CAIRN-INT.INFO : International Edition

Beyond the first years of life, the risk of dying increases with age, but more or less quickly. Early adulthood is marked by a temporary increase of this risk, in particular for young males. The origins of this excess mortality remain poorly understood and are hotly debated. Does it concern all young people? Is it attributable to biological or to social factors? Adrien Remund describes the excess mortality of a cohort of young adults living in Switzerland and focuses on the moderating or amplifying role of socioeconomic factors.

1Young adult excess mortality is one of three components that shape the human risk of death over the life course. It takes the form of a temporary deviation (in the shape of a “hump”) in the age-specific death rates during adolescence and/or early adulthood. This hump is located after the phase of decreasing risks during childhood known as ontogenescence (Levitis, 2011) and before the exponential increase driven by senescence (Gompertz, 1825). This specificity of young adult mortality has been known for almost 150 years (Thiele, 1871) and is now widely incorporated in mortality models and projections, albeit in an often descriptive way, to account for rather than to study this phenomenon for itself.

2In contrast to many studies that approach the hump as a macrodemographic object, this article focuses on a single population and explores interindividual differences in the hump: a cohort of young people born between 1975 and 1979 living in Switzerland and followed from adolescence to young adulthood. Figure 1 displays the age-specific shape of the hump for this cohort, which is characterized in both sexes by a sharp increase from 14 to 22 years of age, followed by a decrease until about 30 when the mortality curve rises again due to senescence. Its magnitude is about three times higher for males than females.

Figure 1

Age-specific death rate for males and females from cohorts born between 1975 and 1979 and living in Switzerland in 1990

Figure 1

Age-specific death rate for males and females from cohorts born between 1975 and 1979 and living in Switzerland in 1990

Source: Swiss National Cohort, author’s calculations.

3One theoretical aspect of the young adult mortality hump that remains uncertain is the extent to which its aggregate shape reflects an intra-individual phenomenon common to all young adults or the exacerbation of interindividual differences during this period of life. Broadly speaking, these two complementary hypotheses concerning the formation of the hump at the aggregate level can be described as endogenous and exogenous, respectively.

4Endogenous theories associated with adolescence and early adulthood are commonly found in the neuropsychological literature that often associates puberty with a supposed turmoil or, in late eighteenth-century literary terms, “Sturm und Drang” (Hall, 1904), also interpreted as a crisis due to the resurgence of Oedipal conflicts (Freud, 1968). Neuroimaging has shown peculiar neurological developments in the adolescent brain, which might explain specific behaviours of adolescents, such as mood disorders and risk-taking (Giedd, 2004; Giedd et al., 1999; Lenroot and Giedd, 2006; Spear, 2000; Steinberg, 2005). These regularities come, however, with large interindividual variations in the age at puberty (Pinyerd and Zipf, 2005) and levels of testosterone (Kelsey et al., 2014), which influence aggressiveness and risk-taking (Herbert, 2015). It is therefore likely that the endogenous changes observed at puberty generate a hump in the individual risk of death, though linked to different causes across individuals.

5The exogenous argument stems from young adults being confronted with several social adaptations that together are called the transition to adulthood. This process is structured around social events, such as leaving home, education attainment, labour market entry, union formation, and parenthood, which are sensitive to changes in socioeconomic conditions (Billari and Liefbroer, 2010; Blossfeld et al., 2005). [1] The quality of social resources at this crucial moment in the life course determines the ability to overcome the challenges posed by the adaptation to these new adult roles (Thomson et al., 2003). After a lowest point between 5 and 14 years of age (West, 1997), mortality differentials by social classes are stronger during early adulthood than at any other moment in life (Valkonen, 2006). Exogenous risk factors such as (but not only) social class may thus contribute to the presence of a hump in age-specific mortality among vulnerable parts of the population.

6In their influence on mortality, endogenous and exogenous factors can manifest themselves indifferently through any cause of death, including but not limited to those often deemed “external”. Each cause of death may thus originate from either an endogenous or exogenous process. This study therefore strictly measures the hump on an all-cause mortality level, although alternative measures using causes of death are also made to interpret the results.

7Large datasets are necessary to study mortality at low levels, such as during young adulthood. Because endogenous biomarkers of vulnerability are hard to gather on such a scale, this article specifically targets exogenous factors and observes the extent to which socioeconomic conditions contribute to young adult excess mortality by asking two questions. First, do exogenous conditions contribute to excess mortality during early adulthood by accentuating mortality differentials during this particular phase of life? This question can be answered by estimating age-specific patterns of the influence of socioeconomic variables and assessing the presence of hump-like shapes. Secondly, can favourable exogenous conditions partly or totally offset the hump in subgroups of the population? To answer this question, this article estimates the shape of the risk of death for different subpopulations and observes the degree to which all of them experience a young adult mortality hump. This analytical strategy uses a unique dataset that covers all deaths in Switzerland between 1990 and 2008.

I – Data

8The Swiss National Cohort (SNC) is a database that links four Swiss administrative sources: the 1990 and 2000 censuses, the vital statistics registries, and the foreigners’ registry (Spoerri et al., 2010). The SNC offers an exhaustive mortality follow-up of the entire resident population from December 1990 to the end of 2008, making it a statistically powerful dataset allowing detailed analysis of low-mortality populations, such as young adults.

9Probabilistic record linkage was used by the SNC’s designers to link individual information from the four administrative data sources, [2] reaching a 96.2% linkage success of death certificates with the censuses. This technique proved less successful on the younger cohorts, with a minimum of 79.7% linkage success for the 1975 birth cohort because of young people’s high mobility and changing characteristics. To compensate for this restriction, all remaining unlinked deaths were attributed to the best candidates based on a limited set of characteristics, an approach found to avoid biases (Schmidlin et al., 2013), while maintaining coherence with aggregated statistics and preserving strong statistical power.

10The study sample includes all individuals born between 1 January 1975 and 31 December 1979 who were present in the 1990 census. These people were between 11 and 16 years old in December 1990 and between 29 and 34 in December 2008 at the end of the follow-up. The initial at-risk population includes 374,833 individuals in the 1990 census, of whom 309,283 (82.5%) are linked to an entry in the 2000 census; 2,073 (0.6%) died; 6,311 (1.7%) emigrated; and 57,166 (15.3%) were censored because they do not appear later in the censuses or the registers. [3] Of the remaining 309,283 individuals observed in 2000, an additional 1,432 (0.5%) deaths were recorded between 2000 and 2008, along with 2,859 (0.9%) emigrations, while the remaining 304,992 individuals were assumed to be present at the end of the follow-up in 2008 (see Appendix Table A.1). The full study dataset contains 6,182,011 personyears of exposure and 3,505 deaths.

