States nationwide are working to strengthen primary care amid workforce shortages, burnout, underfunding, and misaligned incentives. As part of multi-faceted initiatives to address these challenges, states and payers are designing and implementing primary care population-based payment (PBP) models, which shift primary care payment away from volume-based fee-for-service and toward predictable, upfront, per-person payments tied to quality. These models aim to provide revenue to primary care practices in a way that is more predictable and flexible, to better support financially vulnerable practices and enable more holistic care.
Massachusetts has been a leader in Medicaid primary care PBP through its primary care sub-capitation program* — a PBP or capitated model — and its participation in CHCS’ Medicaid Primary Care Population-Based Payment Learning Collaborative. In this blog post, authors from Massachusetts Medicaid (MassHealth) explore lessons from developing the nation’s first primary care-specific Medicaid risk adjustment approach — insights that can help other states pursuing rate-setting in primary care PBP models.
In April 2023, Massachusetts’s state Medicaid and CHIP program, MassHealth, launched a statewide primary care sub-capitation initiative in which MassHealth invests in primary care to help providers shift their care model and operations away from typical fee-for-service medicine and toward more team-based, integrated primary care to improve their patients’ experience and quality of care. Primary care practices participating in MassHealth’s Accountable Care Organizations (ACOs) are paid a fixed per-member, per-month rate for a defined set of primary care services. ACOs already receive a global, risk-adjusted capitation payment from MassHealth for total cost of care. The primary care sub-capitation is a capitation payment for a subset of the total cost of care services (hence “sub” capitation) from ACOs to primary care practices.
In developing the sub-capitation payment approach, MassHealth designed a risk adjustment model specific to primary care, called the Primary Care Effort Model (PCEM). PCEM is now in use across approximately 900 primary care practices serving over one million MassHealth members. The goal in risk adjusting primary care sub-capitated payments is to distribute funding such that practices with members requiring higher primary care effort receive, on average, more funding than practices with members requiring less effort. By effort, we mean chronic conditions and other diagnoses that may increase the complexity, and therefore the time required to provide comprehensive, high-quality care. We believe that this effort is not always captured in claims and utilization and that we should pay more to practices that have more complex members, even if some of those practices have historically seen lower utilization.
Risk adjustment, a statistical modeling tool used to predict health care costs for a specified population, is commonly used in managed care capitation rate setting. However, there are very few primary care-specific risk adjustment models in the literature. MassHealth’s model is specifically focused on the factors that predict primary care costs based on a narrowly defined set of codes that are core to the primary care setting. We found that the variables predictive for primary care spending are meaningfully different from those that are predictive in our total cost of care model.
Key Considerations for Primary Care Risk Adjustment
Following are key findings that policymakers and payers can consider in developing a risk adjustment model tailored to the primary care environment. These insights are based on modeling using claims and encounter data from 590,000 MassHealth members from July 2021 through June 2022.
1. Managing chronic conditions is a major part of the primary care provider’s time and effort — and therefore, spending.
During model development, MassHealth found that as the number of chronic conditions a distinct member has increases, primary care costs also meaningfully increase. To account for this relationship, PCEM utilizes its underlying diagnosis-based model’s (DxCG by Cotiviti) Aggregate Condition Categories (ACCs), a set of 31 clinical systems and conditions. ACCs create several model variables capturing the compounding effect of chronic condition counts across members. This approach borrowed concepts from Colorado’s APM2 Chronic Condition risk adjustment approach, which accounts for 12 chronic conditions.
MassHealth also evaluated clustering for specific conditions based on complexity and cost, but found that to be less predictive of primary care spending. This suggests that the number of chronic conditions a member has is more predictive of primary care effort than the specific conditions themselves or their complexity.
MassHealth found that primary care costs associated with these chronic conditions differ meaningfully between adults and children. PCEM therefore uses separate model weights for each group, which improved prediction for both populations.
2. Adjusting for behavioral and mental health conditions within the model, particularly within Medicaid, is predictive of primary care effort.
Though behavioral health codes are not included in the primary care sub-capitation code set, MassHealth found that behavioral health conditions, particularly opioid use disorder, have a strong impact on primary care effort, demonstrated by improved prediction and significant positive values in the corresponding model coefficients. This suggests that even if behavioral health spending takes place largely outside of the primary care sub-capitation program, members with these conditions require greater primary care effort. These coefficients are separate from and mutually exclusive to ACCs.
As with chronic conditions, MassHealth found a statistically significant difference between adult and pediatric behavioral health costs, resulting in a separate set of coefficients for each population in the risk model.
3. Risk adjustment may not always be the best way to account for health-related social needs in primary care.
MassHealth evaluated the inclusion of several social factors that often impact health outcomes including homelessness, housing instability, and an area-based geographic index that captures various measures of social deprivation. These factors either yielded significantly negative coefficients in the model, indicating that members with these social risk factors generated less spending than other members, or were found to be insignificant in predicting primary care costs.
We were one of the first states to include social risk factors in our total cost of care risk model, but our findings show that social risk in the primary care setting behaves differently. For example, in the case of homelessness, the adjustment in MassHealth’s total cost of care model is highly positive (i.e., these members generate more spending), while the adjustment in PCEM modeling was highly negative, indicating under-utilization of primary care services for these members. As a result, MassHealth decided that primary care payments, and primary care risk adjustment in particular, are not the right lever to improve care for members experiencing homelessness, and chose to exclude these factors from PCEM to avoid penalizing providers for caring for these more socially complex members. Additionally, MassHealth found that the final PCEM model predicted primary care costs for homeless members well without the addition of an independent variable.
Next Steps
MassHealth found the final model to be significantly more predictive of primary care spending than alternatives such as age/sex categories alone or a total cost of care model. The final model has been applied to all Massachusetts primary care practices with MassHealth members, and we found that the model predicted historical costs to within 10 percent for most practices. Most key populations, such as those with higher counts of chronic conditions, behavioral health diagnoses, and disabilities were also predicted within 10 percent of observed costs.
In future iterations of this model, MassHealth would like to explore ways to include primary care effort that is more difficult to quantify and may be non-billable, for example coordinating care with social service organizations. MassHealth heard extensive feedback from primary care provider groups that this kind of work represents significant effort by provider groups, confirming the need for further exploration.
Primary care risk adjustment is only one part of the sub-capitation payment model, which is in turn one part of a broad effort to transform primary care delivery for Massachusetts Medicaid members. MassHealth hopes that its work to transform primary care will inspire other payers to follow suit, leading to more predictable and equitable payment, and increased investment, in primary care.
*CHCS uses the term “population-based payment” (PBP) to refer to payment models that are based on upfront, per-person payments intended to replace some or all fee-for-service revenue and that also have a tie to quality. This can include primary care PBP models, like MassHealth’s primary care sub-capitation program, as well as models focused on different types of care, including the state’s ACO program. MassHealth refers to these types of payment approaches as “capitation” payment.
