Socioeconomic inequalities in physical activity among older adults before and during the COVID-19 pandemic: evidence from the English Longitudinal Study of Ageing

Introduction

The global population of older adults aged 60+ years is projected to double from 1 billion in 2020 to 2.1 billion by 2050.1 The health implications of population ageing have commonly been framed in negative terms, with older adults depicted as a social and economic burden.2 Yet, substantial interindividual variability exists in the health and functional status of older adults, which is only loosely associated with chronological age.2 3 While more work is needed to understand these variations, there is clear evidence of the importance of maintaining healthy lifestyle behaviours in older age, particularly physical activity.2 It is well documented that regular physical activity in older adults plays a critical role in the prevention of chronic disease, preservation of physical independence and improvement of quality of life.4 5 However, the majority of older adults in England do not meet recommended aerobic or muscle-strengthening physical activity guidelines.6

Health disparities in older age are often a consequence of cumulative advantages or disadvantages experienced over an individual’s life course.7 Although a variety of measures have been used to characterise socioeconomic status in the literature, some of the most common individual-level indicators include education, occupational class and income/wealth.8–10 Interestingly, these indicators may be associated with different types of physical activities, suggesting that multiple indicators of socioeconomic status should be considered in physical activity research.11 12 Individuals classified as being of higher socioeconomic status (according to diverse indicators of socioeconomic position) consistently report higher physical activity levels than individuals of lower socioeconomic status across the lifespan.11 12 This socioeconomic gradient in physical activity participation widens in older age.13 Furthermore, research suggests that older adults of lower socioeconomic status experience a greater number of individual and environmental barriers to physical activity than the general older adult population.14 Nevertheless, despite being the least active of all adult groups, older adults of low socioeconomic status remain largely absent from the physical activity literature.6 15

Since the first recorded case of the SARS-CoV-2 in December 2019, the lives of many people have been disrupted.16 Older adults are disproportionately vulnerable to the physiological risks of infection from the coronavirus disease (COVID-19), as well as the psychosocial impacts of distancing and lockdown, such as loneliness and social exclusion.17 18 Moreover, lockdown regulations and social isolation during the COVID-19 pandemic have likely contributed to a decline in physical activity among older adults.19 20 A recent study, using data from a large representative sample of the English population (n=726 257, aged 16+ years) participating in the Sport England Active Lives Surveys, found that the odds of reporting physical activity were approximately 30% lower during the first national lockdown (April to May 2020), compared with the same time period in previous years; however, the magnitude of these declines differed across sociodemographic groups.21

Physical inactivity in older adults is associated with numerous health risks, including more severe COVID-19 outcomes among infected individuals.22 23 Therefore, a deeper understanding of changes in older adults’ physical activity levels during the COVID-19 pandemic is warranted.19 23 Emerging evidence suggests the COVID-19 pandemic may have exacerbated socioeconomic inequalities in physical activity.24 25 While the mechanisms underlying these associations have not yet been explored, it is possible that a range of psychosocial (eg, higher social participation) and environmental factors, known to be important mediators in explaining prepandemic socioeconomic differences in physical activity, may have helped individuals of higher socioeconomic status to maintain healthy lifestyle behaviours during the COVID-19 pandemic.24–26 Importantly, it remains unclear whether the widening socioeconomic gradients in physical activity participation observed in previous studies conducted during the COVID-19 pandemic are mirrored among the older adult population in England.

The aim of this study was to investigate associations between indicators of socioeconomic status (ie, education, occupational class and wealth), physical activity levels and change over time among older adults in England, using data collected before and during the COVID-19 pandemic. We hypothesised that individuals of higher socioeconomic status would present higher physical activity levels at the prepandemic and intrapandemic assessments, and that socioeconomic inequalities would increase during the COVID-19 pandemic.

