Techweek

It’s K Health vs Google

By    |   October 1, 2018, 5:37 am EDT

What if every time you were sick and googled your symptoms, you were not overloaded with impersonal, generic responses? New York-based K Health is trying to add a level of authenticity to your online symptom check.

K Health is a free personal health information app based on a proprietary approach called “People Like Me”. This idea, that could make googling in the dark a thing of the past, recently raised a $12.5M venture round. The participating investors include Mangrove Capital Partners, Lerer Hippeau Ventures, Primary Ventures Partners, the Box Group, Max Ventures, Bessemer Ventures, and Comcast Ventures.

Image may contain: phone

PC: K Health

K Health has built a completely data-driven health app because of its access to a unique dataset of over one billion health interactions, which include physician notes, lab results, treatments and prescriptions. It’s aim is not proactive care like other startups but to help people cut down on the time spent reading confusing and misleading information online. “We all know the feeling of waking up at 2:00 a.m. and researching our symptoms online even though we know it’ll only make our anxiety skyrocket, ultimately making things worse,” Allon Bloch, the co-founder and CEO of K Health said. “Today, if you go online and search for something as simple as a cough, you’ll see millions of results ranging from the common cold to cancer.”

This simple and rather predictable problem statement got Bloch, serial entrepreneur and former CEO of Vroom, Wix, Dolphin Software, thinking over how to convert generic information to personalized, insightful responses and launch K Health in 2016 with co-founder Ran Shaul.

The Siri For Symptoms

K Health operates through an android or iOS app where users can chat with what’s called ‘K’, or an “AI assistant for health” any time they feel under the weather. Users answer K’s questions about their symptoms to discover how people with similar medical histories and biological traits such as age, gender, and body mass were treated in a clinical setting.

K Health’s blog explains how it works: Say you’re a 35-year old man with moderate low back pain that comes on gradually and gets worse when you move. You click through a popular symptom checker to see what it might be, and the top two results are cauda equina, a serious surgical emergency, and lumbar spinal stenosis, a condition that’s extremely rare in men of this age. Tell K about these symptoms, and she’ll show you that 90% of guys at this age with these symptoms had a low back strain.

PC: K Health

Or, for instance, say, you want to know about something more common such as persistent headaches: K will ask you a series of questions such as “how long have you had headaches’ or ‘tap where you feel the pain’ in a diagram in the app. The app would then formulate the results which can show how many people have had the same symptoms. The user might then be advised according to what those with the same symptoms did—say, they went to their primary care doctors—or how doctors treated people like the user when they had the same symptoms. In doing so, K hopes to equip consumers with personalized information that not only cuts the unnecessary time spent researching but also allows users to make better decisions about health.

But should one be asking Dr Google or Dr K in the first place?

Although dangerous, even doctors believe that looking up information online is now a standard. Therefore, the least that can be done is reducing the hysteria associated with it. “I don’t want you to read that you have ‘insert dreaded, invariably painful or fatal disease’ after being sick for 4 hours”, writes Dr Neil Brown, the Chief Medical Officer of Kang Health. “I want you to have access to current, up to date, relevant information on your symptoms.”

Therefore, what a more reliable symptom-check approach does is that it makes it clear to the user that extreme diagnoses are uncommon and that statistically, many people have had those symptoms. It then shows you how others dealt with it and thereby reduces the stress associated with looking up symptoms on a generic platform.

While other symptom-checks are available online, they’re based on rules: rules which show how certain symptoms add up to diagnoses for hundreds of conditions or rules that factor in genetic differences. The problem, however, in a rules-based approach is that one will need infinite rules in infinite possible combinations in order to replicate the complexity of human experience.

K, on the other hand, is not based on rules but artificial intelligence. “We let K learn the rules by observing millions of real cases—just like doctors learn during residency, except faster,” Ran Shaul, co-founder of K Health says. K has had access to 15 years of real anonymized health data, including medical notes from millions of doctor visits which, the company claims, has helped them understand the connections between the symptoms, diagnoses, tests, and treatments people experience. “K turns these learnings into a kind of medical ontology, the language of experience, that she can use to talk to users anywhere in the world.”

This slideshow requires JavaScript.

The other, obvious benefit with automation is that K can operate on a different scale, seeing more patients in a day than a doctor does in his/her entire life. Going forward, it can also tell users what medications were used to treat the symptom they have or what tests were ordered.

But at the moment, K Health, which is HIPAA compliant, has partnered with a network of providers for same-day appointments and remote advice for more efficient care. One such partnership is with Integrity Family Care where the K report can be shared with the medical care provider who can then advice rest, medication, test, or a visit.

Care, the startup assures you, is only delivered by a doctor and rightly so, because K’s AI is at a nascent stage. Numerous K Health app users have left reviews on Google Playstore complimenting the idea but complaining how the symptom-check was just like Google or WebMD, “inaccurate”.

 

  •  
  •  
  •  
  •  
  •  


Leave a Reply

techweek-1