Health trajectories after age 60: the role of individual behaviors and the social context

Introduction

The biological changes underlying the aging process are neither linear nor consistent, and chronological age is an imperfect surrogate measure of this phenomenon. In fact, while some 80-year-olds may enjoy good physical and mental functioning, others may be frail, suffer from a high disease burden, and require significant support in their daily life. Moreover, older people with equal health status at earlier ages may follow different trajectories over time, and their needs will diverge according to which path they are on. Consequently, the health of the older population can be viewed as a continuum and dynamic process that needs to be investigated longitudinally [1].

When looking at trajectories of healthy aging, both what we measure and how we measure it matter equally. Most studies have focused on single impairments or functional limitations [2], however, it is unlikely that any single measure will be capable of reliably capturing the health heterogeneity of aging populations. There is increasing awareness for the need to assess older people’s health in a multidimensional way by using phenotypic indices that capture diseases related to organ systems as well as functional ability, while remaining pragmatically driven by available data in clinical practice and aging studies [3]. Comprehensive assessments of functioning, coupled with measures of more severe disability, are particularly relevant for older adults and are well-aligned with the provision of person-centered care [4]. Moreover, they are much stronger predictors of negative health outcomes than merely the presence of disease [5–7]. The Health Assessment Tool (HAT) used in this study is sensitive to changes in multiple dimensions that people experience as they age. It has a better ability than its individual components in terms of predicting medical and social care needs, self-rated health, or death, and it is based on conventional tests commonly embedded in the Comprehensive Geriatric Assessment (i.e. diseases, physical and cognitive function, and disability) [5, 6, 8].

As for how to measure health changes, capturing distinct developmental health trajectories over time is key, but also a major challenge. Most studies still look at average trajectories of change [9], but more sophisticated techniques, such as latent variable mixture modelling, use multiple long-term observations per individual and allow for improved discernment among alternative health trajectories in terms of rate and pattern of change. In order to apply such techniques, longitudinal data covering different birth cohorts, with low attrition rates, and containing objectively assessed health measures is essential, yet seldom available.

Differences in healthy aging are partly due to genetics and randomly occurring biological dysfunctions, but they are to a great extent influenced by individual health-related behaviors and contextual social determinants (i.e. socioeconomic and psychosocial factors). Despite the strong link of these factors with multimorbidity, frailty, disability, and mortality, [10–14] evidence of their impact on long-term health changes in old age is weaker [15]. Moreover, previous studies have rarely assessed potential risk/protective factors comprehensively, in spite of the increasing recognition of the complex interplay among social and behavioral factors in affecting older individuals’ health status and disease risk [16].

The aims of our study were to identify different health trajectories after age 60, and to explore to what extent health-related behaviors and the social context account for such differences at both the individual and population levels.

Materials and Methods

Results

The distribution of all demographic, socioeconomic, psychosocial, and behavioral variables for the total study sample is shown in Supplementary Table 1. In Table 1, the most prevalent values for each health indicator by baseline HAT score are shown. Figure 1A shows the results from the best-fit growth mixture model and the sample mean curve for the general population; 78% of the population was included in the best trajectory (i.e. green), with a faster decline from age 80 onwards; 18% of the population belonged to the middle trajectory (i.e. yellow) with lower initial HAT scores and a similar decline between 60-80 years, but a slower decline from age 80 onwards; and 4% of the population was categorized in the worst trajectory (i.e. red) with similar initial scores as the best trajectory but a much faster decline. For example, on average, a person aged 60 years in the green trajectory would decline 2.3 points in the HAT score across the subsequent 20 years, while a person of the same age in the red trajectory would decline 6.9 points across the same time period, even if both persons had similar HAT scores at age 60. Results from Laplace regressions showed that, compared to the best trajectory, subjects in the middle and worst ones lost their functional independence (i.e. HAT score <5) 12.0 (95% CI: 11.4-12.6) and 12.1 (95% CI: 11.5-12.7) years earlier, respectively. The age- and sex-adjusted median survival of subjects in the middle and worst trajectories was 4.28 (95% CI: 3.51-5.06) and 3.51 (95% CI: 2.52-4.50) years shorter than those belonging to the best trajectory, respectively.

