Surveillance of development in the incidence of heart disease, particularly acute myocardial infarction (AMI) requires a valid system of registration of all cases of the disease. During the period 1972-1981, all admissions with the diagnosis AMI to hospitals in Greater Copenhagen and Arhus were registered in the Heart Register. From 1978, admissions to all Danish somatic hospitals were registered in the National Patient Register. In order to assess the validity of the two registers as regards registration of AMI, the two registers were compared for the years 1979 and 1980. This investigation includes computerized comparison of the registered admissions with AMI and also re-assessment of the diagnosis based on the summaries for a random sample of the admissions. Comparison of the two registers revealed considerable disagreements. Among the individuals who had been admitted with AMI according to the National Patient Register, only approximately 2/3 were registered with this diagnosis in the Heart Register, while almost all persons registered in the Heart Register could be found again in the National Patient Register with the same diagnosis. Technical problems in coding and electronic data analysis in the National Patient Register only appear to explain a lesser fraction of the disagreements as the diagnoses which were summed up in the upper part of the summaries showed reasonably good agreement with the diagnoses in the National Patient Register. Re-assessment of the diagnoses revealed that approximately 90% of the admissions with AMI could be verified, without modification, by reviewing the text of the summaries.(ABSTRACT TRUNCATED AT 250 WORDS)