How We Built a Better Health App
Updated: Feb 28, 2020
Health is inherently personal, and all the details matter. Your age, gender, symptoms, and underlying biological factors can dramatically change your potential diagnosis and treatment options. Your doctor takes all of this into account. But what happens before you get to the doctor’s office? Research shows that a majority of adults look up information about their health online, and that should come as no surprise. We live in a world where we expect instant access to information, and technology has allowed us to personalize so much of the web.
So why is digital health information still so impersonal?
We’ve mastered curated content and personalized product recommendations, but on the web health information isn’t dynamic, personalized, helpful, or trustworthy. It’s impersonal, overwhelming, and often scary. Search results for symptoms or conditions are seldom relevant for me. I always end up with the lowest common denominator. And I usually freak out.
Meanwhile, the real information about your health, your medical record, is scattered across various EMRs and insurance company systems that you may not know how to access. You can’t utilize your data, and your doctor may not be able to see it to help deliver the care you need.
Fifteen months ago we set out to build a health app that could help consumers better understand and manage their health. Today we’re thrilled to introduce you to K, the world’s first AI assistant for your health that is built from the ground up using real anonymized medical data from people just like you.K will learn your biometrics and symptoms and show you how people like you were diagnosed and treated. You can use K to track your health and to get accurate, personalized information to help guide important healthcare decisions. In short, you can take control of your health.
K isn’t like any other health app out there. Here’s why.
Most health information on the web is static content that doesn’t get updated. For more personalized information, you might try a symptom checker. But most of today’s symptom checkers are rule-based systems. There are rules for how certain symptoms add up to diagnoses for hundreds of conditions. There are rules for factoring in biological traits such as age, gender, blood pressure, and body mass. Add hundreds of symptoms and rules for their attributes and values so you can distinguish between, say, abdominal pain from heartburn versus appendicitis.
The problem is, you need infinite rules in infinite possible combinations in order to replicate the complexity of human experience. To really get it right, you’d need an army of doctors to write rules for years.
With K, we took a totally different approach. Rather than write the rules, we let K learn the rules by observing millions of real cases—just like doctors learn during residency, except faster. We gave K access to 15 years of real anonymized health data, including medical notes from millions of doctor visits. Soon she understood the connections between the symptoms, diagnoses, tests, and treatments people experience. She turned these learnings into a kind of medical ontology, the language of experience, that she can use to talk to users anywhere in the world.
On top of that, we taught her to have intuitive, dynamic conversations with anyone she meets. She follows up to keep learning about how your symptoms are evolving over time. As long as K has experience with the condition and has the symptoms in her ontology, she can point you to highly personalized information about the diagnoses and treatments that doctors recommended to people like you.
If you’re game, you can see how K’s results stack up against other symptom checkers out there. Or just get the app and try it for yourself. In time, K will know your health better than anyone, and you’ll be able to share reports with your doctors. Together we envision being able to predict problems before they arise, making personalized preventive medicine a reality.
By Ran Shaul, Co-founder and Chief Product Officer,
K Health Inc.