Introduction
Value-based payment (VBP) is designed to support collaborative partnerships between patients and physicians, improve the quality of care and reduce health care spending. To achieve these aims, VBP for primary care must support the four key functions of primary care (i.e., first-contact access, comprehensiveness, coordination and continuity), which are essential to meeting the goals of improved quality and reduced spending.1 The success of VBP is highly dependent on alignment across payers and unlikely to work if only a small subset of a practice's patient population is included. Increased investment in primary care across public and private payers using VBP models designed for primary care will contribute significantly to improving health, eliminating inequities, reducing the per capita cost of care over time and improving the well-being of the care team.
To help facilitate the transition away from fee-for-service (FFS) payment and toward VBP arrangements that sustainably support the kind of robust primary care that is essential to a high-performing health care system, the American Academy of Family Physicians (AAFP) has established a set of guiding principles to describe the ideal design for key components of VBP models for primary care.1
This paper and others in the series aim to translate the guiding principles into actionable steps that key stakeholders can use to implement VBP models that sustainably support primary care. The AAFP will provide background information, industry context and stakeholder recommendations, along with our expertise, to provide a broad understanding of the issues unique to primary care in VBP implementation.
This call-to-action brief focuses on the following principles1:
Risk adjustment methodologies should incorporate clinical diagnoses, demographic factors, and other relevant information such as social determinants of health without exacerbating health care disparities or expanding the administrative burden on primary care practices. Social determinants of health should be identified as risk factors and used for risk adjustment of populations. Primary care physicians cannot be held accountable for providing resources to address social determinants of health that do not exist in the community.
The sections below address reasons family physicians should become familiar with risk adjustment, its role in VBP programs, common methodologies, areas of success, areas that need improvement and recommendations for stakeholders to drive greater implementation of these guiding principles.
Rationale for Action
Risk adjustment—a process used to predict health care costs by assigning a risk profile to an individual’s health status—has long been a critical component of health plan operations, particularly for plans paid on a capitated basis like Medicaid managed care organizations (MCOs) and Medicare Advantage organizations (MAOs). Beginning in earnest in 1997 with the first federal requirements, risk adjustment was initially developed to provide adequate financing to health plans that treat individuals with higher-than-average health needs and minimize incentives to selectively treat healthier members (i.e., cherry-picking).
Several decades later, while risk adjustment continues to be a core focus of health plans, it is increasingly relevant to other types of health care organizations that assume accountability for the total cost of care, including primary care practices participating in VBP programs. As physicians and practices transition away from FFS toward VBP—particularly those involving downside risk and/or per-person payments—they can benefit from the same protections risk adjustment provides to health plans. The ability to predict the relative needs and costs of care for patients and populations is important for achieving success in VBP, and risk adjustment is an essential tool/process to help inform that understanding.
As discussed in the Value-based Payment Models for Primary Care Call-to-Action Brief, participation in public and private sector VBP programs—particularly those that include prospectively paid population-based payments (e.g., per member per month [PMPM] capitation)—is slowly increasing.2,3 To further accelerate the adoption of VBP, the Centers for Medicare & Medicaid Services (CMS) set an aggressive goal of bringing all Medicare beneficiaries and the majority of Medicaid enrollees into an accountable care relationship by 2030.4 Physicians’ appetite to participate in VBP, and ultimately, the success of these programs, often hinges on how well the program payments and financial benchmarks account for the variability of health status within a physician’s patient population. Under VBP arrangements—where practice payments are tied to the cost and quality of outcomes for defined populations or services—risk adjustment is an important tool for ensuring payments and the benchmarks used to determine financial performance reflect the acuity of the patient population.
Moving from a predominantly volume-based FFS system that underfunds and overburdens primary care to a system of flexible, prospectively paid, risk-adjusted VBP programs for primary care necessitates action from all key stakeholders. Stakeholders include purchasers (e.g., employers and/or union trusts who purchase health care on behalf of their workforce), payers (e.g., insurance companies and health plans), policymakers (e.g., lawmakers and regulators), physicians, and organizations that employ family and other primary care physicians.
Considering the potential promise and current shortcomings of risk adjustment in VBP programs for primary care, the AAFP recommends the following actions in Table 1 for various stakeholder entities.
Table 1. Actions to Simplify and Improve Risk-adjustment Methodologies
Current State
Risk adjustment is a statistical process used to predict health care costs based on factors known to be reliable predictors of health care spending, such as age, sex and clinical conditions.5 Payers use a variety of methodologies to calculate a risk score that reflects the estimated cost of providing services for each individual. Since patients, on average, differ in their health care utilization based on these characteristics, a patient expected to incur higher health care costs should receive a higher risk score.
The role and impact of risk adjustment are increasingly important as physicians and practices take on greater levels of financial accountability for patient outcomes. Without risk adjustment, entities assuming accountability for patients’ outcomes may be deterred from enrolling or caring for populations with higher needs because they would be unfairly evaluated on cost and quality metrics against risk-bearing entities that care for healthier populations, and they would not receive payment reflective of the cost of providing care to this population.
