David B. Agus
Professor of Medicine & Engineering, USC
CEO, Emerging Businesses, Philips
How does the masses of data flooding healthcare, medicine and science change our relationship to managing and curing disease? What changes in our relationship to doctors and how they work? Will we cure more disease? How big a deal is data-driven medicine?
Kirkpatrick: The reality is, this next session is absolutely pertinent to everyone intimately. David is so good at talking in common parlance about how our lives are changing in the health arena that I think no one will have any trouble identifying with this session. David Agus is Director of the USC Center for Applied Molecular Medicine, so he teaches both engineering and medicine. But he’s also written a number of books, but two really major books. “The End of Illness,” is his first big bestseller. And more recently, “The Lucky Years: How to Thrive in the Brave New World of Health.” These are books that are really indispensable to understanding how healthcare is changing. And to discuss this with him, we have Krishna Kumar, who is the CEO for Emerging Businesses at Philips.
Kumar: Well, thank you.
So to set a little bit of a context. I mean you’re clearly aware that when we think about health as an industry—while we were talking about banking and financial services—I mean, the top three sectors in the world, which are agri and food processing, the health industry, which is a $9.5 to $10 trillion dollar industry globally, followed by oil and gas—we see that the top three sectors are relatively untouched by technology and IT, right? I mean, computing and connectivity hasn’t really had that much of an impact. So we have, however, really in the last 100 years of medicine, seen significant advances in health. But still, at the turn of the century, we have big, unsolved problems in medicine.
So what I want to do is to really focus our conversation on three key topics. The first is really to set a context for this audience on, what are really for the twenty-first century the big, unsolved problems in medicine. Second, you’ve written prolifically about—and you’ve been an advocate for—really how technology can and should create an impact in transforming health, so I’d like to get your perspectives on that. And I know you’ve spent some time on cancer, which is one of the very big unsolved problems.
Agus: I’m against cancer. [LAUGHTER]
Kumar: We all are, right? And I’d like to understand your thoughts on cancer, which is a big data problem. What are your thoughts and really the role that the Internet of Things can really play, and technology in general, in transforming the problem—addressing the problem of cancer?
Agus: Yes. Those are big topics, but I’m ready. Let’s do it.
Kumar: Great. And then I’ll open it up to the audience for a quick conversation. As we look at the last 100 years, we’ve seen everything from vaccines to antibiotics to cardiac care seeing significant improvement, but that still leaves significant unsolved problems as we look at the twenty-first century. So can you just paint a picture for us of really what you see as the big problems in healthcare?
Agus: There are a couple big problems. I mean, the first one is something that is just near and dear to everybody, because it’s in the news every day over the last several weeks, which is something like Zika virus. Remember, in 1918, we lost 100 million people in the world to influenza. At that time, you weren’t allowed to go to church services, you weren’t allowed to go anywhere to congregate more than three people. That’s how they fought it. Everyone was basically confined to their house for a year. It took, on average, 14 years off life expectancy, one flu outbreak, for the entire world. So things like Zika virus are happening. You’re going to be astonished by what happens this summer. Two hundred and fifty pregnant women have Zika now in the United States. That’s going to go into the multiple thousands and many babies are going to be born without a fully developed brain and it’s going to be scary.
So the world now is flat. It is a remarkable time of people going all over the place, and when you have a think like the World Cup happening in Brazil, everyone congregates and then spreads back. That’s what happened with Zika, and we’re going to see that over and over again. So we as a world have to adapt to the changing transportation, to climate change, to all of those, to adapt to what’s in infectious disease.
Number two is, I think the fundamental problem in medicine is a mistake we made 90 years ago, and it literally has lasted with us. There was an experiment where there were 12 people and they each had a significant cut on their leg. Half of them, they left open to the air, and half they took a piece of bread, they dipped it in water, and they wrapped it around the leg. Well, the people with the bread on their leg healed twice as fast, and it screwed up medicine. That bread made mold, the mold made penicillin, and it spawned something called Germ Theory. Germ Theory said as soon as you know what you’re up against, you know how to fix it. So look into the microscope and see what bacteria it is, you know exactly what to give the patient. And it’s right. The problem is Alzheimer’s, heart disease, cancer are from within, not without. And so the way we’re going to treat these diseases is we can measure the whole system and not the cell. Well, I don’t even know the language to measure a system and even describe it. So that next advance in medicine will no longer be targeting a cell with cancer, or targeting the LDL in cholesterol problems and heart disease. It’s going to be modulating the system. You know, you drop a match after it rains, nothing happens. You drop a match in my hometown of Los Angeles, it goes up in flames. Context matters, and we don’t know how to describe context in health. So as we move into the big data revolution, we’re just getting the low lying fruit. If we really want to get what’s up here, we have to start to describe context in the system.
