Context
Managed Care Models. For more than three decades, states have increased their reliance on managed care delivery systems with the aim of improving access to certain services, enhancing care coordination and management, and making future costs more predictable. Across the states, there is wide variation in the populations required to enroll in managed care, the services covered (or “carved in”), and the quality and performance incentives and penalties employed. Most states contract with risk-based managed care organizations (MCOs) that cover a comprehensive set of benefits (acute care services and sometimes long-term care), but many also contract with limited benefit prepaid health plans that offer a narrow set of services such as dental care, non-emergency medical transportation, or behavioral health services. A minority of states operate primary care case management (PCCM) programs which retain fee-for-service (FFS) reimbursements to providers but link enrollees with a primary care provider who is paid a small monthly fee to provide case management services in addition to primary care. While the shift to MCOs has increased budget predictability for states, the evidence about the impact of managed care on access to care and costs is both limited and mixed.5,6,7 In 2024, the Biden administration finalized major Medicaid managed care regulations designed to advance access and promote quality of care for enrollees. These rules are complex and set to be implemented over several years unless overturned or delayed by Congress or the Trump administration.
Capitation Rates and Risk Mitigation. MCOs are at financial risk for services covered under their contracts, receiving a per member per month “capitation” payment for these services. Capitation rates must be actuarially sound8 and are applied prospectively, typically for a 12-month rating period, regardless of changes in health care costs or utilization.9 States may use a variety of risk mitigation tools to ensure payments are not too high or too low, including risk sharing arrangements, risk and acuity adjustments, medical loss ratios (MLR), or incentive and withhold arrangements. When, however, significant enrollment, utilization, cost, and acuity changes began to emerge early in the COVID-19 public health emergency, CMS allowed states to modify managed care contracts, and many states implemented COVID-19 related “risk corridors” (where states and health plans agree to share profit or losses), allowing for the recoupment of funds. States and plans faced another period of heightened rate setting uncertainty when the continuous enrollment provision expired on March 31, 2023, resulting in acuity and utilization shifts within the remaining population that were difficult to predict.
Looking ahead, the 2024 Medicaid managed care rule requires states to incorporate all state directed payments (SDPs) through capitation rate setting adjustments instead of using “separate payment terms” (which provide payments outside of base capitation rates) beginning in July 2027.10 The 2025 federal budget reconciliation law (H.R.1) will also create rate setting challenges for states as the Medicaid provisions impacting enrollment and spending (e.g., work requirements, more frequent eligibility redeterminations, and provider tax and SDP caps and reductions) roll out over the next several years.
Prior Authorization and Artificial Intelligence (AI). MCOs often require patients to obtain approval of certain health care services or medications before the care is provided, an insurance practice commonly referred to as “prior authorization”. Subjecting a service or drug to prior authorization allows the MCO to evaluate whether the care is covered, medically necessary, and being delivered in the appropriate setting, but can also increase the administrative burden on providers and sometimes delay or limit access to care. To reduce administrative costs and processing times and increase consistency of decisions, health insurers are increasingly turning to AI to automate the processing of prior authorization requests. Using AI for this purpose, however, is drawing scrutiny due to concerns that poorly implemented AI can harm patients. In June 2025, the Department of Health and Human Services (HHS ) announced a voluntary initiative where dozens of health insurers pledged to reduce the burden of prior authorizations across insurance markets, including a commitment to expand “real time” responses to electronic prior authorization requests, which may involve increasing the use of AI. In July 2025, the Trump administration released an AI action plan, emphasizing the removal of regulatory “red tape” and enabling faster adoption of AI tools, and in 2026, the Administration plans to launch a new innovation model to test the use of technologies, including AI and machine learning, in the prior authorization review process for select Medicare services. In September 2025, the launch of the Safe AI in Medicaid Alliance was announced, bringing together 32 states and industry leaders to develop frameworks for AI adoption and use in state Medicaid programs.
