top of page
GlucoSense
A simple way to map out your life and make sense of your glucose data
Overview
About
HCDE capstone project sponsored by Microsoft, aimed at helping diabetic patients better manage their conditions and reduce their burden.
Problem
Diabetes management depends on interpreting blood glucose patterns with contextual data like activity, diet, and sleep. Patients often find recording this information burdensome, making it hard for providers to recommend treatment changes.
Solution
An AI-powered tool that seamlessly logs contextual information, maps it with blood glucose data, provides activity-based insights, and automatically captures notes from provider appointments for future reference.
Timeline
January - May 2024
Team
2 UX Designers
2 UX Researchers
Role
User research
Ideation workshop
Prototyping
Usability testing
Logo design
Animation creation
Impact
Received the Capstone Innovation Award. The next step is to leverage AI to further tailor content to each user’s condition and reach out to potential healthcare organizations for product integration.
Research
Why focus on diabetes?
Did you know that diabetes is on the rise globally? In the US alone, 38.4 million people, or 11.6% of the population, have diabetes. That’s one in every ten individuals. This alarming statistic drove us to explore ways to help diabetic patients manage their condition and prevent new cases.
Where we can help?
To identify where we could provide meaningful assistance, we conducted research to better understand our target users and the broader problem space.
Literature Review
We began with a literature review to explore the diabetes management landscape, uncovering challenges and opportunities. We found that effective management requires a multi-faceted approach involving medications, diet, exercise, and monitoring. Many patients feel overwhelmed by the abundance of information, making it difficult to take effective action or adjustments.
Competitive Analysis
We analyzed 10 competitors to assess the current AI-powered diabetes tools and identify gaps. The analysis revealed that most tools focus on specific areas like nutrition or exercise, without offering a comprehensive solution.
Interview
To gain first-hand insights, we conducted interviews with healthcare providers, patients, and AI expert.
Healthcare provider
We interviewed 5 endocrinologists to validate the problem space and understand the treatment processes for diverse patient populations. We found that providers often lack the necessary context to interpret patients’ blood glucose patterns effectively. This missing context includes factors like physical activity, diet, stress levels, and sleep, which are crucial for recommending treatment changes.
"Just because someone's wearing a CGM doesn't mean they're equipped to do pattern recognition...CGMs collect blood glucose data but you have to know what to do about it."
- Provider
Patient
We spoke with 5 patients to learn about the challenges they face in managing their diabetes. They reported that consistently recording contextual information is burdensome, especially when spread across multiple platforms. During consultations, while providers review blood glucose data and suggest treatment changes, patients often struggle to remember all the details discussed and to adjust their routines accordingly.
"It's hard to think back on what I did. Honestly I can't tell you. Sometimes it's unexpected but if there's an event that day, I can probably recall it."
- Patient
We consulted with Dr. David Rhew, Global Chief Medical Officer and VP of Healthcare at Microsoft, to explore the feasibility and potential impact of the AI tool we envisioned. He emphasized leveraging AI to synthesize and present information in an understandable format and highlighted the importance of making data collection as easy and frictionless as possible.
AI Expert
"Consider multimodal data AI and passive ways to capture [patient] data ...the less you can ask the patient, the greater the adoption."
- David Rhew, M.D
With the insights we gathered from our research, we realized that interpreting patterns in blood glucose data is essential for successful diabetes management. However, one critical element — contextual information — is currently missing. This gap led us to focus on the following question to address the problem:
"How might we help patients effortlessly collect contextual data to make sense of their blood glucose levels?
Define
Understand our users
As we progressed, we developed a persona of our target users to clarify who we are helping and ensure we think from the user's perspective throughout the design process.
We also created a journey map to illustrate Elvia's day-to-day life, identifying the pain points she encountered and the opportunities we could tap into.
Elvia
26 • PhD student
"It’s a burden to remember to record everything, especially when I’m busy."
About
Elvia is a busy PhD student with Type I diabetes who visits her endocrinologist every 3 months. Her endocrinologist wants her to track her diet and exercise for better diabetes management, but she’s too busy to do it consistently.
Personal Characteristics
• Willing to learn
• Curious
• Data-driven
• Solution-oriented
Hobbies & Interests
• Hiking
• Skiing
Goals
• Understand what works and
what doesn’t for diabetes
management
• Spend less time managing
and more time living life
Needs
• An easy way to record daily
activities
• An accurate way to carb count
• Visibility into how her daily
activities correlate with her blood
glucose levels
Based on these pain points and opportunities, we established design requirements to guide the subsequent process, ensuring our solutions effectively address patients' needs.
1 Seamless user input: make it as easy as possible to collect data from users
2 Personalized insights: offer personalized insights based on patient activity
3 Effortless notetaking: automatically record information from consultations and
summarize key takeaways
4 Transparency & control: offer transparency on how AI generates information and give
users control to edit or correct
User Journey
We also created a journey map to illustrate how Elvia would use our solution in her day-to-day life.
Design
Turn insight into innovation
Co-designing with patients
We began by developing a solution based on our research insights, visualizing how it could integrate into patients' daily lives using a storyboard. To ensure our product truly meets patients' needs, we conducted a co-design workshop.
