What the Unwinding Required States to Do
The FFCRA continuous enrollment provision, which prohibited states from disenrolling Medicaid beneficiaries during the public health emergency in exchange for enhanced federal matching funds, ended under the Consolidated Appropriations Act, 2023 (Pub. L. 117-328), signed December 29, 2022. States were permitted to begin disenrollments April 1, 2023, with a 12-month unwinding period. CMS reported that approximately 94 million individuals were enrolled in Medicaid and CHIP at the start of the unwinding — a program population that had grown substantially since 2020 and included a significant share of enrollees who had not had their eligibility verified in three or more years.
Every state eligibility agency was required to re-determine eligibility for every enrolled individual within that period. CMS required states to complete renewals in eligibility-date order, use ex parte (automated) determination before sending paper renewals, and apply enhanced contact requirements using multiple contact modalities for returned mail cases before proceeding to disenrollment. CMS published a public unwinding data tracker at cms.gov with monthly state-by-state statistics on ex parte rates, procedural versus substantive disenrollments, and pending renewal backlogs.
Where State Eligibility Systems Failed
Ex parte automation varied dramatically and the gap was architectural, not operational. Federal guidance required states to attempt ex parte renewal — checking eligibility against available data sources (SNAP, employment records, tax data, SSA databases) before sending a paper renewal — before any individual reached the paper renewal step. Ex parte rates measured the share of renewals completed automatically. Across states, these rates ranged widely. KFF data and Georgetown Center for Children and Families tracking documented some states completing well above 90 percent of renewals ex parte while others completed single-digit percentages, with the gap reflecting data source integration rather than program size or demographics.
The variation was not explained by program size or beneficiary demographics. It reflected the state eligibility system's integration architecture. States running systems with real-time or near-real-time access to federal data sources (SSA, SSI, IRS, SNAP) through established API connections or shared data agreements completed ex parte renewals reliably. States whose eligibility systems queried federal data through batch jobs — weekly or monthly data feeds — either ran those jobs on accelerated schedules that introduced load problems, or ran at normal frequency and watched paper renewal backlogs grow faster than ex parte processing could clear them.
Contact data was years stale. Medicaid enrollment records reflect the information beneficiaries provided at enrollment or their last income change. For individuals enrolled during 2020 or 2021 who subsequently moved — a segment that included a substantial share of the pandemic-era enrollment cohort — address records pointed to former residences. States were required to run National Change of Address (NCOA) matching before beginning the renewal cycle; the NCOA hit rate was high enough that millions of cases were updated, but NCOA only covers households that filed a USPS change-of-address card. Households that moved without filing — particularly renters in high-turnover housing markets — received renewal notices at wrong addresses, and returned mail immediately created a backlog of cases requiring multi-modality contact outreach before proceeding to disenrollment.
The contact data problem compounded the ex parte rate problem. A case with stale address data AND a failed ex parte check ended up in a manual exception queue that required both contact verification and outreach — the two most resource-intensive exception types, arriving simultaneously in large batches.
Batch job windows collapsed under concurrent load. State eligibility systems process renewals, applications, income changes, and appeals in the same production environment. Unwinding added a volume of renewal processing that, in some states, exceeded normal annual renewal volume within the first several weeks. State IT teams that had sized overnight batch windows for steady-state renewal volume found those windows extending past business-day start, creating data consistency issues between eligibility system records and worker-facing interfaces. Several states had to implement emergency processing windows or stage the renewal population in waves — which was operationally correct but put some beneficiaries' renewals at the end of a longer queue.
Procedural disenrollments exceeded CMS tolerances. CMS issued corrective action plans to multiple states where audit data showed disenrollment rates inconsistent with expected eligibility outcomes — indicating that cases were being closed for procedural reasons (non-return of paperwork, returned mail) at rates suggesting the beneficiary outreach and contact verification steps were not being completed as required. In August 2023, CMS directed a group of states to pause procedural disenrollments due to a systems error in which household-level eligibility assessments were improperly triggering disenrollments for household members who remained individually eligible, including children. CMS published documentation of its corrective action process through the unwinding period.
The GAO examined the unwinding and identified concerns about state data systems' capacity to support required ex parte determinations, the completeness of contact verification workflows, and federal oversight mechanisms. The GAO report on Medicaid unwinding oversight — GAO-24-106883, "Medicaid: Federal Oversight of State Eligibility Redeterminations Should Reflect Lessons Learned after COVID-19," published July 2024 — found that CMS had identified compliance issues in almost all states and that CMS's oversight structure created conditions where corrective action could lag the underlying procedural errors by months.