This section provides information about:
- Managed care models
- MCO medical loss ratio (MLR) and remittance requirements
- Risk corridors
- MCO capitation rate amendments and rate setting challenges
- State oversight of MCO use of AI in prior authorization processes
Findings
Managed Care Models
Capitated managed care remains the predominant delivery system for Medicaid in most states. As of July 1, 2025, all states except five – Alaska, Connecticut,11 Maine, Vermont,12 and Wyoming – had some form of managed care (MCOs and/or PCCM) in place (Figure 2). As of July 1, 2025, 42 states13 were contracting with MCOs (unchanged from 2024); only two of these states (Colorado and Nevada) did not offer MCOs statewide (although Nevada plans to expand MCOs statewide in 2026). Thirteen states reported operating a PCCM program (with the addition of Missouri).14 Although not counted in this year’s report, following the passage of HB 345, Idaho expects to end its PCCM program by December 2025 and implement comprehensive MCOs by January 2029.
Of the 46 states that operate some form of comprehensive managed care (MCOs and/or PCCM), 33 states operate MCOs only, four states operate PCCM programs only, and nine states operate both MCOs and a PCCM program. In total, 28 states15 were contracting with one or more limited benefit prepaid health plans (PHPs) to provide Medicaid benefits including behavioral health care, dental care, vision care, non-emergency medical transportation (NEMT), or long-term care (LTC).
Capitation Rates and Risk Mitigation
Minimum Medical Loss Ratios (MLRs) and Remittance Requirements
The MLR reflects the proportion of total capitation payments received by an MCO spent on clinical services and quality improvement, where the remainder goes to administrative costs and profits. To limit the amount that plans can spend on administration and keep as profit, CMS published a final rule in 2016 that requires states to develop capitation rates for Medicaid to achieve an MLR of at least 85% in the rate year.16 There is no federal requirement for Medicaid plans to pay remittances to the state if they fail to meet the MLR standard, but states have discretion to require remittances. The 2024 Consolidated Appropriations Act included a financial incentive to encourage certain states to collect remittances from Medicaid MCOs that do not meet minimum MLR requirements. As state Medicaid programs faced heightened uncertainty due to the COVID-19 pandemic (2020) and the unwinding of the pandemic-era continuous enrollment provision (starting in 2023), analysis of Medicaid managed care market data (reported to the National Association of Insurance Commissioners) showed a decrease in the average Medicaid MLR in 2020 – 2022 compared with prior years, followed by an increase in 2023. More recent analysis suggests the average Medicaid MLR continued to increase in 2024. This year’s survey asked states whether they have a state required minimum MLR and whether they require MCOs that do not meet the minimum MLR requirement to pay remittances.
Nearly all MCO responding states (38 of 41) reported a minimum MLR requirement is always in place for MCOs as of July 1, 2025 (Figure 3). Among responding states, responses were unchanged/consistent with last year’s survey. While states must use plan-reported MLR data to set future payment rates so that plans will “reasonably achieve” an MLR of at least 85%, states are not required to set a minimum MLR for their managed care plans. If states set a minimum MLR requirement, it must be at least 85%.17 While most states that described their requirements reported a minimum MLR requirement of 85%, several states reported higher requirements that ranged from 86% to 93%. A few states noted that minimum MLRs may vary by program or population. For example, in Pennsylvania, the minimum MLR requirement is set at 85% for MCOs covering acute care only (hospital and physician services) and at 90% for MCOs that cover acute care and LTC. Similarly, New Jersey reported the minimum MLR requirement is set at 85% for non-LTC populations and 90% for LTC populations covered under MCO contracts. In Indiana, the minimum MLR requirement is set at 85% for MCOs that cover children and pregnant individuals, 87% for MCOs that cover ACA expansion adults, and 90% or higher for MCOs that cover more complex populations such as older adults (that may be receiving LTC) and people with disabilities.
More than three-quarters of responding MCO states (33 of 41) report they always require remittance payments when an MCO does not meet minimum MLR requirements (Figure 4). Thirty-three states reported that they always require MCOs to pay remittances, while three states indicated they sometimes require MCOs to pay remittances (among responding states, responses were generally consistent with last year’s survey18). States reporting that they sometimes require remittances may limit this requirement to certain MCO contracts. For example, Rhode Island reported that the remittance requirement did not apply to all populations.
Additionally, some states (North Carolina, Oregon, and Tennessee) give MCOs that fail to meet the state required minimum MLR the option to either remit funds to the state and/or use funds towards community reinvestments. California reported CMS requires its plans to pass MLR reporting and remittance requirements down to risk-bearing subcontractors.19 Five states do not require remittances (including two states that do not set a minimum MLR requirement). States that do not have minimum MLR and remittance requirements in place may have other risk mitigation strategies such as profit caps or experience rebates and/or risk corridors.
