Are there educational disparities in health and functioning among the oldest old? Evidence from the Nordic countries

Socioeconomic health inequality, i.e. higher life expectancy and better health in higher socioeconomic status (SES) groups, is a global phenomenon and the Nordic countries (Denmark, Finland, Iceland, Norway and Sweden), known for their generous welfare systems, are no exceptions (Mackenbach 2012). While the level of generosity, spending on eldercare and the way of organizing health and social services differ in many respects, the Nordic countries have certain social policies in common, such as universal health- and social care coverage and minimum pensions that support longevity and social and economic security in old age (Lundberg et al. 2008; Szebehely and Meagher 2017). Even with this social protection and class-equalizing potential, several studies have shown substantial socioeconomic health inequalities in young, working age and young old age populations in the Nordic welfare states (Bambra 2012; Lahelma et al. 2001; Mackenbach et al. 2008). Yet, relatively little is known about the extent of health inequalities in very old age.

Besides the similarities in welfare state models, Nordic countries have rather similar patterns of population ageing. The proportion of those who reach old age (e.g. 75+ or 95+) is smaller in Finland, especially among men, than in the other Nordic countries. However, after reaching old age (i.e. 75 years) the remaining life expectancies are very similar (Jørgensen et al. 2018) ranging from 12.04 to 12.67 years (Human Mortality Database 2018). Altogether, the proportion of 75 + year olds of the total population is expected to increase from 7.5% in 2015 to 12% in 2040 in the Nordic countries (Eurostat 2018).

House et al. (1990) have elaborated on the potential mechanisms by which the association of SES with health may vary by age. In young age groups, the prevalence of health problems is lower due to biological robustness, which may lead to smaller health inequalities. In middle and young old age, the impact of work-related psychosocial and environmental risk factors on health is high, causing higher inequality in health between socioeconomic groups. Different hypotheses have been put forward on the mechanisms of the increase or decrease in health inequalities in old age, see e.g. (Hoffmann 2011; Rehnberg et al. 2019). Exit from the labour market reduces exposure to work-related, unequally distributed health risks, and this may lead to a convergence in health inequalities in later life (House et al. 1994). In addition, inevitable biological processes and the equalizing social policies in welfare states are suggested to contribute to the decline in health inequalities in old age. Counteracting with the biological processes, also selective survival, which refers to the higher mortality in lower socioeconomic groups, may result in reduced health inequalities if only the healthiest and most robust individuals from the lower socioeconomic groups reach old age (Dupre 2007). On the other hand, if the accumulation of advantage or disadvantage in material and social resources lasts throughout life, it could lead to an increase in health inequalities in old age (Ross and Wu 1996). Further, if the health effects of poor working conditions are postponed to older ages, it might increase health inequalities in old age.

The evidence of the existence and size of health inequalities by age in later life is inconsistent, and has barely touched upon the oldest old age groups. Minkler et al. (2006) showed a social gradient in functional limitations among 55–84 year olds in the US: the lower the level of education, the worse the functioning. Inequalities were strongest in younger age groups (55–64) and flattened towards older age groups, and finally disappeared at the age of 85+. Huisman et al. (2003) also showed decreasing but persistent inequalities in long-term disabilities and self-rated health (SRH) for men in the age groups 60–69, 70–79, and 80+, but not for women in the oldest age group in a study with 11 European countries. Rostad et al. (2009) who studied inequalities in limiting long standing illnesses and SRH for women in Norway found that inequalities were apparent and even seemed to increase in SRH in the oldest age group (85+) according to education, but not with other health or SES indicators. Arber and Cooper (1999) found persistent inequalities in disability (80+) and in SRH (85+) in Britain with previous occupation as a SES indicator. Schöllgen et al. (2010) studied health inequalities in physical, functional and subjective health with education, income and financial assets as SES indicators in the study sample that consisted of 40–85 year-old Germans. The study showed in general stability or increase in inequalities throughout the age groups (40–54, 55–69, and 70–85) but decrease according to education in subjective health. Other earlier studies that have shown health inequalities among the oldest old have been limited to one cohort or one country (Bootsma-van Der Wiel et al. 2005; Enroth et al. 2013; Fors and Thorslund 2015).

