Care Compass

Inspiration

According to the United States Census Bureau, 8% of the United States population did not have health insurance for the entire year. In the greater Houston area alone, we have that 1.17 million people are uninsured, and even more do not have insurance plans which do not adequately meet their healthcare needs. Our inspiration for developing Care Compass comes from our desire to make healthcare more accessible, especially for marginalized groups who face discrimination in the healthcare world.

What it does

Care Compass serves as a web-based platform for users without health insurance to identify low-cost, accessible clinics that meet their needs. Users can search and filter for clinics and their procedures, receive AI summaries on specific clinics, and receive 24/7 guidance from a fully-integrated chatbot through plaintext communication.

How we built it

We built our project in React + TailwindCSS. Our backend was seeded with data on a week-to-week basis from a database hosted in MongoDB. We used Google Cloud to store images and host our website, which we deployed using Vercel. We used the Google Maps API to display the location of nearby clinics, as well as the Google Gemini API to integrate AI summaries and an AI assistant chatbot.

Challenges we ran into

One of the primary challenges our Backend faced was deploying to Google Cloud Run. We are all new to writing Dockerfiles and configuring containerized applications, so we ran into a issues surrounding authentication in the CLI and getting the build process to work smoothly. For our AI-team, it was their first time working with both Gemini and the Gemini API, and it was their first time engineering prompts for non-personal use. Ultimately, this required a few hours of tweaking our prompts to get the output just right. The other problem that popped up was that our chatbot couldn’t remember the chat history, which we fixed by having the frontend query our database via the chat’s session ID – which the backend creates and maintains when the chat session opens. On the Frontend, there were many, many cosmetic errors (overfull h-boxes, to steal a useful term from LaTeX, in particular) which required significant tweaking to eventually fix.

Accomplishments that we're proud of

Firstly, Alejandro is quite proud of the logo. Generally, we are proud of having made a working project out of several components with which we had minimal prior experience. The AI team is very proud of fixing the chatbot's memory problems, the frontend is immensely satisfied with getting the UI responsive to different user choices and giving the user enough options to target their particular issues. The frontend was also pleased to incorporate a mini map into our website’s filters page.

What we learned

We learned how to properly use the Gemini and Google Maps APIs to create excellent user experiences. We also learned about and implemented best practices for creating a web environment and communicating between front and back end code as well as git best practices.

What's next for Care Compass

First and foremost, we want Care Compass to be built for the needs of the underinsured and underserved groups everywhere. As such we will be adding more clinics across the US to our database as well as getting more accurate pricing data. This data will need to be amalgamated from the chargemasters of multiple hospitals, which legally must be made public, other sites which track this data, and potentially even user reported pricing.