This study employed an observational cross-sectional study design. Residents aged 21 years or above were recruited using stratified random sampling with stratification based on housing type, geographical region, gender, and age group. A door-to-door survey was conducted by experienced interviewers between October 17, 2020, and November 27, 2020, not long after partial lockdown restrictions were gradually lifted [20]. The in-person surveying approach ensured inclusion of study participants who might not have access to the internet to mitigate selection bias. The questionnaire administered by the interviewers was worded in the English language, which is the main language in Singapore. To be included in the study, participants were required to have resided in Singapore during the lockdown phase of the pandemic (known locally as the circuit breaker (CB) period between April 7, 2020, and June 1, 2020), in which spatial distancing measures were imposed (known locally as safe distancing). They were also required to be able to speak English and be aged 21 years or older. Interviewers were trained by a geriatrician to exclude older adults who exhibited possible signs of cognitive impairment (eg, drowsiness, agitation, and incongruent language) during the process of taking informed consent. Additionally, among those aged above 70 years, the Abbreviated Mental Test (AMT) [36] was used to screen for possible signs of dementia and other cognitive impairments, and participants were excluded if they failed 1 of the 3 items in the AMT. Safe distancing rules were adhered to during the data collection period.

Ethics approval was obtained from the National Healthcare Group Domain Specific Review Board (2020/00973), and written informed consent was obtained from all participants. Participants were reimbursed with grocery vouchers worth SGD 10 (USD 7.50) for participation. All data collected were deidentified prior to analysis.

Based on an a priori power analysis (G*Power 3.1.9.7) using a power of 0.80 and an error probability of 0.05, a sample size of 300 participants was required for each group to detect a between-group difference of a small to moderate effect size.

A total of 602 participants (Table 1) were recruited for the study (mean age 53.30 years, SD 16.26 years). Of these 602 participants, 302 were categorized as younger (21-59 years old; mean age 39.87 years, SD 11.46 years) and the other 300 were categorized as older (60 years or above; mean age 66.82 years, SD 5.84 years). The majority of the younger adults completed tertiary education (213/302, 70.5%) and were employed (234/302, 77.5%), while the majority of the older adults completed secondary education (145/300, 48.3%) and were largely retired (145/300, 48.3%) or were no longer in employment (44/300, 14.7%).

Up to a third of older adults (102/300, 34.0%) in the study sample reported not using digital media to communicate or run errands during the CB measurement period compared with only 2.7% (8/302) of younger adults who did not do so. Among older adults who did not use the internet (102/300, 34.0%), an overwhelming majority did not have tertiary education (94/102, 92.2%). For older adults, the top 3 reasons for using the internet (Table 2) were social media (154/300, 51.3%), telecommunication (120/300, 40.0%), and online banking (111/300, 37.0%), whereas for younger adults, the top reasons were social media (253/302, 83.8%), online banking (234/302, 77.5%), food delivery (211/302, 69.9%), and telecommunication (210/302, 69.5%).

Overall, agreement scores on the ability to use digital platforms to meet needs were lower for older adults (mean 3.37, SD 1.14) than for younger adults (mean 4.17, SD 0.78) (t₆₀₀=10.04; P<.001; d=0.82), and the difference was large in magnitude. The ability to use digital resources for information updates related to the pandemic was lower for older adults (mean 3.28, SD 1.20) than for younger adults (mean 4.22, SD 0.75) (t₆₀₀=10.50; P<.001; d=0.86), and the difference was large in effect size.

More specifically, for older adults who were able to use the internet (Table 3), the agreement scores in their ability to do so were lower compared with the scores for younger adults in the areas of using telecommunication (t₄₂₇=7.55; P<.001; d=0.76), obtaining supplies (t₄₇₇=8.97; P<.001; d=0.82), using online services (t₅₀₃=8.17; P<.001; d=0.74), and changing appointments (t₄₉₈=3.53; P<.001; d=0.31). The effect sizes of the differences were large on the whole, except for the last variable.

Among older adults, there were minimal changes in the nonuse of the internet for online shopping for essential items (Table 4) before and during or after the pandemic (186/300, 62.0% and 183/300, 61.0%, respectively). This relationship was similar among younger adults, although a much smaller proportion of younger adults did not use the internet for online shopping for essential items before and during or after the pandemic (45/302, 14.9% and 39/302, 12.9%, respectively).

