Redesigning K Health's pediatric Body Picker through usability research, iterative testing and close collaboration with clinical and technical teams.
A parent reports that their child has a rash. The next question sounds simple: Where is it?
Translating what they see into structured medical information is harder than it sounds, yet precise location can directly affect clinical assessment.
Instructions were confusing, feedback was imprecise, and users could not always tell whether a specific selection had been registered correctly.
To select a specific organ, users had to actively zoom in, adding friction and unnecessary clicks.
Even after selecting a specific organ such as “hand” in zoom mode, the feedback could return the broader “arm”, reducing confidence in what the system had recorded.
Changing a previous selection was tedious and unintuitive.
The illustrations felt sterile and overly medical, misaligned with K Health's warmer, human-centered visual language.
The feature did not meet accessibility and usability standards.
Help parents confidently express the precise symptom location without having to think like the clinical system underneath, while still capturing structured, clinically meaningful data.
The goal was not to copy another picker. I wanted to understand which interaction patterns helped users stay oriented, select precisely and feel confident that the system understood them.
I compared direct manipulation, tags, body maps and zoom patterns across medical products, looking at feedback, accessibility, visual language and how much complexity was exposed at each step.
I combined those patterns with direct user feedback and usability evidence from the existing flow to define the first design hypothesis.
The first redesign translated the research into a new interaction model that aimed to reduce friction, support more precise symptom reporting, and align better with K Health’s visual language.
I ran remote sessions in Lookback and observed how parents interpreted the new pattern, switched between body areas and recovered when a part was difficult to find. I used a multi-location scenario designed to expose where the interaction broke down.
Test scenario “Your child has a rash on the chest, hands, ears and tongue. Show us where it is.”
Testing confirmed the overall direction while exposing a new tension: visual simplicity could make clinically important details harder to find.
"I don't associate 'other place in the face' with the eyes because it's too generic. The eye is a very specific and sensitive organ."
Free-form selection solved part of the usability problem, but raised a product question with developers and data scientists: should users be able to select locations that did not exist in the clinical ontology, and what would that mean for data structure, model training and medical interpretation?
Fewer, broader hit areas keep the first view calm and reduce cognitive load.
Regions are combined in the overview, then separated into clinically meaningful areas in focused views when more precision is needed.
Tags sit alongside direct selection, providing a clear alternative for details that are difficult to represent visually.
After Round 1, I refined hit areas, clarified the hierarchy, and ran a second testing round to see whether the changes made selection easier and more precise.
For handoff, I defined each illustrated layer by its role as a touchpoint or zoom trigger and mapped it to the correct clinical ontology. This helped engineers map interactive zones precisely and turned the redesign into a reusable, scalable component.
More precise body-part selection reduced frustration during symptom reporting. Fewer users dropped off and more completed the assessment.
More specific symptom information reduced the need for clarification. Conversations became shorter and more focused, improving clinical and operational efficiency.
