How to Use Data to Take Control of Your Health [transcript]

Written by Christopher Kelly

Nov. 13, 2018

[0:00:00]

Christopher:    Hello and welcome to the Nourish Balance Thrive podcast. My name is Christopher Kelly and today I'm delighted to be joined by Dave Korsunsky. What a fantastic name you have, Dave.

Dave:    Yeah. Thanks, Christopher. Thanks for having me.

Christopher:    Sun sky. Is it sunny in Truckee, in Tahoe today?

Dave:    We got a smoke, actually, that's blown in as of yesterday. The long weekend was just absolutely epic weather wise and then yesterday the smoke blew in. So, yeah, it's mixed up here. How about there, down in the Santa Cruz mountains?

Christopher:    We've been very lucky, actually, with air pollution. You can see it though. When you take off in an airplane from San Jose you can see the smog that's sitting over the Silicon Valley. I can't smell it and I can't see it and it does worry me a little bit. I was talking to Tommy about this just yesterday that maybe we need to start making some travel plans for the late summer as of 2019 because this is becoming a regular occurrence in the west coast, the forest fires and the very poor air quality.

Dave:    Yeah. We have it up here now. But, otherwise, it's been just wonderful up here this summer.

Christopher:    For people who don't know Dave, Dave is the founding CEO of Heads Up Health and Dave is making some fabulous software that creates a dashboard that you can use to track all of your biomarkers. Dave, we first met at Low Carb Breckenridge very briefly and then recently I heard you interviewed on Robb Wolf's podcast, of which we are a huge fan. I thought, my goodness, I do need to reach out to Dave and find out what he's been up to me because it's been a long time. Low Carb Breckenridge, when was that? It was 2017 so it's been over a year.

Dave:    I think it was maybe two years even since we first met there.

Christopher:    That's right. I think you're right.

Dave:    It's been a while, for sure. I'm glad you reached out because I've been having people come to me and say, "Hey, you got to reach out to Christopher and his team." We've both been getting feedback that we need to connect. I'm glad we did.

Christopher:    That's a very good sign. Before we get into any of that, tell me, I want to know the genesis story of Heads Up Health. How did someone who's used to the Silicon Valley world making software of other types get into making health-related software? What prompted you to get out of bed one morning and start working on this project?

Dave:    Well, we talked about this very briefly before we started recording. I was doing the Silicon Valley thing. For seven years, I was in San Francisco. I was commuting on the 101 every day to Palo Alto. It was miserable. And back. At that point in my life I hadn't really developed any awareness of mind-body, of health, of lifestyle quality.

    I'd always been fit and healthy but very little awareness of what was really going on with my health. Over time things just got to the point where the stress was getting almost unbearable, managing a huge engineering team, long commutes, not enough sleep, partying too hard on the weekends, not having any understanding of nutrition, largely just on the Standard American diet. Just ignorance of all of that stuff that is now front and center in my life. It just got to the breaking point.

    I started working with a functional medicine doctor who was in Texas and I was in San Francisco. I had all this lab test data that the functional doctor and I were looking at every time we would do our consultation just over Skype. A quarter of my records were at UCSF in their patient portal. Another quarter was at Stanford in their patient portal. Another quarter was at St. Mary's because that's where my GP was. And then the other quarter was in some stacks of paper that were in a closet in my room.

    But I needed to see the trends in all of that data because the functional doctor was working on some thyroid issues and he was working on some GI digestive issues, infections in the gut, just gut dysbiosis, intestinal permeability, whatever you want to call it but we did the test on my digestive system and there were a lot of things that needed to be worked on.

    The nerd in me just basically spun up a Rails application and scraped that data out and I was then able to then share it with the functional doctor. What was really amazing was I could see the trends across all those medical facilities. I could look back to 2005 when I was 24 years old and see my lab ranges. Hey, my white blood cell count was perfect back at that time. My platelet counts were perfect.

    And then start looking what was happening over the years in Silicon Valley and seeing these lab tests start to go down every six months consistently. My doctor couldn't see that because he didn't have all the trends. Only I did. He's got a small piece of that window. But sure enough, over the course of ten or 15 years, every one of these biomarkers was trending in the wrong direction and it was not until I put that all into my own system that that trend even revealed itself.

Christopher:    You must have a special sort of guy to even be doing the blood test in the first place. Most guys I've met, especially those just working in Silicon Valley that don't have any issues, they're not going to be regularly running blood tests, right? What prompted you to do those tests?

[0:05:09]

Dave:    I had some lab tests when I would just go for the routine physicals. There was some work there. I called my doctors in Boston when I lived there, "Hey, remember I got a physical there back in 2005? Send me the blood work." It was a pretty watered down panel, admittedly, but it had some numbers on there that actually were really helpful. Those were the oldest numbers I have.

    I can look at where I was in 2005 as a baseline for some of these immune markers and other things that I'm working on now. And then I really got enchanted with the whole idea of self-quantification, fortunately, while I was doing this. I've got really nerded out on WellnesFX, bought a bunch of their panels when they first launched. The functional doctor was running a lot of blood tests for me.

    Even though there was some gaps, I did have a lot of data. Admittedly it was very fragmented, but I had it all. So, when I put it all into my system there was actually some really interesting patterns that started to emerge. Presumably, a lot of the stuff that's interesting in your work as well which is -- what are some of the non-obvious patterns that you can see when you really know what to look for?

Christopher:    And this part, I understand. Engineers working in Silicon Valley, many of them, especially the system ops people, they are looking at dashboards all day long. And so when something goes wrong inside of their body, the first thing they're thinking, "Where the hell is the dashboard?" Okay, so I received an alert in the form of a symptom. Where's the dashboard?

