AI for Healthcare

MGMA: Predictive analytics is now a business necessity for healthcare


In 2011, the Centers for Medicare and Medicaid Services introduced the Fraud Prevention System, a series of predictive algorithms designed to ferret out improper Part B payments that in its first three years prevented an estimated $820 million from being paid to physicians. That success heal a vital lesson for providers: Probe audits are out and advanced statistics are in.

That was the message Monday from Frank Cohen, director of analytics at Clearwater, Florida-based Doctors Management, who spoke at the Medical Group Management Association’s annual conference in Anaheim, California.

[Register Now: Upcoming HIMSS Big Data & Health Analytics Forum]

“We’re using predictive modeling to go away from the clinical side and to use it in an advantageous way on the finance side in healthcare,” said Cohen. “What we want to be able to do is predict which patient is most likely to not show up, because we want to be able to get ahead of that. We want to predict who’s going to be a whistleblower, who’s going to quit, the likelihood of someone suing. We want to go beyond the idea of clinical research and clinical work, to be able to predict beyond that.”

Cohen stressed that it’s important to know the difference between estimation and prediction when it comes to analytics. With estimation, data is used to guess at a parameter of some already-known variable. Prediction uses the data to suss out a random value that’s not part of the known data set — such as identifying which patients are likely to be readmitted to the hospital within 30 days.

Such predictions are especially important when it comes to readmissions, as CMS levies financial penalties on hospitals that perform poorly on that metric.

One of the important facets of predictive analytics, and one that has broad applications on the business side of healthcare, is a regression analysis — a predictive modeling technique that can be used to, for example, forecast how long it will take for a new physician to break even financially, or how many (and which) patients will be no-shows. Such predictive power enables physicians and hospitals the ability to make decisions that are in their best financial interests, said Cohen.

Big data analytics are perhaps the most powerful form of predictive modeling, said Cohen, because huge data sets can be analyzed to see patterns or trends, and use it to predict variable behavior in the future.

“It’s difficult to catch up,” said Cohen. “Most of us are still stuck in the world of Microsoft Excel,” which has limitations on the amount of data it can process. The right technology and infrastructure can make all the difference.

Cohen said predictive modeling is all about probabilities and uncertainty, and doesn’t necessarily work for every issue or problem. But for many things, it’s a must, and just might mean the difference between savings and waste. 

And how accurate do the models need to be?

“Better than chance,” said Cohen. “That’s it. It just needs to be better than the flip of a coin. Those predictive models work.”

Twitter: @JELagasse
Email the writer: [email protected]



Source link