Kentucky’s 2018 work rule was a data system, and a court found its approval failed to weigh the coverage loss
CMS invited Medicaid work requirements on January 11, 2018, and approved Kentucky’s on January 12. In our reading, the terms specified four kinds of eligibility data. A court later vacated the approval because the Secretary had not adequately considered its effect on coverage.
Looking Back revisits past weeks one at a time, up to the most recent, and asks what it meant for the banks, health systems and public agencies we serve. The series was written in 2026, with the benefit of hindsight, and each article says which week it looks back on.
Two-minute summaryThe dek and the first sentence of each section, quoted from the article
CMS invited Medicaid work requirements on January 11, 2018, and approved Kentucky’s on January 12. In our reading, the terms specified four kinds of eligibility data. A court later vacated the approval because the Secretary had not adequately considered its effect on coverage.
- The week On Thursday, January 11, 2018, the Centers for Medicare & Medicaid Services sent a letter to state Medicaid directors.
- Status data: proving an exemption the state might already know The first kind of data answered whether a person was exempt.
- Borrowed data: deemed compliance from other programs The second kind of data came from other programs.
- Reported data: the enrollee as the source of record The third kind of data was what the enrollee had to supply.
- Consequence data: what the state did, and who it lost The fourth kind of data was the state’s own record of what it did.
- How this differed across the organizations involved The four kinds of data fell on different organizations in different ways.
- Objections, in order We take five objections, numbered in the order we address them.
- What a program director could check These checks follow from the four kinds of data.
- With hindsight Kentucky HEALTH was scheduled to take effect on July 1, 2018.
The week
On Thursday, January 11, 2018, the Centers for Medicare & Medicaid Services sent a letter to state Medicaid directors. It was numbered SMD 18-002. Its title was “Opportunities to Promote Work and Community Engagement Among Medicaid Beneficiaries.” It said the agency would support state efforts to test incentives that make work or other community engagement a condition of continued coverage. The vehicle was a demonstration project under section 1115 of the Social Security Act.
The letter described its own scope. It applied to non-elderly, non-pregnant enrollees who qualified on a basis other than disability. It said this was a shift from prior agency policy on work as a condition of coverage. Applications would be reviewed case by case to decide whether the approach was likely to promote the program’s objectives.
The letter set out design considerations. It encouraged states to align their requirements with work rules in the Supplemental Nutrition Assistance Program and Temporary Assistance for Needy Families. It said people enrolled in and complying with a SNAP or TANF work requirement, or exempt from one, must automatically be considered to comply. It noted that states had built integrated enrollment systems, using a waiver of federal cost allocation rules available through calendar year 2018. Where more information technology work was needed, costs would be expected to be reasonable.
The letter put conditions on who could be required to participate. States had to exempt people determined to be medically frail. They should also exempt people with acute medical conditions, validated by a medical professional, that would prevent them from complying. States had to comply with federal disability rights laws and provide reasonable modifications, including exemptions, reduced hours and supports. They also had to take steps for people with opioid addiction and other substance use disorders.
The letter also said what states would have to measure. States would submit monitoring metrics, including enrollment and termination for failure to meet program requirements. Evaluations had to assess people who lost or experienced a lapse in coverage, as well as those who stayed enrolled. States would not be permitted to accrue savings, for budget neutrality purposes, from a reduction in enrollment that might result.
On Friday, January 12, CMS approved Kentucky’s demonstration, called Kentucky Helping to Engage and Achieve Long Term Health, or KY HEALTH. The approval letter said the demonstration was effective January 12, 2018, through September 30, 2023. It covered several initiatives. Within it, a program called Kentucky HEALTH applied to people who did not qualify on the basis of a disability.
The approval described the community engagement requirement. It applied to people aged 19 to 64 in Kentucky HEALTH. Exempt groups included former foster care youth, pregnant women and one primary caregiver of a dependent per household. People considered medically frail, people with a disqualifying acute medical condition and full-time students were exempt too. Non-exempt beneficiaries had to complete 80 hours a month of activities such as employment, education, job skills training and community service.