There was a significant negative relationship between time spent on the internet and age (r=−0.63; P<.001), indicating that increasing age is associated with decreasing time spent on the internet. On average, older adults indeed spent significantly less time on the internet for both work and personal uses per day (mean 146.00 min, SD 9.18 min) compared with younger adults (mean 433.27 min, SD 14.32 min) (t₆₀₀=16.84; P<.001; d=1.38).

Zero-ordered bivariate correlations revealed that there was a negative relationship between age and the well-being indicators of depression scores (r=−0.36; P<.001), anxiety scores (r=−0.22; P<.001), and stress scores (r=−0.19; P<.001).

There was also a significant positive association between time spent on the internet and depression scores (r=0.31; P<.001), anxiety scores (r=0.23; P<.001), and stress scores (r=0.28; P<.001). Time spent on the internet was also positively associated with stress concerns (r=0.08; P=.04).

On interpreting the findings, depression scores suggested that those who spent more time on the internet tended to experience less positive feelings, feeling nothing to look forward to, difficulty in working up initiative, etc. For anxiety scores, those who spent more time on the internet tended to feel worried and scared. Likewise, for stress scores, those who spent more time on the internet tended to experience difficulty in winding down, overreaction to situations, and difficulty in relaxing. The effect sizes for these indicators ranged from low to medium.

Links between internet use patterns in this sample and mental health were explored. The findings showed that mental health was significantly better among those who did not use the internet for social media compared with those who did (Table 5). The effect sizes of the differences however were small (d_depression=−0.27; d_anxiety=−0.17; d_stress=−0.31). This was similarly the case among those who did not use the internet for online shopping compared with those who did (Table 6), with the effect sizes being small (d_depression=−0.44; d_anxiety=−0.17; d_stress=−0.31).

The main effects of time spent on the internet and age on distress were examined using regression modeling. The regression models were statistically significant for depression (F_5,501=39.64; P<.001), anxiety (F_5,501=19.84; P<.001), and stress (F_5,501=31.40; P<.001). The models explained 28%, 17%, and 24% of the variances in depression, anxiety, and stress, respectively (see Table 7 for the adjusted and unadjusted models).

To facilitate comparison of the effect sizes, all variables were standardized. Examining the variables in the models, time spent on the internet and age were significant predictors of depression, anxiety, and stress. This was the case even after controlling for the effects of individual adaptability, social support, and health status. Standardized regression coefficients showed that the magnitude of the effect of time spent on the internet was higher than that of age for anxiety (β_age=−0.16; β_{time spent on the internet}=0.18) and stress (β_age=−0.17; β_{time spent on the internet}=0.21), but this effect was lower for depression (β_age=−0.24; β_{time spent on the internet}=0.20). Overall, the findings suggested that distress was associated with decreasing age, and likewise, this was the case for those who spent more time on the internet.

To examine if the effect of age on mental health was influenced by time spent on the internet, mediation analysis was conducted using the approach advocated by Baron and Kenny [41]. The bootstrapping method with 1000 resamples to estimate the 95% CI was additionally conducted to investigate the significance of indirect effects [42]. A significance level of P<.05 was used for all analyses.

To facilitate interpretation of the effect sizes, all variables were standardized to reflect the strength of correlations. Results on the partial mediating role of time spent on the internet for the effect of age on mental health are shown in Figure 1.

In all models, mediation analyses showed that there was a significant indirect effect of age on mental health through time spent on the internet. For depression, the indirect effect of time spent on the internet was as follows: ab=−0.08; z=−2.66; P=.008. The mediation effect accounted for 23% of the total effect (β=−0.36). For anxiety, the indirect effect of time spent on the internet was as follows: ab=−0.09; z=−2.15; P=.03. The mediation effect was moderate and accounted for 39% of the total effect (β=−0.24). For stress, the indirect effect of time spent on the internet was as follows: ab=−0.11; z=−3.20; P=.001. The mediation effect was moderate and accounted for 42% of the total effect (β=−0.22). Examining the r² of the models, all variables together explained 43% of the variance in depression, 40% of the variance in anxiety, and 40% of the variance in stress.

