How data-crunching is cutting down on massive health-care fraud

Big data is helping investigators unmask fraudsters as they try to hide in plain sight amid hundreds of millions of transactions.

There are no high-tech forensic gadgets or state-of the-art surveillance devices in Mike Cohen’s work station here. The federal investigator’s “office” appears about as exciting as an insurance company cubicle.   

But Dr. Cohen is at the cutting edge of a law enforcement innovation that is helping federal agents level the field in the fight against large-scale health-care fraud. 

“No other reporter has ever seen this,” he says, tapping out a command on his computer keyboard. “But just to give you an idea of the metrics we look at.…”

Line after line of data begins to appear on his computer screen, forming a long list of companies and addresses with columns of related measures and rankings assigned to each business.

“Your standard pharmacy that is just billing Medicare is going to be $300,000 to $1.5 million,” Cohen says. “Maybe $3 million if you have a really intense population.”

Cohen scrolls through the list on his computer screen. Nine pharmacies at the top of his list show Medicare billing of $100 million or more.

“We are not talking about a couple of prescriptions here that are out of sorts,” he says.

Not long ago, it would have taken an entire squad of health-care fraud investigators a decade worth of shoe leather to connect all the dots and compile such a list, Cohen says. Today, he can do it in a few seconds. 

“There is no shortage of ways we can twist and crunch numbers to look for targets,” Cohen says. “And there is no shortage of targets.”

Health-care fraud has become a big, lucrative enterprise in the United States. No one knows the full extent of the drain on Medicare, Medicaid, and private health insurers. Experts suggest it may cost $100 billion each year.

Whatever the actual loss, fraud diverts critically needed resources from patient care, undermining the ability of the government and others to help those most vulnerable. 

For decades, federal agents have struggled to keep pace with growing numbers of health-care fraudsters. Now, the hope is that data analytics can inject a new level of oversight and enforcement into the system.

Depending on commands Cohen types into his computer, the displayed results could be a list of the most suspicious doctors, pharmacies, hospitals, drug companies, medical device manufacturers, or others operating in the US health care industry.

The metrics seek to identify patterns in Medicare billing data that resemble known examples of fraud. Health-care swindlers get rich by finding a way to cheat the system and then using that deceptive practice over and over again. Since every transaction in the government-run health-care system is documented, a successful fraud depends on the ability of the swindlers to hide in plain sight amid hundreds of millions of transactions. 

If those hundreds of millions of transactions can be organized in a way that identifies patterns of fraud, the suspected perpetrators of that fraud are no longer able to hide from federal agents.

“For decades Medicare has been a spigot of money flowing into the hands of fraudsters,” says James Quiggle of the Washington-based Coalition Against Insurance Fraud. 

“The system has finally woken up to the size of the losses and is taking powerful strides to turn off that spigot,” he says. “But it is going to take years before health care turns the corner on these scams.”

$1 billion – and that's just one case

Law enforcement operations provide only a hint of the size of the nation’s health-care fraud problem. 

• In 2007, the Justice Department established a Medicare Fraud Strike Force to battle rampant health-care scams in Miami. In the years since, similar strike forces have been set up in eight other cities. They have charged more than 2,900 individuals accused of submitting $8.9 billion in false bills to Medicare. 
• In June 2016, federal agents conducted a nationwide health-care fraud takedown, charging 301 individuals who allegedly submitted $900 million in false billings. Those arrested included 61 physicians, nurses, or other licensed medical professionals. Officials called it the largest such health-care fraud enforcement operation in US history.
• In July, federal prosecutors in Miami indicted Philip Esformes for running a network of more than 30 assisted living facilities and skilled nursing centers that allegedly billed for unnecessary tests and services for 14,000 poor and elderly patients. Prosecutors estimate the network submitted $1 billion in fraudulent claims, making it the nation’s single largest health-care fraud case – so far.  

There are not nearly enough federal agents to stop all health-care fraud, officials say. But with the help of computer technology, there is hope that the most egregious criminals can be identified and brought to justice.

Cohen is an investigator with the Office of Inspector General (OIG) at the Department of Health and Human Services. He is among 600 OIG agents and staff members who work every day on the front lines in the fight against health-care fraud. 

As an attorney and former physician assistant in emergency and family medicine, Cohen brings a broad perspective to his work from both the health-care side and the law enforcement side. (Editor's note: The story has been updated to correct Cohen's background.)

“They are killing our program financially. This is as big as some of the Wall Street crimes,” Cohen says. But he says his is job isn’t just about locking people up. 

“We have a dual role,” he says. “We have to protect against both patient harm as well as financial harm.” 

Based on the honor system

The high level of fraud in government-run health care programs is partly a function of its founding premise. Both Medicare and Medicaid were intentionally set up under the honor system, officials say, to make it as easy and as fast as possible for health-care providers to receive payment from the government. The problem is that the honor system approach also makes it fast and easy for the dishonest and corrupt to steal from the government. 

“It is a terrible crime,” says Gary Cantrell, deputy inspector general for investigations at HHS, and the top health care fraud investigator in the US government. 

“These are some of the most deplorable individuals doing the worst kind of crime,” he says. “It affects tax payers and in a significant way it compromises the health of patients.”

For many years, too few agents pursued a growing number of swindlers while following a law enforcement strategy called “pay and chase.” In essence, investigators were attacking the problem on the back end, after the fraud had already taken place, rather than addressing it on the front end before the money was paid out. 

“Pay and chase is not the most fruitful model, to pay them and try to recoup the money after the fact,” Mr. Cantrell says. “If you can stop the payment on the front end, you are in a much better situation.” 

