Changes in subjective wellbeing over time
Changes in general subjective wellbeing. A repeated measures ANOVA with time as within-subjects factor on the general subjective wellbeing scores measured on a scale from 0 to 100 (see Fig. 2b) showed a main effect of time (F(3,368) = 73.34, p < .001, \({\eta }_{p}^{2}\)=.37). Post-hoc paired samples t-tests with Bonferroni correction (α = .008) showed that general subjective wellbeing was significantly higher in the period before COVID-19 (M = 79.50) compared to M1 (M = 72.96, t(370)=14.49, p<.001, d = 0.75), M2 (M = 75.11, t(370)=9.75, p<.001, d = 0.51) and M3 (M = 74.14, t(370)=10.67, p<.001, d = 0.55). General subjective wellbeing on M1 (M = 72.96) was significantly lower than on M2 (M = 75.11, t(370)=-4.95, p < .001, d=-0.26). General subjective wellbeing did not differ significantly between M1 and M3 and between M2 and M3 (p≥.010).
Changes in the different subdomains of subjective wellbeing. Repeated measures ANOVAs with time as within-subjects factor and each of the PWI-A subdomains (i.e., general life satisfaction, satisfaction with standard of living, health, achieving in life, relationships, safety, community connectedness and future security) as outcome were conducted. Table 3 provides a full description of all analyses and accompanying post-hoc tests with Bonferroni correction to follow-up significant effects. Generally, ratings of wellbeing for the subdomains were significantly higher before COVID-19 compared to all three measurement moments. Moreover, ratings on M1 were significantly lower than on M2. Ratings on M1 vs. M3 and M2 vs. M3 did not differ significantly from each other for most wellbeing subdomains. Mean ratings for each subdomain of wellbeing for the different measurement moments are included in Table 2.
Table 3
Statistical results of the repeated measures ANOVAs with the PWI-A subdomains as outcome and time as within-subject factor, and the accompanying Bonferroni corrected post-hoc tests.
PWI-A subdomain | Main effect of time | Post-hoc comparisons (Bonferroni correction α = .008) |
General life satisfaction | F(3,368) = 56.42, p < .001, \({\eta }_{p}^{2}\)=.32 | Pre vs. M1: t(370) = 12.20, p < .001, d = 0.63 |
Pre vs. M2: t(370) = 8.18, p < .001, d = 0.43 |
Pre vs. M3: t(370) = 9.78, p < .001, d = 0.51 |
M1 vs. M2: t(370)=-5.21, p < .001, d=-0.27 |
M1 vs. M3: t(370)=-3.15, p = .002, d=-0.16 |
M2 vs. M3: p = .076 |
Standard of living | F(3,368) = 11.34, p < .001, \({\eta }_{p}^{2}\)=.085 | Pre vs. M1: t(370) = 5.66, p < .001, d = 0.29 |
Pre vs. M2: p = .077 |
Pre vs. M3: p = .32 |
M1 vs. M2: t(370)=-3.12, p = .002, d=-0.16 |
M1 vs. M3: t(370)=-3.41, p < .001, d=-0.18 |
M2 vs. M3: p = .47 |
Health | F(3,368) = 14.56, p < .001, \({\eta }_{p}^{2}\)=.11 | Pre vs. M1: t(370) = 5.61, p < .001, d = 0.29 |
Pre vs. M2: p = .089 |
Pre vs. M3: t(370) = 4.28, p < .001, d = 0.22 |
M1 vs. M2: t(370)=-2.68, p = .008, d=-0.14 |
M1 vs. M3: p = .54 |
M2 vs. M3: t(370) = 3.08, p = .002, d = 0.16 |
Achieving in life | F(3,368) = 14.71, p < .001, \({\eta }_{p}^{2}\)=.11 | Pre vs. M1: t(370) = 6.33, p < .001, d = 0.33 |
