Scroll Top

Tom Davenport on the Analytics Gap in Healthcare

Episode Summary

Harry’s guest this week is Tom Davenport, who argues that the healthcare industry is way behind in its use of big-data analytics software to make smarter decisions about business and patient care. “This is a period of lots of opportunities to use new technologies to change healthcare, and God knows we need it, from a value-for-expense standpoint,” Davenport tells Harry. “But we’re not really at the point, at least on the clinical side yet, where we see a lot of direct applications. We’re still in the age of compiling transaction data. We haven’t used it much yet to make decisions and take actions.”

Episode Notes

Tom Davenport knows analytics, big data, and AI—he teaches executive courses on the subject at Babson College, Harvard Business School, the Harvard School of Public Health, and the MIT Sloan School of Management, and is widely known for his books on analytics and AI in business, Competing on Analytics (2007), Only Humans Need Apply (2016), and The AI Advantage(2018).

Davenport notes that a number of life science startups are attempting to use machine learning, big data, and AI to reinvent drug discovery (a subject thoroughly covered in previous episodes of MoneyBall Medicine). But in other areas, progress has barely begun. A few startups are trying to bring machine learning into the world of providers and payers, to offer insight-based recommendations about care gaps and treatment. And a few researchers are studying the use of deep learning for pattern recognition in radiology and pathology imaging. But substantive advances are years away.

On the clinical side, Davenport says, “The biggest changes are in the institutions that have more data—combined provider/payer organizations like Geisinger and Kaiser—who absorb the risk of care and need to make informed decisions about it, and are more focused on treating the entire patient and keeping the patient as well as possible. But even there it’s still early days.”

Healthcare organizations that haven’t already started to implement analytics may never catch up, Davenport warns. “This is not an area where it’s going to be successful to take a fast-follower strategy because it requires so much data, so much learning, and so much trial and error over time.”

This episode is part of a special series of interviews with speakers at the AI Applications Summit produced by Corey Lane Partners. Check out the full show notes and other MoneyBall Medicine episodes on our website.

Please rate and review The Harry Glorikian Show on Apple PodcastsHere’s how to do that from an iPhone, iPad, or iPod touch:

1. Open the Podcasts app on your iPhone, iPad, or Mac.

2. Navigate to The Harry Glorikian Show podcast. You can find it by searching for it or selecting it from your library. Just note that you’ll have to go to the series page which shows all the episodes, not just the page for a single episode.

3. Scroll down to find the subhead titled “Ratings & Reviews.”

4. Under one of the highlighted reviews, select “Write a Review.”

5. Next, select a star rating at the top — you have the option of choosing between one and five stars.

6. Using the text box at the top, write a title for your review. Then, in the lower text box, write your review. Your review can be up to 300 words long.

7. Once you’ve finished, select “Send” or “Save” in the top-right corner.

8. If you’ve never left a podcast review before, enter a nickname. Your nickname will be displayed next to any reviews you leave from here on out.

9. After selecting a nickname, tap OK. Your review may not be immediately visible.

That’s it! Thanks so much.


Harry Glorikian: Welcome to this special series of Moneyball medicine focused on AI machine learning and analytics applied to drug discovery and development. This special series was recorded as part of the AI applications summit produced by Cory lane partners. I’m your host, Harry Glorikian. In this series, I will interview different speakers from the event and we will hear their experiences.

We will dive into the challenges and opportunities they’re facing and their predictions for the years to come welcome to Moneyball medicine.

My next guest is Tom Davenport. He is the president’s distinguished professor of information, technology and management at Babson College, the co-founder of the International Institute for analytics, a fellow of the MIT initiative for the digital economy and a senior advisor to Deloitte analytics. He has written or edited 20 books and over 250 print or digital articles for Harvard business review, Sloan management review, the financial times, and many other publications.

