A learning health system (LHS) is described as creating a virtuous cycle where data collected in routine care is studied for variations and best practices, and that knowledge feeds changes to the system to create better performance. The concept has been gaining momentum for several years, but since 2020, the pandemic has really given the movement a boost, and more health systems are establishing infrastructure and training programs around it.
In a March 2023 talk to the Center for Learning Health System Sciences at the University of Minnesota, Paula Lozano, M.D., M.P.H., of Kaiser Permanente Washington’s Health Research Institute, said KP Washington’s experience during the pandemic might offer some keys to successful scalability of the LHS concept.
“When you think back to March and April of 2020, this was a time when learning health systems around the country sprang into action. At KP Washington, the topics were around modeling incidents of COVID-19, figuring out how to set up, design and evaluate virtual care, looking at predictive models to figure out who was shut in at home and isolated and needing refills on their medications to avoid exacerbations,” recalled Lozano, who is a senior investigator for the Kaiser Permanente Washington Health Research Institute and director of the Center for Accelerating Care Transformation. She also serves as senior associate medical director for research and translation at the Washington Permanente Medical Group.
“We also had a tidal wave of new information coming out about the disease, possible treatments and diagnosis approaches,” she said. “As scientists, we had to help our colleagues in healthcare synthesize all of this to know what to do from one day to the next.” Kaiser researchers also tracked utilization to understand how they could respond to this pandemic, including understanding who is best poised to benefit from new medications.
“A lot of us learning health systems researchers felt validated,” Lozano said. Healthcare executives could see how important being a learning health system is when research is embedded and when the situation is urgent. “It is hard to imagine a more urgent situation, and it was all hands on deck,” she added. “It was researchers elbow to elbow with the care delivery leaders. Our priorities were aligned. It was so clarifying that we needed to get through this moment.”
She stressed that the concept has not reached its full potential at scale. “The fact remains that in academic circles we talk about learning health systems, but there are places in the healthcare industry where they're talking about these concepts, but with much less enthusiasm than the folks in academia,” she said. “Also, when you look at the kinds of systems that have been set up in academic medical centers and integrated systems like the VA and Kaiser Permanente, those are the places where we really have fully developed LHS models or at least models that are going toward being fully developed, but the generalizability to other populations has not really been demonstrated at scale."
Center for Learning Healthcare at Vanderbilt
In January 2023, Vanderbilt University Medical Center launched the Center for Learning Healthcare led by Matthew Semler, M.D., assistant professor of medicine in the Division of Allergy, Pulmonary and Critical Care Medicine, and Cheryl Gatto, Ph.D., research assistant professor of biostatistics and associate director of the Vanderbilt Institute for Clinical and Translational Research (VICTR).
Semler says the culture of a learning healthcare system that had developed at Vanderbilt prior to the pandemic paid off in a huge way. A critical care physician, he spent two years working in the COVID ICU. “I saw firsthand that when COVID arrived, by definition, we had no knowledge of which treatments were effective,” he says. The disease hadn't existed a month prior, so no one could know if a given drug worked. Many health systems across the country administered untested therapies pretty widely in clinical care, some of which turned out to be effective, and some of which turned out to be harmful. “I think that culture of learning healthcare was a key piece of why Vanderbilt was able to get out in front and do two things: one was to be clear about focusing on evidence-based therapies in clinical care, but also very quickly opening opportunities to learn which therapies were effective,” he adds.
Hydroxychloroquine is an example of one of these therapies that was widely administered in the early part of the pandemic, and we now know it was either ineffective, and there may be some concern for harm, Semler says. “While others administered it broadly to patients, we didn't do that. We helped start up a randomized trial 13 days after the pandemic hit our area, and that was one of the trials that helped show hydroxychloroquine actually is no better than placebo for COVID in the inpatient setting. Instead of giving unproven treatments to patients early in COVID, we funneled that uncertainty into clinical trials and rapidly mobilized because we had the infrastructure to support standing up that type of work so quickly.”
Over several years, VICTR developed a learning health system platform that has focused on identifying areas of uncertainty in common treatments or in the way the health system delivers care, Gatto explains. “The platform enables pragmatic clinical trials to understand which therapies or healthcare delivery approaches work best, and then implement those into practice,” she adds. “The platform provides the technical, procedural and human infrastructure required to find and shape those different trial concepts to assess how to most pragmatically implement them in the system and then provide the wrap-around support required to conduct them to find out if we could drive an evidence-based best-practice recommendation.”
One of the first projects the learning healthcare system platform took on was comparing saline vs. balanced crystalloids in the administration of IV fluids, Semler says. “We saw that variation in practice and structured it in the form of a pragmatic trial. Then when we identified that balanced crystalloids were better, we implemented that into our health system in a structured way such that basically all clinicians and all patients were receiving balanced crystalloids, which was a big change, because historically, saline had been the most commonly administered fluid.”
Semler says the era of consolidation may help expand the LHS concept because there will be more health networks and health systems and fewer stand-alone hospitals. “That creates opportunity for standardization and evaluation,” he says. “The shift to value-based care could help, too. When the health system is incentivized to focus on value, then investments in understanding what creates the best outcomes becomes in the health system’s interest. We're likely to see this process initially led by academic health systems, but this is obviously the way you would want your local hospital to work is to be continuously evaluating care and improving it, so I think that it will become much more common globally.”
