Advanced Analytics Is Changing How Agencies Fight Fraud

How the DOJ Is Fighting Healthcare Fraud

At the DOJ’s Health Care Fraud Unit, eight data analysts assist prosecutors with identifying, investigating and prosecuting healthcare fraud cases.

The analytics team uses multiple databases and tools to conduct its analysis, including the Microsoft Power BI data visualization tool, i2 Analyst’s Notebook and Microsoft Excel. The team also uses multiple internal and external models to detect healthcare fraud, and it maintains files related to its analysis on in-house department servers.

The Centers for Medicare & Medicaid Services, part of the Department of Health and Human Services, provides direct access to its portal so the Health Care Fraud Unit’s data analysts can look at its information. The analytics team also collaborates with other government data teams, including those in HHS and the Drug Enforcement Administration.

“Our team builds models analyzing hundreds of different variables that can indicate fraud,” Foster says. “But data is not the truth. While it can point us in the right direction, we have to go out and investigate.”

The DOJ uses two types of models. In the first approach, it finds suspected fraudulent providers by examining the characteristics of medical professionals and others who were prosecuted for healthcare fraud in the past. Through analytics, they find current providers who share those characteristics, Foster says.

LEARN MORE: Understanding the advantages and disadvantages of advanced analytics.

It then seeks out national outliers and ranks providers with a scoring system. For example, the DOJ investigates physicians who order more cancer genetic tests than 99.9 percent of doctors in the country.

The second modeling approach is to analyze billing and other healthcare data to find trends in fraud. For example, in 2019, the DOJ saw a spike in Medicare spending on durable medical equipment.

In its investigation, the DOJ found that some equipment makers were paying illegal kickbacks and bribes to medical professionals and telemedicine companies that were ordering medically unnecessary braces for patients.

The DOJ charged 24 people who it alleges were responsible for $1.2 billion in fraud. The agency’s work resulted in $1.9 billion in Medicare cost avoidance, Foster says.

“That’s an example of how our investment in data analytics returned outsized results,” he says.

Using Pandemic Era Funding to Prevent Fraud

The Pandemic Response Accountability Committee (PRAC), which was established under the CARES Act, has built a data analytics platform to help government agencies identify fraud, waste and abuse of pandemic relief funds.

“It’s what we call ‘pay and chase.’ The money has gone out the door. Now, we are trying to ensure that those who defrauded the government or used the money inappropriately are admonished,” says PRAC Executive Director Jenny Rone, whose organization oversees $5 trillion in pandemic-related spending and includes 20 agency inspectors general among its committee members.

In August 2021, PRAC launched the Pandemic Analytics Center of Excellence, a secure, cloud-based data analytics platform in Microsoft Azure.

PACE pulls siloed data, such as information on loan and grant recipients, and centralizes it in an Azure data lake, Rone says. Through sophisticated data analytics tools — including the use of artificial intelligence algorithms, natural language processing, risk modeling and other capabilities — agencies can perform the analytics and uncover potential instances of fraud or waste, she says.

Once suspicious activity is identified, the PRAC Fraud Task Force collaborates with agency inspectors general and law enforcement to investigate further, she says.

PRAC’s team of data scientists used PACE to identify about 69,000 questionable Social Security numbers that were used to obtain $5.4 billion in Small Business Administration loans, Rone says.

And an investigation by the PRAC Fraud Task Force led to the conviction of a New York man last November for using a false identity to secure $1 million in SBA pandemic loans.

What Is the Treasury’s ‘Do Not Pay’ Service?

At the Treasury Department, the Bureau of the Fiscal Service offers a Do Not Pay (DNP) service, a web-based tool that enables agencies to check multiple data sources to ensure that people, businesses and organizations are eligible for federal payments, including loans, grants and other benefits.

The bureau makes payments on behalf of about 90 percent of federal agencies. In fact, during the 2022 fiscal year, the bureau made 1.4 billion payments valued at $5.3 trillion, says Kevin McDaniels, senior advisor for payment integrity.

The bureau’s Office of Payment Integrity, which operates the DNP service, allows agencies to search a single person or entity or search in batches. It also has a team of 20 data scientists who provide agencies with advanced data analytics services to prevent fraud, says Marshall Henry, director of the DNP Business Center.

“We want to engage agencies early, before a payment is issued, by leveraging the data sources we have so that funds never leave the government that should not be leaving,” he says.

EXPLORE: How federal agencies can future-proof their investment in serverless architecture.

The DNP application and data are hosted in a private cloud within a Federal Reserve Bank data center, Henry says. Today, the bureau pulls the data from other agencies and unifies it in a central repository in the private cloud.

Increasingly, however, the organization is using application programming interfaces to remotely access data, such as death certificates, tax transcripts and databases that show whether people are incarcerated, owe federal debt or are delinquent on child support.

The bureau’s data scientists use cloud-native analytics tools and, in some cases, machine learning algorithms to help agencies identify potential fraud, develop predictive models and spot trends. Analytics prevents citizens from illegal double-dipping, McDaniels says.

“If a person receives a disability payment from one agency and that person applies for disability payments from another agency, the duplication may not be appropriate,” he says.