AI: Finding the Fraud Needle in the Big Data Haystack

PinIt

A systematic AI approach used against fraud, waste, and abuse not only uncovers specific instances, but often provides additional insights.

With fraud losses accounting for about 3% of the nation’s overall health care spending each year, pharmacy benefit management firms would be wise to seize the opportunity to battle endemic fraud, waste, and abuse (FWA). The upside can be significant, according to the National Health Care Anti-Fraud Association, which estimates annual losses at $68 billion nationally.

While the savings are potentially huge, identifying instances of FWA hiding in the massive data volumes generated by prescribing, dispensing, and covering medicines is no easy feat.

See also: AI Brings Real-Time to Fraud Detection and Prevention

Advancing technologies and the shifting fraud landscape

FWA can occur anywhere within the health care system, originating from prescribers, pharmacies or members engaged in myriad schemes like:

  • Phantom pharmacies and ID theft.
  • Diabetic test strip resales.
  • Telemedicine and online dispensing of prescription meds.
  • Compound prescription fraud.

Certainly, among the reasons fraud and abuse are on the rise is the health care industry’s rapidly advancing technology. This embrace of all things digital creates new avenues for criminals’ illicit exploits and enables them to operate with relative ease and anonymity. The silver lining? The related data trail can likewise serve the scammers’ undoing.

Indeed, fighting fire with fire is emerging as the industry’s best defense: using advanced analytics technologies like machine learning and artificial intelligence (AI) to stem fraud losses. Such powerful capabilities make fraud investigators’ jobs easier, detecting and red-flagging suspicious activity – or finding those fraud “needles” in the big data “haystack.”

Some pharmacy benefit managers (PBMs) have deployed anti-fraud analytics platforms to help their health plans wage battle with fraudsters. The most enterprising of these early adopters are wielding the power of AI.

AI’s exponential value

Given the incredible decline in computing costs in recent years, AI has become a viable option for most companies. It can run on low-cost servers instead of supercomputers, as was required in years past.

AI, a subfield of advanced analytics, consists of software that continually learns from experiences, adjusts to new inputs, and accomplishes specific tasks without human intervention. In fraud detection, AI spotlights potential instances of fraud, saving investigators from much of the more tedious, time-consuming work and helping optimize their investigative resources.

For example, the most comprehensive fraud-fighting platforms include data visualization capabilities. When the system identifies anomalous data patterns indicating potential fraud, investigators can create network diagrams that show interconnections between plan members and prescribers or pharmacies. These visualizations help illustrate the links inherent in widespread fraud.

When a systematic AI approach is used against FWA, benefits managers not only uncover specific instances, but larger insights are often revealed. They can, for example, identify emerging fraud trends and institute process changes to stop those trends from spreading.

Importantly, AI won’t replace human investigators. However, when used appropriately, AI can make investigators more effective. Computers are ideally equipped to pore through massive data sets, do complex calculations and automate operations. People, on the other hand, bring uniquely “human” capabilities to the table – common sense, intuition, creativity. Together, experienced fraud investigators teamed with AI-capable anti-fraud technology are a yin-yang combination, each making the other perform better.

A Prime example

Minnesota-based PBM Prime Therapeutics supports nearly two dozen Blue Cross Blue Shield plans across the country. The company processes claims for 66,000 pharmacies, which includes benefits for 28 million members of its clients’ health plans.

Last year, Prime embarked on an ambitious project to implement AI and advanced analytics aimed at slashing fraud, waste, and abuse. Through close partnerships with its health plan clients, Prime’s new SAS analytics platform analyzes not only drug prescribing and dispensing data but also medical claims data. That’s data from members, prescribers, and pharmacies, fully integrated together, creating a holistic data picture that’s unprecedented in the industry.

In the first year since the platform’s deployment, Prime saved the 23 Blue Cross plans it serves $279 million – $51 million in recovered payments, plus $228 million in cost avoidance, where Prime detected signs of FWA before claims were paid.

And the project’s value extends beyond dollars and cents. The platform also helps Prime combat the scourge of opioid abuse among members of its client insurance firms. The National Center for Health Statistics estimates there were 68,557 drug overdose deaths in 2018, most of them due to opioids.

Collectively, members looking to illegally obtain prescription opioids have created a national health crisis – but a single incident might involve just a few hundred dollars, an amount easily overlooked by an insurer or PBM.

Analytics changes the playing field. For example, Prime’s anti-fraud platform can red flag a prescriber who has written an inordinate number of opioid prescriptions, or a plan member who has seen ten doctors in two weeks and filed 20 opioid claims in that same period. Appropriately alerted, human investigators can dig deeper and instigate action.

In the first half of 2019 alone, Prime’s new analytics platform surfaced 721 cases for investigation. In one case, a member saw 53 prescribers in one year and visited various hospital emergency rooms 58 times seeking prescription opioids. Another case centered on a physician-owned pharmacy that only billed for eight different drugs – a data outlier that led to the unraveling of a $352,000 dispensing scheme.

The platform also turns up instances that are not intentional fraud but simply the result of unintentional error – a 30-day drug supply accidentally billed as 30 “units,” for example, and other simple data entry errors. Though benign in intent, the cumulative toll of such mistakes can be significant.

The better way forward

Whether due to deliberate fraud or unintended waste, financial losses drive insurance premiums and health care costs ever higher. Embracing data analytics, there is a better way forward.

And anti-FWA analytics is a great place to start – but, with AI, why stop there? Advanced analytics can help avoid negative drug interactions, monitor drug therapies, and assist drug formulary selection. Analytics can also inform better treatment protocols, better provider education, and better policy decisions.

Wherever health care organizations choose to focus their innovation, robust analytics, and AI can deliver insights that measurably increase the dollars available to the health ecosystem and help save lives.

Jo-Ellen Abou Nader and Steve Kearney

About Jo-Ellen Abou Nader and Steve Kearney

Jo-Ellen Abou Nader, Vice President of Fraud, Waste and Abuse and Supply Chain Optimization, Prime Therapeutics: Jo-Ellen Abou Nader joined Prime Therapeutics in March 2017 and is responsible for network audits and comprehensive fraud, waste and abuse program strategy, as well as oversight of the supply chain. Prior to joining Prime, Jo-Ellen gained over 16 years of experience in the pharmacy benefit management (PBM) audit and fraud industry. During that time, she also worked with specialty pharmacy, clinical product solutions, government affairs, and sales and account management. Steve Kearney, Medical Director, SAS: Steve Kearney helps lead the SAS’ focus on the future of digital health across health care, life sciences and government. An innovator in health outcomes and digital medicine, he works with SAS’ world-renowned team to help solve the most complex health care challenges using advanced analytics, machine learning and AI.

Leave a Reply