The Patient Equation

by Glen de Vries

  Let's Start with the Thermostat

  In the next section of the book, we'll start to look at some examples of applications that are doing things right, using these new technologies to better understand disease, to generate actionable information, and to make a difference in people's health and wellness. But we can start here with a really simple example from outside the world of medicine.

  The Nest Learning Thermostat—now owned by Google—was the first “smart” thermostat to enter the market at scale. The user starts by putting data in—what temperature they want their house to be, what time they leave the house, what time they return—and, after a few weeks, the thermostat starts to be able to make predictions, to learn when users are home and when they aren't, and to switch the system into energy‐saving mode when the house is empty.24 It builds its schedule around you and learns from you. With limited inputs, and with continuous learning, your energy usage and comfort are optimized. This is the climate control version of a patient equation. Inputs, algorithms, outputs, and benefits—for your home, as opposed to for your health. (Or, at least, not yet for your health. Perhaps our smart mattresses will provide inputs one day that, combined with our smart air conditioners, will work to optimize sleep patterns by controlling the temperature of the bedroom. But one step at a time.)

  The world of home heating and cooling seems to have been permanently altered by Nest. A quick search on Google or Amazon shows a huge range of companies that have since entered the marketplace, including the old‐line manufacturers of traditional (a nicer way to say “dumb”) thermostats. Most of these advertise their ability to integrate with Apple HomeKit, Amazon Alexa, or Google Home. If the thermostats are a stand‐in for the sensors and systems that power patient equations for specific diseases or conditions, the analogy can be extended into the theoretical platforms we will discuss later that gather and apply all of these inputs and algorithms to more holistically manage our health.

  With thermostats, it's about the right temperature in the right room at the right time. With the patient equation, it's about giving the right patient the right treatment at the right time. We aspire to use traditional medical measures, digital instrumentation, and math, all combining to figure out who to treat (and how to treat them), and who not to treat. We have trouble predicting the weather when it's more than a week away. But we are starting to have the tools to predict disease, to predict drug success, to predict our medical futures. We are starting to have the tools to expand patient value, extend the years of quality life. We can harness data to deliver more precise therapies that turn death sentences into chronic diseases, and chronic diseases into curable issues. In the next section, we will look at areas where this is already being done well, and which can serve as models for the rest of human health.

  Notes

  1. Mercatus Center, “Atul Gawande on Priorities, Big and Small (Ep. 26),” Medium (Conversations with Tyler, July 19, 2017), https://medium.com/conversations-with-tyler/atul-gawande-checklist-books-tyler-cowen-d8268b8dfe53.

  2. “Wearables Momentum Continues,” CCS Insight, February 2016, https://www.ccsinsight.com/press/company-news/2516-wearables-momentum-continues/.

  3. Caroline Haskins, “If Your Apple Watch Knows You'll Get Diabetes, Who Can It Tell?,” The Outline, February 21, 2018, https://theoutline.com/post/3467/everyone-can-hear-your-heart-beat?zd=1&zi=ixcc7c67.

  4. “Cardiogram—What's Your Heart Telling You?,” Cardiogram.com, 2018, https://cardiogram.com/research/.

  5. Madhumita Murgia, “How Smartphones Are Transforming Healthcare,” Financial Times, January 12, 2017, https://www.ft.com/content/1efb95ba-d852-11e6-944b-e7eb37a6aa8e.

  6. Hamish Fraser, Enrico Coiera, and David Wong, “Safety of Patient‐Facing Digital Symptom Checkers,” The Lancet 392, no. 10161 (November 2018): 2263–2264, https://doi.org/10.1016/s0140-6736(18)32819‐8.

  7. Eric Wicklund, “MHealth Startup Uses a Smartphone App to Detect Sickness in Speech,” mHealthIntelligence, November 13, 2017, https://mhealthintelligence.com/news/mhealth-startup-uses-a-smartphone-app-to-detect-sickness-in-speech.

  8. Dave Muoio, “Machine Learning App Migraine Alert Warns Patients of Oncoming Episodes,” MobiHealthNews, October 30, 2017, https://www.mobihealthnews.com/content/machine-learning-app-migraine-alert-warns-patients-oncoming-episodes.

