Background This review al-no document and analyse aspects of death cer
tification that are relevant to public health. Methods A literature re
view on death certification primarily used the computerized Index Medi
cus (1981 to mid-1995), and concentrated on completing death certifica
tes, accuracy, standards, education and procedural requirements. Furth
er sentinel publications pre-dating this were identified from the main
literature base. Results The uses of mortality data, historical and p
rocedural context for recording death, the philosophy of Underlying Ca
use of Death and its relationship to 'the truth', the extent and impac
t of 'inaccuracy', the certificate and the certifier, and possible way
s forward are discussed. It is argued that the question 'How inaccurat
e are cause of death data?' is harder to answer than the literature su
ggests, Deriving a useful estimate is difficult because of inter-study
differences in (1) definition, measurement (how and by whom?) and pra
ctical importance of error, and standards used; (2) focus (e.g. death
certificate or mortality data), observing everyday practice or simulat
ion exercises, diagnostic and/or semantic issues. Conclusion The tradi
tional perspective on improving the quality of death certification has
not worked. There is a need for reorientated thinking rather than jus
t urging more education. Evidence-based educational interventions are
needed. The flaws in the theoretical framework of cause of death and t
he routine nature of death certification are unavoidable, but require
consideration. Certifiers need practical feedback mechanisms, integral
to continuing quality assurance tall levels and fostering an understa
nding of the construction of mortality data. Continued development sho
uld be a core public health medicine role.