CURRENT DATA INPUTS AND PROCESS USED BY THE CENTER FOR MEDICARE TO VALUATE PRIMARY CARE IN THE PHYSICIAN FEE SCHEDULE

The PFS specifies payments for physicians and other clinicians participating in Medicare Part B, including professional fees, diagnostic tests, and radiology services. Nearly all of these fees are for single services (i.e., fee for service [FFS]), while a very small proportion are for care bundles (CMS, 2024a). By law, payments must be for “services furnished” and cannot be only paid to a specific specialty. Because the PFS must be budget neutral by law, if CMS decides to increase the value of a PFS service code, offsetting savings must be achieved by reducing the value of other services (NASEM, 2021).

Currently, the Resource-Based Relative Value Scale (RBRVS) is the predominant mechanism used by CMS to translate data into recommendations for updates to the PFS. The American Medical Association’s Relative Value Scale Update Committee (the RUC) was established to offer annual recommendations to CMS on PFS changes (the RBRVS and the RUC are described in greater detail below). While not required to do so, CMS typically accepts between 85 and 90 percent of the RUC’s recommendations each year (Laugesen et al., 2012; Moore, 2023). These recommendations greatly affect how physicians are compensated for their work, as the PFS determines not only what CMS pays physicians through Medicare, but also what physicians are paid by the majority of state Medicaid programs and commercial payers (including Medicare Advantage plans) that model their fee schedules on the CMS PFS. Beyond influencing payment rates by government and private payers alike, relative value units (RVUs) are frequently used to monitor productivity and serve as the basis for many alternative, or value-based, payment models as the source of benchmarks or total cost targets, amplifying the broad effect of these valuations on “pricing, profitability and care patterns” (Klepper, 2013; Laugesen, 2014).

RECENT PHYSICIAN FEE SCHEDULE CHANGES IN SUPPORT OF PRIMARY CARE

In recent years, CMS has made changes to the PFS to broaden the scope of valuated services and activities and to increase payments for services provided in primary care (AMA, 2024a). However, these changes have only partially addressed this issue of imbalanced and inadequate capture of primary care services, because of both limited and uneven uptake and other challenges described below.

In 2011, Medicare introduced annual wellness visits that are reimbursed at higher rates than comparable evaluation and management visits and designed to allow primary care clinicians to deliver evidence-based preventive and aging-related care that can strengthen primary care relationships (Davis et al., 2011; Ganguli et al., 2018). However, uptake of the visit has been slow (Ganguli et al., 2017, 2018; Lind et al., 2019; Misra and Lloyd, 2019), and there was disproportionate adoption by already well-resourced practices serving fewer marginalized patients (Ganguli et al., 2017, 2018). In 2013 and 2015, respectively, Medicare introduced transitional care management (TCM) and chronic care management (CCM) codes to incentivize and support evidence-based patient care activities that occur outside of typical office visits, and that primary care teams often already performed without reimbursement (Bindman and Cox, 2018; Jackson et al., 2013; Kim et al., 2025). But uptake of TCM and CCM has been limited as well (Agarwal et al., 2018). Between 2015 and 2019 there was only an increase from 1.1 percent to 3.4 percent of beneficiaries with two or more chronic conditions receiving CCM services (Jang et al., 2024). More recently in 2019, 7,729 practices out of 177,220 (5.2 percent) who billed qualifying evaluation and management visits had at least one attributed beneficiary who received CCM services (Colligan et al., 2022). While TCM services have been associated with reduced readmission rates, spending, and even mortality, they have similar low uptake with an increase in use from 3.1 percent of eligible beneficiaries in 2013 to under 10 percent of eligible beneficiaries in 2022 (Bindman and Cox, 2018; Burdick et al., 2022; Kim et al., 2025). A 2024 study showed a slight increase from those numbers, with 11.3 percent of eligible beneficiaries receiving TCM services (Anderson et al., 2024).

Multiple barriers may account for the poor uptake of these codes including low payment rates relative to the cost of implementing and maintaining these services (Agarwal et al., 2018), complicated requirements for beneficiary consent and tracking of service time, challenges with training staff and allocating resources to support services (Agarwal et al., 2018), and the effect of cost-sharing on patient acceptance of services for TCM and CCM specifically (O’Malley et al., 2017; Reddy et al., 2020). Because the work included under TCM and CCM codes is widely viewed as standard (if not always completed) aspects of primary care delivery, carving out these codes for measurement, billing, and cost-sharing may create cognitive dissonance for many of the primary care clinicians being asked to bill them, further contributing to underuse.

In 2021, based on a recommendation from the RUC, CMS increased the RVU values for office-based evaluation and management service codes while also easing documentation requirements for these codes, such that more visits qualified for higher service codes (Neprash et al., 2023). However, this PFS change has had only a modest effect at decreasing the payment gap between primary care and other specialties owing to such factors as a reduction in the conversion factor to achieve budget neutrality (described above), the fact that many physicians other than those in primary care also benefited from increased evaluation and management code rate increases (Neprash et al., 2023), and inconsistent adoption by private health plans and health systems of this change for their own payment policies.

