In response to COVID-19, health care organizations around the country are mobilizing to identify patients at high risk of getting infected by the virus and/or having poor health outcomes if they contract it. As it becomes clear that this pandemic will have long-lasting effects, the conversation around identifying risk is broadening to include who will be at risk in the coming months (and possibly years) due to the health care disruptions and ongoing emotional and financial toll.

Spurred in part by the pandemic and the recent release of two rigorous evaluations of complex care programs, emerging efforts within the field to better understand what being “at risk” means mirror conversations that select participants in the Center for Health Care Strategies’ Complex Care Innovation Lab (CCIL) have been exploring since 2018. With support from Kaiser Permanente Community Health, eight complex care organizations participating in CCIL kicked off the Rising Risk Initiative to better understand how to identify “rising risk” populations — that is, individuals who are not yet “high-need, high-cost” but are on the trajectory to become so. Additionally, the group sought to gain insight into contributors to this risk trajectory, as well as how to develop interventions that can disrupt it. This blog post highlights lessons from the initiative, as well as opportunities to support health care organizations in identifying and meeting the needs of “at risk” patients in this new COVID-19 era.

Considerations for Identifying Rising Risk Populations

Each of the eight CCIL sites sought to identify “rising risk” in different ways, highlighting that: (1) the concept of rising risk is still in the nascent stages of definition; and (2) “risk” is in and of itself contextual. Some sites found it useful to address this challenge by first defining poor outcomes for this group (i.e., what are they at risk of?) and then building identification models around these parameters. Three of the participating sites — Denver Health, Boston Health Care for the Homeless Program (BHCHP), and the University of California, San Francisco (UCSF) — demonstrate the different ways that teams approached the concept of risk. Each site was interested in better understanding of how homelessness impacted individuals’ risk trajectories, but each used a different approach to assess risk. After realizing that many of its patients who experienced homelessness were categorized as “low risk,” Denver Health partnered with a local housing and health care partner to augment its existing risk-tiering algorithm with housing status data. Building on this, Denver Health and the Colorado Coalition for the Homeless subsequently launched a supportive housing pilot for Medicaid managed care patients identified as rising risk. BHCHP was curious whether applying a “medication trauma” lens — assessing the negative impact that complex medication regimens and fragmented coordination among providers can have on patients — would provide another dimension to evaluate patients’ rising risk. In San Francisco, the county’s Whole Person Care pilot had already identified a group of 4,000 homeless individuals with co-occurring behavioral health needs. UCSF’s rising risk work examined which of these individuals were at risk of moving from the top five percent of high-utilizers to the top one percent.

While the approaches that sites took to identify rising risk populations were diverse, several common themes emerged:

1. Taking a Population-Wide Approach

Because most complex care efforts focus on identifying patients who interact with the health care system the most and/or in the highest-cost ways, rising risk patients are much less likely to be on health care systems’ radars.

Because most complex care efforts focus on identifying patients who interact with the health care system the most and/or in the highest-cost ways, rising risk patients are much less likely to be on health care systems’ radars. To understand who these individuals are, the CCIL sites found that analyzing their entire population was essential. CareOregon, for example, found that a cluster analysis model of its entire population was an important first step for better understanding behavior patterns beyond high-need, high-cost utilization and identifying groups of people who have common patterns of interactions with the health care system but were previously not recognized. This included, for example, individuals with multiple chronic conditions who have lower engagement rates with primary care, but no inpatient admissions.

2. Factoring in Social Risk

Sites considered identifying data points that might be signals of rising social risk, such as a change in marital status indicating that someone has recently been divorced, or multiple changes of address within a short amount of time.

The sites in the Rising Risk Initiative explored ways to better understand how social risk factors contribute to risk trajectories and where best to gather these data. Approaches included collaborating with community partners and identifying publicly available data sources, such as the Area Deprivation Index. Other sites considered identifying specific data points that might be signals of rising social risk, such as a change in marital status indicating that someone has recently been divorced, or multiple changes of address within a short amount of time. Anand Shah, MD, MS, Vice President of Social Health at Kaiser Permanente, shared with the group that in his experience, incorporating social data into already robust risk stratification or segmentation models may not do much to improve these models’ predictive abilities, but can shed light on the types of interventions that might stabilize individuals most effectively.

3. Validating Identification Approaches with Frontline Staff

Sites found it beneficial to validate their data analytics models with frontline staff to confirm that patients flagged by the models were consistent with their perceptions of which patients were likely to become less stable and more costly over time. This was also a critical step in building staff comprehension of the concept of rising risk and getting their buy-in for focusing on patients in this cohort.

Focusing on Rising Risk Moving Forward

The above considerations for identification are foundational in developing tailored interventions for rising risk populations. The following early lessons from the Rising Risk Initiative provide useful insights for both near-term COVID-19 related activities, as well as ongoing efforts to refine complex care models to maximize their impact and efficacy.

Forging Stronger Provider/Payer Partnerships

Partnership conversations can enable payers and providers to work together to share data, identify needs, and deploy services for individuals who need longer-term support due to the pandemic and beyond.

Identifying patients as rising risk has implications for health care providers, systems, and plans. Since CareOregon’s network is contracted, their rising risk work consisted of facilitating conversations with its provider network focused on understanding who is best positioned to lead the care for its population segments. The plan refers to this process as understanding the “yours, mine, and ours” of their membership — that is, which patients are best managed by network provider partners (yours), which are best managed by the health plan (mine), and which need a combination of both (ours). These types of partnership conversations point to ways in which payers and providers can work together to share data, identify needs, and deploy services for individuals who need longer-term support due to the pandemic and beyond.

Refining the Complex Care Model

Just weeks before the pandemic hit the United States, many in the complex care field were focused on the release of a randomized control trial of the Camden Coalition of Healthcare Providers’ care management model. The evaluation showed that the program had no impact on 180-day hospital readmission rates; spurring renewed discussion about which patient populations are best served by complex care programs and what types of interventions are most impactful. Insights gleaned from the Rising Risk Initiative about ways to refine the identification of specific subpopulations and understand the key, modifiable drivers of their risk provide a useful foundation for exploring this topic further.

Continuing the Conversation

In the coming months, CHCS will publish a series of case studies exploring the various approaches and lessons of the Rising Risk Initiative sites. If your organization is focusing on rising risk, we would love to hear from you. Please leave a comment below or email Audrey Nuamah (anuamah@chcs.org).

This site uses Akismet to reduce spam. Learn how your comment data is processed.

0 Comments
Inline Feedbacks
View all comments

Related Resources