s really happening in population health, with us working our way through understanding context enough from all this data to do impactful things for individual patients.
Hal Weitzman: Dan Adelman, how is all this big data that the hospitals and other health-care institutions are using, how is it changing the value equation in health care? You know we’ve all heard about the $500 Band-Aid and sort of stories of overcharging. How is that changing because of big data?
Dan Adelman: Well, I think we don’t know yet. So I agree with the other panelists that things are still in the infancy of big data in health care, but I do think there’s, in my mind, two areas where analytics has great potential. One is accelerating adaptive learning in the medical enterprise. You know, it used to be 100 years ago that physicians would just operate as silos, locally using their own experiences to base their practices on, but now with data being available on a national level—potentially one day, and some of it already is—physicians can share information very quickly to learn about best practices.
So that’s one area. And the other area I think where big data analytics in health care is having a potential huge impact in the future is enabling consumer choice. And so now there’s so much data available. There’s a potential for this data to be used to compare hospitals against one another, and consumers to use this data and this information about outcomes and whatnot to make better, or different choices at least. There’s not a whole lot of evidence yet that that’s happening, but certainly with patient satisfaction data, that’s been around for quite some time on the internet, and that’s had some impetus.
Hal Weitzman: We seem to be talking a lot about potential and trying to predict the future, which is obviously a dangerous game for academics.
Harold, you look at policy, health-care policy. How much is policy and the need to drive down cost to meet, you know, the demand to lower health-care spending in the US? How much has that driven the big-data revolution in health care?
Harold Pollack: It’s been fundamental. We have a $2.9 trillion health system, and if you look at any projection about the future of the federal budget, or the Illinois state budget, any government in the United States, the future of health care is the big variable that is either going to make or break the financial health of our government structures.
And all across the economy, everyone understands that we are changing from a medical system based on what’s called fee-for-service medicine, where doctors and hospitals are rewarded by how much cutting and prodding and imaging they do, to one that is really based on providing real value in patient outcomes. And through health reform and other recent legislation, I think it’s become obvious that the old world of medical care is over, and we’re moving to something new.
We don’t exactly know how it’s going to work, but big data is going to be a fundamental part of it. And so we’re seeing real signs of rapid improvement in American health care right now, not just in terms of ensuring more people, but also in terms of lowering the rate of cost growth, improved patient safety, and in many other areas, quality is improving quite rapidly. And I think that big data is playing a role in that.
Hal Weitzman: OK. So big data has a potential to drive down costs significantly?
Harold Pollack: Or at least to stabilize those costs, which would be a huge win right there.
Hal Weitzman: OK. Jonathan Silverstein, what kind of data are you collecting?
Jonathan Silverstein: Well, that’s a great question, and so, I think as David said earlier, there’s many, many different kinds of data collection. One thing that we’ve focused very heavily on at NorthShore University HealthSystem, particularly in our group, is to get what we call discrete data directly from the clinicians at the point of care.
And so with the electronic medical records coming forward, which we pushed to very early, more than 10 years ago, we are now getting tighter and tighter on the data that we collect. So when you want to understand what’s happening with individual patients or with the population having that discrete data from a known source that’s classified, it helps you tremendously in your analysis, as distinct from text data that’s dictation of what people are doing.
And so we’ve put forward a program that we’re now beginning to move through the medical record community into other institutions that allows us to collect the entire record from the physician’s office in certain subspecialties—it doesn’t work for everything—by collecting it at that point of care in the office, click by click collecting everything, rather than these long dictated things that are often . . . really the problem isn’t so much that we can’t get the data out of those, as you might think. With natural language, processing is quite good. The problem is omissions. There’s data that’s just not there. Things that people didn’t say, and so we’re very focused on what we like to call sushi gray data. It’s consistent. It’s complete. It’s discrete in form coming from all of these various places.
And I would say that is a huge task. That is a very significant change in the way that physicians work, in the way that the medical assistants work. And it’s really a next phase of electronic health records before this first phase of just putting them in to do billing.
In addition to that, the digital health effort is quite strong in our institution, and it’s coming along and others as well, in which things are being collected directly from mobile phones and smartphones. This gives you a context of data collection that is not in the office because you spend most of the time not in the office. And so that’s really where most of the data is.
