Introduction
Theories and measurements of healthy ageing
Ageing can be seen as a success story for public health policies, socio-economic development and medical advancements in relation to disease and injury, but it also challenges countries to adapt in order to maximise older people's health and functional capacities, and to maintain their social participation and security (World Health Organization (WHO), 2002). ‘Healthy ageing’ has become a popular topic worldwide in past decades. The term is often used interchangeably with other terms such as ‘active’, ‘successful’ or ‘productive’ ageing. Healthy ageing refers to the process of optimising opportunities for health, participation and security so as to enhance quality of life as people age, highlighting the impact of social environment on healthy ageing (WHO, 2002).
In the literature, Kuh's healthy biological ageing theory (Kuh et al., Reference Kuh, Richards, Cooper, Hardy, Ben-Shlomo, Kuh, Cooper, Hardy, Richards and Ben-Shlomo2014) and Rowe and Kahn's successful ageing theory (Rowe and Kahn, Reference Rowe and Kahn2015) both suggested that social engagement and mental capacities are as important as biological factors for achieving healthy ageing, indicating that healthy ageing should be measured in a multi-dimensional way. Previous literature reviews also found that physical capabilities, cognitive functions, physiological and metabolic health, and psychological and social wellbeing are fundamental phenotypic components of healthy ageing which have been frequently employed worldwide to measure healthy ageing comprehensively (Lara et al., Reference Lara, Godfrey, Evans, Heaven, Brown, Barron, Rochester, Meyer and Mathers2013, Lu et al., Reference Lu, Pikhart and Sacker2019). The WHO 2015 healthy ageing model (WHO, 2015) and the Baltes and Baltes's ‘selective optimisation with compensation’ model of successful ageing (Baltes and Baltes, Reference Baltes, Baltes, Baltes and Baltes1990) emphasised the concept of resilience, suggesting that the elderly are capable of taking advantage of their current capacities to compensate for any losses and limitations. When considering healthy ageing, a ‘disease-free’ ageing status might not be achievable but a ‘resilient’ status may be and classifying healthy agers by a dichotomous yes/no measure might introduce selection bias (Lu et al., Reference Lu, Pikhart and Sacker2019).
Improving the measurement of healthy ageing has also been emphasised by the WHO. The WHO 2015 healthy ageing model defined healthy ageing as ‘the process of developing and maintaining the functional ability that enables wellbeing in older age’ (WHO, 2015: 28). Functional abilities are the health-related attributes that allow people to do what they have reason to value, which can be determined by intrinsic capacities (the composite of all the physical and mental capacities that an individual can draw on) and social environment (both the individual- and environmental-level social factors in the extrinsic world), as well as the interactions between them (WHO, 2015). The WHO suggested that building and maintaining intrinsic capacity is one fundamental way of enhancing functional ability (WHO, 2015). A literature review identified five domains of intrinsic capacity, including cognition (memory, intelligence and problem solving), psychological (mood and emotional vitality), vitality (hormonal function and cardio-respiratory function), locomotion (balance, muscle strength and gait) and sensory (vision and hearing) (Cesari et al., Reference Cesari, Araujo de Carvalho, Amuthavalli Thiyagarajan, Cooper, Martin, Reginster, Vellas and Beard2018). Another two literature reviews also proposed similar domains constituting the intrinsic capacity construct, including physical capabilities, cognitive function, physiological health and psychological wellbeing (Lara et al., Reference Lara, Godfrey, Evans, Heaven, Brown, Barron, Rochester, Meyer and Mathers2013; Lu et al., Reference Lu, Pikhart and Sacker2019). Besides, researchers suggested that social wellbeing (e.g. social network, functioning or support) should also be considered as a key component for developing metrics of healthy ageing (Hodes et al., Reference Hodes, Insel and Landis2013; Lara et al., Reference Lara, Godfrey, Evans, Heaven, Brown, Barron, Rochester, Meyer and Mathers2013; Lu et al., Reference Lu, Pikhart and Sacker2019), which is an indicator of social environment and interconnected with intrinsic capacities (WHO, 2015).
Regarding the established scales of measuring intrinsic capacities and social wellbeing, according to the literature review based on 50 healthy ageing studies across 23 countries or regions (Lu et al., Reference Lu, Pikhart and Sacker2019), (instrumental) activities of daily living ((I)ADLs) were recommended for community-based studies to predict physical capabilities. It is also better to test direct observations of performance, such as grip strength, walking speed, balance and the chair rise test, to improve predictability. Measures of cognitive function were diverse across countries. But the word recall (immediate and delayed recall) and date naming (month, day of month, year, day of week) tests to measure short-term memory and orientation to time had been universally employed. Questions about self-reported absence of chronic diseases (e.g. high blood pressure, diabetes, cancer, lung disease, stroke, heart problems or arthritis) had also been frequently used to measure physiological health in many studies. For psychological wellbeing, each measure must focus on a clear conceptual domain. For example, the depressive symptoms could be measured using the Center for Epidemiological Studies Depression Scale (CES-D). Moreover, it is recommended that social participation should be measured in terms of specific social roles. Questions about participation in a variety of social activities, such as social or sports clubs, exercise classes, music groups, or Neighbourhood Watch had been frequently used.
