Response to short-term spike in glucose level
Response to longer-term growth in glucose levels
Daily tips & tricks - personalized on specific profile of user
High cost of blood-sugar tests, combined with the incredible importance of monitoring as part of managing diabetes means that low-income-individuals are unable to manage the disease. Over 80% of diabetics live in low-income cities, and there are over 75 million urban slum dwellers suffering from the disease (without the ability to buy the monitoring tools required).
What it does
Matchstick offers several main benefits to users:
- automated monitoring and SMS subscription, where users input the results of their tests directly in through SMS, and a system searches the inputs for long-term trends.
- personalized tips & tricks based on consumers' current diabetes progression timeline
- preventing emergencies with consistent communication and fast-response
- reporting to doctor remotely
Why it's amazing
Matchstick offers several amazing advantages:
- It's cheaper (2% of the cost of blood-glucose monitoring, while between 80-95% as accurate)
- Faster and private, saving long and expensive trips to the hospital / doctor for regular readings
- It's pain-free, removing one of the biggest barriers for monitoring
- Removes the risk of blood-borne diseases, since no finger-pricks are required
- Provides hyper-tailored monitoring and feedback to those who would otherwise not be able to access it.
How it fits what you're looking for
- Using SMS instead of smartphone app to tailor usage toward those who would not otherwise have access to monitoring solutions
- Using urinalysis instead of blood and combining with remote logging, which has never been done before
- Combining data from all users to increase accuracy over time
- Variety of APIs and tools user: Geocoder, Twilio, Google Maps, Dictionary.com
- Advanced multi-regression analysis to allow software to "learn"
- Natural language processing, both custom and with libraries
- Extremely error-proofed and theoretically crash-prone
- Individualized & personalized to every user
- Outlined above: cheaper, faster, pain-free, accessible, no blood-borne disease risk, hyper-personalized
How I built it
Using Ruby on Rails and a PostgreSQL database to store the users. Twilio API is used to interface with SMS numbers, and Dictionary.com, Google Maps, and Geocoder are used to locate hospitals + provide definitions. Custom multiple-regression algorithms were built to help improve the accuracy of the monitoring solution over time. Custom fuzzy-match was created for blood glucose level inputs.
Challenges I ran into
- Tying PostgreSQL database into easy-to-code frontend website
Accomplishments that I'm proud of
- Creating my own code to parse through XML in Ruby and extract the needed information from HTTP requests
- Developing algorithms that can correct for the limited accuracy of urine test-strips versus blood-testing, allowing for close to 95% correlation between the two methods of testing.
- Building custom fuzzy-match codes for medical- and diabetes-specific user inputs
What I learned
While creating Matchstick, I learned:
- Ruby on Rails is an extremely flexible language, with gems for almost everything
- Fuzzy-matching algorithms are blessed for helping make software more error-proof
What's next for Matchstick
Given more time, Matchstick will:
- integrate Uber / Lyft ASAP in case of emergency
- order test-strips directly through SMS
- printable summary reports on customized time basis