IBM Pitches “Dr.” Watson In Asia To Leapfrog Into Big Data Management: An Interview With PharmAsia News (Part 2 of 2)

IBM’s supercomputer Watson, named after the founder of the company, is thinking about how to transform health care in Asia and ready to take questions.

KUALA LUMPUR – IBM sees Southeast Asia as fertile ground to grab the strands of data that flow across a large hospital or public health bureaucracy and feed them into its supercomputer Watson as a way to control costs and make links that can help in diagnosis.

Company executives say doctors and nurses are drowning in information as drug development becomes increasingly targeted using genetic data while treatments develop at a lightning pace, making it imperative to move away from the old model of passed on wisdom in health care to one that acts intuitively to sift large amounts of data.

While it is billed as the first commercial cognitive computing system and one that managed to win a game of Jeopardy against players of the U.S. television quiz show, Watson’s moves into health care are relatively new.

In May 2011, IBM said Watson had the same level of knowledge as a second-year medical student but was learning and storing knowledge at a pace that outstripped the capability of humans. In addition, the system can understand natural language and find patterns in unstructured information by creating hypotheses into hundreds of queries, collecting the evidence to support or destroy each hypothesis, all of it within seconds.

Watson returns the evidence that these multiple queries uncover, accompanied by a ranking by strength of each. In health care that could be hundreds of thousands of pieces of medical evidence gleaned over years from patient charts and records, medical journals and clinical trials. The results provide physicians evidence-based treatment options all in a matter of seconds.

Farhana Nakhooda, IBM’s Asia Pacific Healthcare Solutions Manager for the Health and Life Sciences team, and Paul Grundy, a medical doctor who is IBM Corp.’s global director of IBM Healthcare Transformation, sat down with PharmAsia News on the sidelines of the Asia Health Care Conference March 21-22 in Kuala Lumpur, to talk about how medical care in the future could work with big data driving the changes (Also see "Could Big Pharma Answer The Call From Chinese Health Insurers For Big Data?" - Scrip, 13 Feb, 2013.). Part one of this interview appeared here: (Also see "IBM Pitches “Dr.” Watson In Asia To Leapfrog Into Big Data Management: An Interview With PharmAsia News (Part 1 of 2)" - Scrip, 31 May, 2013.).

PharmAsia News: How do your hospital solutions, say, for administrative systems, work with data for clinical studies?

Farhana Nakhooda: For example, Singapore has a national electronic health record, which is basically a record from birth until death. They recently announced a RF [request for information] to vendors. They wanted to see what’s out there that can actually look at the full population of Singapore, the longitudinal medical record, and basically do analytics across it, for multiple things. So everything from disease surveillance – looking at trending of dengue fever across Singapore – to looking at trends in chronic disease and even chronic disease management.

For care coordination we’re actually looking at analytics at the point of care. What I mean by that is that if you’re managing a group, a population, and you want to prioritize who I pick first, you can use analytics to show me who are my high-risk patients.

And that requires analytics, to look through all of the patients and the history, and look at patients that went on to become very sick or had very bad outcomes or whatever it may be, so you know that and you can find what are those indicators that made them sick.

So now as new patients come in, you say, okay, this is highly likely this patient – we’d better focus on them, because if we look at their trajectory based on other patients, it’s pretty negative. That’s at the point of care. So that’s the kind of thing they’re looking at.

But in the Singapore example, they’re also looking at financials across the different hospitals – how are they each performing, how are they utilizing their resources, are they cost-effective or not? So from a government perspective, they’re doing it nationally now. Singapore is way ahead of the rest of Asia Pacific, and in fact many parts of the world, because they are so digital. They’ve got a lot of data already. Most other countries are just in the process of redefining some of their healthcare systems, and then defining what they want to collect.

Paul Grundy: I’ll give you a specific example, just from my memory, because I think it helps bring it down to the reader. In one particular (U.S.) state, we were doing some work and helping them build the technology, the services, to be able to get at the data around what we call a medical home. That would be health information, basic information about every patient in that practice, using tools like Patient Care and Insight and registry-type information.

And so all of a sudden, this particular state, this particular entity, had a knowledge base of their population that they never had before. So, a heat wave was going to hit this population. And the last time a heat wave hit, they lost 28 people. Now, they were able to identify in the population who the frail elderly were. And then, with a little technology that IBM just acquired called Curam, which provides social services support in that same state, we were able to identify those folks by looking at their billing records and by looking at their electric power records. We were able to identify those folks who most likely didn’t have air conditioning.

