Introduction
The expression of suicide ideation has been longitudinally associated with completed suicide both in psychiatric and general population samples (Hubers et al., Reference Hubers, Moaddine, Peersmann, Stijnen, Van Duijn, Van der Mast, Dekkers and Giltay2016). Suicide ideation is higher among people younger than 50 years in Europe (Bernal et al., Reference Bernal, Haro, Bernert, Brugha, de Graaf, Bruffaerts, Lépine, de Girolamo, Vilagut and Gasquet2007). However, people aged 50 years and older have a higher prevalence of completing suicide (Eurostat, 2020). In addition, suicide ideation in middle and old age is associated with functioning problems (Briggs et al., Reference Briggs, Tobin, Kenny and Kennelly2018), loneliness, poor social support (Almeida et al., Reference Almeida, Draper, Snowdon, Lautenschlager, Pirkis, Byrne, Sim, Stocks, Flicker and Pfaff2012), and most importantly with an amount of psychological suffering that can be managed by existing treatments (O'Riley et al., Reference O'Riley, Van Orden, Conwell, Pachana and Laidlaw2014).
Although there are well-known risk factors for suicide ideation among middle and old age people, the role of ageing in this group is still unclear. Some studies have found that people younger than 65 years show higher suicide ideation than people older than 65 years (Cabello et al., Reference Cabello, Miret, Ayuso-Mateos, Caballero, Chatterji, Tobiasz-Adamczyk, Haro, Koskinen, Leonardi and Borges2019), whereas other studies have reported that suicide ideation increased with age among people older than 70 years (Stolz et al., Reference Stolz, Fux, Mayerl, Rásky and Freidl2016; Fässberg et al., Reference Fässberg, Vanaelst, Jonson, Sterner, Ahlner, Wetterberg, Rydén, Kern, Sigström and Zettergren2020). These results might suggest a U-Shaped relationship between age and suicide ideation, increasing suicide ideation in the middle-aged, decreasing in the young-old and increasing again over the oldest-old. However, this available literature has several limitations that should be considered. First, the use of different measurements. Some measures include thoughts that life is not worth living (passive suicide ideation) (Stolz et al., Reference Stolz, Fux, Mayerl, Rásky and Freidl2016; Briggs et al., Reference Briggs, Tobin, Kenny and Kennelly2018), other instruments assess thoughts of taking own life (Cabello et al., Reference Cabello, Miret, Ayuso-Mateos, Caballero, Chatterji, Tobiasz-Adamczyk, Haro, Koskinen, Leonardi and Borges2019), and finally others mix both concepts (Almeida et al., Reference Almeida, Draper, Snowdon, Lautenschlager, Pirkis, Byrne, Sim, Stocks, Flicker and Pfaff2012). Secondly, the inclusion of samples with different age groups. Some studies include age-heterogeneous samples (Briggs et al., Reference Briggs, Tobin, Kenny and Kennelly2018; Cabello et al., Reference Cabello, Miret, Ayuso-Mateos, Caballero, Chatterji, Tobiasz-Adamczyk, Haro, Koskinen, Leonardi and Borges2019), whereas others only include older people (Almeida et al., Reference Almeida, Draper, Snowdon, Lautenschlager, Pirkis, Byrne, Sim, Stocks, Flicker and Pfaff2012; Fässberg et al., Reference Fässberg, Vanaelst, Jonson, Sterner, Ahlner, Wetterberg, Rydén, Kern, Sigström and Zettergren2020). Finally, a third potential factor is probably related to the lack of distinction among age, cohort and period effects in most of the existing studies. Cohort effects refer to the existence of common cultural and societal values that are shared by people born in similar years. Period effects, in turn, are the environmental factors that are related to living in specific years and that affect all population age groups (for example, an economic recession or medical and technological advances). Cohort and period effects have been largely analysed for depression and suicide deaths (Phillips, Reference Phillips2014; Sullivan et al., Reference Sullivan, Liu, Dodge, Andreescu, Chang and Ganguli2020). However, only one study to our knowledge has identified that suicide ideation trends in the last two decades could be partially attributed to cohort and period effects (Twenge et al., Reference Twenge, Cooper, Joiner, Duffy and Binau2019). In this study, people born after the 1980s reported higher suicide ideation rates than previous generations. In addition, an increasing suicide ideation trend was observed since 2015 across all age groups. Although this study is valuable, it did not include a large proportion of older adults, and it was limited to the US population.
