AI is coming to detect healthcare fraud

Forbes’ insights brings us this article about the future of healthcare fraud detection, How AI Can Battle A Beast—Medical Insurance Fraud. This is a future of fraud detection that is inevitable. With the rise of the amount healthcare data it is almost a forgone conclusion that computers are going to be integrated into battling healthcare fraud. This article only proposes how AI can detect three of the top ways that healthcare fraud is committed but be rest assured they are working on several different aspects. Lets begin with what they said about the top three issues and how AI can fight them.

Billing for services patients never received 

The problem: This could take the shape of healthcare providers billing for services that a patient never actually received during a visit or even billing for entire visits that never happened. This type of fraud requires examining claims and sensitive patient files.

The AI solution: The common saying in healthcare is that “if it isn’t documented, it didn’t happen,” Trzcinski says. AI can help sift through data to see if there is documentation to prove that a patient actually received the service that was billed.

“Upcoding” for more expensive treatments

The problem: A simple procedure might be billed as much more complex, and costly, by providers looking to pad the bill. This is known as “upcoding,” because a different code is used on billing forms for the more expensive service. 

The AI solution: Here, AI could use anomaly detection to understand what the typical treatment is for a specific condition and detect any deviations by a provider from their peer group. It then flags the anomaly as potential fraud.

Kickbacks and corruption

The problem: A kickback involves some form of “quid pro quo,” or something offered in exchange for something else. A physician could be offering patient referrals to one specialist in exchange for referral fees, for instance. (This is illegal in federal healthcare programs.)

The AI solution: AI tools could analyze a mix of behavioral and transactional data to discover corruption. Corruption is based on influence, Clopton notes, and that usually doesn’t show in the general ledger. It might require looking at disparate data sets such as the history of referrals made by a physician as well as travel expenses they had covered by a company or person benefiting from those referrals.

Luckily, the are aware of the exceptions…

As AI advances to handle the complexities of thwarting fraud, human review seems likely to remain a key element in determining whether someone intended to be deceitful.

Proving fraud rests heavily on proving intentional wrongdoing. If something was done incorrectly—there was a wrong billing code applied or there was a typo in a health record—that can still be costly and time-consuming to fix, but an accident is not fraud. And AI can’t always determine on its own—yet—whether something was intentional. 

“It is easy to detect an anomaly, but an anomaly doesn’t mean something bad,” Esman says.

A healthcare provider could simply be participating in a study, for example, that means the patterns in their treatments are different from the typical behavior of other physicians in their field. That could be flagged by an AI algorithm because it is different, but it would require human review to understand the conditions of why that difference is present. 

In this ever changing world of healthcare fraud allow Patient Options to guide you through how to effectively run your clinic in a compliant, cost-effective manor.