In 2024, CMS introduced the add-on code G2211 that may be billed in tandem with an office visit code to represent the complexity inherent in continuity-of-care visits; the G2211 code has an RVU of 0.49, corresponding to a national Medicare allowable fee in 2024 of about $16, and may be billed by any clinician who has a longitudinal relationship with the patient. While evaluations of the uptake and potential impact of this code have not yet been published, there is early anecdotal evidence of billing beyond its stated intent. For instance, societies for specialties not traditionally associated with comprehensive, longitudinal care (e.g., ophthalmology, physical medicine and rehabilitation, urology) have issued formal guidance to its members on using the code (American Urological Association, 2024; AOA Coding & Reimbursement Committee, 2024).

Most recently, the CMS CY2025 fee schedule introduced several changes to support high-quality primary care (CMS, 2024a) including APCM codes. These three new Healthcare Common Procedure Coding System G-codes are tiered by patient medical and social complexity, allowed to be billed as often as monthly, and meant to capture a range of primary care activities, including care management and asynchronous care, while simplifying billing and documentation for care management services. While APCM requires that one clinician (physician or advanced practice provider) can bill for APCM services, the intent is for the payments to cover the activities of the full interprofessional primary care team under the billing clinician’s supervision. Level 1 APCM services are for patients with one or fewer chronic conditions and are valued at 0.17 RVUs (total reimbursement about $15). Level 2 APCM services are for patients with two or more chronic conditions and are valued at 0.77 RVUs (total reimbursement about $50). Level 3 APCM services are for qualified Medicare beneficiaries with two or more chronic conditions and are valued at 1.67 RVUs (total reimbursement around $110) (CMS, 2024a). The method used by CMS to determine the valuation of the APCM codes was based on an assessment of the predecessor CCM codes, including data on time-based billing and frequency of bills.

APCM includes 10 service elements:

1.

patient consent,

2.

an initiating visit for new patients,

3.

24/7 access to care and care continuity,

4.

comprehensive care management,

5.

a patient-centered comprehensive care plan,

6.

management of care transitions,

7.

practitioner, home-based, and community-based care coordination,

8.

enhanced communication opportunities,

9.

patient population-level management, and

10.

performance measurement.

These APCM service elements are critical components of high-quality and advanced primary care that for the most part were not billable under the PFS prior to introduction of these new codes.

In their Request for Information about the program, CMS expressed an interest in understanding the implications of expanding the APCM payments to include additional bundled advanced primary care services (CMS, 2024b). The National Academies, in its 2024 consensus report Response to the Centers for Medicare & Medicaid Services CY 2025 Advanced Primary Care Hybrid Payment Request for Information, suggested that additional services such as the annual wellness visit, some behavioral health and social needs care, and additional asynchronous care and population health services could be added to future payments (NASEM, 2024a). That report also recommended reassessing the appropriate valuation of APCM codes, especially if additional service elements are incorporated into these codes.

The report also raised concern about whether the beneficiary cost-sharing that CMS requires for billed APCM services will constitute a large barrier to their uptake. Medicare beneficiaries must individually consent to their clinician billing for APCM services and be informed during the consent process that they must pay a share of the allowed fee. The share of cost may be a major impediment, especially for billing on the maximum allowed frequency of once every month, resulting in the actual additional revenues to a primary care practice falling far short of the optimal amount that would be generated by full participation among all eligible beneficiaries for billing for APCM on a monthly basis. The report recommended that CMS waive cost-sharing, while acknowledging that congressional authorization would likely be required to do so.

In sum, while CMS has taken several steps to improve how primary care is valued through new billing codes, the uptake has been limited, or for new codes, it is too early to tell. Further, these steps have not fully incorporated the work of primary care teams and nonvisit-based care delivery into the compendium of services that are valued. These changes do not appear to have shifted a sufficient share of physician payment to primary care to adequately fund this critical sector. National data indicate that primary care investment as a percentage of total health care expenditures is decreasing rather than increasing (Jabbarpour et al., 2024). This committee therefore believes that the methodology for valuation of services for the PFS should continue to be revised and carefully reconsidered. The next section suggests new data and methods that could be used to better inform valuation decisions by CMS.

NEW CONSIDERATIONS FOR THE VALUATION OF PRIMARY CARE

Ideally, CMS’ valuation of primary care should include the full scope of work performed by primary care physicians and other primary care team members within synchronous and asynchronous patient encounters as well as in nonpatient-facing activities in support of high-quality primary care. This valuation process requires defining high-quality primary care, identifying which services and activities to measure, identifying a sample in which to assess team configuration so costs can be assigned, and optimizing variation in that sample on attributes such as geographic location, rurality, size, ownership, and characteristics of patients served. Then methods must be deployed to collect data to identify the workforce configurations (types of professionals, and clinical full time equivalent) and the time and cost to do this work. It is worth noting that, while this report is focused on primary care valuation specifically, these new considerations and valuation methods and processes could theoretically apply to all specialties.

