The best predictors of attention deficit hyperactivity disorder (ADHD) were examined using Conners 3 ADHD Index (Conners 3 AI) (teacher and parent reports), Child Behaviour Checklist for ages 6-18 (CBCL/6-18) and Youth Self Report for ages 11-18 (YSR/11-18) in a sample of 350 schoolchildren from the Epidemiological Project on Neurodevelopmental Disorders (EPINED) (n= 2,818). The diagnosis was made on the basis of the DSM-5 criteria and the three presentations of ADHD were categorised as non-diagnosis (n= 175), subclinical (n= 56) or clinical (n= 118). Discriminant analyses showed that the CBCL attention problems scale was the best predictor, correctly classifying almost 80% of cases (78.4% unadjusted model; 79.2% model adjusted for IQ and socioeconomic level). The slow cognitive time scale was the best predictor of inattention presentation (68.7% unadjusted; 71.0% adjusted) and the DSM scale of attention problems was the best predictor of hyperactive-impulsive (71.1% unadjusted; 78.0% adjusted) and the combined (68% unadjusted; 71.0% adjusted) presentation. Predictors did not differ between models for two (non-diagnostic and clinical) or three diagnostic categories (non-diagnostic, subclinical and clinical).