Pre vs. M2: t(370) = 2.75, p = .006, d = 0.14 |
Pre vs. M3: t(370) = 4.41, p < .001, d = 0.23 |
M1 vs. M2: t(370)=-3.11, p = .002, d=-0.16 |
M1 vs. M3: p = .21 |
M2 vs. M3: p = .050 |
Relationships | F(3,368) = 34.97, p < .001, \({\eta }_{p}^{2}\)=.22 | Pre vs. M1: t(370) = 9.63, p < .001, d = 0.50 |
Pre vs. M2: t(370) = 6.31, p < .001, d = 0.33 |
Pre vs. M3: t(370) = 6.30, p < .001, d = 0.33 |
M1 vs. M2: t(370)=-3.24, p = .001, d=-0.17 |
M1 vs. M3: p = .020 |
M2 vs. M3: p = .39 |
Safety | F(3,368) = 60.88, p < .001, \({\eta }_{p}^{2}\)=.33 | Pre vs. M1: t(370) = 12.97, p < .001, d = 0.67 |
Pre vs. M2: t(370) = 9.89, p < .001, d = 0.51 |
Pre vs. M3: t(370) = 9.30, p < .001, d = 0.48 |
M1 vs. M2: t(370)=-3.30, p = .001, d=-0.17 |
M1 vs. M3: t(370)=-3.22, p = .001, d=-0.17 |
M2 vs. M3: p = .97 |
Community connectedness | F(3,368) = 58.59, p < .001, \({\eta }_{p}^{2}\)=.32 | Pre vs. M1: t(370) = 11.74, p < .001, d = 0.61 |
Pre vs. M2: t(370) = 9.77, p < .001, d = 0.51 |
Pre vs. M3: t(370) = 11.22, p < .001, d = 0.58 |
M1 vs. M2: p = .046 |
M1 vs. M3: p = .78 |
M2 vs. M3: p = .032 |
Future security | F(3,368) = 72.54, p < .001, \({\eta }_{p}^{2}\)=.37 | Pre vs. M1: t(370) = 14.17, p < .001, d = 0.74 |
Pre vs. M2: t(370) = 9.81, p < .001, d = 0.51 |
Pre vs. M3: t(370) = 10.40, p < .001, d = 0.54 |
M1 vs. M2: t(370)=-4.67, p < .001, d=-0.24 |
M1 vs. M3: p = .024 |
M2 vs. M3: p = .022 |
To shortly summarize, after the negative impact of the first peak of the pandemic, subjective wellbeing fluctuated with fluctuating severity of the pandemic. Results for cognitive functioning were mixed. While participants indicated a slightly better general subjective cognitive functioning at the end of the study, similar to the level of cognitive functioning before the pandemic, problems in different subdomains of cognitive functioning significantly increased towards the last measurement moment.
The influence of protective and vulnerability factors on the changes in subjective cognitive functioning and wellbeing
Table 4 displays the mean scores for the protective and vulnerability factors included in the current study.
Table 4
Mean scores and standard deviations for the protective and vulnerability factors included in the current study.
| | M1 | M2 | M3 |
| Range | n | Mean (SD) | n | Mean (SD) | n | Mean (SD) |
Cognitive failures (CFQ) | 0–100 | 364 | 22.61 (11.25) | 367 | 24.50 (11.40) | 369 | 26.85 (11.86) |
Depressive symptoms (GDS-15) | 0–15 | 371 | 2.60 (2.66) | 371 | 2.59 (2.80) | 371 | 2.77 (2.80) |
Social network (LSNS-6) | 0–30 | 371 | 17.54 (5.22) | - | - | - | - |
Resilience (BRS) | 1–5 | 371 | 3.40 (0.65) | - | - | - | - |
Anxiety symptoms (HADS) | 0–21 | - | - | - | - | 371 | 4.22 (3.65) |
Note. CFQ = Cognitive Failures Questionnaire total score; GDS-15 = Geriatric Depression Scale-15 total score; LSNS-6 = Lubben Social Network Scale-6 total score; BRS = Brief Resilience Scale mean score; HADS = Hospital Anxiety and Depressive symptoms sum score on the anxiety items. |
The influence of moderators on subjective cognitive functioning.
