Medicaid Enrollment and Unwinding Tracker

Name: National Enrollment in Medicaid/CHIP, February 2020 to February 2026
Creator: KFF

Published: May 29, 2026

Section:

0 / 0

Enrollment Data

Note: The data presented below are updated monthly as new Medicaid/CHIP enrollment data become available.

The Medicaid Enrollment and Unwinding Tracker presents the most recent data on monthly Medicaid/CHIP enrollment reported by the Centers for Medicare & Medicaid Services (CMS) as part of the Performance Indicator Project as well as archived data on renewal outcomes reported by states during the unwinding of the Medicaid continuous enrollment provision. The unwinding data were pulled from state websites, where available, and from CMS.

Medicaid/CHIP Enrollment Trends

Medicaid/CHIP enrollment trends generally use February 2020 as the baseline month because it was the month prior to the start of the COVID-19 pandemic and implementation of the continuous enrollment provision. During continuous enrollment, which was in place during the three years of the pandemic, states paused Medicaid disenrollments. As a result, when the continuous enrollment provision ended in March 2023, national Medicaid/CHIP enrollment had increased to a record high of 94 million enrollees. Beginning April 1, 2023, states could resume disenrolling people after conducting renewals to verify eligibility for the program, though some states delayed the start of their unwinding periods until May, June, or July 2023. Most states took 12 months to complete unwinding renewals and nearly all states completed renewals by August 2024.

The figures below show Medicaid and CHIP enrollment from February 2020 through the most current month of available data. Some figures also include enrollment for adults and children in Medicaid/CHIP. Key enrollment trends as of February 2026 include:

There are 74.9 million people enrolled in Medicaid/CHIP nationally (Figure 1). This represents a 21% decline from total Medicaid/CHIP enrollment in March 2023, but is still 5% higher than Medicaid/CHIP enrollment in February 2020, prior to the pandemic (Figure 2 and Table 1).
Several factors likely explain why national Medicaid/CHIP enrollment is higher than pre-pandemic enrollment. The pandemic may have encouraged some people who were previously eligible for Medicaid but not enrolled to newly enroll in the program. During the unwinding, many states took steps to improve their renewal processes, which reduced the number of people who were disenrolled despite remaining eligible. In addition, some states expanded eligibility for certain groups since the start of the pandemic, such as the Affordable Care Act’s (ACA) Medicaid expansion.
Medicaid/CHIP enrollment is higher than pre-pandemic levels in all but nineteen states (AK, AZ, AR, CO, FL, ID, IA, LA, MA, MI, MT, NH, NM, RI, SC, TN, TX, VT, WV) and DC. Enrollment changes from pre-pandemic baseline vary from a 18% decrease in Montana to a 53% increase in North Carolina (Figure 2). Many of the states with the largest increases in enrollment expanded eligibility since the start of the pandemic. For example, five states (NE, OK, MO, SD, and NC) implemented the Medicaid expansion between October 2020 and December 2023 and Maine increased the income limit for children to qualify for Medicaid.
In the 49 states and DC with complete enrollment data by age, there are 35 million children (48%) and 38.3 million adults (52%) enrolled, a change from pre-pandemic (February 2020) enrollment patterns when children made up a slight majority (51%) of Medicaid/CHIP enrollees (Figure 1).
Child enrollment in Medicaid/CHIP is below pre-pandemic enrollment in 25 states, while adult enrollment is below pre-pandemic levels in 17 states and DC (Figure 2).
There are 67.7 million people enrolled in Medicaid and 7.2 million people enrolled in CHIP (Figure 1). More states report CHIP enrollment above their pre-pandemic baselines compared to the number reporting Medicaid enrollment above the baseline (Figure 2).

National Enrollment in Medicaid/CHIP, February 2020 to February 2026 (Line chart)

Cumulative Percent Changes in Enrollment from February 2020 to February 2026 (Column Chart)

Total Medicaid/CHIP Enrollment, Selected Time Periods (Table)

Unwinding Data - Archived

Note: The data on unwinding renewal outcomes presented below were last updated on September 12, 2024; since most states have now completed the Medicaid unwinding, the information will not be updated again.

As of September 12, 2024 and with nearly complete unwinding data for most states:

Over 25 million people were disenrolled (31% of completed renewals) and over 56 million people had their coverage renewed (69% of completed renewals).
Disenrollment rates varied across states from 57% in Montana to 12% in North Carolina, driven by a variety of factors including differences in renewal policies and procedures as well as eligibility expansions in some states.
Among those who were disenrolled, nearly seven in ten (69%) were disenrolled for paperwork or procedural reasons while three in ten (31%) were determined ineligible.
Among those whose coverage was renewed during the unwinding, 61% were renewed on an ex parte, or automated, basis, meaning the individual did not have to take any action to maintain coverage.

State Data on Renewal Outcomes

The data on unwinding-related renewal outcomes presented in this section rely primarily on monthly reports that states were required to submit to the Centers for Medicare & Medicaid Services (CMS) during the unwinding period. The data also reflect updates to the monthly reports that states submit three months after the original report submission to account for the resolution of pending cases and any other changes in renewal metrics. For 13 states, data were pulled from dashboards or reports published on state websites that provide more complete information, and for a few additional states, updated monthly reports were pulled from state websites because they were more timely than what is reported on the CMS website.

To view archived data for specific states, click on the State Data - Archived tab.

As of September 12, 2024, States Have Reported Renewal Outcomes for Nearly Nine Out of Ten People Who Were Enrolled in Medicaid/CHIP Prior to the Start of the Unwinding (Donut Chart)

Medicaid Disenrollments

As of September 12, 2024, at least 25,198,000 Medicaid enrollees had been disenrolled during the unwinding of the continuous enrollment provision. Overall, 31% of people with a completed renewal were disenrolled in reporting states while 69%, or 56.4 million enrollees, had their coverage renewed.
There is wide variation in disenrollment rates across reporting states, ranging from 57% in Montana to 12% in North Carolina. A variety of factors contribute to these differences, including differences in renewal policies and system capacity. Some states adopted policies that promote continued coverage among those who remain eligible and/or have automated eligibility systems that can more easily and accurately process renewals while other states have adopted fewer of these policies and have more manually-driven systems. In addition, North Carolina and South Dakota adopted Medicaid expansion and other states increased eligibility levels for certain populations (e.g., children, parents, etc.) during the unwinding, which may have lowered disenrollment rates in these states.

At Least 25,198,000 Medicaid Enrollees Have Been Disenrolled and 56,378,000 Have Had Their Coverage Renewed, as of September 12, 2024 (Stacked Bars)

Across all states with available data, 69% of all people disenrolled had their coverage terminated for procedural reasons. However, these rates vary based on how they are calculated (see note below). Procedural disenrollments are cases where people are disenrolled because they did not complete the renewal process and can occur when the state has outdated contact information or because the enrollee does not understand or otherwise does not complete renewal packets within a specific timeframe. High procedural disenrollment rates are concerning because many people who are disenrolled for these paperwork reasons may still be eligible for Medicaid coverage.

(Note: The first tab in the figure below calculates procedural disenrollment rates using total disenrollments as the denominator. The second tab shows these rates using total completed renewals, which include people whose coverage was terminated as well as those whose coverage was renewed, as the denominator. And finally, the third tab calculates the rates as a share of all renewals due, which include completed renewals and pending cases.)

Of All People Who Were Disenrolled, 69% Were Terminated for Procedural Reasons, as of September 12, 2024 (Stacked Bars)

Medicaid Renewals

Of the people whose coverage has been renewed as of September 12, 2024, 61% were renewed on an ex parte basis while 39% were renewed through a renewal form, though rates vary across states. Under federal rules, states are required to first try to complete administrative (or “ex parte”) renewals by verifying ongoing eligibility through available data sources, such as state wage databases, before sending a renewal form or requesting documentation from an enrollee. Ex parte renewal rates varied across states from 90% or more in Arizona, North Carolina, and Rhode Island to less than 20% in Pennsylvania and Texas.

Overall, 61% of People who Retained Medicaid Coverage Were Renewed Through Ex Parte Processes, as of September 12, 2024 (Stacked Bars)

Federal Data on Renewal Outcomes

The data presented here are cumulative unwinding metrics published by CMS. These counts and percentages may differ from the above data, which present renewal metrics reported on state websites when state-reported data are more complete.

Figure 1 below shows cumulative renewal data reported by CMS during states’ unwinding periods. Renewal data for the months after the end of states’ unwinding period are excluded. The data reflect updated unwinding data reported by states three months after the original monthly reports as they become available.

Cumulative Medicaid Renewal Outcomes for Reporting States Through August 2024 (Stacked Bars)

For questions about this tracker, please contact KFFTracker@kff.org

State Data - Archived

Note: The state data presented below were last updated on September 12, 2024; since most states have now completed the Medicaid unwinding, the information will not be updated again.

The data presented here provide state-level data on enrollment trends and renewal outcomes during the unwinding period. Figure 1 shows total Medicaid enrollment by month starting in January 2023 and, once disenrollments resumed in a state, the cumulative percent change in Medicaid enrollment relative to the month before Medicaid disenrollments started (this baseline month will differ across states). Figure 2 shows renewal metrics for each month of a state’s unwinding period (or cumulative data for the unwinding period for some states).

For total national Medicaid enrollment, click on the Enrollment Data tab.

Related Resources

Resources on unwinding data

Resources on state policies and preparations for the unwinding

Resources on pre-pandemic enrollment patterns and coverage transitions

KFF’s unwinding explainer

10 Things to Know About the Unwinding of the Medicaid Continuous Enrollment Provision

A Closer Look at North Carolina’s Implementation of the 2025 Reconciliation Law Medicaid Provisions and Other Changes Amid Medicaid Budget Shortfalls

Authors: Jada Raphael, Amaya Diana, and Anna Mudumala

Published: May 29, 2026

On April 30, 2026, North Carolina Governor Josh Stein signed legislation that includes Medicaid policy changes and closes an estimated $319 million shortfall in funding for the state’s Medicaid program for FY 2026. Many of the legislation’s Medicaid policy changes are related to implementation of the 2025 federal reconciliation law. The 2025 reconciliation law requires states to condition Medicaid eligibility for adults in the Affordable Care Act (ACA) Medicaid expansion group and enrollees in partial expansion waiver programs (Georgia and Wisconsin) on meeting work requirements starting January 1, 2027. The 2025 reconciliation law also limits states’ ability to raise the state share of Medicaid spending through provider taxes, restricts state-directed payments (SDPs) to hospitals, nursing facilities, and other providers, and increases barriers to enrolling in and renewing Medicaid coverage. As states are preparing to implement the reconciliation law provisions, many states are facing more tenuous budget situations with slowing revenue growth and broader reductions in federal funding.

This policy watch examines the current budget context in North Carolina, the state’s recently passed legislation, the state’s Medicaid Advisory Committee (MAC) meetings, and data from KFF’s Medicaid work requirements tracker to provide initial insight into how North Carolina is preparing to implement certain Medicaid provisions of the 2025 reconciliation law and how other policy changes may affect coverage and access to care. While some of the issues North Carolina is facing are unique to that state, others are likely to be faced by other states as they implement federal changes to Medicaid in the midst of other fiscal challenges.

What is the budgetary context as North Carolina prepares to implement the 2025 reconciliation law’s provisions?

North Carolina is facing a more tenuous fiscal climate like in other states, and state legislators have not yet enacted a comprehensive state budget for the FY 2025-27 biennium. In the past year, revenue volatility and rising costs have led to slowing state revenue growth following a period of record-breaking revenue and expenditure growth for states after the initial pandemic-induced economic downturn. In North Carolina, scheduled tax cuts have been projected to drive declines in state revenue, and debates over whether to proceed with the cuts have contributed to a budget stalemate in the legislature.

In August 2025, Governor Stein signed a stopgap funding bill that appropriated $600 million from the state general fund for Medicaid, but it left a $319 million shortfall for FY 2026 in funding for the cost of services for non-expansion (traditional) enrollees. The shortfall and budget stalemate led to rate cuts and the elimination of GLP-1 coverage, both of which were eventually restored. The Medicaid agency ceased “Healthy Opportunities Pilots” program services in FY 2026 due to a lack of appropriations. The pilots covered certain non-medical services that target social needs, including housing, nutrition, transportation, and interpersonal relationship supports to specific and limited enrollees, and evaluations of the “Healthy Opportunity Pilots” 1115 waiver showed lower costs over time and largely positive outcomes. The Medicaid agency also implemented changes to reduce administrative expenses, including reducing temporary staff and contractors, ending certain contracts, pausing quality improvement projects, and scaling back compliance and quality oversight activities. The legislation signed in April 2026 appropriated $319 million to close the shortfall for FY 2026 and made changes aimed at addressing financing pressures associated with new federal limits on provider taxes, which the state uses to help finance its Medicaid expansion and hospital state directed payment program (which increases payment rates for hospitals).

What are some of the Medicaid policy changes included in North Carolina’s recent legislation?

North Carolina’s new legislation includes more restrictive standards for how the state will implement work requirements than is required in current law. At a minimum, the 2025 reconciliation law requires states to look back one month immediately preceding the application month and one month between renewal periods to confirm compliance with the requirements. North Carolina’s legislation requires the state to confirm compliance for the three months preceding the application month. At renewal, the state must confirm compliance for at least three of the six months since the last determination of eligibility. The North Carolina legislation also prohibits the acceptance of self-attestation as the only evidence in verification of eligibility requirements (unless required by federal law or regulation, or pursuant to a court order). States await guidance from CMS as to whether self-attestation of medical frailty, parent/caretaker status, or other exemptions or work status can be accepted, but most states report plans to accept self-attestation if allowed.

The legislation increases the frequency of data checks to identify changes in circumstances for Medicaid enrollees from quarterly to monthly. The state will review information on earned and unearned income, employment status and changes in employment, residency status, enrollment status for other public assistance programs administered by the state and outside of the state, financial resources, incarceration status, and lottery and gambling winnings. States are required to follow up on reported changes that potentially affect eligibility and give individuals an opportunity to respond before taking adverse action. In North Carolina, when data indicates an individual is no longer eligible, enrollees only have 10 days in advance of case closure to submit documentation verifying ongoing eligibility. Increasing the frequency of periodic data checks with insufficient response times can lead to procedural disenrollments and exacerbate churn.

The Medicaid agency will be required to set Medicaid copayments at the highest allowable amounts for both traditional Medicaid enrollees and ACA expansion enrollees. Current federal rules limit cost sharing in Medicaid because of enrollees’ low income and limited ability to pay out of pocket costs. The maximum allowable cost sharing varies by type of service and enrollee income. North Carolina has set current cost sharing amounts, regardless of enrollee income, at $4 per service. Starting July 1, 2027, the legislation requires the Medicaid agency to increase current cost sharing amounts for services where the maximum allowable amount is more than $4 and to increase cost sharing for ACA expansion adults with income 100-138% FPL to up to 10% of the cost of the service, except for prescription drugs and non-emergency use of the emergency department. Beginning in October 2028, when states are required to implement mandatory cost sharing of up to $35 per service for ACA expansion adults with income between 100%-138% FPL, the state will be required to set cost sharing amounts at $35 per service, except for prescription drugs, for all non-exempt services for this group.