“You can be much more proactive as a public health authority, but you could also be much more proactive as a provider of care. And increasingly, governments are going to want to pay those primary care docs to be proactive” - Grundy

And when you think about that, it’s pretty simple, right? And then those two information sets were merged together, and all of a sudden, the state had the ability to proactively identify who the most likely frail elderly were that were going to be the most likely exposed to that heat wave, and we didn’t have a single death. I mean, that’s the kind of thing – putting it in a concrete example of actually using it, it makes it clear too – because it sounds like mush when you talk about technology and data. I mean, its real information that can impact people’s lives.

If you think about that information that exists now in that population, which Singapore has, Malaysia is developing, if you think about that, and you think about dengue. We now know who’s most likely to be exposed to dengue, or most likely to be affected, in terms of their demography and their living places. You can be much more proactive as a public health authority, but you could also be much more proactive as a provider of care. And increasingly, governments are going to want to pay those primary care docs to be proactive.

PharmAsia News: I want to use that as a springboard to public health spending and Health Technology Assessments that are increasingly in focus in Asia. Governments actually want to spend more on health, but in smarter ways and to find savings. And so the health ministry might have a broad idea, like, we need to cut incidence of a certain disease, or we need to manage this disease better. But what they run up against is the financial part. Is that the kind of macro approach you would be looking at, working with governments to produce those kind of outcomes?

Grundy: That’s the kind of thing that we are really working on and helping look at. And helping them understand that there’s a return on investment that’s quite strong, when you really integrate population management with a delivery of health. And increasingly, that’s the expectation of what the delivery system will be. So buyers of care, be they government, be they insurance plans, really are asking the providers to move away from delivering an episode of care, to managing a population. And we’re beginning to see health and sick care integrated.

PharmAsia News: If governments had this kind of data, would they change the way they bought drugs? Would they select different drugs?

Grundy: What happens in some places where they have the data like this is that all of a sudden, the contract with the drug company, with the drug manufacturing company, is no longer about providing a pill. It’s providing end-to-end management of that disease, using that pill. And increasingly, pharmaceutical companies are talking about getting into that business.

“What I see the pharmaceutical industry increasingly moving towards is no longer the blockbuster model, but how are we part of the total health solution?” – Grundy

For example, a solution around not only pills that you take to control your diabetes, but solutions in which monitoring that the pill has been taken is part of what government is buying from that pharmaceutical company. It’s quite an interesting transformation that’s occurring around the world.

PharmAsia News: How would that work?

Grundy: In this particular case, the pharmaceutical company actually worked with IBM to create a solution in which, when you popped a pill from the dispenser, that information gets Wi-Fi’ed into your cell phone, which gets Wi-Fi’ed into your personal health record and becomes actual information. And there’s now devices that actually activate when the pill hits the stomach acid. So increasingly there’s interesting ways of doing that.

And what I see the pharmaceutical industry increasingly moving towards is no longer the blockbuster model, but how are we part of the total health solution, so, much more end-to-end in terms. Increasingly when I talked with them, they’re really thinking about that.

Nakhooda: I know of a case where one of the hospital groups in Singapore was saying that even within the wards of their own hospital, they noticed that the cost of drugs for – I think if I’m not mistaken, it was asthma or something – was very high in one ward compared to all the others, for the same exact treatment, the drug costs. Then when they drilled in, and this is what you can do with analytics, you can look to the trending. And when they drilled in, they found out that it was one department that had used a different drug. It was more expensive. But the patients didn’t come back; whereas the other department’s route, using less expensive drugs, the patients kept coming back. So I think this is where the outcome data, once you have that, you can basically then really compare, interestingly, if the medication is working or not, because just because it’s cost-effective doesn’t mean the quality is good.

PharmAsia News: Would there be other applications for health authorities to make sharper choices, say for efficacy?

Grundy: We’re beginning to understand re-hospitalization, emergency room visits. There’s one solution with asthma that I saw in one geography where it was a pharmaceutical company that provided that solution, and it made it available for all asthma medication. But they were one of the leading asthma medication providers.