New Data Sources and Methodologies

Given the limitation of current valuation methods, as described in this report, diversifying data sources to inform the valuation of health care services would enhance the accuracy, generalizability, and comprehensiveness of payment rate determinations. Ideally, new data sources should capture a fuller scope of primary care services, activities, and cognitive work, and the contributions of the entire care team. And they should do so accurately and from a broad range of practice settings representative of the diversity of primary care settings, with minimal financial cost and a low measurement burden for the individuals performing the work and those extracting the data. Data sources should be transparent, be reproducible, and include the diverse perspectives and values of all those delivering and receiving health care. CMS can consider new data sources for updates or changes to the PFS if they are empirically collected survey or observational data, provided in consultation with physician organizations, and involve a RBRVS-compliant process to receive data per PFS statute. The committee also notes the “chicken-and-egg” problem that is unavoidable in this discussion: any approach to valuate primary care work by capturing the time and effort currently spent on primary care, in the context of current capacity and time constraints and underinvestment, will not reflect the time and effort that may be truly needed to provide high-quality primary care.

Better Survey Data

Broad-based surveys with representation from the full scope of primary care clinicians across the country can provide data that are more complete, actionable, relevant, and useful to consider when updating the Physician Fee Schedule (PFS) (Green Center, 2025). For example, internet or email-based surveys through professional groups offer a mechanism for lower-cost, higher-frequency sampling of key practical questions regarding the intensity, scope, effort, and cost of services provided through primary care. While no survey can be simultaneously cheap, fast, and fully generalizable, the recent experience of mobilizing near-real-time data on primary care practice during and after the acute phase of the COVID-19 pandemic is illustrative (Sullivan et al., 2023).

A network of researchers led by the Larry Green Center at Virginia Commonwealth University distributed a series of online, short-form surveys across a range of policy-relevant primary care practice questions, totaling over 32,000 responses from over 8,100 clinicians representing every state in the United States. While imperfectly representative of the entire scope of primary care practices, this survey sample nonetheless generated quick turnaround data snapshots (Green Center, 2021) from an unusually wide array of clinicians on practice closure, finances, personal protective equipment availability, quality initiatives, and clinician burnout (Sullivan et al., 2023).

Building on this effort, recent surveys from this team focused on primary care clinician perceptions nationwide on the adequacy of current payment with respect to the time, intensity, and presence of intervisit work (Green Center, 2025). A total of 417 survey responses were received from clinicians in 46 states, of which 85 percent were M.D.s or D.O.s, 58 percent female, 75 percent family medicine, 22 percent rural or frontier, 27 percent self-employed, and 23 percent from community health centers or FQHCs. The first section of the survey asked about estimated time needed for additional core components of primary care during an adult well-controlled diabetes follow-up visit, billed as a level 3 CPT code (which is roughly estimated to take 25 minutes) (Green Center, 2025). Notable levels of effort and time were needed to address common scenarios or requirements during these primary care visits, ranging from a mean of 6–26 minutes each, depending on the scenario.

For example, respondents estimated that adding in effort for addressing two clinical preventive services for which the patient is eligible (a common and assumed set of tasks in routine primary care) would take a mean of 7 additional minutes on top of the 25 minutes allotted for the common visit type and reason above. Similarly, surveyed clinicians reported attending to a mental health concern that became apparent during the visit would take an additional mean of 15 minutes, while responding to a social driver of health salient to the patient’s condition would take an average of an extra 10 minutes. Reviewing outside records or taking care of other care coordination tasks related to the visit problem would also take an estimated average of 10 extra minutes each.

In a second group of questions summarized in Table 1, respondents reported an aggregate range of 2–3 additional hours of work per day (mean 176 minutes; median 120 minutes) on primary care tasks outside of face-to-face visits, and not including notes and visit documentation. When asked to prioritize the importance of these nonvisit tasks for patient care, respondents ranked answering patient messages in a timely manner and coordinating care as the most crucial activities they engage in on a daily basis (Green Center, 2025).

TABLE 1

Unreimbursed Primary Care Activities.

Surveys such as these do have a risk of recall bias, and clinicians may intentionally or unintentionally overestimate the time required for activities, but they can allow capture of the full scope of primary care work across team members, are relatively low burden, and can be made transparent. These survey results provide an additional proof of concept that current valuation approaches likely under-capture the work of high-value primary care and begin to quantify adjustments needed to capture this work.

Qualitative Data

Observational studies and semi-structured interviews with clinical leaders, and with clinical leaders and personnel carrying out the tasks, can help to identify primary care functions—like care coordination, which are achieved through specific activities such as team huddles, hallway consults, and multiple communications with patients, care givers, clinicians, and insurers (Cohen et al., 2015, 2022; Coleman et al., 2016; Davis et al., 2015; Friedberg et al., 2020; Patel et al., 2013). Approaches that include some type of observation, even self-observation, may yield more accurate estimates than surveys, particularly for frustrating tasks where bias might affect estimates. While these data may be more difficult to quantify for generating a dollar value, they do provide critical insights as part of a broader valuation process.