  9. Second Opinion Health, Inc., “Migraine Alert,” App Store, August 2017, https://apps.apple.com/us/app/migraine-alert/id1115974731.

  10. Victoria Song, “The Most Intriguing Wearable Devices at CES 2017,” PCMag, January 9, 2017, http://www.pcmag.com/slideshow/story/350885/the-most-intriguing-wearable-devices-at-ces-2017.

  11. Karthik Iyer, “Sence Wearable Band Accurately Tracks Emotional States & Productivity,” PhoneRadar, November 26, 2016, http://phoneradar.com/sence-wearable-band-accurately-tracks-emotional-states-productivity/.

  12. Andrew Williams, “Panic Button: How Wearable Tech and VR Are Tackling the Problem of Panic Attacks,” Wareable, December 3, 2015, https://www.wareable.com/health-and-wellbeing/wearable-tech-vr-panic-attack-sufferers.

  13. Ananya Bhattacharya, “Apple (AAPL) Filed a Patent Application for a New Kind of Heart‐Monitoring Wearable,” Quartz, August 11, 2016, https://qz.com/756156/apple-signaling-a-new-direction-filed-a-patent-application-for-a-new-kind-of-heart-monitoring-wearable/.

  14. Brad Faber, “Skin Patch: New Device Collects Sweat To Monitor Health [Video],” The Inquisitr, November 26, 2016, http://www.inquisitr.com/3746360/skin-patch-new-device-collects-sweat-to-monitor-health-video/.

  15. Jonah Comstock, “Verily's Goal: Make Our Bodies Produce as Much Data as Our Cars,” MobiHealthNews, October 3, 2017, https://www.mobihealthnews.com/content/verilys-goal-make-our-bodies-produce-much-data-our-cars.

  16. Polina Marinova, “How to Lose $700 Million, Theranos‐Style,” Fortune, May 4, 2018, https://fortune.com/2018/05/04/theranos-investment-lost/.

  17. Felix Salmon, “What Went Wrong With IBM's Watson,” Slate, August 18, 2018, https://slate.com/business/2018/08/ibms-watson-how-the-ai-project-to-improve-cancer-treatment-went-wrong.html.

  18. Michela Tindera, “It's Lights Out For Novartis And Verily's Glucose Monitoring ‘Smart Lens’ Project,” Forbes, November 16, 2018, https://www.forbes.com/sites/michelatindera/2018/11/16/its-lights-out-for-novartis-and-verilys-glucose-monitoring-smart-lens-project/#18933d4f51b2.

  19. Eric Wicklund, “MHealth Engagement Issues Still Stand Between Wearables and Healthcare,” mHealthIntelligence, May 13, 2016, https://mhealthintelligence.com/news/engagement-issues-still-stand-between-wearables-and-healthcare.

  20. Lukasz Piwek, David A. Ellis, Sally Andrews, and Adam Joinson, “The Rise of Consumer Health Wearables: Promises and Barriers,” PLOS Medicine 13, no. 2 (February 2, 2016): e1001953, https://doi.org/10.1371/journal.pmed.1001953.

  21. Elizabeth Weingarten, “There's No Such Thing as Innocuous Personal Data,” Slate, August 8, 2016, http://www.slate.com/articles/technology/future_tense/2016/08/there_s_no_such_thing_as_innocuous_personal_data.html.

  22. Ibid.

  23. “The Piezoelectric Effect—Piezoelectric Motors & Motion Systems,” Nanomotion, August 28, 2018, https://www.nanomotion.com/piezo-ceramic-motor-technology/piezoelectric-effect/.

  24. “Nest Learning Thermostat—Programs Itself Then Pays for Itself,” Google Store, 2009, https://store.google.com/us/product/nest_learning_thermostat_3rd_gen?hl=en-US.

