Healthcare Analytics, Population Health Management, Healthcare Big Data


Epic Systems: Machine Learning Is the EHR Usability Solution

Can machine learning, interoperability, and the Judy Faulkner touch bring joyful usability to electronic health records?

Source: Epic Systems

A public appearance by Epic Systems founder and CEO Judy Faulkner is always a big event for the health IT industry, and nothing is bigger for Epic than the electronic health record company’s annual user group meeting.

With 9000 thousand employees and around 8000 users from the US and elsewhere congregating in Verona, Wisconsin, to hear about what’s on Epic’s development docket for the coming year, Faulkner seemed committed to challenging the very definition of what makes an EHR – and an EHR developer – tick.

This year, in a magician-themed extravaganza, Epic executives extolled the value of big data analytics for revolutionizing EHR usability and ramping up physician satisfaction as they teased a number of new developments in machine learning, voice recognition, and natural language processing

But the focus on EHR usability isn’t just to keep Epic’s massive customer base happy. Innovations in EHR form and function are crucial for enabling a new era of patient engagement, population health management, and quality measurement.

“Making software a joy to use is a high bar,” said Faulkner to a packed auditorium, costumed in a red and gold gown with a wizard’s cap perched on her head.

“For the past couple of years, especially after meaningful use, we've been focused on making our software easier to use with fewer clicks for you to do everything you need to do. We want an interface that draws you to it in addition to giving you the functionality you need to deliver the best care.” 

Source: Xtelligent Media

Intuitive data visualizations, sleek drag-and-drop interfaces, and more natural input methodologies such as voice recognition are poised to change the way physicians, nurses, and patients interact with their records, she said.

“The ‘E’ in ‘EHR’ is so last year,” said Faulkner. “‘Electronic’ has to go away as a distinguisher now, because everything is electronic these days. Now we need to go beyond the traditional walls and knock them down.”

J-O-Y from an E-H-R?

“Joy” has become a bit of a touchy word for clinicians in recent years, especially for those who do not believe the American Medical Association’s campaign to restore joy to medicine has thus far borne sufficient fruit.

Physician burnout rates are still high and EHR satisfaction numbers remain grim, Taylor Davis, VP of Business Development at KLAS, acknowledged when he took the stage.

More than a third of all EHR users are dissatisfied or very dissatisfied with their health IT systems, he said, citing a KLAS survey that included responses from more than 12,000 users at 40 organizations.

Source: Xtelligent Media

While the poll found somewhat higher levels of satisfaction among Epic customers, segmented out for the purposes of the presentation, the number of unhappy EHR users is still worryingly high.

“You're all adding up that 24 percent and 5 percent in your heads. Some of you have gotten very long emails from those folks. Some of you are those folks,” Davis said, drawing a ripple of laughter and plenty of nods from the crowd.

“Are these people just whiners? Or are they among the most dedicated providers who simply aren't going to stop pushing until they see EHRs achieve their full potential?  I believe it's the latter,” he said.

“Making software a joy to use is a high bar.”

Reaching that potential is, of course, the health IT development world’s biggest challenge, and one that not even Epic has overcome just yet, said Dr. Christopher Longhurst, Chief Information Officer at UC San Diego Health.

“It’s great to see that Epic is top in satisfaction.  But to be honest, that’s sort of like being ‘cream of the crap,’” he said to journalists in a candid post-show wrap-up.  “One in three doctors is still dissatisfied.”

Part of the problem stems from health IT’s well-known chronic inability to catch up with other data-driven industries, Longhurst said.

“Having spent a long time in Silicon Valley, I know that the IT experts there scoff at healthcare for being ten years behind the rest of the world,” he said.  “We are just starting to apply technologies to healthcare delivery that have been common in the consumer environment for five to ten years.”

The slow pace of health IT adoption is frustrating, but “we have done much of this to ourselves,” Longhurst asserted. 

Multitudinous regulatory requirements and outdated E/M coding for billing purposes contribute to overgrown clinical notes, he said.