Kumar: There is also the added problem of costs of healthcare versus these problems to solve. Can you also describe that economic challenge with healthcare as well?
Agus: You know, we in our country have a right to do whatever we want. You can smoke, you can be obese, you can sit all day, and then society has the obligation of paying for the healthcare ramifications of your decisions. Well, first of all that has to stop. The State of California has the lowest smoking rate in the country, yet almost a third of our healthcare costs are tobacco-related. So we can lower healthcare costs dramatically by changing behavior and by starting to tell people—you know, I was at Cleveland Clinic last night, I came back this morning. And at Cleveland Clinic now, if you smoke, you can’t be hired. So they will not hire a smoker. It’s legal. They do it, and it has a dramatic impact. In the county they’re in, the smoking rate went from 25% down to 12% when they instituted that. So you can impact things that can make change. Technology should dramatically drive down healthcare costs. I mean, it’s as simple as that. It’s scalable, and it works. You know, right now, a pathologist—if you have a biopsy, a pathologist takes a needle, they stick it in, and you look under a microscope, and they use pattern recognition, normal, versus what they call cancer. Which is basically, you know, this is the difference of normal and cancer, right? This is normal, this is cancer. The cell is smooshed a little. That’s the state of the art now, and you pay somebody $1,000 dollars every time they do this. It makes no sense. So with AI machine learning, you know, slides can be scanned it and it can basically democratize reading of pathology across the world and bring down costs dramatically. And we can scale that on imaging, we could scale that on many of the things we do to lower costs. And that’s certainly exciting and encouraging.
There’s also the notion of longer you live. The longer you live, in general, healthcare costs go down in our country, not up. If you live past the mid-eighties, we don’t put you in an intensive care unit, we don’t put you on a ventilator, these crazy things that cost the individual and society a tremendous amount with almost no benefits. In general, the younger you get a disease, the more expensive it is. And so the good thing is, if we make people live longer, costs will come down.
Kumar: That leads me directly into some of the stuff that you’ve written about, which is the big challenge of big data. Big disconnected data, when it comes to healthcare, right? How do we tie the data together to draw meaning out of it? Where would you start when you begin to address the problem?
Agus: So, you know, ovarian cancer, a horrible killer in our country. It turns out, if a woman is on a beta blocker, which is an inexpensive generic drug to treat blood pressure, she lives four and a half years longer than if she isn’t with ovarian cancer. Where do we get that? It was a big data study. Somebody just look at large cohorts of ovarian cancer patients across the world and came up with that. We would have never had it in biology.
So big data can be transformative. The problem now is that most of the data is collected without context, in a non-standardized way, and using non-standardized data elements. And I think that’s what’s challenging. You know, if you call it a broken leg, and I call it a fractured leg. Our databases don’t talk. And so in order to get real utility, we all have to standardize our things. Which seems easy. You know, in the Department of Commerce, there’s the National Institute of Standards and Technology, whose job it is to say, “Hey, here’s the standard. Let’s do it.” And they classically take it from the private sector. The problem is, in our field, we have standards. So somebody needs to stand up—and there’s a lot of careers involved, because if you spent your career working on one and everybody just said use another one, you’re pissed off. But somebody’s got to have the leadership to do it, and we need leadership in healthcare, which we don’t have right now.
So big data will transform. At the same time, you know, you’ve heard a lot about security. Well, security is going to be an issue. You’re going to see it soon where you end up in an emergency room because you have a cut on your arm. You go there, they look at your medical record, they give you an antibiotic, an hour later you’re dead. Because what happened is people are breaking in, obviously, to the Blue Crosses of the world, they’re holding hospitals hijacked, where they have to pay a ransom of millions of dollars in order to get access to their EMRs, but they’re going to go in and change your medical record. So they took your allergy to penicillin and they took it out, so when that doctor saw you, they gave you a penicillin. Or they could take a Donald Trump of the world and give him a disease and actually that will go proliferate out there. Or take a CEO and say they’ve been promiscuous with regard to sex—you name it, that can happen now and that will happen. We have to separate that out from the heroes of the world who donate their medical information and their data to actually be mined and learned. I think both can happen concurrently, and it needs to. But big data will be a revolution in our space if we use it correctly.