The consequences were set out in the same letter. Beneficiaries who did not demonstrate compliance would have their coverage suspended. They could reactivate it on the first day of the month after completing 80 hours in a 30-day period, or a state-approved health literacy or financial literacy course. Beneficiaries still suspended on their redetermination date would have enrollment terminated and would need to submit a new application.
The approval also authorized premiums for some enrollees, consequences for not paying them, six-month non-eligibility periods for some failures to complete redetermination or report changes, and a waiver of retroactive eligibility for some groups. The special terms and conditions attached to the approval ran to 72 pages.
Public debate over work requirements centers on values. We read the Kentucky approval of January 12, 2018, as a systems document first. Its special terms and conditions described what the state had to know about each adult, every month, and what its systems had to do with that knowledge. Whether the policy worked as designed would depend on data more than on any single rule.
The approval also showed where the data would come from. Some of it the state already held. Some it could borrow from other programs. Some only the enrollee could supply, every month. The rest the state would generate itself, as it acted on the first three.
We take the data in four kinds. Each answers a different question, and together they cover what the state needed to run the requirement:
- Status data: whether a person was exempt, and on what basis.
- Borrowed data: whether a person could be deemed compliant from another program’s records.
- Reported data: the monthly hours an enrollee had to document.
- Consequence data: what the state did with the first three, and what happened to the people affected.
Our argument is that the requirement’s effect on coverage was set by the balance between data the state could find and data it had to ask for. Every person the state could not clear from its own or borrowed records had to report, every month, or face suspension. That made the coverage effect a data design question. The court that later vacated the approval did so because the Secretary had not adequately considered that effect. After the four kinds, we look at how this differed across the organizations involved, at objections, and at what a program director could check.
Why frame a social policy this way? Because a program director inherits the systems, not the debate. Once a state chose to pursue a requirement, the remaining decisions were about records: which ones to trust, which to request, on what schedule, from whom, and what to do when they were missing. Those decisions are where an agency’s technology and operations staff have influence. They are also where the coverage effect is set.
A note on our reading. We describe the federal letter and the Kentucky approval from Internet Archive copies of CMS’s own files, captured on January 11 and January 12, 2018, and later events from court records and federal law. Where we say what a provision meant in operation, that is our reading. State agencies should rely on their own counsel and CMS for current requirements.
Status data: proving an exemption the state might already know
The first kind of data answered whether a person was exempt. The Kentucky terms listed the exempt groups in STC 44. They were former foster care youth, pregnant women, and one primary caregiver of a dependent per household. People identified as medically frail were exempt, as were people with an acute medical condition validated by a medical professional. So were full-time students as determined by the state, and people under 19 or over 64.
Exemptions the state could see
Some exemptions sat in data the state already held. Age is on every application. Pregnancy and former foster care status might be recorded in the eligibility system or in child welfare records. Medical frailty was defined in the state’s alternative benefit plan, under the federal rule the terms cited, so the state already had a definition and, for some people, a determination.
For these groups, the state could apply the exemption without asking anything new. The quality of the state’s existing records decided whether it did. An outdated pregnancy flag, or a frailty determination stored in a different system, would turn an exempt person into one who appeared to owe hours.
Exemptions that needed a clinician
Other exemptions needed information from outside the eligibility system. An acute medical condition had to be validated by a medical professional. That meant a document from a clinician, in a form the state would accept, reaching the state in time to matter. The requirement connected the eligibility system to every provider who treated a Kentucky HEALTH enrollee.
The federal letter went further on disability. It said people classified as non-disabled for Medicaid purposes might still have a disability under the Americans with Disabilities Act or section 504. It said states should evaluate individuals’ ability to participate and the reasonable modifications they needed. That evaluation needs information the eligibility file does not hold.
Who holds the evidence of frailty
Medical frailty is a clinical fact before it is an eligibility fact. The evidence sits in diagnoses, treatment and functional assessments held by providers and, for enrollees in managed care, by plans in their claims and encounter records. A state that relied only on a person’s own claim of frailty would need a form, a review and a decision for each. A state that could draw on claims data might identify likely frail enrollees without a form.
Either route has risks. A form route places the burden on the enrollee and the clinician. A claims route can miss people who are frail but have had little recent care, and can flag people whose diagnoses do not reflect their function. A state could use claims to find candidates and a simple attestation to confirm them, with the reasoning written down.
