Introduction

While the physical distancing measures that governments implemented worldwide were effective in curbing the spread of the Coronavirus Disease 2019 (COVID-19), they also changed patterns of social interaction. For instance, in South Korea, strict, phased social distancing measures and policy were implemented from May 2020 to April 20221 (Supplemental Fig. 1). They were adjusted daily based on regional case numbers (e.g., limits on the number of people gathered in one place and checks of infection and vaccination status). These unprecedented restrictions resulted in new forms of social isolation rarely experienced before in Korea. However, it remains unclear how social isolation and loneliness changed across COVID-19 pandemic periods and whether these patterns varied by age, as extant literature is incongruent on these issues. To elucidate these inconsistencies, multilevel modeling (MLM) was employed to examine longitudinal trajectories of social isolation and loneliness over time.

Fig. 1
figure 1

Age-specific prevalence of loneliness and social isolation (%, 95% CI).

Full size image

Social isolation is defined as an objective lack of social relationships, whereas loneliness is the subjective distress arising from a gap between desired and actual connections2. In the context of COVID-19, social isolation and loneliness must be distinguished to better understand their respective effects. This is because, although pandemic-related restrictions reduced physical contact and objectively increased social isolation, not all individuals experienced loneliness. Conversely, some individuals reported heightened loneliness. Furthermore, research has shown inconsistent relationships between social isolation and loneliness during the pandemic3,4, indicating that they are distinct phenomena. This distinction has important implications, as the policies and interventions required to address each condition may differ.

Building on these conceptual distinctions, empirical research has demonstrated significant variations in the prevalence of social isolation and loneliness across different age groups and populations. A meta-analysis reported that the pooled period prevalence of social isolation was 31.2%, whereas that of loneliness was 28.6%; both showed a significant increase compared with pre-pandemic levels5, with South Korea also experiencing elevated risk6. However, patterns of social isolation and loneliness vary by age4. For instance, a large-scale study in Spain found that those in their 50 s had the highest prevalence of social isolation; this was attributed to frequent pre-pandemic social interactions and strict adherence to distancing guidelines, resulting in a marked reduction in their friendship networks7. Likewise, a study in London reported that middle-aged adults had the weakest social networks and experienced the highest levels of social isolation and loneliness compared with younger and older adults8. Similarly, a study in Japan reported comparable patterns among middle-aged adults9. Conversely, a national cohort study in Japan reported the most pronounced increase in social isolation among older adults during the pandemic10.

A survey of UK adults indicated that young adults (18–30 years) faced the greatest risk of loneliness during the pandemic11,12. A Canadian longitudinal study further found that, while loneliness increased across all ages, the increase was the greatest among middle-aged adults and the least pronounced among those aged 75 and older13. These discrepancies underscore that social isolation and loneliness, although related, manifest differently across age groups.

Although many studies have examined age-related differences in loneliness and social isolation, their findings have been mixed. Numerous studies report a nonlinear, U-shaped relationship between age and loneliness, with loneliness peaking during young and older adulthood and declining during midlife14. Some studies note a linear increase in loneliness with age15, whereas others note an increase from middle adulthood16. Early COVID-19 research in Japan found ambiguous patterns for loneliness but an inverted U-shape for social isolation, peaking during middle age17. These inconsistencies highlight the need for further research to clarify age-related patterns, especially across diverse cultural settings.

Age-related challenges play a crucial role in shaping social isolation and loneliness experiences, as the impact of COVID-19 appears to differ across age groups. For youth, school closures, cancellation of extracurricular activities, and restricted peer interaction disrupted critical periods of social development and contributed to increased feelings of loneliness and psychological distress18. For young adults, the pandemic amplified existing pressures related to higher education, employment insecurity, and delayed transitions to financial independence. The declining labor market and inability to socialize with friends or romantic partners compounded these stressors19. Another study reported that economic hardship and disrupted job prospects during the pandemic were associated with greater psychological distress and increased loneliness among young adults20. A global survey of over 21,000 individuals from 152 countries found that adults aged 25 through 49 were more likely to experience housing and food insecurity, worsening family relationships, and post-traumatic stress during the COVID-19 pandemic than those aged 50 and above21. During the pandemic in France, older adults (65+) experienced fewer social contacts and declining intergenerational interaction22 but reported lower anxiety, perceived risk, and fear of death compared with younger and middle-aged adults23.