Applications of Risk Adjustment within VBP
Risk adjustment currently plays the following roles in VBP programs:
Informs the calculation of PMPM payments under VBP programs using full or partial capitation. A higher risk score equates to higher PMPM payments. Risk adjustment is particularly important under capitated arrangements, as monthly PMPM payments are intended to partially or fully replace FFS payments. Under these programs, practices are responsible for providing needed patient care under a fixed, per-person budget for a defined set of services. Therefore, PMPM payments must reflect the anticipated level and cost of care needed.
Adjusts other types of advanced payments, such as care management fees. While not intended to cover the total cost of care for patients (as with full capitation), VBP programs involving these additional PMPM care management fees on top of FFS revenue can help fund important care management programs. By risk adjusting these payments, practices are better equipped to address the enhanced care management needs of patients who are less healthy.
Informs financial benchmarks used to evaluate performance. Higher-risk scores also equate to higher benchmarks (i.e., the financial targets established by payers administering VBP programs to assess actual spending relative to a predetermined baseline). Risk-adjusting financial benchmarks is especially important under VBP programs that are based on an FFS architecture, such as the Medicare Shared Savings Program (MSSP). Under these arrangements, participants continue to receive FFS payments with the opportunity to share in generated savings (or be on the hook for financial losses) at the end of the performance year. Success in these shared-savings models largely depends on accurate and fair benchmarking. Since FFS payments are not adjusted to reflect the disease burden, demographics or social needs of patients, risk-adjusted benchmarks provide an opportunity to account for the acuity of a patient population. For more information on benchmarking in VBP programs, see the AAFP’s Benchmarking in Value-based Primary Care Payment Models Call-to-Action Brief.
Accounts for social risk factors. Health-related social needs are more predictive of health care costs than clinical care,7 but they are typically not included in risk-adjustment models. However, this is rapidly changing as recognition of their impact increases and CMS prioritizes health equity considerations. Some states (e.g., Minnesota, Massachusetts and Maryland)8,9 and federal programs (e.g., Accountable Care Organization Realizing Equity, Access, and Community Health [ACO REACH]),10 have begun incorporating social risk factors into risk-adjustment methodologies, providing larger or supplemental PMPM payments or more favorable benchmarks for organizations caring for patients with greater social needs. Additionally, CMS is soliciting comments on whether and how to enhance the CMS-hierarchical condition categories (HCC) model to address SDoH.11
Evaluates performance on quality metrics. Risk adjustment is also used in quality measurement to ensure fair evaluations, though this brief focuses on risk adjustment as it relates to payments and financial performance. For more information on performance measurement under VBP, see the AAFP’s Performance Measurement in Value-based Primary Care Payment Models Call-to-Action Brief.
Common Risk-adjustment Models
There are currently several statistical models used for risk adjustment. The most prevalent is the HCC model, which CMS developed for the Medicare population. Due to its wide use across traditional Medicare and MA, many private payers also use HCC for non-Medicare populations. The model calculates a risk-adjustment factor (RAF) based on patient characteristics (i.e., age, sex, dual-eligibility status and whether the beneficiary is in Medicare due to age or a qualifying disability) and diagnoses reported in the prior year.12,13 An RAF score of 1.0 represents the average patient. Higher scores indicate higher predicted spending. RAF scores are normalized to the population each year.
While the CMS-HCC model is the most pervasive, other public sector risk-adjustment methodologies are tailored to non-Medicare patient populations, such as the U.S. Department of Health and Human Services-HCC model, which is for patients covered by Affordable Care Act Marketplace plans. In addition to the public sector models designed and implemented by government entities, academic institutions, actuarial firms and vendors design many different private risk-adjustment methodologies. These private-sector methodologies operate similarly to the HCC model conceptually but are often less transparent than public-sector methodologies and can incorporate additional factors into risk adjustment, including pharmaceutical use, patient-functional status and site of care. Some private sector models generate a patient risk score like the CMS-HCC model, including the Chronic Illness and Disability Payment System (CDPS), commonly used for Medicaid populations. Others, called clinical categorical models, assign patients to a specific risk category. Models of this type include 3M’s Clinical Risk Groupers (CRGs) and Johns Hopkins’ Adjusted Clinical Groups (ACGs).
What Is Not Working
Lack of alignment across risk-adjustment models. While there are commonalities among the various risk-adjustment models, the variations between them can lead to inefficient or duplicated workflows if physicians must report different sets or types of information for risk adjustment depending on the population. For example, a process designed to capture needed information for one risk-adjustment model may not apply to another. Ultimately, this is a distraction from providing high-value patient care.
Lack of transparency in the commercial sector. While models introduced by the federal government provide ample information on the methodology for calculating risk scores and applying risk adjustment, many privately developed models rely on proprietary methodologies. They are less forthcoming with information, putting physicians at a disadvantage when contracting with plans in commercial VBP arrangements.
