Accuracy of death certificates of cardiovascular disease in a community intervention in Sweden.

Aim: The aim was to investigate the possibility to evaluate the mortality pattern in a community intervention programme against cardiovascular disease by official death certificates.

Methods: For all deceased in the intervention area (Norsjö), the accuracy of the official death certificates were compared with matched controls in the rest of Västerbotten. The official causes of death were compared with new certificates, based on the last clinical record, issued by three of the authors, and coded by one of the authors, all four accordingly blinded.

Results: The degree of agreement between the official underlying causes of death in "cardiovascular disease" (CVD) and the re-evaluated certificates was not found to differ between Norsjö and the rest of Västerbotten. The agreement was 87% and 88% at chapter level, respectively, but only 55% and 55% at 4-digit level, respectively. The reclassification resulted in a 1% decrease of "cardiovascular deaths" in both Norsjö and the rest of Västerbotten.

Conclusions: The disagreements in the reclassification of cause of death were equal but large in both directions. The official death certificates should be used with caution to evaluate CVD in small community intervention programmes, and restricted to the chapter level and total populations.