Cause-of-Death Contributions to the Compression of Mortality in Switzerland, 2000-2014

Adrien Remund, Université de Genève
Michel Oris, Université de Genève
Bernadette W.A. van der Linden, EU ITN Marie CurieLongpop project
Mathias Voigt, EU ITN Marie CurieLongpop project

Increasing levels of life expectancy have traditionally been accompanied by a compression of ages at death, leading to a rectangularization of the survival curve. This process seems however to have slowed down in Switzerland in the last years and some indicators such as the standard deviation of ages at death above the mode (SD+) even suggest a possible stagnation of the compression process in the first years of the new century. In this paper we will (1) test whether this trend continued in the last years, (2) decompose the changes into cause-of-death components. In order to perform the second analysis, we will accomplish a redistribution of ill-defined causes of death using the heterogeneity between the 26 regions (cantons). The data will come from the Swiss National Cohort, which collects all death records that take place in Switzerland and matches them with individual census data.

Cause-of-deathcontributions to the compression of mortality in Switzerland, 2000-2014

Decreasing mortality levels at any ages generate an increase in lifeexpectancy. Whether these gains are located at young or old ages impactshowever the disparity of ages at death differently  (Vaupel & Canudas Romo, 2003). If improvements atyounger ages are stronger than those achieved at older ages, this generates acompression of the ages at death, while the opposite increases the disparity.

In practice, countries that experienced a rapid increase in lifeexpectancy during the twentieth century have also observed a rectangularizationof the survival curves due to a compression of age-at-death disparities. Thisprocess finds its roots in improvements of living standards and medicalprogress that allowed an ever-larger share of the population to reach (very)old ages, leaving however these old people facing a glass ceiling above whichgains in mortality are more difficult to achieve.

In the last decades however, this process of mortality compressionseems to have slowed down. In Switzerland, the decrease of the standarddeviation above the mode may even have come to a halt in the last years (Figure1). This suggests a possible new stage in the progression of old age survivaltowards a shift rather than a compression of ages at death.


Figure 1: Standard deviation abovethe mode in Switzerland (1876-2005)

Source:  (Cheung, 2009)

The cause-of-death dimension of theseprocesses has been largely ignored. At the moment we know for instance nothingabout the cause-specific contributions to the evolution of the standarddeviation at age at death above the mode in Switzerland. This would help putthese trends in perspective with the theory of the epidemiological transition,which postulates the emergence of an “age of delayed degenerative diseases”,following, according to Olshansky and Ault (1986), the three-phase modeldesigned by Omran (1971), and observed low-mortality countries from the 1970s. This age is characterized by anupward shift of the age pattern for degenerative causes of death to higher age.There is thus a need to estimate the contribution of such degenerative diseasesto the possible stagnation of mortality compression observed in the last years.

We will use data from the SwissNational Cohort (SNC), which is an individual record linkage of census, birth,death and migration registers of all people living in Switzerland. Sincemortality at old age is of primary importance in our analysis, we will payspecial attention to the so-called ill-defined causes, i.e. “symptoms, signs,abnormal clinical and laboratory findings, not elsewhere classified” (ICD-10codes starting with “R”). In the total population, between 2000 and 2014, thesedeaths remain relatively rare (3.6%), but are more common among old people (15%among centenarians). Moreover, these proportions vary across places ofresidence, as shown on Figure 2.

Figure 2: Canton-specificproportions of ill-defined deaths (2000-2014)

Source: Swiss NationalCohort, authors’ own cacluation

We will thus compute cause-specific proportionsfor each canton by sex. Then, we will regress the level of ill-definedmortality against those from each known cause. The coefficients of theseregressions will indicate how much each percentage point of ill-defined causesof death increases the percentage of other causes, providing us withredistribution coefficients that can be applied to ill-defined causes (Ledermann,1955).

Using this corrected dataset we willcompute summary statistics about mortality compression in old age. We will thendecompose the changes in these quantities by age and cause of death usingnumerical decomposition techniques (Andreev et al., 2002).


Andreev, E. M., Shkolnikov, V. M., & Begun, A. Z. (2002)."Algorithm for decomposition of differences between aggregate demographicmeasures and its application to life expectancies, healthy life expectancies,parity-progression ratios and total fertility rates". DemographicResearch, 7(14), 499-522.

Cheung,S.L.K., J.-M. Robine, Fr. Paccaud, A. Marazzi (2009), Dissecting thecompression of mortality in Switzerland, 1887-2005, Demographic Research21(19), 569-598.

HORIUCHI, S., & WILMOTH, J. R.(1998). "Deceleration in the age pattern of mortality at olderages". Demography, 35(4), 391-412.

Ledermann,S.(1955). "La répartition des décès de cause "indéterminée"".Revue de l''Institut International de Statistique / Review of theInternational Statistical Institute, 23(1/3), 47-57.

Olshansky,T. S.. Ault, Br. (1986), The Fourth Stage of the Epidemiologic Transition: Theage of Delayed Degenerative Disease, The Milbank Quaterly 64(3),355-391.

Omran,A. R. (1971), The Epidemiologic Transition: A Theory of the Epidemiologyof Population Change, The Milbank Memorial Fund Quaterly 49(4), 509-38.


Presented in Session 1192: Mortality and Longevity