The journey of healthcare leaders towards optimally leveraging data and analytics for population health strategies remains a long and daunting one for leaders in both Canada and the United States. That was the consensus of discussants during a panel entitled “Data & Analytics: Driving Population Health Management,” which was held on Friday, Sep. 18, during the Health IT Summit in Vancouver, sponsored by the Institute for Health Technology Transformation (iHT2, a sister organization to Healthcare Informatics, under the Vendome Group, LLC corporate umbrella).
The panel was moderated by Trevor Strome, manager, Informatics & Process Improvement, in the Emergency Program at the Winnipeg Regional Health Authority (Winnipeg, Manitoba). Strome was joined by Jat Sandhu, Ph.D., regional director in the Public Health Surveillance Unit at Vancouver Coastal Health Authority (Vancouver, British Columbia), Alyssa Daku, vice president for strategy, quality and risk management at eHealth Saskatchewan (Regina, Saskatchewan); Larry Svenson, director, epidemiology and surveillance, Alberta Health Services (Edmonton, Alberta); and Eugene Kolker, chief data officer at Seattle Children’s (Seattle, Wash.).
One of the initial challenges that discussants talked about was that around defining the population health concept to begin with. “I don’t get the sense that we have a really clear understanding and agreement of what population health management really means,” Strome said, as he opened the discussion Friday afternoon.
“I agree, I think the term population health management is overused, particularly in the U.S.,” Sandhu responded. “In Canada, we think about it rather differently. It’s about the organization or management of health delivery focused on achieving outcomes that are effective and safe,” Sandhu said. “And at the crux, it’s about being proactive about the management of at-risk populations, making sure they don’t become too expensive to the system. So it’s about managing clinically defined populations or those that are frequent users of the system. And to be effective, we need to obtain good longitudinal data.”
Seattle Children’s Kolker said, “I would agree that it’s very important to understand populations at risk, and it’s not good to limit ourselves geographically; at the same time, we need to understand scale. It’s not only about the population that are frequent flyers. In an ideal case, looking at this with a ten-year perspective, I’d like us to keep populations healthy lifelong.”
One of the elements to consider, said eHealth Saskatchewan’s Daku, is that, “When you talk population health, it very much sits in the health sphere. As soon as you add ‘management’ into that,” she said, “there are so many variables that exist outside the healthcare system that exist outside our purview. It becomes incumbent on social services, education, and government, to become involved.”
Public health perspectives are important in that regard, said Alberta Health Services’ Svenson. “Because I work primarily in public health,” he said, “we’re always thinking about these non-health entities and how you integrate that in. Looking at children in care—they have lots of mental health visits and other interventions. Once the interventions take place, their utilization actually drops off. The other part of this is that, in Alberta, 5 percent drive 65 percent of costs; but that is not a stable, permanent group. It changes over time. And there’s valuable data coming from many places. Our biggest challenge in government is how you derive meaningful policy decisions that turn into meaningful care decisions.”
“Yes, as soon as you begin pulling educational, and social services, and correctional data, that adds tremendously to the complexity,” Daku agreed.
“I agree,” seconded Sandhu. “We need that other kind of data. And our challenge is that it’s not necessarily easily collected; so we tend to rely a lot on aggregate geographical data with which to do analysis. But that shouldn’t stop us. The methodologies around managing some of that data have been around for a long time. I’ve heard here a few things about how you incorporate new kinds of data, beyond the in-hospital and EMR data,” he continued. “We do have this much broader ecosystem, from which we need to be able to grab data to much better understand situations. So what are some of the best strategies for making the data analytics work for population health?”
Strategies for pursuing population health: a complex subject
Responding to that question, Seattle Children’s Kolker said that “I think that that’s a big question. I can share some of our experience,” he added. “At some point, we became focused on the backgrounds of our patients, our minority patients, for example. We talked to many people, including in social service organizations, and then our team was working on something completely unrelated, and realized we were seeing an unexpected slowing in our business. And we found that we needed to look specifically at our Asian population, to better understand what was going on with them.” Kolker explained that while the Asian-American population is the fastest-growing ethnic population in Seattle Children’s service area, the organization had a cultural perception problem in that community. “We discovered through research based on focus groups, that we needed to present ourselves as not only serving the sickest of the sick,” in order to better capture market share in that ethnic community. “So we changed our image over time. We accidentally uncovered a problem and were able to combine internal and external data of different types to understand our relationship with a specific population.”