Risk Corridors
Risk corridors allow states and health plans to share profit or losses (at percentages specified in plan contracts) if aggregate spending falls above or below specified thresholds. Under two-sided risk corridors, states and plans may share in plan profits and losses. Although less common, some states may use “one-sided” risk corridors that apply only to profits or losses. Risk corridor thresholds may be tied to a target MLR. Risk corridors may cover all/most medical services (and enrollees) under a contract or may be more narrowly defined, covering a subset of services or enrollees. States may introduce risk corridors on a time-limited basis—for example, following the expansion of coverage to new groups (e.g., ACA Medicaid expansion adults). CMS encouraged states to implement two-sided risk mitigation strategies, including risk corridors, for rating periods impacted by the COVID-19 public health emergency. In 2023, nearly two-thirds of responding MCO states reported implementing a pandemic-related MCO risk corridor (in 2020, 2021, and/or 2022), leading to the recoupment of payments for many states. In this year’s survey, states were asked whether they were using risk corridors as a tool to protect the state and/or MCOs against risk of capitation rates significantly differing from actual experience for MCO contracts in place as of July 1, 2025.
Over two-thirds of responding MCO states (26 of 38) reported using risk corridors for MCO contracts in place as of July 1, 2025 (Figure 5).20 Some of the risk corridors that states described broadly apply to all/most populations and/or costs while other risk corridors apply to specific populations and/or a subset of costs. States frequently reported the use of multiple risk corridors. For example, Arizona reported using a two-sided medical risk corridor (for all programs) which includes benefit costs but excludes administrative costs and a two-sided risk corridor for fixed administrative costs for its largest program with the most population fluctuation (to ensure fixed costs are covered regardless of population fluctuations). California reported several risk corridors including a two-sided risk corridor for its new Enhanced Case Management (ECM) benefit, noting the potential variability (e.g., by plan and region) associated with the implementation and ramp up of ECM supports; a two-sided risk corridor for state directed supplemental payments for family planning services; and a two-sided risk corridor for a new federally qualified health center alternative payment model (APM) program. While the majority of risk corridors described by states are two-sided, at least three states (Nebraska, Washington, and West Virginia) reported using one-sided risk corridors for at least certain populations or MCO programs.
Rate Amendments and Rate Setting Challenges
State Medicaid programs use the most recent and accurate enrollment, cost, and utilization data available to ensure that MCO capitation rates are actuarially sound and that MCOs are not over-paid or under-paid for the services they deliver. Even if risk mitigation strategies are in place (e.g., MLR with remittance and/or risk corridors), states may determine rate amendments are necessary, for example, if their actual experience differs significantly from the assumptions used for the initial certified rates. During a contract rating period, states may increase or decrease rates by 1.5% per rate cell (which apply to population subgroups with one or more common characteristics such as age, gender, eligibility category, and geographic region) without seeking CMS approval for the change (different rules apply for states with approved rate ranges per cell).21 To make a larger change, states must submit a rate amendment for federal approval that addresses and accounts for all differences from the most recently certified rates.
During the unwinding of the pandemic-era Medicaid continuous enrollment provision, millions of people were disenrolled and states and plans faced considerable rate setting uncertainty. Higher member risk and utilization patterns began to emerge by late 2023, and many states sought federal approval to adjust rates to address these shifts in FY 2024. This year’s survey asked states whether they have or will seek CMS approval for a capitation rate amendment to certified rates to address unanticipated shifts in acuity and/or utilization in the rating period that began in FY 2025.
Half of responding MCO states (19 of 38) reported seeking CMS approval for a capitation rate amendment to address unanticipated shifts in acuity and/or utilization for a rating period beginning in FY 2025 (Figure 6). Of the 19 states that reported seeking rate amendments, nearly all reported that the amendment(s) resulted in an increase to capitation rates and about two-thirds reported that the changes applied retrospectively (i.e., adjusted capitation rates for a period that already passed).
During the unwinding period, state actuaries used a variety of approaches to account for changes in cost, utilization, and member acuity.22 This year’s survey included questions to better understand capitation rate setting challenges in the post-unwinding environment. Some states noted making significant changes to the process for developing actuarially sound capitation rates post-unwinding, including the incorporation of acuity adjustments and mid-year reviews of rates to determine if changes are appropriate.