We’re also doing very interesting things with the data that used to be collected and thrown away, which is the streaming data in ICUs and other circumstances in which there are large amounts of data being collected in wave forms and other things that were just too much to collect and keep. We used to, you know, throw away imaging data, and now it’s possible with storage that we can keep much more of it. All these wave forms in ICUs. You know, we have projects working with babies in the ICUs, where we are storing the data that comes from all of these many monitors that they have on them. So we can begin to understand how things change and and begin to look ahead.
And I think the real promise is getting a handle on all of that for prediction. You know, prediction is a big word, but at the end of the day, it’s finding the things that are associated with what’s going to come next and giving people a head’s up.
I don’t think we’re gonna move into some kind of automated medicine, but I think that data collection allows us to provide insights to physicians and other health-care professionals that are the decision makers and really giving them better ideas on what the data shows. “Hey, the data is telling us this. You might want to take that into consideration.” I think that’s what our predictive models really will look like from all that data collection.
Hal Weitzman: And I want to get more into predictive analysis—
Jonathan Silverstein: Sorry—
Hal Weitzman: No, no, it’s a great point.
David DiLoreto, what kind of data right now are proving most useful?
David DiLoreto: Well, clearly clinical data that we can aggregate is important. But the the burden of chronic disease in the United States is fairly significant. So folks with hypertension, diabetes, asthma, obesity-related issues that are really outside of the medical system are more and more important to us. So the ability to understand behaviors is—and the data associated with behavior—is something that we’re very interested in.
So for instance, we take care of lots of folks who are low-income and poor in Illinois, and as we reach out to help them manage diseases, we encounter, for instance, their ability just to communicate with us with their smartphones. They’re often using prepaid cell-phone plans, and we can reach them in the first few weeks of a month, but in the last few weeks of the month, when those minutes have run out, it’s harder for us to actually reach them. So we create incentives around, “Hey we’re gonna help you get more cell-phone minutes, if you can share your data on your blood pressure or your diabetes,” and then enable their smartphones to actually provide us with some of that information.
So creating incentives for people to engage with us and managing chronic diseases means different types of data now have to come, outside of things that traditionally we had in the health system.
Hal Weitzman: Harold Pollack?
Harold Pollack: Well, I love that example because it also illustrates something important, which is that you have to combine high tech and big data with actually understanding how communities work, what’s happening in patients’ lives, and really engage people often outside the medical sector to understand, you know, what are they eating? What are the problems that they have in their lives and how can we help them with the big data resources that we have?
And I think one of the challenges we have in the health-care system is we’ve traditionally been inside the buildings, helping people when they come and visit. And we have to get outside those buildings and really engage people’s lives, and big data allows us to do that. But to be effective doing that, we really have to be much better than we are right now at really engaging communities in different ways. And I think that’s one of the big challenges that we have deploying these big data resources most effectively.
Hal Weitzman: Dan Adelman, what about this question of driving down costs? What are some of the examples of how data helps . . . there are two things, right? There’s improving outcomes and there’s driving down the costs while you’re getting to those outcomes. So what are some examples of how big data is helping us to do both those things?
Dan Adelman: Well, the big thing that many health-care institutions, if not all of them nowadays, are interested in is reducing the readmissions rate, the 30-day readmissions rate. And a lot of this has to do—
Hal Weitzman: Just explain what that means.
Dan Adelman: So CMS . . . there are incentives for, or penalties actually put in place if a hospital discharges a patient with, say, heart failure, and they return to the hospital within 30 days, this results in
Hal Weitzman: Going back this model of: we got rid of the patients so we should be paid. Now it’s about what actually happens to that patient a month later.
Dan Adelman: Yeah, exactly. That’s right. So a great deal of interest, especially in the projects that we do in the Healthcare Lab here at Booth, are oriented toward trying to do things to reduce the readmissions rate. Part of that is trying to build predictive analytic models of what can predict readmissions. And there are obvious things like diabetes and, you know, salt levels, and blood pressure, and those kinds of things.
But to build on what Harold was commenting on: the one area where we lack enough data is on the social determinants, because so much of readmissions is actually determined by behavior and other conditions that are happening outside the walls of the hospital. But we don’t have . . . while we have, you know, electronic medical records and operational data and financial data, we don’t have so much data about the individual patient and their circumstances. They don’t have access to transportation. They don’t have a telephone, all of these things. They don’t have a family support structure.
