[The epidemiology of medical errors: a few issues in methodology].

Epidemiological research is needed to determine how many medical errors are made in hospitals. An estimation of error risk requires knowledge of the number of errors made and the total number of patients treated. But how do we define the term 'error'? The outcome of a medical intervention alone is insufficient; the entire treatment process has to be taken into account. For the estimation of error, two different approaches are available: retrospective chart review or prospective ethnographic observation. Both methods should account for processes and outcomes. Both methods have their limitations and can produce measurement errors. In particular, the low reproducibility of the determination of error is worrisome. To achieve safer health care, emphasis should be placed on control of processes, rather than outcomes.