Socio-economic inequalities in healthy ageing worldwide
A literature review in 2010 identified six future areas for the long-term study of ageing, based on 51 longitudinal studies of ageing worldwide; socio-economic inequality in health and wellbeing among the ageing population was one of those six areas (Stanziano et al., Reference Stanziano, Whitehurst, Graham and Roos2010). During the past decade, research questions regarding socio-economic inequalities in healthy ageing have been discussed in many articles worldwide. In general, older people with disadvantaged socio-economic positions (SEPs) are less likely to achieve healthy ageing than those with advantaged SEPs in many countries (Li et al., Reference Li, Wu, Jin, Zhang, Xue, He, Xiao, Jeste and Zhang2006; McLaughlin et al., Reference McLaughlin, Connell, Heeringa, Li and Roberts2010; Hirai et al., Reference Hirai, Kondo and Kawachi2012; Perales et al., Reference Perales, Martin, Ayuso-Mateos, Chatterji, Garin, Koskinen, Leonardi, Miret, Moneta, Olaya, Tobiasz-Adamczyk and Haro2014; Sowa et al., Reference Sowa, Tobiasz-Adamczyk, Topor-Madry, Poscia and la Milia2016).
However, the magnitude of socio-economic inequalities in healthy ageing across countries may be different due to variations in political, cultural, economic and epidemiological histories (Mackenbach et al., Reference Mackenbach, Stirbu, Roskam, Schaap, Menvielle, Leinsalu and Kunst2008). Previous studies have applied different measures of healthy ageing due to the inconsistency in definitions of healthy ageing (Li et al., Reference Li, Wu, Jin, Zhang, Xue, He, Xiao, Jeste and Zhang2006; McLaughlin et al., Reference McLaughlin, Connell, Heeringa, Li and Roberts2010; Hirai et al., Reference Hirai, Kondo and Kawachi2012; Perales et al., Reference Perales, Martin, Ayuso-Mateos, Chatterji, Garin, Koskinen, Leonardi, Miret, Moneta, Olaya, Tobiasz-Adamczyk and Haro2014; Sowa et al., Reference Sowa, Tobiasz-Adamczyk, Topor-Madry, Poscia and la Milia2016), which made the comparison of the magnitude of socio-economic inequalities in healthy ageing across countries difficult. For example, one study from the United States of America (USA) applying Rowe and Kahn's theory defined healthy ageing as no major diseases or disability, good cognitive and physical functioning, and active engagement in social activities (McLaughlin et al., Reference McLaughlin, Connell, Heeringa, Li and Roberts2010), but another study from Japan included mortality and loss of healthy life to assess healthy ageing (Hirai et al., Reference Hirai, Kondo and Kawachi2012).
Moreover, comparative studies about socio-economic inequalities in healthy ageing are also rare. Each of these is based on a single database, and the regions are restricted to European countries (Perales et al., Reference Perales, Martin, Ayuso-Mateos, Chatterji, Garin, Koskinen, Leonardi, Miret, Moneta, Olaya, Tobiasz-Adamczyk and Haro2014; Sowa et al., Reference Sowa, Tobiasz-Adamczyk, Topor-Madry, Poscia and la Milia2016). One study assessed factors associated with healthy ageing in Finland, Poland and Spain, finding that Finish participants achieved healthy ageing better than Polish and Spanish participants, and that higher education and occupation were commonly associated with higher levels of healthy ageing in the three countries (Perales et al., Reference Perales, Martin, Ayuso-Mateos, Chatterji, Garin, Koskinen, Leonardi, Miret, Moneta, Olaya, Tobiasz-Adamczyk and Haro2014). Another study found a positive educational gradient in healthy ageing in southern and central-eastern European countries, and a positive income gradient in healthy ageing among females in western European countries (Sowa et al., Reference Sowa, Tobiasz-Adamczyk, Topor-Madry, Poscia and la Milia2016). To our knowledge, no studies in the literature to date have compared socio-economic inequalities in healthy ageing among countries from different continental regions in the world.
Researchers have suggested that Europe offers excellent opportunities for comparative research, since good data on health inequalities are often available (Kunst et al., Reference Kunst, Bos and Mackenbach2011). But the conduct of comparative research should not be driven by data alone; countries outside Europe with large ageing populations also need to explore strategies to eliminate socio-economic inequalities in healthy ageing. Evidence shows that for life expectancy at birth, Japan, the USA and China ranked as the top three among countries with populations greater than 100 million in 2017 (GBD Mortality Collaborators, 2018), but the percentages of the working-age population had continuously decreased in all three countries up to that date (World Bank, 2018).
Similarly to high-income European countries, labour force ageing in the USA and Japan is also likely to be substantial over the next decades. Between 1995 and 2030, the share of workers aged 60 years and older was expected to rise, from 12.5 and 5.8 per cent to 30.1 and 16.1 per cent in Japan and the USA, respectively; this is similar to or larger than the increase in the average share (from 4.7 to 17.1%) in Europe (Organisation for Economic Co-operation Development (OECD), 1998). The number of older workers aged 60 years or more is unclear in China, but it too will increase in future decades as the retirement ages for both genders increase (World Economic Forum, 2016). The WHO proposed that healthy ageing assumes that ageing is a valuable process which permits older people to make crucial contributions to society, leading to personal fulfilment and economic growth; healthy ageing also shifts the traditional stereotypes of ‘old age’, and views the phenomenon of ageing as an opportunity (WHO, 2002). Given the phenomenon of labour force ageing among top economies including the USA, England, China and Japan, through conducting this cross-country comparison, this study will help explore universal and region-specific public health practices to support healthy ageing among both Western and Asian countries. Ensuring a successful demographic transition among the world's top economies will also be beneficial to the world economy.
Conducting an international comparsion to identify universal socio-economic determinants of healthy ageing is also in line with the WHO's suggestion: ageing research needs to be better co-ordinated across countries, to discover the most cost-effective approaches to maintain older people's health and wellbeing (WHO, 2011).