And it was a simple solution, where the information around whether the patient picked up their medication in their drugstore was transmitted to the druggist, and the druggist reviewed that information, and if it was of enough concern, they notified the doctor that the patient wasn’t taking the medicine. So if you stop and think about that, everybody but the doctor knew that the patient wasn’t taking his medicine. The patient knew he wasn’t taking it, the druggist knew he wasn’t taking it. But that information wasn’t getting into the clinical hands. And that sort of information, we began to see the rate of hospitalization decrease by about 40%.

Another solution that is really cool is that inhaler devices, hooked up with – by the way, the same technology that we use to track the train wheels, whether they’re hot or not hot, whether they should be replaced or not replaced, which keeps the trains in the United States from derailing … but the same sort of technology, hooked up to a GPS device, which that solution has, which monitors your air quality. And we can now look at when people use the asthma inhaling device.

Nakhooda: In fact, the other area which is a big focus, of course, is this concept of personalized medicine – being able to take genetic information, link it with clinical information, and look at if there is a particular genetic marker that shows the Herceptins of the world, for breast cancer?

It’s the move away from blockbusters; they’re very expensive. Can you look at cheaper therapies? Or what about the rest of the population that doesn’t hit the blockbuster model? Well, that’s where analytics is important. But then that’s again a government decision. It could cost a lot for a pharma to build any drug. And then if you look at the ones that are more targeted, is that helpful or not?

The other area which will be quite interesting as well is this whole area of social media, because that is data as well. So, patterns of what people take. Even my own example, a crazy example in Singapore where I literally thought I had a brain tumor. Turned out it was a side effect of a drug I was taking for my stomach. And the only way I found out about it is I Googled it and I found all this information. And in their site, they’re now saying that 60% of people that take this drug have this side effect. And this one side effect was like intense migraines. And basically, people that were taking it had actually gone and done MRIs and thought they were literally dying, so I was like, “That’s me.” So I just stopped taking the drug and I was fine.

“In the U.S., we’re actually getting clinicians to train Watson to think like a doctor, and to understand the medical context” –Nakhooda

But if I hadn’t gone online and seen this information … I think this is going to be the future, because people just say, “Don’t take this drug.” You know what people are like, they say whatever they want. But at least it gives people – it would give pharma – a very good indication of what people are thinking about their drug. And that’s called sentiment analytics. It’s being used in the world for elections, to see what people think. But you could actually use that to see what consumers think about your drug.

PharmAsia News: So, how do you separate the wheat from the chaff?

Nakhooda: Watson does three things: It’s got natural language processing, which is the ability to understand the data.

Grundy: And speech.

Nakhooda: And speech, that’s right.

Grundy: And written text.

Nakhooda: So any kind of data, it can understand it.

Grundy: It can read 300 million pages of text in like 0.3 seconds or something. It’s just incredible.

Nakhooda: It’s incredible. Massive parallel processing. And then it’s got, which is what makes it all unique, is a lot of algorithms and knowledge to understand context. Because basically, as we said, of the example, which I love, is that if you ask a computer today what’s two plus two, it’ll say four. If you ask Watson what’s two plus two, is two plus two four, it’ll say maybe.

PharmAsia News: My son would say 22.

Nakhooda: Well, there you go. So your son’s like Watson. But the reason it says maybe is that, depending on the context, if you’re a doctor, two plus two is two parents, two children. If you’re in the automotive industry, it’s the front two wheels and the back. So it depends on what you’re referring to. So if it’s – is two plus two four – Watson would still say maybe.

In the U.S., we’re actually getting clinicians to train Watson to think like a doctor, and to understand the medical context. And so that’s all built in. What happens is, when Watson makes a decision, it gives it a confidence level. It’s not just saying some druggie out there took a drug and said, this is the best thing ever, it’s not going to take that as gospel. It’s going to look through the information, and the more information it has, the more confidence it will have to say, this is the correct answer, or this is the correct treatment path.

Grundy: Or, I’m 87% confident that this is the correct answer.

PharmAsia News: Who owns the data?

Nakhooda: That’s a very good question.

Grundy: We think increasingly, societies that are further along in this road are defining the data as being owned by the patient. The example of that is this Danish example, and they probably have the most integrated information at the primary care level of any country on Earth. And they’ve defined it legislatively as the patient’s data. And by the way, in Denmark, the physician has to make their information about the patient available to the patient within two weeks.