Another key issue is sampling. In qualitative data collection, sample sizes tend to be smaller to offset the effort involved in data collection that is richer and possibly also more internally valid. Purposive, iterative sampling and sampling to saturation—common principles in qualitive methods—can ensure transferability of findings (Cohen and Crabtree, 2008). Yet researchers still need to identify a sample of practices that are fully staffed to deliver high-quality, advanced primary care. This can be challenging given the current primary care environment (Jabbarpour et al., 2024). Data collection will require asking practice leaders and clinical team members about their current staffing and what work they need to do but never get to because they are short-staffed in order to identify ideal staffing configurations for primary care valuation. Such work needs to agree on a definition of high-quality primary care. On this point, there is growing consensus and alignment among the National Academies definition of high-quality primary care (NASEM, 2021), and other models such as Bodenheimer’s 10 building blocks of high-performing primary care teams and Wagner’s functions of high-performing safety net medical homes (Bodenheimer et al., 2014; Wagner et al., 2014). However, there are challenges in measuring those qualities and using those measures to identify, select, and recruit practices whose care meets those definitions, and it may be difficult to observe all of the required elements of high-quality primary care when observing care.

Direct Observation and Time-Motion Data

Time-motion data represent the quantification of the duration and movements required to complete a task (Zheng et al., 2011). In the context of health care, time-motion data have been used to estimate the time involved in delivering ambulatory care (Young et al., 2018) and to examine changes in workflows such as following the introduction of EHRs or other innovations (Sinsky et al., 2016; Zheng et al., 2011). As such, time-motion data collection in the primary care setting could supply data for new methodologies to valuate primary care service delivery (Young et al., 2018). For example, time-motion data could provide data on team-based tasks such as coordination and consultation across roles (e.g., pharmacy, nursing, and behavioral health), or the administrative work involved in confirming prior authorizations and referrals.

In conventional time-motion analyses, data are generated by human observers (which could be members of the team itself or outside observers) who document the time spent on distinct tasks according to their observations. While these observational data could improve the precision of time estimates, there are some disadvantages to this approach. The reliance on human effort (1) limits the validity of findings because of the potential for interrater and subject variability, and (2) increases the costs and burden involved in ensuring representativeness across diverse practice settings as well as updating data as practice patterns evolve. As such, time-motion data may become outdated when health care innovations, particularly those targeting efficiency or requiring the performance of new tasks, emerge and permeate practice settings (Lopetegui et al., 2014; Zheng et al., 2011). Rapid advances in AI technology combined with video and sensor technology may mitigate these limitations by reducing the effort involved in repeat time and motion data collection as the tasks performed by clinicians evolve with changing technology and practice patterns (Deeds et al., 2025; Hernández et al., 2021; Rotenstein et al., 2023).

Electronic Health Record System-Event Log Data

The widespread adoption of EHRs provides an opportunity for collecting data to determine the relative value and resources of primary care services for setting payment rates with no added burden to clinicians. The Trusted Exchange Framework and Common Agreement (TEFCA) established as part of the 21st Century Cures Act operates in the United States as the nationwide framework for electronic health information sharing across the health care ecosystem. The TEFCA framework has provided the foundation for the data infrastructure that could support the use of EHR data to guide primary care reimbursement rates (HealthIT.gov, 2025). For example, EHR data extracts could inform patterns of time spent by primary care team members on treatment pathways, preventative care strategies, and management of chronic diseases. In the United States, every certified EHR vendor is required to maintain data logs that track who accesses patient records, when they do so, and their actions within the record (Sinsky et al., 2020). The data from these EHR data logs can be used to describe the activities that different types of primary care team members are doing during a particular time period.

Data extracts can also provide information on the time spent working on tasks in the EHR outside of scheduled patient hours. Specific tasks that can be captured in data logs include documentation, time spent in the patient portal (i.e., sending and receiving physician, staff, and system messages), time spent on clinical review, and time spent placing orders (Rotenstein et al., 2022). Demonstrated consistency between directly observed findings, physician self-report of EHR work, and EHR log data supports consideration of user-event log data as a tool for assessing EHR activities associated with both direct patient care and asynchronous work during and after clinic hours (Arndt et al., 2017).

There are several implementation challenges to the use of electronic health information data including the proliferation of application types (e.g., EHRs, personal health records, Health Information Exchanges, population health platforms, etc.) that contain patient information that can provide full context of a clinical primary care encounter (Walker et al., 2023). The infrastructure to support these applications can be difficult and costly to maintain due to lack of a common data structure resulting in too many places for data to be located, creating interoperability inefficiencies (Walker et al., 2023). Beyond infrastructure challenges, there are also several limitations of EHR data extracts. Many data workflows within the EHR do not fit with the natural flow of longitudinal primary care clinical encounters, such that tasks like care coordination and communication, which are often documented within unstructured clinical narratives, may be missed. Time alone does not capture all the dimensions of those clinical activities, including the appropriateness of the activity tasks and the direct mapping of the time spent on a type of EHR screen to that actual total time required to complete that specific task (Sinsky et al., 2020). In clinical settings, members of the primary care team multitask (e.g., write notes while also calling a patient) resulting in many of these multiple tasks performed not being captured in an EHR audit log. Not all primary care team members use the EHR. Thus, the reliance of EHR audit log data alone to determine the relative value of primary care may lead to the undercapturing of the clinical work provided. Standardization of EHR time stamp-based measures and a national regulatory pathway for EHR vendors to make more granular EHR data widely available may help make EHRs a more reliable source of primary care valuation data (Adler-Milstein, 2024).

