Causes of death

Causes of death

Magali Barbieri, Aline Désesquelles, France Meslé, Laurent Toulemon, Jacques Vallin

The differences in mortality trends observed between countries can only be understood by analysing the medical causes of death. This can be done by using existing databases, such as the World Health Organisation’s mortality database. This offers access to a large number of countries for general analyses. However, these can only be summary because the data are not comparable over time and space. For more detailed research, we need to develop international long-term time cause-of-death series using constant definitions common to all countries. This is being done here for France, the former Soviet republics and a growing number of other countries.

Usually, cause-specific mortality refers only to the underlying cause of death. Our understanding of mortality trends would undoubtedly be enhanced if secondary and contributing causes were also taken into account. However, there are numerous methodological problems with analysing multiple causes. We have started to unravel these and shall continue. An international comparison of cause-specific mortality rates would also make it possible to refine the conventional exercise of turning the lowest mortality risk into a hypothetical life expectancy reflecting the lowest mortality rates observed at a given time. Lastly, the exceptional mortality due to the heat wave in August 2003 led us to take an interest in crises of excess mortality observed in France.

Reconstruction of continuous cause-of-death series

In the late 1970s, INED embarked on a vast project to reconstruct detailed, consistent statistical cause-of-death series. We now have cause-of-death series reclassified according to the detailed list of the Ninth Revision of the International Classification of Diseases (ICD-9) for the years 1925 to 1999. This rich database is available on INED’s website (

The database is being used for a great deal of research on mortality trends in France and more broadly on the health transition. However, the shift to the tenth revision of the ICD in 2000, which introduced automatic coding, has created new discontinuities in the series. In order to overcome these, we needed to wait until we had enough years after the change. As the data for 2000 to 2005 are now available, it is urgent to reconstruct series of consistent data incorporating the most recent years. Unfortunately, the simultaneous change in the coding method used and in the classification of causes of death considerably complicates the issue. France Meslé and Jacques Vallin have undertaken an initial exploration of the discontinuities introduced by the changes in classification and coding.

To contribute to a better understanding of the fundaments of the health transition, it is also important to extend the series into the past, by going back further than 1925. In the case of France, this means working with statistics that are increasingly imprecise in terms of age, increasingly patchy in terms of geographical coverage (whole of France, then towns with a population of more than 10,000, then Paris only) and involving classifications that are increasingly divergent from current concepts. We shall nevertheless attempt this at least for the period from 1880 to1925.

Multiple causes of death

Most studies on causes of death, including those mentioned above are based, out of necessity, on the underlying cause of death only, to which statistics traditionally give priority. We know, however, that, although important, this information is an imperfect reflection of reality. Increasingly, especially as most deaths are concentrated at old ages, we will need to refer to the concept of multiple causes. For France, alongside the main cause, we have information on the immediate and contributing causes since 1979.

Aline Désesquelles and France Meslé have broken down the available data on deaths occurring after age 60. The French data have been compared with US and Italian data. That research will continue in several directions. The indicators developed (Cause-of-Death Association Indicator - CDAI - and Standardized Ratio of Multiple to Underlying cause -SRMU) need to be improved in several ways. So far our analyses have involved examining a specific year. The production of long series is worthwhile in itself. It could also help us to understand and correct the discontinuities introduced by the change from the ninth to the tenth revision of the ICD and automatic coding of causes of death.

Reconstruction work will start with deaths for which the medical certificate indicates Parkinson’s disease or senile dementia (including Alzheimer’s). Analysis of multiple causes - still little used by demographers - is highly promising. Given that the quality of the data is still in question, an international comparison seems valuable. We would therefore like to gather an international research group on this issue. The fact that this approach is starting to interest international institutions should make that easier.

Exceptional mortality: the 2003 heat wave

The exceptional excess mortality in France due to the heat wave in August 2003 highlighted the need to better understand the mechanisms of that type of health crisis (a sudden crisis concentrated in time and space that mainly affects elderly people).

One of the main focuses was to estimate excess mortality due to the heat wave in terms of the decline in life expectancy or years of life lost. The long-term effect of the heat wave on mortality depends on three factors: the total number of victims, the age structure of the victims (for the same number of victims, the years of life lost will increase in inverse proportion to the average age of the deceased) and the existence and size of a possible harvesting effect.

To measure these three factors, Magali Barbieri and Laurent Toulemon analysed data from municipal civil registries provided by INSEE, namely deaths by month, gender, year of age and département of residence for the period from 1997 to 2005. The main finding of the study was that the harvesting effect was small: of the 15,000 victims of the heat wave, a minority, representing 4,000 at most, would have died by the end of 2004 anyway, given their age, sex and state of health, independently of the heat wave in August 2003. The others could have lived another eight to eleven years on average, depending on their sex. The total number of years of life lost is therefore around 100,000. The heat wave in August 2003 was not the first heat wave in France.

In the twentieth century, heat waves occurred in France in 1911, 1928, 1947, 1957, 1976 and 1983. Between the episodes that cause a large number of deaths, there are also hot summers that have a small but real effect on mortality, such as, recently, the summers of 1990, 1994, 1995 and 1998. Research is therefore continuing with a view to describing the 100-year trend in mortality due to heat waves in the twentieth century, based on the related causes of death - underlying causes but also, wherever possible, contributing causes (available since 1968) - and on a cross-referencing of the air temperature recorded at Paris-Montsouris and in the rest of France with monthly death statistics, a process that will make it possible to evaluate the number of deaths due to heat over time.

Some references

Désesquelles A., Salvatore M.A., Frova L., PACE M., Pappagallo M., Meslé F. Egidi V., 2010. « Revisiting the mortality of France and Italy with the multiple-cause-of-death approach », Demographic research, vol. 23, n°28, p. 771 - 806

Meslé France, Vallin Jacques, 2012, Mortality and causes of death in 20th century Ukraine [Mortalité et causes de décès en Ukraine au XXe siècle], Dordrecht (Netherlands) / Heidelberg (Germany) / London (United Kingdom) / New York (USA), Springer-Verlag, 279 p ttp://

Meslé France et Vallin Jacques, 2008a. - Effet de l'adoption de la CIM-10 sur la continuité de la statistique des décès par cause. Le cas de la France, Population-F, vol. 63, n°2

Toulemon, Laurent et Magali Barbieri. 2008. The mortality impact of the August 2003 heat wave in France: Investigating the 'harvesting' effect and other long-term consequences, Population Studies 62(1): 39-54