Getting to high-performance healthcare

BOSTON—Despite “an endless well of metrics,” healthcare is not operating at a high level of performance, said Jason Burke, senior advisor for innovation and analytics, UNC Health Care and School of Medicine, speaking at Medical Informatics World.

There are more than 9,000 measures in healthcare but no one knows which ones actually matter, he said. And, it’s impossible to expect high performance on all of them.

“There is growing evidence that our current conceptualization of quality doesn’t actually measure quality or performance. Good things and important things are two different things. We should not get confused that quality is some kind of analytically driven, evidence-based way of driving better performance.”

It’s not reasonable in today’s healthcare environment to expect performance to be driven solely by the performance of an individual practitioner, he said. “There’s too much complexity to expect any single practitioner to do that unarmed.”

UNC’s innovation center recently did a project to look at hospital readmissions and how to increase the ability to predict them. They brought in more than 30 different statistical models to see if they could get better fidelity of readmission risk using more sophisticated forms. They also studied whether the addition of consumer data made a difference.

“Predictive analytics absolutely outperformed traditional methods,” they found, for a difference of more than $10 million a year, he said. “It makes a lot of sense to start looking at computers as more sophisticated helpers.” Adding consumer data to EHR data results in about 1,500 data points for each patient and the potential for thousands of permutations and patterns that could be actionable. “There’s no way a person could pull that off.”

UNC uses a detailed framework for using analytics to help manage population health which involves a series of questions and perspectives to consider. Do some risk stratification of the patient population impacted around the program, Burke advised. The framework includes an assessment phase and monitoring phase which factor in utilization, health outcomes, performance managed and more. This is much different approach from those organizations that are “just putting poker chips down.”

Healthcare has all kinds of data available, Burke said. The information is not collected under controlled clinical conditions but “it is an opportunity to understand what is actually happening in the real world.” In the past, healthcare was restricted to long timeframes of constrained data collection but that’s no longer the case in the era of EHRs. “We have an opportunity now to use that information not just for care delivery but also for decision support.”

There’s also an opportunity to create up front metrics known to be applicable across the enterprise. “Use of agile methods could help us do that, help us create a more consistent, stable and reusable set of data measures useful across our system. This is a natural byproduct of the rate of growth we’re going through and needs to be managed at the institutional level very carefully.”