Researchers from countries including the USA, England, China and Japan are currently conducting nationally representative longitudinal studies of ageing; these are sister ageing studies, and they commonly incorporate measures of health, economic status, family and wellbeing (Hidehiko et al., Reference Hidehiko, Satoshi and Hideki2010; Steptoe et al., Reference Steptoe, Breeze, Banks and Nazroo2013; Sonnega et al., Reference Sonnega, Faul, Ofstedal, Langa, Phillips and Weir2014; Zhao et al., Reference Zhao, Hu, Smith, Strauss and Yang2014). Employing these four studies, which contain nationally representative samples of older people in the four countries, provides a unique opportunity to conduct a multinational comparison of socio-economic determinants of healthy ageing, on a scale not done before.
Therefore, we created a universal and multi-dimensional measure of healthy ageing, conducted comparative analysis to assess the magnitude of educational, income and wealth inequalities in healthy ageing in the USA, England, China and Japan, and identified influential socio-economic predictors of healthy ageing in each country.
We hypothesised that education is an influential predictor of inequalities in healthy ageing in the USA, as it is a key mechanism involved in raising a person's status in the USA (Bartley, Reference Bartley2004; Hollingshead, Reference Hollingshead2011); but education is not an influential predictor in China, as older Chinese people are generally illiterate or low-educated (Krieger and Fee, Reference Krieger and Fee1994) (Hypothesis 1). We also hypothesise that low- or middle-income countries such as China have greater socio-economic inequalities in healthy ageing (Fang et al., Reference Fang, Dong, Xiao, Liu, Feng and Wang2010), while countries such as England and Japan, which have been covered by free or low-cost national health services, have lesser socio-economic inequalities in healthy ageing (Reich and Shibuya, Reference Reich and Shibuya2015; National Health Service, 2018) (Hypothesis 2). Chinese participants are less likely to achieve healthy ageing than participants from the USA, England and Japan (Hypothesis 3).
Methods
Data
The data were from the US Health and Retirement Study (HRS) (Sonnega et al., Reference Sonnega, Faul, Ofstedal, Langa, Phillips and Weir2014), the English Longitudinal Study of Ageing (ELSA) (Steptoe et al., Reference Steptoe, Breeze, Banks and Nazroo2013), the China Health and Retirement Longitudinal Study (CHARLS) (Zhao et al., Reference Zhao, Hu, Smith, Strauss and Yang2014) and the Japanese Study of Aging and Retirement (JSTAR) (Hidehiko et al., Reference Hidehiko, Satoshi and Hideki2010). For the HRS, the analysis included data from Waves 7–12 (2004–2014). Wave 7 rather than Wave 1 of the HRS was used as the baseline wave. The reason is that some variables, such as social wellbeing-related variables, only started being recorded from 2004 (Wave 7) in the HRS. For ELSA, data from Waves 1–7 (2002–2015) were used. In CHARLS, data from Waves 1, 2 and 4 (2011–2015) were included. Wave 3 of CHARLS was a life history survey, and is not eligible for the current longitudinal analysis. In JSTAR, only participants from the original five cities from Waves 1–3 (2007–2011) were included.
For sample selection details, see Figure S1 in the online supplementary material. Participants were aged 60 years or more at baseline. Individuals without healthy ageing outcomes and booster samples without baseline weights were excluded. We finally included 10,305 (6,056 women) American, 6,590 (3,685 women) English, 5,930 (2,862 women) Chinese and 1,935 (995 women) Japanese participants.
Healthy ageing index
We followed three principles for creating a healthy ageing index (HAI, time-varying). First, healthy ageing should be measured in a multi-dimensional way. Both biological and psycho-social components should be considered. Psychological and social wellbeing are measured in order to examine the effects of self-efficacy, social roles and social support on functional wellbeing (Lara et al., Reference Lara, Godfrey, Evans, Heaven, Brown, Barron, Rochester, Meyer and Mathers2013; Lu et al., Reference Lu, Pikhart and Sacker2019). Second, ‘resilience’ should be taken into account when measuring healthy ageing. Healthy ageing is not simply a ‘disease-free’ ageing status and should not be a binary variable (yes versus no). Older people are capable of taking advantage of their current capacities to deal with their illnesses or impairments (Baltes and Baltes, Reference Baltes, Baltes, Baltes and Baltes1990; WHO, 2015). Third, each scale or method which is applied to measure healthy ageing should reflect a specified conceptual domain. Measuring physical capabilities by (I)ADLs, grip strength and other functional limitations (e.g. mobility, large muscle and fine motor skill), cognitive functions by verbal memory and orientation to time, physiological health by self-reported absence of chronic diseases, psychological wellbeing by the CES-D and questions about life satisfaction, and social wellbeing by participation in social activities were recommended (Lu et al., Reference Lu, Pikhart and Sacker2019).
The HAI included 33 indicators in physical, cognitive, physiological, psychological and social components of healthy ageing, considered resilience and chose established scales with clear conceptual domains (Figure 1). Each health indicator was dichotomised or organised into quartiles or quintiles, and then coded for the interval 0–100 (see Table S1 in the online supplementary material). For each individual, the scores on all indicators were summed and divided by the total number of measured indicators to yield an HAI score ranging from 0 to 100. A higher score of HAI indicates healthier ageing status.
Indicators of constructing the healthy ageing index.
Notes: ADLs: activities of daily living. IADLs: instrumental activities of daily living. lbs: pounds. kg: kilogramme. CES-D: Center for Epidemiological Studies Depression Scale.
For validity and reliability check in detail, see Tables S2, S3 and S4 and Figure S2 in the online supplementary material. The test–retest reliability and internal consistency of HAI were both >0.7, in an acceptable range (McDowell, Reference McDowell2006). The prediction of mortality by HAI was similar to that of phenotypic frailty (PF) (Fried et al., Reference Fried, Tangen, Walston, Newman, Hirsch, Gottdiener, Seeman, Tracy, Kop, Burke and McBurnie2001). However, compared to the measure of PF, the HAI measured more psycho-social components such as social participation and life satisfaction. The loge-transformed HAI was used for analysis due to HAI's left-skewed distribution.