Nakhooda: It comes down to country decisions, typically. In the case of Singapore, I remember when I used to work in biomedical research, and when they did a survey to say, ‘Who would you trust your data with most?” And it was sent out to the population, probably one of the only countries in the world that came back with “the government.” Because they felt like there are no vested interests from the government.

Grundy: They also figured that they already knew everything.

Nakhooda: And this was specifically genetics. So it was even to the point of DNA. This was a long, long, long, long time ago. And it’s a good question, because doctors sometimes think they own the data. And so it makes it challenging, because to do this, that comes back to policy and the way people are looking at care. Because typically, many doctors would say, this is my patient data. I’ve collected it.

And even sharing it becomes challenging, so some of the challenges we’ve had in many countries with these projects have nothing to do with IT. It’s actually to do with getting people to share their data.

PharmAsia News: The greater public good versus confidentiality?

Nakhooda: That’s right, exactly. You could anonymize it, but even then, the question is, do I have the right to get access to that data. And Singapore is very progressive that way. They think that it’s for the better good – and they will position it and market it that way. And as long as they don’t use it for anything negative, I think that people will be fine, and I think it’s about seeing value from it.

Grundy: I want to talk a bit more about Watson. I think that it’s really exciting, because it’s cognitive. It’s no longer dependent upon rules-based thought processing, which most computing is done when you talk about health care. And health care doesn’t respond well to rules, or only to rules, because human behavior and the human condition doesn’t follow rules very well. We don’t do things necessarily logically.

“It’s really about having the knowledge to know what data to collect, how to integrate it, how to structure it, and how to basically make it useful to the various stakeholders” – Nakhooda

And so that breakthrough, the ability to handle parallel processing, massive amounts of information, that is cognitive. It literally learns from its experiences. It takes it past the rules, and looks at what works. We had an example of a case in our learning process where the person had a pretty rare bleeding disorder, and there were eight articles that were written about this disease, because it’s a relatively rare one. Half of them said you should Heparinize, and the other half said you shouldn’t. And so it’s quite confusing as to what you should really do when you look at the literature.

But if you look at the known cases that existed out there, if you went to – as we did, we went to a base of knowledge – it turns out that overwhelmingly, the folks that were Heparinized did well. So it begins to go beyond what the eight studies are in the literature, and begins to look at that whole vast information of, what does this case look like compared to a thousand other cases or a million other cases with the same information. It’s a huge breakthrough.

And fundamentally, that breakthrough is going to change how a practice is practiced. It’s going to change what a doctor is. A doctor will no longer be a master builder.

Nakhooda: Or a data repository.

Grundy: A data repository. The need to take a human being that’s smart and put him in medical school and use their brains to store information, and as the information becomes more specialized, break it into further fragments and bring them, like they’re building a cathedral, to a place at the point of care, to deliver that information, that model is going to be broken. And like every other process, there will be a plan for your patient. Does that make sense?

PharmAsia News: Have you told the American Medical Association this?

Grundy: We talk about it all the time. And by the way, we have medical advisers working with us from all the medical societies who are very excited about this, because they of course realize that their ability to store information and retrieve it is relatively limited as a human being. I think there’s like seven things that you can really manipulate in your head at one time, and Watson can manipulate millions.

PharmAsia News: Obviously getting the information to the right place at the right time is key.

Nakhooda: That’s really where it comes down to the implementation and services. Because as you said, these projects are not really about the products alone. It’s really about having the knowledge to know what data to collect, how to integrate it, how to structure it, and how to basically make it useful to the various stakeholders. Because typically when you do an analytics project, you’re not just doing it for one or two users; you’re doing it for potentially an enterprise.

And so the services is that glue that comes in, understands the data, understands the end requirements, the outcomes you’re trying to achieve, and then goes in and does all that integration work. And that’s typically something we tell our customers too, because it isn’t as simple as buying software. You have to do that work of actually understanding and road-mapping out, because as I said, a lot of times when you go around and ask the clinicians or whoever, “What is it you want out of the data,” you find that they’re not even capturing that data, or it’s sitting in notes. And so as Paul would say, you have to get that data structured first.

And what’s cool about Watson – well, we think it’s cool anyway – is it’s really at the point of care, and that’s where people typically think of analytics as something that you trend, and you watch, and you dashboard, right? Whereas this is something that a doctor can have alongside him and say, right at the point of care, this is a potential good treatment path, based on, as Paul said, millions of pieces of literature, and even the evidence on that individual patient. You bring it all together and it will guide you.