Overall, the effect of age on mental health was partially explained by time spent on the internet. On interpreting the r² of the models and effect sizes, a conclusion could be drawn that the mediating role of time spent on the internet for the effect of age on mental health was relatively substantial, especially for anxiety and stress. However, as full mediation did not occur, for time spent on the internet, which played a substantial role, the effect through age was not a major causal pathway.

This study examined the relationship among age, internet use, and mental health during the lockdown phase of the COVID-19 pandemic. Specifically, we compared internet use between younger and older adults and examined the mediating role of internet use in the relationship between age and mental health. In support of our hypothesis, older adults were reported to spend less time on the internet as compared with their younger counterparts. In addition, internet use was found to partially mediate the effects of age on depression, anxiety, and stress.

Given that ICT is ubiquitous in today’s world, digital exclusion of older adults is emerging as an imperative concern amid rapidly aging populations. Previous studies have reported poor adoption of ICT among older adults [43-46], but findings specific to older adults during the early phase of the COVID-19 pandemic were sparse. The findings of this study extend this empirical evidence by demonstrating that such an age-related digital divide was pervasive during the COVID-19 pandemic. In support of our hypothesis, internet use among older adults during the lockdown period was found to be less as compared with that among their younger counterparts. Specifically, the use of digital platforms for essential services and entertainment purposes was less prevalent among older adults than among younger adults. While age has been known to be a predictor of internet use [47,48], the large effect size (d=1.38) found in this study suggests that the magnitude of such a grey digital divide is substantial and warrants more attention. Given that higher education is also associated with greater internet use [47], our results may be confounded by lower education levels among older adults in the study sample. Such a difference may also be plausibly attributed to the negative relationship between age and technology acceptance [49] or fear of stereotype threats regarding technology use in older adults [50].

Imposed spatial distancing measures during the COVID-19 pandemic have proliferated the adoption of technology to help individuals meet essential needs and stay connected with one another. This has been claimed to amplify the digital divide between younger and older adults [15,51,52]. Indeed, our findings showed that older adults not only used less internet than their younger peers, but also reported to be less able to use online services to run essential activities during the lockdown phase of the pandemic. This suggests that older adults were less capable of adapting to digital means to meet their basic needs. Our findings corroborate the results of another study reporting that the inability of older adults to use digital devices limited their access to transportation, medical care, and food supplies during the pandemic [52]. Consequently, this has raised concerns that the negative impacts of the pandemic on mental health may be disproportionate in certain groups, such as older adults, who may risk being excluded from the society due to such a digital divide [53].

While lockdown measures were well intended to mitigate infections during the COVID-19 pandemic, they had detrimental effects on the psychological well-being of individuals [5]. The results of this study suggest that both age and internet use are significant predictors of mental health, and importantly, such associations persisted even after accounting for adaptability, social support, and health status. As purported by the socioemotional selectivity theory [54], we found that older age was associated with better mental health. Specifically, we previously reported that older adults had lower depression, anxiety, and stress levels as compared with younger adults [38]. Our results are consistent with those of other studies that reported better mental health in older adults during the pandemic as compared with younger adults [54-57]. In contrast, higher internet use during the lockdown period was found to be associated with poorer mental health. While digital technologies certainly could aid individuals in the continuation of their work and daily essential activities during the lockdown, they could also be used in a disordered manner [17,18]. Our results suggest that it could be the case of the latter since mental distress has been observed to be higher in those with higher internet use (ie, spent more time on the internet and used more social media and online shopping). Indeed, the risk of problematic internet use was reported to have increased during the pandemic [2].

