Executive Summary
One major challenge for health care organizations is finding most cost-effective ways to serve patients who have complex health problems and social needs. A CareMore program for such Medicaid patients in Memphis, Tennessee, has generated encouraging results. Its elements include: instead of focusing only on patients who incurred high costs in the prior year, using predictive models, claims data, clinical criteria, and clinician judgment to identify the patients most likely to benefit most from complex-care management; identifying the unique drivers of poor outcomes for each individual patient; and forging strong partnerships with community-based organizations and social-safety-net institutions such as food banks and housing authorities.
For roughly two decades, health care organizations have been aggressively experimenting with programs to improve care for high-need, high-cost patients. Most of those efforts focused on care models for chronically ill and frail elders, but organizations are now increasingly developing programs to better serve Medicaid beneficiaries with complex health and social needs. Although select programs appear to be effective, a recent high-profile evaluation of one of them — the Camden Coalition of Healthcare Providers — has raised concerns over the utility of care models for medically and socially complex patients.
Such skepticism is premature. In an article in the new issue of the American Journal of Managed Care, we published encouraging results from a complex-care-management program serving high-need, high-cost Medicaid patients in Memphis, Tennessee. The program, built on top of an existing integrated–care model, lowered total spending by keeping patients out of the hospital. We believe our experience with this program offers important lessons for others seeking to design and implement their own complex-care-management programs.
CareMore Health — a physician-led integrated-care-delivery system that’s part of Anthem Inc. — began providing comprehensive care services to Medicaid beneficiaries in Memphis in 2015. An early analysis of our patient population revealed that spending was remarkably concentrated: The costliest 5% of patients incurred roughly 70% of all spending. Most had multiple chronic medical conditions, often with co-occurring behavioral health disorders. Many also had significant social needs, ranging from housing instability to food insecurity. As part of an integrated-care-delivery strategy, CareMore developed a program aimed at providing high-touch, comprehensive care for these complex patients.
In designing the program, we drew on our experiences caring for Medicaid patients and from existing programs that targeted medically and socially complex patients. We augmented our primary care physician-led medical home model with a full-time community health worker and greater support from social workers. Patients who enrolled in the program underwent a comprehensive, multi-disciplinary assessment of their medical and social needs, the results of which were used to create a tailored care plan.
Patients in the program received frequent, structured follow-ups. The community health worker contacted patients weekly (via the phone or in person) to check in, evaluate progress, and address barriers to their adherence to the care plan (e.g., transportation or health literacy). The community health worker, social worker, and primary care physician reviewed the care plan weekly, re-prioritizing tasks and assigning new responsibilities. Patients returned to CareMore care centers monthly for in-person visits with the entire team. Additional follow-ups were customized. The community health worker accompanied some patients to specialist visits and social service appointments. The social worker provided counseling for behavioral health needs, helped navigate social services, and arranged for necessary referrals and medical equipment. The primary care physician saw patients in the office to address gaps in care and stabilize chronic conditions.
Results and impact. To disentangle the effect of the program from unrelated changes in utilization (including regression to the mean), we evaluated its impact through a randomized controlled trial. We found that the program led to a $7,732 (or 37%) reduction in total medical spending per patient per year. This was driven primarily by decreases in hospital utilization: Patients were less likely to be admitted to the hospital (50% decrease), and when they were admitted, their hospital stays were shorter (62% decrease). We also saw a small decline in specialist visits, possibly due to more active management of chronic illnesses by the primary care physician. Patients were highly satisfied with the program: Its net promoter score (measured three months after enrollment in the program) was 100 out of 100.
Here are lessons for others.
Precise patient targeting can improve effectiveness and efficiency. We used predictive models, claims data, clinical criteria, and clinician judgment to identify the patients most likely to benefit most from complex-care management, rather than focusing only on patients who incurred high costs in the prior year. Incorporating clinician judgment allowed us to harness the intuition and wisdom of care team members — and we found that patients referred to the program based on clinician judgment were more likely to engage with the program and experienced greater reductions in spending and hospital utilization.
Programs with an integrated model for addressing medical and social risk may be most effective. We augmented our existing integrated-care model with increased staffing, resources, and protocols to identify and manage social risk. Close integration between the community health worker and primary care physician was necessary to rapidly address the social drivers of poor health outcomes and unnecessary hospitalizations. For example, after one of our community health workers discovered that a patient was no longer able to afford his insulin, she immediately alerted his primary care physician, who prescribed a more affordable regimen within an hour.
Focus only on the most relevant drivers of poor outcomes. Our planning process began by identifying the unique drivers of poor outcomes for each individual patient. Specific attention was paid to the drivers that mattered most to patients and those that could be addressed over the following weeks to months. For patients with multiple social risks (e.g., housing instability, poverty, loneliness, food insecurity), this exercise helped clarify where to direct early attention and resources, allowing for rapid stabilization in many of the most complex cases.
Partner with community-based organizations and social-safety-net institutions such as food banks and housing authorities. Blind referrals and attempts to coordinate services in real time are often bureaucratic and cumbersome, resulting in long wait times for patients who need quick attention. Building trusting, longitudinal relationships with organizations in the Memphis area was critical for our patients. For example, we discovered that one patient in our program was frequently utilizing the emergency department due to a lack of safety and support in his group home. Drawing on connections we had built with group homes in the community, we were able to help the patient rapidly transfer to a more supportive home, bypassing traditional administrative hurdles and wait times.
What’s next for the field?
Our results and those of other successful models suggest that carefully designed and targeted programs can improve care and reduce spending for high-need, high-cost Medicaid patients. What’s needed now is a better understanding of which program elements work best for specific patient groups and what it takes to rapidly scale successful interventions. Important work in this area is already underway at places like the University of Pennsylvania. At CareMore, we are standardizing workflows, implementing our model across new markets, and closely measuring its impact.
The United States needs a broader commitment to improving the health and social services provided to complex Medicaid patients. If ever there was a population that demands our persistence, ingenuity, and commitment to finding delivery models that work, it’s the highest-need patients in communities like Memphis.
The authors would like to acknowledge the contributions of Manisha Sharma, Caroline Hagan, Paula Ma, Brisa Samudio, Yolanda Sutton, and Nupur Mehta.