11Descriptive statistics of individual characteristics are shown in Table 1. These include variables from the 1990 census: sex, type of household in which they lived in 1990, [4] urbanization of their municipality, main language spoken in their region, and religion (declared by the parents). Information was also collected on parents at the time of the 1990 census, including their highest level of education and highest occupational category, [5] the age difference between the young adult and his or her father and mother, and migration background. [6] In the 2000 census, additional socioeconomic information was collected on young adults’ educational attainment [7] and professional activity. The latter captures a combination of labour market status and current schooling into 11 categories reflecting the variety of situations during early professional stages. [8] One limitation of this dataset is that it does not provide any early life indicators or direct medical measures of health that life course studies have shown to have important implications on later health (Kuh and Shlomo, 2004). This restriction illustrates the often conflicting objectives of gathering many explanatory variables while keeping a large sample size.

Table 1

Individual and parental characteristics of young adults in Switzerland*

Table 1
Individual variables Category Frequency (%) Sex (1990) Male 51.3 Female 48.7 Household type (1990) Nuclear 87.5 Single-parent 9.6 Other family household 0.6 Non-family household 0.3 Institution 2.0 Type of municipality (1990 & 2000) Urban 23.5 Periurban 46.0 Rural 30.5 Linguistic region (1990 & 2000) German 72.2 French 23.5 Italian 4.2 Religion (1990) Catholic 49.0 Protestant 39.4 Other religion 6.2 No religion 5.4 Economic activity (2000) Employed 50.5 Unemployed 3.0 Benefits 0.4 Compulsory schooling 0.4 Apprenticeship 4.6 Secondary 2.0 Superior (professional) 5.1 University 12.0 NEET 2.6 Unknown 2.0 Missing 17.5 Highest education attained (2000) Compulsory 11.3 Secondary 62.1 Tertiary 5.3 Unknown 3.9 Missing 17.5 Number 374,833 Parental variables* Category Frequency (%) Educational level (1990) Compulsory 18.7 Secondary 52.9 Tertiary 26.2 Unknown 2.3 Occupational category (1990) Manager 2.8 Liberal professions 2.0 Self-employed 17.2 Intellectual and executive 13.7 Intermediate 22.5 Qualified non-manual 12.9 Qualified manual 9.5 Non-qualified 12.9 Others 1.7 Unemployment 0.4 Unpaid work 2.3 Unknown 2.1 Age difference with mother (1990) 15–19 years 4.5 20–29 years 63.3 30–39 years 26.5 40+ years 2.1 Unknown 3.6 Age difference with father (1990) 15–19 years 1.4 20–29 years 43.5 30–39 years 38.4 40+ years 5.9 Unknown 10.7 National origin (1990) Switzerland 64.3 European Union 23.7 Balkans 3.9 Rest of the world 8.0

Individual and parental characteristics of young adults in Switzerland*

* When two parents are present in the household, the highest of each of their educational levels and occupational categories is retained. Regarding national origin, only information about the young adults are used (place of birth and nationality).
Source: Swiss National Cohort, author‘s calculations.

II – Methods

12This study uses survival analysis to answer the first question on whether exogenous risk factors contribute to the formation of the hump by generating stronger mortality differentials around the age of the hump. Most survival studies assume that mortality differentials apply proportionally (i.e. with the same intensity over time) and tend to define time as calendar years as opposed to age, thus ignoring the potential dynamics of risk factors over age (Thiébaut and Bénichou, 2004). To overcome this limitation, age is here considered as the variable of time. Observations are left-truncated before the age at the 1990 census and right-censored at the age of last observation if death did not occur before. Exposure time is divided into two phases (1990–2000 and 2000–2008) to accommodate changes in individual characteristics between the two censuses. Professional activity and highest education attained, which are only known from the 2000 census, are set as “compulsory school” for everyone before 2000 to avoid making these variables “prophetic” (Therneau and Grambsch, 2000). [9] Likewise, during the post-estimation, age-specific effects for these variables are computed only from age 20 when the information about these variables becomes available (Figure 2). Interactions between education and birth cohorts are also tested to control that the effect of educational attainment is not confounded by age at the census. Multiple observations of the same individual at each census are grouped to account for the correlation between these observations in the computation of robust variances.

13Regarding the model itself, the individual risk of death over age μi(x) is defined as the product of a baseline hazard μ0(x) common to everyone and an individual vulnerability vi(x) that varies both over age (x) and across individual profiles (i) (Equation 1).

15This definition of individual vulnerability, which is both individual and dynamic, builds on the preexisting concepts of frailty, which is individual but constant over time (Vaupel et al., 1979), and of vulnerability, which is time-varying but common to all individuals (Vaupel et al., 1988). Therefore, each combination of risk factors yields a specific vulnerability profile, i.e. an age-specific response that can act as a buffer or a magnifier on the underlying risk of death.

16We first estimate a classic Cox semi-parametric model and subsequently approximate age-specific effects using the scaled Schoenfeld residuals (Schoenfeld, 1982). These are defined as the difference between the observed value of a given covariate at the moment of someone’s death and its conditional expectation given the characteristics of the people still alive at this moment. When plotted against age, the shape of these residuals can be interpreted as age-specific effects (Grambsch and Therneau, 1994). These shapes are estimated with splines, and confidence bands are estimated by bootstrapping on 1,000 samples. Including individual-level covariates in the model thus leads to the following extension of the Cox model (Equation 2), where μ0(x) is the baseline hazard, βi(x) are the age-specific coefficients, and Zi(x) are the individual covariates.

18The shapes of the μi(x) indicate the extent to which exogenous processes can buffer the hump from the baseline. Since studying each of the 374,833 individual risks is unfeasible, one can aggregate them into a reduced set of representative subpopulations that together account for the entire population. Instead of using the explanatory variables, these groups are created directly based on the individual risk scores from the Cox model, using it as a multidimensional indicator of vulnerability. Grouping individuals according to these scores can serve as a test of proportionality designed to check whether the hump is present at all levels of vulnerability. This grouping is performed using the CLARA technique (Clustering Large Applications), a type of Partitioning Around Medoids (PAM) algorithm designed for large datasets (Kaufman and Rousseeuw, 1990), which consists in finding the population’s most representative individuals. Clusters are built for each sex by assigning each individual to the nearest representative, thus leaving out no individual. The resulting clusters can be interpreted as subpopulations that share a similar trajectory of age-specific vulnerability to death. Once the dataset is split into these subpopulations, age-specific death rates are computed for each of them, using the first derivative of a smoothed function of the Kaplan–Meier cumulative hazard.

III – Results and discussion

19Table 2 displays the results from a first full model containing all the information available from the individual and parental variables. In the second parsimonious model, several variables and categories are dropped because their effects are not significantly different or because of multicollinearity. [10]

20For this second model, age-specific effects are approximated, as explained in the Methods section and presented in Figure 2. These results confirm the existence of strong mortality differentials during the transition to adulthood and show that these differentials are often unstable with age. The majority of individual characteristics in the model feature a U-shape or inverted U-shape in the age-specific effects, which means that they generate particularly strong differentials during early adulthood. All the results presented here are understood as net of confounding effects and independent of the shape of the baseline risk n0(x). Appendix Figure A.1 shows the brute (uncontrolled) age-specific death rates by categories of each variable. This allows studying the shape of the hump in these different subgroups.

