The issue of co-pay card fraud has been in the news for some time, but with the increased prevalence of high-priced specialty products, the cost of missing this fraud – literally – has become too high for manufacturers to rely only on traditional techniques, such as human investigators.  This isn’t to say that there’s no role for people in stopping co-pay program fraud – they’re absolutely crucial!  They just can’t be the only tool used to combat fraud, as human intervention only is simply too costly and inefficient in the face of today’s fraudulent pharmacy.

So if we can’t rely solely on trained investigators to ferret out fraud, what should we do? The answer is that we need to employ sophisticated analytics to alert us to potential indicators of fraud early on.

Broadly speaking, analytic models work well in fraud detection because they’re adept at identifying patterns in data and at detecting outliers—or values that don’t seem to make sense. At TrialCard, our analytics team has built a tool called “Rx SpotlightTM,” which identifies unusual activities among co-pay program transactions that could be indicative of fraud.

The goal for any detection system is to be perfect—and therefore to identify all cases that are “truly positive” (i.e., fraudulent) and to ignore cases that are “truly negative” (i.e., not fraudulent). Unfortunately, data—as in life—rarely presents us with perfectly clear evidence. Instead, we have to rely on methods of estimating and predicting fraud and work to continuously refine those methods to make them as accurate as possible.

All of which brings me back to my grad school days—which is unfortunate—and one particularly gnarly topic, called Bayes Theorem. While the calculations were mind-bending, the key concept behind this was pretty simple:  How do you know whether something you think is true is really true?

You can visualize this in a table like you see below:

Bayes Theorem Chart

Any detection system seeks to minimize the number of “false positives” (i.e., things you think are fraudulent, but aren’t) and “false negatives” (i.e., things that look legitimate, but aren’t).

Let’s use a real example:

“Say a pharmacy has been averaging 500 fills per month for a particular medication for 12 months. But in the next month, it averaged 1,000 fills. Do we think this pharmacy is engaged in fraud?”

The very unsatisfying answer to this question is that we don’t know. It could be that the manufacturer has deployed a new marketing tactic to increase product sales and it’s working really well. Or it could be that the pharmacist has decided that he can make a bundle of money by faking some scripts. Or it could be something entirely different.

We simply can’t tell from the data alone whether this is fraudulent activity or not—but our Rx SpotlightTM tool can certainly flag this pharmacy as having some strange activity.

This is where the role of the investigator comes in. Our investigators are trained in law enforcement and interrogation techniques that allow them to dig deeply into suspicious cases. But this takes time and therefore costs money—so it requires that our Rx SpotlightTM tool present only the most likely suspicious cases to our investigators. Otherwise, our investigators will be overwhelmed with cases that could look suspicious, but really aren’t (i.e., false positives). Typically systems like these get better over time and the models start producing fewer, but higher quality cases for the investigators to assess.

In the end though, effective fraud monitoring is the essence of a collective effort—using the best of what analytics and human beings can bring to bear in order to provide the best possible means to separate legitimate activity from fraud.

Paul LeVine, Vice President, Strategic Business Partnerships, Corporate Strategy

To learn more about TrialCard’s pharmacy fraud solutions, visit its Resources Library.