Overdiagnosis ADHD

Overview

• Topic: Systematic review and meta-analysis of ADHD medication use in children/adolescents across continents, focusing on undertreatment in diagnosed individuals and overtreatment/misuse in those without ADHD.

• Protocol: Pre-registered in PROSPERO (CRD42018085233).

• Data sources: Broad electronic databases and grey literature; 25,676 abstracts screened; 36 studies retained from 10 countries; 104,305 participants; 18 studies met main analysis criteria for diagnosed ADHD/HKD (DSM/ICD or validated scales) and 14 studies for non-ADHD controls.

• Main question: What are the pooled rates of ADHD pharmacological treatment in (a) school-age children/adolescents with ADHD and (b) those without ADHD, across continents?

• Key findings (main analyses):

  • Diagnosed ADHD/HKD group: pooled pharmacological treatment rate = $0.191$ (95% CI: $0.115 ext{--}0.299)$.

  • Non-ADHD group: pooled pharmacological treatment rate = $0.009$ (95% CI: $0.005 ext{--}0.017)$.

• Core interpretation: Evidence for both undertreatment of youths with ADHD and overtreatment/misuse in those without ADHD, underscoring the need for education and policy discussions on ADHD medications.

• Extrapolations and country comparisons: Authors provide country-level projections (US, Australia, Netherlands) using demographic data and prevalence estimates to illustrate the balance between undertreatment and overtreatment.

Objectives and scope

• Primary aim: Estimate pooled rates of ADHD medication use in two groups across continents:

  • Children/adolescents with ADHD or HKD (diagnosed) undergoing pharmacological treatment.

  • Children/adolescents without ADHD (und diagnos) using ADHD medications.

• Secondary aim: Explore potential imbalance between undertreatment and overtreatment, and discuss implications for public health policies and education.

• Hypothesis: Global pooled estimates would suggest undertreatment of ADHD overall.

Methods: study design and criteria

• Study design eligible: Population-based, cross-sectional, or longitudinal studies with data since inception up to Apr 30, 2020; no language or age restrictions.

• Population definitions:

  • ADHD/HKD diagnosed participants (DSM II, DSM-III, DSM-IV(-TR), DSM-5; ICD-9/10; validated diagnostic instruments or clinical interviews).

  • Non-affected individuals (controls).

• Included pharmacological interventions: stimulants (methylphenidate, dexmethylphenidate, amphetamines) and atomoxetine; also accepted studies reporting ADHD medication use by participants/caregivers.

• Excluded: clinical/insurance-based samples, unclear diagnostic process, or using ADHD pharmacological treatment as a proxy for diagnosis without confirmation; lack of clear data on diagnosis or assessment method.

• Data extraction: Study metadata, participant demographics, diagnostic method, comorbidities, medications/ formulations, dose, treatment duration, and assessment method for medication use.

• Outcome measures:

  • Primary: Rates of ADHD medication use in diagnosed children/adolescents.

  • Primary: Rates of ADHD medication use in non-ADHD children/adolescents.

• Study quality: Modified Newcastle–Ottawa Scale (NOS).

• Data synthesis: Meta-analysis of proportions using the R meta package (Metaprop), with a random-effects model (DerSimonian and Laird). Transformation: Logit (found to be most appropriate). CI estimation via Clopper–Pearson with continuity correction. Heterogeneity: Cochran’s Q and I^2. Publication bias: Funnel plots and Egger’s test. Sensitivity: Jackknife method. Meta-regression: Covariates entered individually (year, quality, continent, country, diagnostic method, medication type, medication assessment method, study design).

• Protocol deviations: Reported in supplementary materials (not itemized here).

Data processing and statistical details

• Transformation for proportions: Tried multiple ordinary transformations; final main analyses used the Logit transformation for proportions. $p = ext{proportion}.$ Back-transformed for reporting and CIs.

• Model: Random-effects due to expected heterogeneity across studies. $ext{Model: } heta<em>i = \beta</em>0 + u<em>i, ext{ with } u</em>i hicksim N(0, au^2).$

• Heterogeneity metrics:

  • Cochran’s Q test;

  • I^2 statistic (percentage of total variation due to heterogeneity).

• Publication bias checks: Visual funnel plots; Egger’s regression test (p-values reported: diagnosed group p = 0.29; non-diagnosed p = 0.10).

• Sensitivity: Jackknife analyses to assess influence of individual studies on prevalence and heterogeneity.

• Covariate/meta-regression findings (main analyses, DSM/ICD ADHD diagnosis in children/adolescents):

  • Quality of the study (p = 0.0349; HAF = 23.54%).

