Day 1 of last week’s Health 2.0 conference in Santa Clara, California was a Provider Symposium, where innovation personnel at some of the nation’s large and notable health systems came together to speak about their experiences around innovation, big data, and patient engagement, among other topics.
Data isn’t enough: Machine learning, big data, and interoperability
One theme that emerged from Sunday’s discussions was that despite all the focus on collecting more data and using machine learning and natural language processing to analyze that data, the missing piece for many providers is how to translate data into action.
“My biggest question to my colleagues when I’m talking to them about a new project is not only 'why do you want to know this?' but also 'what will you do with this information once you have that information?'” Sutter Health Chief Health Information Officer Dr. Sameer Badlani said. “Once you have this piece of information, do you have the ability to make all your floor managers, all your pharmacy benefit managers, react in a certain way to those analytics?”
Without a concrete way to integrate data into workflows, machine learning algorithms can leave front line workers — including nurses and doctors — perplexed.
Venkat Mocherla, director of business development and marketing at Qventus, a startup providing predictive analytics to Sutter as well as other hospitals, used the analogy of Google Maps to describe the problem.
“Let’s say you’re using Google Maps and you're driving home from work, and the app says ‘There’s a 37 percent chance you should take a left turn and, by the way, here’s a standard distribution of traffic in realtime’,” he said. “That’s not useful. What we’ve learned is it’s not enough to predict, you need to be able to prescribe in simple English and persuade people to change their behavior. Which isn’t just machine learning, it’s behavioral science, decision science.”
From what executives have seen so far, machine learning may even be a little overhyped.
“[Machine learning] has theoretically lots of potential, but any time I’ve actually seen it used it hasn’t really been as useful as the proponents thought it was going to be,” Dr. John Lee, chief medical information officer at Edward Hospital, admitted. “It ends up being another form of alert fatigue.”
Alert fatigue is a common problem with systems in hospitals that try to manage the influx of real-time patient monitoring data by alerting doctors or nurses when vital signs cross certain thresholds. The problem is, if too many alerts go off, providers will stop paying attention to them entirely. This can extend to the EHR interface as well.
“According to a June 27th Epic national average for all of their customers, per 100 orders, 50 alerts are fired,” Kevin Baldwin, informatics portfolio manager at UCLA Health Sciences, said. “There’s 50 pop-ups telling you to do something different, or are you sure you want to do X, Y, and Z? And of those alerts that are fired, there’s a 90 percent override rate. So nine times out of ten the physician is clicking the X, not even reading it. They’ve been trained to do something and these popups are not going to change their behavior.”
One thing that will limit data overload is hospitals being more intentional with the data they collect.
“The way to think about this is if you can’t answer the question ‘Why do I need this data,’ what are we doing worrying about it at all?” Sahid Shah, chairperson of HealthIMPACT and moderator of one of the sessions, said. “If you don’t know what the output is, if you can’t walk it all the way back, just focus on the things you can answer that question for. Here’s why I need to know X. Just ask the simple questions before we worry about the harder stuff.”
The other perennial data problem discussed was interoperability. In a short presentation, Dr. Steven Lane, Sutter’s clinical informatics director, said that interoperability actually has gotten better in the last few years. But there’s still a long way to go, and right now it comes in the form of a variety of disparate networks and protocols, some of which are underutilized.
Lane shared a story of recently being awoken at 4 a.m. while on call because a patient’s blood test results had come back, and the patient was at another system’s hospital. He had to call that hospital on the phone to inform them.
“That’s 20th Century interoperability, but that’s all that’s going on there,” he said. “We have all these people involved but it’s still a phone call in the middle of the night. Still dependent on a human being being awake enough. It’s telephone calls and people trying to do the right thing. Imagine all the points of failure in that process. And imagine what it will be like when we are able to truly interoperate. When that critical lab result didn’t come to me in bed, it went right to the workstation that was taking care of that patient in realtime.”
Falling in love with problems: Innovation and partnership strategies
More and more health systems are forming innovation groups of various kinds to bring new ideas and new technology into hospitals. At the symposium, innovation discussions centered on partnership, and how hospitals can pick partners out of a crowded and overwhelming field of potential vendors.
The innovation leaders assembled had different ideas about a lot of things, but all agreed that the best partnerships are formed when a hospital has a clear idea of the problem it needs solved and then selects a vendor to do it, rather than letting vendors call the shots.
“We have a list of tenets that we abide by in the innovation department,” Aaron Franklin, project manager at Providence St. Joseph Health, said. “One is to fall in love with a problem and not a solution. So we’re trying to share that with the outside world and say ‘Here’s a problem we’ve identified, now it’s up to you to show us how crisply and clearly your solution will solve that problem and show us the value proposition, and if you do that, well, it’s a no brainer getting in the door and you’ve saved yourselves and our healthcare organization a lot of phone calls and emails that would otherwise just go into the garbage.”
Vendors need to understand this paradigm as well, however. Many of the panelists expressed frustration at the barrage of vendor-pitches they receive. Sanjay Shah, director of strategic innovation at Dignity Health, offered an alternative outlet for that energy.
