Give Me Some Credibility: Addressing the Challenge of Volatility in Value-Based Contracts—Part 1
By James Pisko, Maria Knox and Keith Passwater
Health Watch, March 2023
The increasing popularity of value-based care (VBC) in the US health care market has encouraged providers and payers to align financial incentives in an effort to reduce costs. As a result, this has led to shifting insurance risk[[1]] from health insurers to health care providers through the use of VBC contracts. However, managing insurance risk is new for most health care provider groups.
Managing risk is one of the core competencies of insurance companies and is rooted in the pooling of many “risk exposures” such that losses can be estimated with greater confidence. The level of confidence related to a particular number of pooled risk exposures is frequently referred to as credibility.
As provider groups begin to take responsibility for managing risk, they too need to think about credibility and the volume of risk exposures in the population for which they are being held financially accountable, to ensure the risk they are taking on is consistent with their goals. Said another way, risk-taking providers need to understand whether sufficient pooling of risk exposures exists within the control of their practice to warrant the level of risk transfer that is being proposed by their partner insurance companies (i.e., payers).
As actuaries design the risk transfer mechanisms, it is important to examine a VBC contract for sufficient credibility. This two-part article discusses key items impacting credibility from the traditional point of view of a payer system vs. a provider system participating in a VBC contract.
For the purposes of this article, our scope of VBC models is limited to risk-bearing models with a shared-savings or shared-loss component. In this first part of the article, we will discuss the following credibility related topics:
- Risk pool size: the larger, the better
- Risk exposures: the more direct the assignment of members to the VBC program, the better
- Metrics of interest: measures aligned to provider profitability
The second part of this article will build on these concepts and apply them to the following VBC components:
- Benchmarks: pros and cons requiring prudent choices
- Measurement period results: possibility of decreasing credibility/increasing volatility, especially segmenting the covered population for measurement
- Limiting risk transfer: various methods that generally reduce the degree of risk transfer and increase credibility
- Methods for assessing credibility: the same statistical methods for payers and providers
Risk Pool Size
The size of provider risk pools are generally smaller than that of a traditional payer, since the provider is held financially responsible only for members that are within the control of the provider organization, also known as attributed members. In this case “size” is measured using traditional metrics such as number of risk exposures and aggregate losses. The reduced risk pool size increases the volatility in expected claims for which providers are responsible compared to the volatility in expected claims for which payers are responsible.
As membership attributed to VBC programs is generally lower than the overall plan enrollment, credibility becomes a significant issue in both developing cost-of-care benchmarks for VBC shared savings programs (“pricing” in payer terms) as well as volatility in actual performance that is ultimately compared to benchmarks to determine profit and loss.
Additionally, as risk pools are generally larger for payers, payers have the ability to further segment members in their risk pools into larger, homogeneous risk groups relative to the populations attributed to providers. This is important because payers might have credible experience even within these subgroups (i.e., rating variables), whereas providers might be forced to use a method of accounting for these risk drivers using adjustment factors that are estimates themselves, further compounding the uncertainty of final projections. These reasons make full risk transfer to providers challenging and should be considered when determining the level of risk to be transferred.
Risk Exposures
Provider systems have a less direct “member assignment” process than traditional payers. The output of this process is a list of covered members for a specified period of time, similar to the enrollment process of traditional payers. In this article, we refer to these members as “risk exposures” in both the provider and the payer case.
Risk exposures for which payers carry financial responsibility are clearly identified—individuals are enrolled in a plan through a contract that specifies the exact period of coverage, the services covered and to what extent those services are covered.
The process of member assignment is more nuanced. There are several methods for arriving at the roster of risk exposures against which providers are ultimately measured and paid:
- Voluntary member assignment. This is when members enrolled at a payer voluntarily elect their provider. A member electing their primary care practitioner or provider directly is probably closest to the reality of who dispenses that member’s care, but there are cases when claim evidence contradicts this selection.
- Attribution. In less direct methods this pairing process of an enrolled member (at the payer) to a provider is done through inferring relationships based on claims history or geographic proximity to a provider. This process is referred to as “attribution.” There are variations in methodologies even within claims-based and geographic-based assignment. Certain VBC models deploy multiple attribution methodologies within the same program, creating additional complexity. Provider attribution methodologies are ultimately an approximation-based approach for assigning risk exposures to something that looks like health plan enrollment.