Figure 1B, 1C show the results from the best-fit growth mixture models stratified by baseline age and sex, respectively. While only two trajectories were identified in the groups aged 60-70 (n=1251) and 80+ years (n=975) at baseline, three trajectories were identified in the group aged 70-80 years (n=882) at baseline. Three trajectories were also found for both males and females in the sex-stratified models. The best (i.e. green) and middle (i.e. yellow) trajectories showed steeper declines in females compared to males. Yet, the prevalence of the worst trajectory (i.e. red) was double in males compared to females. The parameter estimates and fitting criteria as well as their interpretation for all these models can be found in Supplementary Tables 2, 3.

Table 2 shows the odds ratios and 95% CI for all potential socioeconomic, psychosocial, and behavioral determinants. For model I (adjusted by baseline age, sex, and each variable individually), all factors were significantly associated with belonging to the middle/worst vs. best trajectory, but occupation, smoking, and social support lost their statistical significance in model II, which was further adjusted by other domain-specific factors. The effect of physical activity was largely attenuated in model III, when all factors were simultaneously adjusted for. Among all possible pairwise interactions, only those between manual occupation and high BMI (OR: 2.10, 95% CI: 1.06-4.15) and between manual occupation and low social connections (OR: 0.46, 95% CI: 0.23-0.90) were significant. The correlation matrix among all exposure variables is shown in Supplementary Table 4, and the odds ratios and 95% CI for the non-dichotomized exposures in Supplementary Table 5.

In stratified analyses, while the risk effect of low education, occupation, and financial strain; high BMI; and low social connections seemed to decrease with age, it increased for smoking and inadequate physical activity. Financial strain (OR: 4.33, 95% CI: 1.45-12.99) and inadequate physical activity (OR: 3.57, 95% CI: 2.59-4.91) were the factors with the highest risk effect for the youngest and oldest groups, respectively (Supplementary Figure 2A). Furthermore, the effect of financial strain was higher for males (OR: 3.18, 95% CI: 1.54-6.59) than for females (OR: 2.55, 95% CI: 1.46-4.44). However, the opposite was true for physical activity (OR: 3.56, 95% CI: 2.53-5.00 for females; OR: 3.11, 95% CI: 2.02-4.78 for males). Social participation was only significant for females (OR: 1.55, 95% CI: 1.10-2.20, Supplementary Figure 2B).

The study of the PAFs, as shown in Table 3, revealed that insufficient physical activity was the most important contributor (PAF: 27.1%, 95% CI: 20.4%-33.3%) for belonging to the middle/worst vs. best trajectory, followed by low education (PAF: 19.3%, 95% CI: 5.7%-31.0%) and low social participation (PAF: 15.9%, 95% CI: 2.2%-27.7%). Behavioral factors showed the highest domain-specific PAFs (PAF: 37.4%, 95% CI: 26.1%-47.1%). Socioeconomic (PAF: 24.7%, 95% CI: 11.3%-36.0%) and psychosocial factors (PAF: 23.8%, 95% CI: 9.3%-35.9%) had similar domain-specific PAFs, with percentages around two-thirds of that for behavioral factors. In total, 66.1% (95% CI: 54.1%-75.0%) of people in the middle/worst trajectories could potentially belong to the best trajectory, if all significant risk factors were eliminated simultaneously.

Discussion

In this study aiming to explore health changes in an older urban Swedish population, we identified three trajectories with marked health differences both at age 60 and over time, with 78%, 18%, and 4% of the study participants belonging to the best, middle, and worst trajectories, respectively. A mix of health-related behavioral factors (e.g. physical activity) and contextual social factors (e.g. education and social participation) contributed mostly to the levels and rates of change of health trajectories. In an extreme scenario where all associated life-long factors were eliminated, 66% of those in the middle and worst trajectories could potentially follow the same slow health decline as older adults belonging to the best trajectory.

