Utilizing health history and clinical reasoning for healthcare guidance
Disclaimer: This was a 6-week class project for TECH 21: User-Centered Design for AI Applications
My role
Lead Designer
What I did
User and Market Research
Product Design
Presentation Deck
Team
William Callahan
Dawn Martin
Kalyan Kuppa
Shogo Toiyama
Pallavi Mishra
Rupa Chaturvedi, Mentor
Timeline
July - August 2024
Respondents ages 18 - 45
To find out, we turned to patients. Adults, Students, Working Professionals. Everyone has had a question about their health at some point. And we wanted to find out what they do next.
Online, people typically use Google or WebMD for quick answers, but the information is usually generic and overwhelming, and for some, catastrophizing. Which can be scary.
Seeing a doctor provides professional advice but is time-consuming, expensive, and not always immediately accessible.
We found 4 consistent themes
Impersonal
Many existing health resources, like WebMD, offer generic information that don't account for individual differences. This one-size-fits-all approach can lead to confusion and misdiagnosis because it ignores your unique medical history and specific needs.
Incomplete
Resources don't have complete patient history, limiting their predictability and accuracy. This can lead to unnecessary anxiety or a false sense of security.
Costly
Healthcare today is extremely expensive and lacks affordable options
Delayed Access
Waiting for appointments or navigating general information is time taking
Goals
Easy, instantly accessible, live answers
Pain Point
Children fall sick very often
Goals
Affordable alternatives
Pain Point
Expensive physician and urgent care today
Goals
Filling the knowledge gap between health professional interactions, reinforcing learnings
Pain Point
Want more information
We were originally MediChat, but we quickly realized there was already another one of those. So, we went with VitalAnswers.
These are some logo iterations I put together during early explorations. I ended up choosing to use 1 and 3 because of their minimalistic nature and therefore, ability to scale down.
It understands your symptoms in natural language and asks follow-up questions to narrow down possible diagnoses with probability assessments.
Based on your inputs and patient data such as medications, demographics, and other conditions, it provides clear next steps and probable diagnoses.
Our vision for VitalAnswers to be a comprehensive, personalized health management system that integrates directly within Electronic Health Record systems. This is an example of Kaiser's portal- but it very well could be Epic, Athena, or Meditech as well.
Starting small, focusing on a key use cases
A potential direction could be to further define our use cases and perfect the model for those scenarios. Also, what defines an emergent versus a non emergent case? We still have some things to find out.
Probable or certain?
Probability isn't good enough. When it comes to health, people want an answer. Simply displaying probability assessments won't cut it. This is a core principle that we need to keep in mind moving forward.
