We welcome Dr Walshe's review of the study of adverse events.¹ Dr Walshe points out the usefulness of studies of adverse events, but also the need to be cautious when they are used as measures of health care quality. We think that this article underlines the major methodological issue in studies of adverse events—namely, the lack of a standardised definition. Until this point is resolved, the practical applications of the concept of adverse events will be limited. Definitions of this term are frequently grouped into two categories—restrictive and broader. Restrictive descriptions of adverse events are limited to those cases which provoke an evident negative consequence for the patient and therefore can only be identified after the event when harm has already been done. The broader definition includes circumstances or events that are potentially harmful to the patient, regardless of the negative effects they may or may not produce as they are identified earlier. In our experience² the second definition is preferable when the objective is to avoid adverse events as it implies a more realistic outlook. It is well known, for example, that only a small proportion of smokers suffer from chronic obstructive pulmonary disease (COPD) yet we ignore a priori pointers to who will develop the condition. Consequently, all are urged to stop smoking. Likewise, if we wish to prevent negative effects associated with patient falls in hospital we should study all falls, not just those that have resulted in fractures. In this sense the term “adverse event” is confusing and it is preferable to use “potentially adverse event”. The key, then, is to establish which circumstances are potentially adverse and therefore require being monitored and, insofar as possible, controlled. Some, such as nosocomial infection, are already clearly identified and their analysis, surveillance, and control constitute one of the great achievements of hospital epidemiology.³ In other cases it is more difficult to decide whether an event is or is not potentially adverse, and to what extent its surveillance and control are advantageous. This would be the case in management problems—for example, admission to a service other than the one responsible for the patient's care due to a shortage of beds.

In our opinion Dr Walshe's review highlights the need for unification of criteria to reach a consensual definition of “potentially adverse event”. Such a definition should include a list of circumstances classified by the degree of evidence of harm attributed to each, and its capacity for being monitored and controlled. We feel that opening a debate in this direction should be a top priority in the study of adverse events. When all researchers speak the same language, we will be more likely to understand each other and to make progress.