A repeated measures ANOVA was conducted with subjective cognitive functioning as outcome, time as within-subject factor, gender, living assisted or not and living alone or not as between-subject factors and age, monthly net income, frequency of cognitive failures, depressive symptoms, social network, resilience and anxiety symptoms as covariates. This analysis showed significant main effects of frequency of cognitive failures (F(1, 349) = 116.7011, p < .001, \({\eta }_{p}^{2}\)=.25), depressive symptoms (F(1, 349)=6.79, p=.010, \({\eta }_{p}^{2}\)=.020) and anxiety symptoms (F(1, 349)=7.01, p=.008, \({\eta }_{p}^{2}\)=.020). As can be seen on Supplementary Fig. 1, participants with a higher frequency of cognitive failures (panel a), higher depressive (panel b) or anxiety symptoms (panel d) show overall lower ratings of subjective cognitive functioning. This pattern is the most prominent for those participants with the highest cognitive failures and depressive symptoms (i.e., bin 4). Moreover, significant interaction effects between time and frequency of cognitive failures (F(2, 348)=5.03, p = .007, \({\eta }_{p}^{2}\)=.028), time and depressive symptoms (F(2, 348)=19.35, p<.001, \({\eta }_{p}^{2}\)=.10), time and social network (F(2, 348)=4.25, p=.015, \({\eta }_{p}^{2}\)=.024) and time and anxiety symptoms (F(2, 348) = 10.98, p < .001, \({\eta }_{p}^{2}\)=.059), were present. None of the other main effects and interactions were significant (p≥.075). To interpret the four significant interactions, we first used one-way ANOVAs to compare the M3-Pre difference score for cognitive functioning (i.e., subjective cognitive functioning measured on M3 - before the pandemic) between the different bins of the covariate. This allowed us to study whether changes in cognitive functioning from pre-COVID to M3 were related to varying levels of the covariate. Table 5 contains the mean difference scores for cognitive functioning for each bin depending on the protective or vulnerability factor. If this analysis was not sufficient (i.e., the ANOVA was not significant), we explored the difference score for cognitive functioning between pre-COVID and M2 and between M2 and M3 in order to capture the interaction.
Table 5
Mean difference scores (M3-Pre) and standard deviations for the different bins of the significant protective and vulnerability factors for changes in subjective cognitive functioning and subjective wellbeing over time.
| Protective or vulnerability factor | Bin 1 | Bin 2 | Bin 3 | Bin 4 |
n | Mean (SD) | n | Mean (SD) | n | Mean (SD) | n | Mean (SD) |
Subjective cognitive functioning | Cognitive failures | 92 | -0.13 (1.10) | 99 | 0.051 (1.06) | 88 | 0.00 (0.98) | 85 | 0.26 (1.51) |
Depressive symptoms | 163 | 0.055 (0.94) | 80 | 0.28 (1.07) | 42 | 0.24 (0.69) | 86 | -0.78 (1.50) |
Social network | 103 | -0.52 (1.46) | 102 | -0.10 (1.09) | 81 | 0.32 (1.00) | 85 | 0.14 (0.76) |
Anxiety symptoms | 101 | 0.31 (0.88) | 126 | 0.095 (0.98) | 52 | -0.058 (1.36) | 92 | -0.72 (1.30) |
Subjective wellbeing | Cognitive failures | 92 | -5.59 (9.35) | 99 | -5.38 (9.29) | 88 | -4.53 (8.60) | 85 | -5.83 (11.35) |
Depressive symptoms | 163 | -2.81 (6.85) | 80 | -3.46 (8.78) | 42 | -4.97 (10.90) | 86 | -12.16 (11.29) |
Resilience | 114 | -7.56 (12.25) | 35 | -3.35 (7.89) | 101 | -4.53 (8.60) | 121 | -4.58 (7.87) |
Anxiety symptoms | 101 | -1.43 (6.33) | 74 | -2.30 (7.39) | 112 | -5.82 (9.39) | 63 | -13.62 (10.67) |
Interaction between time and frequency of cognitive failures. A one-way ANOVA with M3-Pre difference score for cognitive functioning as dependent variable and cognitive failures bins (i.e. based on the CFQ score) as between-subject factor was not significant, F(3,360) = 1.26, p = .29, \({\eta }_{p}^{2}\)=0.010. Therefore, to further explore the interaction, one-way ANOVAs with the M2-Pre (F(3,360)=5.95, p<.001, \({\eta }_{p}^{2}\)=0.047) and the M3-M2 (F(3,360)=0.91, p=.44, \({\eta }_{p}^{2}\)=0.008) difference score as dependent variable and cognitive failures bins as between-subject factor were conducted as well. Post-hoc independent samples t-tests with Bonferroni correction (α = .008) for the significant ANOVA with difference score M2-Pre showed that participants with the highest frequency of cognitive failures (bin 4, M=-0.51) had a significantly steeper decrease in subjective cognitive functioning from before the pandemic to M2 compared to participants with the lowest (bin 1, M=-0.16, t(132.94)=3.23, p=.002, d=0.50) and second lowest frequency of cognitive failures (bin 2, M=-0.12, t(137.63) = 3.58, p<.001, d = 0.55). This is visually presented in Supplementary Fig. 1a. All other comparisons were not significant (p≥.057).
