Moving Beyond Identifying ‘High Utilizers’ in Complex Care Initiatives

May 18, 2021
Executives from Kaiser Permanente, CareOregon and New York City Health + Hospitals share their experiences trying to better understand populations and tailor delivery models

Many complex care interventions have identified populations based on cost/utilization measures or by identifying “high utilizers” of acute care. But some health systems are working to get better at identifying people with complex health and social needs to develop tailored interventions for targeted subpopulations.

During a recent webinar, executives from Kaiser Permanente, CareOregon and New York City Health + Hospitals shared their experiences trying to better understand populations and tailor delivery models that improve care for individuals who experience a combination of medical, functional, behavioral, and social needs.

The webinar, coordinated by the Better Care Playbook and made possible through the Seven Foundation Collaborative, was led off by Michelle Wong, M.P.H., M.P.P., director of the Care Management Institute at Kaiser Permanente, who described ongoing work in Kaiser Permanente’s Complex Needs initiative.

Because Kaiser Permanente is an integrated system, with a health plan, a medical group, and hospitals in eight different regions, it’s a complex ecosystem, Wong said. “We feel that this integration and complexity uniquely positioned our organization to learn, and that just as there's a need to embrace the intersections in our complex needs patients, there is as well within our organization.” She said KP is working to bridge silos across these entities in order to improve care. “At the Care Management Institute, we have a saying, which is that we need to make the right thing easier to do. What we're doing is focusing on trying to understand the right thing to do for our complex needs patients, and then creating systems to make it easy for our practitioners to do the right thing.”

KP’s approach has been to create a portfolio of learning activities, which are always connected to practice, to care teams, and patients. The health system has a research unit called CORAL (Complex Care Collaboration: Operations, Research and Leadership) that pairs KP research investigators with operational leaders to generate actionable evidence to improve complex needs care.

CORAL tests and evaluates patient-centered interventions and systems across Kaiser Permanente and partners with external organizations to advance the field and evidence-base for complex care.

“This learning approach has been focused on supporting local teams by connecting them with each other, and often supplementing them with support, like evaluation, patient and provider ethnography, standardized screening and care plan tools and workforce competencies and training,” Wong said. Research plays a critical role in this learning approach, she added, and it highlights how their thinking has evolved. It has helped them move from the earliest concepts of the high utilizing 1 to 5 percent to a more refined thinking about the integration of social/behavioral functional needs, and now into work on how to think more clearly about population identification. “We're still learning at Kaiser,” Wong said, “but we have a strong belief that connecting the research to operations and down to our care teams and our patients in both directions will be critical for us to learn how to best provide care.”


CareOregon is a managed care organization that support the needs of many dual-eligible Medicare/Medicaid individuals in the state of Oregon.

During the webinar, Jonathan Weedman, vice president for population health at CareOregon, described its data segmentation approach for identifying impactable “rising risk” populations — individuals who are on the trajectory to becoming complex, but whose needs and utilization may be stabilized through tailored interventions.

They began working with a partner, Collective Medical Technologies, as they looked at data for patients who are utilizing the ED as well as inpatient treatment. They started to see patterns in that data. “It really sparked our initial thoughts around segmentation and how we want to start matching services to the right population,” Weedman said. “We started to look more deeply into the data beyond just utilization data to help guide our process, and that's really what began our journey into population segmentation.”

CareOregon’s historical state, like many organizations, was based on the acute episode for identifying members who needed care coordination. “They would go to the emergency room or to the inpatient setting, and we would then intervene and help provide care coordination,” Weedman described. “The interventions were very reactive, because they were based on an acute episode, and it was very much criteria-based. That was limiting us in terms of doing any kind of proactive identification to prevent an incident from happening, as opposed to helping manage the incident after it occurred,” he explained. CareOregon’s newer proactive state is more about identifying members or patients prior to acute episodes. CareOregon works with its contracted provider networks to determine which things the provider can do and which things the managed care plan can do to help. “You're really using data to identify the physical, social and behavioral health needs of members and patients,” he said.

“In our current model, we are in the process of putting folks into segments to understand behavioral patterns, inform our resource allocation to address their specific needs, and observe population-level trends, Weedman said. “Are members getting better over time? Are they getting worse over time? And how do we identify member-level trends by provider clinic to inform opportunities for quality improvement support?”

