Medicaid work requirements built on invented evidence will harm the people they claim to help

The Centers for Medicare and Medicaid Services (CMS) has issued a new Interim Final Rule with comment period (IFC) implementing the Medicaid “community engagement” requirements included in H.R. 1. The IFC is fundamentally defective in both its substance and its analysis.

With respect to the analysis justifying the rule, the agency’s mandated Regulatory Impact Analysis (RIA) does not conform to the standards required of such analyses, as stipulated in Executive Order 12866 and the Office of Management and Budget (OMB) Circular A-4. It substitutes unsupported assumptions for the best available empirical evidence, counts speculative benefits while omitting well-documented costs, and produces estimates of coverage loss and federal savings that are inconsistent. With respect to the substance of the rule, CMS has added conditions, not required under the statute, that will make it even more difficult than the statute intended for people with mental illnesses and substance use disorders to retain coverage.

These are not technical shortcomings. This rule will determine whether millions of low-income Americans retain health insurance.

H.R. 1 requires certain Medicaid beneficiaries to document 80 hours a month of work or other qualifying activities to retain their eligibility for Medicaid. This IFC sets forth the rules defining which beneficiaries will and will not be exempt from this “community engagement” requirement. Under Executive Order 12866, economically significant regulatory actions, such as this IFC, must be accompanied by an RIA that assesses the costs and benefits of the action. In this case, the key parameter that will drive the costs and benefits of the regulatory action is how many people will become employed or otherwise community-engaged under the threat of losing Medicaid coverage. OMB requires agencies to base regulatory analyses on the best reasonably obtainable scientific, technical, and economic information. CMS acknowledges that it does not estimate this key behavioral response parameter from any such information. Instead, it assigns subjective weights to hypothetical scenarios. In its assessments of costs and benefits, CMS simply assumes that roughly one-third of affected beneficiaries will move into qualifying activities because of the requirement.

That assumption is wildly out of line with a substantial body of empirical evidence. Three distinct, robust lines of research, all consistent with one another, could have been used to estimate the likely impact of the requirement. First, studies of welfare-to-work programs, which condition receipt of cash benefits on employment, generally find small employment effects, with effects of under 5% even in scenarios where the work requirements in the programs are accompanied by childcare, transportation assistance, case management, and other supports—no such supports are included in the Medicaid rule. Second, studies of the Affordable Care Act’s Medicaid expansion—the converse of the new work requirements—find that gaining Medicaid has little or no effect on employment, implying that withdrawing coverage is unlikely to produce a substantial increase in engagement. Finally, studies of recent work requirements in the Supplemental Nutrition Assistance Program (SNAP) and Medicaid—including Arkansas’s Medicaid work requirements waiver—have found substantial coverage losses but no measurable increase in employment.

CMS’ assumption that one-third of those subject to the work requirement will become community-engaged is over six times greater than the largest estimate in any of these literatures. That assumption is an invention from which the agency mechanically generates its estimates of employment, earnings, costs, benefits, and government savings.

Because those estimates carry through the entire analysis, they generate further inconsistencies and distortions. The one-third engagement estimate leads CMS to predict that only about 3.1 million to 3.3 million people will lose Medicaid coverage annually. By contrast, the Congressional Budget Office (CBO) estimated that the statutory work requirement would reduce Medicaid enrollment by approximately 5.7 million people. Yet despite this much smaller headcount, CMS simultaneously projects approximately $350 billion in federal savings over 10 years—10% more than CBO’s estimate of $317 billion.

These figures are difficult to reconcile. If CMS expects fewer people to lose coverage but more money to be saved, it must implicitly assume that those losing coverage are dramatically more expensive than the beneficiaries reflected in CBO’s calculation. CMS’s numbers appear to imply average federal Medicaid spending of roughly $11,000 for each disenrolled person, compared with about $5,600 in CBO’s estimate.¹ But Medicaid spending is highly concentrated. Many of the most expensive beneficiaries have serious health conditions that should qualify them for exemptions or temporary hardship protections. CMS does not adequately explain how its relatively small group of projected disenrollees can generate such large savings.