With the ageing of the populations and recent pressures on important welfare state arrangements, updated knowledge on the linkage between SES and health in old age is pertinent for shedding light on emerging patterns of health inequalities in the Nordic countries. The current study focuses on health and functioning inequalities by the level of education. We assess the extent of health and functioning (self-rated health, independence in mobility and activities of daily living (ADL)) inequalities among people aged 75+ in four Nordic countries (Denmark, Finland, Norway and Sweden). The study aims to provide an overview of the direction and magnitude of health and functioning inequalities in three age groups 75–84, 85–94, and 95+, and in the 75+ population as a whole.

The unadjusted prevalences showed that, in general, individuals in the youngest age group had a higher percentage of being independent in mobility and in ADL than individuals in the middle age group. In turn, the percentage was higher in the middle age group than in the oldest age group (Table 2). The DLSA study tended to show a higher level of independence in mobility than the other studies. For SRH, the variability between age groups and between countries was high, especially in the oldest age group. Participants in the Vitality 90+ Study assessed their health worse than participants in the other studies. Overall, with only a few exceptions, in all age groups, for all three outcomes, and for all countries, individuals with higher education had more favourable health and functioning than those with basic education.

Figure 1 shows a forest plot for good SRH according to level of education separately for all age groups and an overall summary effect for all age groups combined. The age-specific analysis for the youngest age group (75–84 years old) showed significantly higher odds for good SRH among those with higher education than among those with basic education in all individual studies and in the summary effect that combines all studies (OR 1.82 CI 1.38; 2.41). For the middle age group (85–94 years old), the DLSA study differed from the other studies by showing lower odds for good SRH for those with higher education; however, this result was not statistically significant. For the oldest age group (95+ years old) all individual studies showed higher odds for good SRH for those with higher education. However, the summary effect did not reach statistical significance neither for the middle age group nor for the oldest age group. When synthesizing results for all age groups, people with higher education had higher odds of having good SRH than those with basic education (OR 1.51 CI 1.24; 1.84). The overall between-study heterogeneity showed moderate consistency for the studies (p = 0.02, I-squared 54.3%).

Figure 2 shows a forest plot for mobility. For the youngest age group, all individual studies showed higher independence in mobility for those with higher education when compared to those with basic education. Even though the result was statistically significant only for the SWEOLD study, the summary effect of all studies reached statistical significance (OR 1.58 CI 1.22; 2.04). For the middle age group, all studies except the DLSA showed significantly higher independence in mobility for those with higher education. Moreover, the summary effect showed significant differences in mobility between individuals with higher and basic education (OR 1.47 CI 1.07; 2.03). The results were highly similar for the oldest age group (OR 1.89 CI 1.25; 2.86). Furthermore, synthesized results for all age groups showed the same; higher independence in mobility for those with higher education (OR 1.54 CI 1.30; 1.82). The heterogeneity of all studies was moderate (p = 0.18, I-squared 28.3%).

Figure 3 shows a forest plot for independence in ADL. In the youngest age group, individuals with higher education had higher odds of being independent in ADL than individuals with basic education. The differences in ADL between levels of education were statistically significant in the SWEOLD study, and in the summary analysis (OR 1.68 CI 1.17; 2.41) that included results of all studies in the age group. Also for the middle age group, all individual studies showed higher independency in ADL for those with education beyond the basic level resulting in a statistically significant summary effect (OR 1.29 CI 1.07; 1.57). For the oldest age group, Vitality 90+ Study and SE-COOP studies showed higher odds for being independent in ADL for those with higher education. The result from the SWEOLD study differed from the other studies; however, it had less weight in the analysis because of the small study sample. Thus, the summary effect showed higher independence in ADL for those with higher education also in the oldest age group (OR 1.80 CI 1.20; 2.72). When all studies in all age groups were combined, those with higher education had statistically significantly higher odds (OR 1.42 CI 1.22; 1.67) of being independent in ADL than those with basic education. The between-study variability was low (p = 0.63, I-squared 0%).

This study adds knowledge of health and functioning inequalities in very old age by examining inequality patterns according to level of education in three age groups 75–84, 85–94, and 95+ in four Nordic countries. We found that individuals with higher education were more likely to report good SRH, and they were more often independent in mobility and ADL than people with basic education. Stratified analyses across three age groups showed similar findings, except for SRH, where the summary effect was statistically significant only for the age group 75–84. As the study was based on the prospective meta-analysis, it included high-quality comparable data of the oldest old.

