Fraud, waste and abuse (FWA) schemes like unnecessary genetic testing, fake Covid-19 testing and inappropriate billing of behavioral health services continue to plague the healthcare industry — and artificial intelligence (AI) is playing a larger role in such schemes. Considering how quickly AI is transforming other business sectors, it is not surprising that some bad actors are using it to commit various schemes, including generating false claims as well as cloned medical records that set the stage for boilerplate billing. Some may even create “deepfake” providers to support false claims.
Conservatively, FWA schemes cause financial losses that represent 3% of total healthcare expenditures, according to the National Health Care Anti-Fraud Association (NHCAA). Given that U.S. health spending hit a record $4.8 trillion in 2023, payers can expect to see at least $144 billion lost to FWA schemes each year.
What makes combating such schemes particularly challenging is that they are always evolving. Traditional, rules-based fraud detection and investigative analysis can help payers identify known schemes, but rules may be too rigid to identify new and emerging schemes. As a result, payers often need to play catch-up in the fight against fraud, particularly when resources limit their ability to maintain a large special investigations unit (SIU). However, adding sophisticated AI tools like machine learning (ML) to their anti-fraud arsenal can help plans identify the latest schemes more quickly than using conventional, rules-based logic alone.
To optimize their fraud detection efforts, plan leaders should understand how their organizations can supplement their human expertise by deploying these AI applications to detect FWA schemes with greater speed and accuracy.
The potential of ML to detect fraud
To supplement their traditional rules-based fraud detection systems, forward-thinking health plans are harnessing advanced AI approaches like machine learning to uncover inappropriate claims and identify problematic FWA trends. Two main types of ML algorithms can assist plans in these important efforts: supervised and unsupervised learning models.
Supervised learning: Supervised learning models can be trained to detect fraud through an iterative process that requires end-users to label data. In this approach, an experienced investigator validates the tool’s findings (such as labeling a questionable billing behavior identified by the tool as problematic and worth investigating) and provides expert feedback that helps the model learn from their input. Over time, these tools become better at flagging providers with suspect behaviors, so investigators can spend more time pursuing, rather than identifying, potential fraud.
Unsupervised learning: Unlike supervised learning, unsupervised learning models are not trained by an end-user. Instead, unsupervised models learn to identify FWA patterns through methods such as outlier detection. When used to supplement existing rules-based algorithms, unsupervised learning models can help SIUs identify emerging FWA threats more quickly and decrease the likelihood of missed opportunities.Trend analysis, another form of unsupervised learning, can also compare a provider’s billing behavior by code with their peers to uncover potential fraud. This demonstrates the value of detecting emerging FWA trends through multiple techniques including supervised learning when investigators are not limited to rules-based analysis.
Smart strategies for using AI to thwart fraud
Even though AI can add tremendous value to plans’ fraud detection efforts, it is not a panacea. As plan leaders consider integrating ML models into their mix, they should recognize that these tools must complement — not replace — human expertise. Here are some strategies for plan leaders to implement AI responsibly and effectively for fraud detection:
Recognize how AI and ML fit into the overall strategy. FWA prevention cannot be achieved using sophisticated models alone. Instead, it requires multiple investigation methods, complemented by a plan’s SIUs, to achieve the best results.
Look beyond your own plan. Using data-driven tools infused with AI that aggregate data from health plans across the country can help teams identify fraud trends more effectively. By tapping into a wide breadth of data, SIUs can identify schemes that would otherwise go undetected.
Dispel any misconceptions that investigators will be replaced by AI. Plan leaders should articulate the benefits of using AI models to enhance their team’s effectiveness, including the ability to focus their work on investigations and recovery without the burden of unvetted, false-positive leads.
Know the signs of AI-driven scams. Because of the ease with which bad actors can now duplicate medical records, plans need processes to detect telltale signs of fabricated records, such as a high percentage of duplicate diagnosis codes across services or patient ages that conflict with date of birth.
Use AI to verify tips from members. One health plan received a tip that a provider was billing for home health services not rendered and failing to reassess members’ needs. After data analysis confirmed that the provider was an outlier and had billed more than $1 million in improper claims, the plan worked with law enforcement to launch a criminal investigation. As a result, the provider agreed to pay $3 million for allegations of violating the False Claims Act.
Be patient. It may take time to achieve optimal results from machine learning, as training the models is an ongoing process. However, most plans find that having an effective tool to complement their investigative expertise is worth the wait.
A realistic view on the value of AI and ML
Even though plan leaders have reason to be concerned that bad actors will use AI to perpetrate fraud, they should also appreciate the considerable value of machine learning models to help them thwart emerging threats and decrease losses. By harnessing AI for FWA prevention and recovery, plan leaders can help their SIUs get ahead of the latest schemes, improve claims accuracy, mitigate their risks and operate more efficiently.
Photo: Feodora Chiosea, Getty Images
Erin Rutzler is Vice President of Fraud, Waste and Abuse at Cotiviti. In this role, she is responsible for the oversight and strategic direction of Cotiviti’s FWA solution suite. Erin has been integral in the development of Cotiviti’s FWA solutions over the past eight years. Serving as the company’s primary subject matter expert in investigations and FWA for compliance, client training, sales and marketing activities, she regularly represents the company at industry conferences such as the National Health Care Anti-Fraud Association’s (NHCAA) Annual Training Conference (ATC).
This post appears through the MedCity Influencers program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. Click here to find out how.