Artificial Intelligence and Large Language Models

Artificial intelligence (AI) and large language models may provide an important way to make sense of EHR and other data.⁷ Specifically, there is growing interest in how AI systems can organize and interpret the large amounts of unstructured data in EHR documentation. Neural network architectures used in natural language processing (NLP)⁸ AI systems have shown promise as a solution to efficiently extract and classify unstructured clinical narratives used in EHRs (Park et al., 2019; Pyne et al., 2023). These systems also have great potential to modernize risk adjustment models used to predict health care spend and the allocation of resources. Complex neural network architectures could improve the risk adjustment models used in Medicare by reducing statistical bias used in the predictions, reduce errors due to variance, and allow for closer calibration of models to the task of risk adjustment (Weissman and Joynt Maddox, 2023). More recently, transformer architectures have provided a self-attention mechanism to AI supporting the training of these systems on a massive corpus of text resulting in large language models (LLMs), which have performed remarkably better on NLP benchmark tasks of recognition, relation extraction, sentence similarity, natural language inference, and question answering (Omiye et al., 2024; Yang et al., 2022).

LLM digital scribes deployed in primary care have been effective in capturing the breadth of the important clinical details discussed during a clinical encounter while reducing the time spent by the provider documenting that encounter (Bundy et al., 2024). LLMs have also demonstrated promise in extracting information within patient communication that occur outside of a clinical encounter (Yan et al., 2024). Finally, LLMs have the potential to overcome previous manual chart abstraction methods of unstructured EHR data (Ge et al., 2023). Using LLMs for chart abstraction could offer a promising way to accurately capture and report on complex primary care quality measures at scale (Boussina et al., 2024).

LLMs could also be used to inform updated PFS payment rates related to primary care, including APCM. Comprehensiveness is a key indicator of the relative value of primary care service, with more comprehensiveness associated with less care fragmentation, better health outcomes, and lower cost (O’Malley et al., 2019). Medicare claims-based measures have traditionally been used to measure comprehensiveness but are limited in identifying the degree to which the primary care clinician is truly managing the condition listed on the claim (O’Malley and Rich, 2015). EHR data within clinical notes, perhaps derived from ambient documentation systems using AI LLM models, would optimally be included when measuring comprehensiveness but are often not included owing to burdensome challenges with EHR chart abstraction (O’Malley et al., 2019).

Time-Driven Activity-Based Costing

Time-driven activity-based costing (TDABC) is a methodology for calculating the detailed costs of a process or activity (Cidav et al., 2020; Gritz, 2024; Najjar et al., 2017). In health care, TDABC has been proposed as an alternative to cost-to-charge ratios or RVU-based costing, which are often inaccurate and poorly correlated with the costs of inputs (Kaplan and Porter, 2011). While RVU-based costing restricts input estimation to reimbursable services and procedures (Kaplan and Porter, 2011), TDABC is a “bottom up” approach based on the actual processes involved in patient care (Najjar et al., 2017). The approach relies on two core components: time and the cost per time unit for each resource contributing to the process or activity under study. Costs are then summed across each contributing resource according to time required. These resources often include personnel but can also include the equipment and space required to deliver care.

While TDABC has often been used to estimate the costs for an episode of care, such as completion of mammography screening, TDABC can also be applied to a process, such as a clinic visit, or an entity such as a clinic site (Choudhery et al., 2021). TDABC feasibility has been examined with respect to surgery, radiology, chronic disease management, and outpatient services more broadly (Choudhery et al., 2021; Doyle et al., 2022; Leung et al., 2013; Najjar et al., 2017).

As a methodology for primary care valuation, TDABC offers unique advantages. For example, TDABC can estimate the costs associated with staffing diverse professional roles, administrative tasks, and clinical activities (Choudhery et al., 2021; Leung et al., 2013). As a result of this flexibility, TDABC can accommodate heterogeneity of clinical scenarios and services offered in primary care settings such as procedures, behavioral health, chronic disease management, group visits, and team-based care. Of note, direct observation has often been used to map process inputs, which raises questions of feasibility given the heterogeneity and complexity of services contained within the primary care domain. However, other data sources could theoretically be used including EHR audit log data and survey data. In particular, leveraging EHR audit log data could present a novel low-burden data source informing the cost of a wide variety of primary care activities, including capturing team-based approaches. To this point, rough estimation is usually considered sufficient, reducing the administrative burden of TDABC (Cidav et al., 2020; Leung et al., 2013). Finally, TDABC lends itself to cost simulation such as estimating the expected costs of providing new services, expanding interprofessional teams, or implementing new evidence-based practices (Cidav et al., 2020). As such, TDABC can not only support the estimation of current costs, but it can also estimate the expected costs following practice innovations. In this way, TDABC can incentivize cost-effective and value-based care (Keel et al., 2017).