Exemptions that change during the year
Status is not fixed. The Kentucky terms said that a suspended beneficiary who became pregnant, was determined medically frail, became a primary caregiver or a full-time student, or developed a disqualifying acute condition could reactivate eligibility. So the state needed to learn about changes in status as they happened, as well as at renewal.
That put a reporting duty on both sides. Beneficiaries were required to report changes, as all were under the program’s rules. The state needed a route to receive those reports and act on them within the month. A change that sat unprocessed in a queue would leave a person suspended who should not be.
Caregivers and the household
The caregiver exemption was limited to one exempt beneficiary per household, for a dependent minor child or a disabled adult. Applying it needed household data: who lives together, who depends on whom, and which adult is the primary caregiver. Households change, and the eligibility record does not always follow. Two adults in one home might each believe they were the caregiver, and the system could only exempt one.
The cost of a wrong status
An error in status data had an asymmetric cost. A person wrongly marked exempt would face no requirement until the error was found. A person wrongly marked non-exempt would owe hours they did not owe, and could be suspended for not reporting them. The federal letter’s emphasis on modifications and exemptions implied that the second error was the one to design against.
Borrowed data: deemed compliance from other programs
The second kind of data came from other programs. The Kentucky terms, in STC 45, said some beneficiaries would be deemed to meet the requirement through verified participation in specified activities. Three were listed. Meeting SNAP or TANF employment requirements, or being exempt from them, counted. Enrollment in the state’s Medicaid employer premium assistance program counted. Employment of at least 120 hours per calendar month counted.
What deeming saved
Deemed beneficiaries did not have to document their activities. They still had to report changes in eligibility. For a state, every person it could deem from existing records was a person it would not have to chase for monthly reports. For the person, it removed a monthly task with a suspension behind it.
The federal letter made the SNAP and TANF route mandatory in its guidance. It said people complying with or exempt from those programs’ work requirements must automatically be considered compliant. That turned a policy choice into a data integration requirement. The Medicaid system had to know, reliably, who was compliant or exempt in two other programs.
The integration behind it
STC 48(c) made that explicit. Before implementing, the state had to ensure processes and procedures to seek data from other sources, including SNAP and TANF. It also had to have systems to let beneficiaries efficiently report hours and to let the state monitor compliance.
The federal letter pointed to the integrated eligibility systems states had built. Where health and human services programs shared one eligibility platform, the data might already sit side by side. Where they ran on separate systems, the state needed a match: a regular exchange of records keyed on identifiers both programs trusted.
Premium assistance as a deemed route
The second deeming route, enrollment in the state’s Medicaid employer premium assistance program, drew on data the Medicaid agency held itself. A person whose employer coverage Medicaid helped pay for was, by definition, employed. The terms extended the deemed status to a spouse or dependent enrolled in the same program. That route was cheap to apply, because the record was already in the agency’s hands, and it shows the value of designing deeming around data the agency controls.
Employment data and its limits
The 120-hour employment route raised a different question. It needed a record of hours worked in a calendar month. The Kentucky terms said beneficiaries could demonstrate compliance consistent with the federal rule on verification, 42 CFR 435.945. They did not name an employment data source that recorded monthly hours.
So the employment route depended on whatever the state could match. A worker with steady hours at one employer was easier to verify than a worker with several part-time jobs, gig work or seasonal hours. The second worker might work the hours and still have to report them, because no record showed them.
Matching the same person across systems
Borrowed data is only as good as the match that links it. Two programs may hold the same person under slightly different names, dates of birth or addresses. A match keyed on a Social Security number is strong where both programs have it. Where one does not, the match falls back on weaker combinations, and some true matches are missed.
A missed match has a direct cost here. A person who is compliant in SNAP, but whose SNAP record does not link to the Medicaid record, appears non-compliant. The person then has to report hours the program should have deemed. A state can measure its match rate, review samples of non-matches by hand, and give people a simple route to point out that they already participate in another program.
Agreements and confidentiality
Each partner program has its own confidentiality rules and its own reasons for sharing data or not. A data sharing agreement has to say what data moves, for what purpose, on what schedule, how it is secured and how long it is kept. It also has to say who answers when a shared record is wrong. Drafting those agreements takes legal and privacy staff time, and it has to happen before the first month’s matches run.