Dave:    This [0:06:36] [Indiscernible] has reached 4,000 milliseconds and now I need to go fix this problem. Just like, okay, why can't I do that with my health? It just didn't exist. So, that started it. And then we're now collecting a lot of data at home. There's technology now in the home that is awesome, like being able to quantify HRV and some of the new tools that can quantity sleep quality.

    We're getting data in the home and these devices know you better than you know yourself whether you believe it not. The devices are so accurate that they're very insightful. I needed to see the biometric data in the home, next to the data coming out of all these doctors' offices. When you see the whole feedback loop, the food, the sleep, the stress, the blood sugar, the blood pressure, and then you see the clinical records, okay, lipids going down, inflammation going down, that's what I needed to see personally. That's what I built. It's like this is what I need and then I'm like I'm just going to build it and productize it from there.

Christopher:    When you said that's what I needed to see personally, do you know why that was? Do you mean that that's what you needed to see in order to make the change? I'm assuming here that the thing that got you out of that mess are diet and lifestyle modification. It's not like you went to the doctor and the doctor said, "I've got this perfect magic pill."

Dave:    Yeah. I could have done that. They have a place and there's times we need that but I was working with a functional doctor who first put me onto a clean Paleo template. That helped a lot. And then I was looking at the blood tests and then the functional doctor would give me a three-month protocol to work on and we'd retest the thyroid numbers. Not just the TSH. We'd rerun the whole thyroid panel.

    Because mine, it was actually more related to seeing really high reverse T3, some of these subclinical markers that weren't part of the routine physical. So, every three months we'd rerun this whole full thyroid panel. I had no medical training. I didn't know what these numbers mean? But as patients, we don't need to know everything about the biochemistry. We just need to know what numbers need to move up or down. And so I was smart enough to know that. We do the three-month protocol. We'd retest the numbers. I was actually really excited to see the numbers. It was fun.

Christopher:    And do you think that gave you the confidence to continue? You look at that dashboard and you think, "Oh, this is working. I should definitely keep doing this."

Dave:    It was helpful to get me engaged as a lay person into these numbers. Instead of deferring to the doctor to manage that data, the shift I want to see in society is that individuals take ownership of these numbers just like they take ownership of their bank account numbers. There's a different mentality around medical records. It's like, well, the doctors got that data. They'll call me if something's wrong or if I need to know something.

    I want to shift that paradigm where it's like, no, I can pull up my last 15 years of hemoglobin A1C instantly on my phone. That's the paradigm shift I want to see. It got me engaged as a lay person in the numbers. Admittedly, I had some technical knowledge from my years in the data center but it got me engaged as the patient. I want everybody to have this kind of ability.

Christopher:    That's interesting. You pick up on that. You're right though. Most people are not really interested in any of the numbers that could be measured from inside of their body yet they're perfectly comfortable with the dashboard in their car. They can tell you what all of the dials mean. I know that many of the analog dials are going away from car dashboards now but they understand at least what the engine check light is.

[0:10:09]

Dave:    They know how many miles are on the engine until the next service. They know the 52-week high on the stock portfolio. But it's like, okay, we need the 52-week high on your blood sugar. And that's going to keep you from developing -- I don't know what the statistics are, Chris. Maybe you know. But it's like one in three for cancer, one and four or something like for prediabetes or diabetes.

    These things take decades to develop, hypertension, cardiovascular disease. Why not start looking at the trends so you can see? That pattern may already be developing but your doctor only has 24 months of history on you. So, what about the other ten years where those patterns may already be revealing themselves? That's, I think, the importance of pooling all this information together.

    Now, in other countries, it's not this hard as you may know. In other countries, there's just one national system and they have all your blood work for eternity. But here in the US, every system is its own silo of information and I do believe that as patients now we have access to more sophisticated tools and we know more about our health honestly than our doctor does because there's so many ways to measure it now.

    I wake up, I check my HRV. I know what I ate the night before. I know my lowest resting heart rate when I slept last night from the ring. We have incredible power at our fingertips now. There should be no reason for all of this disease.

Christopher:    Yes. So, let's talk about some of the devices that integrate with Heads Up Health. You mentioned, you're talking about the Oura Ring then. That integrates.

Dave:    Oura Ring integrates. That one has been incredibly insightful for me personally. It's changed my sleep habits completely for the better. We can get into how and why later but we pulled that in. Elite HRV is another one that we just onboarded.

Christopher:    Right. And I recently started using the CorSense device which is--

Dave:    Cool.

Christopher:    It's cool, yeah. I mean, the reason I'm hesitating is that it shouldn't be that difficult to put on a strap and measure your HRV first thing in the morning. I think that Jason Moore would concede that it's just enough friction to stop people from doing it on an ongoing basis. They've made this really nicely designed finger trap device. Now all you need to do is stick your finger in there for a couple of minutes, maybe a little bit longer, and that's it. That's all you need to do. No putting on straps or anything like that. I really like that device.

Dave:    Yeah, I would agree with you, just enough friction to make it a little bit painful in the morning. I have a CorSense on order. But for people who have never measured HRV before, this technology is able to tell you when you're getting sick before you even had the symptoms. So, for example, I'll wake up in the morning and the Oura Ring and Elite HRV will both confirm that there's something wrong and they'll do that by saying the body temperature was elevated at night, which is a key symptom.

    So, I hurt my back in the gym a couple of days ago and woke up the next morning and I'd been in a lot of pain and, sure enough, the devices picked it up like that. You need to take it easy today. All of the measurements we've taken said your body is in a state of extreme recovery. Injuries, illness, overtraining, those are things that get picked up by these devices.

