Hundreds of thousands of lawfully present immigrants are beginning to lose full Medicaid coverage after a federal eligibility restriction took effect Oct. 1, putting refugees, people granted asylum, trafficking survivors and some humanitarian parolees among the first groups affected by the 2025 budget law’s sweeping health provisions. New reporting published Saturday shows the change is no longer a future projection: state agencies have started closing cases, while patients and providers are trying to determine which services remain available.

The immediate policy is narrower than the law’s broader Medicaid overhaul, but its consequences are unusually concentrated. The people losing eligibility were in the United States with lawful status and had qualified under long-standing humanitarian categories. Undocumented immigrants were already excluded from ordinary federally funded Medicaid. The new rule does not alter emergency-only Medicaid, and it preserves full federal participation for citizens, lawful permanent residents, Cuban and Haitian entrants, and people covered by Compacts of Free Association.

What changed on Oct. 1

Section 71109 of Public Law 119-21 narrowed the immigration categories for which states can claim federal Medicaid and Children’s Health Insurance Program funding. Beginning Oct. 1, the statute generally limits that federal participation to U.S. citizens and nationals, green-card holders, Cuban and Haitian entrants, and citizens of certain Pacific nations living in the United States under free-association compacts.

That language removed the federal financing basis previously used for full benefits for many refugees, asylees, parolees and trafficking survivors. A detailed CMS letter issued in April instructed states to update eligibility systems, reverify affected beneficiaries and ensure that federal funds were not claimed for excluded groups after the deadline. The agency said states could continue coverage using only state money, but such programs would no longer be treated as Medicaid for federal matching purposes.

There are important exceptions. Emergency Medicaid remains available to people who meet financial and residency rules but lack an eligible immigration status, although it covers only treatment necessary for an emergency medical condition. States may also continue using an existing option to cover lawfully residing children and pregnant people without the usual five-year waiting period. A CRS review confirms that the new eligibility list is narrower than the earlier “qualified alien” framework and that refugees and asylees without permanent residency are among those newly excluded.

The near-term count is larger than one projection

The Congressional Budget Office originally estimated that the immigrant Medicaid and CHIP restriction itself would reduce federal spending by about $6.2 billion over a decade and increase the uninsured population by roughly 100,000 in 2034. That is distinct from CBO’s much larger estimate for the full Medicaid chapter: the combined provisions are projected to increase the number of uninsured people by 7.5 million in 2034 and reduce the federal deficit by $886.8 billion over 2025 through 2034, according to the agency’s official estimate.

State-level case reviews now suggest that the immediate number losing Medicaid eligibility could exceed the earlier national uninsured estimate, though the figures measure different things. A September survey by KFF Health found more than 281,000 people in nine states and the District of Columbia were identified as at risk of losing coverage in October. Florida alone reported nearly 177,000 affected enrollees, while Arizona expected close to 28,000.

Those numbers should not be read as a final count of people who will become uninsured. Some notices may be reversed when beneficiaries submit updated documents. Some people may have acquired a green card or another qualifying status, and some may move to employer coverage, a marketplace plan or a state-funded program. Conversely, the state survey did not include every jurisdiction, and administrative confusion can cause eligible people to lose coverage. The apparent gap between 281,000 identified cases and CBO’s 100,000 additional uninsured people is therefore a difference in definitions, timing and data coverage rather than proof that either estimate is necessarily wrong.

Implementation creates a second layer of risk

The statute changes eligibility, but implementation determines who actually loses care. States must distinguish among multiple immigration categories, query federal databases, send notices in accessible languages and give people an opportunity to correct records. A September KFF analysis warned that people who remain legally eligible may still be dropped if they do not understand or respond to documentation requests. CMS similarly urged states to use clear, consistent outreach and mark notices as time-sensitive.

The timetable compounds that problem. States are preparing for other large Medicaid changes, including more frequent eligibility checks and a community-engagement reporting requirement for many adults beginning in 2027. Each reform requires new computer rules, staff training and beneficiary communication. When several changes arrive together, it becomes harder to separate a lawful termination from a paperwork failure.

That distinction matters clinically. Full Medicaid can pay for primary care, specialists, scheduled procedures, mental health treatment and ongoing prescriptions; emergency-only coverage generally cannot substitute for those services. The loss of insurance does not itself prove that a particular patient will experience a worse health outcome, and the current state counts do not measure diagnoses, hospitalizations or deaths. But a 2026 JAMA study of the post-pandemic Medicaid unwinding found measurable declines in prescription access after enrollment protections ended, with effects varying by state policy. That observational evidence cannot isolate every cause, but it supports concern about treatment interruptions.

What the evidence can and cannot show

This is fundamentally an eligibility and financing change, not a clinical trial. The strongest evidence about who qualifies comes from the statute and CMS guidance. The best evidence about scale currently comes from preliminary state case counts, which are more timely than claims data but are vulnerable to duplication, later reversals and uneven reporting. Final termination totals will arrive after states complete appeals and eligibility corrections.

Broader research gives a more reliable direction of effect than a precise forecast. The randomized Oregon Medicaid experiment found that gaining Medicaid increased health-care use, improved financial protection, reduced depression and increased diabetes detection, while not producing statistically significant improvements in several physical-health measures over two years. The study design was unusually strong because coverage was offered by lottery, but its adult population, geography and expansion context differ from today’s humanitarian immigrant population.

Those limits argue against assigning a specific number of preventable illnesses or deaths to the October change. They do not erase the more immediate and well-supported effects: people without comprehensive coverage face larger out-of-pocket exposure, weaker access to nonemergency care and greater risk of interrupted medication. Language barriers, disability, trauma histories and unfamiliarity with U.S. eligibility systems may make those burdens heavier for some refugees and trafficking survivors.

States and health systems now face the practical consequences

States have three broad choices: end full coverage for excluded categories, finance replacement coverage entirely with state funds, or create narrower assistance programs. California has budgeted state money to preserve coverage temporarily for many affected residents, while New York and Pennsylvania already operate state-funded options for some immigrants. Other states are directing people toward community health centers, charity care or marketplace enrollment, although affordability and eligibility vary.

Hospitals and clinics should expect effects before national datasets catch up. Enrollment staff will need to distinguish emergency Medicaid from full benefits, verify whether a patient has obtained permanent residency, and identify children or pregnant patients who remain eligible under a state option. Pharmacy teams may see refill disruptions first. Safety-net providers may also face more uncompensated care if people defer routine treatment until conditions become urgent.

For policymakers, the most useful near-term measures will be completed terminations, successful appeals, shifts into state-funded or private coverage, prescription interruptions and avoidable emergency use. CBO’s latest baseline projects Medicaid enrollment well below the trajectory expected before the 2025 law, but its nationwide estimates aggregate many policies and cannot reveal how this specific restriction affects individual communities.

The October cutoff therefore marks the start of an implementation test, not the end of the analysis. The legal rule is settled for now, but the actual coverage loss will depend on state choices, database accuracy, outreach and the ability of households to navigate alternatives. Until those outcomes are reported, the clearest conclusion is limited but consequential: a federal program that once covered several humanitarian immigration categories no longer does so, and hundreds of thousands of people are being reassessed under the narrower standard.