Multiple cause of death approach to analyze mortality patterns

Luisa Frova, Instituto Nazionale di Statistica (ISTAT)
Marilena Pappagallo, Instituto Nazionale di Statistica (ISTAT)
Antonio Salvatore, Università di Roma "La Sapienza"
Viviana Egidi, Università di Roma "La Sapienza"

Background: The mortality analysis normally refers to the underlying cause of death but often death can be caused by a variety of factors and not from only one cause. This paper illustrates analytical uses of multiple-cause-of-death data. Data: Multiple cause of death data of 325,229 elderly people (70 years and over) who died in Italy in 2001 were considered in this paper. Certificates who were been automatically codified were considered in this study. Methods: The average number of causes per record, cause-specific frequencies and cause-specific ratio of underlying cause (U.C.) of death to multiple cause of death (M.C.) were calculated. A cluster analysis was carried out to explore the existence of some particular morbid profile. Multivariate odds ratios were computed to evaluate the association between morbidity patterns and U.C.. Main Results: 1,063,994 medical terms were codified representing 3.3 pathologies on average per death record (including the U.C.). The average number of M.C. in a death certificate varied between 3.2 and 3.7. The “degree of agreement” shows a good agreement for ischaemic heart diseases (65%), neoplasms (58%) and cerebrovascular diseases (55%). However U.C. data captured only a fraction of deaths identified by M.C. data for other causes. The cluster analysis led to the identification of 8 groups that presented particular profiles regarding both the causes of death (underlying and associated) and the demographic characteristics. Finally, three different logistic regression models were adapted to calculate the odds ratios of the M.C. and the respective U.C. identified into the clusters 2, 3 and 4 (the U.C. of these clusters were cerebrovascular diseases, ischemic heart diseases and other circulatory diseases). Conclusions: Multiple cause of death data can be seen as a major new dimension to mortality statistics. It can represent a very useful instrument to study the interaction of diseases leading to death also if much still need to be done to increase the quality of death certificate completion.