    We get a lot of requests to integrate the nutrition apps, the things that people are using to track food, predominantly from the keto community because they're carefully measuring protein, fat, carbs, total calories. MyFitnessPal, My Macros+, Cronometer, those are integrated. We're working on a project with Keto-Mojo which will be the first blood ketone meter that will have an API sync. That's actually working in beta where you just test your blood sugar ketones and that will sync.

    And then the wireless scales, I think, are a big part of it, the breath ketone devices, the LEVL, the Ketonix. Just seeing a lot of interest in low carb keto so that's where a lot of the integration work we're doing is focused now. Those are some of the big ones, food, sleep, diet, body composition, heart rate variability, the core stuff. But then there's also the medical facilities and, I think, we can currently tap into over 30,000 providers across the US. You can just sync up everything.

Christopher:    Wow. So, we're talking about blood test results here.

Dave:    Blood tests, yes. Stanford, LabCorp, Quest, UCSF, John Hopkins, Cleveland Clinic. Wherever you have blood work, sign in, link it up, we'll download all that data, which is a complete mess, and then we have to look at it all and try to make sense of it.

Christopher:    I'm laughing because I've tried to do this and I absolutely concur. You just really don't want to go -- If you're an engineer and you're listening to this, you really don't want to go there.

Dave:    We have 30,000 different names for cholesterol in our database because every doctor's office sends us different naming structures. It's a data normalization nightmare. We have to do that so I can show you one trend line of all of those tests across all the different providers.

[0:15:02]

    That's something we have to do on our side so that we can provide the trending that people need to go back and look, okay, I got diagnosed with condition X based on these defined ranges. What was happening for that lab range two, four, six, eight, ten, 15 years ago? We want people to be able to put that history together. That's a big part of the vision for this.

Christopher:    I absolutely recognize your story. I've been through some similar health issues and absolutely I had blood test done at a bunch of different facilities and I don't even have that data. It just disappeared into the ether and I'm not sure I have any way to get it. Now, I get all of my blood work done at Quest and I order that for myself. But when I try and log in to Quest to try and get that data from them it's nowhere to be seen. I don't think I've ever met anyone with a Quest login that had all of their blood data in Quest or LabCorp or any other of these providers before. Am I just missing a trick here? Is it common for people--

Dave:    Well, who are you buying a test from? You're not buying it from Quest. You're buying it from somewhere else.

Christopher:    No, I'm not buying it directly. I'm buying it from a reseller. It's maybe why my data is not showing up.

Dave:    Sometimes, yeah. You can email Quest and say I'm buying my test from this reseller, I want them in the Quest portal. Some resellers, it'll automatically show up in the Quest portal. It seems to be how they negotiate their agreements with each of these companies. I just dealt with this with one of our users. He goes to Quest for everything but the lab he uses, I'd have to go back and find it in my notes, for some reason, they just don't show up in the Quest and LabCorp portal which is super annoying because you know they did the blood draw.

    You actually paid the money to some third party reseller. That part gets a little bit tricky. I use requestatest.com and then my doctor, Grace Liu, you might know her, she uses Ultra Labs. If I buy the test from Ultra Labs they show up in the Quest portal every single time automatically. This is just part of the challenge. It's like, shit, what a mess. That's where it's like, okay, let me get the PDFs. And you can enter those manually into our system so we can fill in the gaps from what we can--

Christopher:    This is another interesting area. When I first started developing the blood chemistry calculator I was sure that nobody would want to do the data entry. I didn't want to do it. And I've met several doctors that had said to me, "Look, if this involves data entry, I'm out. I'm not doing that." And so from the outset, I had it so that you could just upload whatever you had, be that a JPEG or PDF and we would have someone that would do the data entry.

    As it goes, if you're doing this all day long, not that I would recommend that anybody do this all day long for a living, but like anything, you start to get really good at it and you can do the transcription very, very quickly. But it's really not ideal, is it? I understand you have that same concierge service.

Dave:    We do the same thing. Yeah, there's just a lot of people where we can't get access to their portal or they have a lot of functional tests. The functional medicine test we have to enter manually. We provide the same service because we just feel it's an important part of what we want to provide. We do the exact same thing.

Christopher:    We run a whole bunch of other tests as part of our elite performance program, the Precision Analytical, the Dutch and then also the Great Plains organic acids and we're doing a variety of stool tests. In fact, the list of tests that we use is becoming longer and longer although we don't do all of the test on all of the people. It's becoming more of a -- I mean, what do you do about that? Is it possible for you to integrate with all of these functional labs?

Dave:    I'm told some of these ones have access available but I think it's a level of access that would be available to like a hospital or medical, I don't know if it's the standard web based API. We would love to work with Genova and Great Plains and Doctor's Data and BioHealth Labs. That would be incredible to have those integrating as well. I don't even think they have patient portals.

Christopher:    No. I'm pretty sure they don't. No, they still send out PDFs. I've been there as well. I've written some software that will scrape the numbers out of a PDF and it works pretty well until inevitably they changed the format of the PDF and then your puzzle breaks. You're not back to square one but it's an ongoing battle.

Dave:    We're thinking about the same things, PDF scrapings, CSV uploads, pretty much anything we can do to ingest the data, that's what we want to do long term. That's where machine learning comes in and artificial intelligence to now start saying, okay, we were able to pull in their genetic data, the mutation is there, we then went and looked at the lab test from Quest and, sure enough, the mutation is causing a blood test to be out of range high and then we can go look at some other thing in the organic acid.

    Those types of algorithms that the brilliant practitioners can define for us and then those are just programmatized. It's like, man, that's where we're trying to get to with all of this stuff. It's very non-obvious stuff. It's stuff that's outside the purview of 90% of the practitioners out there because there's just so many complexities to it.