He earned his PhD from Harvard university and has taught at the Harvard business school, the University of Chicago, the tuck school of business, Boston university, and the University of Texas at Austin. One of HBR’s most frequently published authors. Tom has been at the forefront of the process, innovation, knowledge management and analytics and big data movements.

He pioneered the concept on competing on analytics with his 2006 Harvard business review article and his 2007 book by the same name since then, he has continued to provide cutting edge insights on how companies can use analytics and big data to their advantage and then on artificial intelligence, Tom’s book co authored with Julia Kirby, Only humans need apply: Winners and losers in the age of smart machines offers tangible tools for individuals who need to work with cognitive technologies. In his latest book, the AI Advantage: How to put the artificial intelligence revolution to work, he provides a guide to using artificial intelligence technologies and business.

Tom, welcome to the show. 

Tom Davenport: I’m happy to be here. Thanks Harry. 

Harry Glorikian: Well, it was great to see you at, uh, the recent AI biopharma conference a few weeks ago. I could see you intently sort of listening to all the different, um, you know, discussions going on about how AI was being used and machine learning and where it was being used, uh, you know, within the life sciences or pharma, I should say.

And you know, this show is, is generally bent towards healthcare and life sciences. And so I love it If you could take your vast experience across different industries and see if you could bend towards, uh, our world of healthcare and life sciences, when, uh, when we’re interacting on this. Cause I’m sure you can talk about this world from, from multiple angles.

Tom Davenport: Yeah, sure. I’m happy to- I do a fair amount of work with providers and payers in healthcare. A little bit less with life sciences and most of the Life science that’s work. I’ve done has been with commercial analytics groups within pharma companies. But, um, you know, I kind of dabble in some of the other areas too. 

Harry Glorikian: Well. It’s interesting because I, you know, I’ve always looked at this capability as a tool and it, and you know, I guess it depends on, you know, if it’s a wrench, it’s a screwdriver, it’s a hammer, but it can be used in a lot of different ways. And some of them, you know, have, have significant overlaps. So, you know, you are known widely in the field for your work on analytics.

You know, if you were to try to summarize some key takeaways for people in the healthcare world, what, how would you, how would you summarize that? 

Tom Davenport: Well, um, you know, it’s interesting. I teach every year, a course at the Harvard school of public health on analytics and AI in healthcare. And, um, You know, despite all the changes in technology, um, my presentation doesn’t change, change all that terribly much because in terms of day-to-day clinical practice, there haven’t been that many changes.

I think, um, this is a period of lots of, uh, opportunity to use, new technologies to change healthcare. And knows we need it, uh, from a value for, um, expense standpoint, but we’re not really at the point in, on the, at least on the clinical side yet where we see a lot of, of direct application, you know, we’re still in the kind of, um, age of compiling transaction data, mostly from, you know, electronic medical records systems.

We haven’t used it much yet to make decisions and take actions. 

Harry Glorikian: In your book, you describe, you know, let’s say the workplace of the future that will be augmented by the introduction of AI and machine learning. You know, can you describe where, where you see this happening today? Um, you know, maybe some examples of, of where that would be and where you see the most significant areas of disruption by this technology.

Tom Davenport: Yeah. I mean, um, as I say, I think we’re poised on the edge, certainly in life sciences, you have, um, a number of mostly startups. I mean, you know, more about this than I do Harry, but mostly startups who are really trying to reinvent the whole process of identifying, um, potentially useful molecules with, um, AI and machine learning, um, in, um, the provider and payer context.

I think there are certainly startups. I was actually just talking with a west coast startup that is, um, trying to pull together all the different types of data about patient care, um, EHR data, and, um, data from outside of a hospital by a, you know, individual, um, a physician’s office or by a clinic. Um, add that to social determinant data.

