You are here

Article Text

Article menu

Download PDFPDF

Original research
Widening educational inequalities in mortality in more recent birth cohorts: a study of 14 European countries
  1. Di Long1,
  2. Johan P Mackenbach1,
  3. Silvia Klokgieters2,
  4. Ramunė Kalėdienė3,
  5. Patrick Deboosere4,
  6. Pekka Martikainen5,
  7. Kristian Heggebø6,7,
  8. Mall Leinsalu8,9,
  9. Matthias Bopp10,
  10. Henrik Brønnum-Hansen11,
  11. Giuseppe Costa12,
  12. Terje Eikemo6,
  13. Wilma J Nusselder1
  1. 1 Department of Public Health, Erasmus MC, Rotterdam, Netherlands
  2. 2 Department of Sociology, Vrije Universiteit Amsterdam, Amsterdam, Netherlands
  3. 3 Department of Health Management, Lithuanian University of Health Sciences, Kaunas, Lithuania
  4. 4 Department of Sociology, Vrije Universiteit Brussel, Brussel, Belgium
  5. 5 Population Research Unit, Faculty of Social Sciences, University of Helsinki, Helsinki, Finland
  6. 6 Department of Sociology and Political Science, Norwegian University of Science and Technology, Trondheim, Norway
  7. 7 NOVA, Oslo Metropolitan University, Oslo, Norway
  8. 8 Stockholm Centre for Health and Social Change, Södertörn University, Huddinge, Sweden
  9. 9 Department of Epidemiology and Biostatistics, National Institute for Health Development, Tallinn, Estonia
  10. 10 Epidemiology, Biostatistics and Prevention Institute, University of Zurich, Zurich, Switzerland
  11. 11 Department of Public Health, University of Copenhagen, Copenhagen, Denmark
  12. 12 Department of Clinical Medicine and Biology, University of Turin, Torino, Italy
  1. Correspondence to Di Long, Public Health, Erasmus MC, Rotterdam, Netherlands; d.long{at}erasmusmc.nl

Abstract

Background Studies of period changes in educational inequalities in mortality have shown important changes over time. It is unknown whether a birth cohort perspective paints the same picture. We compared changes in inequalities in mortality between a period and cohort perspective and explored mortality trends among low-educated and high-educated birth cohorts.

Data and methods In 14 European countries, we collected and harmonised all-cause and cause-specific mortality data by education for adults aged 30–79 years in the period 1971–2015. Data reordered by birth cohort cover persons born between 1902 and 1976. Using direct standardisation, we calculated comparative mortality figures and resulting absolute and relative inequalities in mortality between low educated and high educated by birth cohort, sex and period.

Results Using a period perspective, absolute educational inequalities in mortality were generally stable or declining, and relative inequalities were mostly increasing. Using a cohort perspective, both absolute and relative inequalities increased in recent birth cohorts in several countries, especially among women. Mortality generally decreased across successive birth cohorts among the high educated, driven by mortality decreases from all causes, with the strongest reductions for cardiovascular disease mortality. Among the low educated, mortality stabilised or increased in cohorts born since the 1930s in particular for mortality from cardiovascular diseases, lung cancer, chronic obstructive pulmonary disease and alcohol-related causes.

Conclusions Trends in mortality inequalities by birth cohort are less favourable than by calendar period. In many European countries, trends among more recently born generations are worrying. If current trends among younger birth cohorts persist, educational inequalities in mortality may further widen.

  • INEQUALITIES
  • MORTALITY
  • DEATH

Data availability statement

Data are available on reasonable request.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

https://doi.org/10.1136/jech-2023-220342

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    WHAT IS ALREADY KNOWN ON THIS TOPIC

    • Previous studies that have examined educational inequalities in mortality used a period perspective and found that relative inequalities have increased over time in almost all European countries, whereas absolute inequalities have followed a more variable course.

    WHAT THIS STUDY ADDS

    • This study provides insight into the differences in the trends in mortality inequalities viewed from a cohort and period perspective.

    • This study provides insight into cohort differences in mortality inequalities by investigating both all-cause and cause-specific mortality.

    • The study provides insight into cross-national, sex and age differences in educational trends in mortality across 14 European countries.

    HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

    • More awareness is needed that the period perspective may yield too optimistic results.