- Claims-based attribution. Historical claims data is a reliable indicator of attribution; however, it does not account for more recent diagnoses of health conditions that could dramatically shift provider-patient relationships.
- Geography-based attribution. This method is built on the assumption that members will seek care at the closest provider to their residence. Perceptions of quality are often the biggest reasons for members to break this assumption; they are willing to drive a few extra minutes if they believe their health is in better hands.
- VBC products. This method is similar to payer enrollment and happens when payers create a separate insurance product around a provider system. This is not common and usually for payers that enroll a large portion of the population in the region in which the health system is located. Depending on product design, providers often bear little to no financial risk for members seeking care outside of their system.
These methods matter because the population for which providers are accountable for managing risk in VBC contracts is often not always a one-to-one relationship between the members and services in which those providers actually rendered services.
This can be mitigated with accurate and timely reporting of attribution rosters to providers with features to assess the accuracy of attribution methodologies.
Metrics of Interest
Both payers and providers are interested in assessing the volatility and credibility of covered claims.
To limit variability in claims from the number of underlying exposures, claims per member per month (PMPM) metric is most frequently used to set benchmarks.
It should be noted that payers generally have the same profit margin regardless of the type of claim, whereas profit margin for providers varies by the type of claim. For detailed profitability analysis on the provider side, a further breakdown of claims PMPM grouped by relevant claim types may be necessary.
When defining the metrics of interest, it is important that the following items, specific to the value-based space, are considered:
Ease of Interpretability
There are practical considerations to paying provider systems in radically different ways. Perhaps the most important consideration is the balance between statistical accuracy and ease of interpretation for provider systems.
One example is measuring performance against targets segmented by individual product types instead of introducing what would be considered a “black box” plan benefit design adjustment factor. But these segments need to be large enough to allow for credibility, which may mean the results will not be adjusted with the same level of precision as when using adjustment factors.
Higher Consistency in Unit Cost
Aggregate medical or claim cost is the measure we are trying to predict, whether for a payer or provider as part of a VBC contract. Total costs can be thought of as the product of unit cost of medical services and the utilization of those services. With any covered population at a health plan, the mix of claims is likely to spread across provider organizations with greater variability in unit costs. A high percentage of claims in an accountable care organization (ACO)–style contract occur within that ACO system, which should be more closely aligned in unit costs for the same services among their providers within the ACO.
Subject vs. Relevant Experience
When assessing the credibility of subject experience, insurance carriers usually have relevant experience to blend with the subject experience to increase prediction accuracy. In the case of provider systems, relevant experience may not be practically available, or there may be hesitancy to utilize: views such as “It's not clear or fair that I should be measured based on claims data from other provider systems” are not uncommon. Therefore, it might be best for provider systems to use empirical credibility procedures that strictly focus on the subject experience and don’t require the actuary to assume a prior distribution for aggregate claims.
Without relevant experience or a willingness to use it, providers would have to rely on strategies mentioned in the “Methods for Assessing Credibility” section in part 2 of this article.
Conclusion
With the growing number of shared risk contracts provider organizations are collectively engaging in, it is important to assess the credibility of experience in these contracts. Part 1 of this article highlights key differences between payers and providers when it comes to risk pool size and methods of assigning exposures. There are additional credibility considerations in the design of the metrics both parties monitor. In part 2 of the article, we will build on these concepts to examine approaches to address credibility issues when setting value-based contract benchmarks, measuring outcomes, and techniques for mitigating risk volatility.
Statements of fact and opinions expressed herein are those of the individual authors and are not necessarily those of the Society of Actuaries, the editors, or the respective authors’ employers.
James Pisko, FSA, MAAA, is an actuary at Nuna. James can be reached at jamesp@nuna.com.
Maria Knox, FSA, MAAA, is an actuary at Nuna. Maria can be reached at maria@nuna.com.
Keith Passwater, FSA, MAAA, FCA, is the CEO of Havarti Risk Services. Keith can be reached at kpasswwater@havartirs.com.
Endnote
[1] Insurance risk is the risk associated with the unknown variation in the utilization and cost of services. From Spector, Juliet, Cory Gusland and Carol Kim. Insurance Risk and Its Impact on Provider Share Risk Payment Models. Society of Actuaries, January 2018, https://www.soa.org/49345d/globalassets/assets/files/resources/research-report/2018/insurance-risk-impact.pdf (accessed Feb. 6, 2023).