And so I think that moving forward we’ll be moving into a world where more of that data will be collected by providers. And then the question is how to use it, and that’s really one of the most exciting opportunities.
Hal Weitzman: But is the idea that it would drive down costs because you wouldn’t be wasting time on unnecessary procedures, or you could tell people before they’re gonna have a problem that they should come in for a checkup?
Dan Adelman: Yep. There’s the anticipatory piece of it. There’s the ability to intervene before a large medical expense. There’s other things too. I mean, so for example asthma patients in the uninsurable population. It’s felt that we can provide medical care, especially free medical care to help treat chronic asthma in patients, and if we did that for free, it actually would pay dividends because we’d prevent their longer-range hospitalizations and what not, so. And so all of that, those kinds of interventions can be analyzed, and you know, for financial effectiveness and justified and done with data. If we’ve got enough data, we can actually make the case for more of these kinds of interventions
Hal Weitzman: OK, well, Jonathan Silverstein, you brought up preventive care earlier, so let’s get into it a little more in depth. I think it’s intriguing, you know, a lot of people have these Fitbit devices or other things on their phones, and it’s kind of a game, right? They’re tracking their own progress. I mean, is there really a future in which that will become very significant in the way that you treat patients?
Jonathan Silverstein: So I think there’s two things there. One is that we don’t yet have enough data that’s been organized and compared with medical records to understand what some sense of normal is in that. So there’s sort of the average sense of normal. That’s one thing. We’re not there yet.
The other, though, that I think may be much more impactful is, because the data is relatively continuous, is really comparing patients to themselves and looking for changes. I’m working on a project in which we have a colleague in Urbana who has developed a model to look at gait tracking and look at changes, and we’re starting to compare—
Hal Weitzman: “Gait tracking,” as in the way people walk?
Jonathan Silverstein: The way people walk. From sensors in the phones, and looking at how the way they walk changes as different events occur in health records. It’s literally evaluating the change in the spring in your step, and what that might mean in terms of your health status changing. With chronic disease, that approach could be very valuable at setting a flag and saying, “Hey something’s going on over there. Maybe we ought to talk to this person,” relative to the hundreds of thousands of people that are out there.
So I think that really is an opportunity there. The second. I think . . . we have begun to move . . . We have hundreds of thousands of people that use our patient portal. So they communicate with their physicians and make appointments. They do billing. They ask questions of their physicians through the internet. And we’re now moving that to a model much more of patient questionnaires. So we’re asking questions and we’re getting more and more contextually specific.
So we can begin to predict what social determinants, what characteristics would put people at high risk for various things. And then we need to maybe do them differently. Maybe this person has some features that make them high risk for surgery, or make it necessary to do other things beforehand. Are they at risk for particular postoperative complications based on family history? Are they well-prepared? What is going on in their lives that may be a red flag that they’re not going to be ready for a big operation?
And so, and then postoperatively, following the recovery, in really getting a true knowledge of what their recovery is. In orthopedics, for example, there’s a good long tradition of questionnaires evaluating whether patients can do their activities of their daily lives after surgery and sending them out a questionnaire every week or two. It doesn’t have to be continuous. We can get a very good indication, match up those preoperative things with postoperative ones.
You know, being a surgeon, I’m very focused on that particular model. But I think it applies to any kinds of disease. Anice thing in surgery is that the events happen relatively quickly. In chronic disease, the same model of pre- and post- is over many years and it’s a little bit more complicated to analyze.
Hal Weitzman: Do you think we could get to a stage where we reduce the need for checkups or as many checkups because we’re using these kind of analytics already?
Jonathan Silverstein: There is a lot of evidence already that the annual checkup is not only inefficient but, quite expensive. And so I do think that the notion of changing the model of those is realistic in our future. I think there’s a set of diagnostic activities that have to go on, but the notion that they all happen in an annual visit with the primary-care doc who’s faced with, you know, an unbelievable number of things to try and do in one visit once a year. That model will change. I mean, there are studies that show that if the primary-care doc was to do all the things that they’re supposed to do in an annual visit, it would be you, know order, an order of magnitude longer. And so, they all can’t be done.
And so I do think we do end up in a model like that, and I think the annual visit maybe goes away or maybe becomes something very different.
Hal Weitzman: David DiLoreto, does that also mean for hospitals there’s an opportunity to treat patients far beyond their traditional geographic region?
