Frontline clinicians appreciate the need for data-driven insights, but they're also overwhelmed with the competing mandates of quality improvement and cost-reduction, says Jeanne M. Huddleston, MD, associate professor of medicine at Mayo Clinic. Making analytics work for them – but most especially for patients – requires a broad-based approach.

"We need your great work, but we're tired and confused," said Huddleston, speaking Wednesday at the HIMSS Big Data and Healthcare Analytics Forum in San Francisco.

With a research focus on applying industrial and systems engineering principles to care delivery, Huddleston is acutely aware of the benefits data analytics can bring to the clinical setting, but she also knows how important it is to implement new strategies effectively. She launched the first Clinical Engineering Learning Laboratory within the Mayo Clinic's department of emergency medicine. She also pioneered, alongside clinicians and engineers, the implementation of predictive analytics and machine learning tools for bedside clinical decision support.

For all its medical and technical advancements, the U.S. healthcare system still too often fails its patients, she said. Creating better, safer care depends - not just on smart data-crunching - but effective deployment of those insights.

To truly drive quality and safety improvements, a multi-pronged approach - combining analytics, CDS, stakeholder engagement, design and human factors - has to be brought to bear.

"For every single problem you must have a multidisciplinary, interdisciplinary team," said Huddleston. "The low-hanging fruit in healthcare is gone. Now every problem is going to be hard."

And no question, effectively applying these strategies is very hard.

"From an analytics perspective, the easy things to find in the data are the things that happen, the adverse events and the things that we harm," she said. "The very difficult things to find in data are the things that aren't there."

Huddleston walked through some incidents of Mayo system failures that resulted in serious adverse events and mortality. They were stark reminders of the complexity of care delivery and the sheer number of decisions and actions that can impact patient safety.

But patients and their families "don't care that doctor didn't get the alert or the IT system went down," she said. "Only that the system didn't work."

No one, she said, "should ever suffer or die because of process or system failures."

Where was the system helping the overburdened, exhausted providers? It was recording everything for them just perfectly. But where was the help?

We decided we needed to take a different approach. And so I took my systems architecture book from engineering school and translated its principles to healthcare.

Her approach to quality improvement was to emphasize the unique attributes of a cross-section of health system staff.

"We didn't make everyone sit in a room every week for three years," said Huddleston. "We broke up into teams with different skill-sets, so everyone was just moving theirs as efficiently as possible.

Based on input from those groups, Mayo identified 33 "failure modes" in the space of just five minutes. Among the Top 5: clinical condition not reassessed at the bedside; too many complex things to do in too short a time; physician doesn't review the nursing notes; the care team has the wrong diagnosis; and a clear definition of deterioration doesn't exist.

"We knew that if we were going to solve the problem, we had to hit all Top 10 failure modes and solve each and every one of them," she said. "Or else what we implemented, no matter how great the data, wasn't going to work."

Taking that approach has led to some substantial gains. Mayo tracked nurse input, for instance, assigning a numerical score on their assessment/concern for every patient. Charting this "nursing factor" helped the nurses "feel empowered," said Huddleston.

"We had to get there: Our data team had to understand that they needed to be able to do something to fix the nurses' problems," she said. That "worry factor" is what nurses "have been doing for 150 years," she added. "Analytics was a black box. Bringing them together brought some humanity to our score," thus increasing viability and "implementability" of analytics tools across the health system.

When it comes to systems architecture, "you've to got to build what's going to work," she said. "If it doesn't work for the nurses, it won't be implemented. If it doesn't work for the docs, it won't be implemented. If it hurts patients, it won't be implemented. If it costs too much, it won't be implemented."

Huddleston added a note of caution - but also one of hope.

"There will be times, as data scientists and engineers, when you just want to roll over and give up, because the clinicians in the place are just too difficult to deal with," she said. "But it is a multifaceted problem: Data science isn't good enough, just being a doc isn't good enough, just being a nurse, a social worker or an engineer isn't good enough."

That all-hands-on-deck approach is key to making analytics work for patients and clinicians alike, she said. Those many voices need to help determine what the systems architecture needs to look like, so you the insights derived from data can actually lead to change.

"I encourage you to put data to work," said Huddleston. "For your patients, for your loved ones, for the providers in your facilities. We absolutely have to put it to work, which means working together, across disciplines, to make meaningful change as soon as possible."

This article is part of our reporting on the Big Data and Healthcare Analytics Forum, taking place this week in San Francisco. Other stories include: Atul Butte says 'precision medicine makes doctors nervous and Best practices for healthcare data visualization.

Twitter: @MikeMiliardHITN
Email the writer: mike.miliard@himssmedia.com
Like Healthcare IT News on Facebook and LinkedIn