At the same time, computers won’t take over. AI will not treat patients. It will never happen. I guarantee you that. There will always be an art to medicine. And when you start to see—you know, in the medical literature now, when the computer looks at things in the medical literature, they look at everything relatively equally. You could put a hierarchy based on the journal, etcetera. But an amazing study came out. If you look at the medical literature over a period of multiple decades, 80% of the studies turn out to be incorrect. So you cannot just rely on a study that will change context and be interpreted differently over time. You have to still rely on that art of medicine.
Kumar: So if I put myself in the shoes of the technology sector, which is looking at how do I pick up big problems to solve for healthcare, what are the two or three big areas that you would recommend for—
Agus: Well, first of all, you frame the question right, is you want to solve problems. Most of the technologies I see are not solving problems. So you’ve got to say what is a fundamental problem? And so a fundamental problem is electronic medical records, right? Data in it is unusable. They were written in the ’80s, and if you guys went in and tried to use my electronic medical record, it’s like using MS-DOS. You would be shocked at what’s in there. And so we have to make it usable where there’s an interface, both for the physician and the patient, so they can improve. You know, when you educate people, you get compliance. When they understand why, it happens. Doctors in today’s world are losing the amount of time they can spend with a patient to do it, so the transformation of using technology to educate will be very, very powerful.
You know, a technology comes out every month that people are excited about. Whether it be the new FitBit or the new Apple watch, etcetera, engagement is classically several months and then it goes away. People don’t care about it anymore. And so the main reason for that is that there’s no context, or the information isn’t put in a format that can benefit us over time. You know, if I said to you, “Are you healthy?” you don’t even know how to answer that question, right? Is there a blood test for health? Is there a marker for health? Is it how you look? Is it how do you feel? How do you really define what that is? So we need to develop ways to have these metrics, because with these metrics, we can actually get engagement.
You know, my daughter got her license to drive a car, and our insurance company—and insurance in California is through the roof. They said, “If you put this little device under the steering wheel, we will give you a significant discount on your insurance.” So every time she slams on the brake, I get an email. Every time she accelerates, I get an email. Every time she goes outside a zone or past a time of night, I get an email. And that does nothing. But at the end of every week, she gets a score of 0-100, of how safe she drove. And she competes with her friends. “I got an 88, what did you get this week?” “I got an 89.” “Well, how did you do it?” “Well, after the stop sign, you slowly accelerate instead of doing it quickly.” And they game the system to drive safer, and the engagement is staggering. I talked to the CEO of the company that made it, and this was an afterthought, a little thing, and everything they had planned for years didn’t work. This one thing did.
But we need to do the same in the sphere of health. We need to look at how we get engagement, how we get consumers and doctors to do it. You know, an amazing study came out—this is one of the things that really influenced me is that there are 300,000 appendectomies done every year in the United States, you know, where you take out an appendix for appendicitis. In the 1950s, this Russian explorer who was a doctor was in Antarctica, and there’s an amazing picture of him taking his own appendix out. Because we know—we’re told that whenever you have appendicitis, you have to take it out right away, 24-7, that doctor needs to rush in and take it out. Well, in Europe they did a study where they gave people with appendicitis antibiotics. Seventy percent of the people never needed surgery at all, and of the 30% who eventually needed surgery, there were zero complications for waiting. Well, we’ve been doing this for 70 years. Nobody even questioned it. But the amazing thing is this study came out a year and a half ago. The rate of appendectomies in the United States still hasn’t changed. So on average, it takes a doctor 12 years for 50% of them to adopt a new technology.
So healthcare will change from the ground up, not the top down. And the problem is, the companies, the Philips, the GEs, the Siemens, of the world keep going to see what the doctors want, and a few doctors say it and then the rest don’t want to do it and nothing gets enacted and nothing happens. We have to change the cycle.
Kumar: I want to quickly touch on cancer. How are we going to solve the problem of cancer? What role do you see technology playing?
Agus: Well that obviously is a very broad question, but all of a sudden, for the first time in my career, I can walk into a patient’s room with a true sense of optimism. A year ago Jimmy Carter said, “I have melanoma that went to the brain,” and he was a hero for being transparent about his disease. And literally, the year prior that was a death sentence. He got a drug that blocks the “don’t eat me” signal on the cancer to allow his own immune system to come in and eat the cancer. And three weeks ago he announced he’s cancer-free and no longer on treatment.
