All the death certificates for deaths in 1977 where haemolytic disease of the newborn (HDN) was the principal, an antecedent, or a contributory cause were obtained from the Office of Population Censuses and Surveys (OPCS). The hospital notes of all 54 of the live-born cases and all of the 101 stillbirths were also obtained. The cause of the death indicated by the notes was compared with the cause and coding on the death certificate. In about a quarter of the cases death was not due to haemolytic disease of any type. The commonest errors arose because the International Classification of Diseases (8th edition) stipulates that hydrops without mention of cause should be coded as HDN and because stillbirths to rhesus-negative mothers tend to be attributed to rhesus HDN automatically. Though deaths from HDN may be overestimated in this way, they are also underestimated because rhesus disease, although mentioned on the certificate, is not selected as the underlying cause, which it may be. These cases were found through multiple coding of all the contributory causes of death, which OPCS performs on a 25% sample of all death certificates for research purposes. These two sources of inaccuracy tend to cancel each other out, but statistics from death certificates give a misleading picture of the efficacy of anti-D prophylaxis because anti-D can never prevent cases which are not in fact due to rhesus HDN. Most of the mothers studied had become immunised before anti-D became available, but in those who could have been treated 75% had not received prophylaxis. As this was a sample of deaths, however, it would not be accurate to extrapolate this high figure to the population at risk. Nevertheless, the organisation of prophylaxis is clearly deficient and should be remedied before providing antenatal anti-D to supplement postnatal treatment.