Materials and methods

Results

Associations between socioeconomic variables and physical activity before and during the COVID-19 pandemic

Results for models 1–4 are reported in table 2. In models 1–3, there was a clear gradient in physical activity from highest to lowest education, occupational class and wealth at baseline (all p≤0.01). The odds of engaging in physical activity for participants of high (interaction term: OR=0.64, p=0.006) versus low education (model 1), and in higher (interaction term: OR=0.67, p=0.003) versus routine and manual occupations (model 2), decreased prepandemic to intrapandemic. Baseline associations between socioeconomic variables and physical activity largely remained but were attenuated in the mutually adjusted model (model 4). Patterns of association for interactions between education or occupational class and time were maintained, although statistical significance was lost (all p>0.05). Older participants, respondents of non-white ethnic origin, as well as those reporting a limiting long-standing illness, disability, or infirmity, more depressive symptoms, or shielding during the COVID-19 pandemic had significantly lower odds of physical activity in all tested models (all p≤0.05).

Table 2

View inline|View popup

Multilevel ordered logistic model of physical activity at prepandemic and intrapandemic across socioeconomic groups, adjusted for covariates

The predictive margin probabilities (expressed as percentages) across the four levels of the ordinal physical activity outcome variable before and during the COVID-19 pandemic, by socioeconomic groups, are presented in online supplemental table S8. The collapsed (for ease of visualisation) predicted probabilities of engaging in moderate or vigorous physical activity at prepandemic and intrapandemic across socioeconomic groups are represented graphically in figure 1. Participants with higher education, occupations and wealth had greater probabilities of reporting moderate or vigorous physical activity at both survey periods, suggesting that socioeconomic inequalities persisted during the COVID-19 pandemic.

Request permission

Figure 1

Predicted probabilities of moderate or vigorous physical activity before and during the COVID-19 pandemic by education (A), occupational class (B) and wealth (C). Predicted values were derived from model 1 for education (A), model 2 for occupational class (B) and model 3 for wealth (C). Error bars indicate 95% CIs. Predicted values were computed using the ‘predict (mu fixedonly)’ option to fix the random effects of the respective multilevel ordered logistic model to zero, the ‘asobserved’ option for covariates and the ‘vce(unconditional)’ option.

Discussion

In this study, we investigated associations between socioeconomic variables and physical activity before and during the COVID-19 pandemic (when the second national lockdown restrictions and social distancing regulations were in place in the UK) in a sample of older adults living in England, aged 60+ years. Although there was no statistically significant difference in physical activity levels before versus during the COVID-19 pandemic in the multilevel models, the descriptive statistics suggested a slight decrease over time, with more participants classified as inactive (12.5% vs 5.7%) at follow-up. As hypothesised, we found evidence of pre-existing socioeconomic inequalities in physical activity based on education, occupational class and wealth. Results also suggested that socioeconomic disparities persisted during the COVID-19 pandemic. However, in contrast to our hypothesis, there was inconclusive evidence regarding any socioeconomic differences in physical activity change between the two survey periods. All models controlled for biological sex, age, ethnicity, living status, the presence of any limiting long-standing illness, disability, or infirmity, depressive symptoms, and shielding during the COVID-19 pandemic. The main findings remained unchanged when analyses were performed using multiple imputation.

Our result that older adults with higher education, occupational class and wealth had increased odds of engaging in physical activity before the lockdown agrees with previous work.11 12 This study contributes to the extant literature by providing evidence of persistent socioeconomic disparities in older adults’ physical activity levels during the COVID-19 pandemic, a finding that may extend to other age groups.25 Although the separate models showed some evidence of decreasing socioeconomic gradients in physical activity based on education and occupational class, these results were not observed in the mutually adjusted model, where socioeconomic variables and their interactions with time corrected for one another. Interestingly, previous studies found that socioeconomic inequalities (based on education, occupational class and income) in physical activity were exacerbated during the COVID-19 pandemic.25 36 Nevertheless, these studies used different physical activity measurement instruments and examined associations in samples across the adult age range. It is therefore possible that socioeconomic indicators are less pertinent explanatory variables of physical activity change in older adults. Another plausible explanation is that younger adults were more reliant on the physical activities most affected by the COVID-19 pandemic and showed declines in activity levels due to changes in employment status (eg, furlough) or childcare responsibilities among other factors.21 These postulations warrant confirmation in future studies. In addition, age, ethnicity, the presence of any limiting long-standing illness, disability, or infirmity, depressive symptoms, and shielding during the COVID-19 pandemic were important covariates and may be useful candidates for interaction with indicators of time or socioeconomic status in further research.