So certainly, there are advances happening. There’s immunotherapy, and there’s what we call molecularly-targeted therapy. It’s the notion of identifying an on switch that the cancer has an addiction to and giving drug that can block that. Well, in order to do that, you need DNA sequencing. So right now, DNA sequencing is what’s going on in our space. Every patient I see, we sequence their cancer, and in about 40% of cases, it can guide what I do with treatment.
Well, the next generation will be looking at the proteome, the proteins in the blood, which is conversation in the body. Well, one image of the proteome is about 60 gigabytes of data for one patient. You couple that with about 200 gigabytes of data for a whole genome sequencing. And then there’s going to be the microbiome. So it turns out that you have tenfold more bacteria in you than human cells in you. So these bacteria control your metabolism, what you eat, many things about you. And right now, for the first time, we’re starting to sequence them and look at them. You have 20,000 different bacteria in you. But an amazing study came out last year where they took 20-year-olds and they gave them artificial sweeteners, right? When they came out, they were the greatest food ever, because they hit your sweet tooth and they weren’t absorbed. And after two weeks, they all had markers of diabetes. They gave them antibiotics followed by the artificial sweeteners, no markers of diabetes. So what they do is they change the bacteria to push your system to diabetes. So as we go forward in cancer, we’re going to modulate this system to how we metabolize drugs, how the cancer grows, and change that system.
You know, if I take a woman with early breast cancer, pre-menopausal breast cancer, and give her a drug that builds bone, that doesn’t even touch the cancer, I reduce the rate of recurrence by 40%. So a drug can improve survival in a cancer that doesn’t even touch the cancer, by changing the system. We’re learning ways of changing the system now, whether it be the microbiome, whether it be the proteome or others of those. The problem is it’s staggering data, and they’re data right now in silos. The future is molding that data together into a model, and that’s going to be challenging, but it’s going to be the next advance in what we do.
Kumar: Fabulous. Very inspiring. I want to now turn it over, David, to the audience.
Audience 1: I was struck by the point of 12 years for doctors to update their thinking. Is there any way to shorten that?
Agus: Well, we’re pig headed, we’re stubborn, and the way reimbursements work now, there’s very little incentive to change. You know, if you’re a surgeon and you get paid $25,000 dollars for doing an operation, you get paid nothing if you say, “Listen, let’s observe the condition and we’ll wait a couple years and see if it changes.” And so we’re incentivized to treat. So whether you bring in a technology that makes it better to understand a disease or better to follow a disease, we’re not really incentivized to use it. So you have to change that system. You know, the majority of healthcare costs are to treat disease. Almost nothing is pushed in the front to the prevention side, where a lot of innovations are happening. So the system needs to be changed, but in order to get normative behavior change, you need leadership. I challenge anyone in the room to say right now who is the Surgeon General, to say who is the Secretary of Health and Human Services. Maybe a few of you know. In the days of C. Everett Koop, everybody knew. And so in order to get normative behavior change, you need leadership. We sorely need leaders. When the greatest leader in the healthcare domain over the last decade is the former mayor of New York City, you know there’s a problem in our space.
Audience 2: So I think we do not use technology enough in medicine today. I think we have a lot of functionality sitting out there that we’re not using because, like you say, we just keep doing things the way we’ve always done them. For example, I think when you go to a doctor in most countries in the world, there’s a chip on a card somewhere where they can see your records for the last 10 years and make correlations according to how you’re proceeding with any kind of risk assessment. Additionally, what I’m hearing very hot in the medical industry are garments and clothing that are being made that can monitor your body function, potentially going up to some type of application with a dashboard where somebody calls you and says, “Hey, those French fries you had last night had a lot of salt, your blood pressure is going up right now. You need to do something immediately, before you have a situation,” like that. I mean, across several conditions. I mean that’s only sounds too easy, very uncomplicated, but I don’t think we’re looking to be preventive enough. We’re looking to treat the disease, we’re not looking to prevent the disease, and I think the technology and some of the promise of what we have out there today really needs to come to market a lot faster and be adopted a lot quicker by the medical community.
Agus: I don’t even know how to answer the question, because I’m not sure what the question was. But I think you’re right. Everything you mentioned, and most technologies as I see, they’re not solving problems. So over and over again, I get shown a garment. I get shown them literally every week. But nobody has shown me how that actually makes someone better and changes the disease course. They tell me how it could, and how, “Data is great, use it.” But the challenge is, for the companies out there, is take a problem and show how it improves things. If you show how it improves things, I’ll help you get adoption. So will every other doctor. But you know, monitoring things without showing that there’s a utility to it really doesn’t help us move the needle. Yes, you can show me that the French fries may correlate with a little bit of blood pressure change, but you’ve then got to show that modulating that blood pressure change has an outcome with heart disease, with stroke, with some of those. You know, showing change in number doesn’t mean anything to me, and the companies out there are focusing on that over and over again. Show that you change an outcome.