The effects of social distancing differed within older age groups. Midlife (55–64), young-old (65–74), and old-old (75–84) adults who were used to spending more time outside or with non-family experienced greater changes and potential distress; conversely, the oldest-old (85 +), who already spent more time alone and at home, were less disrupted24. These varied experiences highlight the need for empirical research to clarify whether previously observed age-specific patterns apply to the Korean context.

Additionally, substantial evidence links social isolation and loneliness with adverse mental health outcomes, including depression and anxiety25. Meta-analyses and longitudinal studies have shown that individuals experiencing persistent social isolation or loneliness are at significantly greater risk of developing depression and anxiety disorders compared with those who remain socially connected26. These psychological symptoms contribute to and result from social disconnection, creating a negative feedback loop. Therefore, interventions should target social and psychological factors, as reducing isolation may help prevent or alleviate mental health difficulties.

However, previous studies have several limitations. First, most studies relied on cross-sectional designs, restricting the ability to capture the dynamic and long-term effects of social isolation and loneliness27,28,29. Second, many studies investigated only social isolation or loneliness rather than both concurrently7,30. However, evidence suggests that the trajectories of social isolation and loneliness can differ substantially, and examining them together yields a more nuanced understanding of their respective impacts. Longitudinal empirical research is lacking on these issues in Korea, highlighting a critical research gap.

Therefore, this study aimed to (1) estimate the prevalence of social isolation and loneliness, including age-stratified rates; (2) analyze cross-sectional age-specific patterns of social isolation and loneliness over three consecutive years; and (3) examine how age, depression, and anxiety influence trajectories of social isolation and loneliness over time using MLM, controlling for sociodemographic factors. Ultimately, this study sought to provide an empirical foundation for developing tailored, age-appropriate interventions in the Korean context.

Methods

Participants and procedures

We analyzed data from the COVID-19 Mental Health Panel Survey conducted by the Ministry of Health and Welfare of South Korea. The survey has been investigating mental health conditions among the general population in South Korea annually since 2021. Participants aged 15–79 years were recruited as part of the Population and Housing Census during the study period (2021–2023). The study protocol was approved by the National Center for Mental Health Institutional Review Board (IRB No. 116271-2021-30). All procedures were conducted in accordance with the relevant guidelines and regulations, including the ethical principles of the Declaration of Helsinki.

The survey was conducted each year between September 1 and December 8. It was conducted face-to-face by 126 investigators trained to provide an overview of the survey process, instructions on the survey format, tablet-assisted personal interviewing, and practices. All participants gave informed consent before participating. For minors aged 15–17 years, informed consent was obtained from participants and their parents or legal guardians.

The study utilized a general population sample of individuals aged 15–79 years, including those with and without a history of COVID-19 infection. A stratified sampling design was applied. Within each selected household, priority was given to recruiting members with a confirmed COVID-19 diagnosis by 2021. In households without a diagnosed case, an undiagnosed individual was selected using probability proportional to size sampling.

The final sample for 2021 included 640 participants with a history of COVID-19 and 1963 participants without a history of COVID-19. Data were collected longitudinally, with a total sample size of 2603 in 2021, 2510 in 2022, and 2462 in 2023, yielding a retention rate of 94.6%. The final analytic sample included 2395 participants (mean age = 46.32, standard deviation = 16.47) who completed a face-to-face interview survey.

Variables

Social isolation and loneliness

Social isolation and loneliness were assessed using the Loneliness and Social Isolation Scale (LSIS)31, which comprises six items representing two items each for loneliness, social networks, and perceived social support. Items 1 and 2 assess loneliness, while the sum of Items 3–6 represents social isolation. Item 1: “I feel lonely”; Item 2: “I feel isolated from others”; Item 3: “How many people, including family, relatives, and friends, do you feel close enough to meet privately at least once a month or contact at least once a week?”; Item 4: “On average, how many minutes per day do you spend in private contact with friends or family?”; Item 5: “I can comfortably rely on my family or friends”; and Item 6: “There is someone who can help me with daily tasks.” Responses were recorded on a 4-point Likert scale ranging from 0 to 3. The Cronbach’s alphas for the subscales of loneliness and social isolation were 0.861 and 0.603, respectively. To determine the prevalence of social isolation and loneliness, we applied the LSIS subscale cutoff points consistent with those used in prior empirical studies32. A cumulative score of 3 or higher on the two loneliness items was used to identify high-risk loneliness. For social isolation, a cumulative score of 4 or higher on the two social network items and the two social support items was used to identify high risk. These cutoffs were derived from score distributions and correlations with validated measures (UCLA Loneliness Scale and Lubben Social Network Scale) and demonstrated good sensitivity and specificity for detecting at-risk individuals in the Korean sample.

