Guiding interventions to the specific subpopulations that could most benefit from them may seem like common sense; however, implementation of this process is rare among health systems in value improvement efforts. Health system leaders often find it attractive to reach for new tools and technologies with hopes for quick fixes to reduce expenses. Rather than implementing a new technology or tool across a heterogenous population, however, it was key for UCLA Health’s leaders to roll up their sleeves and understand the spectrum of needs that their patient had first. This exercise helped them to focus often limited resources to where improvements would be expected and address those problems that truly exist locally.
Health system care processes often take a siloed approach, whereby care is often fragmented and costly. Using methods of co-creation was slow initially but also the reason for success in implementation, especially in large-scale care redesign. Implementation of the initiatives required cocreating care models that brought together siloed entities, resources, operations, and data management into one common care pathway. Using population health methods including risk stratification facilitated creating care pathways to better keep high-cost patient subpopulations healthier and out of the hospital.
Investing in ambulatory strategies and a population health infrastructure can develop care models to curb expenses. By viewing the health system beyond the walls of brick-and-mortar hospitals, UCLA identified some common-sense opportunities to reduce unnecessary facility utilization.
The greatest innovations often rest in radical common sense — made after stepping back to identify real problems and deliberately applying viable, sometimes simple, solutions. In light of national and state incentives to improve health care value, health systems have numerous opportunities to implement common-sense care to meet patient needs and these aims. To accomplish this, health system leaders must ask themselves: Why do we currently have processes blind to common-sense problems, why do we lack proactive guidance of patients into appropriate care, and is there more we can do?
For example, it is possible to know when patients with late-stage kidney disease miss dialysis. However, health systems without strategies to capture and act on this information may be allowing these patients to clinically decline until they reappear in the hospital with electrolyte abnormalities or breathing problems. Similarly, it is possible to know when patients on life-saving medications do not pick up prescriptions; yet, often these patients are out of sight until they enter emergency departments in distress.
UCLA Health, motivated by these clinical stories and a need to reduce unnecessary expenditures, aimed to use this common sense in redesigning care for patients with chronic kidney disease. CKD is a leading cause of expenditures nationally with Medicare expenses for CKD beneficiaries (12.5% of total) exceeding $79 billion in 2016, with an additional $35 billion for end-stage renal disease. The UCLA Office of Population Health and Accountable Care leadership realized that there was an opportunity for improvement, as the average expense for kidney injury admissions requiring dialysis topped $42,000 while care centered out of clinic or home was significantly less expensive.
Here, we describe their development of an ambulatory Population Health Value infrastructure to proactively prevent and capture clinical decline among the high-cost population of CKD patients to reduce unnecessary utilization and expenses. Because of the diversity of needs and range of costs associated with different types of kidney care, the organization used a population-based approach that tailored interventions to the needs of subpopulations.
Approaches to Overcome Challenges in Identifying, Stratifying, and Characterizing Patients’ Needs
First, the leadership team identified appropriate patients for outreach and tracked the intervention with longitudinal data developed at UCLA through custom clinician-guided analytics. This early effort to develop valid, actionable data with clinician involvement helped to reduce challenges associated with care team engagement by increasing ownership of data while reducing the feeling of drowning in unclear data. The maintenance of similar definitions across initially siloed teams also facilitated engagement.
The team defined and validated a CKD patient registry using Epic clinical and claims data, clinically meaningful utilization outcome measures, and process or balancing measures (e.g., patient on indicated medication for proteinuria, stage 3b patients seeing nephrology, patients with history of hospitalizations in tiered care management, late-stage CKD patients with advanced directives, etc.). There were 17,172 patients with CKD in the organization’s primary care network that accounted for $5,559 per-member-per-month expense primarily due to hospitalizations.
The organization believed that implementation would require cocreating care models that bring together siloed entities, resources, operations, and data management into one common care pathway. Learning from prior experiences, leaders believed that the extra time required to cocreate would help to prevent drifting off initiative goals while maintaining transparency and continued engagement over time, and engendering trust to facilitate implementation across key stakeholders.
The organization reviewed charts from the CKD patient registry to determine providers and staff with frequent CKD patient interactions. From this group, a representative 15-member, multispecialty leadership team including physicians, nurses, quality and value improvement specialists, informaticists, and care coordinators convened to cocreate a system-wide intervention. Clinicians and staff from primary care, nephrology, dialysis, interventional radiology, hospital medicine, emergency medicine, advance care planning, and palliative care each represented their respective disciplines, while care coordinators focused on understanding CKD patient experiences.
It was vital to invite patient voices to the discussion to avoid predictable challenges associated with developing a relevant, patient-focused intervention. For example, patients preferred one main point-person within the health system to serve as their contact when questions or problems arose rather than members of different siloed divisions contacting them. This team aimed to identify patient needs and structure the improvement intervention.
The team developed a risk-stratification model that placed individual patients in risk tiers based on their associated expenditures (Figure 1). The team believed that understanding and meeting patients’ needs through interventions for all risk tiers would be more effective to achieve short- and long-term returns. The team expected that the most expensive 1% of patients who account for 22% of total expenses alone may be cost neutral because this population requires significant resources to meet their complex needs. The team believed investing in the 19% of middle-risk patients (58% of expenditures) and bottom 80% of patients (20% of expenditures) would reduce long-term costs.