[0:20:06]

    But there are practitioners that know what to look for. And so if we can start distilling out those nuggets and programmatizing those it'd be really, really interesting.

Christopher:    I totally agree. That's not machine learning. Machine learning is learning through example from data. What you're talking about is a hand rolled algorithm of which there are very many, very good algorithms floating around in the heads of all these functional medicine doctors. They're going to give you those algorithms.

Dave:    Exactly. I think they will. Yeah, I believe they will. The ones I talked to, they will. I mean, they've got more patients than they can handle. It's not like they're trying to hold on to the IP. Hopefully, not. But the ones I'm talking to whether they just want it for their patients or whether they want to open source these things, that's a different question.

    But they are the ones that are looking at this stuff all day long. Every single day they're looking at tens of thousands of labs, genetic data. They know. Those are the nuggets we want. And then like you said, and then you can have them more automated learning that could happen as well.

Christopher:    Right. The one place that I think machine learning may have promise is in recognizing the characters that are on these damn PDF forms. And I have been there and tried that. But then basically what happens is you move the problem around. The problem is not now recognizing whether that's a digit or a letter. It's trying to get the data into the right place. If you look at -- For example, doing automatic data entry for till receipt is done really, really well now. If you're buying, you have lots of travel--

Dave:    Yeah. I know what you mean, yeah.

Christopher:    Yeah, the services that do that really well. But you can look at a till receipt--

Dave:    Expensify or something like that.

Christopher:    You can slice that. Imagine you have a guillotine. You can slice it row by row by row and then all the numbers are in the same place every single time and so the data entry task is quite easy. Whereas when you look at an organic acids report, the numbers are all over the place. Even the task of dividing it line by line is quite challenging. We do need Great Plains to get their act together and create an API.

Dave:    Yeah, that would be incredible. I think, for now, guys like you and I are going to hack it together using the best tools we have available. Sometimes we just key it in and just do it the old school way for people.

Christopher:    That's the problem, isn't it? I mean, the automation is only worth it when it's faster than doing it by hand. If I've already got three of this to do, then it doesn't make sense to spend two weeks on a program to enter that data.

Dave:    And there's just not a ton of people with those tests. There's tons of people getting the conventional labs. As much as the nerd in me wants to go solve that problem, I just don't know how far we'll get down the path in terms of automating that stuff.

Christopher:    Talk to me about your reference ranges especially for blood chemistry. That's part of the value proposition of the blood chemistry calculator is that Bryan and Tommy and Megan have spent countless hours in the medical literature looking for health outcomes that are associated with specific ranges for a blood test. People have talked about this ad nauseam now.

    The problem with the conventional lab range is that it's just an average, two standard deviations either side of the mean is your average and, of course, the population on which it is based is getting sicker and sicker so do you really want to be the average sick person? Probably not. And, of course, it varies from marker to marker.

    But generally, it's very helpful to have someone like Bryan or Tommy or Megan do some research and tell you the health outcomes that are associated with the blood level of albumin, say. Is this something you've thought about and where do you get your reference ranges from?

Dave:    We just use the defaults from Quest or LabCorp now so we just use the average ranges. We allow the individual to log in and change those themselves. You can come in and set your own if you know what you're doing. And then we also allow the practitioners to sign up and basically just define their own and then apply those to their clients in the system.

    Each practitioner tends to slice it and dice it a little differently as well. We allow them to come in and say, "Well, I want my reference range for hsCRP, whatever, to be this for my clients." We put some basics in there then we just leave it to the individual to tweak those or to their practitioner. I'd love to have a set of good functional ranges in there. We don't have that yet. We're just using the default conventional for now.

    I'd like people to be able to toggle like here's the conventional Quest, here's the functional medicine ranges for some of these markers. You could actually toggle between those. That would be awesome. It's I think something we get asked about a lot. I was just on the phone today with our CTL thinking about how do we make it so that the system is flexible enough where the doctors can specify their own, individuals an override? And then, of course, we get the ranges from the hospitals that we integrate with as well.

[0:25:02]

Christopher:    It's a mess.

Dave:    Yeah. It gets complicated, really complicated.

Christopher:    It does. In my experience using ranges that came from a functional medicine doctor or software that they had created, typically, they also didn't have a range. I think the blood chemistry calculator is the only place where you can get evidence-based optimal reference ranges. Most of the time, when I look at other people's ranges, they're usually just one standard deviation either side of the mean.

    You said two standard deviations. Before we said glucose was 79 to 120 or something. Let's just call it 90 to 100 or something. We'll just center that more closely around the mean. I mean, it's better but it's not necessarily optimal and it's certainly not health outcome based.

Dave:    I'm on a specific type of diet. I've tested my cholesterol long enough. I know exactly what my personal range is going to be. I'll just set it based on what I want it to be for me and I'm not a medical expert but I just know that, for me, my total cholesterol, I can't remember how I have my ranges set for some of these things, but I've just done enough so I've just gone and set it myself.

Christopher:    Okay. I know you've been hanging out with Dave Feldman. I think I trust--

Dave:    I love that guy. Yeah, we're working on getting all of his data into the system as well. We could talk for hours about that stuff.

Christopher:    Talk to me about tracking. I know that Heads Up Health is really great for tracking a specific marker over time. Have you thought about how a group of markers might come together to form a pattern that might be telling you something? Do you see what I'm saying? You could look at your fasting blood glucose over time and say, "Oh, well, this is getting worse." But then you might look at something else like your level of fasting triglycerides or the triglycerides to HDL ratio or some other--

Dave:    Or waist circumference.

