Smarter health care through Ai
The founders of K Health gave me an interesting challenge – design a product that can become your digital health companion, that will answer questions about your health as accurately as a human doctor can, will be used by millions of users, and possibly save countless lives.
Team
K Health Product and Marketing Teams
Illustration: Ashley Farlow
Logo Development: Christian Rocha
Outcomes
K asks smart questions about your symptoms and accounts for your age, gender, and medical history as it investigates your health. In minutes, you’ll see how people like you were diagnosed and treated by doctors when they were in your situation.
We designed a UI that can dynamically generate any question that K would need to ask about the user's health condition. The product uses a custom set of UI elements to accomplish this task.
K shows you how doctors diagnosed and treated people like you who had similar symptoms.
The 'People Like Me' screen was one of the most challenging views in the application to design, it needed to provide the user with confidence in the system as well as educate them on the various conditions their case matched.
Learn what to watch out for as well as the tests and medications people like you needed to take to get better.
Without the guidance of a trained medical professional. This created anxiety in patients and extra work for health care providers.
The patient's core need was still valid – how can I take charge of my own health? How can I get access to affordable health care without having to make a trip to the doctors office?
When you’re developing a health care app everything is important, but you need to start somewhere. We began by designing the conversation between our Ai doctor K and a patient suffering from a headache.
We wanted to develop a product that felt warm rather than clinical and use levity to diffuse the nervousness of talking to a doctor.
In a 6 month period we designed and built an alpha and tested it on a few hundred users. Our app was polarizing – people either loved it or thought it was a toy. Here’s what we learned:
In the background our Ai team was building an Ai doctor that could diagnose over a hundred primary care symptoms on 20 years worth of medical records. Our product would need to create a dynamic conversation between the patient and the doctor.
Just like a real doctor, K would need to ask you many questions as it needed to rule out all the things that you could be suffering from.
K would dynamically generate the next question that meant we needed a component approach to the system so the application could choose the appropriate UI element.
I worked with our data science and medical team to define and create the different input UI the application needed to gather data from the patient.
Symptom cards would allow us to speed up the conversation by describing a symptom faster than a paragraph of text. They were also memorable.
We developed a library of over 250 symptom cards to help people discuss their symptoms with K.
I sat down with our chief medical officer and had him describe how a doctor asked a patient to identify where the pain is located.
I then developed a comprehensive set of anatomy illustrations that would enable the patient to pinpoint with accuracy where the pain was located.
Our results component needed to display the condition, high level of information about it, confidence level and serve as a call to action to take patients to the treatment page for the condition.
The earlier versions felt cluttered and it was unclear that you could tap on them to go to the treatment pages.We removed the number of cases that the prediction was based on and created a clear call to action.
You can think of the results page as alternate versions of the patient's future. Each condition will have a different treatment path, tests that may be administered, drugs that will be prescribed, and the expected recovery time.
The treatment pages sought to give the patient a clearer picture of what their future would look like.
Our first version relied on the data K could parse from our data set. It turned out to be spotty and insufficient. We were also working with the idea that patients could upvote or downvote these results based on their experience.
We moved away from this practice. When you’re working with a medical system the system should have the gravity of a physician.
In our next iteration we focused on demystifying the content. How could help patients make sense of what the results of a lab meant? What did they need to know about the drugs that they would need to be prescribed.
We quickly spun up a content team that wrote out the treatment paths based on years worth of medical knowledge. The redesign brought together developers, medical staff and writers to rapidly launch a new version.
Patients are only part of the equation. For our K to be helpful we needed to connect you to healthcare providers that could help you get treatment.
I designed both an application for doctors to review conversations sent over by K, and the ability to communicate with the patient.
We worked to get a first version of our Physician portal into the hands of doctors. We started very simple, cutting out any functionality to the point that it was almost painful.
The new version heralded the bigger vision of building a vertical for health care, connecting K with doctors and providing prescriptions and ordering tests through the app.
I helped develop a flow for finding and scheduling an appointments with physicians.
I learned a lot from branding K. The most important lesson is how little I knew about designing a visual identity that scales. I painted us into a corner with some of my color choices.
I would hire a good branding agency in a heartbeat the next time I build a brand from scratch.
With K's logo we wanted express the personality behind K's persona; a friendly, accessible health assistant.
The colored dots gives K a way to communicate when she is thinking and working on a case. The dots also represent the millions of medical cases the system is built on.
We employed a set of playful colors for our symtom cards and second more muted set to handle more sensitive medical content.
In the product we needed to be able to distinguish between the product conversing with the user and larger blocks of medical content. We used Vag for K and Circular as our work horse font for UI elements and content.
We experimented constantly with how to depict people in our marketing collateral. One direction was to keep people anonymous by not displaying their face. This helped emphasize the private nature of our service.