The implausible engagement assumption shows up again in the agency’s claims about the IFC’s benefits. Here, the agency claims that those shifted to engagement through the work requirement will become healthier from being induced into employment. The research basis CMS cites for the assumption that work makes you healthy is entirely based on associational studies of the health of workers and non-workers.

Healthier people are more likely to work. Poor health makes it harder to obtain and retain employment. Education, family resources, housing stability, and other factors affect both employment and health. Studies showing that employed people are healthier than unemployed people in no way establish that requiring a Medicaid beneficiary to work will make that person healthier. Indeed, the limited causal research, which relies on changes in retirement rules and other sources of plausibly exogenous variation, has produced mixed results: some studies find benefits from continued work, while others find that working harms health (for example, by increasing the risk of workplace injuries). There is no credible basis for assigning large health benefits to the hypothetical employment increases assumed to be generated by this rule.

At the same time, CMS largely omits the strongest causal evidence in this area: Medicaid coverage improves access to care, financial security, and health. Randomized and quasi-experimental studies have linked insurance coverage to improved treatment and, in several very credible studies, lower mortality. Other research shows that, at least for some populations, access to effective treatment can improve employment. Randomized studies of depression care, for example, have found increases in job retention, hours worked, and employment.

In other words, health coverage keeps people healthy and may help people work. Taking coverage away has direct harms and may make employment less likely. CMS’s analysis largely reverses this relationship: it assumes that work produces health while ignoring the possibility that health care produces both work and health.

The rule is especially poorly designed for people with mental illnesses and substance use disorders.

In the original statute, Congress provided exemptions for people with qualifying behavioral health conditions. CMS, however, would require many individuals to demonstrate not merely that they have a qualifying diagnosis, but that the condition significantly impairs their ability to comply with the work requirement.

Standard administrative reporting, such as medical claims, does not contain information on functional ability. That means that meeting the exemption will require a new miniature disability determination. States and clinicians will have to assess functional capacity, collect documentation, and decide whether a particular condition prevents compliance with 80 hours of monthly activities. States conduct such capacity determinations in conjunction with Social Security disability determinations; a costly, challenging process that provides applicants with important procedural protections (but nonetheless makes consequential errors).

Requiring such functional status determinations for people with mental illnesses and substance use disorders is particularly damaging because these conditions can impair memory, concentration, planning, and motivation. People with these conditions are likely to find it particularly difficult to navigate bureaucratic systems. The people most impaired may be those least able to prove that impairment.

The functional status criteria also fail to account for the fact that behavioral health conditions are chronic and recurring. A person with opioid use disorder, schizophrenia, bipolar disorder, or recurrent major depression may be stable enough to work for one month, experience a relapse the next, miss a reporting deadline, and lose the Medicaid coverage financing the treatment that helped maintain stability and employment. Current functioning while under treatment is not proof that an exemption is unnecessary. It may instead be evidence that treatment is working.

The statute particularly called for an exemption for those participating in substance use treatment programs. The CMS rule places very restrictive conditions on what participation means. In particular, it limits the exemption to care provided by nonprofit providers. However, nearly half of people receiving substance use treatment are served by for-profit organizations, and the share is particularly high among opioid treatment programs providing methadone. That means that patients receiving substance abuse treatment could lose exemption from work requirements because the providers who happen to be available in their communities are for-profit. That will be particularly damaging in rural areas and other places with thin treatment networks.

The IFC is defective in its substance and its design. A defensible Regulatory Impact Analysis would use the following information:

• The extensive causal literature to estimate employment effects,
• Quantify the health consequences of coverage loss, and
• Reconcile its enrollment estimates with its claimed federal savings.

A defensible rule under the statute would:

• Automatically identify behavioral health exemptions using claims, pharmacy, managed-care, and behavioral health agency data;
• Accept attestations from clinicians, case managers, and treatment programs;
• Recognize relapse and fluctuating functioning; and
• Provide rapid reinstatement when illness contributes to noncompliance.

Instead, CMS’s analysis assumes away the central empirical question, inflates speculative benefits, omits documented harms, and then adds administrative barriers that are likely to exclude precisely the people Congress intended to protect.