Simulation and Modeling

Microsimulation is a modeling tool that can help guide health care decision making. Microsimulation models are developed to synthesize evidence from different sources to demonstrate the implications of certain decisions (Krijkamp et al., 2018). Microsimulation brings together diverse data sources and methods to simulate, for example, the effect of health care policies and programs, to predict health care needs of a particular group (e.g., older adults), and to evaluate the benefits and harms of treatment (Goldman et al., 2004; Hennessy et al., 2015; Mühlberger et al., 2015; Statistics Canada, 2024; Wolfson, 1994). Microsimulation has been used in primary care to evaluate the effects of new funding models for patient-centered medical homes on primary care finances and to estimate the effect of team configuration and team stability on primary care quality (Basu et al., 2016; Hysong et al., 2019). Microsimulation holds promise as an approach for simulating and estimating the valuation of primary care at both a small practice and larger population level. This approach incorporates and integrates a variety of datasets ranging from claims data to workforce estimates, payment rates, and sociodemographic and epidemiological data.

While simulations are always limited by the quality of the data inputs, simulation could reduce the number of primary care practices needed in a valuation sample. Additionally, by drawing on multiple data sources for estimates, methods optimization could be achieved. For instance, EHR data, which are widely available and lower cost to collect on a large scale, could be used to estimate appropriate variables (e.g., number of visits, number of portal messages, patient characteristics) in a valuation model. This could strategically be blended with self-report data collected via survey or LLM models to assess the effect of new PFS payment changes on practice-level finances. Models could make good use of limited observation and semi-structured interview data collected with an intentional focus on particular professional roles and functions. This could greatly improve the internal validity of primary care valuation, and models could simulate costs for different patient populations, for different types of practices, and presumably for different models of primary care, with the right inputs.

HOW THESE INPUTS COULD TRANSLATE INTO HYBRID PRIMARY CARE PAYMENTS

The 2024 National Academies report Response to the Centers for Medicare & Medicaid Services CY 2025 Advanced Primary Care Hybrid Payment Request for Information addressed questions posed by CMS about implementing a hybrid payment model that would include both FFS and prospective payment components (NASEM, 2024a). That report recommended that CMS proceed with implementing a hybrid payment model.

At a conceptual level, this committee agrees with the logic of prospective payment as an important component of a primary care payment model. As noted above, considerable clinician effort is devoted to activities, such as care coordination and asynchronous patient interactions, that are not tightly associated with a specific billable encounter. Team-based services are often provided by unlicensed health workers who are not eligible to bill. A prospective, monthly capitation payment provides a method for compensating a practice for this comprehensive array of services. Details of such a model should be informed by evidence, such as a 2017 microsimulation study that suggested at least 63 percent of such payments need to come from capitated sources (versus FFS payments for the full range of primary care services) to shift practice behavior (Basu et al., 2017).

Although prospective payment has merit, it is important to consider whether and how moving to a hybrid payment (part prospective, part FFS) model might contribute to the goal of more accurate valuation of primary care. One barrier to hybrid payment is uncertainty about whether CMS has the statutory authority under existing law to implement payment methods other than FFS as a universal payment method under the PFS, rather than only having authority under the Affordable Care Act to do so on a more limited basis for Innovation Center demonstration programs. The 1989 OBRA amendment to the Social Security Act adding Section 1848(b)(1) to authorize a PFS specified that fees are to be paid for “services furnished.” Some have pointed out that some existing PFS payment schemes, such as global surgical fees, are not strictly tied to individual visits but incorporate a bundle of pre- and postoperative visits within a specified timeframe; these authors have suggested that this might serve as a precedent for a primary care prospective monthly payment (Berenson et al., 2023).

Because congressional action to further amend Section 1848(b)(1) may be required for CMS to implement fully a hybrid model with prospective payment in the PFS, it is important for CMS to consider strategies to ensure that FFS codes and valuation appropriately reflect the resources needed to deliver high-quality primary care in the meantime. Recent actions by CMS have in essence expanded the boundaries for adding billable service codes to the PFS to enhance the valuation of primary care services. The new APCM codes (described above) represent a bolder reform to allow primary care clinicians to bill for a comprehensive set of services provided in an ongoing manner and not linked to an associated evaluation and management visit. Although technically not a prospective payment, the APCM codes expand the limits of allowable FFS billing by uncoupling many of the services from visits, simplifying documentation requirements from prior CCM codes, and incorporating team-based services into the scope of the new codes. In principal, if CMS developed a new methodology for valuing primary care services that determined that the existing PFS undervalued primary care, it could “load” additional value onto existing billable PFS codes (e.g., increase the G2211 code value to reflect the full amount of longitudinal care effort outside the visit), as well as introduce additional billable codes to pay for more comprehensive primary care (as it did by introducing the APCM codes).