  SECTION 2

  Applying Data to Disease

  4

  Ava—Tracking Fertility, on the Road Toward Understanding All of Women's Health

  It seems like a simple problem. There are just five days in a typical monthly cycle that a woman is able to get pregnant. Identifying those days can be the holy grail for a couple trying to conceive, but, historically, the tools available to make an accurate prediction have been deeply flawed. There's the calendar method, where a woman uses past cycles to predict the length of her current one, with roughly 30% odds of identifying the right day
s—not much better than chance. There's the temperature method, which, for best results, involves taking a rectal temperature reading first thing in the morning and watching for a 0.4‐degree increase—an increase that tends to come at the very end of the fertile window, often too late to act on. And there's the cervical mucus method, which requires a woman to interpret subtle changes (as at least one website puts it1) from tacky to cloudy to slippery.

  No matter the approach, a complex, multivariate process is being reduced to just one set of measurements, just one source of data. Add to that the inconvenience of collecting that data and the potential difficulties in interpretation, and we're left with a real lack of useful and accurate information. Use whichever method you want, you're still not able to be particularly confident that you're identifying the optimal days to try for conception.

  From a patient equation perspective, ovulation—while informed by multiple inputs—is closer to the Nest Learning Thermostat than most of the issues we deal with in health care. It's a closed‐ended system with one yes/no output: is an egg being released? And while it's a complex equation to get results more accurate than traditional univariate measures, it is far less complex than what we'll see in later chapters. Ovulation is a baby step (no pun intended) into the world of applied patient equations, but in taking that step, we can learn a lot about how a business can successfully put these new technologies into practice.

  Enter Ava

  Ava is an ovulation‐tracking bracelet released to market in 2016, approved by the FDA in the United States (and sold in 35 other countries to date), and improving its predictions and insights month after month as it builds what is now the largest database of women's health measurements in the world. Ava noninvasively sits on a woman's wrist while she sleeps and, using temperature sensors, an accelerometer, and a photoplethysmograph (which detects changes in different layers of the skin), it collects information about pulse rate, breathing rate, heart‐rate variability, sleep duration and phases, skin temperature, and perfusion (blood flow)—along with data inputted by the user through a smartphone app about her menstrual cycle and when she has intercourse. Ava uses this multivariate data to predict (as of last report, with 89% accuracy) the five fertile days in the user's cycle—with minimal demands of the user.

  The algorithm behind the scenes is both diagnostic and prescriptive. It identifies the fertile days, and then tells a couple when to act in order to maximize the chances of conception (or when not to act if they're using it to help avoid an unwanted pregnancy, which is a use case the company is working on). Those two sides—the diagnosis, and then the actionable intervention—are what make for a useful application of the data technology. Telling someone that they're, for instance, on the road to Alzheimer's disease is of limited value if there's not also an action to take, an intervention that will alter the default outcome.

  An article in the New York Times described one woman who “discovered she had been missing her fertility window by about a week because her cycles were longer than normal and regular period‐tracking apps didn't pick that up.”2 Within a few months of wearing Ava, she was pregnant. There are countless other stories like that, for sure. In 2019, the company announced that 20,000 babies had been born to women using the Ava bracelet. But the bracelet is just the beginning—and the current algorithm is just the beginning as well. As the company collects and tracks more and more information from women around the world, its data scientists have two aims—(1) to continue to make the fertility predictions more accurate, and (2) to be able to make broader connections about women's health that no one has yet had the data to make it possible to consider.

  As just a few examples: how a woman's cycle impacts mood swings, headaches, and overall health; how to understand and manage the hormonal swings of pregnancy and, later, menopause; how one might even be able to use this data to find early signs of heart disease, cancer, and more. The company has already launched a study tracking a group of women hospitalized with preterm premature rupture of membranes (PPROM), looking to identify data patterns that could detect infection before it presents through traditional methods.3

  From Ava's initial goal of helping couples get pregnant, the plan all along has been to add functionality that can make Ava a value‐add for women no matter their age or reproductive goals. In an interview, Pascal Koenig, Ava's co‐founder and CEO, told me, “Our vision is to understand the influence of hormonal changes on women throughout their lives—from the very beginning, when menstruation starts, up until menopause, and maybe even after.”4

  Koenig continues: “It's about the entire reproductive journey, from issues on the conceiving side, with couples trying to get pregnant, and then also on the contraceptive side, with people who want to take hormones, and then during pregnancy itself. Finally, menopause, which is a completely underresearched area. Our goal is to come up with products and services that can impact people's lives at each of those points.”