    • Practice and policy should be prepared that inequalities in mortality may further widen and that ‘unhealthy homogenisation’ may create an increasingly vulnerable group of lower-educated persons.

    Introduction

    Socioeconomic inequalities in mortality have been extensively analysed. Over the past decades, relative inequalities in mortality have increased in almost all European countries, whereas absolute inequalities have followed a more variable course.1–3

    Most previous studies on health inequalities have adopted a period perspective, in which the unit of time is calendar year. This is useful for monitoring purposes, for instance, to evaluate whether policy targets for reducing health inequalities have been met or to detect interruptions in trends due to major historical events such as the financial crisis in 2008.4 However, not all changes occurring over time are due to true period effects. Instead, changes may also be due to cohort effects, for example, changes between successive birth cohorts in exposure to health determinants during the life course.5 6

    Cohort effects have been found for several health outcomes including lung cancer7 and cardiovascular diseases.8 In addition, cohort effects have been observed for various health determinants including smoking,9 alcohol use10 and physical activity.11 In turn, these developments may have a consequence for trends in educational inequalities in health. While there are several studies focusing on cohort trends in educational inequalities in different health outcomes suggesting growing educational differences in more recent birth cohorts,12–17 most studies were based on a small number of deaths or focused on one single country. Therefore, the generalisability of these findings might be limited to specific national and historical contexts, thus studies covering both a wider geographical area and a larger number of cases are needed.

    Determining whether health inequalities are affected by cohort effects is important for two other reasons. First, it may help to better understand long-term trends occurring over time. For example, changes over time in inequalities in smoking-related mortality are at least partly determined by widening or narrowing inequalities in smoking habits across successive generations.18 Second, it may help to better understand how educational inequalities in mortality might develop in the future. Strong cohort effects in health outcomes indicate that successive birth cohorts may differ importantly in exposure to the determinants of mortality. If we observe more recently born birth cohorts to have larger inequalities in mortality (at the same ages) than previous generations, we can reasonably hypothesise that when they grow older, they might continue to have larger inequalities in mortality. This is particularly true for risk factors such as smoking which have a long-lasting effect on health. Extrapolations into the future based on cohort patterns may, therefore, usefully supplement extrapolations based on period effects only.19

    In this study, we, therefore, aimed to (1) explicitly compare inequalities in all-cause mortality between a period and cohort perspective and (2) explore underlying patterns in all-cause and cause-specific mortality among low-educated and high-educated birth cohorts, exploiting a unique dataset with harmonised information on mortality by the level of education, sex, age and cause of death covering several decades and 14 European countries.

    Data and methods

    Data

    We collected and harmonised data on all-cause and cause-specific mortality by sex, age (in 5-year age groups from 30–34 to 75–79 years), calendar period (various periods between 1971 and 2015) and educational level from 14 European countries. Most data relate to complete national populations, with the exceptions of Italy (Turin only), Spain (Barcelona only), France (a 1% sample of the population), the UK (England and Wales only, and a 1% sample of the population) and Switzerland (Swiss nationals only). Most data are derived from longitudinal mortality follow-ups after a census or based on registries (online appendix 1 table 1). The results are grouped geographically, (3–4 countries per figure) to make the figures easier to read: Northern Europe (Denmark, Finland, Norway and Sweden), Southern Europe (France, Italy and Spain), Western Europe (Austria, Belgium, the UK and Switzerland) and Eastern Europe (Estonia, Hungary and Lithuania).

    Supplemental material

    Socioeconomic position (SEP) was indicated by the highest level of completed education: ‘low’, ‘mid’ and ‘high’ corresponding to International Standard Classification of Education (ISCED) 2011 categories 0–2, 3–4 and 5–8. In the UK, only two groups could be distinguished, namely ‘low’ and ‘high’. Data with missing level of education were excluded. We refer to online appendix 1 tables 2–3 for more information on education. We focused on educational inequalities in mortality because comparable data on educational attainment were available for both men and women in most European populations. In addition, education is the most stable measure of SEP, as it is normally completed early in adulthood,20 and it is not sensitive to fluctuations in economic conditions.21

    Birth cohorts were reconstructed using the information on age and period (birth cohort equals period minus age). Birth cohorts were constructed to retain as much detail as possible. Because data were originally collected in 5-year age groups and 1–5 years calendar periods, the birth year intervals of the reconstructed birth cohorts ranged between 6 and 10 years, with overlaps. For example, those of age 30–34 who died during 1980–1984 belong to birth cohort 1946–1954, whereas those of age 35–39 who died during 1980–1984 belong to birth cohort 1941–1949.

