Home Health Care Lessons from advertising: How AI can empower better self-care

Lessons from advertising: How AI can empower better self-care

9
0
SHARE

Several years ago, after giving birth prematurely to a healthy child, I was beset with severe insomnia. My primary care doctor didn’t know what was wrong with me. He suggested that I was just stressed at work, but he couldn’t prescribe medication because I was breastfeeding. Left to my own devices and desperate for relief, I began tracking my sleep patterns with a wearable sensor connected to an app on my smartphone. I gathered this data, found patterns, and was able to hack myself back to health. A few years later I learned from a psychiatrist that I had gone through an atypical postpartum depression.

At the time, I was the Founder and CEO of Pixability, a Boston-based software firm that helps companies target video advertising on YouTube and other social media platforms. Pixability uses machine learning techniques to optimize outcomes for clients. After my episode of postpartum insomnia, I thought, “Why can’t the same types of artificial intelligence that inform everything from social media advertising to Netflix and Spotify be used to help patients better care for themselves?”

To answer that question, I started looking into the available sources of health data. To no one’s surprise, I discovered that we’re swimming in health-related data, everything from electronic health records to data from mobile apps, wearable sensors and remote monitoring systems. There’s also plenty of data available about environmental conditions that impact health. What’s lacking is the ability to delve into this data at scale to identify patterns that can be useful to the individual, in much the same way that detecting patterns in YouTube media can be useful to advertisers.

Medical records alone are not the solution. While they are essential to understanding a patient’s health history, medical records suffer from having been derived from snapshots in time, such as a medical visit or a procedure. By contrast, continuous monitoring can produce a far more complete view of a patient’s health, whether of vital signs or sleep or activity patterns.

Of course, there’s no shortage of continuous data—it’s streaming right now from a broad array of home and mobile devices, ranging from glucometers, asthma inhalers and digital blood pressure cuffs to smart watches that can measure your heart rate, your oxygen level, the number of steps you take, and when and how long you sleep. The problem with this data is twofold: patients have difficulty making sense of it, and most doctors don’t yet know what to do with it. Apple Health has taken the first step toward integrating mobile data with downloaded EHR data, but the app doesn’t try to help consumers interpret the combined information. It does, however, make the information available, at a patient’s request, to developers of third-party apps that can help the user understand the data.

Big Data at the individual level
That’s where artificial intelligence comes in. AI can throw a wide net over not only EHR and mobile data, but also over other kinds of health-related information. The use of this Big Data in healthcare has already resulted in some advances: for example, the late IBM Watson Health made progress on improving cancer care, and the University of Pennsylvania did pioneering work with machine learning to predict sepsis and pneumonia in hospitalized patients.

However, machine learning has yet to be successfully applied to consumer-generated or mobile health data or to the social context in which people make their health decisions. Doing that could help us to truly personalize care for individuals and help us figure out how to improve patient adherence to their care plans. Equally important, individuals could be empowered to care for themselves to an extent that has rarely been seen up to now.

Eric Topol, MD, the eminent author, editor and director of the Scripps Research Translational Institute, predicted in his 2014 book The Patient Will See You Now that eventually, patients themselves will provide most of their own primary care:

Where once patients could not even access their data, today they can actually generate and own it…Today patients can rapidly diagnose their skin lesions or a child’s ear infection without a doctor. That’s just the beginning…There isn’t a chronic condition that isn’t amenable to sensor, lab and smartphone monitoring. Once the data are captured, it’s just about validated algorithms that provide continuous feedback to the patient.

Social and environmental context

To take Dr. Topol’s vision one step further, consider that the effective treatment of many complex chronic conditions requires more than a review of health data – it requires an understanding of the patient’s social and environmental context. This is true of complex mental conditions like depression, anxiety and bipolar disorder. Some physical conditions also vary greatly from one person to another. Among these are asthma, chronic pain, migraine headaches, and auto-immune conditions.

If a machine learning algorithm could analyze every available source of data, as well as the results of questionnaires administered frequently to patients, it would be able to detect patterns that affect both the individual and other people like them who have similar health conditions. With sufficient data, the AI could begin to provide targeted recommendations to users with complex conditions.

For example, if a person had severe asthma and there was a weather front coming in, they could be told it might be a good idea to postpone a scheduled workout.

Increasing patient engagement

As any doctor will tell you, the biggest obstacle to improving the health of many patients is that they don’t adhere to their care plan. If someone is obese and has diabetes, for example, their physician may tell them to exercise more and eat less or differently. But that doesn’t mean the patient will follow the advice or be able to stick with a diet or a treadmill regimen for long.

That’s where understanding the patient’s social context comes into play. Perhaps the machine learning algorithm has noticed that the person is an art lover. If their doctor has told them to exercise more, the app might advise them to spend a couple of hours walking around in an art museum. Getting those carbs down is also a challenge, but maybe a reminder that the patient is planning a beach vacation may induce them to ditch their evening potato chip habit. Personalized triggers that spur micro-behavioral changes can make a world of difference in self-care.

Data related to health is all around us in quantities that no one ever dreamed of in the past. It’s time to gear our machine learning tools to this data so that people with chronic illnesses—and especially, complex chronic conditions—can get the same kind of recommendations that they use every day to pick their movies or music or buy goods on the Internet.

Photo: mrspopman, Getty Images

Source link

LEAVE A REPLY

Please enter your comment!
Please enter your name here

4 + 10 =