David DiLoreto: Absolutely. So we’re testing now in our own employees who we insure a smart glucometer. So if you have diabetes, you can test your blood sugar daily. That information actually goes to a database. And then you get information back that might say: Your blood sugar is up a little bit today. Take 5,000 more steps. Drink more water.
And so we’re getting real-time feedback to people who formerly might have been testing daily but only seeing a physician three or four times a year to try to manage diabetes. So there is this ability now to engage people more actively and effectively using this technology, and they probably have to worry a little less about their diabetes in that case because they’re actually getting daily support.
The other thing that we’ll find is what activates people. You know, the sensors that people wear now are often somebody who’s already pretty fit, and they wear it for about three months, tests themselves, and they put it down.
But one of the companies recently showed me what they called the whistle, which is actually a sensor that you put on a dog. And they asked me: What do you think about putting a sensor on a dog? And all I could think about is a little dachshund in my house. It has its own closet with leashes and clothes that the family likes to put on it. But what they said is: Doctor, what if we told you we have a validated study that will say that people who should be wearing a Fitbit because they’re obese or inactive and won’t, they will do it if you put it on their dog and they see that the dog needs 5,000 more steps today. They’ll get up and walk the dog.
So the ability to actually engage and change behavior through the analysis of data and validated studies is going to be very, very important. It’ll happen outside of our hospitals, and that’s going to help quite a bit in increasing the value of health-care services.
Dan Adelman: There’s another example, just to go back to Jonathan’s example about orthopedic surgery. So one of the places where value in health care is . . . there’s a lot of interest in assessing it is in orthopedic surgery because many of those surgeries are elective, and so they’ve gotten on the bandwagon pretty early on just to assess it. And so, but with the data that Jonathan is talking about, you know, the gait and all those, you can actually compare, for example, different kinds of knee procedures: partial knee replacement, full knee replacement, which is better. And you can, one way to figure this out—there’s different costs, but then you want to compare them to outcomes. Traditionally it’s been difficult to figure out. How well are they walking two weeks out? Three weeks out?
But now with this data, we can use that to then see if there’s any statistical difference based on what kind of procedure they had on their knee, for example, and so that’s . . . I think that’s one great example of how you can assess value . . . we’ll be in a world where we can better assess value that’s being created in health care.
Hal Weitzman: Harold Pollack.
Harold Pollack: From a policy-makers’ perspective, this is also critical. One of the real challenges, particularly in the Medicare program, is the tremendous variation, local practice variation across the different parts of the United States, often different parts of the same state, where we see, for example, in orthopedic surgery, really different rates of surgeries for similar looking problems.
And no one’s doing anything that’s obvious malpractice. No one’s doing anything that’s obviously, you know, fantastically better. But there’s tremendous differences in cost. And policy makers need to get a handle on that because if we could have more uniform standards that are based on some transparent criteria, particularly in some areas like orthopedics, we could really save a lot of money and also get better patient outcomes. Because we’re overaggressive in some areas that are actually harmful to patients, so that’s one of the critical issues the federal government’s very interested in.
Hal Weitzman: Another big policy I wanted to ask you about is bundled payments. This move to bundle payments. Tell us quickly what “bundled payments” means and how it’s different. And is it really having any effect on, let’s say, patient outcomes?
Harold Pollack: We have known for a long time that if you pay physicians, surgeons, other providers based on a particular service, and then they they get to send the patient on their way that we will have, first of all, they’ll be very aggressive in the services that are well-paid, and also no one is really accountable for the overall well-being of the patient. And so we see excessive imaging and a number of areas that are just, you know, very familiar but difficult problems to uproot in the health-care system.
As part of health reform and also coming in other ways, there’s a whole series of experiments and demonstration projects that are trying to say to health-care systems: Here’s a patient with diabetes with certain clinical indicators. We’re gonna pay you a certain amount of money, and it’s your job to manage this person’s care. I’m simplifying a very complicated arena. But the basic idea is: it’s up to you to figure out how to provide good care for this patient in an efficient way that also provides good quality.
And that good quality part is especially important because we have to make sure that the American public trusts what’s happening. And you know, in the early iterations of managed care, a lot of people and the American public felt, well, if you do this in the wrong way, you’re basically going to provide incentives to do too little for people. And with the kind of data systems we’ve been talking about, we can really hold systems accountable to say, hey, this patient is actually . . . not only are you being more restrained in the areas where you don’t need to be aggressive, but this patient’s actually healthier, and that’s . . . and there’s some good evidence that we’re starting to see, you know, real improvements.
