Evaluation Details Progress on Primary Care First Model

Dec. 20, 2022
Mathematica’s evaluation says preliminary data suggests most practices in Cohort 1 will meet benchmarks for quality measures on diabetes control, high blood pressure control, and colorectal cancer screening

An evaluation of the first year of the CMS Innovation Center’s Primary Care First (PCF) model found that by using care management strategies, practices are seeking to reduce readmissions through improved post-discharge follow-up and fewer preventable hospitalizations among patient subgroups with complex conditions.

A preliminary review of data suggests most practices in Cohort 1 of the model will meet benchmarks for quality measures related to diabetes control, high blood pressure control, and colorectal cancer screening.

The evaluation report describes the basics of the model: The goals of the PCF model are to improve care for Medicare fee-for-service beneficiaries and to lower costs for CMS. As the successor to the Comprehensive Primary Care Plus (CPC+) Model, the voluntary PCF model builds on lessons learned, but differs substantially from CPC+, because it requires participating primary care practices from the outset to take on upside and downside financial risk for the most common primary care services for their attributed Medicare FFS population. Similar to CPC+, CMS partners with Medicaid and commercial payers that offer a payment model that aligns with PCF to their insured populations.

At the heart of PCF is a payment structure intended to reward quality and value and reduce administrative burden. CMS assigns primary care practices participating in PCF to one of four risk groups based on the average Hierarchical Condition Category (HCC) score among the Medicare FFS beneficiaries they serve, as determined by an attribution process.

As of late 2022, the PCF model has enrolled two cohorts of practices: the first began their participation in the model in 2021, and the second began their participation in the model in 2022. Each has a five-year period of performance.

CMS contracted with Mathematica to evaluate the PCF model. Mathematica will assess whether the model reduces hospitalizations and total Medicare expenditures for Medicare FFS beneficiaries served by PCF practices. Mathematica will also examine the changes PCF practices made to their care delivery and the model’s effects on other intermediate outcomes and quality-of-care measures.

Mathematica researchers noted that with PCF, CMS is being less prescriptive in how its requirements are met than in the CPC+ model. This flexibility is attractive to model participants, but it might not inform policymakers about why the model ultimately is or is not effective if strategies vary considerably across practices, the evaluators said.

Because it is too early to expect effects from the model to emerge on lowered hospitalization rates or cost, the report’s focus is on advanced primary care attributes that Cohort 1 practices report they possessed at the start of PCF and the approaches these practices have taken or plan to take to change how they deliver advanced primary care.

The evaluation team focused on the strategies PCF practices undertook to reduce hospitalizations or lower costs, with an emphasis on identifying whether any of these strategies were new to the practices.

Here are a few bullet-point highlights from the first annual report:

• When the model launched in 2021, 846 practices were participating, representing more than 4,000 practitioners and just over 500,000 attributed Medicare FFS beneficiaries. Nearly 85 percent of these practices were affiliated with at least one other PCF participant, such as belonging to the same health system or medical group, according to their applications. Participation increased further after CPC+ ended and CPC+ practices became eligible to join PCF: of the 2,228 practices that joined Cohort 2 in January 2022, more than two-thirds had participated in CPC+. The total number of participants falls short of the 8,000 practices that CMS anticipated would participate in the model. A factor that might have affected the number of practices that applied to the PCF model was the anticipated launch of the Global and Professional Direct Contracting Model, now known as the Accountable Care Organization Realizing Equity, Access, and Community Health [ACO REACH] Model.

• Among Cohort 1 practices, respondents reported they participated in PCF to be at the forefront of care transformation and to improve quality of care. Respondents also appreciated the predictability of revenue from the population-based payments (PBPs). CMS paid more than $190 million in 2021 to PCF Cohort 1 practices for professional PBPs across all risk groups, most of which (83 percent) it paid to practices in risk group 1 that make up the largest proportion of PCF Cohort 1 practices. Median annual PBPs ranged from roughly $144,000 for practices in risk group 1 to nearly $767,000 for practices in risk group 4. Within a risk group, PBPs varied depending on attributed number of beneficiaries. Average annual flat visit fee payments per practice were roughly $60,000 among practices in all risk groups, ranging from about $600 to more than $700,000.

• Although respondents from a few practices—including those that withdrew from the model— expressed disappointment about not being in a higher-paying risk group, the general sentiment was that PCF payments were adequate if not higher than expected. Simulations by the evaluation team suggest that total payments in PCF—adjusted for leakage—are higher than expected Medicare FFS payments if practices had not joined the model—even though CMS intended for the payments to be about the same as remaining in traditional FFS. But interview respondents expressed concern about how payments might change after the performance-based adjustment is applied or leakage adjustments are made. These performance adjustments began in April 2022 when one-fifth of practices received a negative performance-based adjustment (averaging $6,813 per practice) and more than one-third of practices received a positive adjustment (averaging $14,266 per practice). Leakage adjustments began in July 2022; on average, practices received a 34 percent (median: 31 percent) decrease in their population- based payments because of leakage.

• Practices assigned to risk groups 3 and 4 had a somewhat different experience in PCF. These practices often treated homebound patients and were already set up to function as high-touch interdisciplinary teams. Respondents from many of these practices noted that PCF funding allowed them to increase staffing to better meet their patients’ needs. In addition, risk group 3 and 4 practices reported they were more frequently reviewing advance care plans with patients and finding ways to better use data and data analytics to identify patients who need additional services.

This report also synthesizes findings on the 13 payers that are partnering with CMS as payer partners. These findings include details on the characteristics of these payers and why they chose to partner with CMS, supplemented with findings from interviews with payers that did not partner with CMS in the PCF model. Additionally, the report describes payer partners’ efforts to align their payments and other strategies, such as data feedback and quality measures with CMS in the PCF model.

Mathematica said that data from 2021 suggest that care management was one of practices’ leading strategies for reducing acute hospitalizations. Future data collection will help evaluators refine their understanding of the strategies that practices undertake and describe how practices intend for these strategies to result in changes in short-term and long-term outcomes.