PROJECT SUMMARY
Individuals with serious mental illness (SMI) are three times more likely to die prematurely than the general
population. Sixty percent of this premature death is attributable to inadequate care of chronic, co-morbid
medical conditions. A growing proportion of patients with SMI receive care in accountable care organizations
(ACOs) – health care delivery and finance systems, in which global payments and quality benchmarks are
used to incentivize quality care and lower spending. An increasing number of states are implementing ACO
models in their Medicaid programs, the primary source of health insurance for low-income Americans with SMI.
Medicaid ACOs exist in 12 states, caring for over 6% of the Medicaid population. Medicaid ACOs have the
potential to both improve and worsen access to, and quality of, care for low-income adults with SMI. While
financial structures that incentivize care coordination and programs that address health-related social needs
likely benefit those with SMI, inadequate global payments that fail to account fully for social adversities such as
homelessness could result in lower quality care. Certain features of Medicaid ACOs, e.g. leadership structure
or ACO size, may amplify the benefits or drawbacks of the ACO model for patients with SMI. Evidence from
Medicare ACOs has shown that smaller, provider-led ACOs and those serving a lower proportion of socially-
vulnerable patients perform better in terms of quality. However, no evidence exists on how Medicaid ACO
characteristics affect quality of care or the care experience of SMI adults. Since Medicaid ACOs are rapidly
proliferating, filling this evidence gap is critical and can inform the evolution of the ACO model to better achieve
the goal of mental health parity. In this K23 research plan, we propose to identify features of Medicaid ACOs
(e.g. provider-led vs. hospital-led ACOs) that produce the highest quality ACO care for adults with SMI. We
will undertake this research objective in three critical domains. First, we will study whether certain ACO types
tend to drop patients with SMI, a practice known as favorable risk selection (or “cherry-picking”). This
phenomenon can cause instability of ACO enrollment for patients, itself a marker of lower quality care.
Second, we will compare care access and quality among adults receiving care in different ACO types. Lastly,
we will use mixed methods to examine the care experiences of adults with SMI receiving care in different ACO
types, through in-depth interviews and subsequent integration of qualitative and quantitative findings. For the
first two aims, we will use Massachusetts All-Payer Claims Data and leverage the state’s unique auto-
assignment mechanism as a natural experiment to study favorable risk selection, access, and quality. Auto-
assignment refers to Medicaid randomly assigning individuals to ACOs, allowing us to compare outcomes
across ACO types without selection bias caused by patients self-selecting their ACO affiliation. The goal of my
K23 research is to guide policymakers and hospital administrators in shaping the Medicaid ACO model to
produce more stable, higher-quality ACO care for patients with SMI.