An example of one group within the rising risk cohort includes people with at least four chronic conditions. They are engaged with their primary care clinic and do not have any significant ED or inpatient use. There may be interventions available for that person, including making sure they connect with a behavioral health specialist in the clinic if needed, or a referral to specialty mental health, for example. “From our historical perspective, this would have been a person who wouldn't have been on our radar, because they weren't utilizing the healthcare system in the way that we were functioning then,” Weedman said. “We wouldn't really capture them in terms of need, although clearly, they probably need support as well.”

New York City Health + Hospitals

Anne Marie Young, M.B.A., director of complex care at New York City Health + Hospitals, described how her organization created an operational guide to identify, understand and treat high-needs patients. She noted that the guide and predictive model are open source tools that any healthcare systems at any stage in their work on complex care could use in informing their own work.

“In order to better identify complex care patients within our system, we implemented a predictive modeling approach,” Young said. “This allows for proactive population risk scoring, and it allowed us to generate a list of people who are likely to utilize services within our system that cost the most and are really at highest risk for poor health outcomes. We wanted to combine the analytics that we had worked on through the predictive modeling along with clinical insight.”

NYC Health + Hospitals developed a qualitative segmentation framework in collaboration with a cohort of providers. One segment involved people with mixed medical and behavioral health needs, who primarily had acute care utilization driven by recent life events that could include the onset or diagnosis of a serious illness or the exacerbation of a serious illness. Another segment involved a group that is struggling to self-manage. This includes people with both mixed behavioral and medical needs compounded by a limited ability to live independently. People in this group may have functional limitations or the need for skilled nursing services. Still another segment they labeled “basic needs for better health.” This is a mix of behavioral-health-driven and medical need-driven population. “Really what drives much of their complexity is that they lack fundamental resources to get well,” Young said, including things like housing, food security, and social support that have an impact on their health.

“We defined several care models that exists within the field of complex care and we match some of those models to the segments that our providers helped us to develop,” Young said. “This helped us identify gaps in services within our healthcare system, and helped to focus our complex care strategy as a healthcare system moving forward.”

Young’s colleague, Jillian Diuguid-Gerber, M.D., is the lead physician at Woodhull Hospital Primary Care Safety Net Clinic, an intensive primary care clinic that is specifically tailored to meet the needs of patients who are homeless and who have complex barriers to primary care.

“Our  mission is to effectively engage people who are homeless, who have complex barriers to primary care to provide dignified trauma-informed care, focusing on our patients, while addressing their chronic diseases and their mental health needs, which often does include addiction, and lastly to use our interdisciplinary care team model that combines primary care clinicians, social work, and nursing care coordination in order to be able to provide as much in terms of wraparound services for our patients as we can,” she said.

Her clinic took advantage of the predictive modeling and qualitative interviews to help identify cohorts of patients. They used the high-risk algorithm that was created by the data team to distill some of that information. “So all you need for a clinical referral for us is identifying this patient is homeless, and this patient has the high-risk flag in the system,” Diuguid-Gerber explained, “and if they meet those two criteria, we say they qualify, and then allowing for clinical judgment along the way as well.”

Her recommendation for how to replicate a pilot program for patients with complex needs is to start with your needs assessment and understanding what subpopulations you might have within your group of complex defined patients, and then figure out how to make that simple and operational. The second important thing, Diuguid-Gerber said,  is identifying what resources you already have available, and how you can best use them to serve your particular patient population. “Start small, and keep iterating, keep changing,” she said. “You can't do this alone. Our patients are complex. So reach out to the external organizations who might be able to work with you and share your mission and share your values.”

Sponsored Recommendations

Trailblazing Technologies: Looking at the Top Technologies for the Emerging U.S. Healthcare System

Register for the first session of the Healthcare Innovation Spotlight Series today to learn more about 'Healthcare's New Promise: Generative AI', the latest technology that is...

Data: The Bedrock of Digital Engagement

Join us on March 21st to discover how data serves as the cornerstone of digital engagement in healthcare. Learn from Frederick Health's transformative journey and gain practical...

Northeast Georgia Health System: Scaling Digital Transformation in a Competitive Market

Find out how Northeast Georgia Health System (NGHS) enabled digital access to achieve new patient acquisition goals in Georgia's highly competitive healthcare market.

2023 Care Access Benchmark Report for Healthcare Organizations

To manage growing consumer expectations and shrinking staff resources, forward-thinking healthcare organizations have adopted digital strategies, but recent research shows that...