Breakdown of Bill Form Details
Flagged Errors
Home page

Inspiration

Many of us have family members and friends who had to undergo huge amounts of stress to pay off the medical bills. However, many people might not have been aware that some of the high charges listed on the medical bills could have been covered by insurance or reduced. No one should have to carry the stress of any unnecessary high charges to their medical bills. At the end of the day, making sure people are able to receive the care and treatments they need to improve their health is what matters more.

What it does

This website allows user to submit censored medicals bill that can quickly be cross referenced and evaluated by the AI to check for discrepancies in the amount they were charged and notify them if it feels they were overcharged for anything that might be refundable.
It does this by taking the data that the user provides in the form of a fully uploaded medical bill and comparing it to information from hospitals, insurance providers, and any other relevant sites to check if the amount the person was charged lines up with what was done in their visit and their demographic. If it does not line up it will notify you and tell you where the issue showed up and what it thinks the issue is caused by.
This saves the average person a lot of time and ensures that they do not get charged more than they absolutely need to.

How we built it

Frontend: We asked Bob to create a frontend that will recieve a photo upload of the medical bill and send it through fastapi, and then handle the finalBill output.

Backend: We decided that the general framework would go something like this: Reader >> Censor >> Gemini >> Calculator >> Outputter. Of which the pseudocode would look something like this.

userImg = userinput();

userJson = Gemini.py(Privacy.py(userImg));

finalJson = Calculator.py(userJson);

return finalJson;

Having decided that the pseudo code of the program would look like this. We hashed out what each function would look like.

Privacy.py(userImg)

userImg = censor(textalgorithim(userImg));

Gemini.py(userImg)

userJson = pydantis(GeminiApi(userImg)); return userJson;

Calculator.py(userJson);

finalJson.append(userJson); finalJson.append(duckdb.query(userJson.procCode, userJson.procName)); return finalJson;

We decided to use fastapi to call the backend. We used Tesseract and Regex to detect text on the image, and Pillow to manipulate the image for censoring. We used Gemini's api package to gain access to Gemini's doc reading capabilities, and put it through pydantis to make sure it formats correctly. We used duckdb to store hospital insurance information. We then inputted the architecture into Bob and asked it to generate python code for those files.

Challenges we ran into

It was difficult to find a standardized medical bill form to test MedBill Shield with as many medical bill forms vary in format and the type of information they ask for.
It was also a challenge to figure out the local censoring for ensuring that the documents uploaded by the user protect their privacy before they are then sent to the website for evaluation.

Accomplishments that we're proud of

We were able to convert the different screens of MedBill Shield from Figma that included all of the features on there into functional, working screens.
We were able to successfully analyze and audit a trial run medical bill which proves it can work.

What we learned

We learned how to use Figma to prototype MedBill Shield to highlight the different interactions and features users will have to navigate through.
We learned to gather important information from our research that further drove our rationale in developing MedBill Shield.

What's next for MedBill Shield

The current iteration of MedBill Shield has some limitations that have great potential for expansion in the future.
Right now the website is limited in what hospitals and insurance companies it can consider ,most of them being in the bay area, but it is possible to expand it to be nation wide in scale.
It could become more polished as a website allowing for people to use logins to safely and quickly navigate their uploaded medical bills.
It has potential as a business as competitors to this idea run successful companies with similar end goals just without the AI usage.