ABA’s Value-Based Care Experiment Is Showing Results—and Raising Questions

Paying for Progress, Not Hours

DALY CITY, CALIFORNIA – For more than three years, Kyo Autism Therapy has been paid by Magellan Healthcare in a way almost no other ABA provider is: a fixed monthly payment for each member, with quality judged over time by the Vineland Adaptive Behavior Scales rather than by the number of therapy hours delivered. In the years since it began, about 80% of patients have shown progress on Vineland measures even as their care hours were titrated down as they improved.

That sentence contains the whole argument for value-based care in applied behavior analysis. Under the fee-for-service model that funds nearly all of the field, revenue is a function of hours billed, which means reducing a child’s hours reduces the provider’s income. Under capitation, the incentive reverses. Progress that allows a child to need less therapy is no longer a financial penalty.

“We didn’t see a reduction in outcomes, but we saw a reduction in hours, and that’s a powerful example of how those payment incentives were able to shape that behavior,” Deepa Shah, chief growth officer at Kyo, said at an industry conference in March.

The Vocabulary, Defined

Value-based care is not one payment method but a family of them, and the distinctions matter because they distribute risk differently.

Fee-for-service. The status quo. Providers bill in time units, most commonly CPT 97153 for technician-delivered treatment, and are paid per unit delivered. Volume drives revenue; nothing in the payment mechanism registers whether the child improved.

Capitation. The payer sends a fixed amount per member per month, and the provider decides how to deploy it clinically. This is the Kyo arrangement. The provider absorbs the financial risk if a child needs more care than the payment assumes, and keeps the benefit when a child needs less.

Outcomes-based and enhanced-rate contracts. Payment stays closer to fee-for-service, but rates rise or fall against agreed metrics. Cortica, which works with roughly 35 payers, runs several whole-child value-based contracts, including a tiered arrangement in which enhanced rates ride on utilization, speed to care after diagnosis, and patient satisfaction.

What unites them is that someone must define and measure a good outcome, then agree to be paid on it. That requirement, more than any ideological objection, is why value-based contracts in ABA remain rare.

Why the Model Is Arriving Now

The timing is not academic. State Medicaid programs cut ABA rates across at least eight states in the past year, federal auditors have questioned the documentation behind hundreds of millions in payments, and commercial rates have been largely flat for years while labor costs climbed. Payers have begun telling providers, in substance, that higher rates will follow proof of value rather than precede it.

The fee-for-service structure also produces distortions its critics find hard to defend. It rewards maximizing a patient’s lifetime billable hours, gives payers no way to tell an excellent provider from a poor one, and pushes operators toward scale mainly to gain leverage in fee negotiations, which disadvantages small independents. It also treats a heterogeneous population as uniform.

“It’s not one size fits all,” said Themis Gomes, chief executive of Omaha-based Behaven Kids. “A child may come in with a higher acuity level and sometimes needs a two-to-one ratio because their behaviors are actually dangerous to their family, to themselves and their community. How is that considered?” Gomes has proposed billing codes that split by diagnosable severity as a bridge toward more sophisticated arrangements.

“I think value-based care can drive quality, but it’s not guaranteed if it’s not designed well. Incentives shape behavior.” – Deepa Shah, chief growth officer, Kyo Autism Therapy (2026)

What the Kyo Model Proves, and What It Does Not

The California partnership is the field’s most-cited evidence, and it has a track record: Magellan reported in 2023, roughly a year into the collaboration, that 74% of its members receiving ABA showed improved outcomes; by the March 2026 panel, Kyo put the figure at about 80% progressing on Vineland measures. The two companies also set out from the start, in late 2022, to develop outcome standards other providers and payers could use.

The limits deserve equal billing. This is one provider, one payer, one state, and outcomes reported by the parties to the contract rather than by an independent evaluator. The measurement instrument is contested: peer-reviewed work has found that the Vineland-3 has weak psychometric properties for detecting change over time, and that patient-centered measures showed little correlation with adaptive-behavior score changes at 12 and 24 months of ABA. A payment model resting on a scale that may not reliably capture change is a model resting on an open question.

Shah’s own framing is the caution the field should take from the case. Value-based care can drive quality, she said, but only if the metrics are well designed rather than narrowly defined or overly prescriptive, which can distort care. Badly chosen metrics do not fail neutrally; they redirect clinical behavior toward whatever is being counted.

What Comes Next

The infrastructure is being assembled even as the science is argued. The Council of Autism Service Providers released version 3.0 of its practice guidelines with standardized outcome metrics and a new emphasis on long-term tracking, and now runs a clinical-quality accreditation program. Data platforms built to make providers contract-ready have moved into the industry’s dominant software stack. Other models are being pitched in parallel, including outcomes-based contracting from telehealth-native providers.

For now, the honest summary is narrow: one multi-year arrangement, in one state, suggests that paying for outcomes can hold quality steady while reducing hours. Whether that result survives contact with a heterogeneous national population, unresolved risk adjustment, and measurement tools still under dispute is the question the next several years will answer.

AT A GLANCE

SOURCES & REFERENCES