Um, maybe even data from, uh, different lab systems and so on, maybe from your fitness tracker and then start to make insight based, um, recommendations for what a clinician should do. You know, what are the key care gaps and, um, uh, what are the primary issues, um, with this particular patient? Um, we don’t have that now, but, um, this company is creating it and they, you know, they have 35 customers or so on, so they’re starting, but it’s a really big task to integrate all of that information and kind of change the way clinicians do their work. Um, so I think it’s, it’s going to be an interesting period, but one that’s going to take quite a long time to, to settle out. You also see a lot happening in the, um, in the imaging space with, um, radiology imaging, pathology, imaging done. Um, Increasingly well by deep learning algorithms so that, um, uh, a radiologist or a pathologist could at least get a second set of eyes and, you know, perhaps he would get to the point where, um, it’s the only set of eyes.

But, um, again, that’s not in clinical practice yet. It’s still in the research laboratory and my sources, um, in radiology, tell me, you know, it might be a decade before we see much clinical use of that stuff, there are just so many obstacles to overcome in terms of regulatory approvals and getting, you know, uh, a useful inventory of use cases, um, addressing similar things, you know, all the different, uh, radiology pilots address, different aspects of radiology for some it’s, you know, what’s the size of the Legion, the lesions, for some it’s the, you know, the cell structure, et cetera. So a practicing radiologist would need a kind of a standard set of things to work off of. 

Harry Glorikian: Well, let’s, let’s take a step back here and think about this. Cause you were saying providers and payers, and I think that they’re probably almost keenly focused on business processes or places that would make a difference in, in how their operations run or um, where have you seen, you know, the questions or the changes happening on that front? 

Tom Davenport: Well, I, you know, I think the biggest changes are in, um, the institutions, uh, that have. More data. Um, those tend to be combined provider payer organizations like Kaiser and Geisinger and so on. Um, and they, um, absorb the risk of care.

So they need to make informed decisions about it and are, you know, more focused on treating uh, the entire patient and keeping the patient as well as possible rather than addressing, you know, particular, um, diseases that, that the patient may, um, uh, come across. So, um, I think those, those are the places where you see most of the change and even there it’s still early days, but it’s far ahead of the only provider only, only payer kinds of, um, organizations.

Harry Glorikian: But it’s interesting, right? I mean, I think we may say that it’s, it’s a long way off, but I do believe that if you’re not starting to do it now, you’re going to be. You’re going to be way behind the curve from what I can tell, I was talking to somebody the other day, who was saying, um, and this is in astrophysics actually, where they were saying two different groups, same set of data, same group of incredibly smart people.

Um, one uses the tools. One decided to do it the old fashioned way. And the one that did it, the old fashioned way just couldn’t produce the same results and couldn’t keep up and now they’re playing catch up and it’s, it seems like they’re never going to catch up at this, at this point. 

Tom Davenport: Uh, I certainly agree.

And, you know, I say in my work with companies in general, that this is not an area where it’s probably going to be successful to take a path follower or sort of strategy because it requires, you know, so much data and so much learning and so much, uh, um, uh, trial and error over time that it’s going to be really, really hard to catch uo if you’re not uh, being aggressive now, um, now, you know. In pharma, my sense is that’s, um, a little bit different because there’s so much acquisition going on and, uh, companies that figure out great ways to do this. Um, I think are likely to be acquired by larger firms. I mean, you see this already with, um, Flatiron and, and Foundation Medicine being partially acquired by Roche.

Um, those I think were among the more aggressive users of AI and analytics, um, for, um, cancer care. And, um, you don’t see that as much in provider and payer side, I think you have to build it more organically as opposed to buying it. 

Harry Glorikian: Well, that’s interesting. I must expect disruption to come from outside, um, rather than within sometimes, uh, you know, if you look at what’s happening from a value-based perspective where the Walmarts and the CVS’s, and those people are, are trying to change where you might get care.

Um, and how they inter- interact with the populace. Um, I think we’re going to see an interesting shift from sick care to wellness care. And when you have that sort of shift, I think the dynamics of the competitive environment change, 

Tom Davenport: I think that’s true. And you certainly also see it on the part of the Amazons and the Googles of the world.