Epic President Carl Dvorak also pointed out that clinical notes in the United States are an average of three times longer than those in other countries, due in large part to information dumps that attempt to cover all the bases for billing and coding.

While Dvorak readily agreed that clinicians need to spend less time typing and more time focusing on patients, he did take issue with media coverage of a recent AMA study that found physicians spend around six hours per day on EHR data entry, including about an hour and a half of “pajama time” after the traditional workday.

Quality documentation that accurately captures the patient story will always take time, he asserted, and not even the best health IT system in the world can – or should – completely change that. 

“Orders didn't write themselves before the EHR, and they're not going to write themselves after the EHR,” he said.

The key for developers is figuring out how to make those necessary hours interacting with the system a pleasure, not a burden.

More tailored user training with a focus on peer-to-peer specialty training can help, Faulkner suggested.

“We have moved from focusing on physicians to focusing more on project teams, but we need to shift back to focusing on clinician training,” she said. 

“Specialists know the challenges and opportunities of their specialties. It's hard for them to figure out how to address those with generic training, and it takes too long for them to do that. Specialists can train similar specialists in three to four hours, which saves you money and produces a higher quality learning experience.”

Epic is also embracing a number of novel technologies, including virtual assistants such as Amazon’s Alexa and Google Home, to make the documentation process easier and more enjoyable. 

“You'll be able to do a lot more with voice,” Faulkner said, promising a set of Alexa and Google Home skills, including documentation assistance, placing orders, and signing off on notes, by the end 2018. 

“Orders didn't write themselves before the EHR, and they're not going to write themselves after the EHR.”

Ambient computing devices will also play a role in patient engagement, allowing consumers to talk their way through their medication adherence issues, fitness goals, and wellness plans.

When still face-to-face with the system, clinical users will have access to more intuitive visualizations as well as FHIR-based enhancements available from the App Orchard, which is now officially live.

Turning Healthcare Big Data into Actionable Clinical Intelligence

What is EHR Optimization, How Does It Start?

Turning the EHR into a “central nervous system” for patient care

Meaningful use made EHR adoption a requirement, but value-based care and risk-based reimbursements have made health IT use a fundamental necessity.

Adjusting to a longitudinal, collaborative, person-centered care environment that demands extra patient outreach and follow-up hasn’t been easy for most organizations, Faulkner said, especially those whose providers are among the 30 percent who can’t stand their EHRs.

“Change is hard,” she said.  “But obsolescence is worse.  It’s time to create a central nervous system so that when you do things that fall outside of the traditional care processes, you can get paid for it.”

Providers have a responsibility to retool their workflows and develop coordinated care teams to support population health and individual wellness. But the onus is on EHR developers to make sure organizations have the tools to enable practice transformation – and that they are aware of everything that is available to them.

Epic hasn’t done a good enough job of making sure its providers understand what they have and how to use it, Faulkner admitted.

“We're delivering so much technology, but too many people don't know it,” she said. “We don't think there's good enough advertising.”

“A few years ago, I asked people at the user group meeting how many of them wanted to be able to access [Epic’s patient portal] MyChart in Spanish,” she recalled.  “Every hand went up. And then I asked how many people knew they’ve had that capability for the past four years. Only half a dozen people raised their hands.”

“Change is hard.  But obsolescence is worse.”

Endless reams of release notes and explanations wrapped in impenetrable tech-speak have left everyday users unaware of how much their EHRs have to offer, she continued.

“So we're changing the way you learn about new features with short, easy descriptions geared towards end-users,” she said.

“Now, about 80 percent of the new features will just be on when you update. It will be like your smartphone, where new functions are just available to you.  You can choose to switch them off if you want to, but you wouldn’t know about them if they weren’t turned on to start with.”

The increasingly open mentality is extending to the data available for analytics and decision support, as well.

The social determinants of health are becoming critical data points to support comprehensive population health management initiatives, and EHRs can no longer afford to leave socioeconomic and patient-reported health data out of the clinical record.

