Abstract
All incidents of deliberate self harm among patients registered with 16 randomly selected general practices in the inner city area of Southwark, south London, were monitored over 18 months. We recorded details of initial psychosocial management and of any subsequent treatment given by hospital accident and emergency services. We compared the characteristics of patients who repeated self harm with those of patients who did not. We then used factors that were related to repetition of self harm to build a multivariate model. Finally, we used Cox's regression to calculate adjusted hazard ratios for repetition of self harm among groups of patients receiving different forms of treatment.