Medicaid ACOs and Pediatric Behavioral Health Care
Key Takeaways
- Medicaid ACO implementation was associated with higher reported unmet mental health needs among Medicaid- or CHIP-insured children with behavioral health conditions.
- No association was observed for having a personal physician or nurse, mental health treatment, effective care coordination, or family-centered care.
- Subgroup, event-study, and sensitivity analyses were broadly consistent with the overall pattern
Using 2016 to 2023 data from the National Survey of Children’s Health and a staggered difference-in-differences approach, this repeated cross-sectional analysis compared states that implemented Medicaid ACOs between 2017 and 2022 with states that had no implementation during the study period. The sample included Medicaid- or CHIP-insured children aged 3 to 17 years whose caregivers reported at least 1 current behavioral health condition, including anxiety, depression, ADHD, Tourette syndrome, or behavioral problems. The weighted sample represented 29,885,590 children, the unweighted sample included 15,783, and 9.1% lived in implementation states. Analyses were conducted between January 2025 and February 2026, and the five outcomes covered access, unmet need, and care experience.
In adjusted analyses, implementation was not associated with having a personal physician or nurse, with an estimate of -2.48 percentage points (95% CI, -10.10 to 5.14). It also was not associated with treatment from a mental health professional, at -7.27 percentage points (95% CI, -15.53 to 0.98). Estimates for care coordination and family-centered care were similarly unchanged at -5.17 percentage points (95% CI, -15.57 to 5.22) and -0.46 percentage points (95% CI, -8.29 to 7.37). Falsification tests showed no changes in the probability of having a caregiver-reported behavioral health condition or in reported condition severity, reinforcing the largely null pattern across the other outcomes.
Subgroup analyses compared children with a single behavioral health condition with those who had multiple conditions. For unmet mental health needs, estimates were 3.25 percentage points for single-condition children and 5.96 percentage points for multiple-condition children, with no significant difference between groups by Wald test (P=.299). Treatment from a mental health professional was associated with a significant -9.45 percentage-point estimate among single-condition children and a -5.55 percentage-point estimate among multiple-condition children, but the between-group Wald test was nonsignificant (P=.310). Event-study estimates had preimplementation coefficients centered around zero without a consistent directional pattern. A two-stage DID model yielded 4.84 percentage points (95% CI, 1.11 to 8.57), and excluding Idaho and New Jersey did not materially change the finding.