Therefore, the present study is aimed at filling the existing gap in literature, by investigating longitudinally whether suicidal ideation trends in European ageing populations might respond to, age, time period, or birth cohort effects.
Besides clarifying previously mentioned inconsistent results, disentangling the role of age, period, and cohort effects on the prevalence of suicide would hold a number of different implications for suicide prevention. Firstly, given that approximately 35% of adults with suicide ideation continue to report suicide ideation 10 years later (Borges et al., Reference Borges, Angst, Nock, Ruscio and Kessler2008), the existence of a possible cohort birth effect among younger generations may help to predict a greater need for elderly mental health care as these generations age. On the other hand, an increasing prevalence of suicide ideation in the last few years due to a period effect might reveal that effective interventions should be launched to stem the increasing ‘epidemic’ of suicide ideation affecting all age groups. Finally, if the change is mainly due to the ageing process and the prevalence of suicide ideation lessens as the person ages, suicide ideation prevention efforts might be particularly indicated for specific lifetime periods.
Methods
Sample and study design
The Survey of Health, Ageing and Retirement in Europe (SHARE) is a cross-national panel survey including European community-dwelling population aged 50 and older (Börsch-Supan et al., Reference Börsch-Supan, Brandt, Hunkler, Kneip, Korbmacher, Malter, Schaan, Stuck and Zuber2013). A total of 50 782 participants from 20 European countries that participated at least once from Wave 1 to Wave 5 were included. Wave 1 was conducted in 2004–2005. Wave 2 was assessed during 2006–2007. Wave 4 information was mainly collected in 2011 and Wave 5 was carried out in 2013. Data from Wave 3 were excluded as they did not include any suicide ideation measure. Household response rates in baseline ranged from 40.3% (Belgium) to 97.5% (France) with an overall individual response rate of 60.1%. Household response rates by Wave and country can be consulted in supplementary material (online Supplementary Table S1). Wave-to-wave retention rates of Wave 1 participants was higher than 55% in all the countries (Bergmann et al., Reference Bergmann, Kneip, De Luca and Scherpenzeel2017). Data collection was conducted in the respondent's house using a computer-assisted personal interview. All interviewers were systematically trained by local group leaders who received a common train-the-trainer workshop. The study protocol was originally built in English and then translated using a forward-backward methodology. Special efforts were conducted to assure all the translations kept functional equivalence both in concepts and phrases used across countries. Further information on the study protocol development can be consulted elsewhere (Börsch-Supan et al., Reference Börsch-Supan, Brugiavini, Jürges, Kapteyn, Mackenbach, Siegrist and Weber2008). The study methods were approved by the University of Mannheim's internal review board until 2011 and by the Ethics Council of the Max-Planck-Society for the Advancement of Science since 2011. Further details of data collection, sampling methods and study design used are explained elsewhere for Wave 1 (Börsch-Supan et al., Reference Börsch-Supan, Hank and Jürges2005), Wave 2 (Börsch-Supan et al., Reference Börsch-Supan, Brugiavini, Jürges, Kapteyn, Mackenbach, Siegrist and Weber2008), Wave 4 (Malter and Börsch-Supan, Reference Malter and Börsch-Supan2013) and Wave 5 (Malter and Börsch-Supan, Reference Malter and Börsch-Supan2015).
Measures
Suicidal ideation
A binary item from the EURO-Depression scale (Prince et al., Reference Prince, Reischies, Beekman, Fuhrer, Jonker, Kivela, Lawlor, Lobo, Magnusson and Fichter1999) was used. The question included ‘In the last month, have you felt that you would rather be dead?’. Responses were coded as positive if any mention of wishing to be dead or suicidal feelings were reported and negative if there were no such feelings.
Health status
A common latent health score based on self-reported health items related to impairments in body functions, limitations in Activities of Daily Living and Instrumental Activities of Daily Living, and measured tests of cognitive performance and walking speed was created. This variable allowed us to control the effect that the general level of health might have on the presence of suicide ideation over time. Different statistical techniques (including Factor Analysis, Bayesian multilevel Item Response Theory and Machine Learning methods) were employed. The metric has shown a good performance in terms of predictive ability of mortality (de la Fuente et al., Reference de la Fuente, Caballero, Sánchez-Niubó, Panagiotakos, Prina, Arndt, Haro, Chatterji and Ayuso-Mateos2018) and morbidity (Caballero et al., Reference Caballero, Soulis, Engchuan, Sanchez-Niubo, Arndt, Ayuso-Mateos, Haro, Chatterji and Panagiotakos2017). Values ranged from 0 to 100, where higher values reflected a higher health status. Further information on this variable can be consulted elsewhere (Caballero et al., Reference Caballero, Soulis, Engchuan, Sanchez-Niubo, Arndt, Ayuso-Mateos, Haro, Chatterji and Panagiotakos2017).
