The $1.6 billion University of Mississippi Medical Center emphasizes a value-based approach to healthcare. As part of that approach, the organization’s Center for Telehealth uses remote patient monitoring to better manage chronic diseases, including diabetes, hypertension and heart failure – diseases where education and timely interventions can improve individual healthcare and ultimately overall population health.

An important part of remote patient monitoring is identifying patients who are the best candidates for the technology, patients who will respond well to the technology. But how does a healthcare organization best identify these optimal candidates for remote patient monitoring? That’s tricky, said Richard Finley, MD, professor of medicine for infectious disease, professor of emergency medicine, and director of medical analytics for the Center for Telehealth at the University of Mississippi Medical Center.

“We don’t have an answer to that yet, but we are exploring,” Finley said. “Overall, with remote patient monitoring, when you look at the literature, it’s all over the place. Some people say it does not work at all while others have demonstrated solid success. The studies are quite different, and everyone’s ways for doing RPM and identifying patients for RPM are quite different. There are a lot of questions left to be answered.”

[Are EHRs getting better? Readers rank vendors higher than last year in new survey]

Finley’s home state of Mississippi passed a law that requires all payers serving the state to reimburse for remote patient monitoring for chronic disease patients, so Finley is in a better position than many when it comes to use of the technology.

“But given our population, our health problems, and such a big rural and indigent population, we have a lot of difficulties in terms of trying to reduce hospital readmissions and doing so efficiently,” he said. “What I am working on right now is using different analytics tools to conduct pattern recognition to partition patients based on risk factors. So, for example, we know one major risk factor is just the fact that a patient has been hospitalized. That one is easy. And we can go after patients admitted to hospitals.”

But Finley would like the medical center’s remote patient monitoring program and the process of identifying patients to be more aggressive than that, zeroing in on patients before hospitalization.

“We have to recognize factors that we can clearly identify to make reasonable predictions and thus allow us to efficiently do interventions, whether they are simply education or in more extreme cases to use remote monitoring at home, especially for heart failure and diabetes, to try to prevent them from coming into the hospital to begin with,” Finley explained. “We have a fair amount of data we have been collecting. We’ve had an Epic EHR since 2012 and an electronic emergency medicine record since 2000, so we’re trying to leverage all that data, looking back at hospitalizations for patients with heart failure, hypertension, diabetes and COPD to learn how to make predictions.”

The data model the medical center is employing uses a combination of hard data, such as hemoglobin A1C results, and “fuzzy” classification schemes, such as how many times did a patient show up for an appointment in a clinic or the area of the state a patient lives in, which leads to what the medical center knows about that population and its demographics.

“One of the big problems is as the amount of data increases, it overwhelms us by the sheer magnitude of information,” Finley said. “I was trained in theoretical physics so I prefer fundamental analysis. But when you work in the healthcare field we do not have the luxury of even close to the theoretical knowledge of making a fundamental analysis. So really the only game in town is artificial intelligence, whether it’s neural networks or support vector machines or any of the new machine intelligence technologies. They are critical to enabling us to grasp the data.”

There is a danger, though, with Big Data tools like artificial intelligence: Healthcare executives and caregivers can become too ambitious when it comes to how the tools can be applied, Finley said.

“People can get overly ambitious in terms of what they think they can do with the technologies as opposed to the reality,” he explained. “The reality is the U.S. healthcare system is messy in terms of people, psychology, healthcare finance – we in the trenches know the healthcare system is broken, it’s extremely inefficient. And we would like to make some inroads so 25 years from now the healthcare budget isn’t 30% of our gross domestic product, instead of today’s 20 percent. I’d like to see that go down and yet improve our healthcare, but we need more data and more people with good reproducible studies.” 

  Related stories ahead of  Big Data & Analytics Forum in Boston, Oct. 24-25. 

⇒ Charlotte hospitals analyze social determinants of health to cut ER visits
⇒ Big Data: Healthcare must move beyond the hype
⇒ Tips for reading Big Data results correctly
⇒ Small hospital makes minor investment in analytics and reaps big rewards 
⇒ MIT professor's quick primer on two types of machine learning for healthcare
⇒ Must-haves for machine learning to thrive in healthcare

Twitter: @SiwickiHealthIT

Like Healthcare IT News on Facebook and LinkedIn