Health IT, Hospitals

Xerox delves into machine learning for hospitals

We’ve all heard the promise of big data in healthcare and hospitals, but it’s increasingly […]

We’ve all heard the promise of big data in healthcare and hospitals, but it’s increasingly tilting further into the more complex but potentially essential notion of machine learning. As part of an effort to seize on the opportunity, Midas+, a subsidiary of Connecticut-based Xerox, is combining its vast data resources with five years of Medicare and claims data to help hospitals better predict readmissions.

The Midas+ Readmission Penalty Forecaster, as it’s known, is unique in several respects, said Justin Lanning, senior vice president and managing director of Xerox Healthcare Provider Solutions. One such respect is that it’s patient-focused versus provider-focused.

“We took our machine-learning system and we flipped it on it head,” he said. Traditionally, Xerox and other vendors would design a system that looks at provider benchmarks and other provider-focuses concerns. That’s still possible, but with a stronger focus on the patient, hospitals will be in a better position to curb readmissions or perhaps learn if a patient ends up at another hospital.

Given Xerox’s experience in healthcare and data – ranging from government health, commercial payer groups, pharmaceutical and life sciences – Lanning said it’s in a good position to offer vast resources and knowledge. With some 1,900 hospital clients, combined with the Medicare data, the pool of data is large-scale and advantageous, he said.

Having the large data set is a key component, Lanning said, noting that excellent algorithms can be built, but if what they are analyzing doesn’t contain enough information, then they are limited in what they can achieve.

“From a data use stand point, what we have is the brain knowledge – what we can do is learn lessons,” he said.

And with that, a hospital can get a much more accurate prediction, or “near real-time” information on both patient patterns and reimbursement rates. Lanning said the Forecaster has a 1.5 percent margin of error within the predictive model. Quarterly updates are provided to the hospitals.

The software’s measurements include patients who are readmitted within 30 days to a different hospital.

“This isn’t a crystal ball, but the Forecaster can help hospitals get a good look at the future, and do it with enough time to make a change if needed,” Lanning said.

Midas+ also will provide on-site consultation for hospitals to respond to areas of concern and implement targeted improvement strategies.

“Just delivering analytics isn’t enough,” he said.

And hospitals have precious little time to get a handle on readmission rates.

“Hospitals have just until June 30, 2015 to effectively improve readmission performance for fiscal year 2017 penalties,” said Alycia James, vice president of care performance transformation group for Xerox. “We can tell them today where they’re going to stand.”

Several of the company’s hospitals clients are already receiving their first quarterly reports, she said.

Shares0
Shares0