80% of medical bills are wrong costing the average American over $800 per year on average. With 1 in 2 Americans in medical debt and the average $10,000 medical bill containing over $1,300 in errors, we felt like we had to solve this problem. We wanted to empower patients to take back control of their medical care by providing them with powerful patient-focused tools.

There is a trend towards more transparency in the healthcare system. Three new federal laws (Hospital Price Transparency Final Rule, No Surprises Act, and Transparency in Coverage Final Rule) give patients more negotiating power and data. We wanted to use technology to allow patients to benefit from this legislation without effort.

What it does

With Dragonfly, patients upload a copy of their medical bill and get a summary of all the charges on their bill. Once, they have an understanding for what the bill is for and how insurance covered them, patients go through a flow where Dragonfly scans their bills for errors. Patients go through each error type step-by-step and provide information to confirm or deny the possible errors. Once the errors have been detected, patients see a summary of the overcharges on their bill, along with step-by-step guides and instructions for how to correct their bills.

How we built it

We used three sources of data to detect the errors (They are all available to the public)

To detect unbundling we used the NCCI edits from the CMS:

  • The dataset contains all illegal unbundling pairs for CPT codes.

To detect upcoding we used the consumer description files from the AMA (American Medical Association)

  • We analyzed the dataset to create rules and algorithms to automatically flag possible cases of upcoding. User input is then used to determine the severity of the upcoding.

To detect inflated charges we used the provided Turquoise Health dataset

  • We compare the price for each CPT code on the itemized bill to the statistics in the dataset to flag inflated charges. Inflated charges are when the provider charges above appropriate rates for some service. The algorithm can only detect inflated charges for a limited amount of CPT codes as the provided datasets were of limited size. With the Transparency in Coverage rule, the majority of payer rates data is accessible so the algorithm can easily be scaled up by incorporating more data.

Challenges we ran into

The datasets we used are large, it took a while to understand the data to be able to draw useful insights from it. We decided we wanted to detect all 6 categories of billing errors so there was a lot of different data we had to analyze. We solved this by splitting up the datasets between team members where each team member focused on creating an algorithm to detect one type of error.

Accomplishments that we're proud of

Our goal was to detect all 6 categories of billing errors using publicly available data. We're really proud to have accomplished that over the short period of the hackathon.

What we learned

We learned that this problem is even larger than we realized. After reading about the current billing processes in the healthcare system we realize that patients need to be awakened. However, the positive note is that the trend today is to increase transparency and the government is actively creating legislation to solve the issue.

What's next for Dragonfly - Medical Bill Error Detection

After realizing how big the problem is, and how opportunities are opening up to solve this issue through scalable technology, we strive to continue working on our solution and testing it with real patients. We want to continue with the project to use the information to put control back where it belongs, with patients.

Share this project: