Despite an expensive and high profile zero-tolerance crackdown, it’s believed more than $210 billion was lost in the US through healthcare fraud in 2014. It’s a staggering figure based on a 3-7% estimated fraud rate.
Could graphs be part of the answer? Let’s take a look at how graph visualization could be a valuable tool for detecting fraud, focusing on the Medicare Program.
The national Medicare program covers 54 million Americans and accounts for 14% of the annual federal budget, costing $492bn in 2013.
With any program at this scale, the scope for fraudsters is vast. The number and complexity of transactions makes detecting fraud seem like finding a needle in a haystack. The problems are compounded by the desperate state of federal and state-level computer systems. Records are spread across multiple data silos, held by different agencies with no simple way to cross-reference.
But what if Medicare data were to be viewed as a graph?
As with all fraud detection, the key is to understand connections: between patients, procedures, practitioners and clinics.
Using this model, we can identify ‘normal’ patterns:
We can therefore rapidly identify cases of probable healthcare fraud. Let’s take a look at some examples.
One of the most common, and costly, types of Medicare fraud is when practitioners bill for services they haven’t provided. Often this is done using social security numbers that have been stolen from seniors, either directly or using social engineering tactics.
Just last week, the FBI shut down one Medicare fraud ring operating in Florida using stolen social security and provider information to steal $28.3 million.
As millions of claims are submitted each month, Medicare authorities cannot check each bill with the patients concerned. In any case, many patients do not fully know or remember which services they received, and certainly don’t understand the obscure procedure codes used.
Rather than relying on tip-offs from members of the public, analysts could use our graph visualization technology to find potential cases of fraud.
Not using KeyLines or ReGraph yet?
Here we have loaded a dataset of all claims made by practitioners on a certain day. We can see a standard pattern emerging:
Generally, each procedure is connected to one patient and one practitioner. The practitioner is also connected to a location or clinic. This standard pattern makes it easy to spot anomalies, like this very productive doctor who appears to have claimed for 5 of these procedures in a single day.
This should also ring bells, especially if the clinics are geographically spread from one another. Aside from seeking a second opinion, patients tend to stick with a single practitioner.
In the same data, we can find networks like these showing two patients receiving the same procedures from two or three practitioners. This could be another indicator of a fraudulent claim.
If there is significant overlap in patients between two practitioners with no professional relationship, it may indicate two doctors exploiting the same stolen list:
Taking data from a longer time period, we can find these two practitioners share dozens of different patients, despite being based at different clinics. For an analyst, this should be a signal to investigate further.
These are just a few simplified examples of how graph visualization can be used to detect possible cases Medicare fraud. By benchmarking ‘normal’ behavior, outliers and anomalies become rapidly visible when visualizing our data.
Download our ‘Detecting fraud with KeyLines’ white paper to see how one of our graph visualization toolkits, KeyLines, has been used to visualize fraud data. You can also request a free trial of KeyLines and ReGraph, or get in touch.