Primary care practices participating in alternative payment models categorized as hybrid with no downside risk — option 2 of the NASEM framework — showed significant improvement in all measures of accessibility, comprehensiveness, continuity, and coordination of care.19
We had more nuanced findings for models under option 3, involving population-based payments for primary care combined with financial risk. For PCPs in the downside-risk ACO models that relied in part on traditional physician fee schedule payments, we saw significant improvements in measures of access, comprehensiveness, and coordination, as well as the most direct measure of PCP continuity. However, care fragmentation did not significantly improve. Participation in hybrid payment APMs with downside risk and global payment APMs was also not associated with improvement in our care coordination measure.20
Finally, outside the NASEM framework, the Next Generation ACO model (the entity-elected payment option in Exhibit 1) had no significant link to improvements in coordination of care, and fee-for-service, upside-risk-only models were not significantly associated with improvements in new-problem management (an aspect of comprehensiveness), fragmentation, or coordination.21
Discussion
Our findings are consistent with the NASEM report’s recommendation to pay small primary care teams to care for patients rather than providing visit-based payments to individual primary care practitioners. But who to pay to support these teams and how to do so remain in question. In a companion paper, we found inconsistent associations between PCP health system affiliation and the core features of primary care.22 Prior work has found PCPs affiliated with health systems had lower comprehensiveness and continuity of care.23 Data from the Agency for Healthcare Research and Quality Compendium of U.S. Health Systems show that from 2016 to 2022, the proportion of primary care physicians in these systems has grown from 43 percent to 53 percent.24 There also has been rapid growth of primary care practice ownership by larger organizations, such as health plans and other investor-owned corporations.25 In recent interviews on payment reform, PCPs emphasize the importance of directing payments and related resources to the primary care practices responsible for delivering care.26
PCPs participating in the subset of Medicare APMs that relied on hybrid payment without introducing financial risk had significant improvements in all measures of the four key features of primary care. This is compatible with arguments by some primary care payment policy leaders that effective payment reform need not be (and perhaps should not be) tied to downside financial risk.27 In our analysis, other payment model categories had less consistent associations with PCPs’ achievement of all the core features of primary care. Of course, many of these models engage primary care practices affiliated with larger organizations, which could compromise the effective allocation of resources to enhance the provision of high-quality primary care.
The NASEM report also called out the potential importance of enhancing fee-for-service payments to primary care and increasing the proportion of health care spending devoted to primary care. Unfortunately, we could not assess any payment models focused solely on increasing fee-schedule payments for PCPs, as such reforms would be beyond the statutory scope of traditional Medicare APM. Likewise, we did not observe any models that reliably enhanced the total amount of payment to PCPs; the Center for Medicare and Medicaid Innovation has faced statutory constraints developing such models and has only recently begun to test targeted examples (for example, ACO REACH).28
PCPs serving a higher proportion of Medicare beneficiaries with greater social risk were less likely to participate in the traditional Medicare APMs we studied. The numbers were too small to investigate the potential benefits — to practices and patients — of the various NASEM report categories of primary care payment reform.
Like all observational studies, ours had methodological limitations. We attempted to account for selection bias by controlling for observable PCP, practice, beneficiary, and community characteristics, but we cannot control for unobserved characteristics, such as a practice’s predisposition to improve performance on primary care features. Our forthcoming article29 notes various efforts to address these limitations, but we acknowledge these methodological concerns preclude confirming a causal relationship between specific APM participation and improvements in these essential features of primary care.
Conclusion
We find overall that traditional Medicare alternative payment models are associated with improvements in PCP accessibility, comprehensiveness, continuity, and coordination. Applying a framework derived from NASEM recommendations, we find that the most consistent benefits were observed for hybrid payment models that do not subject PCPs to downside financial risk. It is possible that reforming PCP payment to enhance delivery of the essential features of primary care may not be compatible with holding PCPs at risk for costs of care.
We also find signals that models are more effective at improving primary care when the resources are directed to primary care practices instead of larger entities like ACOs. Indeed, disentangling PCP finances and rewards from those of other providers and larger organizations may be essential to restoring primary care to its core role in health care delivery. Increasing payments or program supports to primary care practices will also likely be needed, especially to those who serve patient populations with greater social risks and resource constraints.
How We Conducted This Study
Classifying Traditional Medicare Alternative Payment Models
We reviewed the Centers for Medicare and Medicaid Services (CMS) descriptions of each of the 14 traditional Medicare alternative payment models (APMs) studied and classified them according to the options for primary care payment reform presented in the 2021 National Academies of Sciences, Engineering, and Medicine (NASEM) report.
Analyzing PCP Participation in APMs
Through our access to the CMS Virtual Research Data Center (VRDC), we used Medicare Data on Provider Practice and Specialty (MD-PPAS) to identify physicians with primary care specialties during 2016–2022; Medicare fee-for-service claims and enrollment data to create claims-based outcome measures and obtain beneficiary characteristics for beneficiaries seen by those primary care practitioners (PCPs); and data on physician participation in various CMS initiatives. We relied on additional publicly available secondary data sources to obtain physician practice and area characteristics. For outcome measures, we used previously validated claims-based outcomes of PCP comprehensiveness of care, PCP continuity of care, and PCP care coordination and examined the extent of PCPs’ first-contact care for common problems as a measure of PCP access.30 We examined outcomes across the full range of traditional Medicare patients served by a PCP in each year.
Assessing the Association Between APM Participation and PCP Outcomes
We used combinations of Taxpayer Identification Numbers (TINs) and National Provider Identifiers (NPIs) to determine traditional Medicare APM participation in the 2016 baseline year as well as during the period 2017 to 2022. Using a sample of PCPs (TIN-NPI combinations) not participating in any traditional Medicare APM in 2016, we employed a difference-in-differences analysis to compare changes from baseline to the intervention period in PCP-level outcomes among PCPs who began participating in traditional Medicare APMs during 2017–2022 versus those who did not.
To account for potential differences in characteristics between PCPs who participated in specific APMs versus those who did not, we controlled for physicians’ characteristics, including their participation in other Center for Medicare and Medicaid Innovation models, attributes of their beneficiaries, affiliation with a health system, and area-level characteristics. Tests of parallel trends showed that PCPs entering APMs in 2019 were differentially improving on two measures of comprehensiveness and both continuity measures, so we avoid drawing causal conclusions. To further mitigate unmeasured confounding, we used practice-level fixed effects in all models. Additional details can be found in our full article.31
