How Medicaid Eligibility Software Can Deny Disabled People’s Reality

When Medicaid Software Decides Whose Story Counts

A benefits system does more than process information. It decides which information counts.

That distinction matters in Medicaid, where an error can interrupt access to medication, personal care, transportation, or treatment. NPR’s July 20 report on Deloitte-run eligibility systems describes disabled people being denied coverage through state systems and warns that new policy changes could deepen the problem. The immediate story is about technology and administration. The deeper problem is interpretive power: who gets to define a person’s reality when a rule, data field, or automated notice conflicts with the life that person is actually living?

Automation does not remove judgment

Government software is often described as if it simply applies neutral rules. In practice, every Medicaid eligibility system contains decisions made long before an applicant sees a screen: which facts can be entered, which records override others, how inconsistencies are treated, what triggers a denial, and how much work a person must do to correct the result.

Those choices may be distributed among lawmakers, agencies, contractors, administrators, and software teams, but they remain choices. A computer can apply them quickly and consistently without making them fair.

Michigan’s public contracting records describe Deloitte support for systems involved in real-time Medicaid enrollment and communication across state services. That establishes the contractor’s operational role; it does not, by itself, establish responsibility for every reported denial. The important structural question is broader: when Medicaid eligibility is mediated through a large technical system, where can a disabled person challenge the system’s interpretation before losing care?

Whose bias becomes policy?

The paper Biasology: Toward a Framework of Epistemic Liberation for the Misnamed and Misdiagnosed asks a useful question: whose bias becomes policy?

Biasology names institutional bias as the distortion that occurs when systems erase people they cannot classify. In Medicaid administration, this can happen without anyone announcing a prejudiced intention. A rigid verification rule, an outdated database match, a notice written for administrative convenience, or a workflow that assumes consistent capacity can all turn a limited institutional picture into an official truth.

The applicant then faces a translation tax: the emotional and cognitive labor of converting illness, disability, unstable housing, fluctuating function, or missing paperwork into the exact form the system recognizes. If they cannot do that quickly enough, the institution may treat the failure of translation as evidence that they are ineligible or unresponsive.

The burden is especially severe because the person correcting the record may already be managing pain, fatigue, cognitive overload, limited transportation, or dependence on the very services now at risk. The Medicaid appeals process can become an accessibility test disguised as due process.

The dangerous authority of a clean record

Administrative systems produce tidy outputs: eligible or ineligible, verified or unverified, active or closed. Human lives are not so clean.

A database can be internally consistent and still be wrong about the person. It may faithfully reproduce an incorrect source record, omit context the schema cannot store, or interpret silence as noncompliance when the actual barrier was disability. Once that output enters an official record, workers and reviewers may defer to it because it appears more objective than testimony.

This is where automation amplifies institutional authority. The system’s version travels easily between departments. The person’s correction must fight its way back upstream.

Biasology calls systems like this epistemically incomplete. The phrase does not mean that all classification is useless. It means a system’s inability to represent a person’s reality should be treated as a limitation of the system, not automatic proof against the person.

What to do after a Medicaid eligibility denial

A denial notice should not be treated as proof that the system is correct. Federal Medicaid policy requires states to provide an opportunity for a fair hearing when eligibility is denied, when a person believes an agency action is wrong, or when the agency has not acted with reasonable promptness.

Because procedures and deadlines vary by state, a person receiving a Medicaid denial should:

• read the notice immediately and identify the stated reason, effective date, and appeal deadline;
• contact the state Medicaid agency using the information on the notice or the official state contact listed through Medicaid.gov;
• request the case record or the specific data used to make the decision;
• preserve notices, screenshots, upload confirmations, names, dates, and reference numbers;
• ask how to request a fair hearing and whether coverage can continue while an appeal is pending;
• seek help from a legal-aid, disability-rights, or benefits-advocacy organization when the process is inaccessible or urgent.

This is general information, not legal advice. The important point is structural: a denial produced by software still has to be reviewable by people.

What accountable Medicaid technology requires

A humane eligibility system needs more than accurate code. It needs safeguards for interpretive failure.

• First, every adverse decision should be explainable in language a person can use. A notice should identify the decisive information, its source, and the exact path to correction.
• Second, human review must be meaningful. A worker needs authority to consider evidence the automated workflow could not understand, and applicants need accessible ways to reach that review before care is interrupted.
• Third, agencies should measure the cost of error on the person, not only processing speed or improper-payment rates. Denying necessary care and approving an ineligible case are both errors, but their consequences are not interchangeable.

Finally, contractors and public agencies need clear accountability. Procurement complexity cannot become a fog in which no one owns the interpretation embedded in the system.

The real test

The question is not whether Medicaid agencies should use software. At this scale, they must. The test is whether the technology helps public servants understand eligibility or quietly replaces understanding with classification.

When a disabled person’s account conflicts with a system-generated result, the answer cannot be to trust the cleaner record automatically. The discrepancy is a signal that the frame may be incomplete.

An accountable system pauses, listens, and makes correction possible before an administrative interpretation becomes a medical crisis.

Framework Attribution

The interpretive framework used in this essay comes from Ian P. Pines’s paper Biasology: Toward a Framework of Epistemic Liberation for the Misnamed and Misdiagnosed.

Sources

• [Read at NPR] “Deloitte-run systems denied Medicaid to disabled people. New laws could make it worse,” published July 20, 2026. 
• [Download] State of Michigan, “Change Notice No. 13 to Contract No. 071B2200176 / Deloitte Consulting LLP,” describing continued support for real-time Medicaid enrollment and related MDHHS systems. 
• [Read at Medicaid.gov] “Eligibility Policy,” including federal fair-hearing requirements and state Medicaid contact guidance. 
• [Words.HAIR], “Translation Tax.” 
• [Read the Paper] Pines, I. P. (2025). Biasology: Toward a Framework of Epistemic Liberation for the Misnamed and Misdiagnosed. Knowledge Commons.