For enhanced valuation of FFS billing codes to accomplish the goal of greater investment in primary care specifically, it must overcome several impediments. If CMS substantially increased the relative value for evaluation and management office visits and G2211 codes, the agency would need an approach that only applied the higher value to codes billed by primary care clinicians (e.g., by instituting separate codes for primary care and nonprimary care clinicians) to intentionally focus the additional investment. Such tailoring of new codes to particular clinician groups (i.e., to only apply to primary care clinicians) is not currently acceptable within statute, and thus would require congressional action to amend the authorizing language for the PFS to allow such a bifurcation. Otherwise any outpatient clinicians could bill for these more well-reimbursed codes, inflating costs without targeting the extra spending toward primary care. The Response to the Centers for Medicare & Medicaid Services CY 2025 Advanced Primary Care Hybrid Payment Request for Information also expressed concern that beneficiary cost-sharing is likely to impede uptake of the APCM codes; as noted above, that report recommended that CMS, with congressional authority, should consider waiving cost-sharing for APCM services (and all high-quality primary care services) (NASEM, 2024a).

Should CMS move forward with implementing a hybrid payment model that includes a prospective payment component within the PFS, the agency would need to address several further issues in determining the valuation of the prospective payment. Foremost would be determining whether prospective payment rates should be adjusted based on a practice’s capacity for delivering high-quality primary care. For example, if a PMPM was intended to compensate the practice for the cost of team-based care, CMS might adjust the payment amount based on the extent of team care available at the practice and the quality of care provided. Stated another way, the valuation of the team-care prospective payment might depend on the capacity of the practice for delivering the desired services in an effective way.

As an informative example, Massachusetts Medicaid uses a three-tiered approach for primary care capitation payment in its accountable care organization (ACO) programs (NCDHHS, 2024a,b). The lowest tier requires elements such as routine screening of patients for behavioral health and social needs, and the highest tier requires elements such as integrated behavioral health and opiate use disorder treatment. An alternative to a tiered structure for prospective payment in a hybrid model would be using a single basic prospective payment rate and using FFS payment to incentivize services that are unambiguously high value (e.g., evidence-based vaccinations) or higher capacity, such as paying fees for fully integrated mental health or substance use services.

At the same time, as more robust, higher-paying PPMs are considered, their spillover effect on other CMS programs and priorities must be understood. Should an option for a high monthly payment inclusive of both visit and nonvisit activities be available to primary care practices, many may consider withdrawing participation from ACO contracts whose success is largely mediated by the performance and robustness of primary care activities. There is also the issue that because large prospective payments contribute to ACO spending benchmarks, primary care clinicians and practices within ACOs will face heightened competing demands of billing these newly available codes to generate FFS revenue on the one hand, and meeting these benchmarks to realize shared savings on the other hand. Because of these hard to predict spillover effects, and unclear uptake of these codes, it will be wise for CMS to carefully evaluate the first few years of uptake, code usage, and use within primary care to understand whether the codes are achieving their stated goals of expanding available reimbursement for nonvisit-based primary care services.

RECOMMENDATIONS

Over the past year, subgroups of the Standing Committee on Primary Care have produced recommendations for how to better determine the value of primary care in the context of hybrid payments. The Response to the Centers for Medicare & Medicaid Services CY 2025 Advanced Primary Care Hybrid Payment Request for Information recommended that CMS rely on empirically collected, observational data for physician time and practice expense calculations to help determine hybrid payment rates, rather than use information presented by the RUC (NASEM, 2024a). The Response to the Pay PCPs Act of 2024 Request for Information (NASEM, 2024b) made similar recommendations regarding the composition and functions of a proposed new CMS-directed advisory committee, distinct from the RUC. Alternative valuation methods should address the many issues described above that currently result in problematic appraisal of primary care services by the RUC, including sampling limitations, nonresponse bias, biases introduced by reliance on self-report data, and failure to capture clinician effort outside of discrete patient care encounters and the contribution of the full interprofessional team.

This report builds upon those reports, information gathered at public meetings, and the work of the 2021 National Academies report Implementing High-Quality Primary Care: Rebuilding the Foundation of Health Care. That report presented an implementation plan across five objectives to ensure that high-quality primary care is available to everyone in the United States. To increase the financial capacity to deliver high-quality care in the primary care setting, the report made two key recommendations to increase spending directed toward primary care. The committee agrees with Recommended Action 1.3 from that report, which states:

The Centers for Medicare & Medicaid Services should increase the overall portion of spending going to primary care by:

• Accelerating efforts to improve the accuracy of the Medicare Physician Fee Schedule by developing better data collection and valuation tools to identify overpriced services, with the goal of increasing payment rates for primary care evaluation and management services by 50 percent and reducing other service rates to maintain budget neutrality; and
• Restoring the Relative Value Scale Update Committee to the advisory nature as originally intended by developing and relying on additional independent expert panels and evidence derived directly from practices.