Mismatches between programs
Borrowed data carries the source program’s definitions. A SNAP exemption reflects SNAP’s rules, its household definitions and its review dates. A Medicaid system that relied on it inherited its timing and its errors. A person whose SNAP case closed for an unrelated reason would lose the deemed status, and might not know it.
The federal letter told states to describe how they would communicate differences in requirements to people who left SNAP or TANF but stayed subject to the Medicaid rule. That is a notice requirement driven by a data event. The state had to detect the event to send the notice.
Reported data: the enrollee as the source of record
The third kind of data was what the enrollee had to supply. STC 45 said beneficiaries without an exemption had to document participation in qualifying activities on at least a monthly basis. STC 46 set the requirement at 80 hours per calendar month.
What counted
The qualifying activities were broad. The terms listed job skills training, job search, education related to employment, general education and vocational training. They also listed self-employment, subsidized or unsubsidized employment, community work experience, and community or public service. Caregiving for a non-dependent person with a disabling condition counted, as did substance use disorder treatment.
Breadth helped people meet the requirement. It also widened the range of evidence the state had to accept. A pay stub, a school schedule, a volunteer coordinator’s letter and a treatment program’s record are different documents from different sources. Each needed a way into the state’s system and a rule for how much time it proved.
A monthly clock
The monthly cycle set the pace. Hours did not carry forward. STC 47(c) said hours above the requirement in one month could not be applied to another, except when curing a shortfall. A person with a heavy month followed by a light one would still owe the light month.
The cure provisions added a second clock. STC 47(b) gave beneficiaries the month after a shortfall to make up the missing hours while staying current, or to complete a health or financial literacy course. The course option was available once per 12-month benefit period. A suspension took effect on the first day of the month after that opportunity to cure.
Verifying what is reported
The terms let beneficiaries demonstrate compliance consistent with 42 CFR 435.945, the general federal rule on verifying eligibility information. How much documentation a state demanded for each reported hour was a design choice. A state could accept a statement of hours and check a sample. It could instead demand a document for every activity, every month.
The choice moved burden between the state and the enrollee. Demanding documents for everything produced more evidence and more failures to report. Accepting statements with sampling produced less evidence and fewer suspensions for paperwork alone. Neither is free of risk. A state’s choice, and its reasons, belonged in writing, because they set the coverage effect as much as the hours rule did.
Hours that vary
Work that varies week to week fits awkwardly into a monthly count. Shift workers, seasonal workers, people with several part-time jobs and people in the gig economy may average 80 hours and miss some months. The cure month and the reactivation rules softened that. They also added steps, each with its own notice and its own chance for a record to go astray.
A regional start
STC 46 said the requirement would be implemented on a regional basis following implementation of Kentucky HEALTH. Beneficiaries who had not been subject to the requirement in the past five years would get a three-month period before having to meet it. A regional start gave the state a chance to measure its data flows on a smaller population first.
That chance was only useful if the state looked. The first region’s match rates, reporting rates, suspension counts and appeal outcomes would show where the data design failed. A state that rolled out to the next region on a calendar, without reading those figures, would carry the same faults statewide.
Reporting channels
The terms required the state to make reporting efficient. STC 48(c) required systems to let beneficiaries efficiently report community engagement hours. STC 48(e) separately required that qualifying activities and cure activities be available at a range of times and through a variety of means, such as online and in person, at no cost to the beneficiary. So the state’s obligation reached both the reporting systems and access to the activities themselves.
Reporting channels have their own failure modes. An online portal needs an account, a device and a connection. An in-person office needs transport and opening hours that fit around work. A paper route needs mail that arrives. Each channel the state offered reduced the number of people for whom reporting itself was the barrier.
Notices as data
STC 48(d) listed nine kinds of notice the state had to give. They included when the requirement would start, whether a person was exempt and when the exemption would end, and how hours would be counted. They also covered a month of non-compliance and how to cure it, and how to appeal a suspension. Each notice depended on the state knowing the person’s current status and sending it to a current address.
That made contact data part of the requirement. A notice sent to an old address is a notice the person did not receive. In a monthly program, an address error repeats every month.
