Feature

Risk, Rates, and Reality

Risk, Rates, and Reality

By David A. Quinn

Misconceptions of Actuarial Soundness in Medicaid Capitation Rate Development

The first 110 minutes of the two-hour meeting were cordial. The Medicaid managed care organization (MCO) presented cost drivers, and the state listened and asked clarifying questions, including myself as the state’s representing actuary. Then, the heady topic of losses arose. Suddenly, the last 10 minutes was a flurry of arguments and counterarguments: losses, solvency, surplus, risk compensation, budget—and then it happened: The MCO declared, “The rates are unsound.”

In Medicaid managed care, the health insurer is called an MCO and is paid by the state a fixed and regularly occurring capitation rate to cover the health needs of the state’s Medicaid population. Capitation rates may vary by geography, population, and other rating variables, but we will discuss rates as a single rate cell for simplicity. An actuary forecasts this rate; it is a projection. Often, the forecasting is done by a consulting actuary hired by a state. But note that the MCO does not set the rate; rather, it is forecasted by the actuary and commonly varies between MCOs from health-based risk adjustment.

The contracted rate versus actual costs will determine the MCO’s profit or loss, sometimes called underwriting gain or underwriting ratio. The rate is mostly a function of the data, methods, and assumptions the actuary used in forecasting.

Regulations from the Centers for Medicare & Medicaid Services (CMS), actuarial standards of practice (ASOPs), and other industry best practices guide forecasting practices and sync what stakeholders can expect from the rate development. The actuary certifies and documents the rate, and CMS determines whether the rate is “actuarially sound” as defined in 42 CFR § 438.4(b). If CMS determines the rate is actuarially sound, the state gets a rate approval letter from CMS.

The MCO bears the risk or reward of how different costs emerge versus the contracted rate influenced by the actuary’s forecast. Without intervention, the desired financial outcome is if actual costs come in modestly below the rate; the MCO’s capital at risk is rewarded, the state feels like an effective steward of taxpayer dollars, and costs sum close to the mean of the actuary’s forecast.

Where’s the Bottom Line?

MCOs take on risk in capitation arrangements and expect to profit most of the time. And yet, the MCO loses sometimes. And it feels like something went wrong—that someone should be held accountable. And like in the meeting, it can lead to arguments about the rate’s actuarial soundness. Hence, it’s important to understand the practices, regulations, and statistics underlying actuarial soundness and its relationship to financial outcomes. To start, ASOP No. 1, Introductory Actuarial Standard of Practice, requires the actuary to define the meaning of “actuarially sound” in context.

For the Medicaid context, actuarial soundness is defined similarly in ASOP No. 49, Medicaid Managed Care Capitation Rate Development and Certification, and in CMS regulation at 42 CFR § 438.4(a):

“Actuarially sound capitation rates are projected to provide for all reasonable, appropriate, and attainable costs that are required under the terms of the contract and for the operation of the MCO…” (emphasis added)

However, states, MCOs, and even actuaries forget that “projected to” precedes the word “provide” in the definition.

The MCO contracts only a single rate, a point estimate, compared to the probabilistic distribution of the forecast. The actuary can describe the forecast’s statistical uncertainty with a distribution: all possible forecast values and how often they are expected to occur. When actual costs are observed, it is difficult to parse the effects of the MCO’s business decisions, the (un)luck of how millions of Medicaid beneficiaries interact with thousands of providers across hundreds of facilities, and how it compares to the actuary’s forecast distribution. There will always be risk.