  • Country (p = 0.0391; HAF = 14.97%).

  • Study design (p = 0.0230; HAF = 33.69%).

  • These covariates significantly affected between-study heterogeneity.

• Diagnostic method effects: When combining samples regardless of how ADHD was diagnosed, the method of diagnosis was a strong covariate; sensitivity analyses showed inflated treatment rates when diagnosis relied on a single caregiver-response item and non-diagnosed group rates were deflated under the same condition.

• Adult and preschool data: Separate analyses reported in supplementary materials (Tables S5–S6; Figures S1–S2; S3–S4; S5–S6; S17–S20). Main analyses focused on children/adolescents with DSM/ICD-confirmed ADHD.

Results: study selection and characteristics

• Screening and inclusion: 25,676 records screened; 36 studies retained; 18 studies contributed to the main analysis for diagnosed ADHD; 14 studies for non-ADHD controls.

• Geographic distribution: Most studies from Western developed countries; USA predominates; 10 countries in total.

• Age range and sampling: Predominantly school-age children/adolescents; some adult samples (n = 6,620) and preschool samples (n = 20,174) reported separately in supplements.

• Study designs: Mostly cross-sectional; several retrospective/administrative data sets; some prospective cohorts.

• Diagnostic criteria: Majority used DSM/ICD criteria or validated scales; a subset used caregiver-reported diagnosis.

• Main analytic results:

  • Diagnosed ADHD/HKD group (18 studies; n = 3,311): pooled ADHD pharmacological treatment rate = $0.191$ (95% CI: $0.115 ext{--}0.299)$.

  • Non-ADHD group (14 studies; n = 29,559): pooled ADHD pharmacological treatment rate = $0.009$ (95% CI: $0.005 ext{--}0.017)$.

• Quality: Mean NOS score around $3.7$ for main-analytic studies; key weakness was comparability across studies.

• Publication bias: Not evident for either diagnosed or non-diagnosed groups (Egger’s test non-significant).

Results: detailed findings and interpretations

• Meta-analysis interpretation:

  • Among youths with DSM/ICD ADHD, about 19.1% receive pharmacological treatment in cross-national samples with substantial heterogeneity.

  • Among youths without ADHD, about 0.9% receive ADHD medication, indicating overtreatment/misuse in some cases.

• Heterogeneity: High across analyses, as anticipated in the protocol; explanations include differences in diagnostic methods, treatment practices by country, and study design.

• Sensitivity and robustness: Jackknife analyses indicated no single study drove the heterogeneity; meta-regressions identified plausible sources of heterogeneity (quality, country, design).

• Diagnostic method sensitivity: Using DSM/ICD criteria vs. caregiver-reported diagnosis changed estimates notably; “one-question” diagnoses inflated diagnosed-group treatment rates and depressed non-diagnosed rates in some scenarios (see supplemental figures S17–S20).

• Subgroup extrapolations (contextual): Adult and preschool analyses showed different patterns, but the main conclusions centered on children/adolescents.

Extrapolation to three countries (US, Australia, Netherlands)

• Purpose: To illustrate the balance between undertreatment and overtreatment using population data and ADHD prevalence estimates.

• Common conservative assumption used: At least 70% of properly diagnosed ADHD youths might benefit from a trial of ADHD medication. This threshold is used to project treatment gaps.

• United States (US) projections:

  • Population: Youth aged 5–19 years = $N = 62{,}378{,}000$; ADHD prevalence (DSM) ≈ $P_{ADHD} = 0.095$.

  • Number with ADHD: $N<em>{ADHD} = N imes P</em>{ADHD} = 62{,}378{,}000 imes 0.095 \,.$

  • Diagnosed treated with medication: $N_{treated} = 1.97 ext{ million}.$

  • ADHD youths who might benefit but are not treated:

  • Assumed treated proportion among diagnosed = 0.333 (33.3%), consistent with US data; thus number treated = $N_{ADHD} imes 0.333$.

  • Potentially treatable fraction = 0.70; thus potentially treatable but not treated = $0.70 imes N<em>{ADHD} - N</em>{treated}.$

  • This yields approximately $2.17 ext{ million}$ youths with ADHD who could benefit but are untreated.

  • Without-diagnosis medicated youths:

  • Population without ADHD: $N<em>{noADHD} = N - N</em>{ADHD}.$

  • Prevalence of ADHD medication use in those without ADHD: $p_{noADHD ext{ med}} = 0.012$ (1.2%).