21In addition to the all-cause mortality analyses, Figure 3 illustrates the cause-specific death rates observed in each category. [11] Overall, other (mostly non-external) causes of death are the main killer (about 40% of all deaths), but their intensity varies across categories of people. The mortality rate is especially high among young people not in education, employment, or training (NEET) and those living in institutions, and even more so among people living on social benefits. Traffic accidents, other accidents, and suicides each account for about 20% of deaths on average and display less variation. Suicides are particularly frequent among children of non-family households, unemployed, and recipients of social benefits, but relatively rare among females, in the Italian region of Switzerland, and among non-Christian religions. Traffic accidents are more frequent in the Italian region and among males, and less so among inactive individuals.

1 – Sex

22Young Swiss women have on average 60% less risk of dying between the ages of 10 and 35 than young men. This female advantage starts at age 10 with a 35% gap, increases until about age 22 to 70%, then narrows progressively (Figure 2A). This means that, while young males are always at a mortality disadvantage, they suffer their greatest handicap in the first phase of the transition to adulthood, which also coincides with the peak of the aggregate mortality hump. It is impossible at this stage to say if this differential between males and females is due to biological sex or socially constructed gender roles. However, these effects concern all causes of death and remain after controlling for all the other available dimensions of social life.

2 – Familial context

23Although youth who grew up in nuclear households experience a small hump (Appendix Figure A.1F), those who lived part of their childhood in a non-nuclear household face a higher risk of death, no matter the specific type of cohabitation. In single-parent households, children are exposed to an average 50% higher mortality risk than nuclear households. Vulnerability reaches a peak early on, decreases over time, and eventually disappears by their early thirties (Figure 2F). In other words, single-parent households are a source of vulnerability for the adolescents who currently live in them, but they do not leave a scarring effect. Several possible processes may be the source of vulnerability, such as conflict, lack of support by the parent who lives outside the household, and adverse material conditions generated by the parents’ separation, as well as a reverse effect of the child’s poor health on the parents’ relationship.

24The other, more atypical forms of cohabitation also exert a detrimental effect on the risk of death of young adults, whether they lived at the time of the census in non-family households (such as foster families) or in institutions (such as hospitals and foster residence). Their risk of death increases on average by 50% (Table 2). This excess mortality can be interpreted as a selection effect due to poor health requiring a stay in a healthcare institution, which is supported by the high level of mortality for non-external causes in this category, although the rates for all causes of death are higher for this group compared to nuclear households (Figure 3). The age-specific effect of living in an institution or other types of households shows a clear tendency to peak early on and decrease over time (Figure 2F), supporting the interpretation of the institution as a marker of bad health. Nevertheless, the age-specific excess mortality of young people having lived in an institution rises during their early twenties, which is also the case for young adults who lived in non-family households (Figure 2G). This confirms the well-documented fact that young adults raised outside their biological family find the transition to adulthood a particularly challenging moment (Barth, 1990; Kerman et al., 2002; Yaouancq and Duée, 2014). Among them, suicides are a more frequent cause of death (Figure 3).

Table 2

Effects of individual characteristics on the mortality of young adults in Switzerland (risk ratios)

Table 2
Model 1 Model 2 Sex Male (Ref.) 1 1 Female 0.390*** 0.390*** Household type Nuclear (Ref.) 1 1 Single-parent 1.225 1.448*** Other family household 0.829 0.791 Non-family household 1.242 1.579† Institution 3.316 1.455* Type of municipality Urban (Ref.) 1 1 Periurban 1.059 1.057 Rural 1.182*** 1.178*** Language region German (Ref.) 1 1 French 1.032 1.041 Italian 0.800* 0.808* Religion (Ref = Model 1: Catholic; Model 2: Christian & no religion) Protestant 0.965 1 Other religion 0.760** 0.764** No religion 1.038 1 Economic activity Employed (Ref.) 1 1 Unemployed 1.886*** 1.892*** Benefits 6.487*** 6.504*** Compulsory school 0.936 0.933 Apprenticeship 0.846 0.839 Secondary 0.857 0.856 Superior (professional) 0.892 0.890 University 0.845† 0.849 Unknown 0.898 0.903 NEET 1.658*** 1.663*** Highest education attained Compulsory (Ref.) 1 1 Secondary 0.701** 0.702*** Tertiary 0.556** 0.571*** Unknown 1.026 0.851 Parental education Compulsory (Ref.) 1 Secondary 1.010 Tertiary 1.063 Unknown 0.877
Table 2
Model 1 Model 2 Parental occupation Non-qualified (Ref.) 1 1 Manager 0.842 0.868 Liberal professions 1.263 1.329* Self-employed 1.189* 1.205** Intellectual and executive 1.047 1.085 Intermediate 1.120 1.140* Qualified non-manual 1.065 1.067 Qualified manual 1.206* 1.212** Others 1.360* 1.348* Unemployment 1.684** 1.708** Unpaid work 1.277* 1.268* Unknown 0.455 Age difference with mother 20–29 (Ref.) 1 15–19 1.046 30–39 0.985 40+ 0.944 Unknown 1.260 Age difference with father 20–29 (Ref.) 1 1 15–19 1.303* 1.327* 30–39 1.019 1.012 40+ 1.214* 1.188* Unknown 1.177 0.993 National origin Switzerland (Ref.) 1 1 European Union 1.156** 1.167*** Balkans 1.170 1.195† Rest of the world 1.508*** 1.561*** Birth cohort 1975 (Ref.) 1 1 1976 0.930 0.923 1977 1.007 0.988 1978 0.968 0.968 1979 0.875 0.886* BIC 88,125 87,957 Number 374,833

Effects of individual characteristics on the mortality of young adults in Switzerland (risk ratios)

Significance levels:p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001
Source: Swiss National Cohort, author’s calculations.
Figure 2

Age-specific effects of individual characteristics on the mortality of young adults in Switzerland (risk ratios)

Figure 2Figure 2Figure 2Figure 2

Age-specific effects of individual characteristics on the mortality of young adults in Switzerland (risk ratios)

Note: The scales are identical in all subfigures, except 2B. The reference groups are indicated in Table 2.
Source: Swiss National Cohort, author’s calculations.
Figure 3

Cause-specific death rates by individual characteristics

Figure 3

Cause-specific death rates by individual characteristics

Source: Swiss National Cohort, author’s calculations.

25Another family trait examined here is the age difference between young adults and their fathers. Individuals born to 20–39-year-old fathers had on average a slightly lower risk of death (Appendix Figure A.1G). In comparison, children born to younger or older fathers suffer an increase of 30% and 20%, respectively, in their risk of death (Table 2). All causes of death seem to be similarly affected across paternal age, but the dynamic of this risk factor differs between individuals with young and old fathers. The disadvantage of children born to younger fathers appears to peak in their early twenties, suggesting that they struggle to find their independence as adults, whereas children born to older fathers are mainly exposed to higher risks until about 25 years of age (Figure 2H). This difference may be related to an inability of young fathers to face their new responsibilities, to transfers of social capital, or to other aspects of a parent-child relationship that could be affected by parental age, although these interpretations remain conjectural.