The view from PCORI’s perspective
One of the key organizations driving the nation’s learning health system efforts is the Patient-Centered Outcomes Research Institute (PCORI), which has partnered with the Agency for Healthcare Research and Quality (AHRQ) to fund the training of learning health system scientists. AHRQ and PCORI have co-funded 11 academic institutions to train embedded scientists in the skills needed to conduct patient-centered outcomes research and comparative clinical effectiveness research.
Erin Holve, Ph.D., M.P.H., M.P.P., PCORI’s chief research infrastructure officer, says setting up learning health system infrastructure can be challenging because there are genuine tensions between health system delivery needs and the structures and processes for research, particularly for investigator-initiated research. “There are competing priorities, particularly for highly qualified folks who have a lot of different career opportunities,” she says. “Recruiting and creating a career path for individuals who have the interest and experience to do this multidisciplinary work can be challenging. Not all systems are well aligned to support that.”
Holve says the expanded use of telehealth is another example of the impact of the learning health system during the pandemic. PCORI has funded over 100 different projects in the telehealth space, and there was already a strong evidence base for the use of telehealth, but it was also a tool that could be used to help support research. “I think it's a really good example for learning health system-type questions, because it was both an intervention and a method of doing the research and the ability to look at those things hand in hand, was quite an opportunity.”
Scaling up from projects at one institution to research across multiple health systems adds complexity but also offers potential for greater impact. The PCORI-funded PCORnet network was created almost a decade ago to do research studies that are national in scope, bringing together high-quality health data, patient partnerships and research expertise that can deliver fast, trustworthy answers. It has eight clinical research networks with 70 data sites, thousands of participating providers, and 30 million unique patients who can participate in research a year. “The encouraging message, is that these data sites and these networks are functioning — they regularly work nimbly on research projects together using our common data model,” Holve says. “I was there at the very start of PCORI and PCORnet, and there were real questions about whether or not this type of learning model would work. I would say that now we have very good evidence with the hundreds of studies and manuscripts and lessons that have been learned from the network that this type of collaborative learning model really can work.”
The issues around making those large-scale networks work well are as much about policies as about technology or data standards, Holve says. “There's no question we have increasingly good data, good technology to make these things work. The challenge is really the policy or process interoperability issues.” In her research, she has found that even where people were all using appropriate standards and all have the latest EHR technology, and health information exchange, the policy decisions about where to put data, or how to exchange data with consent are the trickiest issues. “While PCORnet has made great strides in that direction, we still have some distance to go because it's all these local decisions that have to do with local workflows that are still a genuine challenge.”
New center created at the University of Minnesota
Timothy Beebe, Ph.D., interim dean of the School of Public Health at the University of Minnesota, says that health systems often put innovations in place with no sense of the state of the science. Often, there is no systematic review of the literature and no robust measurement infrastructure around it. “With the creation of the Center for Learning Health System Sciences (CLHSS) in 2021, we wanted to take advantage of some of these quality improvement activities, and put more rigor around them. The center is trying to engage the departments and leadership about their priority areas. How can we deploy something that really moves the needle?”
The traditional passive engagement approach by researchers studying healthcare delivery has contributed to the 17-year arc between discovery and deployment to the bedside, Beebe said. PCORI funded a training program to train learning health system scholars and the University of Minnesota was in that cohort starting in 2018, and it got a match from the from the university and Fairview Health System to double the number of scholars. Then COVID hit. The rapid cycle learning that took place during the pandemic required narrowing that period from 17 years down to like 17 days. That really was a use case. The rapid synthesis and curation of evidence, deployment to the bedside, and creation of data structures to facilitate that required researchers and practitioners to interact in real time. That was a good use case to make the case to organizational leadership that this is how research should be done. That's when the leadership decided to invest in the center because they saw the need to do this more systematically. We need to resource it well, and engage the system in this more rapid sort of innovation where the practice provides evidence and evidence informs the practice.”
Clear alignment on the goals
Speaking about the learnings from their work within Kaiser Permanente Washington, Lozano described some other features of the partnerships during the COVID-19 situation that allowed them to be successful. “We were fully engaged partners. Our healthcare colleagues invited us to the table,” she said. “We were all in and they wanted us to be there. As researchers, we really wanted to understand what the needs and priorities of the healthcare organization were. We did not have a grant. There were not hypotheses we were testing. We were showing up trying to help and needing to understand what our partners’ problems were so that we could be most helpful.”
From that came clear alignment on the goals and what their roles were as researchers. “We were able to bring that wealth of expertise that we have, from quantitative to qualitative and all the different methodologies and approaches and study designs that we could apply in these varied and challenging settings,” Lozano said. “Accountability, I think, is really key, because there was a huge focus on getting it right. We held ourselves accountable to act on the findings that we came up with. There was this intense focus on the learning health system as a way to do business.”
This led her to observe that researchers need to understand the key business priorities of their partners. “It's our job to bring those rigorous, analytic approaches and we should make sure they're rapid. If the problem needs to be solved in two weeks, doing something that takes six months is not going to be relevant,” Lozano said. “Being flexible about our methods and bringing other colleagues in who have complementary methodologic expertise and finding the knowledge that exists in the health system about how the whole system is experiencing whatever the problem is — COVID or whatever — is an important part of being responsive researchers.”
