Youth Psychiatric Medication Statistics (2026): Every Number Sourced

This page is a reference sheet. Journalists and bloggers writing about kids and psychiatric medication keep re-deriving the same numbers from press releases, and the press releases routinely drop the two things that matter: the data years and the study design. So here's the compilation I wish I could hand every reporter who calls. One number up front to set the frame: in national all-payer dispensing data from 2006 through 2023, stimulant prescribing to children ages 3–17 peaked in 2011 at 5.9% and sat at 5.3% in 2023 (ASPE Data Point, HHS, April 2025; IQVIA).

Every statistic below carries its data years and its primary source inline. If a number isn't here, I couldn't verify it, and you probably shouldn't print it either. For the narrative version written for parents, see Which Kids Are Overmedicated? What Prescribing Data Show.

Prescribing trends by class: only one class is growing

All rows below measure dispensing (at least one filled prescription), which will run higher than ongoing treatment and says nothing about pills actually taken.

ADHD: diagnosis has outrun treatment, and the gap is widening

In our 2018 study of 2008 national dispensing data, we found stimulant treatment in school-age children (ages 6–12) running at 4.6% against an ADHD prevalence near 8.6% (Sultan et al., J Child Adolesc Psychopharmacol 2018). While public debate assumed excess, those figures described a treatment gap even then. The newer, nationally representative survey data say the gap has grown.

Depression: treatment reach improved, and four in ten teens still get nothing

The nationally representative NSDUH survey was redesigned in 2022, so read the 2021-to-2024 change as a direction of travel; the redesign limits precise year-over-year comparison.

Antipsychotics: the concerns center on sequencing and monitoring

In our 2019 study of 187,563 commercially insured youths newly diagnosed with ADHD, we found 2.6% (95% CI 2.5–2.7) filled an antipsychotic within one year (Sultan et al., JAMA Network Open 2019; MarketScan claims, 2010–2015). Notably, the journal published a correction: the accurate figure is 2.6%, superseding the originally published 2.3%. Commercial claims only, so don't generalize to Medicaid, where antipsychotic use runs higher. For the full breakdown, see the antipsychotic overuse page.

Although some off-label use is evidence-supported (risperidone, for example, has trial evidence for severe aggression after adequate stimulant trials), the sequencing figures above raise concerns regarding how these medications are being started.

Polypharmacy nearly doubled, and preschool prescriptions run long

Treatment effects: medicated periods look better, and the design labels matter

The observational rows compare a person with themselves on and off medication, a design that removes stable confounding (genetics, family) while leaving time-varying confounding in place. None of them is a randomized trial, and two of the three are Swedish; the two Swedish studies are broadly consistent with each other, with the accidental-injury estimate the design-dependent exception. These findings belong to their authors; in our 2025 JAMA Psychiatry editorial we argued about what the Li results mean for U.S. practice, but the numbers are Li's.

Workforce and training: the system prescribing these drugs is understaffed and undertrained

Little is known about stopping these medications once they're started; the final two rows quantify that gap.

How to read these numbers: three distinctions that keep getting flattened

Dispensing vs use vs survey report. The ASPE/IQVIA and Chua numbers count filled prescriptions (at least one fill in the period), which overstates ongoing treatment and says nothing about pills actually taken. NHANES and NSDUH numbers are self- or parent-report surveys, with their own biases. NSCH counts parent-reported diagnoses. These three measurement types will not match each other, and quoting them interchangeably produces fake trends.

Data years lag publication years. A 2025 paper can describe 2010–2015 claims. The Sultan 2019 antipsychotic figures are 2010–2015 commercial claims; the NHANES polypharmacy series stops in March 2020; the MEPS workforce numbers are 2008–2011. When you quote a figure, give its data years.

Observational evidence cannot establish causal relationships, and a rising curve has several explanations. While the within-individual IRRs above are strong observational evidence, they fall short of trial results. Further, a rising prescription curve tracks rising diagnosed illness and a deliberate access push as much as it tracks prescriber behavior; the same datasets show large groups of diagnosed kids getting nothing at all. Accordingly, both patterns warrant vigilance: the sequencing and monitoring problems documented above, and the larger group of diagnosed young people receiving no treatment.

How to cite this page

About the author. I'm a child and adolescent psychiatrist and Assistant Professor of Clinical Psychiatry at Columbia University Irving Medical Center, where I direct the Sultan Lab for Mental Health Informatics. I wrote the 2018 national study of psychotropic prescribing to young people (J Child Adolesc Psychopharmacol, 2008 data), the 2019 JAMA Network Open study of antipsychotic prescribing in 187,563 youths with ADHD, and a 2025 JAMA Psychiatry editorial on ADHD medication and real-world outcomes. I treat children, adolescents, and adults in New York City.

Related reading on this site

• Which Kids Are Overmedicated? What Prescribing Data Show: the narrative version for parents, with the questions to ask at your child's next appointment
• Is ADHD Overdiagnosed and Overmedicated?: what the diagnosis and treatment data support
• Pediatric Antipsychotic Overuse: the full technical breakdown of the 2019 JAMA Network Open cohort
• National Patterns of Pediatric Psychotropic Prescribing: the 2018 study of stimulants, antidepressants, and antipsychotics