We used the same indicators of health and functioning for all age groups across the five studies. When looking at the results of being independent in ADL and mobility, the summary effects show significantly higher independence for people with higher education across all three age groups. For SRH, these differences between education groups were only significant for the youngest age group. SRH is a widely used measure of general health status (Idler and Benyamini 1997), but its interpretation is more complex than for ADL and mobility. The SRH measure is affected by age and culture but also the way of posing the question and answer alternatives (Jylhä 2009). It has been suggested that the oldest old give more positive health ratings compared to younger age groups, which could lead to reduced inequalities. The more positive health ratings may be related to selective mortality, to lowered health expectations in very old age or to downward comparisons, i.e. comparing health to age peers with health problems (Cheng et al. 2007; Tornstam 1975). Another explanation for the slightly deviating patterns of inequality in SRH is that responses to SRH do not include proxy respondents. Smaller sample sizes or greater attrition among those with poor health and basic education would make it more difficult to identify inequalities in SRH in the oldest age groups.

In the individual Nordic studies, the prevalences of being independent in ADL and mobility were very high in the youngest age group especially in NorLAG and DLSA. Both studies showed higher independence for people with higher education but results did not reach statistical significance. The low prevalence of ADL problems in these two studies is related to the strong reliance on community based sampling, which implies a weaker representativeness of individuals not living independently. Kelfve (2017) has elaborated on the representativeness of study populations among older adults, and its effect on the observed magnitude of health inequalities. Her study showed that excluding institutionalised individuals, or those for whom someone else answered the questions (proxy respondent) led to underestimations of the health inequalities.

Our study showed stable inequalities in ADL and mobility until the oldest age group (95+), which is in line with Arber and Cooper (1999) and Schöllgen et al. (2010) but differs from Minkler et al. (2006). This comparison is not optimal since the previous studies had a focus on younger age groups. They also excluded institutionalised, which may explain the difference between this study and Minkler et al. (2006). In SRH analyses, between-study variability was higher, and unlike Rostad et al. (2009), we found slight decrease in inequalities towards higher age groups. One difference between studies was that we categorized education in two groups and the other study used three categories where the highest education group clearly differed from the lowest group. In studying very old people, who generally have a low level of education, the use of a dichotomized measure of education may be justified, although it may ignore some important variation in the social environment.

All five studies had good or excellent response rates. It is, however, known that the study non-respondents tend to have worse health than the study respondents and that poor health is associated with a lower level of education. Thus, it is likely that we underestimate inequalities and as such provide a lower bound for the extent of health inequalities in old age. Moreover, since people with higher education, on average, have longer lives (Moe et al. 2012), their proportion in the study sample is relatively higher than for people with basic education. Selective mortality eliminates frail and unhealthy individuals from a cohort, making initially disadvantaged groups appear compositionally advantaged over time and, thereby, attenuating or even reversing the initial associations (Ferraro and Shippee 2009). If this process occurs at a faster rate among those with lower education, compared to those with more education, this could lead to decreasing health inequalities in old age.

The advantages of this study were the use of synthesized data from four Nordic countries including high number of oldest old and the use of similar variables and methods of analyses. There were also limitations in the study. The study design was cross-sectional and descriptive in the sense that we were not accounting for several possible confounders such as intelligence or personality, that can affect both the level of education and health. Therefore, it is not possible to draw any conclusions about the causal nature of the associations based on our results. Our results show health inequalities across age groups at one time point. Results may reflect continuity of health inequalities with ageing but it can also reflect a cohort effect. In addition, as a methodological consideration, we assessed between-study heterogeneity with Q statistics and I-squared test, which both include uncertainty in a meta-analysis with a low number of studies. High between-study heterogeneity questions whether the summary estimate shows the right magnitude of inequalities since the consistency of the studies is weaker (Borenstein et al. 2010). The heterogeneity statistics from the meta-analysis coincide well with the known differences between the studies. Heterogeneity in SRH, when all age groups were included, is likely to reflect the smaller inequalities in the oldest age groups but heterogeneity that was found in all health outcomes, particularly in the age group 85–94, is related to the healthier study sample in the DLSA study. Thus, it is likely that also summary estimates are somewhat underestimates of health inequalities in this study.

This study showed the extent of inequalities in health and functioning among people aged 75+ in four Nordic countries. Despite the selection processes, health inequalities according to level of education persist in old age and, for mobility and ADL, also in the oldest age groups (95+). In the Nordic countries, known for their generous welfare systems, health inequalities have been observed in younger ages (Huijts et al. 2010), and this study suggests that they continue until the last years of life.