Country
A total of 20 countries were included. Not all the countries participated in all the study assessments. There were some countries such as the Netherlands, Spain, Belgium, the Czech Republic, France, Germany, Italy, Sweden and Switzerland that participated in all the study waves, whereas others such as Hungary or Portugal only participated in Wave 4. The number of participants by wave in each country can be consulted in the supplementary material (online Supplementary Table S2).
Covariates
Age which was grouped into three categories (50–64, 65–79 and 80 and older), gender and level of education, divided into four categories (less than primary education, primary, secondary, and tertiary or higher education) and marital status grouped into four categories (single, married-cohabiting, separated-divorced, and widow) were also included.
Analyses
Descriptive characteristics of the sample were described, using mean and standard deviation for continuous variables, and frequencies and percentages for categorical variables. Chi-squared tests were used in order to assess the association between categorical variables and the presence of suicidal ideation in Wave 1. Similarly, the Wilcoxon test was used for estimating differences between suicide ideation groups in continuous variables.
The measurement occasions were compilated as period effects including the following four groups of years: 2004–2005, 2006–2007, 2011 and 2013. Over this time period, a total of 15 consecutive birth cohorts were made every 4 years, beginning with those born in 1905–1908 and ending with those born in 1961–1964. We selected cohorts based on 4 years period to ensure a homogeneous and sufficient number in all groups. Prevalence rates of suicide ideation were calculated by birth cohort and period groups and graphically presented.
Next, to better disentangle the effects of age, time period, and birth cohort, we performed a mixed-effect model. Suicide ideation was recorded in SHARE as a dichotomous variable, for this reason a logistic approach was applied. The proposed model was a logistic mixed model of repeated measures which could be described as follows:
with $participant\;\sim N\lpar {0\comma \;\sigma_p^2 } \rpar$
, $birth\;cohort\;\sim N\lpar {0\comma \;\sigma_{bc}^2 } \rpar$
and $wave\;\sim N\lpar {0\comma \;\sigma_{wave}^2 } \rpar$
as random effects
Therefore, the random part of the model included three levels: the person, the birth cohort and the time period. Random effects at a personal level were added in order to capture variability due to intrinsic participant characteristics. Cohort birth was added as a random effect to explain the existence of cultural and societal changes that make people born in specific years more vulnerable to suicide ideation. Finally, the period effect was also included as random to account for variability due to the specific circumstances that all participants experienced over the study time. The identification problem, which is the linear dependence between age, period and cohort effects was managed according to Reither's indications (2015). We decided to include periods and cohorts as random since they are contextual variables, that are independent of individual-level variables. These random effects followed a normal distribution with a mean of zero and variance to be estimated, the larger the variance, the higher the impact in the observed variability. The fixed part of the model included sex, educational level, age, health status and country as fixed effects. Age was treated as fixed because it was considered as a biological factor that is related to individual-level characteristics and whose effects would be constant across individuals.
After running the Restricted Maximum Likelihood Model, estimated coefficients for fixed effects were transformed into odds ratios (ORs), with variance components estimations providing an idea of influence in variability since total variance can be described as follows, $\sigma _{Total}^2 = \sigma _p^2 + \sigma _{cb}^2 + \sigma _{wave}^2 + 1$
.
Finally, considering the results of the previous modelled relationships, we simulated the posterior distributions of birth and periods effects on suicide ideation. Predicted OR and confidence intervals (95%) were generated from 500 simulated distributions, estimating the values of suicide ideation for each cohort and period level. Results were graphically presented.
All statistical analyses were done using R (R Core Team, 2013). All graphs were created using ggplot2 package (Wickham, Reference Wickham2016). Mixed-effect model was fitted using the glmer function in the lme4 package (Bates et al., Reference Bates, Mächler, Bolker and Walker2015). Simulations of random effects were calculated using the REsim function in the merTools package (Knowles and Frederick, Reference Knowles and Frederick2016). P values lower than 0.05 were considered significant. All the analyses were conducted with the available information with no data imputation.