“We have to look at who you are, what you eat, how much you sleep, and what your social conditions are like,” Faulkner stressed.

“We know these factors affect health.  You all know that the US spends more on healthcare with worse results than every other nation. What surprised me is that we spend so much less on social care than other countries, too.”

Source: Xtelligent Media

“We won’t be able to afford to continue doing what we’re doing if we don’t change the way we look at social determinants and population health,” she stated. “There is information that is not in the EHR right now that has to be accessible.”  

While Epic is by no means the only EHR developer pushing hard to integrate social determinants data to enable more comprehensive population health management, the company believes that it has a few unique advantages that allow its big data aggregation and exchange platforms to thrive.

“All of our live customers can interoperate thanks to CareEverywhere,” said Faulkner, referencing Epic’s interoperability platform. “We're told we're the only vendor that can do that. The others can't even talk amongst themselves like you do.”

Data exchange through the platform has doubled in the past year, jumping from 1 million transactions per day in 2016 to 2 million each day in 2017.

“What CommonWell has done in a lifetime, you do in 12 minutes,” she added, unable to resist taking a dig at an organization that has produced tensions in the past, although Epic tends to insist that there is no trace of competition in its relations with the Cerner-led network.

One in ten of those record shares is moving across state lines, Longhurst noted, which has been “fundamentally transformative” for the way he cares for his patients as a pediatrician who specializes in infants.

“The first time I saw patients after putting in Epic, I had an infant in California who was born and lived in North Carolina,” he said.  “The baby’s parents were visiting California to show their six-month-old off to family and friends, and the baby had a seizure and ended up in the ED.”

“When I asked the parents to tell me about their baby, they said, ‘Oh, well, he’s got that hole in his heart, and we see a cardiologist at Duke.’”

Prior to using CareEverywhere, a medical student would have had to call Duke and try to get the records released and faxed over, he explained. 

But since Longhurst knew that Duke was a participant in the network, “I was able to query them from the patient’s room in about 26 seconds,” he said. 

Christopher Longhurst, MD, CIO at UC San Diego Health

Christopher Longhurst, MD, CIO at UC San Diego Health

Source: Xtelligent Media

“I got the patient’s last pediatric cardiology clinic note, EKG, and also received their genetics testing results that showed the baby did not have the syndrome I was worried about.  That is a major, major shift in how I can provide care.”

The technology to support interoperability isn’t that complicated, Longhurst said.  What makes CareEverywhere work is the critical mass of organizations that makes it possible to effectively share information across traditional barriers.

And with Epic’s recent announcement of Share Everywhere, the web-based data exchange platform that does not even require users to have an EHR system, as well as an upcoming web-based version of Epic’s population health management platform called Healthy Planet Link, seamless interoperability and collaboration can easily go universal.

“I’m going to go out on a limb and say that from a technical standpoint, interoperability has been solved,” said Longhurst.  

“We have not seen it actualized yet because of processes, policies, and rules of the road, but the technical aspects are no longer a barrier.”

Judy Faulkner: Epic is Changing the Big Data, Interoperability Game

The Difference Between Big Data and Smart Data in Healthcare

Leveraging machine learning to predict, protect, and manage

Increased interoperability and Epic’s ability to aggregate enormous volumes of big data from its user community is enabling Faulkner and her team to go all-in on machine learning.

“Maturing technology holds a promise to help unlock practical solutions to complex problems,” said Sumit Rana, Senior Vice President of Research and Development. “Through time, care processes have evolved from being primarily driven by intuition, to also being driven heavily by data synthesis and algorithmic assistance.”

“More than 1.6 million patients have had Epic predictive algorithms run on them for hypertension, diabetes, asthma, and heart failure, which helps providers create targeted care management programs.”

That is just the beginning of the deep dive into machine learning

The company announced the development of a new Cognitive Computing Platform, built on the Microsoft Azure cloud, that will drive predictive analytics and artificial intelligence into a wide range of workflows.