Depression

Depressive symptoms were assessed using the Patient Health Questionnaire-933. It comprises nine items rated on a 4-point Likert scale ranging from 0 to 3. The questionnaire demonstrated good internal consistency in this sample, with a Cronbach’s alpha of 0.882.

Anxiety

Anxiety symptoms were assessed using the Generalized Anxiety Disorder-7 scale34. It comprises seven items rated on a 4-point Likert scale ranging from 0 to 3. The Cronbach’s alpha in this study was 0.917, indicating a high level of internal consistency.

Demographic variables

Because this study was a population-based survey, demographic variables including sex, age, region, education level, and income were collected. Specifically, sex was categorized as male or female. Age was divided into seven categories: 15–19, 20–29, 30–39, 40–49, 50–59, 60–69, and 70–79 years. Regions were coded from 1 (capital area) to 5 (more distant) using a researcher-defined regional grouping that reflects proximity to the capital and major metropolitan areas: 1 = Seoul, Incheon, Gyeonggi, and Gangwon; 2 = Busan, Ulsan, and Gyeongnam; 3 = Daegu and Gyeongbuk; 4 = Daejeon, Chungcheong, and Sejong; and 5 = Gwangju, Jeolla, and Jeju. Educational status was classified into five levels: elementary school or lower, middle school, high school, current enrollment in college or a master’s program, and a college or master’s degree graduate. Economic status was assessed based on the average monthly household income before tax and categorized into four groups (million KRW): less than 1.5, 1.5 to less than 3, 3 to less than 5, and more than 5. The diagnosis of COVID-19 was recorded dichotomously (yes or no), and the date of diagnosis was collected.

Statistical analysis

All sociodemographic variables were coded as categorical variables. Variables were coded as follows: age (1 = teens to 7 = 70 s), sex (0 = male, 1 = female), COVID-19 infection (0 = not infected, 1 = infected), region (1 = capital area to 5 = most distant), education (1 = elementary or lower to 5 = college or master’s graduate), and income (1 =  < 1.5 million KRW to 4 =  ≥ 5 million KRW). Anxiety and depression were analyzed as continuous variables. To examine whether the data were missing completely at random, Little’s missing completely at random test was conducted. The missing data were handled using a listwise deletion. Of the 2462 participants who completed all three waves, 67 cases with missing income data were excluded. The final analytic sample comprised 2395 individuals with complete data.

Descriptive statistics were analyzed via chi-square tests and an analysis of variance using SPSS version 26.0. Correlation analyses were conducted to assess the relationships among the predictor variables. Cross-sectional regression analyses were additionally conducted to assess the linearity of the association between age and social isolation or loneliness. We performed MLM using R version 4.4.2. The multilevel model accounted for the longitudinal structure of the data, with repeated measures (three time points) nested within individuals. Model selection was based on the Akaike information criterion (AIC), Bayesian information criterion (BIC), and adjusted BIC (aBIC). The main effects for each age group were examined, and pairwise comparisons of slopes were conducted.

Results

Preliminary analysis

Little’s missing completely at random test indicated that the missing data were random (χ2(7) = 12.606, p = 0.082). No significant baseline differences were found between participants who completed all three waves and those who withdrew regarding COVID-19 diagnosis, depression, anxiety, loneliness, or social isolation (ps > 0.05), indicating that the groups were comparable at baseline.

Descriptive analysis

Table 1 presents the descriptive statistics for the demographic variables. This study included 48.4% male and 51.6% female participants. The distribution of age groups was as follows: 15–19 (6.0%), 20–29 (14.5%), 30–39 (14.8%), 40–49 (19.0%), 50–59 (20.4%), 60–69 (18.0%), and 70–79 years (7.3%). Supplementary Table 1 shows the mean and standard deviations of the key variables.

Table 1 Descriptive statistics for the demographic variables.
Full size table

Significant group differences in social isolation and loneliness were identified across all three years. In 2021, individuals in their teens and 20 s exhibited significantly lower levels of social isolation than those in their 30 s–60 s, and those in their 20 s exhibited significantly lower levels than individuals in their 40 s–70 s (F = 18.970, p < 0.001). Individuals in their 70 s exhibited significantly higher levels of social isolation than all other age groups. In 2022 and 2023, individuals in their teens and 20 s reported significantly lower levels of social isolation compared with those in their 30 s–70 s (2022: F = 12.726, p < 0.001; 2023: F = 8.932, p < 0.001; Supplementary Table 1).