3 – Occupational and educational gradient

26Occupation and level of education are both correlated with the risk of death in what is known as the social gradient (e.g. Marmot, 2005). Establishing this gradient represents an additional challenge in the case of young adults who are transitioning out of the educational system and their inherited social position, and towards the labour market and their own social class, the markers of which must incorporate these different sources of disparities (family, schooling, and work). Early adulthood is marked by a sharp increase in social inequalities (West, 1988) that may be due to the loss of parental influence, to rapid social mobility, or to the time it takes before the social environment translates into measurable health inequalities because of latency of the development of diseases or of at-risk behaviours.

27The hypothesis of a weakening influence of parental social class, supported by other studies (Pensola and Valkonen, 2002; Remes et al., 2010), is not systematically confirmed here. Although the disadvantage of young adults whose parents were unemployed (+70%), doing unpaid work (+30%), or in other (+35%) employment (Table 2) was expected, it does not dissipate with time (Figure 2C). Sometimes, it even displays peaks in the late teens or a scarring effect that extends to older ages. Moreover, regarding the other parental occupation categories, children from parents in intermediate occupations or liberal professions or who are self-employed surprisingly suffer from greater mortality than the supposedly more disadvantaged unqualified workers. The others do not differ significantly. Counterintuitively, in Switzerland, the young adult hump is particularly accentuated among those from supposedly advantaged economic backgrounds, possibly because they lack social support due to an unfavourable work–life balance and to their parents’ extended work schedules. Children of parents in the higher-level occupational categories, except managers (before age 30), all display higher excess mortality during their twenties (Figure 2C). Figure 3 suggests that this higher mortality goes in hand with a larger share of suicides, while for other disadvantaged groups all causes are affected. But this conclusion only holds after controlling for all confounding effects. The observed raw death rates for the more privileged groups fall either below or very close to those of children of non-qualified workers (Appendix Figure A.1D), which indicates that young adults from privileged backgrounds are shielded by other protective characteristics.

28The young adults’ own educational gradient, in contrast, points in the expected direction (Table 2). Compared to those with only compulsory schooling, young adults who completed secondary school experience a 30% lower risk on average, and those who go to university or another tertiary institution experience a further 20% advantage. [12] This difference seems to be driven mainly by higher risks of traffic accidents for people with compulsory schooling (Figure 3), which supports findings from other countries such as Sweden, where 18- to 30-year-olds of the lowest educational level suffer a twofold risk of traffic accident and a fivefold risk of being injured or killed (Hasselberg et al., 2005). The age-specific effect of the young adults’ own educational attainment (Figure 2D) displays a growing gap after age 25. Young adults with the lowest level of education thus seem more negatively selected with time. The educational gradient of mortality might be steepest only after a phase of latency during which inequalities build up.

4 – Socioeconomic exclusion

29The social gradient, unequal as it may be, only concerns the majority of the population that is actually part of the educational and professional system. Other people are somehow left behind and excluded first from the educational system, then from the labour market, and suffer much higher mortality risks, in turn. In the context of a more polarized labour market, the transition to adulthood represents a more critical phase of the life course than it used to (Esping-Andersen, 2009), and the consequences of being excluded from the educational system at this stage are probably more important than ever (Blossfeld et al., 2005).

30Socioeconomic exclusion begins at school age and extends later into the occupational and health spheres. The first step of this process begins when young people leave the educational system without obtaining a useful qualification nor entering the labour market. This group of so-called NEETs (ILO, 2013) are exposed to a 70% higher mortality rate than to those who are employed (Table 2). The age-specific effect is not clear, as shown by the wide confidence intervals (Figure 2B). Although all causes but traffic accidents are concerned, it is especially the case for other causes of death (Figure 3), suggesting a possible countermechanism that leads young people in poor health and who present higher risks of death to drop out of school and work.

31Unemployment can be conceived as a second step in the process of socioeconomic exclusion. Unemployed young adults suffer on average a twofold risk of death compared to their employed counterparts. The age-specific effect of unemployment is clearly hump-shaped and peaks around age 25 (Figure 2B). The detrimental effect of unemployment thus coincides with the peak of the hump. Here as well, all causes but traffic accidents display higher levels, which suggests the presence of health selection effects as well as detrimental effects of unemployment (Figure 3).

32The third and most detrimental step of socioeconomic exclusion corresponds to the reception of social benefits, which is associated with an average sixfold increase in the risk of death compared to young adults who are employed (Table 2). Perhaps even more than in previous categories, this may be partly a health selection effect because young people who suffer from life-threatening health issues are more likely to receive invalidity pensions. Figure 3 supports this interpretation by showing very high levels of other causes of death, including causes related to long-term invalidity, but the excess mortality concerns all causes except traffic accidents. The age-specific effect of receiving social benefits is not constant (Figure 2B), although conclusions are difficult to reach because of extremely wide confidence intervals. The median estimates suggest that this group of young people may become more disadvantaged during their twenties, perhaps experiencing a drop to the initial level, thus forming a late hump.

33Together, these three markers of socioeconomic exclusion can thus be understood as successive steps of a general process of marginalization. The absence of a qualifying education often entails unemployment, which in turn can lead to a dependence on social benefits. Health selection also plays a role by forcing the unhealthiest young adults, who already face elevated risks of death, out of the educational and professional systems and into receiving invalidity pensions. The accumulation of these risk factors has the potential to generate a subpopulation that progressively accumulates disadvantages and becomes increasingly marginalized (Dannefer, 2003; Mackenbach, 2012).

5 – Geocultural environment

34A few individual characteristics do not display a curvature in the age-specific shape of their mortality differentials, although these effects are not constant over time. Among these, the impact of national origin is the strongest (Figure 2G). Young adults who have roots in countries other than Switzerland face an increased risk of death that amounts on average to 17–20% for people of European and former-Yugoslavian descent, and over 50% for those of extra-European origin. For all these origins, the shape of the age-specific hazard ratio displays an almost linear decreasing trend that could be an effect of time since migration, rather than age, because both dimensions are indistinguishable in this dataset. In the case of Switzerland at least, immigrants start with a mortality disadvantage and then converge with natives. This process eventually leads to a reversal of the disadvantage, ultimately leaving natives in a worse position, all things held constant. This paradox is due to an advantage of migrants over natives in vulnerable positions (Zufferey, 2016). In brute rates, Swiss youth maintain an advantage over the other groups (except those from the Balkans) thanks to their compositional advantage in other characteristics (Appendix Figure A.1J).

35Two other geographical characteristics have significant effects on mortality differentials during the transition to adulthood, without displaying a hump. First, young adults in rural, and possibly periurban, municipalities have a slightly higher mortality than their urban counterparts (Figure 2I). This observation supports findings from a comparable study in Sweden (Hjern and Bremberg, 2002) and may be explained by higher automobile usage and fatality rates outside urban agglomerations (Zwerling et al., 2005). Figure 3 supports this interpretation because the risk of traffic accidents is higher in rural settings. This residential effect progressively disappears as young adults get older, possibly because of unrecorded mobility to urban centres after the census. Secondly, inhabitants of the Italian-speaking region have a 20% lower risk of death than the rest of the country, which appears only after 20 years of age (Figure 2E). This advantage seems linked to a lower risk of suicide and other causes of death (Figure 3), which is consistent with suicide mortality being lower in Italy than in Switzerland (Liu, 2009).