“I don’t envy the position you’re in trying to crack that nut, but the traditional sales process of yesterday, where it was ‘get your foot in the door until you can land and stand and grow and scale’ is just as resource-intensive and inefficient as trying to do it right,” he said. “If you spend a little bit more time and energy, you can get your entry point, but don’t use that to get to operations. Use your entry point to understand what is the system environment, what is the scale environment, who else needs to be involved in the conversation.”
Dr. Ashish Atreja, chief technology officer at Mount Sinai and founder of NODE Health, suggested that hospitals could also go a long way toward making it easier to choose between vendors if they worked together.
“The bigger problem we have stumbled upon is the problem of evidence,” he said. “Right? How do we choose between 500 apps when there’s so much innovation happening? Which one do we go big with in our health system? None of the health systems can do it alone, none of the providers can do it alone, and we really need to come together to share all the evidence.”
Atreja founded NODE to try to get a handle on that evidence in a way that multiple providers can share. Matthew Fenty, director of technology innovation and strategic partnerships at St. Luke’s University Health Network, also spoke up about the need for hospitals to communicate with each other, noting that it’s a natural extension of a problem-focused approach.
“If you’re one of five or 10 hospitals tackling the same issues and the only difference is what’s your technology stack, what’s your operational structure, we’re all trying to address these billion dollar problems,” he said. “What we do is we work with each other, we have discussions. As a provider system, we have to be more open and collaborative. Less inward thinking. Share with each other what works and what failed.”
Jennifer Wiler, executive director of the UCHealth Care Innovation Center in Colorado, laid out two additional criteria for choosing partners.
“We’re looking for partners with three really basic criteria in mind," she said. “The first is, are companies solving problems that we really have in our health system? … Two, can they really execute on their technology solutions, can they be good technology partners to us, and then finally, do we have a clinical expert who’s going to champion this project and really start to make it part of their own career development and be the lead of this opportunity.”
Finally, Franklin said, making decisions can be easier if you limit the size of the team weighing in.
“We’ve implemented with what we call a two-pizza team,” he said. “We brought that over from Amazon. No decision should be made by a team that’s larger than the number of people who could be fed by two pizzas. As a technologist, I shouldn’t make decisions about legal. There’s a particular person that should be asked about that and they should be asked at a specific time. We see this all the time where people are going, I don’t want to say out of their lanes, but it turns into this very unclear governance structure. The more you can combat that, you’ll see your process go much faster.”
Beyond the portal: Patient engagement and patient satisfaction
The patient portal is still one of the main ways hospitals interact with their patients. But some of them are trying to save it from the relative obscurity into which it tends to fall.
“We track the usage of our patient portal,” UCLA’s Baldwin said. “Ninety percent access it through the web and only 10 percent through the mobile app. So I think there’s an opportunity to make that mobile app easier to use.”
Panelists discussed how the patient portal needed to supply more value to the patient to see stronger adoption, comparing it to apps in sectors like finance and retail. Dr. Danny Sands, chairman and cofounder of the Center for Participatory Medicine, said the very nature of a “sick care”-oriented healthcare system is an obstacle to achieving that kind of ubiquity.
“We think of patient portals as how we get in touch with our doctor when we’re sick,” he said. “And I don’t want to interact with that. So how do we change the psychology of that? I’ve been working with patient portals for a long time, they’re not really built to reach out to healthy patients, and we don’t really have the incentive to do that.”
But if there’s still a difficulty getting patients at home to use the portal, at least it can be reworked to support patients in the hospital, who are already in a patient mindset.
“What we’ve done in partnership with our EHR vendor is look at how we can consolidate that using the patient portal on the inpatient side,” Baldwin said. “The patient has access to the portal, they can read through who their care team is, who’s their RN, who’s their attending physician, are they going to have any labs drawn. If they have dietary restrictions they might be able to place orders for lunch and dinner. If they have spiritual needs they might request those as well. The list goes on of things you could expect someone to want or need.”
Panelists also lamented the limitations of the traditional Press-Ganey survey as a useful device for gauging patient satisfaction.
“We talk about lagging metrics but I would also say that we have a problem with false surrogate metrics,” Lee, of Edward Hospital, said. “We think they represent patient satisfaction, but we really want to know how are patients are feeling in aggregate about the care they’re receiving. From a business perspective, do they feel a connection to your organization, do they want to keep coming back to your organization? … Patient satisfaction most often comes in the form of what I think is a false metric, the Press-Ganey survey result. That, from a business perspective, is not something I actually want to know. And on top of that it’s a lagging metric that doesn’t actually help you drive decisions.”
Badlani agreed, extending the same sentiment to employee satisfaction scores.
“You measure employee satisfaction after enticing them with donuts to participate in the survey and then you process it through a very expensive third party for six months,” he said. “Now at the end of the year, you’re responding to an emotion expressed by an employee nine months ago. And here we are talking about real-time analytics!”
What’s his suggestion instead? Take a lesson from the airline industry.
“Delta customer service used to ask 10 questions. Now they just ask ‘Would you hire this person if you were running the call center?’ Just one very meaningful question,” he said. “If the question was asked as you were leaving in a simple yes or no ‘Would you hire this physician?’ or ‘Would you let this hospital take care of your father?’, you’ve got the answer you need as a caregiver.”