How Highmark Health uses AI to root out fraud, waste, and abuse

Fellow chief compliance officers and chief audit executives need look no further than Highmark Health as a prime example. A national health and wellness group, Highmark Health uniquely serves the dual role of being both a healthcare provider and a healthcare payor, with a consolidated revenue of $18 billion as of year-end 2019. With 35,000 employees across the United States, Highmark Health’s network of affiliates and subsidiaries collectively provides everything from healthcare to dental care, healthcare insurance, reinsurance, and technology-based solutions for the healthcare space.

Since at least 2012, Highmark Health has realized hundreds of millions of dollars in savings through using highly sophisticated data analytic tools to improve efficiencies and to help detect fraud, waste, and abuse in all its forms. “One of the goals I set out for the team is, how do we start to do things bigger, better, and faster? And how can we put that into the work that we do every day? How do we utilize our staff to the highest potential?” says Melissa Anderson, executive vice president, chief auditor and compliance officer at Highmark Health.

The idea was that by taking some of the more tactical work that staff members were doing and, instead, having algorithm and systems to process data, staff is effectively freed up to think more strategically. “So, it was really about how to do more with less, but yet gain more precision as part of the process. It was a win-win,” Anderson says of Highmark Health’s decision to begin leveraging AI.

“Don’t reinvent the wheel. Learn from experts in the industry. Don’t be afraid to reach out to them. That’s how we’ve learned a lot about the programs that we have put into place.” 

Melissa Anderson, EVP, Chief Auditor and Compliance Officer, Highmark Health

One significant benefit afforded by AI capabilities is being able to detect indicators of fraudulent activity much sooner than in the past—such as spotting trends and unusual activity in claims closer to the time they’re paid, or even before they’re paid, with the goal being to stop would-be criminals before money goes out the door as opposed to after the fact. AI also allows for continual analysis of healthcare claim patterns that may be indicative of red flags, such as high-claim utilization in a given day or provider billings that greatly exceed normal billing patterns generated by comparable providers.

“For us, it’s about how do we find these issues before they become large issues and try to mitigate them as quickly as possible?” says Kurt Spear, vice president of Highmark’s Financial Investigations and Provider Review (FIPR) unit, which is tasked with detecting and investigating all alleged cases of healthcare fraud, waste, and abuse in all lines of its business that impact the organization financially. Fraud referrals can come from both internal and external sources—members, employees, and providers, for example.

Aside from wanting to detect fraudulent activity more quickly, Spear says another thing Highmark Health wanted, and has gained, through using AI software is “reasoning” capabilities—in other words, machine-learning software that takes the data and knowledge of forensic investigators and other human analysts and embeds that knowledge into the AI capabilities, essentially turning them into mathematical algorithms processed by computers. And, unlike people, the memory and processing capabilities afforded by AI is nearly limitless.

From an operational standpoint, Highmark has what it calls its “Payment Integrity” program, under which it has deployed 28 unique initiatives to help ensure claims’ payment accuracy, 15 of which are embedded within the FIPR unit and specific to fraud, waste, and abuse initiatives. Healthcare claims go through rigorous reviews, using a combination of automated AI algorithms and a manual assessment process. “It really helps us to have a targeted audited approach, so that we’re looking at all the right places based on the output that we can get much quicker,” Anderson says.

Both Anderson and Spear stress that AI complements human analysis and is not a replacement for it. “It’s a combination of people, process, and technology that enable us to put a program together that is very effective,” Anderson says.

The FIPR unit, for example, utilizes an internal team that includes registered nurses, investigators, accountants, former law enforcement agents, clinical coders, and programmers, complemented by an array of vendors, to complete its objectives. As part of its work, the team performs audits to identify unusual claims, coding reviews, and investigations that assess the appropriateness of provider payments. “It takes a lot of different individuals and entities across the enterprise, as well as outside the enterprise, to have a solid anti-fraud program,” Spear says.

Savings realized

By using AI, Highmark Health has been able to realize hundreds of millions of dollars in savings—$850 million in the last five years alone, to be exact—associated with the prevention of waste, fraud, and abuse. According to data provided by Highmark, it has realized savings of $120 million in 2015; $148 million in 2016; $183 million in 2017; $145 million in 2018; and $260 million in 2019, which included prevented losses, recovered money, and policy savings.

Anderson and Spear don’t intend for Highmark Health’s efforts to remain in a vacuum. In fact, its information-sharing approach has earned it national accolades. In 2019, the National Health Care Anti-Fraud Association honored Highmark’s FIPR department with the “Special Investigation Resource and Intelligence System Investigation of the Year” award.

The award resulted from an investigation involving a specialty pharmacy that was supplying excessive amounts of hemophilia-factor medications to patients. To drive up its reimbursement, the pharmacy and individuals set up a sham employer group that “employed” the recruited hemophiliacs. Within the first several months, the sham employer group (or pharmacy) submitted claims seeking reimbursement for millions of dollars, with $4.5 million in claims in just the first several weeks.

Ultimately, the scheme was shut down. Spear explains that the award to Highmark Health resulted from information it shared with peers across the United States to help stop similar schemes from happening to them.

Anderson says learning from others in the industry is very important. “Don’t reinvent the wheel,” she says. “Learn from experts in the industry. Don’t be afraid to reach out to them. That’s how we’ve learned a lot about the programs that we have put into place.”

Spear says it’s also important to not be afraid of change. Fraud schemes today are a lot more complex and more organized than they’ve ever been in the past, which really forces the hand of healthcare organizations to be as adaptable as the organized criminals themselves.

The increase and proliferation of fraud activity during the coronavirus pandemic is a timely example and makes using AI to identify fraud, waste, and abuse more important than ever before. “We are enhancing our AI software to be able to hone-in on suspect behavior specific to COVID-19 schemes,” Spear says.

Suspect behaviors may include impossible day scenarios whereby providers bill more telehealth services than could have been rendered in a 24-hour period and large volumes of COVID-19 tests for the same patient, for example. “At this time, it’s still too early to report on results from these efforts,” he says.