6 – Accumulation of risks and the shape of the hump

36This broad analysis of the cultural, spatial, social, and economic risk factors of mortality shows that the majority of them display a peak of intensity in people between ages 20 and 30. In other words, exogenous effects contribute to their excess mortality independently from the endogenous risk of death. The peak of the observed hump for the overall population is located at 21.6 years for males and 18.9 years for females (Figure 1). Incidentally, this age corresponds approximately to the average location of the maximum intensity of the age-specific mortality differentials. The mean location of the peak for the 26 characteristics displaying a hump is 22.9 years of age, or 23.1 after weighting by the number of people in each category (Table 1). At the aggregate level, the peak could be at least partly due to the particularly detrimental effect between ages 20–25 of negative socioeconomic conditions surrounding the transition to adulthood.

37Although practically all variables display a higher mortality at these ages (Appendix Figure A.1), this does not mean that all individual profiles do. Testing this question requires observing how net effects aggregate into individual profiles. Notably, adverse risk factors tend to accumulate on the same people, generating strong mortality differentials between individuals or a temporary increase, for only some of them, in their risk of death during early adulthood.

38Since studying each individual hazard is not practical, they are grouped into three subpopulations for each sex (see Methods section). Appendix Tables A.3 and A.4 display the composition of each subpopulation for males and females, respectively; the results are very similar across sexes. Overall, the composition of these groups reflects all the gradients identified earlier, but some peculiarities are worth highlighting because they demonstrate how risk factors tend to accumulate.

39Group 1, which covers two-thirds (69%) of the cohort, aggregates all the other profiles, which, despite their variety, are mostly unaffected by negative risk factors. Group 2, which makes up about a quarter (28%) of the cohort, contains young adults with moderate risk factors generated by their educational level and/or a disadvantaged family background. Group 3, which accounts for 4% of males and 3% of females, reflects the situation of people in acute states of vulnerability because of health issues or an extremely disadvantaged family background. All people receiving benefits belong to the same group (Group 3) but together only make up about 10% of it. The rest of this group often grew up in institutions or single-parent families (80% of cases), without a co-resident father (unknown father’s age 70%, unknown/unclassified parental occupational category 40%), and with extra-European roots (40%). Groups 1 and 2 mainly differ in the proportion of people who finished secondary (ca. 75% vs. 40%, respectively) or stopped after their compulsory education (ca. 15% vs. 50%). Group 2 also contains more people from single-parent families and thus of unknown father age (in both cases about 25% vs. 1% in Group 1).

40The shape of the hazard rate for each of these subpopulations suggests that the young adult mortality hump is not shared by all Swiss youth (Figure 4). Among both males and females, for the least vulnerable subpopulation (Group 1) the risk of death is low in their early teens and progresses regularly, showing little or no sign of a deviation from the general rise in mortality associated with ageing. This shows that the young adult mortality hump is not an unavoidable phenomenon and that, under favourable conditions, socioeconomic resources can act as a buffer to absorb the stress generated by the transition to adulthood.

41The intermediate Group 2 displays a hump similar to that of the total population, but more pronounced. Towards age 30, this group experiences a reduced risk of death and joins up with the more protected group. For these individuals who do not possess adequate resources, the stress generated by the transition to adulthood represents a challenge that translates into a temporary but strong impact on their risk of death.

42The most vulnerable young adults (Group 3) follows an entirely different trajectory. Males first follow the other subpopulations at a higher level, but after the peak of the hump they never converge with them. For females, the trajectory is constantly at a much higher level, despite a drop towards age 30, and does not join the other subpopulations either.

Figure 4

Force of mortality in subpopulations of young adults in Switzerland, 1990–2008

Figure 4

Force of mortality in subpopulations of young adults in Switzerland, 1990–2008

Source: Swiss National Cohort, author’s calculations.

43This analysis of excess mortality across different subpopulations of young adults in Switzerland suggests that the young adult mortality hump is not an unavoidable phenomenon within a given population. In this case, it applies to about a quarter of the young adults. The young adult mortality hump, to which endogenous forces may or may not contribute, can therefore be entirely offset by a favourable environment.


44The young adult mortality hump remains poorly understood, far less than senescence and even ontogenescence. Several untested assumptions still dominate the debate, including the universality of the hump and its endogenous or exogenous nature. On the one hand, the endogenous hypothesis assumes that this phenomenon originates in the largely unavoidable, neuropsychological development related to maturation. The exogenous hypothesis, on the other hand, maintains that the socioeconomic environment may inflict stress on adolescents. This study on young adults in Switzerland was designed to test specifically the validity of the exogenous argument.

45After reviewing the age-specific shape of the mortality differentials, this study finds that exogenous vulnerability contributes to exacerbating the risk of death of certain social groups during the transition to adulthood. Reciprocally, social resources act as a protection against the endogenous or exogenous stress generated by becoming an adult, which explains the absence of excess mortality in sufficiently sheltered subpopulations. In addition, extreme social exclusion reinforces this risk and persists beyond early adulthood.

46This analysis of young adult mortality in Switzerland is a first step towards a better understanding of the factors that shape it. It does not close the debate between the supporters of the endogenous and exogenous interpretations; and even though it confirms the importance of nurture, it does not exclude a complementary natural origin. These conclusions provide a fertile ground for future debate.


I would like to thank Michel Oris, Jamaica Corker, and the anonymous reviewers for their valuable comments on earlier versions of this manuscript. This research benefited from the financial support from the Swiss National Science Foundation (Early Postdoc.Mobility grant and NCCR LIVES IP 213) and a Small Nested Project provided by the Swiss National Cohort.


Table A.1

Evolution of the population at risk

Table A.1
Sample size % of 1990 census % of 2000 census Census 1990 374,833 100 Death 2,073 0.6 Emigration 6,311 1.7 Unlinked 57,166 15.3 Census 2000 309,283 82.5 100 Death 1,432 0.5 Emigration 2,859 0.9 31 Dec. 2008 304,992 98.6

Evolution of the population at risk

Source: Swiss National Cohort, author’s calculations.
Table A.2

Classification of the causes of death

Table A.2
Traffic accidents Other accidents Suicides Other causes ICD-8 800–845, 940, 941 850–929, 942–949 950–959 other codes ICD-10 Chapter V, Y85, Chapter W, X1–X5, Y86 X6–X7, X80–X84, Y87 other codes

Classification of the causes of death

Source: The International Classification of Diseases (ICD) is published by the World Health Organization (WHO), which periodically updates it. In Switzerland, version 8 of the ICD was used until 1994, then version 10 was directly adopted without using version 9.
Table A.3