Generally, a forecast distribution is described by its center and spread, like a mean and standard deviation. This idea of spread, deviation, variance, volatility, or uncertainty is part of the risk MCOs take under a managed care arrangement. In July 2024, the Society of Actuaries summarized MCOs’ profits as part of updating its “Medicaid Managed Care Underwriting Margin Model.” The 2013–2022 profits had a median of 2.0% with a standard deviation of 6.4%.[1,2] These statistics will vary by state, but given these national ones and assuming they follow a normal (Gaussian) distribution, an underwriting ratio of zero or less is a likely outcome at 38%, but the probability of profit is more likely at 62%.[3]

CMS regulation and actuarial standards require the capitation rate to include an underwriting margin component for “margin for risk or contingency.”[4] This margin changes the probability of an MCO profiting, but no guarantee exists. The MCO losing money is still a plausible outcome. For example, assuming the underwriting ratio standard deviation is 6.4%, a capitation rate with a pre-tax underwriting component of 2% is expected to be profitable 62% of the time. In contrast, a component of 3% would be profitable 68% of the time. The prospects are favorable.

However, stakeholders sometimes conflate profits or losses with actuarial soundness. A loss stems from higher actual costs than forecasted costs plus underwriting margin component underlying the contracted rate. The argument goes that if the MCO has a loss, the contracted rate failed to fund all costs; failing to fund all costs is a failure of the actuary’s cost forecast, therefore the rates are not actuarially sound. But this conclusion is a non sequitur.

The main issue is timing. CMS determines actuarial soundness on a prospective basis before observing actual costs because the rates are prospective. No difference in forecasted to actual costs can directly challenge actuarial soundness.

If a cost outcome is implausible given its forecast, it might motivate an investigation into the actuary’s data, methods, or assumptions. Such an evaluation should be constrained to the information known at the time of developing the rate, so decisions and outcomes are separated.[5] If an error or negligence is found on the actuary’s part; if, during rate setting, the actuary was unaware of a reasonably knowable and in-effect or highly likely prospective program, benefit, or policy change; if a subsequent event comes to pass; or if some rate information is found to be erroneous—these are the kinds of events for reevaluating actuarial soundness.[6] Revising rates solely due to an MCO’s losses frustrates prospective capitation rate setting: to transfer risk to the MCO with the aligned financial incentive to control costs and utilization while preserving or enhancing beneficiaries’ health outcomes.[7]

Suppose capitation rates were only actuarially sound if the rate funds the MCO at least a profit. Any loss would require the actuary to update the rate to be at least actual costs. There would be no risk. Moreover, an MCO would be incentivized to maximize service volumes, lavishly spend on administration, and boost the proportional underwriting margin component on trued-up costs. This volume-maximizing scenario should sound familiar. It echoes the unfettered feeforservice incentive misalignment that was the genesis of wider managed care adoption to control costs.[8,9]

No one is obligated to reimburse the MCO’s losses. The National Association of Insurance Commissioners (NAIC) supplies states with standard formulas and reporting to help regulators monitor insurers’ solvency.

One NAIC tool is risk-based capital, a formula for calculating minimum amounts of capital surplus insurers should hold.[10] An MCO can bolster its surplus by retaining past profits, and the surplus lends strength to its balance sheet. If needed, it will not feel good to expend this surplus when losses occur.

That said, for the rating period, a rate can be actuarially sound and the MCO still loses money. It is a valid and statistically expected outcome in the long run.

Close the Flood Gates

What, then, can be done to mitigate losses?

CMS regulation allows states to use optional risk mitigation arrangements. These must be contracted prospectively, which means mitigating future potential losses and not already realized ones.[11] Medicaid managed care limits a state’s financial risk to mostly enrollment volume risk. The basic principle of any risk mitigation arrangement is how much of the remaining capitation risk the state would like to retain.[12]

States can use risk mitigation for any rate cycle, but some situations encourage its use. A major forecasting challenge is when historical data lacks samples of what the actuary is trying to model and forecast, such as the Affordable Care Act’s (ACA) Medicaid expansion for adults; COVID-19 effects before a vaccine is available or public health emergency enrollment policies (maintenance of effort or continuous coverage unwinding) are implemented; or estimating cell and gene therapy use before they come to market. These out-of-sample predictions increase the forecast’s volatility, which modifies the chance of profit or loss and exaggerates potential profits or losses.