Interaction between time and depressive symptoms. The one-way ANOVA with M3-Pre difference score for cognitive functioning as dependent variable and depressive symptoms bins (i.e. based on the GDS-15 score) as between-subject factor was significant, F(3,367) = 16.31, p < .001, \({\eta }_{p}^{2}\)=0.12. Post-hoc independent samples t-tests with Bonferroni correction (α = .008) showed that participants with the highest depressive symptoms (bin 4, M=-0.78) differed significantly in cognitive function difference score from participants with the lowest depressive symptoms (bin 1, M=0.055, t(121.57)=4.70, p<.001, d=0.72), participants with the second lowest depressive symptoms (bin 2, M=0.28, t(153.78)=5.25, p<.001, d=0.81) and participants with the second highest depressive symptoms, (bin 3, M=0.24, t(125.73)=5.25, p<.001, d=0.79). Whereas participants with the highest depressive symptoms showed a negative difference score, indicating a decrease in subjective cognitive functioning over time, all other bins showed slightly positive difference scores, or increases in cognitive functioning over time. This is visually presented in Supplementary Fig. 1b. All other comparisons were not significant (p≥.10).
Interaction between time and social network. The one-way ANOVA with M3-Pre difference score for cognitive functioning as dependent variable and social network bins (i.e. based on the LSNS-6 score) as between-subject factor was significant, F(3,367) = 9.83, p < .001, \({\eta }_{p}^{2}\)=0.074. Post-hoc independent samples t-tests with Bonferroni correction (α = .008) showed that participants with the lowest social support (bin 1, M=-0.52) differed significantly in cognitive functioning difference score from participants with the second highest (bin 3, M=0.32, t(178.72)=-4.65, p<.001, d=-0.66) and highest social support (bin 4, M = 0.14, t(158.91)=-4.01, p < .001, d=-0.56). Moreover, the difference score of participants with the second lowest social support (bin 2, M=-0.10) was significantly different from the score of participants with the second highest social support (bin 3, M = 0.32, t(181)=-2.67, p=.008, d=-0.40). Whereas participants in the two lowest social support bins showed decreases on M3 in subjective cognitive functioning compared to before the pandemic, cognitive functioning increased for the two highest social support bins. This is visually presented in Supplementary Fig. 1c. All other comparisons were not significant (p≥.019).