Christopher:    Yeah, some of the marker metabolism and say, actually, when you look at the big picture I know your fasting glucose looks like it's slightly not optimal but actually the picture here overall is improving. Is that something you thought about?

Dave:    We think about a lot of things like that.

Christopher:    Sorry. I get this as well. You got so many ideas, like which are the ones you implement first?

Dave:    Well, there's two parts to this mission. The first part, actually, is just to solve the data fragmentation problem for people. And we spent several years on that. That's the first thing we did, was like just help me get it on one place so I can see the things I want to see. Phase two of that is what you're talking about, the groupings, the patterns that we can distil down from published research, the intelligence.

    Like you said, hey, fasting glucose is actually 101 but your waist circumference has come down six inches and your weight is down 30 pounds. Those types of intelligence, that's the part we're trying to get to. It's just super hard, man. We're a small startup so that comes down to available capital and resources and engineers. That's the vision for what we want to do.

    We have all the information already and now we want to turn the corner and start building. I would call that an insight that we could deliver to you. You notice this number is high but these other four numbers also associated with it actually got much better. That's an incredibly helpful insight that most people might not make the connection with on their own. They might just look at that number in isolation and freak out and say, "I'm off this diet. It's raising my blood sugar."

    Those insights, I think, are so awesome but it's just been really hard for us to get there just from a business company perspective. You know how it is, boot strapping and building and all this stuff. It's just really been hard. I'm being candid with you on that.

Christopher:    What about tracking symptoms as well? Because you're in a very strong position there to be able to track people's symptoms. I know that you've talked before on your podcast about seizures, I believe.

Dave:    Epilepsy, yeah. That's a big one for us. We've interviewed a lot of people with epilepsy. That's a whole puzzle in and of itself, what is the trigger for the seizure, for example, what is the trigger for the migraine, what is the trigger for any number of brain fog. There's a number of different variables. For some people, the seizures are triggered hormonally based on menstrual cycle. For other people, it's triggered when their blood sugar jumps more than 20 points in a certain window.

    Whatever they eat, it jacked the blood sugar just too high and that triggers it. In other people, it's weather patterns. You know what I mean? That is amazing if we can start getting down that path and being able to look at what happened. Maybe it was something that happens three days before the actual event.

Christopher:    This is another very strong promise of machine learning. This is something I know I could do today, is the idea -- So, if you have the data. Let's say, you're interested in predicting, let's call it afternoon fatigue. Am I going to be tired this afternoon during my important meeting?

[0:30:02]

    If you had enough labled data, people are pouring their data into Heads Up Health, and then you had a symptom tracker, you had somebody record the grand truth which is whether or not I had the fatigue or not, then I could train a model to predict that fatigue. Is it a one or is it a zero or maybe we could have a sliding scale from zero to ten.

    The really interesting thing about some of the models that we'd been using are you can look inside them so they are interpretable. You've got all these things you're pouring in, your blood ketones, how much sleep you got last night, what you weigh this morning, how many calories you ate yesterday, all these features that you don't know which of them are important right now.

    But if I pour them all into a random forest or a boosted decision tree, once I finish training the model, I can then look inside of it and ask it, which are the most important features for predicting afternoon fatigue? And then you would know. Okay, so, I really need to keep an eye on my blood ketones, say. Say that was the most important feature. Then you would know which of the things to keep an eye on.

    But then another cool thing about the random forest and boosted decision tree is you can do a partial dependence plots. You can clamp all the other variables and then just vary one variable at a time and then see how that affects the dependent label. You could say, "Oh, well, if I had got 90 minutes more sleep last night then I wouldn't have been tired in my meeting this afternoon."

    I think that could be a really powerful thing for creating behavior change. You have Heads Up Health ping you saying, "Oh, you need to take a nap this morning." I'm making things up here.

Dave:    That's part of the reason I want to get the food list. A lot of these nutrition apps, they share the macro and the micronutrient data but we don't get the food list. That, I think, a lot of people with food sensitivities, non-obvious food sensitivities would be a huge problem to solve because sometimes there is a 24-48-72 hour delay after you eat a food.

    If these APIs could also -- some of them do share the food list with us. Some don't. But then start looking at, well, actually, it's a 48-hour delay after gluten exposure and that causes the symptoms. You could start figuring out these symptoms people are having related to food sensitivity that they never even expected.

Christopher:    Right. That would be so powerful. At the moment what we really have is the elimination diet. They'd have to go down to something really basic. That is still the gold standard. That's still what we use in our practice. People ask us all the time about food sensitivity testing and we've never gone there and that's partly because my wife is a food scientist that spends some time in the lab looking at food allergies and she's like, "No. there's no way that you can find a food sensitivity like that. Its' ridiculous."

    But the test is still commercially available. Usually what they show is what you've been eating over the last few days before you did the test. That's not helpful information. Wouldn't that be awesome if we could hook up Cronometer to a machine learning algorithm and then have it tell you which of the foods that are triggering your migraine. That would be freaking awesome.

Dave:    Aaron Davidson, if you're listening, we just came up with a brilliant idea for you.

Christopher:    Yeah. I mean, so the other thing that we've been thinking about -- the blood chemistry calculator, one of the things it allows you to do is see patterns that no physicians can see. For example, Bryan recently presented a case study where he showed the prediction for H. pylori. IgG antibodies changed when one of his patients went through a treatment protocol.

    Normally, IgG antibodies, they're not terribly useful because it just means that you've been exposed to the antigen at some point in the past. But it does change. If you care to track the IgG antibodies over time then you will see a reduction in those blood levels of antibodies as you do something like some sort of anti-microbial treatments. It is useful.

