Long-term growth in child mental health expenditures, 2004-2023 Authors: Joshua Breslau, Dar Alon, Alyssa Burnett, Jonathan Cantor, Ryan K McBain, Ateev Mehrotra, Bradley D Stein, Fang Zhang, Hao Yu Published: 2026 ABSTRACT AbstractIntroductionInvestment in mental health treatment for children has grown substantially, but whether this reflects more treatment recipients, higher spending per user, or both has not been examined. To estimate trends in mental health expenditures, probability of use and expenditures per user for US children aged 2-17 from 2004 to 2023.MethodsTwo-part models were estimated to predict inflation adjusted health expenditure data from the 2004-2023 Medical Expenditure Panel Survey (unweighted N = 153 622) to assess overall trends and variation by age-group and insurance coverage.ResultsTotal annual mental health expenditures increased from $10.9B to $37.9B, rising from 4.4% to 8.1% of total health expenditure per child (mean individual-level ratio among children with any health care expenditure). The probability of mental health service use more than doubled (6.3% to 13.0%, +0.35% points/year), while expenditure per user grew from $2816 to $5178 (+3.3%/year).DiscussionRelative to the 2004 baseline those ages 2-5 had the greatest growth in both service use and expenditure per user. Expenditure per user grew substantially faster for privately-insured than for Medicaid-covered children.ConclusionsChild mental health treatment expenditures have grown dramatically over the past 2 decades, with particularly large increases for those ages 2 to 5. Child mental health expenditures tripled in real terms from 2004 to 2023, driven by both a doubling of the percentage of children receiving care and an 84% increase in expenditures per user. Child mental health expenditures tripled in real terms from 2004 to 2023, driven by both a doubling of the percentage of children receiving care and an 84% increase in expenditures per user. Growth in service use relative to baseline was greatest among children aged 2-5, who also showed the fastest growth in expenditures per user. Growth in service use relative to baseline was greatest among children aged 2-5, who also showed the fastest growth in expenditures per user. The probability of mental health service use grew at similar rates for privately-insured and Medicaid-covered children, but expenditures per user grew faster among the privately insured. The probability of mental health service use grew at similar rates for privately-insured and Medicaid-covered children, but expenditures per user grew faster among the privately insured. Introduction Mental disorders commonly begin during childhood and adolescence1 and have adverse impacts on health and functioning across the life course. Evidence of large and persistent treatment gaps has prompted sustained calls to increase investment in child mental healthcare, with the goal of reducing the impacts of mental health conditions during critical developmental periods and thereby limiting lifelong morbidity and impairment.2,3 Policymakers have responded through multiple channels, including expanded insurance coverage for mental health conditions and integration of mental healthcare into primary care pediatrics.4,5 There is evidence that expenditures on mental healthcare have increased substantially. A recent analysis found that US expenditures on mental health and substance use disorder treatment grew from $40.9 billion to $139.6 billion between 2000 and 2021, with growth driven primarily by increases in the number of people receiving treatment rather than by rising costs per case.6 Among children, studies have also found substantial increases in expenditures in recent years,7,8 though whether these increases are due to changes in prevalence of treatment or expenditures per treated child has not been examined. Understanding what has driven this growth, and whether it has been distributed in ways likely to reduce the burden of mental illness, requires first establishing the basic empirical facts: how much expenditure has grown, whether growth reflects more users or higher spending per user, and whether trends have been uniform across subgroups. Using national survey data from 2004 to 2023, we examine trends in total annual mental healthcare expenditures for children and separate expenditure growth into changes in the percentage of children receiving any mental health services and changes in spending per user. We also characterize variation in these changes in treatment receipt and expenditures by age group and insurance coverage. Data and methods The Medical Expenditure Panel Survey (MEPS) is an annual cross-sectional household survey of a nationally representative sample of the US civilian noninstitutionalized population. Unlike administrative