    To reduce random fluctuations in the number of deaths across birth cohorts, while retaining as many birth cohorts as possible, we excluded combinations of country–sex–age–education that had less than 10 all-cause deaths. We further excluded birth cohorts with data for less than three age groups. In total, 18 birth cohorts containing 595 169 966 person-years were included in the analysis, with birth years ranging from ca. 1902–1910 to ca. 1967–1975 (online appendix 2 table 5). We presented the results by the median birth year of each birth cohort. To allow comparability, the same selection of study populations was used for the period perspective.

    In addition to all-cause mortality, we also analysed mortality by four causes-of-death groups (namely neoplasms, cardiovascular diseases, external causes and all other causes of death) and eight specific causes of death (online appendix 3 table 6).

    Methods

    Comparative mortality figures (CMF) were used to measure mortality difference between calendar periods (CMF) and between birth cohorts (cohort CMF, CCMF), based on direct standardisation22 (details in online appendix 4).

    Simply put, CMFs are calculated for each combination of country, sex, level of education and cause of death as follows:

    Embedded Image

    and

    Embedded Image

    As standard population, we used pooled population of all countries, all educational groups and all calendar periods, separate for men and women. Because mortality rates are on average lower among women than men, using the same standard population for both sexes would yield hard-to-interpret comparisons between educational groups (online appendix 7 figures 3–4).

    The CMF and CCMF are index figures, representing differences in mortality between a particular calendar period/birth cohort and the overall average on a relative, dimensionless scale. The approach chosen here is particularly suited for the analyses of this paper because our data had different age ranges and compositions between birth cohorts and this approach can work around it. CCMF has been used in prior research on geographical comparisons.23–25

    Relative inequalities (rate ratio, RR) between the low educated and high educated were calculated as the ratio of their corresponding mortality figures, CMF (period perspective) and CCMF (cohort perspective). For the cohort perspective, the RRs were further adjusted for the fact that more recent birth cohorts included younger age groups than earlier birth cohorts. This could cause an overestimation of inequalities in recent birth cohorts as relative inequalities in mortality are substantially larger in younger than in older age groups.26 This adjustment followed two steps (details in online appendix 5). In the first step, we calculated an expected RR in the standard population based on the age range in each birth cohort. In the second step, we divided each birth cohort’s observed RR by the expected RR and multiplied the result by the all ages RR in the standard population to obtain the adjusted RRs as presented in this paper.

    Absolute inequalities (rate difference) between the low educated and high educated were also calculated. First, we calculated an age-standardised mortality rate by sex in the standard population. Next, we multiplied these standardised rates with CMFs/CCMFs to obtain absolute mortality rates for the low educated and high educated groups. Rate difference is then calculated by subtracting the absolute rates of the high educated from the low educated.

    The 95% CIs were calculated using parametric bootstrapping with Poisson-distributed death counts and 1000 repetitions. Analyses were performed in Stata V.16.0 SE.

    Results

    Descriptive results

    Table 1 and online appendix 1 table 4 present descriptive information about the data included for analysis. For several countries, the analysis covered a wide range of birth years, ranging from 1900s to 1970s. The total number of deaths included in the analysis is 5.9 million. As the analysis covered many older birth cohorts, the proportion of low educated among all deaths was often substantial. Note that the extraordinarily high percentage of low educated in the UK was because the ‘low’ educated in this country also includes those with some secondary education (ISCED 3).

    View this table:
    Table 1

    Birth cohorts and numbers of deaths included in the analysis

    Educational inequality trends from period and cohort perspectives

    Figure 1 shows the absolute and relative inequalities for the period and cohort perspective, for men (figure 1A) and women (figure 1B) separately.

    Figure 1
    Figure 1

    Changes in inequality in all-cause mortality by period and cohort perspectives. RR, (adjusted) rate ratio; RD, rate difference.