  • Number medicated without formal ADHD diagnosis: $N<em>{noADHD, med} = N</em>{noADHD} imes p_{noADHD ext{ med}}.$

  • For US numbers: approximately $677{,}425$ youths without a formal ADHD diagnosis might be using ADHD medication.

• Australia projections:

  • ADHD-treated population: about 45,000 youths (12.8%).

  • Diagnosed but untreated: about 202,000 youths eligible but not pharmacologically treated.

  • Without-diagnosis medicated: about 18,000 youths (0.4%).

  • Ratio of undertreatment to overtreatment: about 11-fold difference (underapplied vs. overused).

• Netherlands projections:

  • ADHD-treated: about 17,000 youths (22.2%).

  • Diagnosed but untreated: about 37,000 yet eligible for pharmacological treatment.

  • Without-diagnosis medicated: about 22,000 youths (0.8%).

  • The under/over-treatment gap is described as less than 2-fold in the Netherlands.

• Formulas used (summary):

  • US: $N<em>{ADHD} = N imes P</em>{ADHD}$; $N<em>{treated} = N</em>{ADHD} imes f<em>{treated} ext{ with } f</em>{treated} ext{ ≈ } 0.333$; $N<em>{untreated ext{ (potential)}} = 0.70 imes N</em>{ADHD} - N<em>{treated}$; $N</em>{noADHD, med} = (N - N<em>{ADHD}) imes p</em>{noADHD ext{ med}}$.

  • Similar structure applied to Australia and the Netherlands with country-specific numbers and the same 70% threshold where stated.

• Key takeaway from extrapolations: Across countries, undertreatment among diagnosed youths and overtreatment/misuse among non-diagnosed youths co-exist, with the US showing the largest gap in absolute numbers, followed by Australia and the Netherlands in relative terms.

Discussion: interpretation and implications

• Main conclusion: ADHD medications appear undertreated in youths with ADHD and overtreated/misused in those without ADHD across continents where data exist.

• Guideline context: International guidelines differ in emphasis:

  • Some American guidelines: pharmacotherapy as first-line treatment for ADHD in many cases.

  • Some European guidelines: non-pharmacological interventions prioritized, or pharmacotherapy reserved for more severe cases or nonresponse to behavioral interventions.

• Evidence on non-pharmacological interventions: meta-analytic evidence supports short-term efficacy for pharmacotherapy; however, non-pharmacological approaches (e.g., parent training) can yield improvements without medication in a subset of patients (estimated 18–23% response in some trials).

• Severity and treatment need: Epidemiological data suggest 70–76% of ADHD cases meet criteria for moderate/severe ADHD, implying a substantial portion could benefit from medication, supporting the conservative 70% threshold used for extrapolations.

• Confounding factors and heterogeneity:

  • Higher treatment rates in the US may reflect practice patterns, health system differences, and access.

  • Within-country differences and sociodemographic factors influence treatment likelihood.

  • Differences in diagnostic methods (DSM/ICD vs. caregiver-reported or single-question diagnoses) substantially affect prevalence estimates and treatment rates.

• Clinical and public health implications:

  • Need for balanced education for clinicians and families about ADHD and treatment options.

  • Policies should address both ensuring access to indicated pharmacotherapy for diagnosed youths and reducing misuse/overuse in those without ADHD.

  • Clinicians should carefully document diagnostic criteria and ensure treatment aligns with guidelines and individual needs.

Limitations and caution in interpretation

• Representativeness: Not a world-representative sample; most data from developed countries, with the US dominating.

• Within-country heterogeneity: Substantial differences in prescribing practices across regions and populations within countries.

• Diagnostic methods: Some studies relied on caregiver reports or single-item questions; temporality (current vs lifetime diagnosis) was not always clear.

• Medication assessment: Several studies relied on self-report or caregiver report without confirmatory prescription records; this may bias estimates.

• Exclusion of under-5s: Pharmacological treatment in preschoolers is controversial; main analyses excluded under-5s.

• Focus scope: This meta-analysis covers pharmacological treatment only; it does not address potential under-treatment involving non-pharmacological interventions.

• Publication bias: Although statistical tests did not indicate significant bias, funnel plots showed scattered patterns; limited numbers of studies in some analyses.

• Heterogeneity: Very high across analyses; meta-regression identified several contributing factors (study quality, country, design), but residual heterogeneity remains.

Conclusions

• The evidence indicates two coexisting issues worldwide:

  • Undertreatment of school-age youths with ADHD with respect to pharmacological therapy.

  • Overtreatment/misuse of ADHD medications in youths without a formal ADHD diagnosis.

• These findings underscore the need for evidence-based medical and parental education and public health policies that optimize ADHD diagnosis accuracy and treatment decisions.