Since the 2021 report was published, HHS has adopted the 2021 National Academies definition of high-quality primary care and has publicly stated that the report and its recommendations are one of the agency’s “blueprints” for primary care policy (HHS, 2023; Money et al., 2024). While federal and nonfederal actors have started to implement several of the report’s recommendations, there has not been any discernable movement toward implementing Recommended Action 1.3. One key barrier to implementation is the lack of relevant and specific data needed to accurately assess the relative value of primary care. To this point, during a May 2024 public meeting of the Standing Committee on Primary Care,⁹ a discussion centered on statutory requirements CMS must follow to make fee schedule valuation decisions based on data through defined mechanisms. However, the data sources CMS can draw from are not prescribed and CMS can review new sources of data that help it better understand the relative resources associated with delivering primary care. Additionally, this committee hosted a public meeting in November 2024 (see Appendix C for the meeting agenda) with experts on the regulatory environment that governs CMS and the PFS processes and alternative data sources and methods to inform future valuation processes.¹⁰

Building on this prior work, as well as the evidence documented in this report, this committee recommends several changes to improve the CMS valuation process to accurately reflect the full scope of high-quality primary care:

Recommendation 1: When valuating physician services and activities for the Physician Fee Schedule, the Centers for Medicare & Medicaid Services should consider a range of data sources that are directly observed (e.g., electronic health record audit log data or time-motion studies) or reported (e.g., high-quality surveys) and that are analyzed using complementary approaches such as time-driven activity-based costing and large language modeling.

Recommendation 2: Using the data sources and methodologies in Recommendation 1, the Centers for Medicare & Medicaid Services should consider the full range of value-enhancing primary care activities and services that are required for high-quality primary care in its valuation process. This includes the efforts of the full interprofessional primary care team and nonencounter-based activities.

Recommendation 3: The Centers for Medicare & Medicaid Services (CMS) should invite other expert and technical advisory organizations beyond the Relative Value Scale Update Committee to contribute to the annual valuation process. This could entail CMS establishing a separate internal or external expert group or involving contractors or researchers to conduct independent analyses.

Recommendation 4: The Centers for Medicare & Medicaid Services (CMS) should annually evaluate Advanced Primary Care Management (APCM) to better understand the accuracy of the valuation of APCM services, determine which practices are and are not using the new codes, assess the extent of beneficiaries giving consent to participate in APCM billing and the sociodemographic characteristics of patient populations receiving APCM services, and identify potential unintended consequences. It should take care to not rush to expand APCM payments without a clear understanding of the uptake issues above, and with full consideration of unintended negative consequences on other important CMS alternative payment models such as accountable care organizations.

CONCLUSIONS

While implementing these four recommendations should greatly improve the accuracy of the relative value assessment process for primary care activities in the short term, longer-term legislative changes would be required for CMS to expand valuation beyond the current formula (physician work, practice expense, and malpractice risk) to consider, from a normative perspective, the value of services and activities to the health and well-being of Medicare and Medicaid populations. Additionally, although the scope of this report and its recommendations are limited to primary care there is no reason why these recommendations could not similarly apply to the valuation process for other specialties.

Future efforts could consider longitudinally incorporating need and access and expected health outcomes into valuation of services. This could be linked with tracking and measuring whether increased valuation resulted in the intended increases in appropriate use of high-value services and activities and associated workforce shifts (Maciosek et al., 2006, 2017).

The committee appreciates the many implementation issues that must be addressed to improve primary care valuation processes. Adopting more rigorous methods for comprehensively measuring and analyzing primary care work and resources brings with it added costs to CMS to perform (either itself or through an agreement with a contracted entity to perform). These are potential costs that CMS has largely been able to avoid under the current arrangement of delegating to the AMA the authority and expense of the RUC. Attention would need to be given to standardization and harmonization of data inputs, decisions on frequency of updating measurements, and related technical issues.

As with any change in policy and processes, monitoring would be needed to detect potential unintended consequences. For example, the report, Response to the Centers for Medicare & Medicaid Services CY 2025 Advanced Primary Care Hybrid Payment Request for Information (NASEM, 2024a), noted that smaller primary care practices that are not part of large health systems and least equipped for practice transformation are often located in underresourced communities serving rural and low-income populations. Reforms in valuation of primary care services must be accompanied by monitoring of uptake of APCM and other new PFS codes, the case mix of patients served by different practices, and related phenomena to detect potential “gaming” of payment reforms and aggravation of health care disparities, to inform any corrective actions that may be needed.

Perhaps that biggest implementation challenges are political ones. The budget neutrality rule for the PFS creates a zero-sum-game dynamic that intensifies opposition to valuation reforms that may benefit one group at the expense of other groups. The recent success of several states (Condon et al., 2022) in implementing policies to increase the proportion of all health care expenditures in the state spent on primary care indicates that sufficient political consensus has been mustered in these settings to drive forward greater investment in primary care, providing potential lessons for valuation reforms at the federal level. A recent public opinion poll found that people in the United States highly value primary care and vastly overestimate the share of health care expenditures devoted to primary care (Ma et al., 2025), suggesting that the public may be a key constituent to engage in policy deliberations on the valuation of clinician services.