The team used case reviews to identify patient needs as opportunities to improve care coordination and access to ambulatory care sites, especially for late-stage CKD patients who accounted for most expenses. High-needs patients at risk of decline (i.e., Tier 1) had issues with missing dialysis due to social influences including limited transportation and food insecurity and concurrent mental health diagnoses left undertreated. These patients also had limited access to catheter placement outside of emergency care to avoid clinical decline. Additionally, these patients had low rates of exposure to advance care planning and palliative care. Patients with middle risk (i.e., Tiers 2–3) often just began developing symptoms of volume overload or electrolyte abnormalities leading to at least one hospitalization or emergency department visit in the prior year. These patients often felt unprepared for symptoms and were unclear how to respond in the ambulatory setting, so they would rush to the emergency department. Patients with lower risk (i.e., Tiers 4–5) were asymptomatic but could benefit through health care screening, condition-specific preventive therapies, and reducing unnecessary care.
Approaches to Overcome Challenges in Care Pathway Development
A common challenge that our health system previously faced was understanding the best ways to direct limited resources to the initiatives with the greatest chance of success and avoiding homogeneous interventions across heterogenous patients. Therefore, after understanding the needs of CKD patients based on their risk level, the team developed three care pathways to align clinical and social service care across primary and multispecialty settings. The pathways focused on proactively developing common-sense interventions to support declining patients to avoid mortality, unnecessary utilization, and expenditures. These patients included those with gradual progression of disease to those who already have late-stage disease and decline rapidly.
Tier 4 and 5 Patients:
These patients received CKD education to better align patient and caregiver expectations about disease progression by leveraging online materials. The team developed best practice alerts in the electronic health record to guide clinicians and staff to reinforce evidence-based kidney preservation therapies. Patients with CKD stage 3b or higher were triggered for a nephrology referral and received early education about dialysis. While this step had previously been a challenge to improvement teams, it was doable because key information technology and informatics stakeholders were present through program planning and, therefore, prioritized the initiative.
Tier 2 and 3 Patients:
Once patients developed advanced disease or inpatient utilization, they entered Tier 3, and the electronic health record triggered a suggested referral to nephrologists. Patients in Tiers 2 and 3 were assigned a CKD care coordinator and focused on case management including communicating with continuity clinicians or staff to proactively manage a patient’s clinical decline prior to requiring hospital utilization. The care coordinator identified patients through mechanisms regardless of care site including patients who: presented to health care with symptoms (within or outside of primary care), called in from home, or who did not present to care (i.e., missed appointments, concerning changes in lab patterns, or facility utilization within or outside of the local medical center).
The care coordinator expedited appointments and coordinated communication between clinicians and staff across specialties and disciplines (e.g., increasing diuretic medications, scheduling outpatient catheter evaluations or palliative care). The care coordinator also monitored cases that fell out of the care pathway to identify opportunities for program improvement including staff training for new staff. The intervention also focused on increasing CKD patients’ receipt of social services, goals of care discussions, and referrals to other specialists when appropriate.
While the CKD care coordinator role was a less costly alternative in staffing (approximately $70K in salary and benefits), one challenge included the care team’s comfort to rely on an unlicensed CKD care coordinator to oversee the intervention. The team overcame this initial challenge after the planning team gained consensus about the scope of the role and protocols to address situations when additional clinical expertise is required. Buy-in improved when an internal training program was developed under a supervising nurse manager and the lead nephrologist for the intervention. Clinicians also began seeing that the role would help to relieve additional work that would otherwise be added to their workload.
Additionally, the role was developed as a centralized glue for the intervention to avoid duplication or omissions due to silos across the health system. The role held the responsibility to keep eyes on the patients in the CKD registry with the most current knowledge of key contacts across the system and available resources. Maintaining this centralized role required continuous training across new staff through the health system so that care team members were aware of the role and to avoid urges to admit patients unnecessarily.
UCLA also enhanced ambulatory infrastructure by outlining community resources to assist with social influences and extending interventional radiology hours to increase access during evenings and weekends to place and declot dialysis catheters. A strategy of staggering work shifts helped to mitigate operational cost concerns of needing new hires to staff the center.
Tier 1 Patients:
In addition to the services described above, the care team reviewed individual patients’ clinical and social needs from the top 1% who often require intensive care management. They created nuanced, creative solutions to coordinate care, address social influences (e.g., delivering food to patients’ homes), reduce high utilization, and initiate palliative care.
Initial evaluation compared monthly mean inpatient hospitalizations per 1,000 patient years between 18 months prior and 12 months after intervention implementation. Findings from a study (publication in process) by the authors showed a nearly 2% reduction in monthly inpatient hospitalizations (p-value<0.004) among late-stage CKD patients accounting for most expenditures. The team is currently using this strategy across other subpopulations with high expenses within the health system.
Acknowledgements: The authors would like to thank Varsenik Papazyann-Gutierrez, Connie Lee, and Lily Roh.