Investigators use data mining tools to claw back billions stolen by crooked doctors and clinics
Late on an April evening in 2010, agents with the Federal Bureau of Investigation broke into a medical clinic in Bath Beach, Brooklyn, an area popular with Russian immigrants. Once inside, they installed a hidden camera in an air-conditioning vent in the ceiling directly above a desk.
Over the next six weeks, the camera recorded a blonde-haired woman stuffing envelopes with $100 bills, which she handed to elderly patients. In exchange, patients handed over their Medicare identification number, which Bay Medical used to bill for services that the patients would often never receive. If there were any doubts about the legality of the scheme, Bay Medical kept a Soviet-era poster pinned to the wall showing a woman with her finger pressed against her lips and a simple message in Russian: “Don’t Gossip.”
Bay Medical fleeced Medicare, the US taxpayer-funded healthcare programme for the elderly and disabled, for $50m in illegal payments, US authorities later proved. Its employees used the proceeds of the fraud to splurge on plastic surgery, luxury cars and vacations.
Sixteen people were arrested, including two doctors and Irina Shelikhova, the mastermind. All but three were convicted or pleaded guilty to charges related to healthcare fraud. Ms Shelikhova, a Ukrainian immigrant, was sentenced to 15 years in prison and ordered to pay $86m in losses and fines.
The Bay Medical case may sound like an example of old-school detective work, and in many ways it was. But before agents could plant their hidden camera, US investigators harnessed the power of “big data” to find the scheme, analysing vast amounts of Medicare billing information for aberrations.
Such data mining has long been used in ecommerce, from music companies trying to anticipate the song you want to listen to next or tech marketers scanning your social media profile to make a pitch. Now it has become a federal fraud-busting tool. Government data analysts are monitoring eight other “hotspot” cities to track potential Medicare fraud, including Detroit, where, in one case, analytics helped confirm a tip that a doctor was prescribing expensive chemotherapy treatments for patients who did not have cancer.
Fraudulent billings are estimated to make up as much as 10 per cent of Medicare spending in the US, according to the most recent study. Medicare, which funds hospital visits, prescription drugs and other services for retirees and the disabled, paid over $600bn to provide medical services to 51m Americans in 2013, according to the US Government Accountability Office.
Healthcare takes the lead
Though still in start-up mode, US investigators’ embrace of big data analysis could save taxpayers billions of dollars a year from healthcare fraud alone. Ultimately, the US Department of Justice hopes to adapt these tactics to crack down on other forms of fraud.
The DoJ says the scheme is already paying off. For every dollar spent to combat healthcare fraud, the US government has collected $8 in recoveries from forfeiture, asset seizures and fines, amounting to $4.3bn in 2013 and a total of $19.2bn over five years.
The propaganda poster advising Russians not to gossip that adorned a clinic where a $50m fraud took place
Data mining was launched as an experiment in 2007 in Miami, a hotbed of healthcare fraud. The DoJ, the FBI and the Department of Health and Human Services’ office of inspector-general wanted to see if data could be as good as a human at rooting out fraudsters. “You might see one medical equipment provider getting referrals for 500 wheelchairs from the same doctor, or a home health agency that’s getting 60 per cent of its referrals from the same doctor,” says Leslie Caldwell, chief of the DoJ’s criminal division. “A lot of this can be gleaned from the data.”
Ms Caldwell said the ability to scour billing data also gives investigators an advantage by allowing them to watch crimes unfold. “The idea of using real-time data to generate fraud cases is unique,” she said. “We have the ability to suspend — [when] there’s reasonable suspicion — [those] who are engaged in fraud even before they are prosecuted and indicted.”
Using big data as a fraud-fighting tactic is the brainchild of Kirk Ogrosky, a 36-year-old prosecutor from Harrodsburg in Kentucky. He worked for five years as a federal prosecutor specialising in healthcare fraud in Miami before a stint at a private firm in Washington.
But in 2006, the DoJ asked him to return to crime-fighting, as the head of the healthcare unit. Mr Ogrosky said he would only do it if they agreed to completely rethink how they prosecute healthcare fraud, with an emphasis on real-time prosecutions. Mr Ogrosky thought the DoJ should take a proactive approach. “I started looking at claims data and tried to identify the worst offenders,” he said. “The questions were how and where is the government taking the biggest beating.”
Culling billing data by postcode, he looked for communities where patient spending was three or four times the national median. From there he narrowed the search for providers in those communities that were the top billers. “Most times, those zip codes would help generate a list of providers that had what I would call ‘medically impossible’ claims. [It was] like peeling an onion ring by ring — and yes, it always burnt my eyes at some point.”
