The Fund supports networks of state health policy decision makers to help identify, inspire, and inform policy leaders.
The Milbank Memorial Fund supports two state leadership programs for legislative and executive branch state government officials committed to improving population health.
The Fund identifies and shares policy ideas and analysis to advance state health leadership, strong primary care, and sustainable health care costs.
Keep up with news and updates from the Milbank Memorial Fund. And read the latest blogs from our thought leaders, including Fund President Christopher F. Koller.
The Fund publishes The Milbank Quarterly, as well as reports, issues briefs, and case studies on topics important to health policy leaders.
The Milbank Memorial Fund is is a foundation that works to improve population health and health equity.
December 14, 2023
Issue Brief
Jonathan Staloff
Max Birger
Lingmei Zhou
Ashok Reddy
Christopher Chen
Judy Zerzan-Thul
Joshua M. Liao
Publication
Sep 16, 2024
May 20, 2024
Back to Publications
Many states have worked to increase primary care investment, which lags in the U.S. compared with other developed countries. These efforts share the need to measure and track primary care utilization and spending. States and groups have used health care claims data to implement methods that define “primary care” through combinations of provider specialty and the place or nature of rendered services. While such work sets a foundation for measuring primary care, claims-based definitions can be improved in several ways. First, definitions should fully capture clinician, organization, and facility types responsible for primary care. Second, definitions should reflect real-world variation in settings where primary care is provided. Persistent gaps in these areas will likely limit policy efforts to track primary care in claims data and generate policy- and practice-relevant insights.
To address these issues, we developed the Value and Systems Science Lab Method, a versatile approach for identifying primary care services. We demonstrate this three-step method using billing and servicing provider taxonomies in Washington Medicaid claims data. This issue brief highlights the feasibility of the approach, which is applicable to a wide range of policymakers and is versatile enough to capture variation in care delivery for populations in a range of datasets, including Medicaid or commercial claims as well as all-payer claims databases. The approach described in this brief offers a way of identifying primary care services that will remain salient as primary care policy evolves over time.
Primary care is central to a well-functioning health care system. At a population level, primary care is associated with lower all-cause mortality, disease burden, and disparities in care.1 Greater investment in primary care is associated with fewer emergency room visits and hospitalizations.1
Many states have worked to increase primary care investment, which lags in the U.S. compared with other developed countries.2-4 Beginning in 2010, Rhode Island required health plans to increase primary care spending as a proportion of total health care spending by 1% per year from 2010 to 2014, with a target of 10.7% by 2015.5 By 2018, primary care expenditures were estimated to be 12.3% of total health care spending among Rhode Island commercial health plans.6 In 2017, Oregon passed legislation requiring health plans to increase primary care spending to at least 12% of total spending by 2023. In 2019, the state’s primary care spending ranged from 10.8% to 16.2% of total medical expenditures across insured populations.7 In 2022, Washington legislators directed a state-sponsored Health Care Cost Transparency Board to provide recommendations on how to increase primary care spending to 12% of total health care expenditures.8
Common to these efforts to track spending is the need to identify primary care use. To do so, a number of states have used health care claims to define “primary care” through combinations of provider specialty and the place or nature of rendered services. While prior methods have set a foundation for measuring primary care (Table 1), claims-based definitions can be improved in several ways.
Table 1. Approaches for Identifying Primary Care in Claims Data9-12
First, definitions should fully capture the types of clinicians, organizations, and sites responsible for primary care. Unfortunately, gaps remain on these fronts. For instance, imprecise estimates have been used to approximate primary care provided by nurse practitioners or physician assistants. Prior approaches have also omitted some sites where primary care is delivered. More work is needed to better incorporate organization types (e.g., group practices) into primary care definitions.
Second, definitions should reflect real-world variation in how and where clinicians provide primary care. Prior approaches either incompletely capture clinician and site combinations or fail to do so altogether. In reality, clinicians can practice in different settings and may deliver primary care in some situations (e.g., a family medicine physician working in a physician office) but not others (e.g., a family medicine physician working in an emergency department). Ideally, primary care definitions would also account for group practices that deliver both primary and specialty care through different clinicians.
Persistent gaps in these areas have limited policymakers’ efforts to track primary care in claims data. Collectively capturing these dynamics would support versatile primary care definitions and generate policy- and practice-relevant insights, such as where and how primary care is delivered.
We sought to address these issues by developing the Value and Systems Science Lab Method, a versatile approach for identifying primary care services that incorporates clinician, organization, and facility types in ways that reflect real-world practice. This brief describes major steps of our approach, which policymakers in other states can use to identify primary care services to support policy goals.