    It turns out that I can predict that level of IgG antibodies if you give me your basic blood chemistry. At the moment, I don't really have that much data. It's very hard for me to acquire that data. That's another idea. We can predict infections from people's basic blood chemistry and you could do that right there inside of the dashboard in Heads Up Health. That would be cool.

Dave:    Yes. We've got lots of data so we can definitely partner. I know you and I have kicked around some ideas. Because we've got these individuals connecting the EHRs and we'd love to be able to run that through your system, for example, and feedback the analysis to people. There's different ways, I think, over time as you and I continue to kick ideas around that we can partner up and start to derive more value from the data that we're collecting and then approach other founders out there.

    The founder of Cronometer is a cool dude and if we come to him with some cool ideas, he's open to it. And we can do some game changing stuff if we had access to the food list and some of these other ways to analyze it. That type of stuff has the potential to completely change the paradigm on how we think about managing our health, treating ourselves and using technology and ways that are far more powerful than anything we imagine or that an individual practitioner could wrap their head around.

[0:35:14]

Christopher:    Yeah. I would love to see it. We're very suspicious of genomics at this point and very much question the value of doing any of their commercially available genome sequencing like 23andMe. We've just not seen it play out in clinical practice. It's an interesting idea that was worth pursuing but not very much has come of it in my opinion. But rather than relying on my opinion, wouldn't it be cool if you've got all that data in Heads Up Health like your MTHFR status. Is that predictive of your homocysteine level? Could I train them on a model--

Dave:    I know for me it is. There's a direct correlation there for me. That's like one o f the main numbers I have to keep an eye on.

Christopher:    But wouldn't it be -- What I'm saying is like if we had 10,000 labeled examples and I was to train a model -- because that happens quite a lot with a chemistry calculator. I really like to predict Epstein-Barr virus. Let's see if I can do that and I can't. I just can't predict Epstein-Barr virus. There's just not enough examples of people that haven’t been exposed to it for me to be able to predict it. We just don't publish those models because I can't do it. I'm wondering if the same thing might happen if I try to predict homocysteine levels based on your MTHFR status. Whether or not I'll be able to do that, that would be a very interesting question to answer.

Dave:    Yeah. I know, n equals one, for me, the answer would be a definitive yes. I mean, I have that mutation and my homocysteine is like way out of range high if I'm not taking the right supplement. That's what I personally keep a close eye on. I had a woman on the podcast, Carrie Brown, recently who had severe bipolar and suicidal depression.

Christopher:    I listened to that. That was a great episode, actually.

Dave:    That started with addressing MTHFR and methylated B12 deficiency. It was like the first thing that just gave her enough of a reprieve to get some breathing room to say, okay, wow, I feel a little bit better. I can take some of these next steps. Man, really just fascinating types of stories that come from data that's commercially available to consumers these days.

Christopher:    Talk about the practitioner and friend tracking. I think this is a really interesting and useful feature. If somebody grants you permission you can see somebody else's data, is that right?

Dave:    Yeah, that's for family and loved ones. My functional doctor has access to my profile at all times, functional doctors. They have access to it at all times. I think that's just a necessary part because you have to look at health as almost like -- there is a time to go to the conventional doc for acute care and there is a time to work with functional and there is a time to work with nutritionists.

    I just want to be able to share it with these people. They can see everything, do what you need to do and then disconnect them and then just take my data and go about my business. So, there's that feature. We just call it care team access where you can just enable or disable people that you want to have access to your information.

Christopher:    What I'm thinking is I can sneak into my dad's house and blow dart him with a--

Dave:    I like where this plan is going. Yes, continue.

Christopher:    With a little hair-like thing that's part of a continues--

Dave:    This is under the cloak of darkness, 3:00 in the morning.

Christopher:    And then I'll have on my dashboard my dad's continuous glucose monitor. And then I can ping him if he eats breakfast cereal or anything else.

Dave:    He will just love you to death. Micromanage the crap out of the old man. I've got access to my dad's profile too. He has a few dances with the devil. But I need that data. I need to be able to pull that up in an emergency. It's all in there, everything in Heads Up Health, every single blood test, all his blood pressure readings. Everything is in there. I'm in California. He's in Scottsdale. I have access to that information if I have to and events of an emergency. Everything.

Christopher:    Have you figured out how to do that? So, for example, I wore the Dexcom 5 for a little while and I remember going through this dance to try and get the data out of it. I found this -- I think it was the Quantified Self has some app that allows you to extract the data from the Dexcom and then you could put it into a spreadsheet or something. Have you done the work with any of the continuous glucose monitor?

Dave:    Dexcom opened their API. So, that's on our road map to integrate that one. Yeah, they did a pretty good job with it. It's should be pretty straightforward. The FreeStyle Libre, I don't know that they have an easy access point yet. I think the days when we could just put an Apple Watch on and it's got some way to do CGM, getting five-minute glucose measurements, I'm getting the five-minute HRV measurements from the Oura Ring, all just continuous and being sent to me in the background, I don't have to do anything, it's being passively collected, and then the technology is saying, "Hey, Chris, whatever you just ate, your HRV and blood sugar went to the roof."

[0:40:09]

    That, I think, CGM in general, making it affordable so you don't need a prescription. And then lay people who are not super into their health at least can get some notifications on which foods are sending them through the roof because it happens and you don't feel it necessarily. You just get fat and then you get a disease ten years later.

    But if you can start notifying people when these spikes in glucose are happening for kids sometimes, 20 times a day, and just building that awareness. The awareness you get when you first put on a sleep track -- crap, I am actually only getting five hours a night. I thought I was better. The awareness you could get from the watch telling you you're at 190 milligrams per deciliter, I think that will have huge implications for the health of our society because it's just not in our consciousness to -- It is for me and you and everybody listening. But for most people, the associations are just not there. So, CGM, in general, I think has incredible promise.