Cohen agrees. “We’ve done this back-end stuff for years and it is not working,” he says. “We cannot prosecute our way out of this problem.” 

That’s where Cohen’s computer comes into play. 

Sniffing out pharmacies with questionable billing

To illustrate this capability, Cohen uses data from a 2012 research project conducted by the OIG’s Office of Evaluation and Inspections. 

The report examined billing records of pharmacies participating in the Medicare Part D program, which helps senior citizens pay for their prescription drugs. The program started under President George W. Bush and has become a prime target for fraud.

Of the 59,000 retail pharmacies in the US that billed the government under the Part D program, researchers identified 2,600 that had questionable billing.

Most of the problem pharmacies are independently owned. It is rare for a CVS or a Walgreens to show up on such lists, Cohen says. 

The researchers examined metrics such as the total amount billed to Medicare, the amount billed per beneficiary, the number of prescriptions written per prescriber, and the percentage of prescriptions for various types of drugs.

With his computer keyboard, Cohen zeroes in on pharmacy sales of controlled substances, drugs that include highly addictive prescription painkillers such as oxycodone and fentanyl. These are among the drugs that have fueled the nation’s epidemic in opiate addiction and drug overdose deaths. 

Such controlled substances account for only about 6 percent of pharmacy sales, he says. 

“So when we are looking at a pharmacy here,” he says, “and we see 61 percent [of bills are for] controlled Schedule II drugs – that’s a problem.”

He adds: “Here are pharmacies dealing with almost 75 percent controlled drugs. That’s crazy,” he says. “Sixty percent, 50 percent, these are really high [for billing] in scheduled drugs.”

But that is only part of the problem, he says.

In addition to policing the sale of highly addictive drugs, OIG investigators are also responsible for the sale and distribution of all other prescription drugs. 

While the market in controlled drugs is about $8 billion a year, the national market for other prescription drugs is $129 billion, Cohen says. From where he sits, that’s a lot of potential fraud.

Cohen is careful to qualify his analytic results. They are not evidence of fraud, he insists. Data analytics is collecting and organizing information in a way that should raise a number of red flags. 

“Overall, the stuff I am filtering to the top is probably fraud, but I’m not going to say that 2,600 pharmacies are fraudulent,” he says. “They just have aberrant billing.” 

What that means is that these businesses are good candidates for a federal investigation. An investigation may uncover fraud, or it may reveal a specialty pharmacy that is involved in an expensive – but legitimate – niche business, he says.  

Prioritizing and advancing key investigations

Cohen uses additional data filters to try to identify the worst of the worst. In that way federal agents can be directed to the largest and most urgent cases.

“Years ago, what we would do is poke around, basically. Someone would say, ‘Let’s go look at Joe’s Pharmacy.’ It may or may not pan out. So there was a lot of wheel-spinning,” Cohen says.

“Here,” he says, pointing to his computer screen, “this is like – BOOM.” 

Fast access to computerized billing information can also help agents advance ongoing investigations, says Cantrell. 

When he started work as an investigator in Atlanta more than 20 years ago, it would often take 90 days or more to obtain Medicare billing records related to the subject of an investigation, the deputy inspector general says.

If in the middle of the investigation more data was needed, they would make another formal request followed by another long wait for the information.

“We didn’t have direct access to data, much less direct access to real-time or near-time data,” Cantrell says.

Such quick access to data also helps investigators identify and contact witnesses who might provide evidence against a crooked doctor or other fraud suspect.

Data analytics also allows agents to discover patients who were referred to a particular doctor or health-care service by another physician or service. That information can uncover kickback schemes involving patient brokers. 

It helped make the case against Dr. Jacques Roy, a Texas physician who was convicted in April 2016 of providing false certifications authorizing home health-care services for some 11,000 Medicare beneficiaries, officials say. Prosecutors estimate that $375 million in fraudulent claims were submitted to the government. 

Most of the raw data used by Cohen has already been organized by researchers in the OIG’s Office of Evaluation. Their role is to identify and study vulnerabilities in the health-care system.

“Criminals migrate to this area because the rewards are high and the risks are relatively low,” says Erin Bliss, assistant inspector general in the Office of Evaluation. “We are trying to change that equation.” 

Similar to credit-card safeguards

Some observers have likened data analytics in law enforcement to the safeguards used by credit card companies to protect against fraudulent purchases. When they suspect fraud, credit card companies refuse payment.  

The model is similar, but not an exact match. “We are trying to use big data and analyze it as quickly as possible to detect an anomaly and check it out immediately,” says Ann Maxwell, who is also an assistant inspector general in the Office of Evaluation.

But paying for health care is different than someone trying to use a stolen credit card, Ms. Maxwell says. “If I am paying for something [by credit card], it is being adjudicated immediately and they know right away if this is me and I don’t normally buy two televisions at a time,” she says.

“But your doctor may bill one week later, two weeks later, up to a year later, so you don’t have that instantaneous feedback loop to immediately detect and stop [a fraudulent health-care payment],” she says.

While officials have been successful in using analytics in the Medicare system to identify trends in fraud and to direct task force operations, those same techniques are not yet possible under the Medicaid program.

That’s because Medicaid, which provides health care for low income individuals, is run by each of the 50 states and the states do not yet have a uniform system of data collection. 

Thus someone who is identified as a fraudster in one state may be able to move his illegal operation to another state without the new state learning of his criminal past.

Efforts to close this gap are under way. Currently 35 states have agreed to submit uniform information about their Medicaid transactions to a national database. Federal officials are encouraging the remaining 15 states to comply as well.