Guided by a conceptual framework, the Value and Systems Science Lab Method involved a three-step process for identifying primary care services. In Step 1, we identified individual clinicians, facilities, and organizations that conceivably provide primary care services in different places of service. In Step 2, we used an iterative process to identify combinations of clinicians, facilities, and organizations and categorized these combinations as reflecting primary care versus non-primary care services. In Step 3, to ensure completeness and validity of provider specialty data, we incorporated publicly available information from a national registry of provider taxonomies.
As a proof-of-concept analysis, we implemented our approach on Washington Medicaid claims for a select set of outpatient visits. Though this work was undertaken amid ongoing efforts in Washington to measure primary care, the three-step approach was designed to enable replication or adaption by policymakers in other contexts and using other datasets.
This work was guided by several principles. First, a clinician can practice in different settings. Second, a clinician can provide primary care in certain situations and non-primary care in other situations. Third, a given organization (e.g., physician group) or facility (e.g., doctor’s office) can provide different combinations of primary and non-primary care services through different clinicians and specialties. As a result, our conceptual framework for identifying primary care was based on the relationship between three factors: clinician, organization, and facility (Figure 1).
Figure 1. Conceptual Framework
The Washington Medicaid claims used in this work included information about the clinician providing a given health care service (called servicing provider) and the clinician, organization, or facility billing for that service (called billing provider). Servicing providers are the individual clinicians who provide care, while billing providers can be either clinicians billing for services as individuals, group practices billing for services as organizations, or facilities billing as sites of care.
Information available in Medicaid claims about servicing and billing providers included servicing provider and billing provider taxonomies. Provider taxonomies are nationally standardized methods of identifying the specialty (e.g., family medicine, general surgery) and type of clinicians (e.g., physician versus nurse practitioner), as well as the type of organizations (e.g., single- versus multi-specialty group practices) and facilities (e.g., rural health clinic, community health center).
To test our approach in a proof-of-concept analysis, we defined visits using the following Current Procedural Terminology (CPT) codes representing common outpatient patient visits: new problem-focused visits (CPT 99201-99205), return problem-focused visits (CPT 99211-99215), office or other outpatient consultation services (CPT 99241-99245), and telephone visits (CPT 99441-99443).
We operationalized our method by implementing a series of steps for each claim using servicing provider and billing provider taxonomy variables to capture primary care versus non-primary care visits.
Step 1: Designate Clinicians, Groups, and Facilities Providing Primary Care
First, we constructed an inclusive list of all clinician, organization, and facility types that could conceivably deliver primary care services. We began by identifying places where primary care services could be provided and generating a draft list based on provider taxonomies.
We refined our draft list by comparing it to taxonomies used in prior work to identify primary care and non-primary care services in Washington.9 Through this process, we (1) removed clinician types and specialties that were designated as primary care in prior work, but which were unlikely to deliver primary care (e.g., neonatal and critical care nurse practitioners, surgical physician assistants); and (2) added facility types (e.g., community health clinics/centers) and organization types (e.g., multi-specialty group practices) through which patients could have received primary care, but which were excluded in prior work.
We then used the refined taxonomy lists to designate a set of clinician and facility types as primary care, with others defaulting to non-primary care. We designated groups as single-specialty or multi-specialty to reflect the reality that patients can receive primary care through single-specialty groups consisting solely of primary care clinicians or through groups consisting of both primary care and non-primary care clinicians (e.g., a multi-specialty group). This process yielded six designations: (1) primary care clinician; (2) non-primary care clinician; (3) single-specialty group; (4) multi-specialty group; (5) primary care facility; and (6) non-primary care facility.
Step 2: Identify Clinician, Group, and Facility Combinations Reflecting Primary Care
These six designations enabled us to create and interpret combinations of billing providers and servicing providers as reflecting primary care versus non-primary care services. This approach provided a more flexible way to identify primary care services in real-world practice.
For instance, under other approaches, all encounters with primary care clinicians, such as internal medicine physicians, are often designated as primary care visits. However, context matters: visits with an internist should be interpreted differently when they occur at ambulatory surgical centers versus community health centers. Approaches based predominantly on clinician specialty overlook such information and create the risk of incorrectly identifying primary care visits.
To overcome these barriers, our approach defines primary care visits as only those occurring with a primary care clinician as the servicing provider (e.g., internal medicine physician) and having a primary care facility as the billing provider (e.g., community health center) (Table 2, Claim 1). In contrast, visits provided by a primary care clinician but billed to a non-primary care facility (e.g., ambulatory surgical center) are not considered primary care (Table 2, Claim 2).
