MedTrack

Inspiration 

MedTrack is more than just a medication tracker that sends you reminders. We wanted to create a solution that also includes how medication interaction can be harmful and wanted to create a one stop shop for it.

What it does

MedTrack is built as a website with 4 main pages after you log-in to your own account: Today, Calendar, Medications, Interactions. The Today page lists your medication and allows you to mark them as taken. You can also enable reminders for taking your medication. The Calendar tab is a visual option to see what medications have been taken and what has not been taken per day. The Medication tab lists all your current medication and gives you the option to add more medication. When adding a medication, the user will provide details like the medication name, the dosage, the timeline it needs to be taken, and the frequency (like once per day, twice per day, etc). Lastly, the interactions page shows a graph of all the medications that might have possible interactions. It ranks the interactions on severity and suggests next steps for the user to take like seeking medical advice.

How we built it

We used React, TypeScript, Vite, Tailwind, and Supabase for storage. We split the work up three ways (tracker, interaction engine, safety intelligence) and each used Claude Code to help us with the actual programming portion of this assignment. Interaction works by classifying the drugs into broad types (NSAIDS, anticoagulants/antiplatelets, SSRIs/SNRIs/TCAs/MAOIs, benzodiazepine, etc) and deriving the severity based on the types that the drugs are from

Challenges we ran into

We struggled a lot trying to use an interaction database. At first we wanted to pull the data from a reliable source by using APIs but a lot of the research proved difficult due to cost constraints. So we ended up hard coding our interaction database for 250+ commonly described medication. The interaction data itself which includes which drugs interaction and at what severity level were not completely verified against a clinical source. For the purpose of the hackathon, we wanted to have a finished product that connected to our vision so we will mark the interaction database as a feature to improve and verify in the future.

Accomplishments that we're proud of

We were able to ensure maximum security for the medication data using supabase's authentication and an extra recovery key that only the user has access to in case they forget their password. Nothing is stored on the browser or in supabase's data other than the hashed password (which cannot be backtracked without the extra salt).

What we learned

 Learning to use Claude Code was definitely something we became weary of. Using its functionality to edit and create code made our work easier, but when it came to combining our work there were some issues with some of the assumptions that Claude made for each person. So working with AI and learning to be skeptical of everything it produces was something we learned. Technically, we learned how to ensure security of health data, and to make a simple website using AI tools while accounting for all our requirements!