Hospitals must be on constant alert to identify and prevent fraud. In December, the Holzer Clinic in southeastern Ohio settled False Claims Act charges that some of its physicians routinely overcharged Medicare for patient care. And in April, a Dallas area doctor was convicted in a $375 million scam that recruited thousands of healthy patients and then billed Medicare for home care that they didn’t need.
Increasingly, organized rings are colluding to commit healthcare fraud, requiring hospitals and health systems to adopt increasingly sophisticated tools crack them. A 2005 report by HHS’ Office of the Inspector General on implementation of Medicare’s Hospital Outpatient Prospective Payment System stressed the need for hospitals to pay particular attention to their computerized billing, coding and information systems.
In fiscal year 2015, the U.S. government netted $2.4 billion in healthcare fraud judgments, settlements and administrative impositions in fraud cases and proceedings. More recently the Centers for Medicare & Medicaid Services reported that its predictive analytics technology helped save Medicare more than $1 billion in 2014 and 2015. The agency is working on next-generation predictive analytics that will feature an improved design to enhance usability and efficiency, Healthcare Dive reported.
Meanwhile, FCA penalties for healthcare fraud are set to double, thanks to an interim final rule from the Railroad Retirement Board, according to The Advisory Board Company. While the agency’s rule doesn’t directly apply to healthcare entities, it suggests raise penalties for fraud against agencies with jurisdiction over the healthcare industry, under the inflation adjustment. Civil monetary penalties would jump from between $5,500 and $1,000 per claim to between $10,800 and $21,500.
Whether they involve high-level executives or low-level staff, fraud accusations can tarnish a hospital’s reputation and bring costly lawsuits and fines. So how can healthcare organizations better detect and prevent fraud?
IT solutions
While rules-based and anomaly detection are still important, healthcare organizations increasingly are turning to predictive modeling and relationship analysis to weed out fraud, says Louis Saccoccio, CEO of the National Healthcare Anti-Fraud Association. “You can’t really be effective in going after healthcare fraud without some sort of data analytics.
Predictive analytics uses mathematical algorithms to develop risk scores and other metrics that can predict the likelihood of fraud. But there are other solutions, such as graph databases, that can also help to cut down on fraud.
Neo Technologies’ Neo4j software gathers data and presents as a set of related events. “The idea is that if you know how two data points are related, you can use that information to make a variety of decisions and, in this case, identify fraud,” says Utpal Bhatt, vice president of marketing at the San Mateo, CA-based firm.
“A lot of our customers in this space are finding that they always had the data. What they didn’t have was the ability to see the connections in the data,” says Bhatt. With the aid of Neo4j, they can chart that relationship and gain insights that help them detect fraud rings and prevent it from occurring, he adds.
Graph databases are particularly useful in capturing large fraud rings, such as those targeting Medicare and Medicaid, because of the ability to conduct entity-link analysis, Bhatt says. “That’s the kind of fraud detection technique that the graph database can bring to the healthcare IT organization.”
Neo Technology’s customers include fraud detection firm Pondera Solutions, which is using Neo4j to enhance its products. Its customers include CMS. In addition, Mitre Corp. is using Neo4j to identify patterns in the relationship between doctors and their patients that could point to drug abuse.
While Bhatt is not aware of any hospitals currently using Neo4j, he says there is nothing that would preclude that if they have a strong IT organization. The software solution is available in an open-source version and an enterprise version. “Essentially, if you have an organization that has database skills, like a relational database, and people who understand how to work with data, how to model it and query it, that’s all that’s really needed to use Neo4j and graph database,” he says.
Communicate findings
It’s also important for organizations to share information that they’re gathering from their data analytics, says Saccoccio, noting that this can help to target emerging schemes and trends.
“It’s a constant fight, because what happens is the folks that commit fraud, once you catch them in one place, they’ll move into another area — whether geographic or subject matter — and you’ve got to catch them there,” he says. Having good data analytics, information sharing and targeted resources are key to getting that done.
For hospitals, especially smaller ones with less sophisticated tools and resources, it’s also a matter of understanding what qualifies as fraud, says Lisa Gingerich, an attorney at von Briesen & Roper. With the shift to value-based payment, there’s a lot of pressure on health systems to maximize reimbursement and stretch margins, she says. And smaller hospitals “don’t necessarily appreciated that a mistake uncorrected and not disclosed becomes fraud in 60 days.”
“As more and more of these transactions are going online, and with EMR happening, the fact that more and more information is going to get discharged, the connection between the patients, the providers, the insurance companies, the drugs they’ve been prescribed and the care centers is … is going to be very valuable in making healthcare costs lower by identifying inefficiencies and fraud,” Bhatt says.