Results
The overall sample consisted of 50 782 participants that were interviewed at least once (n = 22 621; 44.5% were men). Descriptive statistics by the presence of suicide ideation are provided in Table 1. A total of 4031 participants (7.94%) showed suicide ideation the first time they were interviewed. Prevalence of suicide ideation across gender, age, level of education, country groups and health status can be consulted in Table 1. Unadjusted prevalence rates of suicide ideation were descriptively higher for cohorts born during the first two decades of the 20th century than the cohorts born since 1941 (Fig. 1). Two cohorts 1909–1912 and 1913–1916, reported descriptively an increasing trend of suicide ideation in the year 2013 (Fig. 1).
Table 1. Characteristics of the SHARE participants interviewed in the baseline
Notes: Baseline year varied depending on the country. Baseline assessment in Austria, Belgium, Switzerland, Germany, Denmark, Spain, France, Greece, Italy, Netherlands, Sweden, Israel was in 2004. It was in 2007 in the Czech Republic, Ireland and Poland. It was in 2011 in Estonia, Hungary, Portugal, Slovenia and it was Luxembourg in 2013
Fig. 1. Unadjusted prevalence of suicide ideation by Birth Cohort groups over study time.
Notes: Wave 1 (2004–2005), Wave 2 (2006–2007), Wave 4 (2011) and Wave 5 (2013). Birth cohort is denoted by the first year of each cohort.
The results of the mixed-effect regression model are presented in Table 2. As can be seen in the fixed part of the model, youngest-old people (65–79 years) showed a lower likelihood of suicide ideation (OR = 0.88; CI 95% = 0.81, 0.96) than middle-aged people (50–64 years) (Table 2). In contrast, people aged 80 + years reported similar levels of suicide ideation than middle-aged people (OR = 0.95; CI 95% = 0.91, 1.01). In addition, higher suicide ideation was related to being a woman, having a lower level of education and lower health status. Population with married or cohabiting marital status reported lower suicide ideation than singles (OR = 0.72; CI 95% = 0.65, 0.79). Being widowed (OR = 1.21; CI 95% = 1.09, 1.35) and separated/divorced (OR = 1.27; CI 95% = 1.13, 1.42) were associated with higher odds of suicide ideation than being single. Significant differences in suicide ideation estimates were also found across countries (Table 2). Regarding the random part of the model, most of the variation in suicide ideation was explained by inter-participant characteristics, 57.8% of the total variance (σ 2 = 1.40; CI 95% = 1.24–1.55). Birth-cohort random effect explained a smaller percentage 0.85% but still had an effect on the variability in suicide ideation (σ 2 = 0.02; CI 95% = 0.0019–0.0023), whereas the time period level explained only 0.13% of the variation (σ 2 = 0.003; CI 95% = 0.001–0.006) (Table 2). Predicted estimates for suicide ideation across birth cohort groups are shown in Fig. 2. People born in 1941–1944 (OR = 0.89; CI 95% = 0.80–0.99), showed a decreased likelihood of reporting suicide ideation (Fig. 2). In addition, people born between 1953 and 1955 (OR = 1.17; CI 95% = 1.05–1.31), 1956–1960 (OR = 1.21; CI 95% = 1.08–1.35) and 1961–1964 (OR = 1.32; CI 95% = 1.16–1.49) showed a higher likelihood of reporting suicide ideation. Predicted intervals and values of time period effect are displayed in Fig. 3. In comparison with other time periods, participants assessed during the 2006–2007 period reported a decreased likelihood of suicide ideation (OR = 093; CI 95% = 0.87–0.99). The rest of the birth cohorts and period levels did not show any significant effect on suicide ideation.
Table 2. Parameter estimates of the multilevel mixed-effects models assessing the prevalence of suicide ideation in the SHARE participants
Fig. 2. Predicted ORs and 95% CIs of suicide ideation for each birth cohort.
Note: Births cohorts are denoted by the first year of each cohort.
Fig. 3. Predicted ORs and 95% CIs of suicide ideation for each time period.
Notes: Wave 1 covered the 2004–2005 period; Wave 2 was conducted in the 2006–2007 period; Wave 4 was in 2011 and Wave 5 was in 2013.