    For men, the period perspective showed stable absolute inequalities in Northern and Southern Europe and declining inequalities in Western and Eastern Europe (from 2000s and onwards). The cohort perspective generally showed that absolute inequalities have increased in almost all countries between generations born earlier in the 20th century and those born later, except for Italy (constant) and the UK (decreasing). Both perspectives showed increasing relative inequalities for men in Northern Europe (except for Sweden), Spain and Eastern Europe (except for recent birth cohorts), but these increases in relative inequalities were larger in magnitude in the cohort perspective. The relative inequalities showed no or less clear increases in the other countries. In Belgium and the UK, inequalities decreased among men in the cohort perspective only.

    For women, absolute inequalities were stable or reduced in the period perspective, whereas absolute inequalities were higher for the more recent birth cohorts as compared with earlier birth cohorts in several countries, including Eastern European countries, Finland, Denmark, France and Spain. The cohort perspective showed increasing relative inequalities in these countries as well, except for France. The period perspective showed only increases in relative inequalities in Northern Europe, UK and Hungary.

    Mortality trends by education across birth cohorts

    Figure 2 shows the birth cohort trends in mortality from all-cause and four groups of causes of death among high-educated men (figure 2A), high-educated women (figure 2B), low-educated men (figure 2C) and low-educated women (figure 2D). The mortality trends among high-educated men and women decreased consistently: all-cause mortality and all four cause-of-death groups showed decreasing mortality, with the largest reductions for cardiovascular disease mortality.

    Figure 2
    Figure 2

    Changes in mortality by birth year among the high educated and the low educated. CCMF, cohort comparative mortality figure.

    Figure

    Figure 2 Continued.

    For all-cause mortality among low-educated men, a downward trend in successive birth cohorts born in the first decades of the 20th century often slowed down or even reversed in birth cohorts born since the 1930s. In Eastern Europe, mortality increases were seen over all successive birth cohorts. Among low-educated women, the reversals were more pronounced, with mortality generally considerably higher in birth cohorts born in the 1950s and 1960s than in those born in the 1920s and 1930s. The reversals among both men and women were particularly striking in Hungary and Finland. The only countries in which reversals appear to be absent, were the UK and Italy, with continuous declines, particularly in men.

    The slow down of decreasing trends or reversals among recent low-educated birth cohorts were found for several cause-of-death groups, among both men and women. Analysis of eight specific causes of deaths (online appendix 6 figure 1-2) provides more insight into causes with initially large mortality reductions that have slowed down or turned into an increase and causes with increased mortality in recent birth cohorts. Declines in mortality from ischaemic heart disease and cerebrovascular disease, and chronic obstructive pulmonary disease have slowed down in recent birth cohorts of low-educated men. Among low-educated women, decreases from these causes have reversed into increases in many countries. Mortality from alcohol-related causes gradually increased across successive birth cohorts, especially among recently born cohorts of low-educated men and women. In addition, mortality from road-traffic injury and suicide stabilised or increased among the low educated in more recently born generations.

    Discussion

    Summary of main findings

    The cohort perspective offers a less favourable picture of changes in educational inequalities in mortality as compared with the period perspective. Educational inequalities in mortality across birth cohorts have widened in many European countries, particularly among recent birth cohorts, women and in Eastern Europe. The widening was mainly driven by the unfavourable trends among the low educated where a mortality decrease was often followed by stabilising or even returning to the initial (high) mortality levels. This pattern is particularly evident for mortality from cardiovascular diseases (ischaemic heart disease, cerebrovascular disease), lung cancer, chronic obstructive pulmonary disease and alcohol-related causes.

    Strengths and limitations

    This is one of the first studies to apply a cohort perspective on changes in mortality inequalities across Europe using cross-national data. We were not only able to analyse a very large number of deaths (5.9 million) but also to ensure a certain degree of representativeness for Europe as a whole. The range of birth cohorts covered by the analysis is also substantial, including three-quarters of the 20th century.

    Besides well-known limitations inherent to international comparisons of mortality by SEP and cause of death,1–3 we discuss some limitations specific to this study. First, because of data availability, we were not able to compare the complete life course of the cohorts. The approach we have chosen is suited for the explorative purposes of this paper and enables us to include cohorts with different age groups. Nonetheless, we cannot rule out that the parts of the life course outside our observation windows differ in important ways. Confirmations of our results based on longer and more detailed times series that allow formal age-period-cohort modelling is needed to mathematically distinguish between effects that are due to age, cohort or period.27

    Second, our analysis often included only a limited number of age groups for each birth cohort. Particularly for countries where data were only available for two recent decades, the analysis had to rely on three to four age groups for most birth cohorts. The validity of the results thus depends on whether the empirical pattern can be extrapolated to higher ages and younger birth cohorts. In addition, this study covered data until the year 2015. Thus, more recent global challenges in healthcare and public health systems, with potential impact on mortality patterns, are not covered.