Audience 3: I love the idea that you mentioned about healthcare shifting things from the top down method of innovation to bottom up. With this idea of hacking health, where you have people now taking control of their own situation, you have websites like Patients Like Me, you have groups who are coming together and trying to think about ways to address either using technology or else, like crowdsourcing their own problems. Do you see a movement in that direction? Because that direction gets around the whole FDA—there’s all kinds of things that that approach could benefit from. Do you see that direction?
Agus: I’m not sure it gets around the FDA. I mean, so—because there are recommendations made classically that will go under regulatory oversight. But I do think it’s a staggeringly important movement. Remember, the doctor’s office now, you come to see me, I draw your blood, I check your blood pressure, I listen to a bunch of things, and then classically, a few days later, he or she calls the patient with the data. Well, very soon, we’re going to collect our own data and then going to the doctor will put it into context and actually have a conversation about the data. And those are the areas, whether it be crowdsourcing and discussing with others before you go to your data, that you’re going to get a lot more information and go in with the right questions and I think generally improve healthcare. And that’s that movement from the bottom up. But there will be, for most of these things, and there should be, some regulatory oversight. You know, the Patients Like Me concept is great, but it’s very easy for a salesperson for a drug to get in there and pretend to be a patient and talk about an experience, and push patients one direction or another. It’s very different for patients to think that the condition that they have and somebody else has are the same, and they could change something based on what someone tells them, but they have a nuanced difference there. And so we really need to be careful. But if it’s done right in a moderated fashion, I think it’s very powerful.
Audience 4: What you’re saying is making the assumption that humans are logical, that we use data to drive our decisions, but I see a guy over here is wearing two watches and there are motivations that extend beyond just logic. It’s perhaps, in the opening they were talking about what’s facile, what’s safe, what leads to trust with product development, but the engagement factor, the psychology of fun, how does that drive—and how can that become more incorporated into medicine and the product development when you’re thinking of this moving forward in a predictive nature for longevity of a product in the healthcare sector?
Argus: It’s a critical question. I think that, you know, in my mind there are two answers to it. The first is information. One of the big technology companies said to me, “Help us change our cafeteria.” So we did this stupid idea of having the salad subsidize the burgers. So it was half as expensive for salad, twice as much for a burger. People got pissed off, nobody changed what they did. It was uniformly negative. Then we did something clever is we put a little sign next to each thing in the cafeteria just explaining what was good and what was not as good with the product, and behavior changed literally overnight. You know, with physicians, we put—in our hospital, I started to put at the bottom of it, when a doctor biopsies—the patient biopsy rate, how many biopsies are positive. So if one doctor biopsies, if 80% of his or her biopsies are positive and the other is 20%, one of them probably biopsies a little too much and one not enough. So we took the doctor positive biopsy rate and the community positive biopsy rate on the pathology report that the doctor got, and when we started, there were three and 3.5 standard deviations. After 30 days, we were down to 0.8 standard deviations. We didn’t tell doctors what to do. We just gave them information and feedback. And so that’s critical.
And then we need to start to develop—you know, there was a concept call real age a while ago, where you had one number that could look at a lot of your healthcare parameters and give you that metric. I think we need to improve those concepts and have actual numbers so we can game the system and make it fun. So you can start to say, listen, this week—just like my daughter and driving—I’m an 88 on health. I want to get to a 91. And so I think if we could pull in data from various sources into one number, we can start to compete. And then at the same time, reward people, whether it be the insurers or whether it be the companies. You know, there’s two health plans in Florida that made a deal with the supermarkets and when you buy 80% of the foods in their, quote, “good” category, they give you a discount on your health insurance premium that month.
And so we can reward people for what they do, and it’s another way to game the system in a positive way and to really show them. Remember, one case of colon cancer costs a company $160,000 dollars. One heart attack costs a company $92,000 dollars. So there’s a lot of incentive, on a corporation part or an insurer part to make people to do the right thing.
Kumar: If I were to ask you one final question, your one wish for technology to address the problem of health?
Argus: I think the one wish is, you know, go to the clinicians, the physicians out there and say, “What are the problems?” go to the patients and say, “What are the problems?” and make a hierarchy of the problems and focus on the problems. Don’t just focus on the technology.