Socio-economic indicators
The socio-economic indicators were baseline education, and time-varying income and wealth. The 1997 International Standard Classification of Education was applied (UNESCO Institute for Statistics, 2014). Educational categories were first stage of tertiary education or more), post-secondary non-tertiary education, upper secondary education, lower secondary education, and primary education or less and others. Total household income was divided by the square root of household size to give income per capita (OECD, 1995). Wealth was the total family assets. Both income and wealth were organised into quintiles, ranging from the highest to the lowest level in each country.
Covariates
Both baseline and time-varying covariates known to be associated with SEP and healthy ageing were included (Li et al., Reference Li, Wu, Jin, Zhang, Xue, He, Xiao, Jeste and Zhang2006; Hirai et al., Reference Hirai, Kondo and Kawachi2012; Perales et al., Reference Perales, Martin, Ayuso-Mateos, Chatterji, Garin, Koskinen, Leonardi, Miret, Moneta, Olaya, Tobiasz-Adamczyk and Haro2014; Sowa et al., Reference Sowa, Tobiasz-Adamczyk, Topor-Madry, Poscia and la Milia2016). Baseline variables were gender (male and female), ethnicity (white, black and others in HRS, white and non-white in ELSA, and Han and minority in CHARLS) and self-rated health in childhood (excellent, good, fair, poor and very poor). Time-varying variables were age in years, cohort (birth year), marital status (married/partnered, separated/divorced/single and widowed), smoking (current, previous and non-smoker) and drinking (frequency of drinking). Occupation (time-varying) and father's occupation (baseline) were also included as important socio-economic predictors during adulthood and childhood, respectively. Full harmonisation of occupational measures across the four countries was not achievable due to the disparities in societal background. Table S5 in the online supplementary material shows semi-harmonising strategies for occupation in the four countries.
Statistical analyses
The two-fold fully conditional specification (FCS) algorithm was applied to deal with missing values in socio-economic indicators and covariates (Welch et al., Reference Welch, Bartlett and Petersen2014). Age in each wave was used as the timing variable. Compared to the wave number, which assumes that every respondent is measured at the same time-point, age is more accurate in measuring the changes of HAI over time as it specifies entry and exit time for each individual differently. Records with imputed values for non-respondents in each wave were automatically excluded as the two-fold FCS algorithm only imputed non-responded items within each wave rather than non-responders in that wave. For percentages of missing values in each study, see Table S6 in the online supplementary material. Fifty imputed data-sets were generated in each country to ensure the number of imputations was large enough not to impact conclusions or inhibit analysis reproducibility. Baseline weighting adjustment was employed to account for complex survey designs.
To compare socio-economic inequalities in healthy ageing across countries, socio-economic rank scores (0–1, from the highest to lowest) were derived based on distributions of the education, income and wealth in each country (Regidor, Reference Regidor2004). For a hypothetical example using the educational classification to derive a socio-economic rank score, see Figure S3 in the online supplementary material.
Associations between socio-economic rank scores and loge-transformed HAIs were assessed using multi-level modelling to calculate the Slope Index of Inequality (SII), accounting for the entire socio-economic distribution and the sample size of each category of socio-economic indicator to make results comparable across countries (Regidor, Reference Regidor2004). An advantage of applying a multi-level approach is that the methodology is capable of handling attrition and wave non-response, unequal time spacing, and the inclusion of time-varying covariates that are either continuous or discrete measures (Raudenbush and Chan, Reference Raudenbush and Chan1992). Age–cohort (AC) models were estimated adjusting for all covariates and relevant interactions, and allowing for random intercepts and slopes. The AC model does not constrain the linear effect of period. Rather the period effect is integrated into slopes in the age and cohort dimensions (Nielsen and Nielsen, Reference Nielsen and Nielsen2014). SII was interpreted as the percentage of change in predicted mean HAI when individuals moved from the most to the least advantaged socio-economic groups (score changed from 0 to 1). Larger values reflect greater inequality. Coefficients for socio-economic rank scores indicated the cross-sectional relationships between socio-economic rank scores and HAIs at 60 years. Coefficients for interactions between socio-economic rank scores and age indicated the trend of change in the SII thereafter.
Additionally, healthy ageing trajectories were predicted based on the results of AC models, to compare older people's healthy ageing profiles across countries. The ‘observed value’ approach was applied to draw conditional trajectories, which holds each covariate at the observed value for each individual in the sample, calculates the relevant predicted marginal effect for each individual and averages over all cases (Hanmer and Kalkan, Reference Hanmer and Kalkan2013).
Besides, in each country, trajectories of loge-transformed HAIs by different categories of education were also drawn, based on the fully adjusted AC model. The AC model in each country was estimated from a non-standardised base since the original variable of education was used for longitudinal analysis. We illustrated these trajectories in order to set an example for visualising socio-economic inequalities in healthy ageing during the entire later life and helping interpret the interaction terms between SEP and age.
Stata 15.1 was applied for data analysis and p < 0.05 was considered statistically significant.
Results
Table 1 shows baseline characteristics. There were more females in the USA, England and Japan, but more males in China. Participants were mainly white in the USA and England, and Han in China. Most participants had upper secondary education in the USA, primary education or less in England and China, and upper or lower secondary education in Japan. The majority of American participants had already retired at baseline. English participants mainly had intermediate occupational positions. More than 70 per cent of Chinese participants were in unpaid agricultural work only; 22.25 per cent of Japanese participants were in the most disadvantaged occupations. American participants’ fathers had mainly been in disadvantaged occupational positions (around 70%), while English fathers’ occupational positions were mainly at intermediate levels (around 70%). Most Chinese and Japanese participants’ fathers were farming workers (78.30%) and self-employed (52.46%), respectively. Fewer American and English participants consumed alcohol every day and reported poor childhood health than Chinese and Japanese participants. Moreover, there was a great proportion of current smokers among Chinese and Japanese participants.