Two common arrangements for these scenarios are non-risk payment arrangements and risk corridors. In a non-risk payment arrangement, the state retains the risk instead of putting it in the MCO’s contract. The MCO still processes the benefit but then invoices the state for its full cost. A similar non-risk approach is if the benefit is fully “carved out” and the state processes it through its fee-for-service program. The MCO is not involved except for applicable care management tied to the carved-out benefit.

Alternatively, risk corridors shift some of the risk to the state by transferring a percent of profit or losses between the state and MCO after an aggregate cost threshold is met. The state retains a gradient of risk. The state will receive some money back if the MCO has above-average profits. If the MCO has above-average losses, the state pays more to the MCO. CMS encouraged corridors and other arrangements during the ACA expansion, and again during COVID-19 rating periods; both examples of large out-of-sample forecasts.[13,14]

Risk mitigation arrangements create administrative work for all parties. Settling these arrangements takes data and time. Minders of state budgets would prefer to avoid having a potential payment due to the MCO after the end of the rating period or state budget. MCOs would like similar downside protection as states’ upside protection from remittance on minimum medical loss ratios.[15]

States have other ways to influence underwriting outcomes. Suppose a state averages around a 60%/40% probability that the MCO will profit. The state may want to modify that to 70%/30%, and the actuary could calculate how much additional margin for risk or contingency to add to the underwriting margin component. This calculation will depend on the volatility of the forecast. States could also share with the MCOs if they have a target or what the underwriting margin component amount is.

A loss in one year might be manageable, but sustained losses become problematic to the MCO and the state’s goals of competition, provider access, and beneficiary choice. CMS also allows states to adjust contracted rates ±1.5% for subsequent program changes within the rating period.[16] The state must submit the new rate to CMS in an updated contract but is exempt from including an updated certification from the actuary. As mentioned before, the actuary’s certification is part of CMS’s determination of actuarial soundness. Thus, states can use their ±1.5% independent of actuarial soundness because there is no updated certification for CMS to review. Yet, CMS may ask the state what benefit or population has changed to justify an adjustment. To ameliorate an MCO’s loss may be an invalid justification. However, CMS has shown flexibility in extraordinary cases. For example, despite championing prospective at-risk capitation, CMS has allowed states and actuaries an option to retrospectively adjust rates for the public health emergency unwinding.[17,18]

Transparency Tango

If the MCO and actuary discuss the rate, it would be better to happen before the state submits it to CMS. This conversation can be difficult but productive.

The state is often the actuary’s principal and sole intended user, per definition in ASOP No. 41, Actuarial Communications, of the actuarial finding and subsequent communications.[19] The state then decides what is shared with the MCO from the actuary. This sharing can look like information about rate assumptions and components, or a rate review presentation.

All stakeholders should also be mindful of the asymmetry of questions. Questions are good. They clarify or prompt a collective checking of what is discussed. However, sometimes the question is easy to ask but creates much work to answer.

Medicaid managed care capitation rate development is complex, so what are the salient data sets, and what assumptions, and methods to focus on are not always straightforward. Stakeholders are concerned that sharing something new will set a precedent—a perennial upkeep of work. But, if all stakeholders were more willing to share and display new information, with an understanding that they had no obligation to provide it again, this flexibility might create enough room to innovate and find statistics or charts that add discussion value while cycling out less valuable information.

All stakeholders should also be mindful of the asymmetry of questions. Questions are good. They clarify or prompt a collective checking of what is discussed. However, sometimes the question is easy to ask but creates much work to answer. Being patient with no answer or willing to forgo some curiosity may increase the overall volume of questions while easing the willingness of stakeholders to answer them.

