Cross-Informant Agreement and Discrepancies in ADHD Assessment During Adolescence: A Systematic Review and...

Assessment of attention-deficit/hyperactivity disorder (ADHD) requires information across settings, but reports from adolescents, caregivers, teachers, and clinicians are often only partly concordant. This review synthesized cross-informant agreement and discrepancies in ADHD assessment among adolescents aged 10-19 years.

MEDLINE, CINAHL, ERIC and Academic Search Complete were searched through EBSCO, together with PubMed and Scopus, on 19 July 2026. Eligible primary quantitative studies compared at least two independent informants assessing the same adolescent ADHD domain. Methodological quality was appraised with an adapted 11-item QAREL framework. Pearson correlations were pooled after Fisher z transformation using restricted maximum-likelihood random-effects models and Hartung-Knapp-Sidik-Jonkman confidence intervals.

The searches identified 1447 records; 30 reports (282 outcomes) were included. Caregiver-adolescent pooled r was 0.33 for inattention (k = 7, 95% CI 0.19-0.46; I² = 88.1%; 95% prediction interval -0.07 to 0.64) and 0.32 for hyperactivity/impulsivity (k = 7, 95% CI 0.12-0.50; I² = 94.8%; prediction interval -0.28 to 0.74). Despite the positive pooled point estimates, both prediction intervals were wide and crossed zero. Caregiver-teacher estimates were r = 0.29 for inattention (k = 4, 95% CI -0.02 to 0.55; I² = 94.1%) and r = 0.27 for hyperactivity/impulsivity (k = 4, 95% CI -0.03 to 0.52; I² = 83.0%); both confidence intervals crossed zero. Diagnostic agreement was generally low and sensitive to impairment and consensus rules.

Informant reports were positively related on average, but the magnitude of agreement was highly heterogeneous and imprecisely estimated. Wide intervals, including caregiver-adolescent prediction intervals crossing zero and caregiver-teacher confidence intervals crossing zero, indicate that little or no agreement remains plausible in some comparable settings. Reports should therefore not be treated as clinically interchangeable, and pooled estimates should not be interpreted as universal levels of agreement.