Identification of incident myocardial infarction (MI) cases in a defined population using hospital discharge data and mortality data in combination has been suggested. This method of case identification was compared to that of use of MI community registers set up in accordance with principles adopted in a World Health Organization collaborative programme. The comparison comprised data for four Swedish cities over a number of years. On average 81% of incident hospital-treated cases below 65 years of age identified through MI community registers were found by the retrospective use of the method based on hospital discharge data and mortality data. Of hospital-treated cases identified by the latter method, 83% were also found by the MI community registers. For cases fulfilling the diagnostic criteria employed by the MI community registers this proportion would be higher, probably 87%-92%. Several reasons for cases being missed by either method were suggested by the results. According to the findings of this study, the case identification of the method based on hospital discharge data and mortality data seems to be somewhat less efficient compared to use of MI community registers. This may be of importance in descriptive epidemiological studies, but is of less significance in analytical studies. The relative efficiency of the former method could be improved by a more reliable system for the recording of hospital discharges. If supplemented by a validation procedure, it could yield sufficiently accurate data for many epidemiological applications at a fairly low cost.