Discussion
This study is, to our knowledge, the first to disentangle age, cohort and period effects on the prevalence of suicide ideation among people aged 50 years and older in Europe, shedding light on the ongoing debate about the role of ageing in suicide ideation. To do so, we analysed the prevalence trends of suicide ideation, including the age effect and the variability effects of birth cohorts from 1905 to 1961, and of four time periods between 2004 and 2013. In this study, the main effects of suicide ideation prevalence were primarily due to individual-level factors. In comparison with youngest-old (65–79 years), middle-aged (50–64 years) reported higher suicide ideation, regardless of period and cohort effects. Middle-aged might be a period associated with economic, health and social burdens that could rise the prevalence of suicidal ideation (Jo et al., Reference Jo, Jeon and Oh2017). In contrast, no differences were found between people aged 80 years and older and those in middle-aged. Oldest-old people might more likely have to face with difficulties such as physical diseases and bereavement than youngest-old people (Koo et al., Reference Koo, Kõlves and De Leo2017), which may increase their likelihood of suicide ideation. Our results support that the relationship between age and suicide ideation may not be linear in elderly people, and suggest that different age groups should be separately analysed in the study of suicide ideation in older people.
In line with previous literature, married reported a lower likelihood of suicide ideation than single people (Smith et al., Reference Smith, Mercy and Conn1988). In contrast, being single was associated with a lower level of suicidal ideation than being widowed and separated-divorced.
Our study has also revealed that there was a small but still relevant cohort effect in the prevalence of suicide ideation. Decreasing trends of suicide ideation were observed since the earliest 20s, with cohorts born between 1941 and 1944 reporting the lowest likelihood of suicide ideation. This result is congruent with other studies that have shown that people born just before and during World War II were the generation with the lowest prevalence of suicide deaths and depression (Phillips, Reference Phillips2014; Sullivan et al., Reference Sullivan, Liu, Dodge, Andreescu, Chang and Ganguli2020). Some authors argue that people from this generation could have learned to adapt personal aspirations to persevere through difficult times and be more resilient to frustrations (Phillips, Reference Phillips2014). Other authors have also supported that selective mortality could have largely affected this generation, i.e. those with health problems died younger and hence they are not presented in middle and old aged population studies (Hasin and Link, Reference Hasin and Link1988). Further studies should explore the reasons why so many people from this generation have shown low levels of suicide ideation.
On the other hand, an increasing trend of suicide ideation was observed for post-War generations (called baby boomers), reaching the significant level for people born from 1953 to 1964. This finding is also congruent with previous studies that report that baby boomers are showing higher psychopathology than their ancestors (Phillips, Reference Phillips2014). According to Durkheim (Reference Durkheim2005) societal and cultural changes that occurred from the post-war generations onwards might have weakened societal bonds, making these generational groups more vulnerable to suicide. Other authors have added that baby boomers were the first ones who imposed a youth culture where ageing was marginalised. Therefore, it is possible that some of them became victims of the culture they created themselves (McCue and Balasubramaniam, Reference McCue and Balasubramaniam2016). Although further studies are necessary, our findings suggest that baby-boomers are more affected by suicide ideation. Therefore, there will probably be a higher need for elderly mental health and social community support in Europe in the future since this generation is already entering the old age.
Finally, our study also showed that a very small amount of suicide ideation changes might be attributed to period effects. A small period effect was observed for participants assessed during the 2006–2007 years, who reported a lower prevalence of suicide ideation in comparison with other time occasions. Some authors have highlighted the optimism experienced during the 2006–2007 years in the European Union due to the inclusion of new countries and to good economic stability which met or even exceeded expectations (Schrogl et al., Reference Schrogl, Mathieu and Peter2008). Conversely, no increases of suicide ideation were observed during the economic downturn period in the 2011 and 2013 years. Although we do not have a clear interpretation for this result, it is possible that our cross-country approach blurred some differences, since some European countries were greatly affected by the economic recession, whereas others were not. In addition, the effect of a selection bias cannot be ruled out either, since people may be less likely to participate in a study if they were affected by the events.