This study was strengthened by the large, nationally representative sample of older adults in ELSA, as well as the longitudinal design which enabled the comparison of prepandemic and intrapandemic associations. Nevertheless, our findings should be interpreted in light of several limitations. Notably, in longitudinal studies with older adults, considerable attrition occurs due to death, which is only partially corrected by applying weights to statistical analyses.33 37 Second, the follow-up assessment for physical activity took place in November/December 2020. However, the starkest changes in physical activity may have occurred earlier in the COVID-19 pandemic during the critical phase of the national lockdown.20 38 Although we performed hierarchical logistic regressions to examine departure from consistent patterns of association between the explanatory variables and physical activity, we could not formally test whether the proportional odds assumption for models with ordinal outcome variables was violated. However, analyses with the binary physical activity outcomes, as well as the relaxed proportional odds models (available in the online supplemental materials), suggested a similar pattern of results to the multilevel ordinal logistic proportional odds models.

This study relied on self-report measures, which are prone to recall and social desirability biases. In particular, the ordinal categorisation of physical activity obscures small fluctuations within categories over time, whereas fluctuations between categories are reported as behavioural changes.25 Future research may therefore benefit from using a range of subjective and objective measures, to enable a more nuanced understanding of older adults’ physical activity behaviour.39 40 Moreover, while we controlled for numerous covariates, the influence of unmeasured or residual confounding variables cannot be dismissed. We treated socioeconomic factors and covariates as time-constant variables. Although this is common practice, it is plausible that participants’ income or occupational position fluctuated during the COVID-19 pandemic.33

ELSA participants were predominantly white, limiting generalisability to other populations. We found preliminary evidence that non-white participants displayed lower physical activity levels than white older adults. Nevertheless, research with minority ethnic and racial groups is necessary to replicate these findings.37 This is particularly important given that ethnic minority groups in the UK have been disproportionately affected by the COVID-19 pandemic.41 Furthermore, as years of educational attainment (which would have allowed the inclusion of older adults with foreign qualifications) were not available in ELSA, data relating to participants’ highest educational qualifications were used. While the three-class ordinal National Statistics Socio-Economic Classification necessarily excludes participants who had never worked and were long-term unemployed from analyses, which could have affected our results, it is worth noting this response only applied to 0.5% of our sample.

Our findings have several implications. Notably, the results inform clinicians, policymakers and practitioners about socioeconomic subgroups that require targeting via intervention. In the context of restrictions on social contact, remotely delivered physical activity interventions may be an area of future research.20 A recent analysis showed that the magnitude of change in adults’ physical activity levels during the COVID-19 pandemic in England differed across activity modalities and demographic groups.21 Therefore, studies should explore whether these findings are applicable to older adults of varying socioeconomic status. Given the sustained influence of the COVID-19 pandemic on individuals’ lives and behaviours, it will be essential to develop policies and health promotion strategies to support older adults to perform physical activity at home or in a limited space.42 43 Moreover, the results reiterate the need for further work eliciting the mechanisms underlying associations between socioeconomic status and physical activity. In a qualitative study, health limitations, neighbourhood safety and lack of knowledge of physical activity guidelines were cited as the most prominent barriers among a subsample of older adults of low socioeconomic status in the UK.14 As such, identifying modifiable psychosocial and environmental factors differentially associated with physical activity behaviour in older adults of varying socioeconomic status, during and beyond the COVID-19 pandemic, should be a research priority.

Overall, the current study provides evidence for socioeconomic disparities in the physical activity levels of older adults in England, which largely persisted during the COVID-19 pandemic. Our findings emphasise the need to instate policies for the provision of targeted interventions to support physical activity in older adults, considering differences across socioeconomic groups in their design and implementation. Future research should replicate these findings over a longer follow-up period in cohorts with varying ethnic and sociocultural backgrounds to improve generalisability.