Individual and parental characteristics of males by cluster

Table A.3
Individual variables % by cluster 1 2 3 Household type (1990) Nuclear 98.8 72.1 18.9 Single-parent 0.6 24.4 39.8 Other family household 0.5 0.7 0.3 Non-family household 0.0 0.7 1.1 Institution 0.0 2.1 39.9 Type of municipality (1990 & 2000) Urban 30.2 24.9 28.7 Periurban 45.6 40.9 38.8 Rural 24.3 34.2 32.5 Linguistic region (1990 & 2000) German 74.7 68.9 63.3 French 20.9 28.6 34.0 Italian 4.4 2.5 2.7 Religion (1990) Christian & no religion 93.3 94.5 90.7 Other religion 6.7 5.5 9.3 Economic activity (2000) Employed 57.2 40.5 23.7 Unemployed 1.4 6.4 7.4 Benefits 0.0 0.0 9.3 Compulsory schooling 7.5 32.5 46.5 Apprenticeship 5.0 3.0 1.7 Secondary 2.2 0.9 0.4 Continuing 7.6 4.1 1.4 University 16.1 7.7 4.4 NEET 0.9 3.0 4.1 Unknown 2.1 1.8 1.2 Highest education attained (2000) Compulsory 15.3 47.0 68.4 Secondary 74.7 44.4 16.6 Tertiary 6.8 3.8 3.1 Unknown 3.3 4.8 8.9 % by cluster Parental variables 1 2 3 Occupational category (1990) Manager 4.0 0.6 0.2 Liberal professions 1.8 2.4 1.0 Self-employed 16.9 19.2 8.2 Intellectual and executive 16.6 8.2 3.0 Intermediate 24.9 18.4 10.4 Qualified non-manual 13.8 11.7 6.4 Qualified manual 7.1 15.4 6.5 Non-qualified 13.7 12.2 6.3 Others 0.7 5.9 44.5 Unemployed 0.0 0.6 5.1 Unpaid work 0. 5 5.3 8.5 Age difference with father (1990) 15–19 years 0.7 2.7 1.9 20–29 years 50.9 32.2 10.7 30–39 years 44.1 30.1 9.4 40+ years 3.5 11.0 6.1 Unknown 0.8 23.9 71.9 National origin (1990) Switzerland 72.5 49.9 36.6 European Union 21.3 29.4 21.7 Balkans 3.6 5.1 4.4 Rest of the world 2.6 15.6 37.3

Individual and parental characteristics of males by cluster

Note: Only the variables retained in the second Cox model are displayed.
Source: Swiss National Cohort, author’s calculations.
Table A.4

Individual and parental characteristics of females by cluster

Table A.4
Individual variables % by cluster 1 2 3 Household type (1990) Nuclear 98.8 72.1 17.5 Single-parent 0.6 24.9 47.3 Other family household 0.6 0.7 0.4 Non-family household 0.1 0.7 0.8 Institution 0.0 1.6 34.1 Type of municipality (1990 & 2000) Urban 32.9 25.9 29.4 Periurban 45.2 39.6 39.2 Rural 21.9 34.5 31.3 Linguistic region (1990 & 2000) German 73.5 69.3 61.4 French 21.9 28.0 35.7 Italian 4.6 2.7 3.0 Religion (1990) Christian & no religion 93.7 95.0 92.3 Other religion 6.3 5.0 7.7 Economic activity (2000) Employed 58.6 35.7 20.9 Unemployed 1. 5 5.2 5.7 Benefits 0.0 0.0 8.5 Compulsory schooling 9.2 37.3 47.8 Apprenticeship 5.8 3.2 1.4 Secondary 3.0 1.0 0.4 Continuing 4.6 2.5 1.5 University 13.2 6.7 4.2 NEET 1.9 6.5 8.0 Unknown 2.2 1.8 1.6 Highest education attained (2000) Compulsory 18.1 52.3 69.3 Secondary 72.9 40.2 20.6 Tertiary 5.7 3.3 2.4 Unknown 3.3 4.2 7.7 % by cluster Parental variables 1 2 3 Occupational category (1990) Manager 4.0 0.6 0.2 Liberal professions 2.1 2.2 0.9 Self-employed 16.2 20.2 8.4 Intellectual and executive 17.4 7.4 3.1 Intermediate 25.3 18.3 11.4 Qualified non-manual 13.8 11.7 6.6 Qualified manual 6.7 15.3 7.5 Non-qualified 13.2 13.0 6.9 Others 0.7 5.1 39.0 Unemployed 0.0 0.8 5.1 Unpaid work 0.5 5.3 11.0 Age difference with father (1990) 15–19 years 0.9 2.5 1.3 20–29 years 50.9 31.6 9.5 30–39 years 43.9 30.4 9.1 40+ years 3.6 11.2 5.9 Unknown 0.8 24.2 74.1 National origin (1990) Switzerland 72.1 51.5 35.1 European Union 21.8 28.2 21.0 Balkans 3.3 4.5 3.7 Rest of the world 2.8 15.8 40.3

Individual and parental characteristics of females by cluster

Note: Only the variables retained in the second Cox model are displayed.
Source: Swiss National Cohort, author’s calculations.
Figure A.1

Age-specific mortality rates of young adults by individual characteristics in Switzerland

Figure A.1Figure A.1Figure A.1

Age-specific mortality rates of young adults by individual characteristics in Switzerland

Source: Swiss National Cohort, author’s calculations.


  • [1]
    That this transition has become longer over the last decades is not incompatible with the observation that the peak of the hump occurs at progressively younger ages (Goldstein, 2011) because the hump has also become wider and affects older people simultaneously (Kostaki, 1992).
  • [2]
    The variables used for the linkage are sex, date of birth, civil status, place of birth, nationality, language, religion, and occupation. In the case where no match can be done in the same municipality, the search continues at progressively larger levels of aggregation (Bopp et al., 2009). A sensitivity test was performed by re-estimating the models without the deaths that were initially unlinked; similar coefficients were found.
  • [3]
    These are Swiss citizens whose emigration between 1990 and 2000 was not registered (for one-third) and people that could not be linked to their record in the 2000 census (for two-thirds). These 57,166 people were censored at the time of the 2000 census. Since most emigrations happened when people were in their twenties, we probably slightly underestimate the death rates in younger ages and overestimate them in the older ages under observation.
  • [4]
    Non-family households are those in which young people share a dwelling with people other than their parents (e.g. roommates, foster families). Institutions correspond to all kinds of collective dwellings (e.g. hospitals, boarding schools). Other households, besides nuclear and single-parent, are composed essentially of couples without children, as well as a few lone adults and adults who live with their parents.
  • [5]
    The occupational categories are based on the typology developed by Joye and Schuler (1995), itself inspired by Wright (1985), which combines educational level and organizational competencies, resulting in a strong correlation between education and occupational categories. The socio-occupational category, including unemployment, only records the highest level among the parents.
  • [6]
    Four variables were used to define the origin: nationality in 1990, place of birth, mother’s and father’s place of birth, and, in case of multiple origins, with the following priorities: Balkans > Rest of the world > European Union > Switzerland. People from the Balkans, who are often subject to negative prejudices (Baumann, 2012), are treated separately.
  • [7]
    Education in Switzerland is compulsory until about age 16. The measure of current activity further differentiates between those in high school (secondary), superior (professional) education, and tertiary education (universities, federal institutes of technology, and universities of applied science).
  • [8]
    Recipients of social benefits constitute a particular category, which comprises people that receive social assistance and disability insurance. Unfortunately, the data cannot distinguish these two groups.
  • [9]
    If a specific status, such as “too young”, had been used for the period 1990–2000, it would have predicted the fact that this person was not observed after 2000. This bias is avoided by using a category that is present after 2000 (compulsory school), which probably reflects the real state of most people in 1990.
  • [10]
    The status of the non-cohabiting parent is by definition unknown in non-nuclear households because parental ties are derived from cohabitation.
  • [11]
    See Appendix Table A2 for the detailed grouping of the ICD items.
  • [12]
    The interactions with the birth cohorts are not significant, thus excluding a confounding effect of age at the census.