• The authors emphasize the importance of harmonizing guidelines and improving education about ADHD medications to maximize benefits and minimize harms.

Key numerical references (summary)

• Overall diagnosed ADHD treatment rate: p_{ ext{diag}} = 0.191 ext{ (95% CI: }0.115 ext{--}0.299)

• Overall non-ADHD treatment rate: p_{ ext{non-diagnosed}} = 0.009 ext{ (95% CI: }0.005 ext{--}0.017)

• Exemplary extrapolation assumption: 70% of properly diagnosed ADHD youths might benefit from medication (threshold used for population projections).

• US population projections (illustrative numbers):

  • Youth 5–19 years: $N = 62{,}378{,}000$

  • ADHD prevalence: $P_{ADHD} = 0.095$

  • ADHD youths: $N<em>{ADHD} = N imes P</em>{ADHD}$

  • Proportion treated among ADHD: $f_{treated} ext{ (≈ }0.333)$

  • Treated ADHD youths: $N<em>{treated} = N</em>{ADHD} imes f_{treated}$

  • Untreated but eligible: $N<em>{untreated} = (0.70 imes N</em>{ADHD}) - N_{treated}$

  • Youths without ADHD using meds: $N<em>{noADHD, med} = (N - N</em>{ADHD}) imes 0.012$

  • Example results: approx. 1.97 million treated with ADHD meds; ≈ 2.17 million ADHD youths who might benefit but are not treated; ≈ 677,425 without ADHD on ADHD meds.

• Country-specific differences in undertreatment vs. overtreatment expressed as relative ratios (US largest gap; Netherlands ≤ 2-fold difference; Australia ≈ 11-fold difference in under vs. over-treatment).

References (selected topics cited in notes)

• ADHD prevalence and outcomes; pharmacotherapy evidence base; guidelines (CADDRA, NICE, etc.); global trends in ADHD medication use; methodological references (DerSimonian-Laird, Clopper-Pearson, Egger’s test, NOS).

• Notable supporting data include: population estimates, prevalence in DSM-defined ADHD, and cross-national prescribing patterns; cross-reference to supplementary materials for tables and figures not reproduced here.

Limitations and caution in interpretation

Representativeness: The meta-analysis, despite its broad search strategy, is not globally representative. The vast majority of the included data originated from developed countries, with a significant overrepresentation of studies from the United States. This geographic bias means that the findings may not be directly generalizable to lower-income countries or regions with different healthcare infrastructures, cultural contexts, and diagnostic/treatment practices.

Within-country heterogeneity: Even within individual countries, substantial heterogeneity in prescribing practices exists across different regions, urban vs. rural settings, and specific populations (e.g., based on socioeconomic status, insurance coverage). The pooled national estimates might mask these important sub-national variations, limiting the granularity of interpretation.

Diagnostic methods: A notable limitation is the reliance on varying diagnostic methods across studies. Some studies based their ADHD diagnosis on less robust measures, such as caregiver reports or simple single-item screening questions, rather than comprehensive clinical evaluations or validated diagnostic instruments. This variability can introduce bias and reduce the comparability of diagnoses, potentially affecting the accuracy of both ADHD prevalence and medication use rates. Furthermore, the temporality of diagnosis (e.g., current diagnosis vs. lifetime diagnosis) was not always uniformly clear, which could impact the interpretation of who is currently experiencing symptoms and might benefit from treatment.

Medication assessment: Several studies assessed medication use through self-report by participants or caregivers. While this method can capture real-world use, it lacks the direct confirmation offered by prescription records or administrative databases. Self-report may be subject to recall bias or social desirability bias, potentially leading to an over- or underestimation of medication prevalence.

Exclusion of under-5s: The main analyses specifically excluded pharmacological treatment data for preschool-aged children (under 5 years). This was a deliberate decision given that pharmacological treatment in this age group is often considered more controversial and is generally approached with greater caution, with non-pharmacological interventions typically prioritized. However, this exclusion means the review’s primary findings do not directly apply to this important developmental stage.

Focus scope: This meta-analysis was exclusively focused on estimating and interpreting the rates of pharmacological treatment for ADHD. It does not address the prevalence, efficacy, or potential undertreatment of non-pharmacological interventions (e.g., behavioral therapy, parent training) which are also crucial components of comprehensive ADHD management. Therefore, it provides only a partial picture of overall ADHD treatment patterns.

Publication bias: Although formal statistical tests (Egger’s test) did not indicate significant publication bias, visual inspection of funnel plots did show some scattered patterns. This suggests that while there wasn't a statistically significant pattern of missing