Immediate and Longer-Term Changes in the Mental Health and Well-being of Older Adults in England During the COVID-19 Pandemic

Key Points Question How have the mental health and well-being of older adults in England changed during the COVID-19 pandemic compared with prepandemic levels? Findings This cohort study including 5146 older adults participating in the English Longitudinal Study of Ageing found that levels of depression, loneliness, and poor quality of life increased significantly during June and July 2020 compared with prepandemic levels and continued to deteriorate during the second national lockdown in November and December 2020, with further increases in anxiety symptoms from June and July 2020 to November and December 2020. Inequalities in experiences of mental ill health during the COVID-19 pandemic were evident, with women, individuals living alone, and those with less wealth being particularly vulnerable. Meaning Older individuals did not adapt well to the new psychosocial stressors introduced by the pandemic; policies should be in place for the immediate provision of targeted psychological interventions to support older people, and access to digital mental health services should be improved.

education ("low" = Compulsory School Leaving/ "medium" = A-levels & College/ "high" = Degree or above), employment status (employed/ retired/ other not working), home tenure (owns outright/ owns with mortgage/ rents), and limiting longstanding illness (no/ yes). Age and partnership status were measured at the first COVID-19 assessment, while wealth, education, employment status and limiting longstanding illness were determined in pre-pandemic assessments (i.e. wave 9 or 8). Further, we derived a binary variable indicating whether the participant had experienced COVID-19 at the first or second COVID-19 wave. The following criteria were applied to identify confirmed or suspected cases of COVID-19: participants were found to be COVID-19 positive on testing, or were hospitalised due to COVID-19, or reported two of the three core symptoms as defined by the UK National Health Service (NHS) (i.e. high temperature, a new continuous cough, and loss of sense of smell or taste).

Missing data
The percentage of missing data in the variables ranged between 0 and 6%. In addition, due to a survey error, for around 75% of the sample the last item of the CESD-8 questionnaire was not administered at the first COVID-19 wave. This type of missing data is classified as missing completely at random (MCAR), and can be dealt efficiently with multiple imputation (MI). 6 We used MI by chained equations with all variables of the analysis included as predictors of the imputation models as well as auxiliary variables. We created twenty imputed datasets, and then pooled the regression estimates across the imputed datasets using Rubin's rules. 7 The distribution of the variables in the imputed and observed data was similar, suggesting that the MI procedure produced accurate model estimates.

Equations of the fixed-effects regression models
(1) Changes in the outcomes before and during COVID-19
Note. Fixed-effects models do not estimate regression coefficients for time-invariant independent variables as their effects are controlled for by the individual fixed-effects.

Sensitivity analyses
First, we tested changes in depression and anxiety during the COVID-19 pandemic and variations with sociodemographic factors using the total CESD-8 and GAD-7 scores, rather than the binary scores representing cases of depression and anxiety. Second, we reran all models presented in the main imputed data analysis using the sample of participants with complete data on all variables. Third, we restricted the main analyses to participants who did not experience COVID-19 at either the first or second COVID-19 wave. Fourth, the changes in the mental health outcomes before and during COVID-19 were estimated using fixed-effects models with robust standard errors to account for potential heteroskedasticity and autocorrelation. Lastly, we examined changes in depression, quality of life, and loneliness before and during COVID-19 accounting for pre-pandemic trends in mental health from wave 4 (2008/09) to 9 (2018/19).

eResults. Sensitivity analyses
First, we tested changes in depression and anxiety during the COVID-19 pandemic and variations with sociodemographic factors using the total CESD-8 and GAD-7 scores, rather than the binary scores representing cases of depression and anxiety. The results mirrored those of the analysis with the binary scores (SI Appendix -sTable6 and sTable7). The change in the total score of depression before and during COVID-19 was smaller than the change in the binary depression score, but still considerably larger than the change observed for the total scores of quality of life and loneliness. The changes in the total and binary scores of anxiety were broadly similar (SI Appendix -sTable4 and sTable6). It is also worth noting that the predicted probabilities of depression and anxiety derived from the fixed-effects linear probability models were all within the range of 0-1 (see Figures 1 and 3), thereby providing further evidence for the adequacy of these models. Second, we reran all models presented in the main imputed data analysis using the sample of participants with complete data on all variables. The pattern of changes in mental health before and during COVID-19 and interaction effects with sociodemographic factors aligned closely with the results found in the main imputed analysis (SI Appendix -sTable8 and sTable9). Third, we restricted the main analyses to participants who did not experience COVID-19 at either the first or second COVID-19 wave. A total of 5ꞏ4% experienced COVID-19 on the basis of testing, hospitalisation, or symptoms, leaving 4867 (94ꞏ6%) participants in these analyses. There were no substantial changes from the primary analyses in the magnitude and statistical significance of the associations (SI Appendix -sTable10 and sTable11). Fourth, the changes in the mental health outcomes before and during COVID-19 were estimated using fixed-effects models with robust standard errors to account for potential heteroskedasticity and autocorrelation. The results were almost identical to those found in the main analysis (sTable12). Lastly, we examined changes in depression, quality of life, and loneliness before and during COVID-19 accounting for pre-pandemic trends in mental health from wave 4 (2008/09) to 9 (2018/19). Anxiety was not included in this sensitivity analysis since the GAD-7 scale was not administered in the regular ELSA survey. The observed percentages/means of the mental health outcomes from wave 4 through to COVID-19 wave 2 are presented in sTable13. The estimated changes in mental health during COVID-19 (wave 1 and 2) versus before (wave 9) adjusted for earlier trends (wave 4 to 8) aligned closely with those observed in the main analyses (sTable14, Model 1). sFigure3 shows the estimated trajectories of depression, poor quality of life, and loneliness from wave 4 to COVID-19 wave 2. As can be seen, the increase in psychological distress during COVID-19 occurred against a slight downward trend in all mental health outcomes over the preceding years. Of note, the estimated prevalence of depression in the sample during COVID-19 was also considerably larger than the average depression prevalence over the preceding 11/12 years (sTable14, Model 2), providing corroborative evidence of marked increases in depressive symptoms during COVID-19. eTable 1. Comparison of the characteristics of ELSA participants included in the analytical sample vs those not included at wave 9