Sample characteristics at baseline
Notes: USA: United States of America. SD: standard deviation. 1. Japan does not have ethnicity and self-rated health in childhood variables. The ethnicity was recoded as one for all in data analyses. The sensitivity analysis confirmed that not adjusting for the self-rated health in childhood in Japan would not bias the comparison of socio-economic impact on HAI across countries. 2. In the USA, 1 = White/Caucasian, 2 = Black/African American, 3 = Others; in England, 1 = White, 2 = Non-white; in China, 1 = Han, 2 = Minorities. 3. For detailed categories of occupation and father's occupation in the four countries, see Table S3 in the online supplementary material. 4. In the USA and England, frequency of drinking = days of drinking per week (0 = none, 1 = 1 day, 2 = 2 days, 3 = 3 days, 4 = 4 days, 5 = 5 days, 6 = 6 days, 7 = 7 days); in China, frequency of drinking = times of drinking per month (0 = none or less than once per month, 1 = one to several times per month, 2 = one to several times per week, 3 = most days of the week, 4 = every day of the week); in Japan, frequency of drinking = times of drinking per month (0 = none, 1 = a few times in month, 2 = 1–2 in a week, 3 = 3–4 in a week, 4 = (almost) every day).
In Table 2, the linear coefficients of education, income and wealth present the SII at 60 years: the proportions of average change in HAI at 60 years if individuals had moved from the most to the least advantaged socio-economic groups in the four countries; the linear coefficients of interactions between education/income/wealth and age present the trends of SII thereafter: the gap in average HAI changes after 60 years between the most and the least advantaged socio-economic groups. Figure 2 summarises the results in Table 2, by illustrating the predicted SIIs for education, income and wealth at 60 years in the four countries, to compare the magnitude of healthy ageing inequalities by education, income and wealth within and across countries.
Predicted Slope Index of Inequality (SII) of the healthy ageing index by education, income and wealth at 60 years in each country.
Notes: The 95 per cent confidence intervals are shown. US: United States of America.
Results of fully adjusted linear multi-level models for associations between socio-economic rank scores and healthy ageing index (loge-transformed) across countries
Notes: CI: confidence interval. SD: standard deviation. Each model was adjusted for other socio-economic rank scores, age, age2, cohort, cohort2, gender, ethnicity, self-rated health in childhood, father's occupation, occupation, marital status, smoking and drinking, as well as interactions between gender and the main socio-economic rank scores, age and the main socio-economic rank scores, age and cohort, age and marital status, and age and smoking.
There was a non-linear relationship between age and loge-transformed HAI among American, English and Japanese participants (Table 2). The decline of healthy ageing accelerated with increased age. Among Chinese participants, both linear and quadratic age terms were non-significant.
Relationships between educational rank scores and HAIs at 60 years were significant in the four countries (Model 1). Participants with lower levels of education were less likely to achieve healthy ageing than those with higher levels of education. Moving from the highest to the lowest educational groups was associated with a 6.7 per cent (5.2–8.2%), 8.2 per cent (6.0–10.4%), 13.9 per cent (11.4–16.3%) and 6.1 per cent (3.9–8.2%) decrease in average HAI at 60 years in the USA, England, China and Japan, respectively. This inequality in HAI between the highest and lowest educational groups kept increasing with age after 60 years in the USA and China, due to significant and negative interactions between age and education.
Relationships between income rank scores and HAIs at 60 years were significant only in the USA and China (Model 2). Participants with lower levels of income were less likely to achieve healthy ageing than those with higher levels of income. Moving from the highest to the lowest income quintiles was associated with a 1.4 per cent (0.7–2.2%) and 3.2 per cent (1.7–4.8%) decrease in average HAI at 60 years in the USA and China, respectively. However, this inequality in HAI between the highest and lowest income groups did not change with age after 60 years in the USA and China, due to boundary or non-significant interactions between age and income.
Relationships between wealth rank scores and HAIs at 60 years old were significant in the USA, England and Japan (Model 3). Participants with lower levels of wealth were less likely to achieve healthy ageing than those with higher levels of wealth. Moving from the highest to the lowest wealth quintiles was associated with a 3.3 per cent (2.4–4.3%), 6.2 per cent (4.9–7.5%) and 1.5 per cent (0.1–3.0%) decrease in average HAI at 60 years in the USA, England and Japan, respectively. However, this inequality in HAI between the highest and lowest wealth groups only increased with age after 60 years in the USA, due to a significant and negative interaction between age and wealth.
Figure 3 shows predicted healthy ageing trajectories after 60 years across countries. The gradient in healthy ageing across countries was clear. Japanese participants were healthier after 60 years than participants in any other country. Chinese participants had the worst health profiles in each year after the age of 60. English and American participants’ healthy ageing ranked second and third, respectively. However, the four trajectories might aggregate in very old age. The rates of decline in healthy ageing accelerated with increased age in the USA, England and Japan. However, in China, the slope did not change across ages, due to the non-significant non-linear effect of age, suggesting a constant rate of decline in HAI across ages after 60 years.
Predicted trajectory of healthy ageing after 60 years in each country. In Japan, the trajectory before 79 years is presented since all participants were aged 79 or less.
Notes: The 95 per cent confidence intervals are shown. HAI: healthy ageing index. US: United States of America.