The great variety of countries covered, the use of a longitudinal panel design – which allowed us to simultaneously manage age, period and cohort effects – controlling for important effects such as general health or country and the use of a measure that recorded the presence of suicide ideation in the last month – which reduced the likelihood of recall bias effects – are probably the main strengths of this study. However, we acknowledge that the results of this study should also be interpreted with caution due to the following limitations. Firstly, although European countries share common cultural and societal values, it is possible that the aim of searching a global European birth and period cohort effect may have hidden the variability that each birth cohort and period group had in the different countries. Unfortunately, our data did not allow specific country analyses due to the variability of time periods that each country included in the study. Secondly, our study did not include information of severity and covered a concept of passive suicide ideation, which has sometimes been considered as a less severe form of suicide ideation in clinical settings. However, wishes of death or that life is not worth living have been strongly associated with psychopathology and they did not seem to be normative in older populations (Van Orden et al., Reference Van Orden, O´Riley, Simning, Podgorski, Richardson and Conwell2015). Thirdly, the use of a self-reported measure might have introduced biases related to underreporting wishes of death due to social desirability. Fourthly, household response rates were globally higher than 45% in all the study waves. However, there were some countries with rates lower than 40%. This fact has to be carefully considered in the interpretation of the study results. Nonetheless, it is important also to consider that response rates in the SHARE study are in line with the numbers of comparable European surveys conducted in the same period (Koen et al., Reference Koen, Loosveldt, Vandenplas and Stoop2018). Fifthly, the method on how to best deal with point identification problem is still an open debate in the literature with not a single answer (O'Brien, Reference O'Brien2017). We acknowledge there are other approaches based on mathematical solutions that could have been applied (Tu et al., Reference Tu, Krämer and Leec2012). However, we managed the point identification problem according to explicit theory-based assumptions that have also been shown as valid (Reither et al., Reference Reither, Masters, Yang, Powers, Zheng and Land2015). Moreover, we categorised age, periods and cohorts into unequal width groups in order to also decrease the possible linear dependence between age, period and cohort effects. Finally, although the study included a follow-up period of 8 years (2004–2013), middle ages could be underrepresented for earlier generations.
In spite of all these limitations, our study is the first to analyse the role of ageing including the age, cohort and period effects on the prevalence of suicide ideation in the European population. The findings in our study support that the relationship between age and suicide ideation is non-linear. Middle-age may be a particularly high-risk time period of suicide ideation in comparison with early elderly (65–78 years). However, the protective effect of older age might not apply for people age 80+. Our results have also suggested that there is a certain cohort effect, with baby boomers being more likely to report a wish to die than the generation born just during War World II, showing this later generation the lowest levels of suicide ideation. Thus, it is possible that a greater need for community mental health services by older European people could follow in the near future.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S2045796021000020
Acknowledgements
The authors gratefully acknowledge the participation of the interviewed participants, without whom this study would not have been possible.
Financial support
This work was supported by the Ageing Trajectories of Health: Longitudinal Opportunities and Synergies (ATHLOS) project which was funded by the European Union's Horizon 2020 (No. 635316). The present work also received support from the Instituto Carlos III-FIS (project PI16/00218) and by The European Union Regional Development Fund (ERDF) ‘A Way to Build Europe’ (project PI16/00218). The first four Waves of the SHARE study have been mainly funded by the European Commission by the 5th framework programme (project QLK6-CT-2001-00360 in the thematic programme Quality of Life)), the 6th framework programme (projects SHARE-I3, RII-CT-2006-062193, COMPARE, CIT5-CT-2005-028857, and SHARE-LIFE (CIT4-CT-2006-028812)) and by the 7th framework programme, projects SHARE-PREP (No 211909), SHARE-LEAP (No 227822) and M4 (No 261982). Additional funding from the German Ministry of Education and Research, the Max Planck Society for the Advancement of Science, the US National Institute on Aging (U01_AG09740-13S2, P01_AG005842, P01_AG08291, P30_AG12815, R21_AG025169, Y1-AG-4553-01, IAG_BSR06-11, OGHA_04-064, HHSN271201300071C), and from various national funding sources is gratefully acknowledged (see www.share-project.org).
Conflict of interest
None declared.
Ethical standards
The authors assert that all procedures contributing to this work meet the ethical standards of the relevant national and institutional committees on human experimentation and the Helsinki Declaration of 1975, as revised in 2013.
Availability of Data and Materials
This paper uses data from SHARE Waves 1, 2, 4, and 5 (DOIs:10.6103/SHARE.w1.700, 10.6103/SHARE.w2.700, 10.6103/SHARE.w4.700, and 10.6103/SHARE.w5.700), which are available via the SHARE project's website.
Open access