Often it turned up a false lead, such as a major hospital, but just as frequently it revealed the name of a small clinic. Mr Ogrosky would then do some old-fashioned sleuthing: driving past the clinic to see if it existed, checking whether the doctors and patients were alive and requesting bank records to see where and how the money was flowing from federal agencies. The results were striking.
In Miami-Dade county, billings for durable medical equipment — wheelchairs, iron lungs and the like — went from $73m in late 2006, before the new data-led programme started, to $40m by the first quarter of 2007. Billings fell to just above $10m by early 2013, according to the office of inspector-general, the federal watchdog. The drop was linked to the rise in prosecutions.
Investigators scouring billing data in 2009 discovered some serious anomalies from a clinic in Brooklyn. It was billing physical therapy sessions three times a week, with nearly all patients receiving special extensions to continue the care. This was their introduction to Bay Medical. “They were billing $20m a year. Everything popped off the charts,” said Thomas O’Donnell, a health agency special agent overseeing the case.
Breaking the silence
But data would only get them so far. They also needed witnesses, but Bay Medical served a very tight-knit Russian-speaking community. “The Russian population as a whole usually doesn’t talk,” Mr O’Donnell said. “We couldn’t do traditional methods. You have to have undercovers.”
A New York state investigator originally from Russia worked with the “strike force” to recruit two elderly residents in November 2009. They agreed to wear a wire and carry cameras embedded in satchels. The next month, one of them, working undercover, recorded Bay Medical employee Elena Girenko saying patients could receive $50 to $150 for agreeing to receive certain treatments, according to court filings. Using their cameras, the patients recorded their own payouts, every other Friday, from the “kickback” room.
The evidence was used to secure a court-authorised wiretap that allowed the installation of the secret camera. Nine months after identifying the clinic, agents made their initial arrests early one morning in May 2010. (Ms Girenko later pleaded guilty.)
Lewis Morris, a former chief counsel at the federal watchdog, called the data-based model “highly effective” at catching fraud cases, but says “the real smart fraud schemes don’t steal a lot. They steal a little bit a lot of the time.”
In Detroit, agents say, the data led them to focus on home healthcare agencies, which provide service to elderly or disabled patients at their residence. From 1984 until 2002, the number of home healthcare agencies registered in Michigan remained relatively static, dropping slightly from 187 to 185. Over the next 10 years, the number of registered agencies exploded to 652, a 250 per cent increase. The Medicare population in the state grew by 18 per cent over the same period — raising suspicions among investigators, particularly the FBI.
“We can use [data] to lead us in the right direction, investigatively, and we can also use it to confirm and corroborate leads that we have from more traditional sources,” said Paul Abbate, the special agent in charge of the FBI’s Detroit office. “It’s made us more efficient overall in terms of what we do.”
In August 2013, investigators received a tip about Dr Farid Fata, an oncologist, who was allegedly ordering unnecessary medical treatments for patients. Prosecutors and agents analysed his billings — and saw a crime scene in the making.
Agents interviewed current and former employees. A doctor at one of his practices said Dr Fata had ordered 56 doses of a treatment for non-Hodgkin lymphoma in two years. The normal dosage, he said, was 12 doses over the same period, court documents say.
A nurse told agents that Dr Fata ordered chemotherapy for patients whose cancer had already gone into remission. Within a week of that tip, Mr Fata was arrested. A year later, facing decades in prison, he pleaded guilty to committing healthcare fraud. Federal authorities says Dr Fata falsely submitted $225m in Medicare claims between August 2007 and July 2013.
David Korff, a retired financial analyst for ITT Corp, said the doctor did more than deceive the government.
Mr Korff went to see Dr Fata after he was diagnosed in 2006 with myelodysplasia, a blood condition that can be a prelude to leukaemia if not treated. Dr Fata started him on chemotherapy for three days a week and then more. “He had me coming in seven days in a row getting an IV [intravenous drip] that would last for about one to two hours,” Mr Korff said. At consultations, Dr Fata told him: “If you stop chemo you’re dead.” The treatment continued for seven years.
“It was like a factory,” Mr Korff recalled. “They had about 30 chairs and it was people sitting in a waiting room for half an hour. As soon as somebody got up the next guy came in.”
Mr Korff believes excessive chemotherapy impaired his body’s ability to manage an unrelated artery disease. When he got a bruise on his leg, he said, it turned into a sore that would not heal. While he was seeing Dr Fata, Mr Korff had one leg amputated and then the other. “At the time we thought this was a terrible case of artery disease,” Tracy, Mr Korff’s wife recalled, until a different “doctor started to say that they had never seen that before”.
Mr Korff’s last appointment was on the same day agents arrested Mr Fata. He has since switched to another doctor, who has taken him off all drugs. Mr Korff has filed a civil malpractice lawsuit against Mr Fata but he also faults the government. The data should have told them something, he said.
“At the rate the man was billing, someone should have known,” he said.