AIM: This study investigated the diagnostic accuracy of the Strengths and Difficulties Questionnaire (SDQ) in a population of children born extremely preterm (<26wks gestation). METHOD: Parents and teachers of 219 extremely preterm children (118 females, 101 males; age 11y) were asked to complete the SDQ to screen for psychological problems. Multi-informant ratings were aggregated using two methods: combined (parent or teacher rated the child with problems) and pervasive (parent and teacher rated the child with problems). Psychiatric diagnoses were assigned using the Development and Well-Being Assessment. RESULTS: Pervasive ratings had the greatest diagnostic accuracy for emotional disorders (89%), conduct disorders (94%), attention-deficit-hyperactivity disorder (ADHD; 90%), and autism spectrum disorders (ASDs; 94%), but were associated with low sensitivity (=50%). For clinical use, combined ratings were best for detecting emotional disorders (sensitivity 77%, specificity 75%), conduct disorders (83%, 88%), and ADHD (85%, 72%). Parent ratings were best for ASDs (93%, 66%). Teacher ratings significantly improved prediction over parent ratings alone for conduct disorders (chi(2) =9.3, p=0.002) and ADHD (chi(2) =24.1, p<0.001) only. INTERPRETATION: Multi-informant data are preferable for assessing most mental health outcomes using the SDQ. As an outcome measure, pervasive ratings have the best predictive accuracy. For screening, combined ratings are best for detecting ADHD and emotional and conduct disorders. For ASDs, parent ratings were best.