We studied the validity of the Finnish hospital discharge register data on coronary heart disease (CHD) for the purposes of epidemiologic studies and health services research. The Finnish nationwide hospital discharge register (HDR) was linked with the FINMONICA acute myocardial infarction (AMI) register for the years 1983-1990. The frequency of errors in the HDR was assessed separately. Between 8% and 13% of hospitalized AMI events registered in the AMI Register were not found in the HDR with an ICD code for CHD. Problems with the register linkage and the use of some ICD code other than one of the codes for CHD explained these missing events. The frequency of errors in the personal identification number was about 5% in the early 1980s. After 1986 errors were found only occasionally. The diagnosis recorded in the HDR was the same as that in the discharge sheet in about 95% of hospitalizations. The positive predictive value of the ICD code 410 (AMI), compared with the FINMONICA definite+possible AMI category, was very high and stable, about 90% in all areas and all hospitals, but the sensitivity varied from 50% at local hospitals to 80% at central hospitals. In summary, data on CHD obtained from the Finnish hospital discharge register give, on average, a correct picture on changes in the occurrence of AMI in Finland and can, with necessary caution, be used in epidemiological studies and health services research. However, the classification of individual cases is not standardized in the HDR, but varies over time, between geographical areas and the levels of care. Therefore, these data should not be used without confirmation in studies where correct classification of individual outcomes is of crucial importance, such as follow-up studies and case-control studies.