Quantifying cause-related mortality by weighting multiple causes of death.

Introduction Good understanding of mortality data is essential for developing and evaluating health policies. The causes of any death are usually reported on parts I and II of a death certificate, in accordance with the international form presented in the International classification of diseases and related health problems, tenth revision (ICD-10), (1) and data are usually collected in a standardized and consistent way. (2) In part I, the physician describes the causal sequence of events that led directly to the death. In part II, the physician can report any other significant morbid condition but only if that condition may have contributed to the death. Generally, cause-of-death statistics are derived from the so-called underlying cause of death in a process hereafter referred to as the classic method. (3) The World Health Organization (WHO) defines the underlying cause of death as "the disease or injury which initiated the train of morbid events leading directly to death or the circumstances of the accident or violence which produced the fatal injury". (1) However, deaths are often caused by more than one disease. Moreover, in a world characterized by an ageing population and decreasing mortality and fertility, death due to infectious disease is progressively being replaced by death due to chronic and degenerative diseases. (4-6) As a result, the classic method discards potentially useful information about the contribution of other conditions to a death. Today, analysis of mortality data increasingly uses a multiple-cause-of-death approach, (3,4,7-12) which is defined as any statistical treatment that simultaneously considers more than one of the causes of death reported on a death certificate. In particular, such approaches have been used to recalculate mortality attributable to specific conditions. In practice, when cause-specific mortality is re-evaluated to take into account multiple causes of death, the number of mentions of a specific cause is usually considered--here the statistical unit is the cause of death rather than the death itself, which raises serious questions about interpretation. For example, studies examining the influence of several diseases on mortality may count a single death two or more times if two or more causes of death are mentioned on the certificate. The resulting apparent increase in mortality could yield an artificial increase in statistical power and possibly result in misleading inferences. An additional problem is that each cause of death mentioned on a certificate is given an equal weight, even though its individual contribution may not have been equally important--the relative importance of each cause of death is not considered. In this study, we investigated an experimental approach that assigns a weight to each cause of death listed on a death certificate by analysing French death certificate data using three multiple-cause-of-death weighting methods. This approach conceptualizes death as the outcome of a mixture of conditions, as we described elsewhere. (13) Consequently, each death contributes only a fraction, rather than a unit, when calculating standardized mortality rates for each cause of death --the fraction depends on the weight assigned. The approach accepts that multiple factors may contribute to a death but also reflects the relative contribution of each cause of death. (13) Use of a multiple-cause-of-death weighting approach could help us identify conditions whose contribution to mortality is underestimated by the classic method. Methods We examined data on all deaths reported in France during 2010. We had access to information on all the causes of death declared on death certificates, including the underlying cause of death, as coded using the ICD-10 by CepiDc-Inserm--the epidemiology centre on medical causes of death of the French National Institute for Health and Medical Research. We used the 2012 version of the European shortlist for causes of death to analyse mortality by cause-of-death category, (14) though the list was modified slightly for the analysis. …