During the workshop, we collaborated with 3 participants to gather feedback on our overall concept. Participants highlighted concerns and identified problematic areas. Together, we brainstormed solutions to address these issues. Key takeaways from the workshop include:
● Offer various data input methods to minimize user effort
● Provide flexibility in notifications
● Allow input of qualitative info such as events (e.g., traveling, illness), stress levels, and mood
● Give users control (e.g., edit AI-generated content, turn off ambient listening)
Used a storyboard to depict how our solution supports users in their daily routines.
Conducted an online workshop with participants to gather feedback and refine our solution.
Building the solution
Leveraging insights from the co-design workshop, we developed a prototype to visualize our solution and its key features. We began by crafting the information architecture, which guided the app design and illustrated user flows.
Information architecture outlining the app's structure and user flow.
Feature 1: contextual map + flexible logging options
Pain point
Patients often struggle to remember to log meals, activities, and other important details due to their busy lifestyles, making the process feel tedious and overwhelming.
Solution
Our app integrates data from existing apps and wearables, utilizing ambient listening to identify activities through background noises. This approach allows patients to effortlessly map relevant life activities without additional effort.
To address any gaps the app might not detect, we offer flexible manual input options. Patients can take photos, record voice notes, or type their entries. For example, by using the camera to photograph their meals, our app automatically calculates the carb count and logs the nutritional details, simplifying the tracking process.
Offers manual data entry options and maps logged information with CGM data for comprehensive tracking.
Capture a photo of the meal before and after eating, and the app will calculate the calories and carbs.
Users can also record daily activities by typing entries or using voice notes.
Feature 2: personalized insights
Pain point
While Continuous Glucose Monitors (CGMs) provide general patterns but lack personalization. Patients often feel frustrated not knowing what has changed or what they can do differently based on their current condition.
Solution
Our app analyzes patients' logs to generate tailored insights, helping them understand their blood glucose trends and offering actionable steps to improve their management.
Offers daily, weekly, and monthly insights to help users understand their progress and identify areas for improvement.
Feature 3: ambient assist
Pain point
Patients often struggle to recall key details from their healthcare consultations due to the number of topics discussed.
Solution
Our app includes an ambient note-taking feature that patients can activate during their consultations. This feature automatically captures key takeaways and notes from discussions, allowing patients to easily review important points and action items later. This ensures they can follow through on recommendations and better manage their care.
The Ambient Assist feature summarizes discussions and organizes the content into categorized sections, making it easier for users to review and understand key points.
Validating with patients
We conducted unmoderated user testing of our prototype to gather feedback on the overall concept and identify areas for improvement. 6 participants took part in the tests, and we discovered several important insights:
● Some terminology, such as "ambient assist" and "events," was unclear to users
● Users preferred varying levels of detail in content, highlighting the need to align content
depth with individual preferences
● An onboarding flow is necessary to set clear expectations and provide users with an overview
of the app's purpose and features
Participants' feedback was categorized by feature, grouping similar themes together to summarize what worked well and areas for improvement.
Refining the design
Based on user testing feedback, we improved our prototypes and updated the design system to develop high-fidelity prototypes and an onboarding flow that better meets user expectations and needs.
Feature 1: context mapping + flexible logging options
Feature 2: personalized insights
Feature 3: smart notetaker
Onboarding flow
Context Mapping
Personalized Insights
Smart Notetaker
Responsible AI
Mitigating AI risks in healthcare
When incorporating AI into our product, responsible design is essential. Two major risks we identified are: a) inaccurate information leading to user misjudgment, and b) privacy concerns due to data collection and storage. To mitigate these risks, we've put safeguards in place.
To address inaccurate information:
● Implement a feedback loop to collect user input on AI-generated content
● Include disclaimers during onboarding and within the app to caution users
● Allow users to modify or delete inaccurate data
● Highlight potentially unreliable or high risk information
To address privacy concerns:
● Maintain transparency around data collection and storage practices
● Provide users with full control over what data is collected and stored
What's next
As we look ahead, there are several potential areas for future development:
01 Provider dashboard
Develop a provider-facing dashboard to give healthcare professionals access to patient insights and help them evaluate patient progress.
02 Increased interactivity
Enhance the app's interactivity to boost patient engagement and encourage consistent usage.
03 Implementation feasibility
Collaborate with AI experts and developers to evaluate the technical feasibility and potential challenges of implementing our solution.
04 Tailored content
Introduce personalized content options that adjust to the user's health literacy and native language, improving accessibility and effectiveness.
Reflection
Key takeaways
01 Time management
One area for improvement would be better time management. Finalizing our solution and its scope earlier could have given us more time for thorough designing, testing, and iteration.
02 AI product design
Designing with AI was a crucial learning experience. It highlighted the importance of responsible design, ensuring transparency, safety, and reliability for users.
03. Value of diverse perspectives
Healthcare is complex, and incorporating diverse perspectives was invaluable. Engaging with patients, providers, and experts helped us validate our findings and identify clear opportunities for innovation.
bottom of page