Small Data’s Big Impact on Healthcare — Full
Deborah Estrin, Professor of Computer Science and Director at the small data lab@CornellTech, chats about how “small data” — our digital patterns and interactions — can help physicians by creating a more meaningful representation of our personal health.
Key takeaways from this episode of StartUp Health NOW can be found here.
[00:05] Unity Stoakes: Welcome to StartUp Health Now! The weekly web show that celebrates the Healthcare Transformers and changemakers reimagining health. My name is Unity Stoakes and today we are here at the Wearable Tech and Digital Health Conference in New York City. We’ve got a very special guest, Deborah Estrin who is the director of the small data lab at Cornell Tech where she’s also a professor. Stick around. It’s going to be a great show.
[00:29] Intro Music
[01:07] Unity: Welcome to StartUp Health Now! Deborah. I thought we’d start out just with a big question which is, everybody knows what big data is. What is small data?
[01:19] Deborah Estrin: Absolutely. Big data is very important. It’s getting a huge amount of hype across all sectors and it really deserves it. By big data, it’s across large numbers of consumers, patients, clients, big end.
[01:39] Unity: There’s tons of money flowing in from DC
[01:42] Deborah: You look at data across all of those and what can you learn? That you never had the data across all of them before. What can you learn by looking across that data? Small data is just complimentary to it. It says as an individual, or for an individual, for n of one. For me, for you. For your aunt, uncle. If you look at all of their different data streams over time how can you leverage that data over time to learn about their patterns and their responses?
[02:18] Deborah: So, it’s really in some sense it’s the same data. It’s that person’s row from all the different sources of big data. Their column whichever way you look at it.
[02:29] Unity: Are these just practical examples, steps people are taking? What are some examples, types of small data that exist in people’s everyday lives?
[02:42] Deborah: Absolutely. It’s the steps you take. Can be measured by Fitbit, can be measured by your phone in the background, because accelerometers are everywhere. But any one data stream doesn’t tell that rich of a story, doesn’t provide very rich of a feedback loop. It’s really in combination. So when I begin to combine the steps I take with the words I use with the busyness on my calendar. With my browsing history. With my Netflix binging, or having just moved I was binging on StreetEasy.
[03:23] Deborah: It’s those detailed patterns across our digital interactions and now so many of our interactions, period. And our transactions are digitally mediated.
[03:35] Unity: Why is this important in the context of health and healthcare? What does this mean for every day, either patients or just consumers?
[03:45] Deborah: I like to think of it as helping to fill the gap between what we’re born with and what we suffer from. Yes, we are born with certain genetic makeup. That does determine our tendencies, our proclivities, but a lot of, just listen to every talk, every paper about genomics, it all depends on exposure to X, Y, or Z. Behavior X, Y, or Z. It’s very important for health to capture that other half that’s determining whether we actually develop type two diabetes. Whether we move from pre-diabetic to diabetic. Understanding what your triggers are as a migraine sufferer. Understanding what your triggers are for flares as a multiple sclerosis suffer. Even understanding what form of physical therapy works for you as somebody who’s dealing with a herniated disc. From across our spectrum of chronic disease, in particular, the role of our behavior and also just bringing good feedback to doctors requires measurements outside of a clinical setting.
[05:04] Deborah: You can’t just measure things in the doctor’s office at a time of a clinical visit and doctors know that. They take patient histories all the time.
[05:12] Unity: This is measuring your life. Things going on in your everyday life
[05:15] Deborah: This is measuring your life. If you look at the nature article that Sage Bionetworks published off of the first year of data, off of the empower study, which is their Parkinson’s study, they already show the level of variability in the symptoms you see, there’s never been that kind of measure. Even within-person variability of the symptom. There’s been lots of data at the point of clinical visit. But how do the symptoms vary within a person? It matters because if you’re suffering from Parkinson’s you’re always balancing how much of the medication you’re taking, how much you can take when to optimize it for the functional part of your day and you’re balancing the effectivity of the drug versus any side effects.
[06:06] Deborah: It’s all about this sort of personal optimization and that kind of optimization can’t happen without good feedback. Now these data streams give us an opportunity to really have that high-resolution feedback from individuals.
[06:19] Unity: What are some of the big challenges in terms of really making sure consumers get access to the data or that this data actually becomes valuable or useful to everyday people?
[06:35] Deborah: A couple of things. First of all, to have it be useful a diff clinically, clinicians need to know how to interpret it. We are just now on the cusp of people systematically gathering this kind of evidence so that we can figure out how to pull out of it what is really a digital biomarker or behavioral biomarker. To have all this step data to have all the step data and your Netflix data and your language use data it’s way too much data in the raw. There’s tons of noise confounding factors. The research community is now in the process of taking these data streams and figuring out how do you pull out what is actionable for the doctor, and how do you then pull out what is actionable for the patient. But it has to have a clinical basis and we have to be able to turn these data streams which are so impressive into something that is actionable. Then from the perspective of the consumer they need access to their digital traces not because most people have time to be a quantified selfer, but because we will increasingly see apps and services that are personalized because they consume your personal data and then personalized feedback to you. In order to do that you have to be able to authorize, ok, just like when you download an app to your phone you say okay I’ll give it access to my location, history, and my contacts because I know it needs that to do its job. You will permission access to your health applications to have access to these different data streams both from your phone and from Google and from Netflix in order to have that app be able to be really smart about you.
[08:19] Unity: I know you’re you’re focusing on these issues every day at Cornell Tech. Maybe share with the audience what’s going on at Cornell Tech and some of the exciting, just the mission overall and some of the exciting developments there.