Figure 4 illustrates predicted trajectories of healthy ageing by categories of education in each country. Generally in the four countries, the educational gradients in healthy ageing were clear after 60 years. Participants with higher levels of education had better health profiles than those with lower levels of education in each year after the age of 60. In both the USA and China, due to significant and negative interactions between education and age, trajectories for primary education or less declined much faster than any other educational trajectory; the inequality in healthy ageing between the highest and lowest educational groups kept increasing during the entire later life. In either England or Japan, even though the declining rates of trajectories were similar at each age, the trajectory for participants with primary education or less kept staying lower than any other trajectory in later life.
Predicted trajectories of healthy ageing by categories of education after 60 years in each country. In Japan, the trajectory before 79 years is presented since all participants were aged 79 or less.
Notes: FI: first stage of tertiary education or more. PO: post-secondary non-tertiary education. UP: upper secondary education. LO: lower secondary education. PR: primary education or less. OT: other. HAI: healthy ageing index. US: United States of America.
Discussion
In summary, participants with advantaged SEP were more likely to achieve healthy ageing than participants with disadvantaged SEP. Education was the strongest predictor of healthy ageing in the four countries (Hypothesis 1 for the USA and China was accepted and rejected, respectively). The inequality in healthy ageing between the highest and lowest educational groups kept increasing with age after 60 years in the USA and China. Wealth was an influential indicator in the USA, England and Japan, while income was an influential indicator in the USA and China. The educational inequality in healthy ageing in China was distinctly larger than any other socio-economic inequality in healthy ageing in the four countries (Hypothesis 2 for China was accepted). The wealth inequality in healthy ageing in England was larger than that in the USA and Japan. The magnitude of socio-economic inequalities in healthy ageing was relatively small in Japan (Hypothesis 2 for Japan was accepted). Japanese, English, American and Chinese participants’ healthy ageing in later life ranked first, second, third and last, respectively (Hypothesis 3 was accepted).
Educational inequalities in healthy ageing
Education was the strongest predictor of healthy ageing in the four countries. Education is widely used as a measure of social status, since it is a key mechanism involved in raising a person's status (Hollingshead, Reference Hollingshead2011). Many studies have found a strong association between education and later-life health. For example, education-related differences in mortality and life expectancy have widened over the past 20–25 years in the USA (Hummer and Hernandez, Reference Hummer and Hernandez2013). In England, having lower education was related to reporting poorer later-life health (Grundy and Holt, Reference Grundy and Holt2001). In China, persons who were illiterate and low-educated had significantly higher levels of depression (Li et al., Reference Li, Liu, Xu and Zhang2015a). In Japan, older people with lower education had a higher risk of experiencing early mortality than people with higher education (Fujino et al., Reference Fujino, Tamakoshi, Iso, Inaba, Kubo, Ide, Ikeda and Yoshimura2005).
Researchers believed that attaining more education makes people pay greater attention to their health (Wang and Yu, Reference Wang and Yu2016). More importantly, a higher level of education may lead to a higher level of income and a more advantaged occupational position. People with better socio-economic conditions are more likely to be healthy (Marmot, Reference Marmot2005).
The educational inequality in healthy ageing among the Chinese participants was distinctly larger than any other socio-economic inequalities in healthy ageing in the four countries. Education in China is a ladder to social success, especially among the older generations. China started making tremendous efforts to improve the quality of education only after the late 1950s; the major transformations of the educational system only started in the early 1980s (Rong and Shi, Reference Rong and Shi2001). Therefore, most Chinese participants in our study ended up illiterate or with an education at less than secondary-school level; few of them went to university for bachelor's or higher studies (see Table 1). However, enterprises and governments desperately needed highly educated ‘talent’ to contribute to economic acceleration and capital accumulation in the late 20th century in China (Rong and Shi, Reference Rong and Shi2001). In this context, educational inequalities created significant income and occupational gaps, since persons with bachelor's degrees or more during that time quickly achieved upward social mobility, gaining higher incomes and occupational positions and greater asset accumulation (Rong and Shi, Reference Rong and Shi2001). From a lifecourse perspective, this is intra-generational mobility – a change from a disadvantaged SEP to an advantaged SEP within one's own lifecourse (Hallqvist et al., Reference Hallqvist, Lynch, Bartley, Lang and Blane2004). During this time, highly educated people might have had a healthier working life, a less-deprived living environment, more positive social participation, and stronger economic and social security. Therefore, among the older generations, compared with those who did not even go to secondary school, people with bachelor's degrees or more in China were far more likely to achieve healthy ageing.
Wealth versus income inequalities in healthy ageing
Wealth was another influential predictor of healthy ageing among American and English participants. A study found that compared with education and income, wealth was the strongest predictor of mortality among the American elderly (Hoffmann, Reference Hoffmann2011). Researchers also suggested that total net non-pension household wealth was the most robust indicator of current SEP in England since it captured financial and other material resources at older ages the most accurately (Steptoe et al., Reference Steptoe, Breeze, Banks and Nazroo2013). A US/UK study found that those in the lowest wealth quintile had higher risks of death and disability than their counterparts in the highest wealth quintile (Makaroun et al., Reference Makaroun, Brown, Diaz-Ramirez, Ahalt, Boscardin, Lang-Brown and Lee2017).
However, wealth was not an influential predictor of healthy ageing in China. Wealth inequality in healthy ageing was unclear in China in this study. In the literature, evidence for wealth inequality in health among the Chinese elderly is limited. We only found one English publication suggesting that more luxury items and better housing quality were associated with less depression among the rural Chinese ageing population (Li et al., Reference Li, Liu, Zhang and Xu2015b). Wealth inequality in property ownership in China has been rising dramatically during the past decades. In 2012, 78.7 and 60.9 per cent of household wealth consisted of housing assets in urban and rural China, respectively; from 2013 to 2015, property income rose by 9.9 per cent, while salary income increased by only 8.9 per cent (National Bureau of Statistics of China, 2016). However, some researchers believe that compared with developed countries, the achievement of high levels of population health in China might not require a generally high level of national wealth. Social investments to eliminate illiteracy, improve the quality of education, protect farmers’ benefits, provide universal primary health-care services and meet basic living needs were more important for achieving healthy ageing in China (Schweiger, Reference Schweiger1997).