The legislative text implementing the changes to Medicaid eligibility for certain lawfully residing immigrants effectively ends the state’s long-standing optional Medicaid coverage for lawfully residing children and pregnant immigrants without a five-year waiting period. The law limits Medicaid coverage for immigrants to coverage that is required under federal law. However, North Carolina is one of 40 states that have taken up the option to extend Medicaid and/or CHIP coverage to children and/or pregnant adults who are lawfully residing and waive the five-year wait for these groups. The 2025 reconciliation law imposed additional eligibility restrictions for many lawfully present immigrants but allows states to maintain the option to cover lawfully residing children and pregnant adults. By limiting coverage for immigrants to only what is required by federal law, the state law effectively ends this optional coverage as of October 1, 2026. In a recent Medicaid Advisory Committee (MAC) meeting, the Medicaid agency indicated it was working with the legislature to make “corrections” and restore coverage for these populations.

The legislation requires the Medicaid agency to report certain Medicaid applicants and enrollees for whom it cannot verify citizenship or “satisfactory” immigration status to the Department of Homeland Security. These include applicants and enrollees who, after a reasonable opportunity period, have not verified satisfactory immigration status or whose final verification indicates that they do not have a satisfactory immigration status and are not lawfully present. This group would include those found ineligible based on immigration status and individuals receiving Emergency Medicaid services (where Medicaid pays hospitals for emergency care provided to ineligible immigrants who would otherwise be eligible for Medicaid based on their income).

Medicaid Financing

The legislation increases intergovernmental transfers (IGTs) from public hospitals, reducing reliance on the state’s hospital taxes for financing the nonfederal share of Medicaid spending. The 2025 reconciliation law imposes significant new restrictions on states’ ability to generate Medicaid provider tax revenue, including prohibiting all states from establishing new provider taxes or from increasing existing taxes, as well as reducing existing provider taxes for states that have adopted the ACA Medicaid expansion. North Carolina uses provider taxes to help finance the nonfederal share of Medicaid spending. State law requires the nonfederal share for the expansion program to be fully funded by certain non-general fund sources, including hospital taxes and hospital IGTs, and requires the end of expansion coverage if those sources cannot fully fund the nonfederal share. The state estimates $14.3 million in one-time administrative costs for the current state fiscal year and $44.4 million in recurring annual administrative costs (including both state and county expenditures) for the expansion program to implement work requirements and six-month eligibility redeterminations that existing financing mechanisms did not account for. The increased public hospital IGTs aim to offset the financing of some of the existing costs under the state’s hospital tax, as well as to help finance the new administrative costs.

By increasing reliance on IGTs as a financing source, the state may aim to retain higher hospital SDPs under new federal provider tax limits, but new federal requirements for state directed payments are expected to require further changes. North Carolina’s Healthcare Access and Stabilization Program (HASP), a hospital SDP program launched alongside Medicaid expansion in 2023, is also financed through hospital taxes and IGTs. An earlier state report indicated the new federal provider tax limits would eliminate all or most of HASP SDPs. The state has been using HASP payments to incentivize hospitals to relieve medical debt, and as of October 2025, more than $6.5 billion in debt had been relieved for more than 2.5 million North Carolinians under the initiative.

There is significant uncertainty about how federal regulations and state legislation may affect the state’s plan for financing the nonfederal share of Medicaid spending, including for the Medicaid expansion and HASP. New proposed rules on state directed payments and forthcoming provider taxes may affect the state’s financing plans. The state’s legislation created a “trigger” to end the new funding should HASP payments fall below certain thresholds or a change in federal law or regulation result in at least a 20% decrease to IGTs.

Medicare Advantage Out-of-Pocket Limits: Variation and Trends

The Average Medicare Advantage In-Network Limit is $5,421 in 2026, But Nearly 1 in 5 Enrollees Face In-Network Limits Higher Than $7,000

Authors: Nancy Ochieng, Meredith Freed, Jeannie Fuglesten Biniek, Juliette Cubanski, and Tricia Neuman

Published: May 28, 2026

For coverage of Medicare benefits, people face a choice between traditional Medicare and private Medicare Advantage plans. While there are many distinguishing features between these coverage types, one key benefit of Medicare Advantage is an annual cap on out-of-pocket costs for medical benefits, which traditional Medicare does not include. In 2026, the out-of-pocket limit for Medicare Advantage plans may not exceed $9,250 for in-network services and $13,900 for a combination of in-network and out-of-network services, but plans can have lower out-of-pocket limits than the maximum.

Policymakers have long considered whether to add a similar out-of-pocket cap in traditional Medicare. In 1988, Congress enacted a Medicare out-of-pocket cap, but the law was repealed one year later principally due to concerns about the financing mechanism. The Medicare Payment Advisory Commission (MedPAC) has recommended changes in traditional Medicare to give beneficiaries better protection against high out-of-pocket spending, including an out-of-pocket cap as part of a broader redesign that also includes combining Part A and Part B deductibles and simplifying cost sharing. Traditional Medicare remains the only major form of health insurance that does not include a cap on out-of-pocket spending, though most beneficiaries in traditional Medicare have additional financial protection through supplemental coverage, including Medicaid, employer- or union-sponsored retiree coverage, or a Medigap policy (which may require additional premiums).

This brief analyzes out-of-pocket limits in Medicare Advantage plans in 2026, variation by plan type, the distribution of enrollees facing different out-of-pocket limits, and trends over time. Of note, this analysis does not show the share of enrollees that reach their plan’s out-of-pocket limit in any year because spending data are not available. The analysis includes Medicare Advantage plans generally available for individual enrollment, reflecting coverage for 21.3 million Medicare beneficiaries in 2026, excluding special needs plans and employer-and union-sponsored group plans (See Methods for details).

Takeaways

In 2026, the average enrollment-weighted out-of-pocket limit for Medicare Advantage enrollees is $5,421 for in-network services and $9,825 for in-network and out-of-network services combined. The average out-of-pocket limit for in-network services is higher for preferred provider organizations (PPOs) ($6,592) than health maintenance organizations (HMOs) ($4,636).
Just over one in ten Medicare Advantage enrollees (13% or 2.8 million people) are in plans with limits of $3,000 or less for in-network services in 2026, more than two-thirds (68%) are in plans with limits above $3,000 and up to $7,000, and about one in five (19%) are in plans with limits above $7,000.
About one in 10 (9%; 1.8 million) Medicare Advantage enrollees are in plans with the maximum out-of-pocket limit for in-network services ($9,250). Among the 8.6 million enrollees in PPOs, roughly one in five (22%; 1.8 million) are in plans with the maximum out-of-pocket limit for in-and out-of-network services combined ($13,900).

The average out-of-pocket limit for in-network services decreased by nearly $600 from 2017 to 2023 ($5,253 to $4,685), before increasing by about $700 from 2023 to 2026 ($4,685 to $5,421).

In 2026, the Average Enrollment-Weighted Out-of-Pocket Limit for Medicare Advantage Enrollees Is Lower for HMOs Than PPOs

The average out-of-pocket limit faced by Medicare Advantage enrollees in 2026 is $5,421 for in-network services and $9,825 for in-network and out-of-network services combined (Figure 1). These averages are lower than the maximum allowable out-of-pocket limits ($9,250 and $13,900, respectively).

Differences in out-of-pocket limits may reflect how Medicare Advantage plans choose to allocate rebate dollars, which are extra payments they receive from the federal government beyond the cost of providing Part A and Part B services. Plans may use rebates to reduce beneficiary cost sharing, including lowering out-of-pocket limits, but they can also use these funds for other purposes, such as offering supplemental benefits not covered by traditional Medicare, offering a rebate against the Part B premium, or lowering Part D premiums.

In 2026, the Average Out-Of-Pocket Limits for Medicare Advantage Enrollees Are ,421 for In-Network Services and ,825 for In-Network and Out-Of-Network Services Combined (Bar Chart)

Out-of-pocket limits for in-network services vary by plan type. HMOs, which have 12.8 million Medicare Advantage enrollees in 2026, generally offer no coverage of services from out-of-network providers, but offer a lower out-of-pocket limit for in-network services than PPOs. PPOs, which have 8.6 million enrollees in 2026, cover services delivered by in-network and out-of-network providers, but require higher cost sharing for out-of-network providers. While PPO enrollees have broader access to out-of-network providers than HMO enrollees, they also face a higher out-of-pocket limit, even for in-network services. Specifically, the average enrollment-weighted out-of-pocket limit for in-network services is $6,592 for PPOs and $4,636 for HMOs.

The distribution of enrollment between HMOs and PPOs, as well as the average out-of-pocket caps for in-network services by type of plan, vary across states (Appendix Table 1). Due in part to PPOs (which have higher average in-network out-of-pocket limits than HMOs) comprising a larger share of enrollment in rural areas, the average out-of-pocket cap is about $800 higher for Medicare Advantage enrollees in rural areas than in urban areas ($6,078 vs $5,291).

In 2026, Roughly One in Ten Medicare Advantage Enrollees Are in Plans With Limits of $3,000 or Less for In-Network Services and About One in Five Are in Plans With Limits Above $7,000

Among all Medicare Advantage enrollees in individual plans, just over one in ten (13%, 2.8 million) are in plans with out-of-pocket limits of $3,000 or less for in-network services (Figure 2). Nearly all of these enrollees (99%) are in an HMO. More than two-thirds (68%, 14.4 million) of Medicare Advantage enrollees are in plans with out-of-pocket limits above $3,000 and up to $7,000. About one in five (19%, 4.1 million) are in plans with limits above $7,000, including 1.8 million enrollees who are in a plan with the maximum in-network out-of-pocket limit of $9,250. PPOs enrollees account for the majority of these enrollees at the maximum cap (66%).

Among the 8.6 million Medicare Advantage enrollees in PPOs, about one in ten (9%) have a combined limit for in-network and out-of-network services at or below $6,000. More than two-thirds (66%; 5.7 million) of enrollees are in plans with a combined limit for in-network and out-of-network services between $6,000 and $12,000. One quarter (25%; 2.1 million) are in plans with an out-of-pocket limit for in-and out-of-network services combined above $12,000, including 1.8 million enrollees in plans with the maximum out-of-pocket limit for in-and out-of-network services combined ($13,900) (Figure 2).

Roughly One in Ten Medicare Advantage Enrollees Are in Plans With Out-Of-Pocket Limits of ,000 or Less for In-Network Services and About One in Five Are in Plans With Limits Above ,000 (Column Chart)

Average Out-of-Pocket Limits for In-Network Services Generally Declined Between 2017 and 2023 but Have Increased Since Then

The average limit for in-network services decreased by nearly $600 from 2017 ($5,253) to 2023 ($4,685), before increasing by about $700 from 2023 to 2026 ($5,421) (Figure 3, Appendix Table 2). The average limit for in-network and out-of-network services combined has fluctuated over time but increased overall by about $750 between 2017 ($9,073) and 2026 ($9,825).

The maximum allowable out-of-pocket cap has generally increased over time, consistent with projected beneficiary out-of-pocket spending in traditional Medicare, which CMS uses to calculate maximum out-of-pocket limits in Medicare Advantage. However, the maximum cap decreased by $100 between 2025 and 2026. The average out-of-pocket limit faced by enrollees each year has always been lower than the maximum allowable limit. This gap has generally widened over time, but shrank somewhat between 2025 and 2026.

Average Out-Of-Pocket Limits in Medicare Advantage Declined Between 2017 and 2023 But Have Increased Since Then (Line chart)

Methods

This analysis uses data from the Centers for Medicare & Medicaid Services (CMS) Medicare Advantage Enrollment, Benefit and Landscape files for 2017-2026. The analysis excludes Special Needs Plans (SNPs), employer- and union-sponsored plans, PACE plans, and cost plans. These plans serve distinct populations and some may have different enrollment requirements than Medicare Advantage plans (e.g., may be available to beneficiaries with only Part B coverage) and in some cases, may be paid differently than Medicare Advantage plans, limiting comparability with individual Medicare Advantage plans for general enrollment. These exclusions are reflected in both current data as well as data displayed trending back to 2017.

The total number of enrollees in individual Medicare Advantage plans in this brief (21.3 million) may be slightly different than the number reported in other KFF briefs because this analysis further excludes a small number of plans without an in-network out-of-pocket amount specified in the plan benefits files.

The average for PPOs in this analysis includes two types of plans: local PPOs, which cover individual or multiple counties, and regional PPOs, which cover an entire state or multiple states. The average for HMOs includes two types of plans: HMOs that primarily cover services provided by in-network providers only and therefore do not have a limit for out-of-network services, and HMOs that are Point-of-Service plans (HMO-POS), which allow out-of-network care for certain services but typically charge higher cost sharing than for in-network services.

In previous years, KFF’s analysis of average out-of-pocket limits in Medicare Advantage excluded HMO-POS because they represented a relatively small share of HMO enrollment at the time (e.g., 10% in 2017). However, HMO-POS enrollment has grown substantially and now accounts for nearly half (46%) of HMO enrollment in Medicare Advantage. As a result, these plans are included in the current analysis to better reflect the experience of a substantial share of Medicare Advantage enrollees in HMOs.

This analysis determines urban and rural analysis based on the 2024 Urban Influence Codes (UIC) published by the U.S. Department of Agriculture (USDA) Economic Research Service. See Methods of KFF, “Key Facts About Medicare Beneficiaries in Rural Areas” (June 2025) for more details. Connecticut is excluded from the analysis by rurality because of differences in FIPS codes in the CMS Medicare Advantage data and the USDA 2024 UIC.

Appendix Tables

Average Out-of-Pocket Limits in Medicare Advantage by State, 2026 (Table)

Average Out-of-Pocket Limits and Maximum Out-of-Pocket Limits, 2017-2026 (Table)

VOLUME 47

May 28, 2026

Hantavirus Outbreak Revives COVID-Era False Health Claims

Joel Luther , Hagere Yilma , and Irving Washington

Highlights

A hantavirus outbreak linked to a Dutch cruise ship in early May was followed by false health claims that mirror patterns documented in previous outbreaks, including unsupported claims that ivermectin is an effective treatment, that the outbreak was planned in advance, and that it was caused by COVID-19 vaccines.

The Monitor also examines a new analysis of Americans’ relationship with health and wellness influencers, finding that most who get health information and advice from them express skepticism about what they hear.

