Deciding whether interventions for antisocial behaviour work: principles of outcome assessment, and practice in a multicentre trial.

Factors influencing the reported effectiveness of interventions for antisocial behaviour are reviewed, with illustrative findings from a multicentre controlled trial of parenting groups. The merits and validity of different types of measuring instrument are discussed and it is concluded that it is desirable to have more than one informant (parent, teacher, observer) and at least one instrument with objective criteria, for example a semi-structured interview or direct observation. In the illustrative trial, the effect size was larger with semi-structured interviews than questionnaires, for conduct problems compared to hyperactivity symptoms, and for trial completers compared to the total sample it was intended to treat. The discrepancy between the encouraging results of trials conducted in university centres and the disappointing findings of those carried out in 'real life' clinics is discussed. The trial discussed here was conducted in health service clinics by regular staff but still reported good effect sizes. Measuring likely mediators and moderators so mechanisms of change can be investigated is discussed. In the trial, change was mediated through acquisition of parenting skills but not through maternal depression or criticism directed at the child, even though these were reduced by the intervention. This suggests that programmes that impart skills are more likely to be effective than those that offer counselling only.