Algorithms used to determine Medicaid eligibility for home care vary widely among states

STAT News – Friday, October 24, 2025
By O. Rose Broderick

The massive tax cut bill President Trump signed in July is forcing states to grapple with a pivotal question: With federal Medicaid spending projected to drop drastically, which patients should get care at home, and how much care should they get?

In the United States, 3 million elderly and disabled people use home and community-based services, ranging from an hour or so a day up to around-the-clock help with various therapies, bathing, dressing, and other needs. States are not required to fund these programs, but many have chosen to use Medicaid funds to cover such care, accounting for two-thirds of all home care spending in 2022, according to a KFF analysis.

Public health experts say states will likely cut back such optional coverage now that the federal government is projected to send $1 trillion less to states for Medicaid over the next decade, undercutting the disability community’s decades-long push for people to receive care in their communities, not institutions.

In recent years, states have increasingly turned to private vendors to craft algorithms to determine eligibility for these services. Some researchers worry that too many of these decisions are being made without public input or scrutiny and that these determinations are based less on health and more on economics.

“These tools, they get kind of talked about as if they’re measuring people’s needs, but they’re actually attempting to limit how many people are eligible for home care, to try to keep the costs down,” said Emily Paul, a project director at Upturn, a research and advocacy nonprofit focused on technology.

To understand how states determine who is eligible for home and community-based care and the algorithms underpinning the selection, STAT spoke with Paul about a report she co-authored and published Tuesday that analyzed five states’ algorithms. She found that supposedly clean-cut lines of who should receive care are a bit murkier and politically motivated than at first glance, and that the algorithms varied widely among the states.

This interview has been edited for length and clarity.

How are these determinations made?

There’s two steps of the eligibility determination: a face-to-face standardized assessment, which is a federal requirement that states use. … I believe it’s over 250 questions, and the assessor gives ratings. And then the ratings get fed into an algorithm.

What questions do assessors ask?

Functional status, so the activities of daily living. Bathing, dressing, and asking how much support someone needs on those things. There’s a scale, and it starts from being able to do it totally independently, and goes to needing complete assistance. It’s based on the last three days. So it’s like, in the last three days, when have you taken a bath? And have you been able to do that by yourself? Have you needed assistance? If you needed assistance, what kind of degree of assistance? They also assess cognition.

If you get enough points to meet or exceed a certain threshold, then you’re eligible. And then you’ll move to the next step, which could be a waitlist and then would be some kind of care planning process. So after you’re eligible, there’s still the question of figuring out, like, how many hours of care are you going to get budgeted for? The eligibility determination is just like, yes or no, are you eligible? And then there’s the budget allocation, which does often use the same assessment as input, but it’s a different determination. You can sit on a waiting list for years until there’s a spot in the waiver program.

Who makes the decisions? The vendors or the state?

The state agency, we saw this in our case studies, they go through a process with the vendors of being like, ‘Okay, how do we want to determine eligibility in this state?’ And the vendor might make certain recommendations, but it really varies.

In some of the states we looked at, incontinence has a really significant impact on eligibility. In other states, it’s not even included in the algorithm as a question. … What we saw is almost the idea of a dial, where the state and the vendors are discussing what percentage of currently eligible people are going to lose eligibility with this new algorithm and finding a tolerable percentage.

How do they determine which dials get turned up or down?

Honestly, we don’t fully understand why New Jersey does not consider incontinence, or D.C. does not consider incontinence, and Mississippi considers it weighs really heavily. A lot of it seems to be vendor recommendations.

In Nebraska, we have meeting minutes from the agency and the vendors which are very revealing. The vendors are quoted as saying to the Nebraska agency, like, ‘Hey, we looked at your algorithm and compared it to other states, and you guys actually are essentially being more generous with eligibility than these other states. Do you want to tighten up your eligibility? Because you’re kind of counting people as needing help if they need supervision help, which some states won’t even consider as needing help.’

Did agency officials take their recommendation?

In that case, the agency was like, ‘No, we don’t want to. We would rather more people be found eligible than people not be found eligible who need the services.’ But it’s like this everywhere, these behind-closed-doors conversations … where the vendors come in with recommendations that have worked in other states, and then there’s this back-and-forth to get to an algorithm that is, like, striking the right balance, essentially, in their view, of people who are going to be ineligible.

One of the vendors was also quoted as saying, the idea of eligibility and the level of care being clear cut is a myth, but the materials about the determinations really emphasize that this is a valid and reliable assessment, and we’re implementing this automation so that we can determine the right people who needs services, that it’s consistent and reliable. I can’t say it was shocking to read, but it was somewhat surprising to see it so explicit.

Is there evidence that these types of conversations are happening in other states?

It’s hard to say. Missouri was transparent in a way that no other state we’ve seen has been because when they were revising their algorithm, they published the first version of it for public comment, and [my colleague] did work with them and did this audit of the algorithm, and that found that 66% of currently eligible people in the sample that they had would be ineligible. And the advocates took that to the state, and they were like, ‘Oh, whoa. We didn’t mean to do that.’ So then they revised it and by the time they were actually implementing, they were anticipating 8% to 9% of people to be cut off.

What’s the history of these determinations?

This expansion of Medicaid home and community-based services started in the early ’80s, when there was political pressure from disability rights activists on independent living and de-institutionalization. And there was also this big concern in the federal government about the rising cost of long-term care. So the federal government did this pilot project to see, okay, can we try to replace institutional care with home care so that it will be cheaper? But the only way it would actually be cheaper is if they limited it to only the people who would otherwise be in a nursing facility.

From the outset, they wanted to try to limit who could get home care to only people who would actually be in a nursing facility and be costing the government money in a nursing facility if they weren’t getting that health care. It’s not actually about need, right?

It seems what you’re saying is that these determinations are not made for health reasons. What’s the right solution, more federal regulations?

We support something like the [Home and Community Based Services] Access Act [in Congress], which is trying to say, ‘here’s this more basic definition of eligibility that’s based on your physician.’ But it’s not like we think, ‘Oh, if this were standardized at the federal level, it would be fine or better.’ We don’t see a technical solution as the way to resolve this.

Our concern is that a state’s reliance on automation depoliticizes this question of who gets home care. And so with the report, we’re trying to demystify the tech and expose it as a political tool. The agency, the vendors know that it’s a political tool. They know that it’s not this objective, scientific exercise, and so we want advocates to also know that and engage with the algorithms in that way and not get sucked into this technical conversation and these questions of accuracy.

Access this article at its original source.