Interaction between time and presence of anxiety symptoms. The one-way ANOVA with M3-Pre difference score for cognitive functioning as dependent variable and anxiety symptoms bins (i.e. based on the HADS score) as between-subject factor was significant, F(3,367) = 15.40, p < .001, \({\eta }_{p}^{2}\)=0.11. Post-hoc independent samples t-tests with Bonferroni correction (α = .008) showed that participants with the highest anxiety symptoms (bin 4, M=-0.72) showed a significant negative difference score indicating a decrease in cognitive functioning compared to participants with the lowest (bin 1, M=0.31, t(157.55)=6.34, p<.001, d=0.93) and second lowest (bin 2, M=0.095, t(161.96)=5.03, p<.001, d=0.72) anxiety symptoms, who even show a slightly positive difference score indicating an increase in cognitive functioning. Moreover, participants with the highest anxiety symptoms (bin 4, M=-0.72) showed a significantly more negative difference score and thus a steeper decrease in cognitive functioning compared to participants with the second highest anxiety symptoms (bin 3, M=-0.058, t(142)=2.87, p=.005, d = 0.72). This is visually presented in Supplementary Fig. 1d. All other comparisons were not significant (p≥.084).
To shortly summarize, changes in subjective cognitive functioning from before the pandemic to M3, during the second peak of the pandemic, were negatively influenced by a high frequency of cognitive failures, depressive and anxiety symptoms and low social support.
The influence of moderators on subjective wellbeing.
The same repeated measures analyses conducted on general subjective wellbeing as outcome showed significant main effects of frequency of cognitive failures (F(1, 349) = 13.03, p < .001, \({\eta }_{p}^{2}\)=.036), depressive symptoms (F(1, 349)=37.00, p<.001, \({\eta }_{p}^{2}\)=.096), social network (F(1, 349)=17.73, p<.001, \({\eta }_{p}^{2}\)=.048), resilience (F(1, 349)=21.44, p < .001, \({\eta }_{p}^{2}\)=.058) and anxiety symptoms (F(1, 349)=28.87, p<.001, \({\eta }_{p}^{2}\)=.076). As can be seen on Supplementary Fig. 2, higher frequencies of cognitive failures, more depressive and anxiety symptoms and lower resilience, were related to lower overall subjective wellbeing scores. Again, this effect seemed to be the most pronounced for participants with the lowest scores on resilience (i.e. bin 1) and the highest scores on frequency of cognitive failures, depressive and anxiety symptoms (i.e., bin 4). Moreover, significant interaction effects between time and frequency of cognitive failures (F(3, 347)=3.90, p=.009, \({\eta }_{p}^{2}\)=.033), time and depressive symptoms (F(3, 347) = 22.18, p < .001, \({\eta }_{p}^{2}\)=.16), time and resilience (F(3, 347)=4.12, p=.007, \({\eta }_{p}^{2}\)=.034) and time and anxiety symptoms (F(3, 347)=30.29, p<.001, \({\eta }_{p}^{2}\)=.21) were present. None of the other main effects and interactions were significant (p≥.075). Again, these interactions were further interpreted by comparing the M3-Pre difference score for cognitive functioning (and M1-Pre, M2-M1, M3-M2, M2-Pre and M3-M1 if the M3-Pre comparison does not allow us to interpret the interaction) between the different bins or levels of the covariate. In Table 5 the mean M3-Pre difference scores for subjective wellbeing for each bin depending on the protective or vulnerability factor can be found.
Interaction between time and frequency of cognitive failures. The one-way ANOVA with the M3-Pre difference score for wellbeing as dependent variable and cognitive failures bins (i.e. based on the CFQ score) as between-subject factor was not significant, F(3,360) = 0.29, p = .83, \({\eta }_{p}^{2}\)=0.002. Therefore, to further explore the interaction, one-way ANOVA’s with the M1-Pre (F(3,360)=1.42, p=.24, \({\eta }_{p}^{2}\)=0.012), the M2-M1 (F(3,360)=0.24, p=.87, \({\eta }_{p}^{2}\)=0.002), the M3-M2 (F(3,360) = 0.54, p=.65, \({\eta }_{p}^{2}\)=0.005), the M2-Pre (F(3,360)=0.55, p=.65, \({\eta }_{p}^{2}\)=0.005) and the M3-M1 (F(3,360)=1.03, p=.38, \({\eta }_{p}^{2}\)=0.008) difference scores as dependent variable and cognitive failures bins as between-subject factor were conducted as well. However, none of these comparisons reached significance implying that they could not aid in further explaining the significant interaction in the repeated measures ANOVA. Even though this interaction between time and cognitive failures proved to be significant in the repeated measures ANOVA, none of the post hoc tests could explain this significant interaction and visually (see Supplementary Fig. 2a) the interaction is not clearly observable in the data as well.