What We’re Watching

As Hantavirus Cases Emerged, So Did Familiar False Claims About Causes and Cures

False health claims tend to follow recognizable patterns across outbreaks, including distrust of official sources, promotion of unproven treatments, and accusations of hidden profit motives. The hantavirus outbreak linked to the Dutch cruise ship MV Hondius was no exception, with several familiar false narratives spreading alongside the official response from the World Health Organization (WHO) and other agencies:

Unproven Treatments: Within hours of the first headlines, a Texas otolaryngologist who became a prominent promoter of ivermectin during COVID-19 posted without evidence on May 6 that ivermectin “should work” as a hantavirus treatment because it “is a RNA virus,” claiming that ivermectin “blocks RNA viruses” from replicating in the cell nucleus. The claim overstates available evidence from some laboratory and animal studies about ivermectin’s effect on virus replication. There is no evidence that ivermectin would be effective against hantavirus, which replicates outside of the nucleus, and there is no approved vaccine or antiviral treatment for hantavirus. Former U.S. Representative Marjorie Taylor Greene shared the post, recommending ivermectin, vitamin D, and zinc to her followers as potential hantavirus treatments. The following day, the same physician offered that she would sell ivermectin without a prescription to her followers in Texas.
False Claims that the Outbreak Was Caused by COVID-19 Vaccines: A Pfizer document listing adverse events that had been monitored during COVID-19 vaccine trials was also misrepresented as evidence that the vaccines caused hantavirus, a claim that was also made about COVID-19 itself. The document listed conditions that researchers designated in advance as worth monitoring closely during trials, not events that were observed or confirmed to have occurred. Hantavirus pulmonary infection appeared on that list, but its inclusion reflects standard safety surveillance, not causation.
The “Plandemic” Narrative: While less widespread than other false health claims about the outbreak, some social media accounts framed the outbreak as a “plandemic” and evidence of a depopulation campaign, reviving narratives that circulated during COVID-19. Their posts misleadingly presented early-stage research into a potential hantavirus vaccine, announced in 2024, as proof that the outbreak was planned in advance and referred to potential vaccines as dangerous or as primarily vehicles for pharmaceutical companies to profit. In reality, research into hantavirus vaccines has been ongoing for decades because it is a known pathogen with no available vaccines.

Why This Matters: The narratives that accompanied the hantavirus outbreak are not new. Researchers and fact-checkers have documented nearly identical claims in response to COVID-19, mpox, and avian flu, often with the same accounts recycling framing across new outbreaks. Each cycle may make it harder for accurate health information to reach audiences before these false narratives spread.

Most Americans Who Follow Health Influencers Are Skeptical of What They Hear

A new Pew Research Center analysis offers a detailed look at who is producing and consuming health and wellness content on social media. Four in ten U.S. adults, and half of those under 50, say they ever get health and wellness information from social media influencers or podcasts. Yet among those audiences, trust is limited. Just one in ten say they trust all or most of what they hear from these sources, while nearly a quarter (24%) say they trust not too much or none of it. The majority, about two-thirds, say they trust some. While most (54%) say the information has helped them better understand how to be healthy, about one in eight (12%) say it has left them more confused.

Of the nearly 7,000 health and wellness influencers Pew identified on YouTube, TikTok, and Instagram, about 41% describe themselves as some kind of health care professional, a category that includes not only physicians and nurses but also chiropractors, naturopaths, and functional medicine practitioners. The remaining health and wellness influencers describe themselves as coaches, entrepreneurs, or offer little biographical information at all. Most people who get content from influencers say they come across this content passively rather than seeking it out. Certain groups, including Black, Hispanic, and Asian Americans, and those without health insurance, are more likely to turn to health influencers for information, a finding that may point to gaps in the formal health care system as a driver of influencer engagement.

The findings add context to KFF polling, which found that 55% of the public use social media to find health information or advice at least occasionally, and 14% report getting health advice regularly from social media influencers. But trust in influencers was limited in the KFF survey as well: about four in ten (39%) of those who regularly got health information or advice from influencers said those influencers were primarily motivated by serving the public interest, compared to six in ten (61%) who said they were primarily motivated by their own financial interests. As KFF President and CEO Drew Altman wrote in a “Beyond the Data” column last August, the relatively small share of people who say they regularly get health information from influencers suggests that health communicators should keep the role of influencers in perspective, at least for now.

A Closer Look at the State Level: Nearly Half of California Residents Distrust the Health Care System, But Most Trust Providers

As the 2026 midterms approach, health information and trust dynamics vary across states, shaped by local politics, demographics, and access. The Monitor will periodically examine state-specific data and trends as part of our broader tracking of the health information environment.

A survey of California residents conducted by NORC at the University of Chicago for the California Health Care Foundation (CHCF) found that 46% say they have “not much” trust in the health care system or “none at all.” But, trust varies sharply depending on who within the system is being evaluated. Nine in ten Californians say they trust nurses and more than eight in ten say they trust their personal doctor, while fewer than half (49%) trust hospital administrators and roughly a third trust health insurance companies (33%) or pharmaceutical companies (30%). The survey also found that Californians who had skipped care due to cost were significantly less likely to trust the system overall, with 35% of that group expressing at least a fair amount of trust, compared to 54% of Californians overall.

These findings are in line with past KFF polling showing that doctors and health care providers are consistently the public’s most trusted source of health information. More recently, KFF’s latest April Health Tracking Poll finds that most of the public (70%) say they trust doctors and health care providers at least “a fair amount” to act in the public’s best interest, a sentiment that is shared across partisanship.

AI Legal Roundup

Pennsylvania Lawsuit Challenges AI Chatbots Presenting as Licensed Medical Professionals

The state of Pennsylvania has filed a lawsuit against Character.AI, alleging that its platform enabled chatbots to present themselves as licensed medical professionals and provide medical advice without proper credentials. According to the complaint, the chatbot claimed to be a psychiatrist licensed in Pennsylvania and offered mental health guidance, including discussion of diagnosis and treatment options, while using a fabricated license number. State officials argue this constitutes the unlawful practice of medicine. The company has said that its characters are fictional and accompanied by disclaimers indicating they are not real professionals, but the case raises questions about the effectiveness of such safeguards when users engage with highly personalized AI systems in vulnerable moments. The lawsuit arrives as the safety and accuracy of AI chatbots in the context of mental health is drawing increased scrutiny. KFF’s March 2026 Tracking Poll on Health Information and Trust found that about one in six adults had used AI chatbots in the past year for mental health information and advice, and a majority (58%) of those who did said they did not follow up with a mental health professional.

AMA Calls for Legal Protections Against AI Deepfakes Impersonating Physicians

The American Medical Association (AMA) released a new policy framework calling for stronger legal protections against AI-generated “deepfakes” that impersonate physicians through fake videos, audio, or images. The group warned that manipulated content falsely showing doctors endorsing treatments or giving medical advice could mislead patients, damage trust in physicians, and spread unproven or harmful health information, particularly as AI-generated misinformation becomes more realistic and harder to detect, even by trained medical professionals.

The AMA’s proposal calls for explicit opt-in consent before a physician’s likeness, voice, or identity can be used in AI-generated content, mandatory labeling and digital watermarks for synthetic media, and faster takedown and enforcement mechanisms. It also calls for shared responsibility among hospitals, platforms, and AI companies for preventing impersonation. The policy reflects growing concern within medicine about how generative AI tools could be used to exploit the credibility of health professionals online.

More From KFF

KFF Issue Brief: Federal Vaccine Injury Compensation Programs: Overview and Current Issues
KFF Quick Take: Supreme Court Leaves Mifepristone Access Unchanged – For Now
KFF Health News: A Danish Couple’s Maverick African Research Finds Its Moment in RFK Jr.’s Vaccine Policy
KFF Health News: Trump and Kennedy Seek To Relax Safeguards for AI Healthcare Tools

About The Health Information and Trust Initiative: the Health Information and Trust Initiative is a KFF program aimed at tracking health misinformation in the U.S., analyzing its impact on the American people, and mobilizing media to address the problem. Our goal is to be of service to everyone working on health misinformation, strengthen efforts to counter misinformation, and build trust.

View all KFF Monitors

The Monitor is a report from KFF’s Health Information and Trust initiative that focuses on recent developments in health information. It’s free and published twice a month.

Support for the Health Information and Trust initiative is provided by the Robert Wood Johnson Foundation (RWJF). The views expressed do not necessarily reflect the views of RWJF and KFF maintains full editorial control over all of its policy analysis, polling, and journalism activities. The data shared in the Monitor is sourced through media monitoring research conducted by KFF.

News Release

Drew Altman, Founding President and CEO of KFF, Announces Retirement Plans; Board Appoints Larry Levitt and Mollyann Brodie as Next Leadership Team

Dr. Altman has led the organization for nearly 40 years; Levitt and Dr. Brodie to assume leadership roles next year, after a combined six decades at KFF in senior positions guiding policy research and polling

Published: May 27, 2026

After nearly 40 years shaping the national health policy landscape and leading it through pivotal debates, KFF announced today that Dr. Drew Altman will retire from his role as Founding President and Chief Executive Officer on December 31. KFF’s Board of Trustees has appointed Larry Levitt and Dr. Mollyann Brodie as the new leadership team. Beginning in 2027, Levitt and Dr. Brodie will assume the leadership positions of CEO and President, respectively.

Since Dr. Altman founded the modern-day KFF, coming to California in 1990 to establish it from what once was the Henry J. Kaiser Family Foundation, the organization has become the nation’s leading source of health policy analysis, polling, and news. Under his leadership, KFF has evolved into a one-of-a-kind information organization, bringing together policy research, polling and survey research, and journalism, most recently through the creation of KFF Health News, the nation’s largest health policy newsroom. There is no organization in the country that has played a more pivotal role in informing policymakers, the media, and the public with data-centered analysis about U.S. health policy.

“Building KFF over the last three plus decades has been a mission for me,” said Dr. Drew Altman, Founding President and Chief Executive Officer of KFF. “My whole purpose has been to build an institution that could be a force for people and for truth in health care, and we’ve achieved that and more. Together, we have built an organization that combines policy research, polling, and news all in one, a new kind of information organization. I’m immensely proud of that legacy and KFF’s role at the highest levels of health policy. Larry and Molly each play a central role as KFF’s executive team now and have for years, and I have complete confidence in their leadership, in KFF’s remarkable staff who do our work, in our Board, and in KFF’s future.”

“With two small grandkids in Sacramento and Atlanta and another on the way, after what will be almost 37 years of very hands-on leadership since I founded KFF, and with KFF at the apex of its effectiveness, and the opportunity for a perfect transition, the clock in my head is telling me it’s time,” Dr. Altman added.

“KFF is the most trusted and respected voice in health policy, in large part because of Drew Altman and his vision and leadership. During my time in the U.S. Senate, the organization’s data and research were invaluable in shaping key health care decisions, underscoring the impact of KFF’s work, and the organization continues to play a unique and monumental role today,” said former U.S. Senator Olympia Snowe, Chair of KFF’s Board of Trustees. “Recognizing this is a pivotal moment for KFF, the Board engaged in a lengthy succession planning process, which was deliberative and thoughtful. Based on our governance considerations, the Board identified two exceptional leaders, Dr. Mollyann Brodie and Larry Levitt, who not only know this organization, but who have helped it grow and evolve into what it is today. We therefore selected Larry and Molly to serve as CEO and President, respectively. As we look to the future of KFF, Molly and Larry’s deep experience, understanding of KFF, and their unwavering commitment to its mission make them the best team to guide us into an exciting next chapter.”

Under Dr. Altman’s leadership, KFF has been a cornerstone of the nation’s health policy and health journalism landscapes—anchoring pivotal debates with rigorous, nonpartisan data analysis and reporting. Its work has informed understanding of landmark policies, including key provisions of the Affordable Care Act, Medicaid, and Medicare, and provided essential context during moments of crisis, such as the COVID-19 pandemic, when its COVID Vaccine Monitor and reporting became vital resources used around the country. Since the 1990s, KFF’s data and analyses have helped illuminate the real-world implications for people of proposals like Medicaid block grants, ensuring that policymakers and the public alike are equipped with credible, evidence-based insights.

Following the recommendation of a search committee led by Board Vice-Chair Jim Canales, KFF’s Board unanimously appointed Levitt and Dr. Brodie as incoming CEO and President, respectively. Both have served as Dr. Altman’s executive team for over a decade. Levitt, who will take over as CEO, currently serves as Executive Vice President for Health Policy, overseeing KFF’s policy analysis of programs including Medicare, Medicaid, the Affordable Care Act, women’s health, racial equity, global health, and more. Dr. Brodie currently serves as Executive Vice President and Chief Operating Officer, overseeing KFF’s operations and leading its nationally recognized Public Opinion and Survey Research Program, and recently received the American Association of Public Opinion Research’s Award for Exceptionally Distinguished Achievement for her professional accomplishments in the field. She will assume the role of President. Levitt and Dr. Brodie will both serve on the Board.

“We are grateful for Drew’s vision in establishing KFF as a leading source for independent and nonpartisan health policy information and look forward to continuing that mission,” said Dr. Molly Brodie, incoming President of KFF, and Larry Levitt, incoming Chief Executive Officer of KFF, in a joint statement. “We appreciate the vote of confidence in us by KFF’s Trustees and are excited to work together to lead an organization of immensely talented and professional staff.”

“With ongoing debates over affordability and other important health policy issues, KFF will continue to bring to the table timely and credible facts through our trusted research and journalism,” said Levitt. “We will build on the credibility KFF has established as an independent and nonpartisan source of health policy information in an increasingly polarized environment.”

“KFF will continue our laser focus on how health policies and the health care system affect real people,” said Dr. Brodie. “We are committed to leading KFF as a trusted source of information in an era of declining trust in public and private health care institutions.”

Dr. Altman will continue to work closely with Dr. Brodie and Levitt to ensure a smooth transition. He will continue in his role as President and CEO through the end of 2026, with Levitt and Dr. Brodie assuming their new leadership roles in January 2027.

Contact

Ann DeFabio

AnnD@kff.org

Global COVID-19 Tracker

Published: May 27, 2026

Editorial Note: The Policy Actions tracker will no longer be updated as the data source has ceased tracking government responses to COVID-19. For more information, please visit the Oxford Covid-19 Government Response Tracker.

Section:

0 / 0

Cases and Deaths

This tracker provides the cumulative number of confirmed COVID-19 cases and deaths, as well as the rate of daily COVID-19 cases and deaths by country, income, region, and globally. It will be updated weekly, as new data are released. As of March 7, 2023, all data on COVID-19 cases and deaths are drawn from the World Health Organization’s (WHO) Coronavirus (COVID-19) Dashboard. Prior to March 7, 2023, this tracker relied on data provided by the Johns Hopkins University (JHU) Coronavirus Resource Center’s COVID-19 Map, which ended on March 10, 2023. Please see the Methods tab for more detailed information on data sources and notes. To prevent slow load times, the tracker only contains data from the last 200 days. However, the full data set can be downloaded from our GitHub page. While the tracker provides the most recent data available, there is a two-week lag in the data reporting.