Many studies conducted during the earlier phase of the COVID-19 pandemic focused on the effects of internet use on younger adults and adolescents who were often conveniently sampled [25,28,30-32]. In this study, a door-to-door survey using a stratified random sampling method ensured that potential sampling bias issues were circumvented. This study ensured that there were adequate responses sampled from participants in the older age group (ie, over 60 years of age) and therefore contributed to the literature by showing the effects of internet use in not only younger adults but also older adults. This study was therefore able to address calls [34] for research comparing the linkages among age, internet use, and mental well-being during the COVID-19 pandemic. Importantly, our statistical modeling suggests that internet use acts as a partial mediator for the effects of age on depression, anxiety, and stress levels. This suggests that the inverse relationship between age and mental distress could be partially explained by the amount of time spent on the internet by younger adults. As mentioned earlier, younger adults were found to spend more time on the internet as compared with older adults. Such increased use may be problematic and consequently have negative effects on mental health. Indeed, problematic internet use during the COVID-19 pandemic has been reported to be associated with poorer mental well-being [58,59]. Studies have reported that pandemic-related stress is associated with tendencies toward problematic digital use [59-61]. Given that younger adults were found to have greater stress concerns during the pandemic [38], it may be the case that younger adults experienced greater stress and thus engaged in greater problematic internet use, which resulted in poorer mental health. More research is recommended to investigate this. Our findings support the recommendations to mitigate the risk of problematic internet use during the pandemic [2].

Despite the negative relationship between internet use and mental health, it is important to note that not all types of digital use had detrimental consequences on mental health, and ICT use for the right purposes could be potentially beneficial. For example, a recent scoping review revealed that internet use for communication purposes was associated with better mental health for older adults during the COVID-19 pandemic [62]. Indeed, older adults have reported the importance of ICT to help them maintain their social connections during the lockdown period [63-65]. This study showed that social media use was the highest among the various uses of the internet even for older adults, and a future study can consider investigating how older adults use social media and assessing the benefits they derive from it during the pandemic. Given that some older adults may experience loneliness due to lack of physical contact, especially under the exceptional circumstances imposed by the pandemic, the use of ICT can help to allay the negative effects of social isolation. Thus, in view of the large digital divide between younger and older adults found in this study, greater efforts are necessary to close such a divide, including advocating for age-sensitive design of technologies [49,52] and deterring stereotype threats associated with technology use [50]. Considering the potential negative relationship between internet use and mental health, it is also imperative to educate older adults in using ICT appropriately.

Notwithstanding the contributions of this study to existing literature, it is important to acknowledge some limitations. First, the study sample consisted of community-dwelling adults, and thus, the findings may not be generalizable to specific subgroup populations who may be more vulnerable. The relationship between age and mental distress observed in this study may differ if other groups of older adults are included, such as those residing in nursing homes or experiencing diminished mental capacity.

Second, the interpretation of this study’s results is limited by its outcome measures. Although this study established the partial mediating role of internet use for the effects of age on depression, anxiety, and stress, the study did not collect measures that would be important indicators of disordered use of the internet in younger adults, including online gambling, viewing pornography, and playing video games. Nonetheless, there is evidence from various studies indicating that increased time spent on the internet during the earlier phase of the COVID-19 pandemic was the main contributory factor for a number of mental health problems, such as depression, anxiety, and stress [66-68]. The findings from this study are therefore aligned with the existing research and add to this field by presenting data from a developed country with high digital adoption rates. However, a future study can consider examining how the pandemic could have possibly exacerbated problematic internet use and whether time spent on the internet is indeed an adequate proxy. The digital divide and its relationship with mental health can be influenced by a myriad of factors going beyond problematic internet use. There is a possibility that increased internet use in younger adults during the lockdown period was attributed to having to work remotely from office, and the distress experienced could therefore be attributed to this new form of work arrangement rather than problematic internet use per se. Given that the study was making reference to the first lockdown period, younger adults may not have adapted to this form of working arrangement despite the known benefits of ICT, including continuation of their work and running of other essential activities.

Finally, all measures of this study were collected from participants’ retrospective recollection of their experiences during the lockdown period. Since time spent on the internet was not based on any objective indicator, we cannot rule out the possibility that this or other attitudinal responses are subject to recall bias.

This study showed that older adults lagged behind younger adults in the use of digital resources during the pandemic, which could have helped them in communication and socialization, and the findings support existing literature on the poor adoption of ICT among older adults. This study further contributes to the literature by showing how through mediation modeling, the negative relationship between increasing age and mental distress appears to be partially explained by the amount of time spent on the internet by younger adults. Without moderate use, the benefits brought by digital technologies could have been attenuated during the lockdown phase of the pandemic. It is imperative to educate both young and old adults in the appropriate use of ICT.