Consequence data: what the state did, and who it lost
The fourth kind of data was the state’s own record of what it did. Suspensions, reactivations, appeals, terminations and payment changes were all events the state generated. They were also the evidence of what the requirement did to coverage.
Suspension and payment
STC 48(a) required the state to maintain system capabilities to suspend and deny eligibility and to lift suspensions. STC 48(b) required mechanisms to stop capitation payments to a managed care organization when a beneficiary was suspended, and to restart them when the suspension lifted. So a suspension was also a payment event for the plan.
That reached the plans’ own systems. A managed care organization needed to receive suspension and reactivation records, adjust its membership and handle claims for services delivered during a suspension. The plan’s data and the state’s data had to agree.
Screening before termination
STC 48(g) required that disenrollment or denial occur only after a person had been screened and found ineligible for all other bases of Medicaid eligibility. It also required review for other insurance affordability programs. That is a data-intensive check. It needs enough information to test every other eligibility group before ending coverage.
Appeals as a data stream
STC 48(f) required full appeal rights before suspension and observance of due process for anyone whose eligibility would be suspended, denied or terminated for failing the requirement. A hearing let the beneficiary raise additional issues and provide additional documentation. Each appeal was a case file, with deadlines, evidence and a decision.
Appeals are also a signal. A rise in appeals that end in the beneficiary’s favor points to a data problem upstream: a missed match, a status flag not updated, a report lost in a channel. An agency that reads its appeal outcomes as data can find those faults. One that treats appeals only as a legal workload will keep repeating them.
Reactivation
A suspension was designed to be reversible. A beneficiary could reactivate after 80 hours in a 30-day period or, once per benefit period, a literacy course. Reactivation needed the same data flow in reverse: a report received, verified and acted on, a plan notified, capitation restarted. A state that could suspend quickly but reactivate slowly would leave people without coverage longer than the rules intended.
Keeping the record
Every suspension, reactivation and appeal had to be kept in a form someone could later analyze. The federal letter’s evaluation requirements and the monitoring metrics depended on it. So would any later review by an auditor or a court. A system that overwrote a person’s status each month, keeping only the current value, would lose the history the evaluation needed.
Measuring the people who left
The federal letter required states to track outcomes for people who lost coverage. Its monitoring section named metrics on enrollment and termination for failure to meet program requirements. Its evaluation section said states should be prepared to track and evaluate outcomes both for those who remained enrolled and for those who lost or experienced a lapse in coverage.
Tracking people after they leave a program is hard. The state’s records stop at disenrollment. Following outcomes requires surveys, other data sources or matching with other programs, all within privacy law. The letter said as much, conditioning tracking on what federal and state privacy laws permit.
Budget neutrality and enrollment
The letter closed one tempting route. States would not be permitted to accrue savings from a reduction in enrollment that might result from the demonstration. That kept the budget calculation from rewarding coverage loss. It did not stop coverage loss from occurring.
Kentucky’s own application, as the district court later described it, projected savings by reducing its Medicaid population by an estimated 95,000 people. The coverage effect was in the state’s own numbers before the program began. Whether the federal approval had weighed it became the central question in court.
How this differed across the organizations involved
The four kinds of data fell on different organizations in different ways.
State Medicaid agencies
The state agency owned all four. It had to classify status, run the matches, receive reports and act on them. Its eligibility system was the place where the requirement became real. A state with an integrated eligibility platform covering SNAP and TANF started with borrowed data in reach. A state with separate systems started with an interface to build.
Human services and workforce agencies
SNAP and TANF agencies became data suppliers to Medicaid. Their records decided who was deemed compliant. Workforce agencies, which run job training and job search programs, became potential sources of reported hours. Neither had been designed to send monthly participation data to a health program. Each would need to agree the fields, the schedule and the identifiers with the Medicaid agency, and to correct its own errors when Medicaid relied on them. A workforce agency whose attendance records lagged by weeks would pass that lag to every person whose hours depended on those records. The partner’s data quality became part of Medicaid’s eligibility accuracy, with no change in who managed it.
