ePROM+ | Devpost

Login screen.
Questions are shown in a social-media style layout with sliding bars instead of fixed buttons.
When a decline in wellbeing is detected, we actively ask the user for more information through voice recordings, which is analyszed via NLP.
The built-in voice recorder gives the patient the possibility to give more personalized feedback to the doctor.

Inspiration

Studies show that (electronic) Patient Reported Outcome Measures (ePROMs, basically regular surveys) impact the lifespan of cancer patients very positively. However, the response rate is rather low due to multiple pain points in the customer journey:

link

"It is too difficult to answer"
"The layout and overall design is not attractive"
"The intervals of the surveys are too frequent and cause survey fatigue"

link "Your surveys are not personalized" "Your surveys aren’t powered by conversational AI"

link

"It’s badly timed — it’s either presented to the consumer immediately"
"It’s unfocused — the survey asks too many questions about too many topics, the results are confusing and difficult to prioritize"

link

"questions are too complex"
"Personal questions at the beginning"

By tackling some of these challenges, and thus increasing the response rate of these ePROMs, we can have a direct impact on the patient’s health. We plan to achieve this by leveraging the full potential of technology.

What it does

Our app encourages the patient to regularly fill out their ePROM by (1) having a pleasant UX and making it simple to fill out the questions (2) providing immediate feedback on symptoms via a dashboard (3) Leveraging NLP (Natural Language Processing) to enable the patient to do a voice recording of their symptoms

(1) User Interface

social media style scrolling
modern UI elements
intuitive gestures (in the future)

(2) Dashboard

see your last reports
statistics about your well-being

(3) Natural Language Processing

speech to text to edit your response later
summarisation for a fast overview of your experience
highlighting of keywords to emphasize important parts of the answer

How we built it

React frontend: Using React to build the frontend UI and deploying the to GitHub pages.
Flask (Python) backend
Integration with external APIs (Google Cloud Speech-to-Text, NLP Cloud, Monkey Learn)

Challenges we ran into

different experiences & knowledge of team -> find a common ground
communication in remote team -> Slack, Google Hangout, Figma

Accomplishments that we're proud of

organized regular meetings across different time zones
collaboratively implemented several conceptual improvements over the current ePROM standard
created a working prototype under time pressure

What we learned

assembling a team takes a lot of time -> start well before hackathon to be ready to go
deciding for technology and setting up everything (deployment, repos, ...) takes a lot of time -> do before start of hackathon

What's next for ePROM+

building the necessary security into it (health data)
polishing frontend for a clean user interface
set up proper backend with database
evaluate if external APIs for NLP are financially feasible or if proprietary models & pipeline should be run in cloud