We found that income was less influential in the USA, and not influential in England. For older people, the source of income was mainly pensions and other public benefits. Researchers found that in the USA, compared with the working-age population, government transfers had been more equally distributed among the ageing population (Bosworth et al., Reference Bosworth, Burtless and Zhang2016). Similarly, pension income inequality was substantially lower in the UK than that in other developed countries (Sefton et al., Reference Sefton, Evandrou and Falkingham2007). Equally distributed government transfers might buffer the effect of income on inequalities in healthy ageing. Therefore, fewer income disparities in healthy ageing were found among American and English participants.
However, previous studies still found income inequalities in health among the elderly in the USA and England. For example, a US study found that low income was a more influential risk factor for mortality than low education (Hoffmann, Reference Hoffmann2011). Another UK study found worse self-reported health among older people on low incomes (Grundy and Holt, Reference Grundy and Holt2001).
Researchers found that the distribution of private pension was unequal in the USA: private pensions increased annual income inequality by 21 per cent among low-income workers (Benedict and Shaw, Reference Benedict and Shaw1995). In the UK, researchers held the opinion that the pension system was effective in preventing the ‘very bottom’ but not ‘low to moderate’ poverty (Sefton et al., Reference Sefton, Evandrou and Falkingham2007). Unequal pension income might still contribute to disparities in healthy ageing. Disaggregating income into pension or unearned income versus earned income and assessing their associations with healthy ageing, respectively, might be more instructive for finding out income inequalities in healthy ageing in each country.
Differently in China, income was an influential indicator. Income inequality in healthy ageing tended to be larger than that in any other country. A study found that income was the dominant risk factor for inequalities in health-care utilisation among outpatients in both developed and developing provinces of China (Wang et al., Reference Wang, Wang, Maitland, Zhao, Nicholas and Lu2012). Pension income is an important socio-economic determinant of health in China. However, in 2013 the ratio of average benefits in China was estimated at 50:25.5:1 for civil servants’, workers’ and residents’ pensions, respectively; more than 400 million people in China had no old-age pension at all (International Labour Office, Reference Ortiz2015). The huge gaps among pension schemes, and between pension receivers and non-receivers, greatly contribute to inequalities in living standards and in the utilisation of health services in China.
Equity in healthy ageing in Japan
The magnitude of healthy ageing inequalities in Japan was relatively small. Japanese participants had the best healthy ageing profile. Since 1986, Japan has ranked first in the world for women's life expectancy in childbirth (Reich and Shibuya, Reference Reich and Shibuya2015). Low-cost health services in Japan during the past decades have maintained people's health and increased social equity among the general population (Reich and Shibuya, Reference Reich and Shibuya2015). Japanese society became more economically egalitarian after the Allied occupation of Japan (1945–1952). Japanese people focused on productive outcomes, and on societal rather than market or individual opportunities, which had profound health effects among the general population (Bezruchka et al., Reference Bezruchka, Namekata and Sistrom2008). By 1970, the income ratio between the top and bottom income quintiles had decreased to 4.3:1 in Japan, while in the same year the ratio was 7.1:1 in the USA (Vogel, Reference Vogel1979).
However, we still found significant educational and wealth inequalities in healthy ageing among Japanese participants. A previous study found that older people with lower levels of education had a higher risk of experiencing early mortality than people with higher levels of education (Fujino et al., Reference Fujino, Tamakoshi, Iso, Inaba, Kubo, Ide, Ikeda and Yoshimura2005). Socio-economic inequalities in life expectancy and mortality increased continuously between 1995 and 2000 in Japan (Fukuda et al., Reference Fukuda, Nakao, Yahata and Imai2007). Therefore, Japan's achievement in promoting healthy ageing among the general population in a more equal society cannot be denied. However, more empirical evidence is needed regarding socio-economic inequalities in healthy ageing in Japan, based on data with fewer missing values, less skewed distributions of variables and a wider age range.
The great gap in healthy ageing across countries
Japanese, English, American and Chinese participants’ healthy ageing in later life ranked first, second, third and last, respectively. This rank is similar to the rank of life expectancy after 60 years in 2017 in the four countries (GBD Mortality Collaborators, 2018). Chinese participants had the worst health profiles on average. This lag in achieving healthy ageing might be affected by great health disparities in China. The educational inequality in China was distinctly larger than any other socio-economic inequality in healthy ageing across countries in this study. However, the socio-economic gap in health is still enlarging due to unequal distribution of income, wealth and health-care services in China (Fang et al., Reference Fang, Dong, Xiao, Liu, Feng and Wang2010).
The US government spends more on health care than any other developed country (OECD, 2015). However, the American population is still less healthy than the Japanese and English populations. Compared with England and Japan, adults in the USA were more economically disadvantaged in attaining health-care services: 37 per cent of adults in the USA did not see a doctor or failed to fill a prescription because of high costs (Commonwealth Fund, 2016). The socially produced inequalities in health status in the USA have made the achievement of healthy ageing more difficult than in the UK and Japan.
Strengths
First, we referred to multiple theories for the measurement of healthy ageing. The HAI was developed to include both biological and psycho-social components of healthy ageing, and to consider social opportunities and resilience, thus measuring healthy ageing in a comprehensive way. The HAI can be applied as a preliminary screening of healthy agers after 60 years of age in the four countries, which might help clinicians and researchers identify patients’ healthy ageing profiles.