  The data creates a feedback loop enabling continuous improvement—the more they collect, the more they can refine the algorithm, the more accurate the predictions can be, the more hypotheses that can be tested, the more insights that can ultimately be learned. With the opportunity, of course, to add more data sources should they prove to add incremental value in the algorithm, or should technology make them easier to capture.

  Koenig explains: “We started at the beginning with a smaller data set, and then we went through the data to really try to understand it, and see what we could add to make the prediction even better. Now we're using more and more factors no one would have expected. Of course our data scientists are making sure the correlations are real and not just effects we should be filtering out…but we see a lot of connections that we wouldn't have expected.”

  Some of those connections have been easy to understand, Koenig told me, and some have been more difficult to explain. They have looked at the hormonal effects of traveling to different time zones, exercising at night, sleeping badly, and more—all in the spirit of trying to isolate the layers that make a difference, and where the device can add value to end users.

  One of the studies the company cites on its website is from all the way back in 2000, a paper on blood pressure and heart rate and how they affect menstrual cycles5—but two decades ago, there was no practical technological way to measure that at home, to make Ava a reality. I wrote in the previous section about how we're at an inflection point in history: for the first time, we have the sensors to collect the data—and we have the computing power to turn that data into actionable information. In many areas, we had the sensors but we just didn't know what layers to look for. Here, we've known some relevant layers for a while—but we didn't have the necessary wearable technology until recently.

  The Changing Role of the Patient

  What something like Ava does is give the patient power that was never before possible with this level of accuracy. Instead of the fertility specialist holding all of the information in the patient‐doctor relationship—being the one who tells the patient what is happening in her body, and when to act in order to maximize the chances of conception—the patient gets to be part of her own care and own her information stream. The patient becomes empowered to proactively make decisions and treat herself, instead of just making sure she gets herself to the doctor. It's a small example here—just a few bits of information being transferred from the clinic setting to becoming trackable at home—but the idea extends further when we look at other conditions. The patient no longer has to be just a passive recipient of information.

  But there's another side to it. In exchange for owning that information, the patient is forced to take on more responsibility for his or her care. In a technologically‐enabled health care world, it's not enough just to make the doctor appointment. The patient is expected to use Ava consistently, to check the results, and to act on the information. And the life sciences company, for its part, is expected to make the information accessible not just at a level that can be und
erstood by an educated medical professional but also by a relatively unsophisticated (perhaps) end user. It's not that the doctor's role is eliminated—if someone gets pregnant using Ava, of course the first step is going to be to see the OB/GYN. But it is the case that the patient has more tools in her toolkit.

  As an industry, we've never really had to think much about information presentation, or about the accessibility of our equipment and devices. It was okay to be hard to use, to some extent, because our products were going to be used by trained professionals. That isn't necessarily the case anymore. Testing one's cervical mucus or basal body temperature is invasive and time‐consuming. You have to remember first thing in the morning. It interrupts your life. Ava is just a bracelet. It syncs to the phone automatically. It, to its creators' credit, has a spectacular engagement model to keep it charged—it syncs when you plug it in, creating a virtuous cycle of engagement and compliance. This doesn't matter in a clinic setting—no one worries that a blood test is too much of a bother when the patient is already at the doctor's office. But it matters a tremendous amount at home, without the supervision of a health care provider.

  From a business perspective, it's possible to see a range of different models playing out. Ava's website addresses potential customers—but there's also a section aimed directly at health care professionals. If someone is having a fertility issue, it's not out of the realm of possibility that a doctor could pull this out of his or her toolkit and recommend that the patient try Ava before moving to further treatment. It's also not out of the realm of possibility, from a payer perspective, that an insurer might say that before they pay for IVF or another expensive intervention, a patient must try something like this, to make sure it's not simply an issue of incorrect timing.