Many models use prospective risk adjustment. Some of the most popular risk-adjustment models, including the CMS-HCC model, use prospective risk adjustment, meaning that prior-year diagnoses inform the predicted costs. This can lead to a misrepresentation of patients with advanced illness and a rapid decline of patients who require more resources than their prior-year diagnoses suggest. Some models account for this. For example, the risk-adjustment methodology used for the High Needs Population ACOs in ACO REACH uses a concurrent model, where current-year diagnoses are used to risk adjust.
Risk-adjustment models are still statistically weak. Current methodologies, particularly the commonly used prospective models, are only able to explain a small share of spending and tend to overpredict costs and risk for healthier-than-average patients while underpredicting the cost of caring for sicker-than-average patients.14,15 Limitations also arise when determining which diagnoses or other attributes to include in a model or how they are operationalized.
Risk-adjustment models are more limited for pediatric populations. Since the most commonly used risk-adjustment models are calibrated for adult populations, research is limited for risk-adjustment methodologies for pediatric populations. While researchers are making efforts in this space, they cite the historical lack of focus on this population as a reason for little widespread use. 16,17 Currently, when children are included in VBP models (e.g., Medicaid managed care), they are often treated as “little adults,” using models built around adult populations despite the fundamentally different health care needs.18
Social risk factors in risk-adjustment models are still in the early stages. The slow uptake of incorporating social risk factors into VBP is a positive development. However, several barriers remain to effectively incorporate social risk factors into risk-adjustment models, including the following:
Risk adjustment still involves burdensome and inefficient processes. While technological advancements and workflow adjustments can help, there are still aspects of risk adjustment that remain burdensome for physicians, like the annual recoding of lifetime events and the extensive detail required for documentation of diagnoses—already a subjective and inconsistent process.15
The economics of risk adjustment create the potential for gaming. Since patients with higher risk scores generate higher payments, the economics of risk adjustment can create an incentive to “game” the process by inflating risk scores. This problem is particularly notable in the MA market. While CMS is cracking down on this practice through increased auditing,23 the potential for gaming inherent in the risk-adjustment process remains a concern.
What Is Working
Risk adjustment is necessary for a well-functioning value-based care ecosystem, but some elements are clearly working better than others. The appropriateness of risk adjustment is an area of debate among policy leaders and health care stakeholders. While there are still several outstanding issues with risk-adjustment methodologies, the following are actively being addressed:
Demographic and diagnostic data are generally reflected in risk-adjustment models. Commonly used public and private models are typically risk adjusted along demographic and diagnostic lines, providing an accurate view of a patient’s health status when the information provided vis à vis the submission of claims is complete and accurate. Furthermore, many payers and CMS VBP programs are improving their methods for capturing diagnoses through claims and incentivizing the administration of health risk assessments, which, when performed by physicians or other clinicians associated with a practice, can lead to more precise risk adjustment that better reflects the health status of a patient panel without disrupting established physician-patient relationships. Risk adjustment for these types of data should not be assumed when entering into risk-based contracts, so it is crucial for physicians or contract negotiators to ensure these factors are reflected.
Alternative payment models (APMs) are becoming more advanced. Payers are improving the application of risk adjustment in VBP arrangements, working to combat the “gaming” of risk scores to garner higher payments while still factoring in rising risk in patient populations year over year. For example, CMS historically placed a 3% cap on risk score growth in MSSP and the Center for Medicare and Medicaid Innovation (CMMI) models, such as Direct Contracting, to limit the artificial inflation of risk scores. However, this cap may punish ACOs whose patient populations show legitimate risk-score growth. Recent changes to the MSSP and the successor model to Direct Contracting, ACO REACH, account for changes in patient demographics when applying growth-cap calculations.
New technologies are reducing the physician burden. The marketplace for artificial intelligence-driven risk-adjustment solutions is expanding. New tools are helping to improve productivity and diagnosis capture rates, alleviating some of the burden placed on physicians to document.
Physicians better understand the importance of risk adjustment. In addition to understanding how risk adjustment impacts payment, more physicians are realizing how risk adjustment enables the opportunity and ability to provide excellent care to patients who most need it by providing an understanding of the health status of their patient panel and allocating sufficient resources to adequately address patient needs.
Call to Action
The AAFP calls on payers and policymakers to focus on reducing the administrative burden for family physicians by minimizing data-reporting burdens and aligning on a shared framework for risk adjustment incorporating diagnostic, demographic and SDoH data. The AAFP further asks that risk-adjustment methodologies be made transparent and provide risk scores at the patient level to ensure physicians have the insight needed to perform well in VBP contracts.
When incorporating social-risk factors, the AAFP calls on payers and policymakers to ensure that social-risk adjustment uses validated, evidence-based indices for population-level factors and does not punish physicians if resources to address social needs are unavailable in the community.
The AAFP also calls on payers and policymakers to continue improving risk-adjustment methodologies by improving the predictive power of risk-adjustment models and utilizing concurrent risk adjustment wherever feasible.
For family physicians and employers of family physicians, the AAFP suggests developing efficient processes to ensure accurate risk adjustment by training physicians and support staff to document and code accurately and appropriately, exploring ways to improve the risk-adjustment process, and ensuring the capture of risk-adjustment-relevant information is incorporated into patient visit workflows.
(April 2024) (September 2024 COD)