The ante-mortem (in vivo) and post-mortem diagnoses of coronary heart disease (CHD) were compared in necropsies at a university hospital for 1965, 1975, and 1985. The secular trends showed gradually rising proportions in both true positive and false negative ante-mortem diagnoses. Both types of change in diagnostic error would progressively lower the counted numbers of CHD in official vital statistics. With suitable statistical corrections, the occurrence rates of CHD in Connecticut became much higher and showed less dramatic trends in secular decline. The results indicate that major improvements in accuracy are needed before vital statistics data are accepted at face value and analyzed for biologic explanations of the changing numbers.