More on Analytics

Hospitals hope 'Moneyball' analytics approach can help bend the cost curve

By incorporating factors like risk adjustment and utilization rates, organizations can essentially create their own population health box scores.

Jeff Lagasse, Associate Editor

Anyone who's seen the movie "Moneyball" knows the story: Despite lower salaries and less capital than most teams, the Oakland A's baseball franchise utilized box score calculations to amass a winning team without breaking the bank.

According to some health executives, that approach can work in healthcare, too. By incorporating factors like risk adjustment and utilization rates, organizations can essentially create their own population health box scores to improve clinical care and lower costs.

Larry Schor, senior vice president of corporate development and analytics at Medecision, a population health management company focused on risk-bearing entities, first came across the concept during discussions with a friend who worked as a statistician at STATS Inc., the organization responsible for compiling statistics for Major League Baseball.

[Also: MultiCare Health System to save $25 million in Health Catalyst analytics deal]

What the new approach does, said Schor, is allow health systems to monitor their performance in real time, as opposed to "cramming for the test" --  in this case, the star ratings assigned by the Centers for Medicare and Medicaid Services.

"It's pretty transparent, what a box score means," said Schor. "It's focused on performance, not just on looking back and seeing how we performed for a game. It's focused on right now."

Real-time analytics, that's the aim, said Schor. "We don't have that analog in healthcare yet. We don't have a culture of performance."

But outliers like Neil Kudler, vice president and chief medical informatics officer at Baystate Health, a Massachusetts-based nonprofit healthcare system and one of the largest integrated delivery networks in New England, are seeing good results using the Moneyball approach.

"Right now the model itself is something on the order of small data as opposed to big data, but really taking it to another level that is generally considered 'small' or 'flat' data," said Kudler. "Traditionally, medical management uses data that relies on condition-specific registries -- of 2,000 patients, 200 are diabetic, and so forth. I would import metrics there that are used for pay-for-performance and other quality programs, but the problem with them is they're very process-oriented."

[Also: 'Moneyball' namesake Paul DePodesta joins Scripps Translational Science Institute, targets healthcare]

What the new model does is expand those types of data across the time spectrum. Rather than singling out the last posted value, the focus is on values posted over time, which allows staff to analyze the details in a way that helps identify the practices, individual providers and patients who are doing well, poorly or otherwise.

"The other side of the equation here (is) clinical knowledge and intervention," said Kudler, "but realistically speaking, those interventions are going to be based on programming. And it goes without saying that programs cost money. Healthcare is not rolling in cash these days, particularly for uncompensated services. Creating accountable care systems creates multiple uncompensated services."

Integrating more information into a practice, of course, has the potential to suck up more of a physician's time -- which isn't always feasible. Making a new approach to data seamless for any given practice may require an upfront investment.

But Schor said that cost pales in comparison to what many systems incur in order to score the highest possible ratings.

"The country spends $15 billion just cramming to pass the test," said Schor. "There's a tremendous opportunity to make sure they not only pass the tests, but do what they have to do to optimize performance as a system.

"Cramming for the test encourages gaming, which in turn encourages cynicism about what really matters," he said. "The first step is to create trust in the data we're using. Once we establish credibility in the data itself, the physicians are the most active promoters of this new approach, because they want to do well."

Still in its early stages, the new data approach seems to be working for the Baystate Health system. Nine months in, Kudler said he's seen a drop in emergency room visits, readmissions, and overall utilization. He said the approach is critical in bending the cost curve.

As long as the system can make the approach seamless for clinicians he sees it as the way forward, not just for his organization, but for healthcare generally.

"Pay-for-performance is not going away," said Kudler. "But what also isn't going away is the physician's primary responsibility, and hopefully, passion -- which is helping patients."