The Dose That Never Moves

Ian will join HTN for a live discussion on Thursday, November 19th at 3 p.m. Eastern time. You can register to join here.

The first time I dosed a diuretic on my own was at three in the morning during my intern year. The patient was an older man with heart failure who had come in short of breath, fluid backing up into his lungs, and legs that pitted when you pressed on them, where 40 milligrams of furosemide, IV, was a reasonable starting dose.

Then you watched. Urine output over the next few hours. Weight in the morning. Creatinine and potassium on the morning labs. If nothing happened, you might double the dose. If his kidneys worsened or his blood pressure dropped, you backed off. Nobody knew what the exact right dose was when we started. The right dose was whatever let him breathe without wrecking his kidneys, and we would know it when we saw it.

Dosing is a hypothesis in action. You test it against the patient, and you reevaluate. You never look up the state cap on diuretics.

Applied behavior analysis (ABA), the therapy most autistic children with intensive needs end up in, works differently. The number is set at initial assessment, by the provider, often at 30 to 40 hours a week for high-needs children. It goes into the first authorization request, the payer approves it for six months, and it often rolls over. It may roll over again and again. A child’s hours in year three can look a lot like the hours in month one. Industry practice guidelines put comprehensive ABA at 30 to 40 hours a week.

Why it never moves is multifactorial but not mysterious. Fee-for-service pays by the hour. Reauthorization is a documentation exercise more than a clinical one, where goals get updated and the hours often come through unchanged or trimmed at the margin. Reauthorization asks whether the child still has deficits, and for a child with autism, particularly profound autism, the answer is almost always yes, even after a strong response to therapy. The question it should ask is whether the child is responding, and how much, and whether this many hours is what the response requires. Nobody is paid to ask that question or recommend fewer hours. And families, reasonably, fear that fewer hours might be less help and sometimes push back.

The consequences and backlash of such dynamics are predictable, and they have arrived as ABA costs have exploded. Across the US, ABA therapy spending is up 421 percent between 2021 and 2025, and federal audits of individual states have found millions in improper payments. Nebraska capped Medicaid ABA at 30 hours a week in 2025. Indiana put a 4,000-hour lifetime limit on comprehensive ABA this spring. Virginia’s new budget holds ABA to 20 hours a week, with exceptions on documented medical necessity, pending federal approval. Colorado, where I live and where Soar was founded, just proposed its own cap.

The payer’s logic is understandable. Medicaid directors across the United States are watching ABA costs go up, cannot always see whether their members are making progress, and naturally reach for caps.

But a cap is just the same mistake from the other side: a dose that never moves, but set lower. The first kids it reaches are the ones who may actually need the full 30 or 40 hours, children with profound autism and high support needs, whose families will now have to win an exception, a process that falls hardest on the families with the least slack.

What does the research say? The National Academies’ 2025 review of TRICARE’s autism benefit found that the general pattern across studies is that more intensity and longer duration produce stronger outcomes but that lower-intensity programs also produce real gains. It concluded that 10 to 20 hours a week may produce good outcomes for many autistic children, that some with profound autism may need 30 to 40, and that it is not appropriate to specify an amount that fits all, or even most, of them. The committee declined to endorse a blanket hours recommendation. The report’s own suggestion is regular evaluation of progress toward goals to decide whether intensity should change.

Overall, a balanced reading of the literature says that clinical benefit rises with hours but then flattens as dose goes up, and sits at a different height for different kids. Lower hours can work for many children; others make more progress with more hours. The dose should be individualized. This is the type of pattern that our health system should titrate and adjust against. It is a verdict against both numbers: 40 for everyone and 20 for everyone are the same error just set at different heights. And both are unlike the furosemide order I wrote at 3 a.m., which always came with a plan to monitor and adjust by the morning and wasn’t clouded by incentives.

That is the proposal for ABA: treat hours as a dynamic dose. Start at a defensible number for the particular child, given age, support needs, and whatever else is in the picture, and attach a plan to move it up or down as the child responds.

Each side of the table owns half of that plan.

Providers own the clinical judgment. The first test of it is whether provider recommendations actually vary. If every child leaves an intake assessment with 35 hours, that is a template, not a judgment. Hours should spread with a child’s age, support needs, learning rate, and commitments. Before the first hour is authorized, the provider and the family should set a baseline and write down what a good response looks like in 90 days, in terms a parent and a reviewer could both check: communicating needs more reliably, getting through daily routines with less support, using a skill at home that was only ever seen in the clinic. They should write down what too much looks like, too: regression, exhaustion, interference with school or family life, no added gain since the last increase. Those criteria are what let hours go up or down without a fight. And when a child is not benefiting after a fair trial, providers should change something in the care plan rather than extend it: fewer hours, a different model or therapeutic style, or an evaluation for something medical that might be getting in the way.

Payers own the cadence, the data, and the dollars. Reassess high-hour starts at 90 days, not six months, against the measures agreed up front. Let authorizations move in both directions, both up and down, which does not happen today. No caps. Use claims data to identify age- and severity-adjusted dosing patterns: providers whose hours barely vary across their caseload, a child whose hours have not moved in two years, or a family that got 20 hours from one assessor and 40 from the next. Finally, pay in a way that doesn’t punish stepping down, through case rates, performance-based payments, and network status tied to whether a provider's hours actually move.

The one thing neither side can do alone is agree, before the first hour, on what response will mean for this child and how it will be measured. Without that, a 90-day reassessment is a 90-day argument. With it, the hours finally have something to answer to.

Autism care has spent twenty years setting the number and then defending it, first on the way up and now on the way down. The people who wrote 40 into the guidelines and the people writing 20 into budget caps have more in common than either would admit: both decided the number before meeting the child. The children are three and four years old, and what each of them needs is a dose found with patience and then defended, however many hours it turns out to be. The useful question for the states writing budgets this year isn’t 40 hours or 20 hours. It’s whether the number can move.