Renton, Wash.-based Providence health system participates in 140 value-based care agreements. During a panel session at the Value-Based Health Care Congress this week, Deepak Sadagopan, Providence’s senior vice president of value-based care and population health informatics, described how critical a strong data infrastructure is to the transition away from fee for service.
Providence is a nonprofit system of 51 hospitals across seven Western states. Sadagopan’s role is to lead the transition from volume to value across the portfolio of different payment models, and in the process, support that transition to a data-driven approach. “That is where the importance of data infrastructure emerges in the space,” he said, “because you cannot drive a transition into value-based care without using information to empower every aspect of decision making — on the clinical, administrative and the business side.”
Sadagopan began by providing some context about the pace of change. “Looking back at 2010, 99 percent of our revenue and payment models were fee for service. When you fast forward to 2020, we have about 140 unique value-based care agreements,” he said. “We operate the fifth-largest MSSP program in the country covering over 120,000 beneficiaries. Overall, across all of these programs, we have north of 1.7 million patients covered under one form of risk arrangement or another, so we have seen an overwhelming shift over the past 10 years.”
As Providence tries to manage total cost of care, key areas it is addressing include things like acute care admissions, avoidable readmissions, avoidable ED visits, and use of brand-new pharmaceuticals.
The data infrastructure is critical to that work, he said, pointing to three categories of information he called essential to support executing on these arrangements at the field level. One is as real time as possible clinical information about the population that you're serving; second is claims information that provides a holistic view around expenditure and acuity of the populations you serve, and third is the performance information, which is critical because you need to understand at any given point in time how you're performing with respect to your assigned populations from a quality standpoint, from a total cost of care expenditure standpoint, all of which is usually provided to you retrospectively. “That makes it very challenging to manage populations on a day-to-day basis,” Sadagopan said. “At Providence we are working on taking real-time clinical information and combining it with claims data into what we call a population health data model that we use to drive all our operations downstream as it relates to supporting care delivery in a team-based model.”
He noted that acquisition of information from hundreds of sources is a challenge. “Over the past few years, our team has done an exceptional job in transitioning to a single electronic health record,” Sadagopan explained. “Inside the system, we have homogeneity in that space, with all our all our clinical information emerging from one electronic health record source, but as we partner with affiliates, as we partner with other community providers, then that equation gets complex really quickly. There's a diversity in every community that we serve in terms of the clinical information sources.”
In terms of sourcing data, Sadagopan believes traditional HIEs [health information exchanges] don't really work well in this space. “They provide a service that I'd say is like using a sledgehammer to drive a nail into drywall. We have got a very limited set of information that we need to effectively manage our populations, and getting a vast amount of information about every patient that is served by all our affiliates doesn't really help. We need targeted information that enables us to measure quality of care delivered, efficiency, efficacy of care delivered, and transitions of care.”
He said Providence also needs to gather reference data such as community-level social determinants information, as well as industry benchmarks. “We constantly need to keep our eye on our total cost of care target,” he explained, “and in many of these contracts, our obligations require us to manage care within a particular budget. Often you don't know what the target is until the end of the performance year. Imagine if you're shooting at a target but they don't actually tell you where the target is. That's kind of what we have to do under many of these arrangements. We have to use historical information to develop our own internal perspective on the type of targets that we have to meet in order to meet our commitments under this portfolio of arrangements.”
Once they gather data, Providence data analytics teams want to be able to meet the needs of three types of users or personas: the clinical persona — the team that is actually managing care for the patients in the offices and acute care spaces; the operational managers who are managing utilization, and coordinating care; and the business managers who are responsible for supply of resources.
“These three personas use the same information, but in very different ways,” Sadagopan noted. “Providing the relevant information to these different personas requires both an aggregation and simplification of information that we see about each population, as well as inferences about how we're going to perform at the end of the year. This is where advanced technologies like machine learning and artificial intelligence really help in enabling us to say how we may be performing at a particular level and total cost of care. Does it put us at risk of meeting the target or exceeding the target? And if so, what interventions would we need to undertake to be able to meet the needs?”
By looking at clinical and claims data together, Providence is able to drive its prioritization in terms of key areas of utilization compared with national or regional trends, Sadagopan explained. It could be underutilization or overutilization in specific areas such as skilled nursing facility days or inpatient days. They can then identify if it's isolated to one or more facilities and prioritize where their focus needs to be.
“When we combine expenditure information and clinical information, we're able to get to a point where we can see where we need to focus in on doing better with managing care pathways for our joint patients or other types of cohorts that may be seeing different levels of utilization compared to what we generally refer to as well-managed populations that have better outcomes,” Sadagopan added. “It is about driving a better patient experience. Our success in this space is almost entirely dependent on our ability to deliver high-quality care and retain our patients over a period of time, so we know from evidence that the longer we are able to retain our patients and the more comprehensive view we have of their care, the better we're able to help them manage their outcomes and get to a point where they are better from a self-managed standpoint. That's really the goal — to empower our patient populations to achieve better outcomes by using all these different data sources as signals that are funneled to our caregivers.”
Sadagopan closed by saying the next area of focus in this space is a concept Providence calls healthcare 2.0. “Essentially, we are envisioning how to redesign a system of care that is focused mainly on primary care and prevention and a system of care that is entirely designed around the patient.”