Effective methods are needed to identify and manage risks in healthcare to improve patient safety and hence the reputation of health care. Analysing safety related records and learning from ‘touch and go’ situations as well as accidents is one possible way of preventing hazardous conditions from occurring in healthcare, both public and private. The eventuality of an incident or an accident may markedly be reduced if the risks connected to it are efficiently diagnosed. This paper explores how a data mining technique can be applied to healthcare in order to discover harm causing factors in vast amounts of patient safety records.