Note: The data in this tool were corrected on March 18, 2024, to clarify that they represent new cases and deaths over a full week rather than the average per day over a seven-day period.

Policy Actions

This tracker contains information on policy measures currently in place to address the COVID-19 pandemic. Policy categories currently being tracked include social distancing & closure measures, economic measures, and health systems measures. Policies are tracked at the country-, income-, and region-level. Please see the Methods tab for more detailed information on data sources and notes.

As countries continue to implement policies to prevent the transmission of SARS-CoV-2, the virus that causes COVID-19, these tables and charts show which social distancing and closure measures are currently in place by country.

Economic Measures

The COVID-19 pandemic has placed an unprecedented strain on country economies. These tables and charts show which economic-related measures, namely income support and debt relief, are currently in place by country.

Health Systems Measures

The COVID-19 pandemic continues to strain and disrupt global health systems. These tables and charts show which health systems measures are currently in place by country.

Methods

Cases and Deaths

SOURCES

As of March 7, 2023, all data on COVID-19 cases and deaths are drawn from the World Health Organization’s (WHO) Coronavirus (COVID-19) Dashboard. Prior to March 7, 2023, this tracker relied on data provided by the Johns Hopkins University (JHU) Coronavirus Resource Center’s COVID-19 Map, which ends on March 10, 2023. Population data are obtained from the United Nations World Population Prospects using 2021 total population estimates. Income-level classifications are obtained from the latest World Bank Country and Lending Groups. Regional classifications are obtained from the World Health Organization.

Policy Actions

NOTES

Policy actions data include the measure that was in place for each indicator at the country-level as of the end of 2022. Policy actions data will no longer be updated as the data source has ceased tracking government responses to COVID-19. For more information, please visit the Oxford Covid-19 Government Response Tracker.

Social Distancing and Closure Measures

Under 'Stay At Home Requirements', exceptions for leaving the house may include anything from being able to leave for daily exercise, grocery shopping, and essential trips, to only being allowed to leave once a week, or one person may leave at a time, etc. Under 'Workplace Closing', partial closing includes instances in which a country recommends closing the workplace (or working from home); businesses are open but with significant COVID-19-related operational adjustments; or when workplaces require closing for only some, but not all, sectors or categories of workers. Under 'School Closing', partial closing includes instances in which a country has recommended school closures; all schools are open but with significant COVID-19-related operational adjustments; or some schools, but not all, are closed; full closing includes schools that are in session but operating virtually. Under 'Restrictions On Gatherings', partial restrictions include restrictions on gatherings of more than 10 people; full restrictions include restrictions on gatherings of 10 people or less. Under 'International Travel Controls', partial restrictions include screening and quarantine requirements for those entering the country. Values for ‘Cancel Public Events’ were not recodified.

Economic Measures

Under 'Income Support', narrow support includes instances in which a country's government is replacing less than 50% of lost salary (or if a flat sum, it is less than 50% median salary); broad support includes instances in which a country's government is replacing 50% or more of lost salary (or if a flat sum, it is greater than 50% median salary). Under 'Debt/Contract Relief', narrow support includes instances in which a country's government is providing narrow relief, such as relief specific to one kind of contract.

Health Systems Measures

Under 'Vaccine Eligibility', partial availability includes availability for some or all of the following groups: key workers, non-elderly clinically vulnerable groups, and elderly groups, or for select broad groups/ages. Under 'Facial Coverings', recommend/partial requirement includes instances in which a country's government recommends wearing facial coverings, requires facial coverings in some situations, and requires facial coverings when social distancing is not possible.

SOURCES

Data on and descriptions of government measures related to COVID-19 provided by the Oxford Covid-19 Government Response Tracker (OxCGRT). For more detailed information on their data collection and methodology, please see their codebook and interpretation guide.

The Medical Frailty Exemption from Medicaid Work Requirements: Key Issues to Watch for in Upcoming CMS Guidance

Authors: Amaya Diana and Jennifer Tolbert

Published: May 27, 2026

The 2025 reconciliation law requires states to condition Medicaid eligibility for adults in the Affordable Care Act (ACA) Medicaid expansion group and enrollees in partial expansion waiver programs (Georgia and Wisconsin) on meeting work requirements starting January 1, 2027 or sooner at state option. The law specifies mandatory exemptions, including individuals who are “medically frail.” To ease the burden on individuals, the law directs states to use available information “where possible” to verify compliance with Medicaid work activities or exemption status, without requiring additional documentation from individuals. Given the abbreviated implementation timeline, states are moving forward with key decisions over how to implement the medical frailty exemption even as they wait for formal guidance from the Centers for Medicaid and Medicare Services (CMS), which is required to issue an interim final rule by June 1, 2026. This brief describes early state plans to operationalize the medical frailty exemption and identifies key issues that they are facing and may be answered in the forthcoming guidance.

What is the Medical Frailty Exemption?

The reconciliation law requires states to exempt from work requirements an individual who is “medically frail or otherwise has special medical needs.” The law specifies this includes individuals who are blind or disabled; have a physical, intellectual, or developmental disability that limits their ability to perform one or more activities of daily living (ADL); have a substance use disorder or a “disabling” mental disorder; and those with “serious or complex” medical conditions. This definition closely aligns with an existing federal medical frailty definition that CMS uses for states choosing to set “alternative benefit plans” that differ from the traditional Medicaid benefit package. In that context, CMS used the definition as a minimum standard for medical frailty but allows states flexibility to define medical frailty beyond the statutory and regulatory definitions. To identify individuals who are medically frail, CMS noted in the 2013 final rule that it expected states to use Medicaid claims data and encouraged them to use self-attestation and health screeners for applicants and new enrollees where the state does not have information on their current health status or historic encounter data.

How do States Plan to Operationalize the Medical Frailty Exemption?

Most states have not yet finalized a medical frailty definition, likely reflecting ongoing uncertainty over how much flexibility states will have as they await June CMS guidance. In a recent KFF survey (fielded January-March 2026), about half of states (22) subject to work requirements report having a current state medical frailty definition, though it is unclear whether these definitions align with the medical frailty provisions in the reconciliation law (Figure 1). Most of these definitions were not developed with work requirements in mind and may need to be updated to reflect the new federal requirements. Thirty-three states indicated they had not yet determined what definition they plan to use at the time the survey was fielded. If states are given flexibility by CMS to define who is medically frail and exempt from work requirements, six states would prefer to use a state definition (either an existing definition or a new definition,) while four states would use a federal definition.

Medically Frail Definition States Plan to Use (if CMS Allows Flexibility), March 2026 (Choropleth map)

States plan to use a variety of methods to verify medical frailty status, including using data to automate the process where possible. Most states reported plans to use Medicaid claims data (32) to verify medical frailty exemption status, while Georgia indicated it would not use claims data, and the remaining ten states had not yet made a decision at the time of the survey. New applicants and recent enrollees may not have claims data on file, requiring states to use other forms of verification. All 32 states reporting plans to use claims data also reported plans to use other sources, such as data from other programs (e.g., enrollment in a behavioral health managed care plan or HCBS program), managed care utilization or claims data, and/or managed care case management data. Twenty-nine states said they will seek confirmation from a treating provider. Many states are developing health assessment screeners to collect information to identify medical exemptions, and 11 states said they will use the health screeners to verify medically frail status. Most states (30) also reported wanting to allow applicants and enrollees to self-attest to their medically frail status if verification data are not available and self-attestation is permitted by CMS. However, some states reported they do not plan to accept self-attestation (AR, HI, IN, MT) or have since passed legislation barring the use of self-attestation (NC and UT).

Verification Sources States Plan to Use for Medical Frailty Exemptions, March 2026 (Stacked Bars)

Nebraska, which became the first state to implement new work requirements on May 1, 2026, has released resources outlining how it will process medically frail exemptions, as well as an index of diagnosis and procedure codes. Examples of conditions covered in the thousands of codes include types of cancer, HIV, diabetes, heart disease, and certain types of mental health conditions and substance use disorders, among others. For existing enrollees, Nebraska is reviewing medical claims data prior to renewal and will auto-exempt an individual if indicated based on diagnosis/procedure codes. If Nebraska cannot determine medical frailty through claims data or other data sources, existing enrollees will need to submit a self-declaration form. Not all individuals will be able to be auto-exempted due to limitations of claims data such as:

A lack of data on file for new applicants and recent enrollees;
A lack of data on file due to providers not consistently coding or not using the codes identified by the state;
Selected diagnoses and procedures potentially not being comprehensive of all individuals with target conditions or not capturing functional limitations;
A lack of claims within the look-back period despite an individual having an exempted chronic condition or disability; and
Delays between clinical care and claims presenting.

The state also intends to use the self-declaration form to identify medically frail exemption status for new applicants. The self-declaration form requires individuals to provide a description of their health condition, the contact information of their treating provider, and a note if services were provided while enrolled in Medicaid. No additional documentation is required at this time (e.g., a note from a provider or other proof of a condition). Nebraska indicates that it plans to move from self-declaration to more automated verification methods consistent with federal guidance over time, and plans to continue to review and update the index of ICD-10, CPT, and HCPCS codes as federal direction is clarified.

Early insights from other states highlight the variety of ways states are planning to use multiple sources of information to identify medical frailty. For example:

California plans to automatically exempt individuals who are eligible for certain programs (e.g., home care—also known as home- and community-based services or HCBS) and has been exploring more timely data sources such as managed care plan care management systems in conjunction with evaluating which diagnostic codes could be used to establish medical frailty.
Pennsylvania has outlined a plan to first check health care claims, followed by information from electronic health records exchanged through regional health information organizations, managed care organization (MCO) health assessments and case management information, and finally information provided by the applicant. Pennsylvania explains its approach is designed with data limitations in mind, and aims to reduce paperwork burden and streamline the process for beneficiaries.
South Dakota has shared early plans to use clinical code lists and a medically frail health screener. Indicating lingering uncertainty on CMS’s stance regarding self-attestation, the state notes it may have to use physician attestation at renewal if screeners cannot be repeated and claims data is still not available.
Several states also have reported that they would be interested in adding new categories such as homelessness to the medical frailty category but understand from communications with CMS that states may not have flexibility to do so.

How Will the CMS Guidance Address Key Issues and Potential Implications?

Although states are moving forward with operationalizing the medical frailty exemption, the interim final rule that CMS is required to release by June 1, 2026, is expected to address key issues related to defining and verifying medical frailty. Because of the time required to make system changes, states have been moving forward with implementing work requirements based on informal communication and guidance CMS has shared. Any significant changes to that informal guidance could require states to change course and could have implications for states’ readiness to implement work requirements as well as the costs associated with implementation. Key issues that may be addressed in the guidance, and that states report waiting on, include:

Defining Medical Frailty

Defining medical frailty. CMS may provide a federal definition of medical frailty in the guidance. The law suggests the need for a Secretary-provided definition in its list of exemptions by including anyone “who is medically frail or otherwise has special medical needs (as defined by the Secretary).” It is not clear if CMS will rely on the existing definition or develop a new definition that more directly relates to work requirements. The statute includes people who are both medically frail and those with “serious or complex medical conditions,” which includes people with significant medical needs that may or may not limit their ability to engage in qualifying activities. For these individuals maintaining access to health coverage protects against serious health consequences if treatment is interrupted and may enable them to work.
State flexibility to define medical frailty. Whatever federal definition CMS establishes, another important consideration is whether the definition serves as a minimum standard that states can use or expand on or whether CMS will require states to use the federal definition with little or no flexibility to interpret how it will be applied. For example, CMS may specify a set of diagnoses codes for states to use in their claims data analysis, but the effect of that list will depend greatly on whether states are allowed to add additional codes to CMS’ list. This may be particularly relevant in states that use state-specific diagnosis or treatment codes in their Medicaid billing systems.
Identifying people with functional limitations and mental health conditions. As noted, the law required people “with a physical, intellectual or developmental disability that significantly impairs their ability to perform 1 or more activities of daily living” be exempted. Guidance may provide additional insight into how states are expected to capture functional limitations in medical frailty definitions, as it may prove more difficult to identify these individuals using claims data. While there is a diagnostic code that could help to capture this, it is generally underutilized, with one barrier being a lack of financial incentive. More comprehensively identifying enrollees with these functional disabilities may require technical assistance for states and providers (i.e. in terms of utilizing different codes). Guidance may also provide clarity on identifying individuals with mental health conditions who may also be difficult to identify using claims data, as well as provide clarity on the exact diagnoses that would qualify.

Verifying Medical Frailty

Data sources for verifying medical frailty. States will likely be expected to use claims data to automate identification of medically frail individuals. The guidance may provide information on other data sources states can use or would be expected to use, such as electronic health records, MCO health assessments and case management information, and pharmacy data. States may also be required to use SNAP data and other program data, such as HCBS enrollment or enrollment in a behavioral health managed care program, to identify medically frail individuals. Accessing some of these data sources may raise privacy concerns that the guidance may or may not address.
Data look-back. The guidance may also specify the length of time states are permitted to look back to identify individuals with qualifying conditions when using data or verification from providers, including whether there are different look-back periods permitted at application versus renewal. The length of the look-back period will affect how many individuals are captured under the medically frail exemption (longer look-back periods are likely to pick up more people who may qualify as medically frail).
Exemption re-verification. The legislation does not currently clarify how frequently medical frailty exemptions may last before states are expected to re-verify individuals’ exemption status. States may need to reverify exemption status at every renewal, or states may be able to create an internal flag and/or permanently exempt some people if their health condition or disability status is unlikely to change.
Self-attestation. It is unclear if states will be permitted to accept self-attestation or if they will be prohibited from using it (altogether or in certain circumstances). Some state officials have noted the limitations of existing data sources, particularly claims data, for identifying new applicants who are medically frail as well as enrollees at their first renewal. Self-attestation could be subject to CMS audits, and guidance may outline under what circumstances states are expected to use self-attestation (e.g., only initial application, at renewal). If self-attestation is permitted, questions remain whether attestation alone will suffice or whether enrollees will be required to also provide supplemental information (e.g. provider sign-off, clinical or pharmacy records, etc.).
Health screeners. States may look to use health screeners at application and renewal, especially in the absence of self-attestation. States may have flexibility to create their own health screeners, or there may be a federal template.
Confirmation from providers. When using provider verification, the guidance may specify any requirements, including what information must be collected and whether providers will be required to assess whether an individual’s condition limits their ability to work. Relying on provider confirmation could increase administrative burden (on the clinical workforce, individuals, and states), particularly for providers that treat large shares of Medicaid patients, and, depending on the information requested, could raise ethical concerns among providers.
Exemption hierarchy. When individuals qualify under multiple exemptions or when they would qualify for a medical frailty exemption but also are engaged in qualifying work activities, states will need to develop standards for how to operationalize which status is prioritized (e.g., checking exemption status and work status in a specified order, choosing to use whichever is longer lasting), especially if certain conditions will have longer exemption periods. The guidance may establish a hierarchy for states to follow or may give states the ability to set their own hierarchies.