Christopher:    Yeah, I think you're absolutely right. As Dr. Simon Marshall, our performance psychologist, has said before on the podcast, self monitoring is the cornerstone of behavior change. You can't really hope to change something when you're not monitoring it. It's very, very important. And for me, even though I know that was one of the things that came out of me wearing the continuous glucose monitor, was it changed the things I was willing to eat. I just wouldn't want to eat that store bought cupcake even if it was gluten free.

Dave:    Sometimes ignorance is bliss. Let's be honest about it, right?

Christopher:    It's true. I've taken this thing off so I can eat the cupcake.

Dave:    Exactly. One of the things to me was just having a little bit of data made some really big changes in my lifestyle. I didn't need fancy algorithms or anything to tell me I was not getting enough sleep. I just needed actually to measure it for a few weeks and it was insightful. It was like, "Oh, man, six hours. That's it?" And I've totally changed my sleep hygiene, my lifestyle, the way I think about. And that was just from some data that I didn't have before and then one day I had it and it led to some very healthy changes.

Christopher:    What do you think the greatest challenges are for you in the near future? What are the things you are most worried about at the moment?

Dave:    From what perspective?

Christopher:    From a technical perspective. I'll give you one example. You talked about how most of the data are locked up in silos that don't talk to each other. That seems to me to be a very significant challenge. But maybe you can think of something more challenging.

Dave:    I think that there's obviously risks of health organizations that decide they don't want to share the data anymore. They want to cut off the API access because that's somehow part of their internal systems and intellectual property. I think that's one risk. Although there's a lot of movements -- you may be familiar with this FHIR movement that's starting to solve this problem. I don't know that that's really going to be a blocker.

Christopher:    What do you think about the trust issues? So, we talked about this before.

Dave:    Yeah, that's probably something as well.

Christopher:    Pedro Domingos, who is a professor of computer science, we did a podcast on machine learning and Pedro talked about how he does trust Facebook with some of his demographic and personal data in return for entertainment but he's not sure that he would trust them with his medical data. I think that's a valid point that you could make of any of the big Silicon Valley companies like Google or Apple or Facebook. Do you think that people are going to trust those big organizations with their data?

Dave:    I don't know. I know Apple is looking at putting health records on the phone and they've been very clear that they have no access to that data whatsoever. That's on your device. They have no ability to analyze it or do anything with it. They've been very open. I think that's how they're structuring it. It's been a while since I looked at their terms on there. They're very open. They're like, "We just want to show it to you. We don't have access."

    Other companies just don't have the trust factor. They've lost that consumer trust through mistakes, situations that have happened. Facebook, for example. You can look at examples in other industries which you look at consumer finance ten years ago when mint.com came along. They said nobody is going to put their financial data into a startup.

Christopher:    You've had to be completely bonkers.

Dave:    You're crazy, yes. Startup? You're going to put your money data in there? Well, two million people did that in under two years because they're getting incredible value out of the service. Okay, I get a few extra loan offers in the mail box. Do you know what I mean? I throw those in the garbage. I don't know. It's different for everybody because I don't have a lot of really sensitive health data. I'm just a healthy guy.

Christopher:    Yeah. I totally agree with that.

Dave:    I don't give a shit. But if I had a really sensitive disease that could preclude me from insurance or something I don't want disclosed publicly, totally different set of considerations.

Christopher:    Has anyone talked to you about the idea of putting the data everywhere and nowhere, an encrypt it somewhere on a blockchain, for example.

[0:45:02]

Dave:    That's, I think, the best way, is like some blockchain based solution that's completely anonymized and also allows you to choose who you want to share it with. The right researchers can get access to it. I think there's a lot of promise there. I know there's companies working on it. I don't know how far along they are. That one is definitely intriguing.

Christopher:    Is it something you thought about implementing?

Dave:    Well, we already have all of the information so if we could somehow distil it down into a format that could be shared on a blockchain based system. You could use Heads Up Health, get all your data at all the health systems and the devices and the apps, package it up and then put it on blockchain and monetize it or not monetize it, or share it with a doctor or don't share it with a doctor.

    We could be the front end so that you can at least get it all and then put it into some type of blockchain based wallet and then share it in whichever way you choose. I just think that that's a priority that's probably number three for us. First is give people the data. Two is build the intelligence. Three, at some point, is share it on some of these new open technologies like Blockchain, secure, encrypted, although that's -- yeah.

Christopher:    This might be a good moment for you to explain to people where the data is stored. I know you talked about this on Robb's podcast but since we're here I think it's worth reiterating how the storage is done in Heads Up Health.

Dave:    We have to abide by the same rules as everybody. We've architected everything inside of Amazon AWS. The nice thing about AWS is pretty much most of their services are HIPAA compliant right out of the box. That makes the infrastructure and architecture very easy and secure to build. We use them to handle all of our encryption and data storage and those types of things.

    We try to further reduce exposure by actually not storing the clinical data. If we connect your account from Quest, we will display it for the duration of your session and then when you log out, it's destroyed. It stays in their system. For a lot of the clinical data -- Obviously, if you enter it manually into Heads Up Health, that's in our database. But for the electronic health record connections, we don't actually store that information. It's just another way for us to further protect and de-risk. If you connected with Quest, there's really nothing in our system except for the duration of your session, actually.

Christopher:    Yeah, this makes sense from a security standpoint. I'm not sure it makes sense from an efficiency standpoint especially if -- So, for the blood chemistry calculator, you enter your blood data into the model and it does a significant amount of work.

