Workers’ compensation fraud in America is a lucrative business, with fraudsters raking in billions of dollars annually, according to the National Insurance Crime Bureau. According to the Coalition Against Insurance Fraud, some of the biggest and costliest offenders in the fraud system include medical providers who exploit vast amounts of data in order to slip in phantom bills and counterfeit expenses, and perform unnecessary medical procedures on injured workers. Fortunately for workers’ compensation carriers, the application of artificial intelligence (AI) through fraud detection algorithms, machine learning, and other emerging processes can assist in helping identify fraudulent activity.

One of the biggest challenges for insurance fraud investigators is the race against the clock. The longer it takes to find evidence that a claim is fraudulent means more time a carrier is unknowingly paying for a fraudulent claim. This makes the road to recovering payments difficult. However, according to CLARA Analytics, AI insurance applications can mitigate and thwart four of the top tactics that medical providers use in filing fraudulent claims: gaps in provider information, fragmented payments, lag time, and long supplier chains.

For a long time, spotty, inaccurate, and constantly changing medical provider information has thrown insurance companies off the trail of fraudulent activity. Some fraudulent providers will switch states and change names in order to cover their tracks, and insurance carriers then struggle to gather sufficient evidence of fraudulent activity and claims. Data is closely tied to key identifiers, including names and addresses; medical providers who constantly change these details make rooting out fraudulent claims confusing and time consuming.

Analysts hardly see the whole picture when sifting through data related to fraudulent medical activity. Depending on state insurance claim database laws, carriers do not always make their fraudulent claim data public. Other carriers may then continue to make payments to the fraudulent claimant. If there are state mandates that ensure fraudulent claim information is made public, providers can simply switch their services to a different state with tighter legislation on how claim information can be shared.

Claim lag time can vary from company to company, but the length of time that passes between the time a claim is made and when it is discovered to be fraudulent can be a year or more. However, the time between when the bill for services is submitted to and paid by the carrier may be only 1 to 2 months. This means that fraudsters can receive multiple payments from insurance carriers before they even have the chance to notice red flags regarding evidence of fraud.

“A recent study found an AI program was
able to code 1.2 million claims in under
three hours, a task that would have taken a
human approximately 4.5 years.”

It can be difficult to decipher which specific physician or agency is at fault in a fraudulent claim. What’s more, attempting to gather data on each entity involved in the treatment process can be difficult for claims analysts. Determining the correct data sets that will capture all of the entities involved, and writing the algorithms to do it, has made investigating fraudulent claims tedious and challenging.

One of the biggest advantages that AI has over traditional methods of fraud discovery is the amount of time it takes for AI to search through data and discover patterns. A recent study conducted by the National Institute for Occupational Safety and Health and the Ohio Bureau of Workers’ Compensation integrated AI tools to auto-code workers’ compensation claims. The program was able to code 1.2 million claims in nearly three hours. It is estimated that it would have taken a human approximately 4.5 years to code that many claims. This is a great example of how effective AI implementations can be used to interpret and organize data efficiently.

AI can help track down gaps in provider information with supplemented data that is not attached to names and addresses. This includes taxpayer information, federal employee identification numbers (FEIN), national provider index numbers, and medical licenses. This aggregated information and tracking of multiple years of account information allows AI programs to effectively fill in the gaps in provider identification.

AI can effectively navigate privacy laws and access data on providers receiving multiple payments from insurers. By using multipayer databases and having a strict scope on the billing patterns of clustered networks, AI integration can measure suspicious and fraudulent activity while staying within the realms of Health Insurance Portability and Accountability Act laws and other state and federal privacy laws regarding employers and employees.

AI offers insurance carriers real-time bill tracking over multiple years and tracks procedure codes, diagnoses, and other subsequent claim actions. When a fraudulent provider tries to slip through the cracks by ordering treatment for an old claim, AI programs let fraud investigators know in real time.

AI programs help expose fraudulent patterns between entities by tracking sequences through accumulated data instead of looking at individual entities. These programs can also track the sequencing between in-house and inter-provider services by monitoring the patterns of billing between separate provider entities.

“AI effectively measures suspicious
activity while maintaining compliance
with laws and regulations.”

One main difference that separates an AI program from a run-of-the-mill fraud algorithm is that AI can find and root out patterns and make discoveries by learning and determining rules and instruction from the data that it receives. Even though many AI systems are still set to obey fixed system rules, the future of using AI to combat fraud within the insurance industry looks very bright.

© 2021 Applied Underwriters, Inc. All Rights Reserved.


Return to Almanac Main Page