And obviously those companies are um, much more aggressive in terms of their use of AI and data and analytics than, um, most companies, uh, organizations in healthcare and life sciences. And so. If they do decide that they’re going to enter headlong into these fields. I mean, now they’re kind of sticking a toe in, but if they do decide to address it more directly, I think it’s going to be quite scary for all of the existing players.

Harry Glorikian: If you were to sort of, you know, give somebody advice, like, what are the, what are the essential skills that you think someone in the healthcare arena needs to think about relevant, you know, for the future in this AI augmented workplace. And there’s probably a set of general skills, but you know, there may be different skill sets for different settings that that may be needed.

But if you were talking to a group of people and sort of advising them, what would you say to them are the important areas that they need to focus on. 

Tom Davenport: Well, you know, I, I see a long-term particularly in the payer provider space and life sciences will be affected by this as well. A long-term trend toward, rather than, you know, doing care and then recording it as we’ve done for a couple of decades now, um, we are going to have systems that, you know, have data across a whole variety of, uh, health related contexts and they systems will make recommendations, uh, kind of, um, Do care planning and recommended actions and so on.

Um, and then of course, you know, capture what the outcome of that is. Hopefully they will be personalized to your genome and your proteome and the various other kinds of Omex that might be relevant. And, um, so that means. Hey, you have to understand all of those data domains, um, how they might be combined.

You have to understand, um, the fundamental approaches to AI, you know, is it going to be, um, uh, traditional machine learning that does this, is it going to be some sort of deep learning is natural language processing involved? Can we get by with rules? You know, rules have been the primary basis for care recommendations and clinical informatics now for, uh, a couple of decades, but most people. And I would agree with this, seem to  think that kind of top-down there, there may be some cases where rules are still useful as well, and then start putting it together. You know, I, one of my favorite lines about AI is think big, but start small. AI is fundamentally a you know, it’s a task related technology, AI doesn’t do, you know, um, jobs or entire business processes.

It does particular tasks. And so start stringing together some AI informed, uh, tasks so that you can ultimately have a transformation in, in care. 

Harry Glorikian: Yeah. I think sometimes people, you know, I always hear people expecting AI to do everything and I’m like, it really only needs to do a few things really well, and it can be very disruptive.

Tom Davenport: Yeah, which means that you have to think pretty carefully about where do you want to apply it within your, um, care process or your R and D process or, or whatever? Um, I think, um, most organizations they don’t understand enough about AI to have made an intelligent decision about where do you, where do you put it?

Um, you know, where do you try it out? Um, and you know, when I think in life sciences, and pharma, there, there was this initial flurry of excitement about how Watson was, we’re going to really be the key to identifying new drug targets and molecules. And obviously that’s kind of faded, it probably set the field back by five years or so.

Harry Glorikian: Stepping back you know, and because you’ve, you’ve worked across a number of different areas is how do companies embed in AI strategy into, you know, sort of their organization or into their own DNA where they think about this, not as a point change, but as a systemic change. 

Tom Davenport: Well, you do have, um, a relatively small number of companies that have decided that AI is so important to them that they’re, they’re going to embed it into almost everything they do. And, um, you know, Google, is probably the single best example. I think they coined the AI first term and, um, they, I don’t know, 2016, they had. 2,700 AI applications, um, underway. Um, they don’t even count them anymore, but it’s, it’s really spread throughout the entire company. In healthcare um, you know, I don’t think we have any organizations that are quite that aggressive, but Anthem the largest, the second largest payer in the United States has decided they are going to be AI first. And they’re going to focus on patient wellness and use all the AI they can to implement pieces of that capability. And, um, so I, I think it really demands on the part of any of these organizations that senior executives, you know, they know their business, but they don’t know how AI might fit into it. They don’t know the different types of use cases available. They don’t know what’s practical now and what isn’t. And so it really requires that all of them, all of the senior leaders of the organization, I think get an idea of what might be done with AI.