    Third and finally, we had to work around the limitations of the data that were not initially aimed at birth cohort analyses. Because data were collected in 5-year age groups and 1–5 years calendar periods, we had to construct and compare sometimes overlapping birth cohorts. We also had to exclude combinations of country–sex–age–education with less than 10 deaths to reduce random fluctuations in recent birth cohorts. Lastly, there were only two levels of education in the data from the UK. Therefore, mortality inequalities in the UK might be underestimated, and the time trends are less comparable to those in other countries.

    Interpretation

    Our study highlights the importance of analysing educational inequalities in mortality from a birth cohort perspective, in addition to a period perspective. Our analysis by birth cohort shows a less favourable development, with generations born after the 1930s at risk of developing larger mortality inequalities in mortality than their parents. This is mainly due to the unfavourable mortality trends among the low educated for more recent birth cohorts. Our analyses suggest that, in many European countries, low-educated men and women born after the 1930s no longer benefit from the continuous decreases in mortality seen among most previous generations. This applies particularly to mortality from ischaemic heart disease and cerebrovascular disease, and smoking-related and alcohol-related causes. The results may suggest stagnation—or reversal—of improvement in many cardiovascular risk factors, such as smoking and harmful alcohol use.28 29

    One possible explanation for this unfavourable picture for the low educated is that in successive generations the low-educated group has become much smaller. The share of high educated in the population has increased noticeably because of the rapid expansion of education opportunities in most European countries. Thus, the low educated may have become a more adversely selected subgroup of their generation with fewer opportunities for cognitive advancement and more adverse behavioural or personality profiles, and in turn, have less opportunity for socioeconomic advancement, creating an ‘unhealthy homogenisation’.30

    A second explanation is that social factors have become increasingly important, particularly for disease categories where humans have the capacity to prevent or treat disease.31 32 The high-educated individuals are more likely to adopt and maintain healthy practices or profit from new technologies and knowledge as they have the money, power and prestige to do so, while the low educated are more often laggards.33 It is also worth noticing that there is an increasing diversity in population composition due to immigration, which may also impact the magnitude of health inequalities, although the effect may depend on how (im)migrants are distributed over socioeconomic groups and their (dis)advantage compared with natives.34

    The upswing in mortality among recent-born birth cohorts, particularly among low-educated women, may be explained by, first, healthy homogenisatio since women have increased their educational level more than men during educational expansion. Second, the processes of diffusion of uptake and regress of (un)healthy practices differ by sex, SEP, region and risk factor(s). Women followed men with some delay in the uptake and regress several originally ‘male dominant’ risk factors for so-called ‘diseases of affluence’ (eg, cardiovascular disease, some cancers and injuries).35 This is well documented for smoking in the literature on the stages of the cigarette epidemic.36 For obesity, a similar stage model with opposite sex patterns is described in the literature.37 Over time, both of these risk factors show a reversal of the socioeconomic gradient from higher prevalence in higher SEP groups to higher prevalence in lower SEP groups. But for obesity, women are ahead of men, with initial higher obesity prevalence and earlier increases. In most European countries, low-educated women now experience the highest prevalence of obesity (compared with other groups and past periods), with no signs yet for a decrease. This may add to the mortality associated with initially ‘male dominant’ risk factors.

    Conclusions

    Analyses of changes by birth cohort give a different, and less favourable, picture of trends in educational inequalities in mortality than analyses of changes by calendar period. In many European countries, trends among more recently born generations are worrying. If current trends among younger birth cohorts continue, mortality inequalities are likely to further widen in the future, particularly among women.

    Data availability statement

    Data are available on reasonable request.

