Arkansas ARChoices / ARIA

Arkansas's Department of Human Services used an algorithm called Resource Utilization Groups, or RUGs, from 2016 to set Medicaid home-care hours. It turned a nurse's answers on a standard assessment into a weekly number of hours. From 2019 the state replaced it with ARIA, the Arkansas Independent Assessment, built by Optum.

What changed for people

Before 2016, nurses used their own judgment to set each person's hours. The algorithm replaced that judgment, and the nurses kept assessing. The lawsuits over the cuts say the assessors were bound by its number, with little room to change it.

The sources put the program at about 8,000 older and disabled people. About 47 percent saw their weekly hours cut, some by nearly half. Bradley Ledgerwood, who has cerebral palsy, had been assessed at 56 weekly hours. In February 2016 the state cut him to 32.

Documented accounts say neither the people affected nor even the assessors could reconstruct the explanation for a cut.

What went wrong with the formula

RUGs assigned too few hours at every level of medical need. It did not account for certain medical conditions and other factors. That mistake caused incorrect calculations for hundreds of people.

Because one formula set everyone's hours, the switch changed thousands of people's hours at once. The flaw about medical conditions gave hundreds of them wrong hours.

What the federal court found

Legal Aid of Arkansas filed a federal lawsuit, Jacobs v. Gillespie, after hearing directly from people who had lost hours. The federal court found the state had violated due process. Due process is the right to fair notice and a fair chance to challenge a government decision. The state had not explained the reasons for the cuts, so people could not know what they needed to prove to keep their care.

The federal court's due-process finding centred on people being unable to understand or meaningfully contest their hours.

The state-court case

Ledgerwood and six other residents sued in state court. They alleged the switch had cut their care by an average of 43 percent.

In 2017 Arkansas's Supreme Court upheld an order protecting those seven people's hours. It held they were likely to show the state had adopted the method without following the rulemaking process state law requires. The ruling reached the procedure, not the algorithm.

A state court later invalidated the algorithm on rulemaking grounds. Once the state adopted the rule properly, the injunction in the case was dissolved in 2019. An injunction is a court order that stops or requires an action.

What replaced it

Starting in 2019, the state replaced RUGs with a new system for assessments, eligibility, and care allocation. Its assessment tool, ARIA, was built by Optum. The state uses ARIA to set eligibility and an individual budget cap on the cost of a person's services, whatever their actual care needs. A separate time-task tool sets a number of minutes for each care task.

ARIA left over a quarter of the program's participants determined wholly ineligible for care. People lost benefits immediately, although rules required them to keep benefits during an appeal.

Beneficiaries sued again, in Elder v. Gillespie, and the state lost on appeal in 2022. The settlement required the state to pay nearly $500,000. It also required that people keep their benefits while they appeal and receive a detailed explanation of the reasons for a cut.

The central dilemma

A tool meant to support the nurses' judgment replaced it instead. The people whose care was cut could not understand or contest the number that cut it.

What the available tools can and cannot address

A failure pathway is a link between two parts of the network, where a mistake made by one part can be passed on to the other. A tool closes a pathway when mistakes stop passing along it.

This case has a budget of 12 units. Explore (No Targets) sets no targets. There, and under Service Targets Only, two tools costing 4 of the 12 units are enough to keep mistakes from building on one another across the network. Under Service Targets Only, the same two also meet the service target. The service target asks that the algorithm stay useful for setting care hours. One such pair is Mark AI-written records with Escalate checks, at 2 units each.

Under Service and Safety Targets and All Governance Targets, you must also close every failure pathway. There the targets are not fully addressable with the available tools. Within the budget, the Lab tried all 14,222 allowed combinations of the eleven tools and their settings under each of those two levels. None meets every target.

Money is not what stands in the way. With the budget set aside and every tool applied at its stronger setting, one failure pathway stays open. It is the pathway where the allocation on the record becomes the care hours a person receives.

No tool offered here closes it, because it is how the program delivers care. No tool offered here adds the step that would let a person understand and contest a change first. This is a finding about the deployment, not a gap in your approach.