Seventy-four studies (2,193 participants) were included: 68 published journal articles and 6 unpublished theses. Nine of the included studies were treatment-control, 34 were ABA and 31 were crossover studies.
The mean ES for the various interventions was 0.48, or 0.42 when participants classified as 'mentally retarded' or with Tourette's syndrome were excluded from the analysis.
The effects were greater for students than adult age groups, with an average ES of 0.40 for kindergarten-aged children, 0.37 for elementary school students, 0.57 for middle school students and 0.25 for adults. For the few studies that reported socioeconomic background, the ES was 0.34 for low-socioeconomic status (SES) students, 0.35 for middle-SES students and 0.15 for upper-SES students.
The effects were greater when physicians made the diagnosis (0.46) than when psychologists (0.28) or teachers (0.30) made the diagnosis. The ESs for the different methods of diagnosis were 0.45 for the DSM, 0.38 for the DSM-IV in conjunction with an unspecified battery of tests, 0.35 for an unspecified battery of tests, 0.23 for the Conners test and 0.51 for the Diagnostic Instrument of Childhood and Adolescence.
There was a significant relationship between the length of treatment and the ES (correlation, r=0.44). The effects were similar whether the treatment was managed by a physician (0.41), a psychologist (0.35), a teacher (0.47) or by parents (0.39), although the effects were lower when managed by a physician and psychologist together (0.24). The effects were greater when managed in the home (0.47) rather than in clinics (0.36) or school (0.33).
The effects were greater for pharmacological interventions (0.45) followed by school-based psychological/educational interventions (0.39), non-school-based psychological interventions (0.39), parent training interventions (0.31) and, finally, multimodal interventions (0.28). Within the pharmacological interventions, antidepressants had the largest ES (0.85), followed by stimulants and antidepressants (0.43), stimulants (0.35) and antipsychotic drugs (0.12). The effects of antidepressants were greatest on behavioural outcomes (1.58), although there were also negative effects on physical outcomes (-0.44). Within the school-based psychological/educational interventions, behavioural training had the largest ES (0.50), followed by cognitive/self-regulation (0.49), educational interventions (0.29) and other school-based psychological/educational interventions (0.25). Within the non-school-based psychological interventions, cognitive/self-regulation had the largest ES (0.58), followed by biofeedback (0.50), social-skills training (0.31), behavioural training (0.29) and other non-school-based psychological interventions (0.18).
The effects were greatest for behavioural outcomes (0.56), followed by social (0.38), cognitive (0.28) and personal/emotional (0.22) outcomes, with a very small negative effect on physical outcomes (-0.03). Within the behavioural outcomes, the effects were highest on hyperactivity (0.68), followed by better behaviour (0.66), impulsiveness (0.45) and attention (0.32). All the physical effects were associated with pharmacological treatments, primarily stimulants, while the negative effects related to side-effects such as a reduction in fine motor speed and skills, weight loss, sleep patterns, and an increase in nausea and shakiness.
The interaction of treatment and outcomes indicated that pharmacological and multimodal treatments had the greatest effects on behavioural outcomes; multimodal treatments had the greatest effects on social outcomes; school-based treatments had a slightly better effect on cognitive outcomes than the other interventions; and the interventions were equally effective on personal/emotional outcomes. The interactions of treatment with subcategories of outcomes were also presented in the report.
The effects were greatest if teachers made the assessment (0.60), followed by self-assessment by the student (0.42), parents (0.39), published scales administered by an independent trained person (0.33) and, finally, physicians or counsellors (0.31). The interactions of who made the assessment with separate outcomes and subcategories of outcomes were also presented in the report.
There was no correlation between the ES and the number of participants in a study, the year of publication, the affiliation of the author, where the article was published, or the major thrust of the article, other than one article with a combined neurological and behavioural thrust, which had a lower ES.
The effects were greater for ABA design studies and crossover studies than treatment-control studies. A between-group homogeneity statistic indicated no statistically significant difference between the three means. The authors stated that the reason for the lower treatment-control effect may relate to the nature of the outcomes measured: one of the treatment-control studies used an electroencephalogram to measure neuropsychological functioning and the ES was -0.69; the other three treatment-control studies had an average ES of 0.35.