Early adulthood is often characterized by a phase of excess mortality. It is not clear whether this temporary increase in the risk of death occurs because of biological or contextual forces, nor whether this threat concerns all individuals of the same cohort. Age-specific mortality differentials from 10 to 34 years of age are calculated using a unique dataset that includes all individuals living in Switzerland born between 1975 and 1979. Certain risk factors act with variable intensity and follow patterns similar to the hump observed in the overall age-specific mortality risk. The results suggest that socioeconomic mortality differentials partly shape the hump. The division of the cohort into multiple subpopulations representing various levels of vulnerability shows that although a minority of Swiss youth experience a phase of temporary excess mortality, this is not the case for all groups of individuals. Overall, the results indicate that a favourable social context offsets the stress associated with the transition to adulthood and helps avoid the phase of heightened risk of death during this period of life.


  • young adult excess mortality
  • transition to adulthood
  • vulnerability
  • Cox model
  • time-varying effects
  • Switzerland

La surmortalité des jeunes adultes en Suisse : quel rôle joue la vulnérabilité socioéconomique ?

Le début de l’âge adulte est souvent caractérisé par une phase de surmortalité. On ignore encore si cette augmentation momentanée du risque de décès est le produit de forces biologiques ou contextuelles, ni si cette menace concerne uniformément tous les individus d’une même cohorte. Grâce à un ensemble de données unique incluant tous les individus vivant en Suisse nés entre 1975 et 1979, des taux différentiels de mortalité par âge de 10 à 34 ans sont calculés. Certains facteurs de risque agissent avec une intensité variable qui suit une évolution similaire à la forme du risque général. Les résultats suggèrent que les facteurs socioéconomiques de mortalité expliquent au moins en partie la surmortalité des jeunes adultes. La division de la cohorte en plusieurs souspopulations représentant différents niveaux de vulnérabilité montre que si une minorité de jeunes Suisses connaît cet excès temporaire de mortalité, ce n’est pas le cas pour tous les groupes d’individus. Dans l’ensemble, les résultats indiquent qu’un contexte social favorable compense le stress associé à la transition vers l’âge adulte et permet d’éviter la phase de risque accrue de décès durant cette période de la vie.


La mortalidad excesiva de los jóvenes adultos en Suiza: ¿qué influencia ejerce la vulnerabilidad económica?

El comienzo de la edad adulta está frecuentemente caracterizado por una fase de mortalidad excesiva. Todavía se ignora si este aumento momentáneo de la mortalidad se debe a factores biológicos o contextuales, ni si esta amenaza concierne uniformemente todos los individuos de una misma cohorte. Gracias a un conjunto de datos que incluye todos los individuos residiendo en Suiza y nacidos entre 1975 et 1979, se calculan las tasas de mortalidad por edad de 10 años a 34 años. Ciertos factores de riesgo actúan con una intensidad variable que sigue una evolución similar a la forma del riesgo general. Los resultados sugieren que los factores socio-económicos de mortalidad explican al menos una parte de la mortalidad excesiva de los jóvenes adultos. La división de las cohortes en varios grupos que representan diferentes niveles de vulnerabilidad, muestra que, si una minoría de jóvenes suizos conoce un exceso temporal de mortalidad, este no es el caso de todos los grupos. En conjunto, los resultados indican que un contexto social favorable compensa el estrés asociado a la transición a la vida adulta y permite evitar la fase de riesgo acrecentado de muerte durante este periodo de la vida.


