A mobile interaction checker leveraging deep learning to predict drug and food interactions, empowering and educating users on multi-drug use.
Through EasyMed's homepage, users can effortlessly check for drug-drug and drug-food interactions, access curated daily articles on polypharmacy, and access personalized pill reminders.
Effortlessly search or scan a drug by barcode to check interaction details with meds in your list, stay updated on side effects and food interactions, and add to list for future reference.
Just snap a photo of your food items or product barcode, and the app performs a thorough analysis, alerting you of any potential adverse interactions with your list of medications.
Keep a detailed drug list for future reference during doctor visits, and receive customized medication reminders to ensure timely and accurate dosing.
Access personalized articles to enhance your understanding of polypharmacy, with the option to filter articles by category and enjoy interactive quizzes for added engagement and learning.
First Place & Bonus Prize
EasyMed secured first place among 760 global participants and earned the Best Integration of Responsible AI bonus prize for best exemplifying Microsoft's commitment to Responsible AI principles.
Featured OnlineEasyMed has been showcased on the Microsoft Azure Developer Community platform, and received recognition on Facebook (@MSFTDev.US) and Twitter (@MSTCommunity).
High User Satisfaction
Task success rate: 94.44% (34/36, benchmark: 78%)
NPS: 80% (4 promoters, 1 passive)
User satisfaction: 100% (3 highly satisfied, 2 satisfied)
To understand the current landscape of the problem, I researched extensively online and found that many adults mixed drugs together and the problem is even worse for older adults since they are often taking more than 5 drugs for various medical complications.
To collect qualitative data on how different stakeholders perceive adverse drug interactions, we conduct 6 in-depth interviews with 2 older adults, 1 adults with multi-drug use and 1 without, and 2 healthcare practitioners and here are the key insights:
“Build a platform that allow users, especially older adults who are taking multiple drugs together, to check and educate themselves about poly-drug use side interactions so that they can avoid damage to their health.”
After evaluating 5 websites and 5 mobile competitors, it's evident that most of them exhibit usability issues such as information overload, poor UI design, and limited functionality. Recognizing the importance of this, the inclusion of severity levels and explanations for these interactions would greatly enhance usability and user experience.
Synthesizing research findings, I brainstormed 17 features and narrowed down to 6 P0 features we want to implement first based on impact, hackathon requirements and implementation cost. (See Brainstorm Doc)
Keeping our core users in mind, I utilized the Usability Heuristics and the Microsoft Responsible AI principle as the guideline for our design.
Accessible DesignProper font size, use of icons and visual imagery
Privacy & Security
Password protection & Username resetting
Help & Support
Onboarding guide and confirmation windows
Transparency Include disclaimers and warnings
I sketched quickly on iPad for quick iterations and discussed with team members and our instructor for feedback.
Participants
To validate our research and identify relevant problems, we invited 5 participants to test our prototype including 2 older adults, 2 healthcare professionals, and 1 oridinary adult.
Mythology
We first asked them a set of semi-structured interview questions to gain general background information. We then asked them to finish 6 tasks, observe their performance, and asked follow-up questions to learn their thoughts. The plan and results are attached here (See User testing plan)
Insights
We used an affinity map to group the feedback and determined what we can improve given the limited time and resource. (See map and priority matrix)
Participants
To validate our research and identify relevant problems, we invited 5 participants to test our prototype including 2 older adults, 2 healthcare professionals, and 1 ordinary adult.
Mythology & Analysis
We conducted semi-structured interviews to gather background information and evaluate task completion. We then use affinity map and priority matrix to group feedback and identify key areas for improvement. (User Testing Plan, Map and priority matrix)
Research Insights
Before
After
Participants reflect that it's hard to understand the complicated medical terms.
Side effects are presented without explanations
Offer explanations for medical terms from notable sources
Why I use a bottom sheet?
Bottom sheet offers context and information without disrupting the reading experience.
How can users learn more?
We obtain definitions from authoritative sources, and users can access them with a single click.
Research Insights
Before
After
Older individuals had a hard time reading the long text and understanding the severity
Long paragraphs of text for users to read and comprehend
Users can listen to the text and visualize severity levels using graphs and tags.
How is text-to-speech implemented?
We utilize Azure's Cognitive Services for greater accuracy.
Colors Indicators
Color indicators are employed to aid user comprehension of severity levels, with red signifying severe interactions and orange indicating mild ones.
Research Insights
Before
After
Users feel at lost about what the app does when they first entered
First time users go to the sign up page directly.
We will start with an onboarding tour to guide users on how to make the most of the app.
Since the logic of our app is complicated involving ML model caculation, back-end storage and front end demonstration, my role is to thread all components together into one compact app. This figure shows how our app works technically.
Refine AI Model
We would like to refine our machine learning model to support analysis of allergen and dosage information in relation to interactions.
Personalized Support
Add more features such as medication reminders, search by text, and food image recognition.
MyChart Integration
Integrate with digital health record system like MyChart for more transparency in data sharing between patients and physicians
.svg)