Key Facts About Health Care Affordability for People With Medicare

Authors: Nancy Ochieng, Meredith Freed, Juliette Cubanski, Jeannie Fuglesten Biniek, and Tricia Neuman

Published: May 27, 2026

Health care costs and affordability recently topped the public’s list of economic anxieties in the U.S., according to KFF polling. Even people with insurance say they struggle to afford health care costs, including people with Medicare. While Medicare provides health insurance coverage to 70 million people age 65 or older and younger adults with long-term disabilities, having Medicare coverage does not insulate beneficiaries from health-related affordability challenges, such as not getting needed medical care due to costs or incurring medical debt. In a 2026 KFF Health Tracking Poll, half (49%) of all Medicare beneficiaries ages 65 and older say they expect their health care costs to become less affordable in the next year.

This brief presents key facts and analysis about affordability of health care costs among people with Medicare, including younger adults with long-term disabilities, drawing on data from various sources (see methods for additional information).

Many Medicare Beneficiaries Have Low Incomes and Modest Savings, Which Can Make It Hard To Afford Medical and Long-Term Care Services

One in four Medicare beneficiaries (16.5 million people) had income below $24,600 per person in 2024, or about 160% of the federal poverty level that year (Figure 1). These estimates take into account income from Social Security, pensions, retirement account (IRA) withdrawals, and other sources. At the upper end of the income spectrum, the top 5% of Medicare beneficiaries (3.3 million people) had incomes above $169,700 per person. Many beneficiaries also have relatively low levels of savings, with one in four Medicare beneficiaries having savings below $18,950 per person in 2024. Financial resources are even lower among some subgroups of beneficiaries, including Black and Hispanic beneficiaries. For example, one in four Black beneficiaries had income below $20,150 per person in 2024, while one in four Hispanic beneficiaries had income below $14,150 per person.

One in Four Medicare Beneficiaries Lived on Incomes Below ,600 Per Person and Half Lived on Incomes Below ,200 Per Person in 2024 (Line chart)

About 6 million people ages 65 and older (10%) were living in poverty in 2024 under the official poverty measure, meaning they had income of $15,050 or less that year (if single), while 17.3 million (28%) had incomes below 200% of poverty. Poverty rates among older adults are higher under an alternative measure (the supplemental poverty measure, or SPM) that accounts for certain expenses that are higher among older people, including out-of-pocket medical expenses, with 15% of older adults living in poverty in 2024 under the SPM.
In 2025, nearly a quarter (23%) of all Medicare beneficiaries who received Social Security benefits relied on Social Security income for 90% or more of their total per capita income, while about one-third (32%) relied on Social Security for at least 75% of their income (Figure 2). In 2025, the average per capita Social Security income was $20,168 among those who depended on Social Security for at least 90% of their income and $20,670 among those who depended on Social Security for at least 75% of their income.

Figure 2

Medicare Premiums and Other Health Expenses Consume a Sizeable Portion of Income and Household Budgets for People With Medicare

Out-of-pocket health care spending for Medicare premiums and out-of-pocket costs for health care services accounted for 36% of Medicare beneficiaries’ average Social Security income per person in 2023 (Figure 3). In 2023, Medicare beneficiaries spent an average of $6,459 out of pocket on health care costs and had average Social Security income of $17,718 (though most older adults have other sources of income beyond Social Security that can be used to pay their out-of-pocket health care costs). Out-of-pocket spending includes premiums for Medicare supplemental coverage (Medigap) for those who enroll in traditional Medicare and buy private supplemental insurance.

Figure 3

Health care spending accounted for 14% of Medicare beneficiaries’ total household spending in 2024. Medicare households spent a larger share of their total household spending on health care than non-Medicare households in 2024 (6%) (Figure 4).

Health Care Accounted for a Larger Share of Total Household Spending for Medicare Households Than for Non-Medicare Households in 2024 (Stacked Bars)

More than 7 million Medicare beneficiaries enrolled in Medicare Part B in 2024 spent more than 10% of their annual per capita income on Part B premiums alone. The standard Part B premium is paid by beneficiaries in both traditional Medicare and Medicare Advantage and has roughly doubled in the last decade, from an annual amount of $1,259 in 2015 to $2,435 in 2026.

The Choice Between Traditional Medicare and Medicare Advantage Has Cost Implications for Beneficiaries

Out-of-pocket spending for Medicare-covered hospital and physician services is not capped in traditional Medicare, while Medicare Advantage plans are required by law to provide an out-of-pocket limit for these services. In 2025, the average out-of-pocket limit for Medicare Advantage enrollees was $5,320 for in-network services and $9,547 for in-network and out-of-network services combined. The maximum allowed limits were $9,350 for in-network services and $14,000 for in-network and out-of-network services combined in 2025, and increased to $9,250 and $13,900, respectively, in 2026. (Figure 5). This cap protects Medicare Advantage enrollees from unlimited health care costs for Medicare-covered services, but cost-sharing requirements can lead to high out-of-pocket expenses before reaching a plan’s annual limit. People enrolled in Medicare Advantage plans typically pay no additional premium.

In 2025, the Average Out-Of-Pocket Limits for Medicare Advantage Enrollees Were ,320 for In-Network Services and ,547 for In-Network and Out-Of-Network Services Combined (Bar Chart)

Medicare Advantage plans can reduce cost sharing, offer benefits not covered under traditional Medicare (such as dental and vision services), and lower Part B and Part D premiums, using extra payments they receive from the federal government beyond the cost of providing Part A and Part B services. In 2026, these extra payments (known as rebates) amounted to about $2,660 per person on average in 2026. For Medicare-covered services under Part A and Part B, Medicare Advantage plans can charge different amounts than what someone in traditional Medicare would pay, though they often reduce cost sharing for Medicare-covered benefits, and they are prohibited from charging more for certain services, such as Skilled Nursing Facility (SNF) stays. Plans are required to provide coverage that is at least as generous overall as traditional Medicare, even though what enrollees pay out of pocket can vary depending on the type of plan, the type and quantity of services they use, and whether they see a provider that participates in the plan’s network.

Medigap limits traditional Medicare beneficiaries’ exposure to out-of-pocket medical costs, and the most popular Medigap policies cover virtually all cost sharing for Medicare-covered services. However, Medigap underwriting rules make these policies difficult, if not impossible, for beneficiaries with pre-existing conditions to obtain outside of limited enrollment opportunities. For beneficiaries with modest incomes, the monthly Medigap premium ($217 per month, on average, in 2023) could be unaffordable (Figure 6).

The Average Monthly Medigap Premium Across All Current Medigap Policyholders Was 7 and Varied Across States, Ranging From 1 in Alaska to 7 in New York (Choropleth map)

More than 3 million Medicare beneficiaries in traditional Medicare have no additional coverage that helps with Medicare cost-sharing requirements, such as Medigap, employer coverage, or Medicaid. This leaves them at risk of facing high out-of-pocket costs if they need a lot of medical services, or high-cost services, because there is no limit on out-of-pocket spending for Parts A and B in traditional Medicare.

Paying for Health Care, Including for Services Not Covered by Medicare, Can Be a Financial Burden

Medicare beneficiaries can face high out-of-pocket costs for dental, hearing, and vision care services, which are not covered under traditional Medicare (Figure 7). In 2023, traditional Medicare beneficiaries spent $1,107 on dental services and $564 on hearing services, on average (Figure 7). While most Medicare Advantage plans cover dental, hearing, and vision benefits, Medicare Advantage enrollees also face out-of-pocket costs for these services, spending an average of $571 on dental services and $212 on hearing services.

Traditional Medicare Beneficiaries Can Face High Out-Of-Pocket Costs for Services Not Covered by Medicare (Bar Chart)

About one in five (22%) Medicare-age adults reported having some type of debt as a result of medical or dental bills in 2022 (Figure 8). Among older adults with medical debt, about 4 in 10 say they cut back on other household spending and used up or all or most of their savings as a result of their health care debt.

One in Five Adults Age 65 and Older Report Experiencing Debt Due to Medical or Dental Bills (Bar Chart)

Long-term services and supports, which are not covered by Medicare, are unaffordable for all but the highest-income Medicare beneficiaries. In 2025, the median annual costs of common long-term services and supports in the U.S were $80,080 for full-time non-medical caregiver services (including home health aide and homemaker services for people who need help with instrumental activities of daily living such as preparing meals and doing laundry), $129,575 for a private room in a nursing home, and $305,760 for round-the-clock home health aide services (Figure 9). These costs greatly surpass median income ($43,200) and savings ($110,100) among people with Medicare. Long-term care services are generally not covered by Medicare, regardless of whether a beneficiary is enrolled in traditional Medicare or Medicare Advantage.

Median Annual Costs of Common Long-Term Services and Supports, Generally Not Covered by Medicare, Exceed the Median Income of Medicare Beneficiaries (Column Chart)

Medicaid Helps Make Medicare Affordable for About 12 Million Low-Income Medicare Beneficiaries

In 2025, about 12 million people with Medicare also had Medicaid coverage (“dual-eligible individuals”).
- More than seven in ten (72%) dual-eligible individuals, or 8.5 million people, are eligible for the full range of Medicaid benefits not otherwise covered by Medicare, including long-term services and supports, vision and dental services, and non-emergency medical transportation (also known as wraparound services) (Figure 10).These “full-benefit” dual-eligible individuals also usually receive additional help through the Medicare Savings Programs (MSPs), which cover Medicare premiums, and in most cases, cost sharing for people with limited income and assets.
- The remaining 3.4 million dual-eligible individuals (“partial-benefit” individuals) do not receive coverage of the full range of Medicaid benefits, but do receive payments to cover Medicare premiums, and, in most cases, cost sharing through the Medicare Savings Programs. Under federal guidelines, Medicare beneficiaries with income up to 135% of the federal poverty line and assets below specified levels ($9,950 for an individual and $14,910 for a couple in 2026) qualify for financial assistance from their state Medicaid program through a Medicare Savings Program. States can choose to cover Medicare beneficiaries with income or assets above the federal limits, and in 2026, 18 states do so. Without financial support from Medicaid, the Part B premium alone would consume 15% of income for a dual-eligible individual with monthly income of $1,350, not including other out-of-pocket costs.

Medicaid Helps Make Medicare Affordable for About 12 Million Medicare Beneficiaries, Most of Whom Are Eligible for Wraparound Services Not Covered by Medicare (Donut Chart)

Medicare Part D Offers Protection Against High Out-of-Pocket Prescription Drug Costs, but Prescription Drug Costs Remain a Concern for Older Adults

Virtually all Medicare beneficiaries (96%) use prescription drugs and most are enrolled in Medicare Part D drug plans for prescription drug coverage. On average, Medicare Part D enrollees who did not receive low-income subsidies spent an average of $872 on prescription drugs in 2023, and people in fair or poor health spent roughly twice as much, on average ($1,671) (Figure 11). The Part D low-income subsidy (LIS) provides substantial financial support to income below 150% of poverty (under $23,940 for an individual in 2026) and limited assets (under $18,090 for an individual). Part D enrollees with LIS had average out of pocket drug costs of $156 in 2023, roughly one fifth the amount paid by those without these low-income subsidies.

Medicare Part D Enrollees Who Did Not Receive Low-Income Subsidies Spent an Average of 2 on Prescription Drugs in 2023, and Those in Fair or Poor Health Spent Roughly Twice as Much (Split Bars)

Under changes made by the Inflation Reduction Act, Part D enrollees have a cap on out-of-pocket drug spending equal to $2,100 in 2026. Prior to this change, enrollees could have faced several thousand dollars in out-of-pocket costs for expensive medications each year. The Inflation Reduction Act also includes other changes that make prescription drugs more affordable for people with Medicare, including a $35 monthly cap on insulin, expanded eligibility for full benefits under the Part D Low-Income Subsidy, and $0 cost sharing for adult vaccines under Medicare Part D. Additionally, the law requires the federal government to negotiate lower drug prices for some high-cost drugs covered under Medicare through the Medicare Drug Price Negotiation Program.
Even with a cap on overall out-of-pocket costs, prescription drug costs are a concern for Medicare beneficiaries, with half (50%) of Medicare beneficiaries age 65 and older worried about being able to afford prescription drugs for themselves or their family in 2026, according to a March 2026 KFF health tracking poll.

The Growth in Medicare Spending Will Lead to Higher Medicare Premiums, Deductibles, and Copayments Over Time

People with Medicare will face higher premiums and other out-of-pocket costs for Medicare benefits over time since Medicare premiums and deductibles are directly tied to increases in spending on Medicare-covered services. Higher Medicare spending is driven in part by rising prices, higher utilization, and growth in spending on Medicare Advantage. Between 2026 and 2034, the Medicare Part B premium and deductible are projected to increase by roughly 70%—from $2,435 to $4,170 for the Part B premium (annualized) and from $283 to $486 for the Part B deductible, while the Part A deductible is projected to grow by about 30% (from $1,736 to $2,212) (Figure 12).

People With Medicare Are Projected to Face Rising Cost Sharing, With the Part B Premium and Deductible Projected to Increase by About 70% From 2026 to 2034 (Line chart)

Rising Medicare premiums and cost sharing will consume a growing portion of Medicare beneficiaries’ financial resources. On average, premiums and cost sharing for Medicare Part B and Part D accounted for 25% of the average Social Security benefit in 2025, and this share is projected to increase to 33% in 2040. Including other premium and cost-sharing amounts would increase these shares even further, though as noted earlier most older adults can tap other sources of income beyond Social Security to pay for health care expenses.

Methods

This brief draws on data from various sources. Data on income and savings of Medicare beneficiaries, including the share of beneficiaries who relied on Social Security for at 75% or 90% or more of their total per capita income and the number of Medicare beneficiaries who spent more than 10% of their annual per capita income on Part B premiums, are drawn from the Urban Institute’s Dynamic Simulation of Income Model (DYNASIM4). See Methods of KFF, “Income and Assets of Medicare Beneficiaries in 2024” (August 2025) for more details.

Data on out-of-pocket health care spending are from the Centers for Medicare & Medicaid Services (CMS) Medicare Current Beneficiary Survey, 2023 Cost Supplement File (the most recent year of data available). For the analysis on average out-of-pocket spending as a share of average per capita Social Security income, see the methods of KFF, “Health Costs Consume a Large Portion of Income for Millions of People with Medicare” (August 2025) for more details.

The 2024 Consumer Expenditure Survey (CE) by the Bureau of Labor Statistics is used to assess the financial burden of health care spending among households where all members are covered by Medicare (referred to as Medicare households) compared to households where no members are covered by Medicare (referred to as non-Medicare households). The CE is a survey of households (“consumer units”), excluding people residing in institutions such as long-term care facilities. The estimates presented in this analysis are averages for demographic groups of consumer units, not per capita estimates, and thus are not comparable to estimates based on other surveys that report per capita estimates, such as out-of-pocket health care spending reported in the Medicare Current Beneficiary Survey.