Second, this research fills a research gap by comparing socio-economic inequalities in healthy ageing among Asian, European and North American countries. The use of four national longitudinal studies of ageing, with around 25,000 representative older adults, has provided a unique opportunity to conduct a Western–Asian comparison of healthy ageing, which to our knowledge has never been done before. Identifying influential socio-economic predictors of healthy ageing in each country is instructive for exploring universal and country-specific public health practices in supporting healthy ageing among both Western and Asian countries.
Third, advanced statistics were employed appropriately. The two-fold FCS algorithm is able to specify an entry and exit time for each participant; automatically consider interactions between age and other variables; and impute non-responding items only, but not non-responders, in each wave. We also calculated the SII based on a multi-level linear regression equation with full adjustment. Confounding and random effects were taken into account. This multi-level approach allows the prediction of the SII at 60 years and of changing rates of SII after 60 years.
Limitations
First, this study included US data from 2004 to 2014 (11 years) and English data from 2002 to 2015 (14 years), while Chinese data were only available from 2011 to 2015 (five years) and Japanese data from 2007 to 2011 (five years). The data from the USA and England had stronger statistical power for conducting longitudinal analyses than the data from China and Japan, due to the larger sample sizes and longer follow-up durations.
Second, to make results comparable, only variables common to the four studies were included in the HAI and for conducting data analyses. However, some country-specific variables, such as Index of Multiple Deprivation scores in England, financial support from children in China and home ownership in Japan might also be markers of healthy ageing and explain variations in SEP–healthy ageing relationships. Besides, geographical variables are unavailable in JSTAR; and for other countries, variables measure region at various area levels due to different degrees of data confidentiality. However, disparities in healthy ageing across regions within each country could exist. Future study with detailed geographical information could explore the regional inequality in healthy ageing in each country.
Third, an occupational rank score was not derived since occupation does not have a strict hierarchical ranking (Regidor, Reference Regidor2004), although occupation was included as a covariate.
Fourth, a selective survival bias might exist. When we conducted the data analysis, participants without valid HAIs at baseline were excluded. We only imputed item non-response for main exposures and covariates in each wave. Those respondents without HAIs were more likely to have severe illness (Delgado-Rodríguez and Llorca, Reference Delgado-Rodríguez and Llorca2004). Moreover, this research excluded individuals aged less than 60 years at baseline. Distributions of some covariates among respondents might be altered and variation in risk factors might also be reduced due to survival selection.
Last, SIIs only represent the average change in healthy ageing by SEP, accounting for less than 20 per cent of the variability in HAIs. Moreover, variations in HAIs were small, especially in Japan (standard deviation = 6.89).
Conclusions
In conclusion, Japanese and English participants achieved healthier ageing than American and Chinese participants. A positive socio-economic gradient in healthy ageing existed in all countries. Socio-economic inequality in healthy ageing was relatively small in Japan, but more evidence over time is needed. In China, inequality in healthy ageing, especially by education, was daunting.
Education was a universally influential socio-economic predictor of healthy ageing across countries. After 60 years, the educational inequality in healthy ageing in the USA and China kept increasing, indicating that this early life socio-economic factor could affect individuals’ healthy ageing later in the lifecourse. Wealth inequality in healthy ageing was greater in England than in any other country. Wealth was more influential than income in predicting inequalities in healthy ageing in the USA, England and Japan, while income was more influential than wealth in China. More evidence is needed for the effects of pension income on healthy ageing in the USA and UK, and the effects of wealth on healthy ageing in China.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0144686X19001740
Data
The HRS was developed by a team of researchers based at the University of Michigan, supported by the National Institute on Aging (NIAU01AG009740) and the Social Security Administration. The HRS data were made available through the HRS website at http://hrsonline.isr.umich.edu/. ELSA was developed by a team of researchers based at NatCen Social Research, University College London and the Institute for Fiscal Studies. The data were collected by NatCen Social Research. The funding was provided by the National Institute on Aging in the USA and by a consortium of UK government departments co-ordinated by the Office for National Statistics. The ELSA data were made available through the UK Data Archive at http://www.data-archive.ac.uk/. CHARLS was supported by the Behavioural and Social Research division of the National Institute on Aging (grant numbers 1-R21-AG031372-01, 1-R01-AG037031-01, 3-R01AG037031-03S1); the Natural Science Foundation of China (grant numbers 70910107022, 71130002, 71273237); the World Bank (contract numbers 7145915, 7159234); the China Medical Board; and Peking University. The CHARLS data were made available through the CHARLS website at http://charls.pku.edu.cn/en. JSTAR was conducted by the Research Institute of Economy, Trade and Industry (RIETI), Hitotsubashi University and the University of Tokyo. JSTAR data-sets are distributed by RIETI in Tokyo (see https://www.rieti.go.jp/en/projects/jstar/). Harmonised data-sets from the Gateway to Global Aging Data website (https://g2aging.org/) were used where possible. Not all of the variables in the analysis came from that website, since the harmonisation of some variables needed for this study had not been completed, and some variables were only semi-harmonised, despite notable differences in measures and categories. Instead, many variables were taken from the four original databases and harmonised for analysis in this study.
Acknowledgements
We thank the developers, funders and archives of HRS, ELSA, CHARLS and JSTAR, and the Gateway to Global Aging Data. All of them bear no responsibility for the analyses or interpretations presented here.
Financial support
This work was supported by a University College London Overseas Research Scholarship; the China Scholarship Council (file number 201608060385); the European Union's Horizon 2020 Research and Innovation Programme (WL, grant agreement number 635316); and the Economic and Social Research Council (AS, award number ES/R008930/1).
Conflict of interest
The authors declare no conflicts of interest.
Ethical standards
This study is based on secondary data analysis. No ethical approval is needed.