[08:32] Deborah: Sure. If we look at genomics it was really led by as it needed to be by biologists, by clinical researchers, and they’ve led and continue to lead very strongly there. On this side of the equation where it’s about understanding and interacting with people in their everyday lives it’s now an opportunity for the technologists and the designers, the behavioral psychologists, to really come together and lead this side of that revolution in collaboration with clinicians and all of that. That’s really what Cornell Tech is about. It’s about creating research jobs that bring together people around the digital disciplines training master students supporting advanced Ph.D. research and moving it forward with a very engaged faculty. [09:21] Unity: Within that context, I think New York City itself is such an interesting ecosystem because of the collaborative environment. There’s so many different industries here. So many different types of people. Maybe speak to that point. Why New York for Cornell Tech?
[09:42] Deborah: We’ve seen lots of advances in digital health. Pretty much focused more on the clinician’s problem and on the institution’s problem. The perspective on the patient’s problem is very much like a perspective on a consumer’s problem. Where else is the right place to seriously solve consumer-facing problems? Here we have an increasing digital competency and excellence coming together with a long history of excellence in things that are consumer and media and communications focused. It’s really just an incredible time and all under laid by the fact that consumers have adopted mobile technology at a rate and a prevalence that nobody ever expected.
[10:30] Unity: Since you’re seeing so many exciting things and you’re around so many, both students, but also engineers, entrepreneurs, scientists, people focusing on this innovation, what are you most excited about today?
[10:47] Deborah: To me, it’s that coming together of deep understanding of how to put the technology to work. The algorithms, approaches to machine learning, understanding of computer vision, natural language processing, in the presence of patient facing and consumer-facing problems, and this appetite for doing rapid iteration on problems. Twenty years out we all agree where things will be. It’s not so clear how we’re going to get there. What I’m excited about it being in New York City, it being at Cornell Tech, is the appetite to engage in rapid iterations on things we can do not just in twenty years but in twenty weeks in twenty months not only twenty minutes not only twenty years but that really valuable space in between when you have time to build experiment get out in front of people learn from what they do and keep going.
[11:44] Unity: I think one of the criticisms is often at these early stages especially the question of how will this impact everybody? Do you think it’s a broad question of what this is, but you think these digital health innovations will end up impacting whole populations, underserved communities, people that may not necessarily be able to afford the most expensive tools or technologies at the beginning? What’s your view on that and where things are going and how this will innovation in general in digital health will impact people?
[12:29] Deborah: If you look at mobile penetration and smartphone penetration people who cannot afford high-speed cable at home, that smartphone is their point of connection to the internet. They use it for their commerce. If you go down to a point where you’re talking about people who aren’t eligible for credit cards there you begin to have some digital divide which we need to pay attention to. But I’d say the greater digital divide is probably based on age. Across socioeconomics, the generation that did not grow up with digital are those who are less comfortable with it and if you go to today’s elderly today’s seventy eighty-nine-year-olds the match of mobile is not so great. But for those of us who in another fifteen years will be seventy, we have come through this digital basis to the way we run our lives and it’s pretty broad across socio-economics. In some ways, if you’re at the very top one percent you can afford a lot of personal assistance in the form of people and assistance in your life and the scalability an opportunity of personalization through mobile and through small data and through automation make some of that personalization affordable for the first time.
[13:57] Unity: Is this also a design challenge? It seems like if there’s an age divided, maybe enough folks aren’t thinking or designing within the context of who their audience really is.
[14:16] Deborah: It’s a very important question also from the perspective of when you design only for that endpoint and consumer you sometimes miss the fact that important things happen for people in some social context. It might be the adult, child, caregiver, that you really should be designing for, of the elder patient, as opposed to only thinking about that older patient. It might be somebody who’s an Alcoholics Anonymous sponsor rather than the sponsoree. It might be somebody who’s getting tutoring first-generation college student, emerging young adult, and is getting some kind of tutoring or coaching from a senior or another.
[15:00] Deborah: We sometimes overly design for an individual assuming they’re on their own and data and apps are just to serve them, whereas if we begin to design for small groups we also avoid that problem of missing a generation or missing people who are vulnerable in a way that might make the technology less accessible to them directly.
[15:22] Unity: Right. What would your advice be to entrepreneurs building today? People building today, innovating today. Maybe lessons learned or things that they should be focusing on?
[15:36] Deborah: A lot of people start out and are very concerned with scaling up. I think in this space of human-facing technology it’s really important to develop a design and focus on something that scales down. If it doesn’t scale down you can’t get out there quickly and get feedback on your idea from real people. Because you never launched at scale. How are you going to get to scale with the right product and service that is actually useful to people if it doesn’t scale down? If it doesn’t provide utility. At small numbers and at small penetration it will be much harder to get to a successful product at scale that’s actually useful.
[16:16] Unity: Right. Facebook didn’t even launch at scale. Started small. That’s very interesting. A couple of fun questions. Do you have a favorite book that you would recommend to innovators, students, entrepreneurs today that would help them?
[16:33] Deborah: ‘Thinking Fast and Slow’ by Kahneman. Yes, written by a Nobel prize winner, but it’s the sort of bible of behavioral economics and behavioral science. It’s hard to think about any application in the context of health where you don’t need to understand what’s going on in people’s brains and psyches.
[16:54] Unity: How about a favorite technology or tool or app that you love to use?
[17:02] Deborah: Pretty low-tech but again with health in mind I would say it’s podcast while running. You bundle these things. I get to listen to that next episode of a podcast while I’m running around Central Park.
[17:19] Unity: You beat me to my last question which is what do you do to stay healthy, so sounds like you run.
[17:24] Deborah: I do run and in general I try to plan my life so that it has these times to run planned in. As important as a meeting with my students or my Dean, is getting in there on my calendar my runs.
[17:41] Unity: I love it. Well, thank you so much for your time today. It’s been a wonderful conversation.
[17:45] Deborah: Thank you.