The Challenge: Inefficient Data Analysis
Medical fraud accounts for a huge cost annually - approximately $68 billion, according to the National Health Care Anti-Fraud Association, or about 3% of national healthcare spending. Detecting this fraud is extremely difficult though, and analyzing the data takes a long time due to investigators needing to search then parse through results to find individual data points.
A venture capital firm was looking to build an AI product to find doctors committing billing fraud. However, they needed an effective way to display data to help analysts and investigators to see patterns that could indicate fraudulent activity.
Our Process: Human-Centered Design Agile
Our team worked closely with subject matter experts on how they currently investigate fraud and what data patterns paint a picture of potential fraudulent activity. After many interviews and observations, a major pain point the team saw was the need for a summary view of the doctor data. The team then went through a series of conversations with different mockups to home in on the best way to present the data so that it’s easily understood.
The Solution: A Simplified View to Detect Potential Fraud
The team created a single profile page that showed aggregate statistics so an investigator could get a quick snapshot of the doctor. The profile provides a total landscape across procedures, prescriptions, diagnoses, and more to see if anything stands out as a warning sign.
Given that there was a lot of information on the page, the team needed to determine what information was most important and presented that above the fold so that investigators could find it faster.
The focus on the most important information also helped the team roll out the solution in a piecemeal fashion, so value was delivered quickly and iteratively, rather than having to wait for a return on investment.
The Outcomes: Efficient Analysis of Data
Significant Time Savings
It previously took investigators asking analysts for a summary view of the doctor data. This caused interruptions for the analysts and required them to manually run different queries and send the results as they were computed. The investigator then needed to create their own aggregate view manually.
Now the investigator can look up summary statistics for any doctor in the system instantly, without involving the analyst and having to wait up to a day.
Potentially More Effective Fraud Identification
With the most relevant data front and center in a single location, investigators had a single view of doctors. They could now look at different data points at once to see if any - or several of them - showed signs of fraud.