What an Architecture Built for Policy Shock Looks Like
The unwinding failures point to specific architectural deficiencies — not to the eligibility rules themselves, which were unchanged, but to the technical infrastructure for applying them at scale under a time constraint.
Continuous background eligibility verification, not batch renewal. Systems designed around annual batch renewal cycles treat eligibility as a snapshot: the case is accurate as of the last verification date and untouched until the next. Systems designed for continuous verification maintain current-state alignment with data sources (income, SNAP, SSA benefits, household composition) as changes occur, flagging cases for proactive review when a data source change suggests eligibility may have changed. This is architecturally different from batch renewal; it requires event-driven integration with federal and state data sources and a case management model built around exception handling rather than scheduled processing. The ex parte rate differential between states during the unwinding is the argument for this architecture: states that already had good data source integration performed it automatically; states that didn't had to clear a paper backlog.
Multi-channel contact verification as a maintained attribute, not a renewal-time task. Beneficiary contact data should be verified continuously as part of normal case maintenance — address confirmation at annual renewal, mobile number verification at benefit change events, email address collected at enrollment and re-verified at first renewal. The contact data failures during the unwinding were not the fault of the renewal process; they were the consequence of three years of case records with no active contact maintenance. An eligibility platform that treats contact verification as a task triggered only at renewal will face the same address decay issue in the next high-volume scenario.
Disaggregated processing pipelines with clear priority ordering. Mixing renewal processing, new application processing, and change-in-circumstance processing in a single overnight batch window creates a resource contention problem when any one of those input streams experiences surge volume. The architectural fix is disaggregated pipelines: separate queues with independent capacity headroom and a priority model that can route urgent cases (appeal deadlines, eligibility expiration dates) independently of lower-priority bulk processing. This is a fairly standard distributed systems design pattern, but state eligibility systems built in the 2010s frequently pre-date it and run on mainframe or monolithic batch architectures that don't decompose cleanly.
API-first federal data source integration. The ex parte rate problem is fundamentally an integration architecture problem. States with direct API access to SSA, IRS, and federal benefit data sources (through the Federal Data Services Hub and individual agency agreements) can query those sources in near-real-time and complete ex parte checks within seconds. States relying on weekly batch data files from those sources are operationally dependent on the data being current as of the last file delivery — which, for a beneficiary whose income changed 10 days ago, may mean ex parte fails and the case routes to paper. Modernizing federal data source integrations to API-first reduces ex parte failure rates directly.
The Governance Question for State Health IT
The unwinding is over for most states. Corrective action plans have been resolved or are in their final monitoring periods. CMS has wound down the public data tracker's update cadence. The operational pressure has lifted.
The risk in that outcome is treating it as confirmation that the system worked — eventually, imperfectly, with federal intervention where required. The more accurate interpretation is that the system revealed, under stress, the specific failure modes that have been building since enrollment management was last re-architected.
The next policy shock that requires mass re-verification of benefit eligibility will not be Medicaid — it might be SNAP, it might be CHIP, it might be a new federally-mandated income verification for a state-administered program. The eligibility architecture that failed to sustain 12 months of high-volume renewal processing will face the same constraints in that scenario. State health IT modernization roadmaps that don't include ex parte rate improvement, contact data architecture, and federal integration modernization as explicit items are not planning for the next failure — they're scheduling it.
Sources and further reading
- Consolidated Appropriations Act, 2023 (Pub. L. 117-328) — The statutory basis for ending continuous enrollment, signed December 29, 2022
- CMS Medicaid Unwinding and Returning to Regular Operations — CMS's main unwinding page, including public data tracker and state corrective action information
- GAO-24-106883: Medicaid — Federal Oversight of State Eligibility Redeterminations Should Reflect Lessons Learned after COVID-19 — GAO report on CMS oversight of state redetermination processes during the unwinding, published July 2024
- CMS Ex Parte Renewal Guidance (SHO letter, 2021) — CMS's guidance on ex parte renewal requirements, which established the expectations states were measured against during unwinding
Spartan X's work on state benefits and eligibility systems has consistently surfaced the gap between what a system was designed to handle and what policy change eventually demands of it. The eligibility architecture problems documented during the unwinding — stale contact data, batch processing ceilings, federal integration gaps — aren't unique to Medicaid; they are characteristic of state eligibility platforms built in the 2000s and 2010s that haven't had the kind of sustained architecture investment that would prepare them for surge operation. Addressing them requires a modernization approach that prioritizes continuous data integration and disaggregated processing, not another point upgrade to the existing monolith.