“Here’s a story about a project we did,” offered Alberta Health Services’ Svenson.”I’m involved with a lot of the evidence to support a lot of our immunization programs, how safe and effective they are. We had heard about an increase risk of seizures over a particular vaccine, for MERS. We were able to bring data from different sources; we followed half a million kids longitudinally, and came up with a way to look at that data in a more sophisticated way; and we found that there was a risk of 1 in about 2,800-2,900 cases, and were able to show a real risk, and share that with our frontline clinicians and with parents. And that became part of a package deal. And we were able to actually estimate a numerical risk. We said, this is a real risk, but that the risk of disease far outweighed that of the vaccine; but we gave parents information with which to make an informed decision, and some parents asked that the MERS vaccination be separated from others. And that combined public health data and other data, and that doesn’t happen routinely.”
Winnipeg Regional Health Authority’s Strome emphasized that “In all the analytical work we do, it’s important to understand that we still have a bit of disconnect between the analytics and some of the discussions happening on the front line. Alyssa,” he asked Daku, “what are some concrete steps we can take to make sure the insights we get from the analytics actionable to population health management on the front line?”
“I think there are two pieces here,” Daku said. “One piece is the connection to the front line; the other is the connection to policymakers. On the front line, in the chronic disease and complex patient space, we’ve been able to do some work around cascading metrics, ways to take some of those higher-level outcomes and bring them to the front lines—for better management in diabetes care, for example. We’re looking for certain outcomes for patients to achieve. So understanding what your objective is, and being able to cascade your metrics down to the front line, is very important, and we’ve had some real success in that area. Now, per policy, it’s not always intuitive as to how we get policymakers to take action around population health management. But our board of education was able to craft a decision to ban all soda pop and junk food machines from our high schools in Regina; and that was driven by statistics we were able to derive from analytics. So it involves bringing the evidence forward. So we have to be able to cascade up and cascade down, and present at both levels.”
Vancouver Coastal’s Sandhu said, “I agree with what Alyssa says, and sometimes, that disconnect between the analytical and front line comes out of a lack of communication; but also, we need to see what outcomes the analytics can generate. And there are cases where metrics have helped drive decision-making for front line decision-makers as well. Also,” he noted, “a lot of the clinical information tends to. And as Rick Skinner [chief information and technology officer at the University of Virginia Health System] said yesterday, we need to involve people who can derive nuance from data. We have a lot of resources focused on the health information management (HIT) management, and not on drilling down to some of the nuances in the data; we need data scientists.”
“And we’re drowning in data,” Strome emphasized. “And when we’re coming up with metrics and indicators, sometimes it’s hard to know what data is important to measure and include. How do we work with data analytics people to get to the right data and support pop health management?
Daku shared what she called “a recent experience with alternative public health data in Saskatchewan.” As she noted, “We certainly know that there are a lot of individuals in our acute-care facilities who do not require that level of intensity of care. And when we looked, we found we were capturing tons and tons of data, but not the right data. The systems of data capture weren’t designed to actually capture the data we needed. So we did a rapid process improvement process to capture the right data. We had clinicians, HIM [health information management] folks, data coders, and we worked very intensively for a week; and we found that 30 percent of people sitting in hospitals weren’t waiting for long-term care; they were waiting for home care or for rehab care services they couldn’t access, or for other services they couldn’t access, or had social factors affecting them. So that was a great example of how that kind of process can result in data that is usable and actionable and meaningful.”
Svenson noted that “One of our challenges is that the change management side is often one of the biggest challenges. The other is understanding what drives those metrics. What do policymakers really need? So the policymakers will say, we want all these performance measures. And we have immunization coverage rates—how well are we doing? We just show the rates. And when the rates appear to go down, people say, oh wow, that sucks—what next? It’s really a signal. We think of it as signal detection. We need to then dig deeper; and we need to look at communities with lower rates of immunization, for example. And we started looking more closely and found that single parents are less likely to have their children immunized. Single mothers are 50 percent less likely to have their children immunized, and single fathers are 80 percent less likely to do so. So part of this ends up being, how do we ensure that working single parents can access immunization services? Because it might be that we don’t have clinics open when they’re able to take their children. So we need to continually determine how we link our data-gathering and analytics processes to other systems of processes related to how policy-makers and clinicians use the data we are able to share with them.”
And for those kinds of outcomes to emerge, Daku noted that “Lean [management] has driven a lot of accountability at senior levels, around governance. People are afraid of the linking of data to some processes, which I understand. At eHealth Saskatchewan, we have statisticians, data analysts, etc. We make sure to de-identify data. And we have to make sure we do a good job explaining and doing our due diligence to demonstrate that we are following the right policies and correctly using the data, with regard to data privacy and other policies.”
“ It’s similar in Alberta,” Svenson reported. “We do have very strong privacy practices, in terms of how we go through processes to allow for the sharing of data within the health sector without any issue. In the Canadian context, of course, the provincial governments are the payers. The challenges have to do with non-healthcare-sector data, actually.”
