In order to measure the coverage and accuracy of the Hospital In-patient Enquiry Scheme data in a large acute hospital in Dublin, 793 patient charts were selected randomly from all the discharges in 1990. The capture rate was 56%. Whether a diagnosis was captured for the Hospital In-patient Enquiry Scheme depended partly upon geographical factors within the hospital and partly upon the nature of the diagnosis. "Non-capture" was not random and was particularly likely to occur where the case was complex, e.g. oncology, HIV infection. The coding accuracy of primary diagnoses was 59% and the completeness of recording of secondary diagnoses was 56%. Both were significantly associated with the presence in the hospital chart of discharge summaries and with the clarity and source of written diagnoses. The factors that affected accuracy were not the same as those affecting the capture rate.