Interaction between time and presence of depressive symptoms. The one-way ANOVA with the M3-pre difference score for wellbeing as dependent variable and depressive symptoms bins (i.e. based on the GDS-15 score) as between-subject factor was significant, F(3,367) = 22.17, p < .001, \({\eta }_{p}^{2}\)=0.15. Post-hoc independent samples t-tests with Bonferroni correction (α = .008) showed that participants with the highest depressive symptoms (bin 4, M=-12.16) had a significantly more negative difference score, and thus a steeper decrease in wellbeing since before the pandemic, compared to participants in the lower bins (bin 1, M=-2.81, t(118.88)=7.03, p<.001, d=1.08; bin 2, M=-3.46, t(159.12)=5.56, p<.001, d=0.86; bin 3, M=-4.97, t(126)=3.42, p<.001, d = 0.64). This is visually presented in Supplementary Fig. 2b. All other comparisons were not significant (p≥.23).
Interaction between time and resilience. The one-way ANOVA with M3-pre difference score for wellbeing as dependent variable and resilience bins (i.e. based on the BRS score) as between-subject factor was significant, F(3,367) = 3.02, p = .030, \({\eta }_{p}^{2}\)=0.024, but the post-hoc independent samples t-tests failed to reach significance after Bonferroni correction (α = .008). Therefore, to further explore the significant interaction between time and resilience, one-way ANOVAs with the M1-pre (F(3,367)=4.46, p=.004, \({\eta }_{p}^{2}\)=0.035), M2-M1 (F(3,367)=0.50, p=.68, \({\eta }_{p}^{2}\)=0.004) and M3-M2 (F(3,367)=1.67, p=.17, \({\eta }_{p}^{2}\)=0.013) difference scores for wellbeing as dependent variable and resilience bins as between-subject factor were conducted as well. Post-hoc independent samples t-tests with Bonferroni correction (α = .008) for the significant ANOVA with difference score M1-pre showed that participants with the lowest resilience (bin 1, M=-8.68) had a significantly steeper decrease in subjective wellbeing from before the pandemic to M1 compared to participants with the second lowest (bin 2, M=-4.00, t(135.64)=-3.70, p<.001, d=-0.48) and highest resilience (bin 4, M=-5.16, t(207.87)=-2.81, p=.005, d=-0.37). This is visually presented in Supplementary Fig. 2c. All other comparisons were not significant (p≥.012)
Interaction between time and the presence of anxiety symptoms. The one-way ANOVA with M3-pre difference score for wellbeing as dependent variable and anxiety symptoms bins (i.e. based on the HADS score) as between-subject factor was significant, F(3,367) = 42.96, p < .001, \({\eta }_{p}^{2}\)=0.26. Post-hoc independent samples t-tests with Bonferroni correction (α = .008) showed that participants with the highest anxiety symptoms (bin 4, M=-13.61) had a significantly more negative difference score, and thus a steeper decrease in wellbeing since before COVID, compared to participants from the lower bins (bin 1, M=-1.43, t(145.17)=9.53, p<.001, d=1.41; bin 2, M=-2.30, t(152.23)=8.75, p<.001, d = 1.27; bin 3, M=-5.82, t(142)=4.39, p<.001, d=0.76). Moreover, participants with the second highest depressive symptoms (bin 3, M=-5.82) had a significantly steeper decrease in wellbeing since the pandemic compared to participants with the lowest (bin 1, M=-1.43, t(75.56)=3.04, p=.003, d = 0.59) and second lowest depressive symptoms (bin 2, M=-2.30, t(176)=2.67, p=.008, d = 0.44). This is visually presented in Supplementary Fig. 2d. All other comparisons were not significant (p≥.35).
To shortly summarize, changes in subjective wellbeing from before the pandemic to M3, during the second peak of the pandemic, were negatively influenced by high depressive and anxiety symptoms and by low resilience, although the latter was clearly less prominent.