“I’m going to go out on a limb and say that from a technical standpoint, interoperability has been solved.”

Soon, users will be able to have their documentation errors corrected, their care tasks optimized, and their decisions supported by the experiences of millions of other patients.

Natural language processing will extract meaning from free text and voice input to suggest next steps for patient care, how to best document them, and how organizations can ensure that they are getting properly paid for the services they provide.

Machine learning, alongside the integration of social determinants data into the patient record, will fuel a new generation of tools that can predict and close gaps in care. 

More accurate quality and performance benchmarking will be available by individual providers or by custom groups such as a specific scheduled shift, and a personalized EHR user experience will become second nature for providers.

“We don't have time to make everyone else's mistakes.  If you want to improve, you have to measure, collaborate, and implement changes,” said Davis from KLAS. 

“Personalization of the user experience is a big part of improving satisfaction.  Only 30 percent of users have set up personalization profiles. But when people customize their tools, their satisfaction skyrockets. We think there is a huge opportunity here to use emerging technologies to help that process along.”

Epic’s investments in machine learning aren’t rare by any measure.  Vendors of all sizes and types have been gleefully embracing the idea of artificial intelligence as a panacea for everything that ails the healthcare industry, with varying degrees of effectiveness so far.

Despite the magical theme of the week, Faulkner does not believe that machine learning is a fool-proof potion to solve all of the health system’s ills.  

While she has previously stated that she “has always felt that artificial intelligence was going to be big” in healthcare, she also expressed reservations about the idea that machine learning could or should supercede the experience and problem-solving skills of talented clinicians.

“It’s our responsibility to use AI to help highlight what a physician should look at next, and say, ‘Here's what we think,’” she told at HIMSS17 earlier this year.

Epic’s push into machine learning with its new cognitive computing platform will help to do that – and will, perhaps, represent a turning point for providers. 

For the most part, healthcare organizations have only dabbled in machine learning pilots and cognitive computing proof-of-concept projects, said Longhurst, as a result of niche vendors only offering AI-driven tools to solve very specific problems.

“At UC San Diego, we’ve had a couple of small predictive analytics projects for revenue cycle management and scheduling,” he explained. 

“Much of the machine learning landscape is currently very driven by use cases.  That has been great for showing that machine learning is feasible, but I’m excited to see this deployed across a suite of tools from a vendor with so much influence in the direction of the industry.”

“I believe that Epic’s eagerness about machine learning is going to help transition those local efforts into more general initiatives as we make better use of the tools on offer.”

Creating a seamless, joy-to-use EHR environment with machine learning at its core will require a great deal of effort from providers and their technology developers. 

“We don't have time to make everyone else's mistakes.”

Strong governance principles, a nimble and engaged organization, and attention to the details of data integrity will be crucial for organizations that wish to leverage innovative technologies to their fullest potential, cautioned Davis.

“Bureaucracy kills innovation and paralyzes decision-making.  Of course you need processes in place, especially to measure opportunities for improvement, but too much will make it impossible to implement change.”

“And you need physician builders to get involved and stay involved,” he added.  “When you allow physicians to participate in making the EHR work for them, they will actually do it.”

It also helps to have an innovative and responsive technology partner that doesn’t promise vaporware without delivering on the goods, Longhurst said.  

Many developers make very attractive promises about their innovations in machine learning or other cutting-edge developments, and it is important to make certain that their concept art and sales pitches really do translate into day-to-day clinical care.

Making the leap from idea to implementation is difficult, Faulkner acknowledged, and requires much more than a sorcerer’s incantation or a witch’s brew.

“I wish there was magic so we could just wave a wand and fix everything,” Faulkner said. “But there isn’t.”

“It's just us: our caring and our hard work. We have to be the wizards.  The possibilities are boundless, but there is still so much to accomplish.”

This article was originally published on September 27, 2017.


Join 25,000 of your peers

Register for free to get access to all our articles, webcasts, white papers and exclusive interviews.

Our privacy policy

no, thanks

Continue to site...