As part of a large clinical and epidemiological study of inflammatory bowel disease (IBD), we examined the hospital case records of a sample (255) of the 1257 patients aged 0-20 years, recorded in Scottish Hospital In-Patient Statistics (SHIPS) for 1968-1983 as Crohn's disease (CD) or ulcerative colitis (UC). The coded diagnosis was incorrect at the time of coding in 47 instances (18.4% of the sample), for the following reasons: clinical diagnosis wrong (24 cases); IBD correctly diagnosed but wrongly coded as CD for UC or UC for CD (seven cases): various other clerical or computer errors (15 cases). One case that did not meet standard diagnostic criteria for CD at the time of coding was shown to be correctly labelled when confirmatory pathological information became available. In view of the influence of such statistics on the social and clinical management of chronic illnesses such as IBD, and in view of the impact of analysis of these data on the provision of services and allocation of resources, it is suggested that some indication of the degree of confidence in the clinical diagnosis (possible, probable, definite) should be incorporated in the information submitted for coding and should be reflected in the derived data.