claims data, MEPS captures self-pay expenditures, which are particularly important for mental health services where out-of-pocket payments are common.9 Expenditures on treating medical conditions in MEPS include payments from all sources for hospital inpatient care, ambulatory care provided in offices and hospital outpatient departments, care provided in emergency departments, paid care provided in the patient's home, and the purchase of prescribed medications. Expenditures were converted to 2023 dollars using the Consumer Price Index for medical care. All analyses accounted for the MEPS complex survey design and sampling weights. We focused on children aged 2-17 from the 2004-2023 samples (unweighted N = 153 622). Following methods developed by AHRQ,10 we identified mental health utilization and expenditure using the ICD-9 based Clinical Classifications Software (CCS) codes before October 2015 and the ICD-10 based Clinical Classifications Software Refined (CCSR) codes since October 2015 (see detailed CCS, ICD-9, CCSR, and ICD-10 codes in the online Supplement I). We used the MEPS Prescribed Medicines Files to identify medications prescribed for mental health conditions, including stimulants, anxiolytics, antidepressants, antipsychotics, and mood stabilizers. In addition to descriptive analyses of trends, we estimated a series of 2-part models. The first part used logistic regression to estimate the probability of any mental health service use each year. The second part used a generalized linear model with gamma family and log link to estimate annual expenditures among children with any mental health service use. Both parts included calendar year as a continuous predictor. We used marginal standardization to generate predicted probabilities of service use and predicted expenditures per user for 2004 and 2023, along with the estimated annual change for each. For any treatment use, annual change is expressed as an absolute change in percentage points per year. For expenditure per user, annual change is expressed as the proportional change relative to the prior year, analogous to a compound annual growth rate. We conducted 2 sensitivity analyses by (1) dropping 2020, a year with massive disruptions in healthcare during the COVID-19 public health emergency, and (2) applying Winsorization to the cost data at the 99th percentile. To examine heterogeneity in trends by age group and insurance coverage, we specified models for each outcome with interaction terms between year and subgroup indicator. Age groups were defined as 2-5, 6-12, and 13-17 years. Insurance coverage was classified as private insurance, Medicaid, or other/uninsured, based on coverage status at the time of the survey. Statistical significance of subgroup differences was assessed via the interaction term. A sensitivity analysis was conducted using models that adjusted simultaneously for age group, insurance coverage, race/ethnicity, and sex. All analyses were conducted in Stata version 18.0 (StataCorp). This study was determined not to constitute human subjects research by the Harvard Pilgrim Health Care Institute IRB. Results Figure 1 shows trends in total annual child mental health expenditures and mental health expenditures as a proportion of total annual health expenditures for children from 2004 to 2023. Total annual mental health expenditures grew about $1.4 billion per year from $10.9B to $37.9B over the study period, rising from 4.4% to 8.1% of per-child health expenditures (mean individual-level ratio among children with any health care expenditure). Child mental health expenditures declined in 2020, increased sharply in 2021 and 2022, and then returned to the pre-pandemic trajectory by 2023; the 2020 decline in mental health expenditures was commensurate with declines in overall child health spending, as the share of total spending attributable to mental health continued to rise without interruption. Child mental health expenditures, overall, and as a proportion of total child healthcare expenditures, 2004–2023. Service use and expenditures per user Increases in both the percentage of children receiving mental health services and average expenditures among those who received services contributed to the trend in total child mental health expenditures (Table 1). On average, the probability of any mental health service use increased by 0.35% points per year (95% CI: 0.29-0.42), rising from 6.3% in 2004 to 13.0% in 2023. Among children with any service use, expenditures per user grew by 3.26% per year (95% CI: 1.84%-4.68%), rising from $2816 in 2004 to $5178 in 2023. Trend results were similar in sensitivity analyses excluding 2020 and applying Winsorization (Supplement II). Table 1.Trends in child mental health service use and expenditures per user, MEPS, 