SummaryIt is notoriously difficult to obtain evidence from clinical randomised controlled trials for safety innovations in healthcare. We have developed a research design using simulation for the evaluation of safety initiatives in anaesthesia. We used a standard and a modified scenario in a human-patient simulator, involving a potentially life-threatening problem requiring prompt attention -either a cardiac arrest or a failure in oxygen supply. The modified scenarios involved distractions such as loud music, a demanding and uncooperative surgeon, telephone calls and frequent questions from a medical student. Twenty anaesthetics were administered by 10 anaesthetists. A mean (SD) of 11.3 (2.8) errors per anaesthetic were identified in the oxygen failure scenarios, compared with 8.0 (3.4) in the cardiac arrest scenarios (ANOVA: p = 0.04). The difference between the combined standard scenarios and the combined modified scenarios was not significant. The mean rate of errors overall was 9.7 per simulation, with a pooled SD of 4.46, so in future studies 21 subjects would provide 80% statistical power to show a reduction in error rate of 30% from baseline with p £ 0.05. Our research design will facilitate the evaluation of safety initiatives in anaesthesia.