Variability of American Society of Anesthesiologists' (ASA) physical status scores attributed to the same patient by multiple physicians has been reported in several studies. In these studies, the population was limited and diseases that induced disagreement were not analysed.To evaluate the reproducibility of ASA physical status assessment on a large population, as used in current practice before scheduled surgery.Multicentre, randomised, blinded cross-over observational study.During a 2-week period in nine institutions, ASA physical status and details of assessment performed routinely by anaesthesiologists for patients who underwent elective surgery were recorded. Records were blinded (including ASA physical status) by an independent statistical division and returned randomly to one of the nine centres for reassessment by accredited specialist anaesthesiologists.The level of agreement between the two measurements of the ASA physical status was calculated by using the weighted Kappa coefficient.During the study period, 1554 anaesthesia records were collected and 197 were excluded from analysis because of missing data. After the initial evaluation, the distribution of ASA physical status grades was as follows: ASA 1, 571; ASA 2, 591; ASA 3, 177; and ASA 4, 18. After the final evaluation, the distribution of ASA grades was as follows: ASA 1, 583; ASA 2, 520; ASA 3, 223; and ASA 4, 31. Two per cent of the patients had an underestimation of their physical status. The degree of agreement between the two measures evaluated by the weighted Kappa coefficient was 0.53 (0.49-0.56). No difference was observed between public and private institutions. Patients with co-existing diseases, obesity, allergy, sleep apnoea, obstructive lung disease, renal insufficiency and hypertension were least likely to have been graded correctly.The degree of agreement between two measures of the ASA physical status grade is moderate and influenced by staff characteristics and the complexity of diseases.