2004-2023.Part 1: probability of any mental health service use (%)aPart 2: expenditure per user (2023$)bSubgroupPredicted2004 (%)Predicted2023 (%)Annualchange(pp/yr)b95% CI P Predicted2004 ($)Predicted2023 ($)Annualchange(%/yr)c95% CI P Panel A. Overall Full sample6.3%13.0%0.350.29-0.42<0.001$2816$51783.26%1.84%-4.68%<0.001 Panel B. By Age Group Ages 2-51.3%6.6%0.280.22-0.35<0.001$2687$76495.66%2.64%-8.68%<0.001 Ages 6-12 [reference] 7.8% 13.4% 0.29 0.22-0.38 <0.001 $2453 $5452 4.29% 2.02%-6.57% <0.001Ages 13-178.4%17.4%0.470.37-0.58<0.001$3388$43031.27%−0.64%-3.18%0.194p for interaction (age × year)————<0.001————0.027 Panel C. By Insurance Type Private insurance6.2%13.1%0.360.29-0.44<0.001$2429$57154.61%2.70%-6.52%<0.001 Medicaid [reference] 7.2% 13.4% 0.33 0.22-0.43 <0.001 $3757 $4340 0.76% −0.97%-2.50% 0.390Other/uninsured3.9%4.2%0.02−0.14%-0.17%0.844$1730$31873.27%−2.77%-9.30%0.288 P for interaction (insurance × year)c————0.276————0.004Predicted values derived from 2-part models applied to mental health expenditure data from the MEPS, 2004-2023. All models incorporate MEPS complex survey weights and account for the survey's stratified cluster design. Reference categories shown in gray italics. Interaction P-values test whether the annual rate of change differs significantly from the reference group. All models include year, subgroup indicator, and their interaction only. Results were similar in sensitivity analyses adjusting for age group, insurance coverage, race/ethnicity, and sex simultaneously. aPart 1: absolute change in the probability of any mental health service use per calendar year, in percentage points. This is an additive quantity. bPart 2: exp(β)−1, the proportional change in expenditure per user per year, expressed as a percentage. This is a multiplicative (compound) quantity analogous to an annual growth rate. cDue to very small sample size for the Other/uninsured category, the interaction tests differences in trend between the privately-insured and Medicaid-covered. Trends in child mental health service use and expenditures per user, MEPS, 2004-2023. Predicted values derived from 2-part models applied to mental health expenditure data from the MEPS, 2004-2023. All models incorporate MEPS complex survey weights and account for the survey's stratified cluster design. Reference categories shown in gray italics. Interaction P-values test whether the annual rate of change differs significantly from the reference group. All models include year, subgroup indicator, and their interaction only. Results were similar in sensitivity analyses adjusting for age group, insurance coverage, race/ethnicity, and sex simultaneously. aPart 1: absolute change in the probability of any mental health service use per calendar year, in percentage points. This is an additive quantity. bPart 2: exp(β)−1, the proportional change in expenditure per user per year, expressed as a percentage. This is a multiplicative (compound) quantity analogous to an annual growth rate. cDue to very small sample size for the Other/uninsured category, the interaction tests differences in trend between the privately-insured and Medicaid-covered. Variation by age group The average annual increase in the probability of mental health service use was larger for adolescents aged 13-17 (0.47% points per year, 95% CI: 0.37-0.58) than for children aged 6-12 (0.29% points per year, 95% CI: 0.22-0.38) or 2-5 (0.28% points per year, 95% CI: 0.22-0.35). However, given the substantially lower starting point for the youngest age group in 2004, 1.27% compared with 7.78% and 8.40% for the older groups, proportional growth was greatest among 2-5 year olds. Predicted prevalence of service use in 2023 was 5.2 times the 2004 level for ages 2-5 (1.27% to 6.63%), compared with 1.7 times for ages 6-12 (7.78% to 13.4%) and 2.1 times for ages 13-17 (8.40% to 17.4%). The pattern for expenditures per user differed from the pattern for service use. Annual growth in expenditures per user was statistically significant for ages 2-5 (5.66% per year, 95% CI: 2.64%-8.68%) and ages 6-12 (4.29% per year, 95% CI: 2.02%-6.57%) but not for ages 13-17 (1.27% per year, 95% CI: −0.64%-3.18%, P = 0.194). Variation by insurance coverage The annual increase in the probability of mental health service use did not differ significantly between children covered by private insurance and those covered by Medicaid (interaction P-value = 0.276). Both groups of children showed significant growth: 0.36% points per year (95% CI: 0.29-0.44) for the privately insured and 0.33% points per year (95% CI: 0.22-0.43) for Medicaid-covered children. Annual growth in expenditures per user did differ significantly between insurance groups (interaction P-value = 0.004). Among privately insured children, expenditures per user