Estimates of Medigap premiums are based on KFF analysis of Medicare Supplement Market Data from Mark Farrah Associates Health Coverage Portal^TM, 2023. See methods of KFF, “Key Facts About Medigap Enrollment and Premiums for Medicare Beneficiaries” (October 2024) for more details.

Data on annual Medicare premiums, deductibles, and other cost sharing over time are based on KFF analysis of the 2025 Annual Report of the Boards of Trustees of the Federal Hospital Insurance and Federal Supplementary Medical Insurance Trust Funds.

This work was supported in part by The John A. Hartford Foundation. KFF maintains full editorial control over all of its policy analysis, polling, and journalism activities.

Tracking Key Mental Health and Substance Use Policy Actions Under the Trump Administration

Authors: Nirmita Panchal and Heather Saunders

Published: May 26, 2026

In 2024, over 61 million adults in the U.S. experienced a mental illness and deaths due to suicide, gun violence, and drug overdose remained high. Additionally, the COVID-19 pandemic and necessary public health responses exacerbated an already existing mental health and substance use crises. At the same time, many people experience difficulties affording mental health treatment or finding providers. Among insured adults who described their mental health as fair or poor, 43% reported at least one time in the past year when they needed mental health services or medication but did not receive them; some groups – including communities of color, youth and young adults – experience greater barriers.

Many policy actions were initiated in response to these rising mental health and substance use concerns. During the first Trump administration, the SUPPORT Act – legislation that expanded access to opioid treatment and overdose prevention – was passed along with legislation that created the 988 crisis hotline. During the following Biden administration, federal policies focused on expanding coverage, improving access to care, implementing evidence-based treatments, and strengthening support for federal agencies, such as the Substance Abuse and Mental Health Administration (SAMHSA). Recent data shows that some opioid and mental health related indicators have stabilized or improved.

The second Trump administration, beginning in 2025, marked a change in federal mental health and substance use policy. The administration moved toward a heavier law-and-order approach and simultaneously narrowed the scope of federal leadership capacity in mental health and substance use services, while also continuing some treatment-focused initiatives (such as the SUPPORT Act reauthorization). Many of these policy directions are consistent with themes highlighted in President Trump’s campaign materials and are aligned with proposals in Project 2025.

This tracker lists and briefly describes key actions during President Trump’s second term, organized into the following four broad categories: Opioids (for example, signing the HALT Act); Mental Health (e.g., canceling school-based mental health grants); Federal Infrastructure/Data/Guidance (e.g., proposals to reduce and reorganize SAMHSA under another agency); and Gun Violence (e.g., rescinding community violence intervention grants). It will be updated as new changes occur. This tracker is not meant to be exhaustive; other state and federal policy changes may also affect mental health and substance use but are not captured here.

The tracker can be viewed in the order that each mental health or substance use policy action was implemented. Alternatively, the tracker can be filtered by category (Mental Health; Opioids/Substance Use Disorder; Federal Infrastructure/ Data/Guidance; and Gun Violence).

The Business of Health with Chip Kahn

AI at Scale: Does It Deliver?

May 26, 2026

Video

Transcript

Audio

Listen On:

Apple

Spotify

YouTube Music

About this Episode

Episode 5, AI Series: How is AI applied to clinical care and hospital operations across a real health system at full scale? Chip Kahn talks to Dr. Michael Schlosser, Senior Vice President and Chief Transformation Officer at HCA Healthcare, about how AI is developed for everyday use, starting with careful testing and customization, with clinicians and nurses engaged from the very beginning as end users.

The Host

Chip Kahn

Sr. Visiting Fellow

Charles N. Kahn III is a senior visiting fellow at KFF. He is also a visiting senior fellow at the American Enterprise Institute and a nonresident senior scholar at the University of Southern California’s Schaeffer Center for Health Policy & Economics. He serves as co-chair of the international Future of Health collaborative.

Guest

Michael Schlosser, MD, MBA

Senior Vice President and Chief Transformation Officer at HCA Healthcare

Michael Schlosser, MD, MBA, is the Senior Vice President and Chief Transformation Officer at HCA Healthcare. Previously, he served as Senior Vice President for Care Transformation and Innovation, launching HCA Healthcare’s initial digital transformation initiatives. Before that, he was Group Chief Medical Officer, National Group, overseeing clinical operations for 100 HCA Healthcare hospitals with a focus on quality, patient outcomes, and clinical strategy. He has also held the role of Chief Medical Officer at HealthTrust.

A neurosurgeon by training, Dr. Schlosser completed his residency and fellowship at Johns Hopkins. He has served as a medical officer with the FDA, holds a degree in chemical engineering from the Massachusetts Institute of Technology and earned his Master of Business Administration from Vanderbilt University.

Transcript

AI Usage Disclosure: This transcript was created with assistance from AI tools. It was reviewed and edited by KFF Staff.

Chip Kahn: Today we move from the wide angle to the front line. We turn to the actual application of AI to clinical care and hospital operations across a real health system at full scale. My guest is Dr. Michael Schlosser, Senior Vice President and Chief Transformation Officer at HCA Healthcare. 190 hospitals; 2,500 ambulatory sites; more than 47 million patient encounters a year. No other private sector operator is that large and no one else deploys AI at that scale. What you will hear is how AI in health care actually gets developed for everyday use. It starts with careful testing and customizing with clinicians and nurses engaged as end users from the very beginning. Then comes the path from a promising pilot to enterprise-wide deployment. And then comes what is often overlooked, the work that begins after deployment. When navigation, feedback and course correction take over. You cannot get away from the miraculous technology of AI. But where the rubber hits the road, two things matter. The first is data. Its breadth, its integrity, the infrastructure that holds it. To borrow a decades old phrase, it’s the data, stupid. The second is the human factor. What determines in the end whether a solution actually works to further the health care mission.

Let’s get started. Mike Schlosser, welcome to KFF’s Business of Health.

Dr. Michael Schlosser: Glad to be here Chip, thanks for inviting me.

Chip Kahn: This is going to be a great conversation. You and I have talked before, and I know that our audience is going to learn so much today. Let’s start off with the obvious, which is HCA Healthcare is a huge organization. You’re deploying AI across 189 hospitals with 47 million patient encounters a year. When a AI solution company hands you sensitivity, specificity and accuracy numbers on a product, what criteria does that solution have to meet before you deploy it across this sort of massive set of hospitals and ambulatory surgery centers and other kinds of care settings?

Dr. Michael Schlosser: Yeah, it’s a great question. And yes, the scale of HCA is a superpower for us, but it also can be a challenge. But to answer your question directly, the model performance metrics themselves, the sensitivity, specificity, F1 score to me really is a small piece of the story. We really think more about sort of the human AI system, because most of the tools that we’re deploying, the human is still in the loop, right? The people are taking advantage of whatever the AI is producing for them or enhancing their workflow, whatever the product might be. And so in the end it’s really the impact or the outcome of that system that we really care about. So said a much simpler way, are we actually solving the problem? Like, do we see a better clinical outcome? Do we see a better orchestration of care or operational outcome? Are we more efficient? Are we reducing cost? And so those are the KPIs that we track very carefully. Now, the model metrics matter. In fact, where we end up using them more often than not is in ongoing monitoring of models once we’ve scaled them. So, once we’ve gotten comfortable that that system is working as designed and we’re getting the outcomes we’re looking for, those metrics, which can be calculated in a much more automated fashion, can be a way of to make sure that the model continues to perform over time and at scale. And so, they’re very useful from that perspective. But really we want to focus on business problems and clinical problems, and we want to measure those problems directly, more so than just say, oh, your model performs well, therefore we will deliver it. I’m sure we’ll talk a little bit about our nurse handoff tool, but when we developed that solution, we had multiple different metrics around the conciseness, the factuality, the helpfulness of the model that we paid very close attention to, which were much closer to the problem in the bedside than just sort of the F1 score, if you will.

Chip Kahn: Our focus here is AI, but in a sense we miss it, I think, because it isn’t, AI, it’s the data, stupid. And, you called data a strategic asset. What does that mean in a practical sense for your system?

Dr. Michael Schlosser: You mentioned the 47 million patient encounters a year that we have the privilege of caring for. All those encounters create data, and there are incredible patterns inside that data that we can harvest the value from. So, I’ll throw another number out. We do over 210,000 deliveries in our hospitals every year. That’s in the ballpark of how many occur on the continent of Australia. So, we have these massive populations and data describing everything about them. You know, the clinical aspects of the care, but also operations, supply chain, you know, what products were used, you name it. And so that is a strategic asset. And in fact, it’s where the value and the intelligence comes from. The AI models are kind of like shells. And I’m not trying to take anything away from the brilliant people who build these things. They’re incredible, but they don’t become intelligent until you fill them with data. So ChatGPT is trained on the entire Internet. It’s pretty knowledgeable about everything, but it’s not an expert in anything. You give it 47 million patient encounters worth of data and it can become an expert in health care delivery. You then have to pair that with the human insight because it’ll learn the patterns. Some of those are good patterns, some are not good. And the humans have to help identify which is which. But in the end the power comes from the data. And so we treat data like an asset. We’ve invested heavily in building out a new data infrastructure, a data lake house on Google Cloud that allows us to capture all that data, organize it, harmonize it, and then manage it continuously as a product so we can make it available to AI models, to our analysts, to really everyone in the organization. and ultimately that’s where the real step forward is going to come from.

Chip Kahn: Well, you’re a leader in the field, but you also work a lot with others. And how do you compare your notion of this strategic asset and how does that compare with academic health centers or other large systems that either you’re working with or that you are following as developments go in this area?

Dr. Michael Schlosser: Well, and I think there’s great synergy here and there’s, there’s several academic systems that we work with directly we partner with on research. I was actually hearing about this in detail yesterday from the gentleman who leads our trauma research organization that a lot of academic centers come to us because of the density and the velocity of our data. And we can do studies at scale that that they couldn’t do on their own because while they may have incredible experts, they do, they have the most brilliant minds in the field usually working there. They’re often one hospital or a small hospital system. So they just don’t have the depth and breadth of the experience that we do. Also, our data is interesting because it comes from sort of every corner of health care delivery. We have rural hospitals, we have urban and academic centers, we have quaternary care facilities. You know, we have 61 bed hospitals that are truly community hospitals. And so we can see patients and care delivery in sort of every flavor of health care across the country. Which again makes the data set that much more interesting when you talk about training or testing models. So, I’d say we are really synergistic with the academic centers in terms of all of us trying to move health care forward. I will say we tend to come at this from a little bit of a different bend. I think a lot of academic centers see their purpose as advancing the science. Like how do we make the science better, how do we find the next standard of care? We tend to focus more on how can we deliver care in our current standard at the most reliable, the most efficient, the most effective level. I think of if everyone who entered an HCA hospital or any hospital got exactly what the best standard of care treatment was in a timely fashion every single time, how much better the health care system would be than the current state where we’re reliant on a lot of manual processes and local decision making. And they get to the right answer pretty darn frequently. But I think with partnering the humans and AI together, we could do it at a different velocity.

Chip Kahn: When you look at new technologies, whether it’s in the AI area or in other areas, you have innovation hubs, UCF Lake Nona Hospital and TriStar Hendersonville Medical Center. There you have both clinicians and operators who are sort of in a development mode for your enterprise. How does that model work and what has it taught you?

Dr. Michael Schlosser: Yeah, and we actually have three now. We just launched a third one which is HCA Florida Aventura, down in the South Florida market. They’re critical to our strategy. So, when we started this journey of transformation, one of the things that I recognized from my prior experiences as a hospital operator, a group CMO, was that a lot of technology was sort of hoisted upon the care teams, the operators in the hospital. And many times they found it just didn’t work, it didn’t fit into their workflow, it didn’t really understand the nuances of care delivery. I’d say the best or worst example of this is the electronic health record, which achieved the goals it set out to achieve. We created a lot of data, but it did not make the process of delivering care better, faster or more efficient. And so we decided from the very beginning like we were going to do the opposite of that. We are going to design products and solutions or co-develop or bring them into our system that our caregivers really say are solving problems for them, making their lives better, making care for patients better, making patients’ lives better. And to do that we knew we were going to have to get them into the middle of the process. And so we stood up these innovation hubs as a way to bring the technology and the innovation to the bedside. So, we were doing forward deployed engineering before it was cool. We would take our data scientists and our software engineers and other innovation leaders and we would make them go to the hospital, actually insist that all of my DT&I team wear scrubs when they go into our hospitals so that they blend in with the care team, they become part of the milieu and they just sit there in the nurses station, following doctors around in the ERs. And they learn from them, they understand their workflows, their lives, how they do their job, what they care about, what they’re scared of. And it’s not to say that we want to anchor on the current process. I mean, sometimes we need to blow up the current process, but you have to really understand where you’re starting from if you’re going to drive that kind of change. And so the innovation hubs were designed, if you will, to be a place where we could do that work. And it’s a two-way street. So we, you know, we had to get the engineers and the teams to want to go in and spend substantial time in the hospitals. We also had to train the folks in the hospitals to be part of an innovation process. They were not used to this. They’re used to a product being done when we turn it on in a hospital, because of the nature of that environment, it’s high risk. And so the fact that we were going to bring them products that not only weren’t done, but they might just be a prototype on an iPhone was a totally new experience. And that they were going to talk to data scientists who didn’t know anything about health care. These were all new experiences that they were going to go through multiple iterations of change management. So we spent a lot of time upskilling the people, the leaders and the care teams that work in these hospitals to really be an innovation hub. And it’s been an incredible return on investment. We get so much amazing feedback from them. It changes the trajectory of the work that we do every time we go. I wouldn’t do it any other way. And we have a very robust discovery and design and process engineering team that goes in and understands problems at a very deep level. Even if we’re just buying a solution, we want to understand the problem in a really detailed way before we ever start trying to bring a solution in.

Chip Kahn: Let’s take a little deep dive then on some of the solutions that you’ve actually put in place and deployed at scale. You’ve got ambient clinical documentation, you got Timpani staffing and nurse handoff. Can we take each of those, and see what they mean when they leave these innovation centers and go out into the real world across the entire system?