    Ethics statements

    Patient consent for publication

    Acknowledgments

    This study was conducted as part of the Centre for Global Health Inequalities Research (CHAIN), which is funded by the Norwegian Research Council (project number: 288638) and Erasmus Medical Center, Rotterdam, the Netherlands. Data were partly collected as part of the LIFEPATH project, which has received financial support from the European Commission (Horizon 2020 grant number 633666), and the DEMETRIQ project, which received support from the European Commission (grant numbers FP7-CP-FP and 278511). The permission of the Office for National Statistics (ONS) to use the Longitudinal Study is gratefully acknowledged, as is the help provided by staff of the Centre for Longitudinal Study Information & User Support (CeLSIUS). CeLSIUS is supported by the ESRC Census of Population Programme (award reference ES/K000365/1). The authors alone are responsible for the interpretation of the data. This work contains statistical data from ONS, which is Crown Copyright. The use of the ONS statistical data in this work does not imply the endorsement of the ONS in relation to the interpretation or analysis of the statistical data. This work uses research datasets, which might not exactly reproduce National Statistics aggregates. The mortality data for Switzerland were obtained from the Swiss National Cohort, which is based on mortality and census data provided by the Federal Statistical Office and supported by the Swiss National Science Foundation (grant nos. 3347CO-108806, 33CS30_134273 and 33CS30_148415). Pekka T. Martikainen was funded by the Academy of Finland and the MINDMAP project, which received financial support from the European Commission (Horizon 2020 grant number 667661). Mall Leinsalu was supported by the Estonian Research Council (grant PRG722). We acknowledge Bjørn Heine Strand, Carme Borell, Chris White, Eva-Maria Asamer, Gwenn Menvielle, Katalin Kovács, Olle Lundberg, for providing data.

