Predicting Outcome in Anaesthesia: Understanding Statistical Methods

METHODS: anaesthesia, outcomeA patient to undergo surgery may ask "What is my chance of dying under anaesthesia?"or "What is my chance of having a heart attack?".As anaesthetists, we may ask "What will be the impact of a chosen anaesthetic technique on this particular patient?".Surprisingly, after 150 years of anaesthesia we are still unable to answer such questions accurately.Large surveys conducted in the past give us an estimate of outcome for groups of patients 15.The results from such studies have been remarkably consistent.For example, if 20,000 ASA-l patients were anaesthetised, one patient would be expected to die as a result of anaesthesia.But which patient?Such precision is lacking for the individual patient: their outcome remains an all-or-none phenomenon, either survival or death.Each patient's risk is affected by many factors, both known and unknown.The difficulties are made worse when attempting to predict the impact of any particular intervention, or variation in technique, on that individual patient.The aim of this paper is to describe some of the background to outcome studies, the types of statistical analyses used in assessing outcome and the techniques used in assessing their predictive ability.It is not meant to be a comprehensive review of outcome studies in anaesthesia.We have included examples from the recent literature, not to openly criticise their methods, but to highlight the difficulties associated with their design and interpretation.This should allow the interested anaesthetist, unfamiliar with the current statistical procedures employed, to better evaluate future studies.Does the anaesthetist or choice of anaesthetist make any difference to outcome?Most studies have failed *Dip.R.