Dr. Michael Schlosser: And three very different problems and very different solutions. So, I think some good illustrative examples here. I want to start with ambient clinical documentation. So this is one that, I would say most people are probably familiar with or getting familiar with, using artificial intelligence and large language models to listen to a conversation between a doctor and a patient or any provider and a patient, understand what’s being talked about and then structure that data into a draft note that can go into electronic health record. And then the doctor is the human in the loop. They review that note and ensure it’s all factual and accurate, make edits, and then they can sign it. The idea being that we would create better documentation, that the AI would be more thorough in that it would capture everything that was talked about, that it would be faster for the doctors. They would spend much less time editing notes or writing notes, and so they could spend more time focused on their patients and on critical thinking and care delivery. And that also the medical record would be more timely, that we would have complete notes, complete H&P in the chart. Much faster for the rest of the care team, but also for payers and authorization that are dependent on timely medical record documentation. And so a number of different value streams there that we saw ambient clinical documentation could provide. We partnered with a company called Commure. And full disclosure, we have a small investment in Commure as well. And we’ve actually been working with them now for over a couple of years refining this approach. And so rather than buying a solution off the shelf, so, you know, going back to this, care teams being in the middle of the product, we partnered with someone who would do the forward deployed engineering model and they would bring their product, but also bring their people and learn from our physicians. In this case, for a few reasons, we actually use some hospitals in Dallas as our guinea pigs. But we had some great physicians, some hospitalists and ER doctors there, who spent a tremendous amount of time with Commure continuously refining their product. And, the CEO of Commure told me he thinks they’re on version 300 of the product and the model, which is why I think we’re getting the performance we get. And so we’re now rolling this out. We’re live in, I think, 67 hospitals as of today. We’ll hit somewhere in the neighborhood of 105 by the end of 2026. We’re doing about 200,000 notes a month on the ambient documentation platform. A lot of H&Ps, a lot of progress notes, a lot of ER provider notes. We’re about to turn on automated discharge summary, broadly here shortly. We’ve got cardiology coming up quickly. So very rapidly bringing this across our entire platform. And we’ve seen tremendous results. Our doctors are averaging between an hour and an hour and a half of time saved every 12-hour shift. We’re getting great feedback, from our patients, including in our patient experience surveys that the doctors are spending more time at the bedside and they’re more communicative with them because they have to verbalize more in the room so the AI can hear them. So the patients are benefiting. And we’ve seen a pretty significant improvement in how quickly the notes are getting signed. That’s a metric that we track. And so I think it went from something like 67% within 24 hours to almost 90% within 24 hours. So the medical record’s getting completed faster, which helps with care delivery downstream. But the best part of this entire rollout has been the quotes that I get emailed to me all the time from the physicians. I’m constantly getting these nuggets of doctors saying, this has changed my life. I go home, I don’t have to do notes, I can spend time with my family. Like, this is one of the first times that a technology has truly made my life as a physician better. So that’s one of the best parts of my day, is when I get to read those quotes that we’re achieving. One of the missions I set out to achieve when we started this, which was to reduce the administrative burden in health care for providers. So, it’s been a tremendous success, but it came at the end of a lot of hard work. We really worked with our doctors to refine this platform and this model and the application, to get it to the point where it’s being adopted at this rate. I’ll throw out one more number. We have an 81% adoption rate in our emergency rooms amongst our ER doctors and other ER providers, which I’m fairly confident is the highest number I’ve seen reported anywhere. A lot of folks are reporting something more like 25 to 50% adoption of ambient in various care settings. Most of those are outpatient clinics. So, it really has made a tremendous impact.

I’ll pivot and talk a little bit about nurse handoff. This was a problem that our nurses brought to us and basically told us they’ve been working on this for a long time with the EHR companies. Lots of different solutions and really had never been able to crack this. And handoffs, as you know, are just a critical part of care delivery. When the responsibility for a patient is transferred from one care team to the next, there’s a lot of risk involved. We do 60,000 handoffs a day, 24 million of them a year in our acute care setting. And so, you’ve got to get that right. Aand right now, it really falls to the nurses. They do this on their own. They scrape data from the EHR, they write handwritten notes, they communicate verbally to the oncoming shift. and they’re working very hard, doing the best they can. But it’s a process that absolutely can be improved. And so we saw this as a great use case for a large language model. We thought we can teach an LLM to read the chart and figure out how to think like a nurse. And actually, the Google research team was really excited about this. And so they partnered with us to help us build out the system that is around the Gemini model to make it good at this. And it took some time. We worked with nurses. We would have them do mock handoffs, real patient data deidentified, and we would have them read the handoff note, tell us everything that was wrong with it. We’d feed that data back to the engineers and we iterated like that dozens and dozens of times with hundreds and hundreds of examples until we got to the point where the model was really performing at a high level. So that’s the pilot. Then we put it in the real world, tried it out in eight different hospitals and got a whole other set of feedback. I think we got 7,000 comments, about how they wanted the data organized differently. They wanted to view it the way they were used to viewing data. And so we went back and we iterated again. And now we’re at a beta pilot in 12 hospitals and getting just tremendous feedback from our nurses that it’s making their job easier. They feel more confident in taking care of patients. And we’re tracking our safety events. We have a safety event reporting system and we’ve seen an 80% reduction in handoff-related safety event reporting. Some of those are near misses, a lot of them, but nonetheless we’re seeing it in the data that this is making patient care safer and that brings confidence to our nurses. You know, nurses are always concerned about the patients and their safety. So it’s been a real home run. We’re in the process of hardening the data product that sits underneath it. Back to your previous comment about it’s all about the data. We’ve got to make sure that that data product is really ready for primetime if it’s going to be used 24 million times a year reliably. So a little work to do there and then we’ll start rolling out beyond those 12 hospitals probably in the middle of this summer, on a pretty aggressive strategy. That’s, I think, a really great example of how we’ve worked at the bedside built by nurses for nurses, and that is a sense of pride for them as well.

The last one you mentioned is Timpani. And this has been a really interesting journey. It’s maybe a little bit of a cautionary tale also. This was one of the first products we actually started working on. And we did that for a very specific reason. We knew that if we didn’t have the right care team on every unit every day, if we couldn’t deploy the right people, right place, right time, right set of skills, it would be hard to innovate on top of that platform. If we had uneven staffing, that’s not a strong platform to stand on and try to drive even more innovation. And so we said, we’ve got to make sure that we’re deploying our clinical staff uniformly and also doing it with the right set of skills, the right predictive algorithms to know how busy the units are going to be. And so we built all that. And it’s a great piece of software. It’s a really sophisticated application and it’s programmed, I think, with really sort of altruistic goals, to take whoever the team is you have on this unit in this hospital, know who they are, their skills, their preferences, how they like to work, when they like to work, and deploy them on a schedule and maintain that schedule such that we have the best likelihood of providing the right team for the patients who are going to show up. We have great machine learning algorithms that predict how busy the units are going to be. And we rolled it out to 130 hospitals at this point.

The cautionary tale piece of this is around change management. And honestly, we haven’t talked about it yet, but change management is probably the hardest part of the job that I do on any given day, because, as I mentioned, there’s always people involved here. We’re not replacing people with AI, we’re augmenting them, which means the people have to change. And that’s hard. It’s definitely hard at scale. And so what we didn’t recognize was all of the nuances that went into building a schedule, the personal relationships between the nurse leaders and their team and how scheduling and staffing was part of that personal relationship, just all of the things coming out of the pandemic and since then that they’d done to try to maintain that staff, you know, there’s a workforce crisis. And while the system had, I think, really good rules built into it, it didn’t quite understand the nuance at that level. And so, we had trouble getting people to really accept the output of the model and they would go in and edit the schedules to look more like what they would have built on their own. Which means we were losing some of the time savings it was supposed to create. We were losing some of the value in spreading the team out more evenly. So, we’re still learning. That’s one of the things that I would say is the most interesting about AI, is that it’s always about learning, the AI learning and us learning. And so we still think that this is going to be a good solution and it’s going to solve a big problem for us in the long term. But it turned out to be a little more complex than we knew what we were getting into.

Chip Kahn: So it really sounds like you just can’t take it off the shelf when you bring these solutions on. It’s a process. It’s not just a ready-made item to implement. What’s next? What’s sort of in your stack that you’re working with your clinicians and the other staff or the members of the team. Is there anything you can talk about that’s sort of in the wings?

Dr. Michael Schlosser: We’re doing a lot of work in supply chain, and so this is all about making sure that we have the right products in the right place at the right time. We run a lot of our own supply chain and distribution centers. But a lot of it’s manual, a lot of it’s driven by people and their experience. And so we’re bringing intelligence to those systems. So that’s exciting to me because there’s a good financial ROI attached to the supply chain use cases. And so if I’m going to continue to get funding from the organization, we have to focus on things that also improve the finances of HCA in addition to the care experience and the clinical outcomes and the patient safety. Because AI is expensive, as you know. Another one that I’m really excited about is our partnership with OpenAI. We’re just a handful of months into this strategic partnership. but they’ve already brought their engineers to work with our teams on a few really high impact use cases, one of which we refer to as our moonshot, where we’re working with them to build a team of agents that can do the orchestration of care delivery in a hospital. So, you know, if you got 100 patients in a hospital, and they each have dozens of things that they need to have done on any given day. Meds, delivered procedures, scans, therapy. The way that gets done in a hospital right now is sort of by the people like the doctors, the nurses, the therapists. They all sort of just figure it out. Like, they have their work lists and they just get through it. But if we had a team of agents that knew everything that was going on and could behind the scenes, organizing, and orchestrating what’s the most efficient and effective way to get things done. What’s the next priority that has to happen? Who needs to go to MRI right now? Like, you know, just all of those individual details could be organized by agents. And then because they’re agents, they can also take action. They can message PT and say, hey, divert from where you’re headed, go up to the sixth floor and see this patient because they’re waiting on discharge. You could actually make a hospital run way more effectively. The patients would have a better experience. They might even know, like, when their MRI scan is going to occur next. We could get them through their plan of care faster, which means we have more capacity to care for our communities. There’s a tremendous upside that we could create here, clinically, operationally, financially. It’s a moonshot because this is a highly complex use case. Incredible amount of data that we have to organize a team of agents that we have to get to work with each other and then people on either side of those. So, it’s a heavy lift, but so far so good. We’re making incremental progress. I think we’ll have something that we can pilot later this year. So, we’re really excited about that sort of operating system of a hospital of the future, if you will.

Chip Kahn: You’re sort of getting at it now. And I’m going to ask a conceptual question of you. We had Bob Wachter on. He asked whether AI is really changing what it means to be a clinician, and he wrote a whole book about that. And from your view, I mean, you’ve got 44,000 either affiliated or employed physicians across this gigantic system. What is your perception of the gains, risks, or both for the quarterback? I mean, historically and traditionally and in terms of the practice of medicine, those physicians are the quarterback of the care. Agentic AI really coming online at some point begins to either change that or provide some kind of new factor that hadn’t been in the physician’s workflow before.

Dr. Michael Schlosser: Yeah, my short answer to your initial question of is it going to change what it means to be a physician is, I really hope so, because I see this as an incredibly positive change if we want it to be. And I already talked about ambient documentation and the positive change that’s happening. I see now that this is making our physicians’ lives better. And I think when the workforce is happy, the patients are happier. So there’s a lot of downstream positive impact we could have. I think the more we can use AI to make the health care system run better, the more they can be that quarterback. They can spend their time thinking critically, studying, reading articles, talking to their patients, engaging in research and other scientific endeavors. A lot of the things that we became doctors to do and maybe not have to do any of the things that we don’t have to do. I remember back to my days practicing, getting on a phone call with an insurance company explaining to them why my patient needed six more weeks of physical therapy. Like, that’s not what I went to school and to become a neurosurgeon to do. So, I think there’s a first wave here where we can massively clean up a lot of the complexity and the burden we’ve injected into our health care system, which I’m excited about and we’re focused on. And then I think once we build trust with doctors and clinicians, once they come to see AI as a tool and a partner and maybe less of a threat, then I think it opens up for a lot of other possibilities. I do think in the future we will have AI and humans working together to make most of the important critical decisions in our health care industry. That when you go to make a decision for a patient, a treatment or diagnostic decision or whatever it might be, that you can tap into your own knowledge and experience, but then you can also tap into AI and the data and the patterns. And as a second opinion, like a real time, always on second opinion, I want the doctors to stay in control. I don’t think a health care system where AI makes the choices is one that I want to help develop, at least not at this point. I don’t think the AI models are ready for that, to be honest, but I think they’re very capable of being a copilot. That health care system that I just described, where you don’t spend time on administrative burden, you get to spend time thinking and maybe enjoying your life a little bit, and then you have AI helping you do your job better, sounds like one that people would want to be a part of.

Chip Kahn: You know, this podcast is built on the premise that patient care and ultimately patient outcomes really depend on the business model. And in some ways our conversation has covered that. So when AI improves clinical quality or promotes operational efficiency for HCA Healthcare, that’s a good. But where’s the line between that and what makes the world go round, which is volume and payment frankly?

Dr. Michael Schlosser: You’re spot on. It has to do both. And so we think of our AI investments and you and I are saying AI, it’s really digital AI data, it’s the whole technology ecosystem. But we think of our investments in digital transformation as a portfolio. And really it sort of functions like an investment portfolio with, you know, your mutual funds and individual stocks or what have you. And so we have some high risk bets, some lower risk, but much more likely to perform. We have some things that are not going to return a financial return, but they’re really important and we want to do them, we want to support them. And then we have some that we really expect significant financial ROI from like supply chain and revenue cycle. And the idea is that we spread our investments across that entire portfolio and then we manage it as a portfolio and the whole thing collectively has to perform. So it’s got to be ROI positive, right? We’ve got to have a financial return, but every individual use case doesn’t have to. So nurse handoff doesn’t have a big financial ROI. I think there’ll be a lot of downstream benefits. Nurses are our most important workforce that we hire directly. We don’t employ a lot of our physicians. And so if, you know, if we shore up that workforce and people want to work for us, maybe down the line there’s a financial impact, but in the immediate term it’s about patient safety. But then to offset that, I’ve got other use cases that are much more focused on driving an immediate financial ROI. We call this value tracking. But the discipline of value tracking I think is intrinsic to a digital transformation agenda. And it’s not a side gig for us. Like we have people who focus all of their energy on tracking the value of each individual use case, whatever that value might be. It comes in all kinds of different forms, but it’s got to be measurable at scale, not just in a pilot. Like we’ve got to be able to see the value occurring, either the money or the safety or whatever it is, numbers, progressing as we roll it out. Otherwise, we don’t keep going.

Chip Kahn: I guess that you must see every day really miraculous solutions and you’re sort of hitting at it, I think, with the end of that last question. So what are the characteristics or criteria that make you decide not to deploy? To say, wow, that really must be a wonderful thing. But not here.