    References

    1. ↵
      2. Mackenbach JP ,
      3. Kulhánová I ,
      4. Artnik B , et al
      . Changes in mortality inequalities over two decades: register based study of european countries. BMJ 2016;353:i1732. doi:10.1136/bmj.i1732
    2. ↵
      2. Mackenbach JP ,
      3. Kulhánová I ,
      4. Menvielle G , et al
      . Trends in inequalities in premature mortality: a study of 3.2 million deaths in 13 European countries. J Epidemiol Community Health 2015;69:207–17; doi:10.1136/jech-2014-204319
      OpenUrlAbstract/FREE Full Text
    3. ↵
      2. Mackenbach JP ,
      3. Rubio Valverde J ,
      4. Bopp M , et al
      . Progress against inequalities in mortality: register-based study of 15 european countries between 1990 and 2015. Eur J Epidemiol 2019;34:1131–42. doi:10.1007/s10654-019-00580-9
      OpenUrlPubMed
    4. ↵
      2. Mackenbach JP ,
      3. Valverde JR ,
      4. Artnik B , et al
      . Trends in health inequalities in 27 European countries. Proc Natl Acad Sci USA 2018;115:6440–5. doi:10.1073/pnas.1800028115
      OpenUrlAbstract/FREE Full Text
    5. ↵
      2. Frost WH
      . The age selection of mortality from tuberculosis in successive decades1. Am J Epidemiol 1939;30-SectionA:91–6. doi:10.1093/oxfordjournals.aje.a118570
      OpenUrl
    6. ↵
      2. MacMahon B ,
      3. Pugh TF
      . Epidemiology: principles and methods. Boston: Little, Brown and Company, 1970.
    7. ↵
      2. Bray FI ,
      3. Weiderpass E
      . Lung cancer mortality trends in 36 European countries: secular trends and birth cohort patterns by sex and region 1970-2007. Int J Cancer 2010;126:1454–66. doi:10.1002/ijc.24855
      OpenUrlCrossRefPubMedWeb of Science
    8. ↵
      2. Malvezzi M ,
      3. Carioli G ,
      4. Bertuccio P , et al
      . Relation between mortality trends of cardiovascular diseases and selected cancers in the european union, in 1970-2017. focus on cohort and period effects. Eur J Cancer 2018;103:341–55.:S0959-8049(18)30913-4. doi:10.1016/j.ejca.2018.06.018
      OpenUrl
    9. ↵
      2. Raho E ,
      3. van Oostrom SH ,
      4. Visser M , et al
      . Generation shifts in smoking over 20 years in two dutch population-based cohorts aged 20-100 years. BMC Public Health 2015;15:142. doi:10.1186/s12889-015-1481-3
    10. ↵
      2. Lui CK ,
      3. Kerr WC ,
      4. Mulia N , et al
      . Educational differences in alcohol consumption and heavy drinking: an age-period-cohort perspective. Drug Alcohol Depend 2018;186:36–43.:S0376-8716(18)30115-7. doi:10.1016/j.drugalcdep.2017.12.046
      OpenUrl
    11. ↵
      2. Cauley JA ,
      3. Donfield SM ,
      4. Laporte RE , et al
      . Physical activity by socioeconomic status in two population based cohorts. Medicine and Science in Sports and Exercise 1991;23:343 doi:10.1249/00005768-199103000-00013
      OpenUrlPubMedWeb of Science
    12. ↵
      2. Morciano M ,
      3. Hancock RM ,
      4. Pudney SE
      . Birth-Cohort trends in older-age functional disability and their relationship with socio-economic status: evidence from a pooling of repeated cross-sectional population-based studies for the UK. Social Science & Medicine 2015;136–137:1–9. doi:10.1016/j.socscimed.2015.04.035
      OpenUrl
    13. ↵
      2. Masters RK ,
      3. Hummer RA ,
      4. Powers DA
      . Educational differences in u.s. adult mortality: A cohort perspective. Am Sociol Rev 2012;77:548–72. doi:10.1177/0003122412451019
      OpenUrlCrossRefPubMedWeb of Science
    14. ↵
      2. Delaruelle K ,
      3. Buffel V ,
      4. Bracke P
      . Educational expansion and the education gradient in health: a hierarchical age-period-cohort analysis. Social Science & Medicine 2015;145:79–88. doi:10.1016/j.socscimed.2015.09.040
      OpenUrl
    15. ↵
      2. Klokgieters SS ,
      3. Huisman M ,
      4. van Groenou MB , et al
      . Socioeconomic pathways to inequalities in mental and functional health: a comparative study of three birth cohorts. BMC Public Health 2021;21:155. doi:10.1186/s12889-020-10154-0
    16. ↵
      2. Martikainen P ,
      3. Ho JY ,
      4. Preston S , et al
      . The changing contribution of smoking to educational differences in life expectancy: indirect estimates for Finnish men and women from 1971 to 2010. J Epidemiol Community Health 2013;67:219–24. doi:10.1136/jech-2012-201266
      OpenUrlAbstract/FREE Full Text
    17. ↵
      2. Martikainen P ,
      3. Elo I ,
      4. Tarkiainen L , et al
      . The changing contribution of childhood social characteristics to mortality: a comparison of finnish cohorts born in 1936-50 and 1961-75. Int J Epidemiol 2020;49:896–907. doi:10.1093/ije/dyaa041
      OpenUrlPubMed
    18. ↵
      2. Gregoraci G ,
      3. van Lenthe FJ ,
      4. Artnik B , et al
      . Contribution of smoking to socioeconomic inequalities in mortality: a study of 14 european countries, 1990-2004. Tob Control 2017;26:260–8. doi:10.1136/tobaccocontrol-2015-052766
      OpenUrlAbstract/FREE Full Text
    19. ↵
      2. Tabeau E ,
      3. Ekamper P ,
      4. Huisman C , et al