Table 2. Identifying Primary Care Visits (Examples)
This approach also accounts for situations in which primary care was delivered through single- and multi-specialty groups — either of which could encompass primary care, but neither of which necessarily do. In particular, we interpreted single- or multi-specialty groups as providing primary care when the servicing providers were primary care clinicians (Table 2, Claims 3 and 4), but not when the servicing providers were non-primary care clinicians (Table 2, Claims 5 and 6).
We extended this approach across all possible combinations of billing providers and servicing providers, applying our collective clinical and care delivery expertise to interpret each combination as reflecting primary care or non-primary care visits. This approach can be replicated by local groups and stakeholders to ensure that combinations and interpretations capture nuances and features of different settings.
Step 3: Address data issues
After categorizing combinations of billing and servicing providers, we observed problems with the consistency of information in the billing provider taxonomy variable for group practices and facilities (in particular, a given group or facility could be associated with multiple taxonomies across different claims). We addressed this issue by importing data from the National Plan & Provider Enumeration System (NPPES), a publicly available national registry operated by the Centers for Medicare and Medicaid Services.13 In particular, we used NPPES data to fill in missing billing provider taxonomy values for groups and facilities in our data — a step that helped ensure taxonomical consistency and data completeness for provider variables.
We applied this method to analyze claims data for 328,314 adult and 372,565 pediatric beneficiaries continuously enrolled in Medicaid managed care and residing in Washington in both 2019 and 2020. The goals of this analysis were to demonstrate the feasibility of our approach in capturing how much primary care was provided and to generate insight about patterns of primary care delivery (where and through what entities primary care was provided).
Our approach revealed that among adults enrolled in Medicaid in Washington who accessed outpatient care, primary care visits occurred twice as frequently as non-primary care visits. Most primary care visits (nearly 70% in 2019) occurred through federally quality health centers and multi-specialty groups. Similarly, our approach yielded insights about primary care for children enrolled in Medicaid: nearly 80% of outpatient visits among this population were primary care, and, as observed for adults, pediatric primary care occurred most frequently through federally quality health centers and multi-specialty groups (Figure 2).
The Value and Systems Science Lab Method — which builds on prior work to identify and describe primary care using claims data in a more precise and comprehensive way — has three key policy implications. First, it highlights the feasibility of improving claims-based definitions. In this work, we captured the realities of clinical practice and designated primary care based on simultaneous consideration of clinician, organization, and facility types. Though no method is perfect, the Value and Systems Science Lab Method retains the benefits of prior methods while being likely to achieve greater validity and resonance with practitioners, health care delivery organizations, and health care facilities.
Second, our method can inform a wide range of policymakers. Because it uses billing and servicing provider and taxonomy data as ubiquitous claims variables, our approach is versatile enough to capture variation in care delivery for populations in a range of datasets, including Medicaid or commercial claims from other states as well as all-payer claims databases. Indeed, the Value and Systems Science Lab Method has informed the efforts of a primary care committee created in Washington to track primary care utilization and spending as part of a statewide initiative to measure health care spending.14 A range of other policymakers can replicate or adapt our approach for use in their primary care policy efforts.
Third, the Value and Systems Science Lab Method will remain salient as the focus of primary care policy evolves over time. For many states and groups, the most important near-term goal is to identify overall trends in primary care utilization and spending. Long-term policy questions will require insight about specific care patterns — that is, where and through what entities different populations receive primary care. By defining primary care in view of clinicians, organizations, and facilities, our approach also yields care pattern insights that can support targeted policies and programs.
Our method cannot overcome inherent limitations of claims data, including the inability to capture continuity of care. Capturing that information would require broad reformulation of claims data fields or the collection of supplemental data from patients and provider organizations. Second, like all claims-based definitions, the Value and Systems Science Lab Method cannot identify activities that do not result in billable services, such as certain types of asynchronous care coordination. Policymakers must use other data collection approaches to capture these activities, which have been increasingly implemented under value-based payment. Third, replication of our approach must consider data completeness and ability to compare data elements to external sources.
Our work suggests several areas of future exploration for state policy work. States can replicate our steps using an expanded set of codes. Policy leaders can easily incorporate a more comprehensive set of primary care services and procedures — such as vision, dental, or behavioral health services — into our method. Second, states can adapt the method to fit a particular policy purpose. For instance, our method identifies all nurse practitioners and physician assistants as possible primary care providers. However, based on their policy needs, states may wish to designate nurse practitioners and physician assistants according to the scope and orientation of their group practices (e.g., distinguishing between those in single- or multi-specialty groups encompassing primary care versus those in groups that do not encompass primary care).
Ultimately, our versatile approach for identifying primary care services incorporates clinician, organization, and facility types in ways that reflect real-world practice. This approach can be used or adapted by a range of policymakers and states to inform primary care policy and programming.