Previous studies comparing trajectories of health changes over time have investigated the development of single-domain disorders, such as chronic diseases [22, 23] or disabilities [24, 25]. In this study we were able to incorporate several of these measures into a composite metric of healthy aging, truly reflecting people′s capacity to do what they value most, which is being able to move around, take care of themselves, and decide independently [1]. When looking at changes over time, we found that, compared to the best trajectory, which comprised approximately three quarters of the study population, subjects in the middle and worst trajectories became functionally dependent 12 years earlier, around the age of 78 years. A similar time gap was found between centenarians in SNAC-K, an example of successful agers, and their shorter-lived cohort counterparts concerning the occurrence of multimorbidity, disability, and cognitive impairment [26].

The age- and sex-stratified health trajectories found in our study were comparable to those seen in the general population, and further unraveled some interesting traits. In terms of age, we found that the health heterogeneity widens considerably in the 70 to 80 years' decade, as indicated by the increase in the number of classes identified in this age range. While previous research from our group [27] and others [28] has shown that until age 80, most people are free from functional impairment or disability despite the presence of chronic disorders, age 70 seems to be a key period when minority groups start to significantly deviate from average trends. The lower number of health trajectories identified in octogenarians and nonagenarians may be due to survival selection characterizing individuals reaching old ages. In terms of sex, females in the best and middle trajectories exhibited steeper health declines than males, which is in line with the disability-survival paradox whereby women suffer more frequently from nonfatal but disabling diseases, whereas men have fewer diseases in total but more of these are life-threatening [29].

The impact of socioeconomic, psychosocial, and behavioral factors on health changes in old age has already been examined, but most past work focused on single exposures and specific health dimensions [23, 25]. According to a recent systematic review on modifiable risk factors for healthy aging (understood as a multidimensional construct) by Kralj et al. [15], educational level is the most frequently explored socioeconomic determinant, with the majority of studies reporting a beneficial association between having higher levels of education and healthier aging. Consistently, elementary/high school vs. university education accounted for 19% of all cases belonging to the middle and worst trajectories; the highest PAF value among all studied determinants. However, we found an even stronger association between financial strain and belonging to the middle/worst vs. best trajectory. Besides entailing a better economic status, higher education may enable a more effective use of power, money, general and health-related literacy, prestige, or social support to protect oneself from health risks or mitigate the consequences of poor health [30]. Our results suggest that older people’s health is partly shaped by material factors such as income and wealth, which allow people to secure goods and services needed for a healthy life (e.g. housing, food, healthcare), but also by non-material resources and capabilities afforded by education. Our findings differ from two previous longitudinal multi-cohort aging studies, where education and wealth were consistently associated with cross-sectional health differences in old age, but not with their evolution over time [31, 32]. Notably, both these studies focused on average health changes over time, without taking interindividual differences into consideration.

Findings on the association between occupation and healthy aging remain inconclusive [15]. In our study, the association was no longer significant after including education and financial strain into the models. This is probably because, unlike the measures of wealth that reflect a process of life-long accumulation, occupation refers to more distant periods of time already accounted for by education.

More than one third of cases belonging to the middle and worst trajectories could be attributed to health-related behavioral determinants including BMI and physical activity (i.e. PAF: 37%). Despite being clearly influenced by economic and cultural status, health behaviors are often analyzed separately from other social determinants [16]. Inadequate levels of physical activity were strongly associated with belonging to poorer health trajectories in our study, and its effect size increased with increasing age. This is in line with studies indicating that, even in very advanced years, physical activity can have powerful benefits for health improvement and well-being [33, 34]. It is also in agreement with a recent study that found that health behaviors, even more than socioeconomic status, play a major role in the risk of dying among subjects with cardiometabolic multimorbidity [35]. Despite such evidence on the health benefits of physical activity, we are aware that reverse causation may also play a role in explaining our strong association, even if the longitudinal design of our study should help minimize this possibility. Moreover, previous population-based longitudinal research has shown reasonable stability in physical activity performance among older adults [36].