Dr. Michael Schlosser: The most common reason is actually because it ends up not being an interesting problem to solve and the value isn’t there. Right. So a lot of times people will pitch us an idea and they’ll say, hey, this is going to solve this amazing problem for you. And when we go in and we start looking at it, we realize, well, that problem actually isn’t really something that’s strategic for us, or it’s not top of our list, or, yes, you can make that better, but we don’t really see that that’s going to create the return for us. It’s rarely that the technology doesn’t work. I mean, most of the times the technology is solid. It’s because it doesn’t sort of fit, it won’t fit in our workflow, or it doesn’t integrate well with our systems, or we don’t see people adopting because it’s creating a new app that they have to go to that’s outside of their normal workflow and that just creates burden instead of solving for it. So it’s usually at that level, at that sort of technology, human interface level, that things fail more so than the technology. So the way we avoid spending too much time going down those rabbit trails is that discovery process. We make sure we really understand the problem we’re solving at a very detailed level. And where does the problem originate from and what does the current workflow look like? If you’ve got a manual paper process, if people are already working around a piece of technology with a paper process, enhancing that technology is not going to do anything. So you’ve got to understand sort of the people and how they’re doing their jobs if we’re going to pick the right solutions. I would say the other way is that it doesn’t scale. And so there are a lot of processes or products, technologies that you can kind of force to work in a pilot. You can sort of stand up everything you need to make it work. You can get the people on board, you can sort of get the data pipes doing what you need them to do. Then you see a result in a pilot. But it just, there’s no way to take that to 189 hospitals. It becomes prohibitively expensive. People don’t adopt it because you had this really controlled environment where you generated all this excitement around it. But at beta or at scale, the excitement’s not there. And people decide it’s really not solving the problem. For me, that’s sort of the second most common way. And so we see those through the data.That’s where the value tracking and the adoption tracking comes in. And so hopefully we catch those quickly and we pivot.

Chip Kahn: You know, no matter what happens in the systems that affect clinical care directly, humans are going to be involved. So as we get into more clinical decision support and other kinds of, I’ll call them bells and whistles, but alerts, we see in hospitals and other areas, you know, something called alert fatigue. Actually it predates AI and the new solutions, but it’s going to become a real issue in terms of the human role. Across your system as you adopt all kinds of new solutions and particularly, eventually you will get heavily into clinical decision support because it’s coming, how do you deal with the human factor, which is really what alert fatigue is all about?

Dr. Michael Schlosser: The simple answer is you’ve got to directly manage it. So we are very sensitive to alert fatigue. I talk with my team all the time that we really can’t deploy any solutions that add alerts right now. Like, if you were to go talk to our nurses in our hospitals, they’ll tell you they’re already past the saturation point. All of our nurses have what we call imobile. They all have iPhones that we provided them and then software on those iPhones that help them communicate with their colleagues and others. And so they have a platform where we can deliver alerts to them. And we deliver a lot of alerts. And they’ll tell us like, you know, it’s work to sort through all that and find the signal from the noise. And so we have to actively manage that. And so I would say we have as much or more work going on figuring out how to reduce the number of alerts going to those phones as we do people trying to design like, new AI solutions that could potentially throw off new alerts. The nursing team did a big project around bed alarms. Like, how do we clean all that up so that we’re not overloading them with bed alarms all the time? So, you have to actively manage it. You have to have a discipline around it. Now the future state. I have a really innovative design team that has this concept that they refer to as nugentic. It’s a combination of nudge and agentic. When they talk about nugentic, I sort of, I almost think of like the old-style telephone, systems where there was a person plugging in, like the, you know, I’m talking about plugging in, like the connections. Could you have an agent that’s sitting in the middle of all of these communication streams and figuring out what does Dr. Schlosser actually want to know right now? Like which of all of these signals that are coming in that are screaming for attention are the ones that really matter? And how does he like to receive that information? Does he like to get a text message, a phone call, an email, an overhead page? Smoke signals? And the agent could be figuring out how to parse out the information in a way that it’s not overloading, that we’re not missing critical things where patients are going to get hurt if we miss them. But we’re also filtering the noise as much as we possibly can. So that’s a future state, that’s a long-term solution. But we’re starting to build that technology because this is a problem that’s going to have to be solved because the alert fatigue is a very, very real thing. And so, we’re putting some effort behind it.

Chip Kahn: And I guess along those same lines, you’ve described a very comprehensive process of design and development and then integration into your system very carefully done. But when I had one of my other guests, Elad Walach from Aidoc, he mentioned something called drift. In a sense, his system’s a medical miracle. Over time, there’s some, either statistical or data or technological things that happen. And whether it’s drift or whether it’s other kinds of degrading, what are the feedback loops that you all are anticipating to try to keep an eye on all these new systems that are dynamic? I mean, they’re going to change over time and there may be unintended consequences.

Dr. Michael Schlosser: Yeah, and sometimes it’s the patient population that changes. Right? I mean, if, God forbid, we ever had another pandemic, the patient population your model would be working on now would not be the same as the one it was trained on, because that disease didn’t exist. So there’s lots of reasons why model drift can occur. And yeah, your point is a great one. You have to have this ongoing monitoring. And so, I’ll go back to what was the first question about sort of the model scores. So we’ve set up a model registry with model cards. Lots of folks have done that, but we also have invested, and this is early days, but in a technology where we can bring that kind of data in in real time. I mean, it wouldn’t be like literally minute by minute, but at whatever the right cadence is, we could ingest those kinds of metrics back into the model cards and have an AI system that’s watching that so that we can see, okay, has the performance changed, like month over month? Is the accuracy of his model changing, and if so, what do we need to do about that? How do we flag that? With the large language models, we do have to rely on the people. And so because you can’t measure an F1 score, you can’t measure a model performance the way you can with like a machine learning model, like his systems use when you’re talking about a big foundation model. So we built into the nurse handoff tool, like UI that makes it really easy for them to tell us if they see something wrong. If they see a hallucination or an error or an omission that they think is material, they basically just press a button on the interface and that data goes back to us. And we capture that. In our ambient clinical documentation, we worked with Commure to build multiple layers of safety nets. So we have a hallucination checker, which is an LLM that’s really good at seeing the hallucinations from other LLMs. Then we have a subset of notes that are just hand reviewed by people on a regular basis. And then we archive de-identified data so that we can use it as a QA testing environment in the future. And so it depends on the type of solution, what kind of, program you have to put into place. But this concept of having sort of ongoing machine learning operations, LLM ops, ongoing monitoring, I think is part of the challenge of scale and part of why going from pilot to scale is actually really hard.

Chip Kahn: What I understand here is first there’s choices about even looking at solutions. Then there’s development and design and bringing the staff in. The evolutionary process that you go through with your staff, with the solution producers, and then there’s implementation and scaling, and then there’s the feedback loop. HCA is a very large organization that can do these kinds of things. What’s your view about how smaller systems, because if we look across the country, there are a few, but not that many systems at your scale, that have the management structure and the development structure. You know, what’s your observation about how the solutions that you’re talking about that are so necessary for the future of health care to ensure the quality, to bring the cost down. What’s your observation about what smaller systems should do in this environment when these solutions are coming out every day?

Dr. Michael Schlosser: It’s gonna be a challenge. I would agree. I think that we are sort of the tip of the spear right now. And I would hope that they will benefit from our learnings that as we build these solutions, as we work with partners like Commure, we’re working with Google, with GE, with OpenAI, that we’re helping sort of the whole ecosystem of health care technology get better at this. And therefore the products will be more complete products. They won’t just be an algorithm or, you know, an AI dictation system, but they’ll come with the kinds of belts and suspenders, safety systems that you need, because that’s going to be necessary for these smaller health systems to be able to adopt that technology. The other thing that I think about, sometimes, and I’ve talked with some folks in Washington D.C. about this, is should we be thinking about the next version of what meaningful use was, what the HITECH bill did? Is this important enough that there should be some ways that government supports funding this kind of technology to get down to sort of every nook and cranny of health care? And we don’t need that. Like we, as you said, we’re big enough and successful enough that we can afford this, but we are definitely the outlier. The vast majority of systems would struggle to make these kinds of investments. And so, if in the next few years we really see health care changing for the better because of AI and technology, should there be some way of trying to ensure that all patients and all systems have a way to access that kind of technology? That might end up being the case. We’ll see. But in the meantime, I’m hoping what we learn can benefit all healthcare systems in the future.

Chip Kahn: So you’re actually bringing up one, but let me ask, are there other changes, one or two in payment regulation, what AI and developers do, that you think you’d like to see, to accelerate the dissemination of all these really medical miracles that AI and health care can produce?

Dr. Michael Schlosser: I’ll give you two. One is something that we’ve been pretty consistent about in our talking about this, and that is I think we would benefit from some common sense federal regulations or standards, just some guardrails, around what good looks like for deploying AI across health care systems. If for no other reason then it would help us not have to deal with 50 different sets of rules that every state will feel like they need to enact in order to protect patients or whatever the reasons are. I’m not trying to wade into a political space here. I can just leave it at that. But I think that would actually help drive innovation if we had some agreed-upon guardrails. The second thing I would say is that I wish the conversation and reporting around AI was a little more transparent and a little less hype, cycle driven and that this is never going to happen because it’s just the way the media works. But the constant sort of frenzy around what AI is capable of or could do versus what it actually takes to deploy agentic AI in a health system, there’s an enormous gap between those two things. And so every day there’s some new article or something about what these AI agents are capable of and the dramatic change it’s going to create and all these things. And it completely ignores the fact that you have to do all those steps that you really clearly outlined here just a minute ago if you’re going to put AI in a heavily regulated, high-risk environment like health care delivery. And so I’m incredibly bullish about AI. I’m a huge tech optimist. I think it’s going to create enormous change. It is not going to come overnight, it’s not going to be easy, but we’re going to do it because I think it’s worth doing. I wish that gap was a little more narrow so that people could work on reality of AI rather than fantasy.

Chip Kahn: Mike to close out, we’ve really gone over the whole waterfront, I think, in terms of implementation of AI at HCA healthcare. What keeps you up at night, in this AI area, either in terms of your environment, your ecosystem or even larger scale.

Dr. Michael Schlosser: I’m going to cheat and give you two here also. So the number one has to be change management because we have to adopt these changes and all the people who make up health care have to become open to change and open to innovation if this is going to work. And that’s a challenge. And so that worries me. The second thing that worries me is that I don’t really know that we fully understand AI even yet at this point. I’ll feel schizophrenic where sometimes I feel like, hey, these models make mistakes all the time and are they really going to be any good? And then I’ll read something where I’m like, oh my gosh, humanity is already in trouble because the AI is smarter than us. And the reality is somewhere in between. But that keeps me up at night. I don’t know that we fully understand the technology that we’re using and it might surprise us one day. So, we’ll see.

Chip Kahn: Well, hopefully we can avoid the downside and enjoy all the value from everything you’ve described here today. Mike, this is terrific and I just so appreciate you coming on the podcast.

Dr. Michael Schlosser: Chip, thanks for having me. Enjoyed the conversation.

SERIES

The Business of Health with Chip Kahn

This weekly podcast features insightful conversations between host Chip Kahn and his guests, who discuss the business of health care, connecting the dots between the health care business, policy, and patients.

The podcast’s first series on AI in health care illuminates how AI is changing health care, and features guests who are deploying this technology, managing its consequences, and designing policy around it.

Medicaid Enrollment and Unwinding Tracker

Enrollment Data

Medicaid/CHIP Enrollment Trends

Unwinding Data - Archived

State Data on Renewal Outcomes

Medicaid Disenrollments

Medicaid Renewals

Federal Data on Renewal Outcomes

State Data - Archived

Related Resources

Resources on unwinding data

Resources on state policies and preparations for the unwinding

Resources on pre-pandemic enrollment patterns and coverage transitions

KFF’s unwinding explainer

Also of Interest

A Closer Look at North Carolina’s Implementation of the 2025 Reconciliation Law Medicaid Provisions and Other Changes Amid Medicaid Budget Shortfalls

What is the budgetary context as North Carolina prepares to implement the 2025 reconciliation law’s provisions?

What are some of the Medicaid policy changes included in North Carolina’s recent legislation?

Eligibility and Cost Sharing

Medicaid Financing

Related Content

Medicare Advantage Out-of-Pocket Limits: Variation and Trends

The Average Medicare Advantage In-Network Limit is $5,421 in 2026, But Nearly 1 in 5 Enrollees Face In-Network Limits Higher Than $7,000

Takeaways

In 2026, the Average Enrollment-Weighted Out-of-Pocket Limit for Medicare Advantage Enrollees Is Lower for HMOs Than PPOs

In 2026, Roughly One in Ten Medicare Advantage Enrollees Are in Plans With Limits of $3,000 or Less for In-Network Services and About One in Five Are in Plans With Limits Above $7,000

Average Out-of-Pocket Limits for In-Network Services Generally Declined Between 2017 and 2023 but Have Increased Since Then

Appendix Tables

Related Content

Hantavirus Outbreak Revives COVID-Era False Health Claims

Highlights

What We’re Watching

As Hantavirus Cases Emerged, So Did Familiar False Claims About Causes and Cures

Most Americans Who Follow Health Influencers Are Skeptical of What They Hear

A Closer Look at the State Level: Nearly Half of California Residents Distrust the Health Care System, But Most Trust Providers

AI Legal Roundup

Pennsylvania Lawsuit Challenges AI Chatbots Presenting as Licensed Medical Professionals

AMA Calls for Legal Protections Against AI Deepfakes Impersonating Physicians

More From KFF

Drew Altman, Founding President and CEO of KFF, Announces Retirement Plans; Board Appoints Larry Levitt and Mollyann Brodie as Next Leadership Team

Dr. Altman has led the organization for nearly 40 years; Levitt and Dr. Brodie to assume leadership roles next year, after a combined six decades at KFF in senior positions guiding policy research and polling

Contact

Global COVID-19 Tracker

Cases and Deaths

Policy Actions

Social Distancing and Closure Measures

Economic Measures

Health Systems Measures

Methods

Cases and Deaths

Policy Actions

The Medical Frailty Exemption from Medicaid Work Requirements: Key Issues to Watch for in Upcoming CMS Guidance

What is the Medical Frailty Exemption?

How do States Plan to Operationalize the Medical Frailty Exemption?

How Will the CMS Guidance Address Key Issues and Potential Implications?

Defining Medical Frailty

Verifying Medical Frailty

Related Content

Key Facts About Health Care Affordability for People With Medicare

Many Medicare Beneficiaries Have Low Incomes and Modest Savings, Which Can Make It Hard To Afford Medical and Long-Term Care Services

Medicare Premiums and Other Health Expenses Consume a Sizeable Portion of Income and Household Budgets for People With Medicare

The Choice Between Traditional Medicare and Medicare Advantage Has Cost Implications for Beneficiaries

Paying for Health Care, Including for Services Not Covered by Medicare, Can Be a Financial Burden

Medicaid Helps Make Medicare Affordable for About 12 Million Low-Income Medicare Beneficiaries

Medicare Part D Offers Protection Against High Out-of-Pocket Prescription Drug Costs, but Prescription Drug Costs Remain a Concern for Older Adults

The Growth in Medicare Spending Will Lead to Higher Medicare Premiums, Deductibles, and Copayments Over Time

Related Content

Tracking Key Mental Health and Substance Use Policy Actions Under the Trump Administration

AI at Scale: Does It Deliver?

About this Episode

The Host

Guest

Transcript

PREVIOUS EPISODES