  • OnlineAmato, P. R., & Keith, B., 1991, “Parental divorce and the well-being of children: A meta-analysis”. Psychological Bulletin, 110(1), p. 26.
  • OnlineBarth R. P., 1990, “On their own: The experiences of youth after foster care”, Child and Adolescent Social Work Journal, 7(5), pp. 419–440.
  • Baumann R., 2012, “Versicherungen schröpfen Ausländer – bis zur Einbürgerung”, Tages Anzeiger, 14 July.
  • OnlineBillari F., Liefbroer A., 2010, “Towards a new pattern of transition to adulthood?”, Advances in Life Course Research, 15(2-3), pp. 59–75.
  • OnlineBlossfeld H.-P., Klijzing E., Mills M., & Kurz K. (eds.), 2005, Globalization, Uncertainty and Youth in Society, London, Routledge.
  • OnlineBopp M., Spoerri A., Zwahlen M. et al., 2009, “Cohort profile: The Swiss National Cohort – A longitudinal study of 6.8 million people”, International Journal of Epidemiology, 38(2), pp. 379–384.
  • OnlineDannefer D., 2003, “Cumulative advantage/disadvantage and the life course: Crossfertilizing age and social science theory”, The Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 58(6), S327–S337.
  • Esping-Andersen G., 2009, Incomplete Revolution: Adapting Welfare States to Women’s New Roles, Cambridge, Polity.
  • Freud A., 1968, “Adolescence”, in Winder A. E., Angus D. (eds.), Adolescence: Contemporary Studies (2nd ed.), New York, American Book.
  • OnlineGiedd J. N., 2004, “Structural magnetic resonance imaging of the adolescent brain”, Annals of the New York Academy of Sciences, 1021(1), pp. 77–85.
  • OnlineGiedd J. N., Blumenthal J., Jeffries N. O. et al., 1999, “Brain development during childhood and adolescence: A longitudinal MRI study”, Nature Neuroscience, 2(10), pp. 861–863.
  • OnlineGoldstein J., 2011, “A secular trend toward earlier male sexual maturity: Evidence from shifting ages of male young adult mortality”, PLOS One, 6(8), e14826.
  • OnlineGompertz B., 1825, “On the nature of the function expressive of the law of human mortality, and on a new mode of determining the value of life contingencies”, Philosophical Transactions of the Royal Society of London, 115, pp. 513–583.
  • OnlineGrambsch P. M., Therneau T. M., 1994, “Proportional hazards tests and diagnostics based on weighted residuals”, Biometrika, 81(3), pp. 515–526.
  • OnlineHall S., 1904, Adolescence: Its Psychology and its Relations to Physiology, Anthropology, Sociology, Sex, Crime, Religion and Education, New York, D. Appleton and Company.
  • OnlineHasselberg M., Vaez M., Lucie L., 2005, “Socioeconomic aspects of the circumstances and consequences of car crashes among young adults”, Social Science & Medicine, 60(2), pp. 287–295.
  • Herbert J., 2015, Testosterone: Sex, Power, and the Will to Win, Oxford, Oxford University Press.
  • OnlineHjern A., Bremberg S., 2002, “Social aetiology of violent deaths in Swedish children and youth”, Journal of Epidemiology and Community Health, 56(9), pp. 688–692.
  • International Labour Organization, 2013, Global Employment Trends for Youth 2013. A Generation at Risk, Geneva, International Labour Office.
  • Joye D., Schuler M., 1995, La structure sociale de la Suisse. Catégories socio-professionnelles, Bern, Federal Statistical Office.
  • OnlineKaufman L., Rousseeuw P. J., 1990, Finding Groups in Data: An Introduction to Cluster Analysis, Vol. 344, New York, Wiley.
  • OnlineKelsey T. W., Li L. Q., Mitchell R. T. et al., 2014, “A validated age-related normative model for male total testosterone shows increasing variance but no decline after age 40 years”, PLOS One, 9(10), e109346.
  • OnlineKerman B., Wildfire J., Barth R. P., 2002, “Outcomes for young adults who experienced foster care”, Children and Youth Services Review, 24(5), pp. 319–344.
  • OnlineKostaki A., 1992, “A nine-parameter version of the Heligman-Pollard formula”, Mathematical Population Studies, 3(4), pp. 277–288.
  • OnlineKuh D., Shlomo Y. B., Ezra S., 2004, A Life Course Approach to Chronic Disease Epidemiology, New York, Oxford University Press.
  • OnlineLenroot R. K., Giedd J. N., 2006, “Brain development in children and adolescents: Insights from anatomical magnetic resonance imaging”, Neuroscience & Biobehavioral Reviews, 30(6), pp. 718–729.
  • OnlineLevitis D. A., 2011, “Before senescence: The evolutionary demography of ontogenesis”, Proceedings of the Royal Society B: Biological Sciences, 278(1707), pp. 801–809.
  • OnlineLiu K., 2009, “Suicide rates in the world: 1950–2004”, Suicide and Life-Threatening Behavior, 39(2), pp. 204–213.
  • OnlineMackenbach J. P., 2012, “The persistence of health inequalities in modern welfare states: The explanation of a paradox”, Social Science & Medicine, 75(4), pp. 761–769.
  • OnlineMarmot M., 2005, “Social determinants of health inequalities”, The Lancet, 365(9464), pp. 1099–1104.
  • OnlinePinyerd B., Zipf W. B., 2005, “Puberty – Timing is everything!”, Journal of Pediatric Nursing, 20(2), pp. 75–82.
  • OnlineSchmidlin K., Clough-Gorr K., Spoerri A. et al., 2013, “Impact of unlinked deaths and coding changes on mortality trends in the Swiss National Cohort”, BMC Medical Informatics and Decision Making, 13(1), 11 p.
  • OnlineSchoenfeld D., 1982, “Partial residuals for the proportional hazards regression model”, Biometrika, 69(1), pp. 239–241.
  • OnlineSpear L. P., 2000, “The adolescent brain and age-related behavioral manifestations”, Neuroscience & Biobehavioral Reviews, 24(4), pp. 417–463.
  • OnlineSpoerri A., Zwahlen M., Egger M., Bopp M., 2010, “The Swiss National Cohort: A unique database for national and international researchers”, International Journal of Public Health, 55(4), pp. 239–242.
  • OnlineSteinberg L., 2005, “Cognitive and affective development in adolescence”, Trends in Cognitive Sciences, 9(2), pp. 69–74.
  • Therneau T. M., Grambsch P. M., 2000, Modeling Survival Data: Extending the Cox Model, New York, London, Springer.
  • OnlineThiébaut A. C. M., Bénichou J., 2004, “Choice of time-scale in Cox’s model analysis of epidemiologic cohort data: A simulation study”, Statistics in Medicine, 23(24), pp. 3803–3820.
  • OnlineThiele T. N., 1871, “On a mathematical formula to express the rate of mortality throughout the whole of life, tested by a series of observations made use of by the Danish life insurance company of 1871”, Journal of the Institute of Actuaries and Assurance Magazine, 16(5), pp. 313–329.
  • OnlineThomson R., Henderson S., Holland J., 2003, “Making the most of what you’ve got? Resources, values and inequalities in young women’s transitions to adulthood”, Educational Review, 55(1), pp. 33–46.
  • Valkonen T., 2006, “Social inequalities in mortality”, in Caselli G., Vallin J., Wunsch G. (eds.), Demography: Analysis and Synthesis, London, Academic Press.
  • OnlineVaupel J. W., Manton K. G., Stallard E., 1979, “The impact of heterogeneity in individual frailty on the dynamics of mortality”, Demography, 16(3), pp. 439–454.
  • OnlineVaupel J. W., Yashin A. I., Manton K. G., 1988, “Debilitation’s aftermath: Stochastic process models of mortality”, Mathematical Population Studies, 1(1), pp. 21–48.
  • OnlineWest P., 1988, “Inequalities? Social class differentials in health in British youth”, Social Science & Medicine, 27(4), pp. 291–296.
  • OnlineWest P., 1997, “Health inequalities in the early years: Is there equalisation in youth?”, Social Science & Medicine, 44(6), pp. 833–858.
  • Wright E. O., 1985, Classes, London, Verso, Acta Academica.
  • Yaouancq F., Duée M., 2014, “Les sans-domicile en 2012: Une grande diversité de situations”, in Guedj, H., Morin T. (coord.), France, portrait social, Paris, INSEE, pp. 123–138.
  • OnlineZufferey J., 2016, “Investigating the migrant mortality advantage at the intersections of social stratification in Switzerland: The role of vulnerability”, Demographic Research, 34(32), pp. 899–926.
  • OnlineZwerling C., Peek-Asa C., Whitten P. S. et al., 2005, “Fatal motor vehicle crashes in rural and urban areas: Decomposing rates into contributing factors”, Injury Prevention, 11(1), pp. 24–28.
Adrien Remund
University of Geneva, INED, University of Groningen
Correspondence: Adrien Remund, University of Geneva, Institut de démographie et socioéconomie, Swiss National Centre of Competence in Research LIVES Overcoming Vulnerability: Life Course Perspectives, 40 boulevard du Pont-d’Arve, 1211 Geneva 4, Switzerland, tel: (41) 22 379 89 23.
Uploaded on on 29/01/2019
Distribution électronique pour I.N.E.D © I.N.E.D. Tous droits réservés pour tous pays. Il est interdit, sauf accord préalable et écrit de l’éditeur, de reproduire (notamment par photocopie) partiellement ou totalement le présent article, de le stocker dans une banque de données ou de le communiquer au public sous quelque forme et de quelque manière que ce soit.
Loading... Please wait