And then maybe as I suggested earlier, have a broader vision of when care is transformed with AI. Um, what’s it gonna look like? What’s the vision that we’re trying to achieve here. 

Harry Glorikian: So maybe they, uh, I think they all need to go back to school. 

Tom Davenport: Yeah. Um, unfortunately, you know, there aren’t that many places you can even go back to school to learn all of this in, you know, one nice package, but I think, um, universities will provide it before long.

And if you’re really motivated now, you know, there are. Uh, both these sort of, um, Coursera type, um, offerings that you can learn a lot about AI or, you know, you could probably hire somebody to come in and teach you about this. I’ve done it for a few companies, but not so much in healthcare. 

Harry Glorikian: Yeah. It’s, it’s, it’s tough to find the right pieces of information depending on what you’re trying to learn. If, I mean, if you’re not going to code it, but you need to manage it, it’s a different, different skill set or a different set of information that you need to make good decisions. 

Tom Davenport: Yeah. You know, I was talking with this vendor and, um, it’s hard now because the technology categories cut across things that were traditionally, you know, familiar with.

They’re not just data warehouses or data lakes, they are not just AI. They’re not, um, Uh, you know, there is no kind of Gartner Magic Quadrant for all the things that are necessary to move to this kind of, uh, um, um, data and AI based care planning capability. It really cuts across a lot of different categories.

Harry Glorikian: What are your predictions for the future of say, I hate using the word AI, because it’s just one big, giant umbrella of other stuff. And, you know, there’s AI and machine learning, but, but all these analytic capabilities, what do you see predict for the future of its impact on well, specifically healthcare, but you know, even the business of healthcare, if you think about it, 

Tom Davenport: Uh, you know, I think the situation’s going to get a little bit better in terms of your you’re right. We now have these relatively, um, disconnected pieces and we all, we call it all AI, but it’s, um, different kinds of use cases than components and so on, but I think there is a move toward greater integration of those. So, um, I, I, uh, {inaudible} um, AI and biopharma conference, I think you had had to get on an airplane, but I talked about, um, robotic process automation, um, and that’s a fairly boring AI technology.

Um, uh, one of the, the vendors who, um, I supposedly introduced to things that kindly of my {inaudible}. Um, comment, it’s the most boring form of AI, but, um, it is quite useful and can save a lot of money, but it’s also increasingly being combined with machine learning. So as you know, robotic process automation, um, does a, uh, set of digital workflows, um, machine learning can be used to create um, algorithmic based decisions along the way, not just the kind of rules, simple rules that are in an RPA system. Um, we’re already seeing that, um, physical robots are getting more AI, like brains in them, so they can kind of see and perceive and, and act more intelligently. So I think all these technologies are starting to come together.

My guess is, and you know, IBM Watson was a kind of an early pioneer of this, there’ll be all these kind of AI based microservices. And we’ll choose from the ones we need and they’ll kind of snap together pretty well. And so, you know, AI will, I think look more like a coherent whole than a collection of misfit toys that it, that it is now.

And then you have to think about, well, you know, what are we going to do with these things? And, um, I do think that there is an awful lot of boring, but useful AI stuff. I mean, if you think about, um, in supply chains, whether you’re talking about healthcare or life sciences or whatever, um, there’s often a lot of money to be saved in matching up what was supposed to be delivered to you with what actually appeared at the loading dock. And AI is really good for extracting information from one system and comparing it to another and seeing, oh, these things don’t match up really well. So, um, lots of opportunities and kind of, you know, contracts and, um, uh, inventories and so on that, um, can.make a company far richer, if it really exploits that kind of boring stuff, um, there is some exciting stuff as well, like, you know, developing new drugs or, um, treating cancer, but, um, that’s gonna take a long time and, um, I don’t think you should count on a huge amount of financial return from any of those really ambitious kind of moonshot projects in the short run.