      . Improving overall mortality forecasts by analysing cause-of-death, period and cohort effects in trends. European Journal of Population/ Revue Europenne de Dmographie 1999;15:153–83. doi:10.1023/A:1006109310764
      OpenUrl
    20. ↵
      2. Galobardes B ,
      3. Shaw M ,
      4. Lawlor DA , et al
      . Indicators of socioeconomic position (Part 1). Journal of Epidemiology & Community Health 2006;60:7–12. doi:10.1136/jech.2004.023531
      OpenUrlAbstract/FREE Full Text
    21. ↵
      2. Heggebø K ,
      3. Tøge AG ,
      4. Dahl E , et al
      . Socioeconomic inequalities in health during the great recession: a scoping review of the research literature. Scand J Public Health 2019;47:635–54. doi:10.1177/1403494818801637
      OpenUrlCrossRefPubMed
    22. ↵
      2. Gardner MJ ,
      3. Osmond C
      . Interpretation of time trends in disease rates in the presence of generation effects. Stat Med 1984;3:113–30. doi:10.1002/sim.4780030205
      OpenUrlCrossRefPubMedWeb of Science
    23. ↵
      2. Bridger CA ,
      3. Helfand LA
      . Mortality from heat during july 1966 in illinois. Int J Biometeorol 1968;12:51–70. doi:10.1007/BF01552978
      OpenUrl
    24. ↵
      2. Karch DL ,
      3. Hall HI ,
      4. Tang T , et al
      . Comparative mortality among people diagnosed with HIV infection or AIDS in the u.s., 2001-2010. Public Health Rep 2015;130:253–60. doi:10.1177/003335491513000312
      OpenUrlCrossRef
    25. ↵
      2. Kim I ,
      3. Lim H-K ,
      4. Kang H-Y , et al
      . Comparison of three small-area mortality metrics according to urbanity in korea: the standardized mortality ratio, comparative mortality figure, and life expectancy. Popul Health Metr 2020;18:3. doi:10.1186/s12963-020-00210-7
      OpenUrl
    26. ↵
      2. Huisman M ,
      3. Kunst AE ,
      4. Bopp M , et al
      . Educational inequalities in cause-specific mortality in middle-aged and older men and women in eight Western European populations. The Lancet 2005;365:493–500. doi:10.1016/S0140-6736(05)17867-2
      OpenUrl
    27. ↵
      2. Clayton D ,
      3. Schifflers E
      . Models for temporal variation in cancer rates. II: age-period-cohort models. Stat Med 1987;6:469–81. doi:10.1002/sim.4780060406
      OpenUrlCrossRefPubMedWeb of Science
    28. ↵
      2. Keyes KM ,
      3. Li G ,
      4. Hasin DS
      . Birth cohort effects and gender differences in alcohol epidemiology: a review and synthesis. Alcohol Clin Exp Res 2011;35:2101–12. doi:10.1111/j.1530-0277.2011.01562.x
      OpenUrlCrossRefPubMed
    29. ↵
      2. Marcon A ,
      3. Pesce G ,
      4. Calciano L , et al
      . Trends in smoking initiation in europe over 40 years: a retrospective cohort study. PLoS One 2018;13:e0201881. doi:10.1371/journal.pone.0201881
    30. ↵
      2. Simons AMW ,
      3. Groffen DAI ,
      4. Bosma H
      . Socio-economic inequalities in all-cause mortality in europe: an exploration of the role of heightened social mobility [in eng]. Eur J Public Health 2013;23:1010–2. doi:10.1093/eurpub/ckt145
      OpenUrlCrossRefPubMed
    31. ↵
      2. Link BG
      . Epidemiological sociology and the social shaping of population health [in eng]. J Health Soc Behav 2008;49:367–84. doi:10.1177/002214650804900401
      OpenUrlCrossRefPubMedWeb of Science
    32. ↵
      2. Mackenbach JP
      . The persistence of health inequalities in modern welfare states: the explanation of a paradox. Soc Sci Med 2012;75:761–9. doi:10.1016/j.socscimed.2012.02.031
      OpenUrlCrossRefPubMed
    33. ↵
      2. Rogers EM
      . Diffusion of innovations. Simon and Schuster, 2010.
    34. ↵
      2. Mackenbach JP
      . Health inequalities: persistence and change in modern welfare states. USA: Oxford University Press, 29 August 2019. doi:10.1093/oso/9780198831419.001.0001
    35. ↵
      2. Mikhail GW
      . Coronary heart disease in women. BMJ 2005;331:467–8. doi:10.1136/bmj.331.7515.467
      OpenUrlFREE Full Text
    36. ↵
      2. Pampel F ,
      3. Legleye S ,
      4. Goffette C , et al
      . Cohort changes in educational disparities in smoking: France, Germany and the United States. Soc Sci Med 2015;127:41–50.:S0277-9536(14)00401-8. doi:10.1016/j.socscimed.2014.06.033
      OpenUrlCrossRefPubMed
    37. ↵
      2. Jaacks LM ,
      3. Vandevijvere S ,
      4. Pan A , et al
      . The obesity transition: stages of the global epidemic. Lancet Diabetes Endocrinol 2019;7:231–40.:S2213-8587(19)30026-9. doi:10.1016/S2213-8587(19)30026-9
      OpenUrlCrossRefPubMed

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Footnotes

    • Contributors Material preparation, data were collected by JPM and WJN and were prepared by BHS, CB, CW, E-MA, GC, GM, HB-H, ML, KK, MB, OL, PD, PM and RK. Data analyses were performed by DL. The first draft of the manuscript was written by JPM and revised by DL, WJN and SK. All authors commented on previous versions of the manuscript. All authors read and approved the final manuscript. WJN is the guarantor.

    • Competing interests None declared.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.