Statistical methods for designing and assessing the effectiveness of community-based interventions with small numbers.

Community-based interventions are, by their nature, tailored to the particular needs of the members and stakeholders of the community. In the area of safety promotion, the multifaceted aspects of violence and unintentional injuries necessitate complex intertwining of strategies and approaches. The evaluation of the effectiveness of any 'intervention' programme, either as a whole or of its discrete components, is thus challenging. In addition, standard experimental research designs involving controlled comparisons and the use of randomisation are often not practical. The evaluation is complicated further when the numbers of events in the community that constitute outcomes to be measured, before and after the interventions, are small. This manuscript reviews some of the statistical considerations that impact the evaluation of the effectiveness of such intervention approaches, and proposes a meta-regression methodology to address this complex issue. The application of the proposed methodology to the 'Safe Community' Model is provided as an example.