As in the present study, a consistent negative association between higher BMI in late life and healthy aging has been reported across the literature [15]. Data from the Whitehall II study suggests that the odds of aging successfully with a BMI ≥30 is almost three times lower compared to someone of normal weight [37]. The decreasing effect size of BMI at older ages could be related to its different clinical meaning in the youngest versus oldest old. In fact, a U-shaped association between BMI and mobility decline, disability, and all-cause mortality has been previously described among the oldest old [38]. In old age, a lower BMI may reflect sarcopenia, defined as the reduction in muscle mass that increases subjects’ risk of falls, disability, and mortality [39, 40]. Moreover, low BMI may indicate malnutrition and can be an expression of chronic disease severity [41].

As for smoking, the significance of the association between current smoking and belonging to a poorer health trajectory was lost when further adjusting for the rest of the behavioral factors, even though the direction was consistent throughout all models. According to the systematic review by Kralj et al. [15], up to 25% of studies reported a non-significant negative association of smoking with health changes in older age. This may be due to the health selection seen for current smokers across aging cohorts, even if, in our study, no significant differences were seen between age groups concerning the strength of the association.

Psychosocial determinants such as social connections, support, and participation are increasingly relevant to older people's views of successful aging [42]. It has been shown that these affect health via pathways such as provision of help with daily activities, material assistance, advice and information on particular needs, and cognitive and emotional stimulation [43]. Social relationships may also buffer the negative impact of stress on health [44]. In our study, a lower participation in social activities was the psychosocial factor most strongly associated with worse health trajectories, which is supported by a recent study looking at older people’s participation in community groups [45]. Women received a stronger benefit from social participation than men, which may indicate that more and different aspects related to social capital impact healthy aging among women than men [46].

Social connections, the second strongest psychosocial determinant, have consistently shown a strong protective association with health in old age [15]. Our results also revealed a significant association between low social connections and worse health trajectories, but the association became weaker in older age groups. This is probably due to the socio-selectivity theory whereby, as the end of life approaches, priorities shift from network size and diversity towards emotional components [47]. As for social support, the association with health trajectories lost significance in the fully adjusted model even if its direction was consistent across models, most likely due to the high correlation between social support and connections.

Considering that aging populations will comprise a growing proportion of our global world, we need to adapt the ways we live to maximize the opportunities that longevity presents. The COVID-19 pandemic has unraveled how vulnerable aging populations can become when faced with unexpected threats, highlighting the need to shift in the way we think about global aging. Ours and other studies provide evidence that can guide societies to develop healthy, empowered and resilient communities of older people throughout the life course. Of the behavioral and contextual factors assessed, the largest numbers of cases within the worse trajectories were potentially attributable to physical activity, education and social participation, suggesting that the most effective public health strategies and campaigns for ensuring healthy aging in Sweden should focus on enhancing the opportunities to build up on these factors throughout life. Still, the prioritization of the different socioeconomic, psychosocial, and behavioral determinants on which to act upon may vary by country or community, depending on their prevalence and strength of association with the health trajectories. Future research should inform when these interventions should take place throughout the life span, in order to optimize their effectiveness.

Conclusions

Poor health is not an inevitable consequence of survival to older ages. A number of factors linked to individuals’ health-related behaviors and social context, such as physical activity, financial strain, social participation, education, social connections, and BMI lead to important variations in older peoples’ health trajectories. By intervening on all these factors throughout life, up to 66% of people in the poorest health trajectories could be steered towards a healthier aging. Addressing the social determinants of health in its broadest sense, complementarily considering life-long factors belonging to the socioeconomic, psychosocial, and behavioral dimensions, should be central to any strategy aimed at fostering health in older age.

Supplementary Materials

Supplementary Figures

Supplementary Tables