Regarding loneliness, in 2021, individuals in their teens and 20 s reported significantly lower levels than those in their 50 s–70 s (F = 7.882, p < 0.001). In 2022, loneliness scores were significantly higher among individuals in their 70 s compared with those in their teens and 20 s (F = 4.121, p = 0.001). In 2023, those in their 20 s exhibited significantly lower loneliness scores than those in their 50 s and 70 s (F = 5.206, p < 0.001; Supplementary Table 1).

Correlations

Pearson’s correlations showed that depression and anxiety were more strongly associated with loneliness than social isolation at all time points (ps < 0.001). Loneliness and social isolation demonstrated significant stability over time and were moderately correlated (ps < 0.001; Supplementary Table 2).

Social isolation and loneliness prevalence

The prevalence of social isolation was 2.0% in 2021 (n = 49), 3.6% in 2022 (n = 86), and 5.3% in 2023 (n = 127). Conversely, the prevalence of loneliness was 12.7% in 2021 (n = 303), 10.3% in 2022 (n = 247), and 7.7% in 2023 (n = 184) across all age groups (Table 2; Fig. 1).

Table 2 Prevalence (%) of high-risk social isolation and loneliness over three years (N = 2395).
Full size table

Chi-square tests indicated insignificant sex differences in social isolation across all years. However, female participants exhibited a significantly higher prevalence of loneliness than males in 2021 (χ2 = 6.996, p = 0.008).

There were significant age differences in social isolation and loneliness in 2021 (social isolation: χ2 = 13.972, p = 0.030; loneliness: χ2 = 17.098, p = 0.009) and 2023 (social isolation: χ2 = 13.654, p = 0.034; loneliness: χ2 = 22.505, p < 0.001) but not in 2022. However, post hoc pairwise comparisons did not reveal significant differences between specific age groups in any year.

In 2021, the lowest prevalence of social isolation (0.0%) and loneliness (5.6%) was among those aged 15–19. Social isolation was more common among those in their 40 s (3.1%), 60 s (3.3%), and 70 s (3.4%), whereas loneliness peaked in the 60 s (15.8%) and 70 s (14.8%).

In 2022, although age differences were nonsignificant, social isolation and loneliness remained the highest among those in their 60 s (social isolation: 4.4%; loneliness: 12.3%) and 70 s (social isolation: 5.1%; loneliness: 15.3%). The lowest rates were among those in their 20 s (social isolation: 1.7%; loneliness: 7.8%) and those aged 15–19 (social isolation: 2.1%; loneliness: 8.4%).

In 2023, social isolation was the most prevalent among those in their 30 s (7.1%), 40 s (7.5%), and 50 s (5.5%). Loneliness was highest among individuals in their 70 s (14.2%) and lowest among those in their 20 s (4.0%). Post hoc comparisons indicated that participants in their 20 s had significantly lower loneliness than those in their 50 s and 70 s.

Yearly cross-sectional analysis of social isolation and loneliness by age

The cross-sectional regression analyses for each year showed a significant linear association between age and social isolation, with higher age linked to greater isolation (2021: b = 0.038, p = 0.002; 2022: b = 0.051, p < 0.001; 2023: b = 0.041, p = 0.006). The quadratic term reached significance only in 2022 but was negligible, indicating no meaningful curvilinear pattern. For the other years, the quadratic term was nonsignificant (ps > 0.10; Supplementary Table 3; Fig. 2).

Fig. 2
figure 2

Cross-sectional associations of age with social isolation: 2021–2023.

Full size image

Regarding loneliness, age showed only a marginal association in 2021 (b = 0.016, p = 0.069) and no significant association in 2022 or 2023. No curvilinear relationships were found for loneliness in any year (Supplementary Table 3; Fig. 3).

Fig. 3
figure 3

Cross-sectional associations of age with loneliness: 2021–2023.

Full size image

Moderation analysis

The intraclass correlation coefficients were 0.43 and 0.34 for loneliness and social isolation, respectively, indicating substantial individual-level variance and supporting the use of MLM.

The model comparisons showed that the random slope model, which allows individual variation in time slopes, provided the best fit for both outcomes based on the AIC, BIC, and aBIC values (Supplementary Table 4). The aBIC was particularly useful as it accounts for hierarchical data structures35.

Furthermore, multilevel analyses were conducted. The model included four two-way interaction terms (Age group × Time, COVID-19 infection × Time, Depression severity × Time, and Anxiety severity × Time). The covariates included sex, education, income, and region. This approach enabled the assessment of both interaction and fixed effects while controlling for the influence of each variable.

