Inspiration
Because of a doctor shortage, Indian telemedicine, which has become an essential component of healthcare in rural areas, uses less trained "Spokes," often nurses, to screen patients. Patients with more serious illnesses then meet a doctor. But, doctors believe they still spend too much time asking questions the nurses could ask.
What it does
Spokes currently use a largely static list of questions. We devise and build a demonstration of a more dynamic algorithm that suggests questions to Spokes.
This algorithm aggregates only existing data from medical records and does not need extra work from medical workers. This is important since the Indian telemedicine system is in high demand, with 20+ people often in line for a single Spoke. This demand also precludes slower algorithms and longer question lists.
How I built it
We built the demonstration in the Flet Python UI framework with a Supabase backend.
Challenges I ran into
The constraints of the problem limit the scope of solutions. In rural India, where the telemedicine system is the most beneficial, networks are slow and unreliable, people speak many languages, and access to technology is varied. This makes solutions requiring strong internet connections or computers, such as large neural networks, impractical.
Accomplishments
The demonstration contains an authentication and user setup flow made using a new frontend framework in only several hours.
What I learned
Through Flet, we built a desktop app and used an imperative UI framework for the first time, and practiced polymorphic object-oriented Python.
What's next
Run a controlled simulation to compare our algorithm with existing systems. We have already written the framework for the simulation.
There is currently unused information in medical records, such as the author of an entry, that we could use to improve the algorithm's scoring.
Built With
- flet
- python
- supabase
Log in or sign up for Devpost to join the conversation.