grew by 4.61% per year (95% CI: 2.70%-6.52%), more than doubling from $2429 in 2004 to $5715 in 2023. Among Medicaid-covered children, the annual change of 0.76% per year (95% CI: −0.97%-2.50%) was not statistically significant (P = 0.390), with expenditures per user rising from $3757 in 2004 to $4340 in 2023. Results were similar in magnitude and statistical significance in sensitivity analyses that adjusted simultaneously for age group, insurance coverage, race/ethnicity, and sex (Supplement II). Discussion From 2004 to 2023, expenditure on child mental health treatment grew more than 3-fold in inflation-adjusted terms. This growth reflected increases in both the percentage of children receiving any mental health services, which more than doubled from 6.3% to 13.0%, and expenditures per user, which grew by approximately 84%, from $2816 to $5178. While reducing the persistent treatment gap11 remains a priority, the magnitude and consistency of this 2-decade trend suggest that questions about the distribution, quality, and effectiveness of care deserve equal attention. The consistency and magnitude of this 2-decade trend is unlikely to reflect any single cause. While increases in the prevalence of depression and anxiety disorders12 and policies expanding access to mental healthcare, including the Patient Protection and Affordable Care Act13 and mental health parity legislation,5 have contributed, neither aligns well with the steady upward trend: the increase in prevalence began around 2013 and the major policy changes took effect no earlier than 2010, when the upward trend in expenditure was already well established. Longer-term social and cultural changes, particularly decreasing stigma related to mental health treatment,14,15 are likely more consistent drivers of this trend, along with the gradual integration of mental health services into pediatric primary care settings.16 The sharpest growth occurred among the youngest children, though the pattern differs depending on whether absolute or relative change is considered. Among adolescents aged 13-17, the absolute increase in the probability of mental health service use was largest (0.47% points per year), reflecting the fact that this group started from a relatively high base and continued to expand. Among children aged 2-5, absolute growth was more modest (0.28% points per year) but relative growth was by far the greatest: the probability of any service use grew more than 5-fold, from 1.3% to 6.6%. Moreover, expenditures per user for the youngest age group grew at 5.66% per year, significantly faster than older groups. These findings likely reflect a growing clinical focus on identification of psychiatric conditions, including autism, in early childhood, expanding availability of child psychiatry consultation through pediatric primary care,17 and developing systems for classifying mental disorders among very young children.18 The parallel growth in privately insured and Medicaid-covered children, with annual increases of 0.36% and 0.33% points respectively, is notable given the large differences between these groups; Medicaid-covered children who receive mental healthcare tend to have higher levels of social vulnerability and are more likely to have externalizing conditions such as ADHD, while privately insured children are more likely to have internalizing conditions, such as anxiety and depression.19,20 In contrast, growth in expenditures per user differed markedly by insurance type, growing 4.61% per year among privately insured children with small and nonsignificant growth among Medicaid-covered children. This difference could reflect growth concentrated among clinical conditions and treatment modalities more prevalent in privately insured children—such as anxiety and depression treated with psychotherapy—though differences in reimbursement rates, visit frequency, and provider type may also contribute. The divergence in per-user spending across insurance types warrants investigation in future work. This study has several limitations. Expenditure data do not permit assessment of undiagnosed conditions or quality of care. Expenditure estimates from the MEPS and other surveys differ from estimates from the National Health Expenditure Accounts, which are based on provider revenue data.21 The information collected by MEPS is subject to recall error. MEPS does not include children in institutional settings. Two decades of sustained growth have substantially expanded the child mental health system, making it an increasingly significant component of US health spending. The distribution, quality, and effectiveness of child mental health care deserve sustained research and policy attention commensurate with the scale and trajectory of child mental health spending. Supplementary Material