Reenvisioning the fateful meeting, the MCO still had a loss, but actuarial soundness never comes up. The MCO has insight into the actuary’s data sets, assumptions, and methods from the current and upcoming rate forecasts. Questions and answers flow. And the state is aware of its options to influence the current and future MCO financial outcomes. Risk remains, but the misconception is gone.[20,21]

Endnotes

[1] “Medicaid Managed Care Underwriting Margin Model”; page 37; Society of Actuaries (SOA); July 2024.
[2] “Medicaid Managed Care Organizations: Considerations in Calculating Margin in Rate Setting”; page 23; SOA; 2017. The underwriting ratio median of 2% (or mean of 1.6%) is consistent with common underwriting margin components used in rate setting.
[3] The 2013–2022 profits are approximately normally distributed. However, instead of a single normal distribution, the profits are modeled better as a 22%/78% mixture of a normal distribution with mean 0% and standard deviation of 12% with another normal distribution of mean 2.0% and standard deviation 3.5% (this mixed distribution is displayed as the bolded average density in Figure 1). Given this mixed distribution, the calculated loss or profit of 38%/62% would be more like 33%/67%. For simplicity, the article will continue assuming a single normal distribution for underwriting ratio probabilities.
[4] ASOP No. 49, Medicaid Managed Care Capitation Rate Development and Certification; page 9; ASB; March 2015.
[5] “Are You a Bad Decision-Maker?”; Kozyrkov; May 2020.
[6] ASOP No. 41, Actuarial Communications; page 5; Actuarial Standards Board; December 2010. See “Subsequent Events.”
[7] “Managed Care”; Medicaid.gov.
[8] “Health Insurance and Managed Care, Fifth Edition”; page 121; Kongstvedt 2020.
[9] “Economics for Healthcare Managers, Fifth Edition”; page 46; Lee 2023.
[10] “Risk-based Capital Background”; NAIC.org. The NAIC also has surplus tiers that trigger different oversight actions.
[11] 42 CFR § 438.6(b)(1).
[12] States may have different motivations or philosophies on how much risk to retain. It is not always done out of the kindness of their hearts. Removing risk from a service may encourage its use, like making some cell and gene therapies non-risk. It may also make the state’s Medicaid program attractive for MCO procurement to foster competition or beneficiary choice.
[13] “2014 Managed Care Rate Setting Consultation Guide”; CMS; September 2013.
[14] “Medicaid Managed Care Options in Responding to COVID-19”; CMS; May 2020.
[15] 42 CFR § 438.8(j)” Remittance is optional.
[16] “42 CFR § 438.7(c)(3).”
[17] “In the proposed rule, we noted that we reached a similar conclusion in our review of SDP [state directed payment] proposals which use reconciliation of historical to actual utilization; if States are seeking to remove risk from managed care plans in connection with these types of SDPs, it is inconsistent with the nature of risk-based Medicaid managed care. As further noted in the 2016 rule, ‘[t]he underlying concept of managed care and actuarial soundness is that the [S]tate is transferring the risk of providing services to the MCO and is paying the MCO an amount that is reasonable, appropriate, and attainable compared to the costs associated with providing the services in a free market. Inherent in the transfer of risk to the MCO is the concept that the MCO has both the ability and the responsibility to utilize the funding under that contract to manage the contractual requirements for the delivery of services.’”; 2024-08085 Section P-866; Federal Register.
[18] “2023-2024 Medicaid Managed Care Rate Development Guide”; page 10; CMS; May 2023.
[19] “ASOP No. 41, Actuarial Communications”; page 6; ASB; December 2010. See the last paragraph of “Responsibilities to Other Users.” Under this standard, naming an intended user except the state creates obligations, including communications, that may violate the contract between a consulting actuary and the state. The author also sought guidance from the Actuarial Board for Counseling and Discipline (ABCD) on intended users. The ABCD was of the opinion that the communicating actuary has unilateral authority over who is an intended user. A non-intended user cannot become an intended user by any means except the actuary naming them as one.
[20] All analyses were performed using R Statistical Software (v4.3.1; R Core Team 2024); libraries: tidyverse, svglite, mixtools.
[21] The meeting story is a dramatization intended to be a structural arc to this article.

print
Next article Actuarial Professionalism Considerations for Generative AI
Previous article Spotlight on Tomorrow’s Leaders

Related posts