And the, you know, other companies that the providers that tried that MD Anderson and Houston is the most prominent example, really stubbed their toe and spent a huge amount of money and got relatively little to show for it 62 million, I think was the final tab of their, their attempt to cure cancer with Watson.

Harry Glorikian: So you sort of touched on my, my last question. I mean, I, I spend an awful lot of time reading, you know, everything that’s coming out of the tech world, you know, new chip sets, new methods of combining analytic techniques. Um, all sorts of stuff that I think will have applicability to the world of healthcare.

And so if you step back for a moment and it doesn’t have to be healthcare specific, but where do you see the industry going? Next with the advancements that we see sort of barreling down the, you know, barreling down the highway. 

Tom Davenport: Well, you know, I think if we’re not careful, Harry, they’re going to barreled on the highway and then all get stuck at the, um, in the traffic jam of implementation.

So. Um, you know, you, you see lots and lots.- this is across all industries, not just healthcare, but particularly true in healthcare. Um, you see lots and lots of pilots and proofs of concept and, um, lab experiments and so on. Um, that, that, um, show that AI really is potentially quite useful in certain aspects of healthcare, what you don’t see is clinical implementations of it yet.

Um, in part, because of regulatory issues and PURPA codes, you know, um, you didn’t work carefully with the different professional associations that physicians have to, to make sure they understand what’s going on. So I think it’s very important for any organization that wants to transform care to start at an early stage and look at that process.

You know, if we’re talking about life sciences, that would be, you know, working with the FDA from the beginning, you know, a lot of people say, yeah, the FDA. It’s very resistant to change in, um, the analytical approaches for clinical trials. And, um, you know, they don’t understand some of the new technologies that are available.

I think that’s, uh, a valid complaint, but smart companies and other industries are working closely with regulators at a very early stage to say, let us introduce you to some of these, let’s have a conference where academics and companies and regulators all come together and talk about this stuff and, you know, can we get these things implemented more quickly than we might otherwise?

So, you know, I think the- as you say, lot of exciting technologies coming out at a very rapid rate, But we need to look at deployment, um, as a, as a really important issue. 

Harry Glorikian: Yeah. I, I think that the ones that are that lag on the, on the deployment side, I just see some of these technologies as if whoever gets it in and gets it right really does have a super headstart. 

Tom Davenport: Yeah. And I think, you know, um, patients and physicians and so on will gravitate toward those organizations because these technologies have a huge potential to transform care it’s um, -and I think, you know, um, we, um, are going to probably overestimate their potential to transform in the short run, but under estimate their potential in the in the long run, they, they, um, offer a huge amount of potential for doing things differently in this field that has, you know, largely depended upon and many cases, um, experience and intuition, believe it or not. Um, you know, database decisions have been in short supply and, and healthcare in the past. And um, AI analytics, big data can, can really change that situation. 

Harry Glorikian: Well, I think it’s, it’s, it’s typical of, of when a new technology comes along. I’m not, I, you know, I can’t imagine that many people accurately you know, figured out the disruption that an iPhone would would have, or some of the other technologies that we’ve seen come through that have caused a sea change in how things are done. And so if I take technology shifts, plus how we’re paying, you know, for care in healthcare, those two shifts at the same time, I think we’ll have more of a impact than I think most people anticipate, 

Tom Davenport: And knows we need it. 

Harry Glorikian: Well, Tom, it was great to talk to you. Um, I look forward to staying in touch and, um, I’m sure that there’s another book or another article coming out soon that we all need to, uh, keep our eyes and ears out for 

Tom Davenport: I’m working on one.

Um, maybe we’ll talk about that in a year or two. 

Harry Glorikian: Okay. Excellent. 

Tom Davenport: Thanks Harry. 

Harry Glorikian: Thank you. 

Tom Davenport: Bye bye. 

And that’s it for this special series of AI and machine learning and analytics. If you enjoyed Moneyball medicine, please head over to iTunes, to subscribe, rate, and leave a review. It is greatly appreciated.

Hope you join us next time until then farewell.



Related Posts