A researcher at the US Department of Energy’s Oak Ridge National Laboratory (ORNL) thinks he can save $50 billion per year in fraud, waste and abuse in the nation’s public health care system. And that’s just for starters.

Andrew Loebl, a researcher with ORNL’s Computational Sciences and Engineering Division, is proposing to use the lab’s “Jaguar” supercomputer to analyze the country’s public health data in order to reduce wasteful or criminal billings, while at the same time increasing the quality of care. According to Loebl, using the 2.6 petaflop machine — currently the world’s most powerful computer — would reduce the time needed to process the entire public health claims database from months to minutes. That would allow the government to catch criminals in that act of defrauding the system, while also making real-time decisions about quality of care possible.

This would represent a completely new paradigm for the public health care claim processing. The current system uses what is referred to as a “pay and chase” strategy, where submitted payments are automatically paid as long as the data on the form is filled out correctly. It is only after the fact — usually months after — that those payments may be challenged. Not only is the model inefficient, it also encourages criminal activity by essentially paying all claims with no questions asked up front.

According to Loebl, the sheer size of the public health sector has forced the government into this inefficient model. Public health care encompasses Medicare, Medicaid, and 14 other programs, which together generate an average of 4,900 claims per hour, 24 hours a day, 365 days per year. Up until now, no single computer system was able to analyze this level of transactional intensity. The closest analogy is the credit card processing system. But in that case, the dataset is only around 50 characters per record. Health claim transaction records are 10-20 times that size and more complex in form. “It’s not possible with prior technology for anybody to do any meaningful analysis of the data,” says Loebl.

Today, all claim records pass through the computers of three processing subcontractors in the US. Loebl says these computers don’t do any waste, fraud and abuse analysis because they are incapable of doing so. Claim fraud is carried out manually, mostly by the FBI and the two largest public health care agencies, Medicare and Medicaid. The most ambitious program is managed by the Center for Medicare Services (CMS), which is the organization tasked with managing the recovery of Medicare fraud and waste. CMS pays five contractors, at $100 million per contractor per a year, to post-process the claim records to look for irregularities like double payments. The contractors also collect a 10 percent fee on the amount they recover.

But this system only collects about $1 billion per year, which represents just a fraction of the $150-450 billion that it is estimated to be lost each year due to waste, fraud and abuse for the entire public system. Worse yet, CMS spends $600 million to recover that $1 billion. The current health care bill winding it way through Congress has a provision to increase the recovery 1.6 billion.

With Jaguar, Loebl thinks he can increase the recovery figure to $50 billion, and perhaps even more. Using the machine’s massively parallel throughput processing capability, medical claims could be analyzed in real time, thanks to the supercomputer’s 250,000-plus processing cores and 362 terabytes of memory. The advantage of real-time processing means invalid payments that are caught are never sent out.

By aggregating all public health care transactions into a unified database, relatively straightforward software can be developed to detect the type of anomalies that represent criminal fraud and billing errors. Irregularities such as duplicate claims for same diagnosis or procedure, claims for people who are already dead, and diagnoses that are impossible for the patient (e.g., a hysterectomy for a male patient), are easily detectable once all the data is in one place. “It’s not rocket science,” says Loebl.

As the software matures, Loebl thinks they will be able to increase the recovery rate as well as anticipate new types of fraudulent behavior. Beyond that, the program can be used to improve health care quality by correlating patient outcomes with specific medical protocols. For example, if physicians in New England are executing a successful protocol for ulcerative colitis while another group of physicians in California using a different protocol are having less success, the system could find that difference and report it to the entire medical community.

All of this can be attained at a fraction of the cost of the current system. Loebl estimates it would cost about $6 million in the first year to organize the data so that it could be fed into the Jaguar machine at Oak Ridge. Beyond that, he expects the program to cost between $50 to $75 million per year to keep the system operational. He estimates a return on investment of 2,500 percent.

But at this point, Loebl is hoping for just a few hundred thousand dollars to develop an explanation of the concept and scope out the program. So far he has been unsuccessful in attracting any support outside of ORNL’s contribution of Jaguar time and resources. The basic problem, according to Loebl, is that the data is siloed in the multiple agencies running the various public health programs, and no one is very interested in giving up their data for the greater good. Loebl has sought out a few Congressmen — both Republicans and Democrats — but according to him, there’s always disbelief that this is possible to accomplish with a single machine, and some suspicion that the program is just another way to fund supercomputers.

For his part, Loebl seems to be motivated by a desire to help solve the public health care funding crisis, which threatens to swallow federal and state government budgets over the next several decades. With the US Administration and Congress on the verge of enacting legislation that broadens health care coverage, while keeping any new spending deficit-neutral, Loebl believes his proposal can help realize that ambition. “That’s my goal,” he says, “to make good on the President’s promise to fund health reform with no additional costs to my children or grandchildren.”