Mortality statistics are an important source of information concerning variations in time and place, identification of risk factors and the evaluation of treatment programs. In this study, a new death certificate was completed "blind" on the basis of hospital records from the last episode of care, across a random sample of 1,376 cases. The results showed that the overlap between the official register's underlying cause of death and that of a panel was 72 per cent at the three-digit level. The official underlying cause of death from cerebrovascular diseases (CVD) was 72 cases in this sample, while 93 were deemed to have CVD by a panel. Additionally, of the 1,233 cases originally reported as non-CVD, the panel deemed non-CVD to be the true underlying cause in 1,176 cases. The paper concludes that CVD was most often correctly reported as the underlying cause of death in the investigated ages up to 75 years but plain differences were found between specialities and in different hospital size.