Managed care organizations
Plans received the consequence data as membership changes and capitation changes. They also held something the state needed: claims and encounter records that might show treatment, frailty or acute conditions. Whether plans could share those signals to support exemptions was a question for the state’s contracts and privacy rules. Plans also run member services lines. A member who learned of a suspension from a denied pharmacy claim might call the plan first. The plan’s staff would need to know the state’s rules well enough to point the member to the right step.
Hospitals, clinics and health centers
Providers were the source of medical validation for acute condition exemptions and part of the evidence for frailty. Their front desks also saw the effects of suspensions as coverage checks that failed. A clinic serving a large number of Kentucky HEALTH enrollees would need a process for the exemption paperwork and for patients who arrived suspended.
CMS as the receiving end
The federal agency received the consequence data. The letter said CMS would review written state reports quarterly and work with each state on metrics for quarterly and annual monitoring reports. It said a final evaluation design was due for CMS approval no more than 180 days after approval. CMS’s own capacity to read and act on those reports decided whether monitoring was oversight or filing.
Evaluators
The letter required independent evaluations, using comparison groups, claims data and surveys of current and former enrollees. Evaluators needed the same consequence data the state generated, at the level of individuals, under privacy protections. An evaluation designed after launch would find that some of the data it needed had not been kept.
Legislatures and budget offices
The budget neutrality terms limited what the state could claim as savings. State budget offices still had to fund the systems work, the call centers and the notices. They also had to plan for enrollment that would move month to month with suspensions and reactivations. A legislature that expected savings from fewer enrollees would find that the federal terms did not let the state count them toward neutrality.
County and community partners
In states where counties run eligibility, the county office would hold the casework. Community organizations that offered volunteer placements would become sources of evidence. Each needed a way to document hours that the state would accept.
Objections, in order
We take five objections, numbered in the order we address them.
1. The requirement was about work, not data
The first objection is that the policy’s purpose was employment and health, and data was a means. That is the stated purpose. But a requirement enforced through suspension operates on what the state can verify. A person who worked and could not prove it faced the same suspension as one who did not work. The data design set who bore that risk. A state that wanted the policy to reward work had a reason to make proof of work as cheap as possible. That aligns the policy’s stated purpose with the data argument rather than against it.
2. Exemptions and deeming protected the vulnerable
The second objection is that the exemptions and deeming routes were broad. They were broad on paper. Each worked only where the state’s records, or a partner program’s records, showed the qualifying fact. An exemption the state could not see had to be claimed, documented and processed. People who needed an exemption urgently, such as those with a new illness, were also less able to complete that process quickly.
3. Monthly reporting is a small burden
The third objection is that reporting 80 hours a month is simple. For a person with one steady job and an online account, it might be. The terms required the state to make activities available at a range of times and through a variety of means, such as online and in person, which suggests CMS and Kentucky expected that one route would not suit everyone. A small burden repeated monthly, with suspension behind it, is a different thing from a one-time form. The relevant measure is not the burden for the typical enrollee, but the number of enrollees for whom any one month goes wrong.
4. Coverage loss was a policy choice for the state
The fourth objection is that a state may accept lower enrollment as a trade-off. The court did not decide that question in the abstract. It held that the Secretary had to consider the effect on coverage and had not done so adequately. Our point is narrower. Whatever the state’s view, the size of the effect was a function of its data design, and that design was within its control. A state that accepted some loss could still choose to minimize the part caused by records rather than by conduct.
5. Better systems would solve it
The fifth objection is that the problem was technology that could be fixed. Better matching would reduce the number of people who had to report. It would not reach people whose qualifying activity appears in no record. Some reporting burden would remain however good the systems were, and the requirement’s coverage effect would depend on how the state handled that remainder. Better systems also take time and money to build. The assurances we read in STC 48 named capabilities the state had to have before starting. They did not set measured thresholds, such as a match rate, that it had to reach first.
What a program director could check
These checks follow from the four kinds of data. They are our suggestions, not requirements.
- For each exemption, which record shows it, how current is that record, and who updates it?
- For each deemed-compliance route, which program supplies the data, on what schedule, and what happens when its case closes?
- What share of people subject to the requirement can be cleared without any report from them, and is that figure measured each month?
- Does every notice draw on current status and contact data, and are returned notices tracked?
- Does the state measure what happens to people after suspension or termination, within privacy law, as the federal letter asked?