Hal Weitzman: So just to be clear, the bundled payments is paying for the whole arc of treatment?
Harold Pollack: Yes
Hal Weitzman: And in your mind it has . . . it is making an effect or it will have an effect?
Harold Pollack: It’s already having an effect and is clearly in some form the way that more and more people are going to be paid. I believe that in the next few years the majority of all the money that Medicare is paying out is going to be for some form of bundled payment.
Hal Weitzman: Jonathan Silverstein, bundled payments?
Jonathan Silverstein: Yes, so I think, you know, in surgery the bundled payment has been a thing around for a much longer period of time because one of the early switches was that if you operated in an elective case or a particular case, you’re responsible for something for . . .you get a global fee it’s called, which is you get paid for that, and you know, if you have to reoperate on the patient or bring them back, you don’t get paid for that because it’s in the global fee, period that that happens afterwards.
And so I think it is familiar. It is coming along. I think in the readmissions process that Dan was mentioning it’s come pretty quickly, and people have actively put data behind it. You know, our experience with that was that looking at many different factors as to which patients are likely to be readmitted was easy; finding the intervention that prevents them from being readmitted is much harder. And interestingly, to your point, one of the things that we found that’s quite effective in that and actually can move the dial on the data a little bit is telling the primary-care doc, your patient is leaving the hospital today, and they’re in our high-risk category, so you better pay attention to them.
And we’re not really sure exactly what they do, but we do know that when we tell them that your patient is leaving the hospital and you’re their primary-doc, fewer of them actually come back and get readmitted. So I think that knowledge of figuring out exactly how to change population health based on these incentives that come from bundled payment makes a lot of sense. I do think it’s tough on the health systems. I mean, everyone in business would most love to have their major payers telling them: you now are gonna take more risk, and we’re going to pay you less. How do you like that for your business? And so that’s what’s going on in the health-care-provider industry
Hal Weitzman: OK. David DiLoreto, what are you seeing of bundled payments?
David DiLoreto: Yeah, I would agree that reimbursement reform is probably the short-term major driver of US health-care transformation right now, whether it’s episodes of care bundles, accountable care organizations, patient-centered medical homes: it’s essentially paying providers differently. To your point earlier, saying that we’re gonna pay you for the value that you create, not necessarily just for what you do. It is a big transformation. It is a difficult one for the industry.
On the other hand, I’m encouraged by the amount of innovation I’m seeing among providers on the front lines as this happens, as they begin to think about an episode of care, not just what we do in the operating room or the 90 days postoperative in my office, but what happens in other places, in the laboratory or in a skilled nursing facility? What happens preoperatively? They’re getting very innovative.
So our orthopedic surgeons in a total joint bundle decided that pain management is extraordinarily important. That can lead to longer hospital stays. That can lead to readmissions. And so why don’t we have a tool in our office where we can identify patients who have knee pain who might be tolerant to their pain medicines, meaning that they’re not quite as effective as they might be. And if they find somebody who falls in that category, they’ll get a pain-management consult preoperatively, not after surgery, and that way the pain-management specialist can design a treatment regimen that will be more effective for that patient.
So that type of innovation really comes from just changing the reimbursement model a lot so that providers begin to work together and collaborate in ways that historically they didn’t.
Hal Weitzman: Well, it seems like this big data revolution of health care is just beginning, but unfortunately, our time here at The Big Question has come to an end.
My thanks to our panel, Dan Edelman, David DiLoreto, Jonathan Silverstein, and Harold Pollack.
For more research, analysis, and commentary, visit us online at chicagobooth.edu/capideas and join us again next time for another The Big Question.
Goodbye.
(light piano music)
Harvard’s Dani Rodrik visits the podcast to discuss changing attitudes toward globalization.
Capitalisn’t: The New Economics of Industrial PolicyNew methods of measuring racism and sexism find a larger, systemic impact.
We’ve Been Underestimating DiscriminationAn expert panel discusses how A.I. is reshaping entrepreneurship, and how founders and investors are responding.
Is A.I. Startup Funding a Rerun of the Dot-Com Bubble?Your Privacy
We want to demonstrate our commitment to your privacy. Please review Chicago Booth's privacy notice, which provides information explaining how and why we collect particular information when you visit our website.