Most responding MCO states reported experiencing or expecting to experience new or notable challenges setting capitation rates for rating period(s) that begin in FY 2026. Many of these states reported challenges due to higher acuity and utilization trends. Some states reported challenges with projecting future pharmacy trends and costs. A few states also mentioned rising medical costs (e.g., inpatient hospital costs) as well as state budgetary pressures and uncertainty. Many states anticipate challenges with projecting potential impacts of federal policy changes effective after FY 2026. This includes work requirements and more frequent eligibility redeterminations for expansion adults under the recently passed reconciliation law, which has implications for member enrollment and acuity (on average). Several states also mentioned challenges with calculating SDPs stemming from regulatory changes (e.g., the 2024 managed care rule’s prohibition on separate terms), and a few states mentioned uncertainty regarding the reconciliation law’s limits on SDPs.
Prior Authorization and Artificial Intelligence (AI)
While health insurers are increasingly using AI to automate parts of the prior authorization process, there is limited information available about its use and impact within Medicaid managed care. The Medicaid and CHIP Payment And Access Commission (MACPAC) found that while there are potential benefits of automation in prior authorization such as administrative efficiencies and faster processing times, it may also pose potential risks or challenges depending on how it is administered and monitored. In the absence of comprehensive federal policy governing AI use and oversight in prior authorization, some states have taken steps to regulate or monitor use of AI by health plans. A November 2024 report from the National Association of Insurance Commissioners highlighted that transparency to consumers, providers, and regulators is an important component of AI oversight. This year’s survey asked states whether the MCOs with which they contract use AI in their prior authorization processes as of July 1, 2025.23
Nearly half of responding MCO states (17 of 38) reported knowledge of at least some of the MCOs with which they contract using AI in their prior authorization processes as of July 1, 2025 (Figure 7). At least two states (Oklahoma and South Carolina) reported that AI is used only for prior authorization approvals (vs. use for denials/adverse determinations). Some states that did not report MCO usage of AI may not know or currently track this information.
Less than one-quarter of responding MCO states (7 of 38) reported requiring MCOs to disclose the use of AI in prior authorization processes. States were asked if they require MCOs to disclose the use of AI in prior authorization processes (to the state Medicaid agency, enrollees, and/or providers) as of July 1, 2025. Seven states (California, District of Columbia, Georgia, Indiana, Nebraska, Tennessee, and Virginia) reported requiring disclosure to the state Medicaid agency. Five of those states (District of Columbia, Indiana, Nebraska, Tennessee, and Virginia) indicated MCOs must submit a request to use AI to the state technology officer or Medicaid agency for review before implementation. Three states (California, Georgia, and Indiana) reported requiring disclosure to enrollees and providers.
State examples of AI disclosure requirements include:
- In California, MCOs are required to disclose the use and oversight of AI tools in their written utilization management policies and procedures. These documents must be made available to providers, enrollees, and the public upon request.
- Indiana has adopted an AI policy governing the use of AI technologies across all state agencies. In alignment with this policy, the state Medicaid agency requires plans to submit any AI tools or systems for formal review. Indiana’s State Agency AI Systems Standard requires MCOs to conduct a readiness assessment prior to implementation or use of any AI tool or system as well as annual follow-up or ad hoc assessments when significant changes are made to the AI tool.
- In Tennessee, MCOs are required to contact the state Medicaid agency’s AI Governance Committee when the use of AI is contemplated in any capacity. MCOs must share what vendor is being considered, what purpose the AI is serving, how outputs are being verified, what system risks and vulnerabilities exist, and how data is being safeguarded.
Many states reported concerns and challenges with the use of AI in MCO prior authorization processes. When asked to describe their top concerns or challenges (if any) with the use of AI in MCO prior authorization processes, states frequently cited potential for bias, improper denials, privacy and security risks, and inadequate human/clinician oversight. Some states also reported concerns with ensuring compliance with federal and state requirements, complexities related to oversight, and transparency of AI decision-making processes.
Several states reported implementing new or expanded oversight activities or adopting other safeguards in FY 2025 or 2026 to support appropriate use of AI in MCO prior authorization processes. For example, five states (California, Maryland, Nevada, New Hampshire, and Ohio) reported introducing or plans to introduce language in MCO contracts regarding the use of AI. Texas reported working to develop a standard process to review MCO AI tools prior to implementation.