The third check is the one that decides coverage. A state that can answer it knows how much of its population depends on monthly reporting. A state that cannot does not know the size of the effect it is about to have.
The figures behind these checks should be read monthly, by region and by group. A monthly program changes monthly. A quarterly report to CMS, read once, will describe problems after several cycles of suspensions have already run. A program director who sees the first region’s figures each month can change notices, channels or matches before the next region starts.
The checks also need owners. Status data belongs to the eligibility policy and systems teams. Borrowed data needs a data sharing agreement owned jointly by the Medicaid agency and each partner agency. Reporting channels belong to the agency’s customer operations. Consequence data belongs to the Medicaid program’s monitoring and evaluation staff, with the privacy office.
With hindsight
Kentucky HEALTH was scheduled to take effect on July 1, 2018. On June 29, 2018, the federal district court in Washington, D.C., vacated the approval and remanded it to HHS. The court held that the Secretary “never adequately considered whether Kentucky HEALTH would in fact help the state furnish medical assistance to its citizens.” That, it said, was “a central objective of Medicaid.” It said the Secretary never once mentioned Kentucky’s estimate that 95,000 people would leave its Medicaid rolls. The court issued its opinion on an expedited basis because of the July 1 start.
HHS reapproved the program on November 20, 2018, effective April 1, 2019. The court vacated that approval too, on March 27, 2019. The government’s new argument was that coverage lost under the program would be dwarfed by the coverage lost if Kentucky ended its expansion altogether, as the state had threatened. The court held that the Medicaid Act did not leave the Secretary so unconstrained, and found the reapproval contrary to the Act and arbitrary and capricious.
Challenges to approvals for other states followed. One reached the Supreme Court as No. 20-37, with the federal government and the State of Arkansas as petitioners. On December 4, 2020, the Court granted review and consolidated it with No. 20-38. In February 2021, the federal government asked the Court to vacate the D.C. Circuit’s judgments and remand. The cases were removed from the March 2021 argument calendar and, in April 2021, held in abeyance. On April 18, 2022, the Court granted the motion and vacated the D.C. Circuit’s judgments. It sent one pair of cases back with instructions to direct the district court to vacate its judgment and dismiss the case as moot, and the other pair with instructions to direct the district court to remand to the Secretary. So no Supreme Court ruling on the merits followed.
Congress then wrote a requirement into the statute. Section 71119 of Public Law 119-21 was approved July 4, 2025. It requires states to make community engagement a condition of eligibility for adults in the Affordable Care Act expansion group and certain waiver adults aged 19 to 64. States must begin no later than the first quarter starting after December 31, 2026. The Secretary may exempt a state that is demonstrating a good faith effort to comply, and such an exemption expires no later than December 31, 2028.
The statute speaks directly to our data argument. It requires states to verify compliance, deemed status and exclusions using reliable information available to the state, such as payroll data or Medicaid payment and encounter data. Where possible, the state must do so without requiring the individual to submit more information. It also requires a notice and 30 calendar days to show compliance before the state acts on a failure to verify, with coverage continuing during that period. Before disenrolling anyone, the state must first check for any other basis of eligibility, as Kentucky’s STC 48(g) had required. In our reading, both provisions put the burden of finding data on the state first.
CMS published an interim final rule on June 3, 2026, effective July 31, 2026. Its summary says states must implement the requirement no later than January 1, 2027. It says the rule covers who must demonstrate community engagement, qualifying activities, exceptions, verification of qualifying activities, outreach, steps for noncompliance and new state reporting.
On October 29, 2026, CMS published notice of a re-established computer matching program between CMS and state administering entities. Its summary says the program supports eligibility determinations for coverage programs under the Affordable Care Act, including Medicaid, and also supports implementation of statutory requirements on Medicaid community engagement. In our reading, that is a formal channel for the borrowed data of our second section.
What held up is that the coverage effect depends on the data design, and that, as the court held twice, the Secretary had to consider the effect on coverage before approving. What changed is the legal footing. Kentucky’s 2018 requirement rested on demonstration authority, and both of its approvals were vacated in court. The 2025 requirement rests on statute, with verification rules that start from data the state already holds.
Spotted an error? Tell us and we will check it and correct the article where needed.