Dave:    It's not efficient from a computational perspective, for sure. And we may revisit that in the future but that's how we built it today.

Christopher:    Well, this has been awesome. Tell us about the podcast first. Tell us why you started the podcast.

Dave:    I started Dave-Driven Health Radio because I wanted to help the average person understand how to interpret the numbers in a more intelligent way. We just get on the show and we talk about what these different numbers mean and how to use them and different professionals will come in, medical professionals. I also like to speak to the entrepreneurs who built the tools and the technology. So, the guys building the devices and the apps that we're using every day from Oura and Cronometer and Nutrition Genome and all these cool companies.

    I want to bring the founders on and then bring the medical professionals on. And then I also try to just bring on people who are actually kicking ass and they've conquered their condition. So, somebody who has beaten epilepsy with a keto diet, someone who has beaten sugar addiction, someone who has put cancer in remission. So, the war stories from the people. Those are the three types of things all really centered around data, much I'm sure like your guests are centered around data.

    And so I really, really love doing that part of it. That's what we do with the podcast. It's called Data-Driven Health Radio. It's just something I wasn't sure I would enjoy but I just tried it and I get to speak with guys like you all the time and have these great conversations and it's really rewarding.

Christopher:    Yeah. What's not to love about podcasting? I'd recommend it to anyone. It's absolutely fantastic. I'm already a listener of your podcast so I know that you can find the podcast if you just search in Overcast.  Just remind me of the name again. Once I've added something, I realize I don't really look at it.

Dave:    Data-Driven Health Radio.

Christopher:    Data-Driven Health. I just look at the logo. I scroll down the list and I'm just looking at that image to decide which one I want to look at. I will link to that in the show notes. If you search inside your podcast, you will find it. What's the best way for people to get started? How much data do they need in order to get started on Heads Up Health? Do you think it's going to be meaningful if you've only got one or two blood tests or just a little bit of historical HRV data? Is it still going to be meaningful to create a Heads Up Health account? How does that work?

Dave:    The typical pattern is people will sign up. They'll typically connect the nutrition tracker or a Fitbit or something like that and bring in some of that information.

[0:50:04]

    And maybe they'll enter a few blood sugar, blood ketone readings. Maybe they'll get their most recent doctor connected. It is something that builds up over time. Right out of the box, you're working with an empty profile until you link up some devices or apps, link up some medical facilities, enter some data, and then it starts to get more utility over time. So, me being able to look at three years of weight data from my scale and three years of blood ketone data and see, okay, there's a clear pattern here.

    That takes time to build up the history in there. It just depends how fast you ramp up and how much tracking you're actually doing. I don't think it can hurt anybody to make sure that they have access to their most recent blood test. They probably do through their patient portal. But for me, I had a ton of PDFs. I just started with transcribing some of that data.

    Everybody's trajectory is different coming in whether they're managing a disease or just coming in to kick the tires. It's a little bit different for everybody. Chris, as I'm sure you know, onboarding for a startup is also something we have to spend a lot of time on. We can show people the right things at the right time and ask why they're here and that's just stuff that's going to come with our maturity as a company. We're continuously trying to work on that.

Christopher:    You reminded me of that old saying, and I think that Heads Up Health would fit quite nicely into this, the best time to start Heads Up Health was maybe when you're a child and the second best time to start Heads Up Health is today.

Dave:    Yes. Just get it out there. I'm sure with you guys, too, you're getting it, you're iterating, you're learning, your product just keeps getting better and better and better.

Christopher:    What you don't want to do is you run into a problem. Okay, say, you're feeling good right now and you run into a problem in two year's time then you're scrambling trying to reach the doctor that you went to in 2005 trying to get old medical records. I think I want that all to be in order before I run into some sort of, maybe not emergency, but problem that I would like to solve.

Dave:    Yeah, it happened to me. I was on a business trip in Boston and I just had this awful, awful stomach thing. I had to go to Beth Israel at 1:00 in the morning and I'm like, "Yeah, here's everything." I had it all right there. I was prepared in that situation. It can be hard to keep track of all this stuff especially if you're sick, hunting down the medical records and making sure they get entered and transcribed. It's a lot of work. That's where our concierge service comes in. They can do a lot of that work for people.

Christopher:    Awesome. So, headsuphealth.com is the main website. Is there anywhere else that you would like to send people?

Dave:    No. They can find everything there. I'm pretty open and you can reach me at dave@headsuphealth.com. You can try out the app. There's no cost associated to come in and link up resources and kick the tires on it. And then there's our podcast. Those are the main ways that we're putting our name out there.

Christopher:    Awesome. Well, this has been fantastic, Dave. Thank you so much.

Dave:    Yeah. I didn't know you are in California. I'm looking forward to maybe grabbing a coffee some time when I'm down your way.

Christopher:    Yeah. I mean, it's been a while since I've been -- That was one of the main reasons I moved to the US, was so that I could go snowboarding at the weekends.

Dave:    If you're coming to Taho, I'm happy to entertain any time, man.

Christopher:    Yeah, definitely, at some point. Now, I've got a four-year old. It's a little bit tricky. So far, I resisted the temptation to put her on skis because it just seems a bit unnecessary to me. I learned to snowboard when I was 20 years old and I did just fine. Maybe we'll wait to see what Ivy says about that before we decide. But I will be up to Truckee and doing some snowboarding sometime soon. Thank you. I look forward to meeting you in person.

Dave:    Yeah, sounds awesome. This is really fun conversation.

Christopher:    Thank you.

[0:54:00]    End of Audio

blog comments powered by Disqus

Register for instant access to your FREE 15-page book, What We Eat


© 2018 nourishbalancethrive