Social isolation

Social isolation significantly increased over time (b = 0.326, p = 0.001). In 2021, greater social isolation was associated with being older than teens (20 s: b = 0.671, p = 0.010; 30 s: b = 1.019, p < 0.001, 40 s: b = 1.319, p < 0.001; 50 s: b = 1.323, p < 0.001; 60 s: b = 1.080, p < 0.001; 70 s: b = 1.327, p < 0.001), not having been infected with COVID-19 (b =  − 0.295, p = 0.027), having higher depression severity (b = 0.088, p < 0.001), being male (b =  − 0.277, p < 0.001), having lower education (b =  − 0.256, p < 0.001), having lower income (b =  − 0.187, p < 0.001), and living closer to the capital region (b =  − 0.150, p < 0.001). Anxiety severity was a nonsignificant predictor. Although social isolation increased over time, the interactions between time and age, COVID-19 infection, depression, or anxiety were nonsignificant (ps > 0.05; Supplementary Table 5; Fig. 4).

Fig. 4
figure 4

Changes in social isolation over time by age group.

Full size image

Loneliness

The multilevel model analyses showed that loneliness did not change significantly over time in the total sample (b = 0.016, p = 0.777). At baseline, all age groups older than teens reported greater loneliness (20 s: b = 0.350, p = 0.035; 30 s: b = 0.453, p = 0.007; 40 s: b = 0.552, p < 0.001; 50 s: b = 0.576, p < 0.001; 60 s: b = 0.618, p < 0.001; 70 s: b = 0.433, p = 0.021). Greater loneliness was also associated with higher depression (b = 0.092, p < 0.001), higher anxiety (b = 0.036, p = 0.016), lower education (b =  − 0.103, p < 0.001), lower income (b =  − 0.142, p < 0.001), and residence closer to the capital region (b =  − 0.031, p = 0.006). Sex and COVID-19 infection status were nonsignificant predictors. The interaction analyses revealed a significant decrease in loneliness over time for those in their 40 s (b =  − 0.137, p = 0.030) and 60 s (b =  − 0.211, p = 0.001). No other interactions were significant (ps > 0.05; Supplementary Table 6; Fig. 5).

Fig. 5
figure 5

Changes in loneliness over time by age group.

Full size image

The simple slope analyses showed significant reductions in loneliness over time for those in their 40 s (b = − 0.089, 95% CI [− 0.154, − 0.024]), 50 s (b = − 0.068, 95% CI [− 0.130, − 0.006]), and 60 s (b = − 0.162, 95% CI [− 0.228, − 0.096]). Other age groups exhibited nonsignificant changes (Supplementary Table 7). Post hoc pairwise comparisons revealed that only the difference between teens and those in their 60 s was significant (slope difference = 0.211, SE = 0.064, z = 3.309, p = 0.016). No other age group differences in the rate of change were found (ps > 0.05; Supplementary Table 8).

Discussion

The unprecedented COVID-19 pandemic significantly affected social interactions, heightening public health concern regarding social isolation and loneliness. However, in Korea, empirical research on the prevalence and long-term trends of these phenomena with nationally representative, multi-year data has been lacking. This study addressed this gap by estimating the prevalence of social isolation and loneliness across age groups, examining patterns over three consecutive years, and using MLM to identify the longitudinal effects of age, depression, and anxiety.

Our findings revealed a steady increase in high-risk social isolation, whereas loneliness gradually decreased. These findings may be interpreted in light of the evolving quarantine policies in South Korea. In 2021, when strict level 4 social distancing measures were in place, social networks were heavily restricted, which likely contributed to the observed increase in high-risk social isolation. Although regulations were relaxed under the “With COVID-19” policy in 2022–2023, previous social ties might not have been quickly restored, resulting in isolation levels remaining elevated. By contrast, loneliness gradually declined. This divergence highlights the importance of distinguishing between objective social isolation and subjective loneliness. A possible explanation for this difference is South Korea’s advanced digital infrastructure36, which might have allowed continued online connection and helped mitigate loneliness despite greater isolation37. Moreover, Korea’s collectivist culture, with its strong family and community ties, may buffer against loneliness38, whereas individualistic cultures may heighten vulnerability during periods of disruption39. Previous studies examined social isolation40 or loneliness17,41 separately and relied on early pandemic data (2020–2021)5,17,41, thus capturing only short-term effects. Conversely, our longitudinal data (2021–2023) provides insight into longer-term trajectories.

The cross-sectional analyses demonstrated a primarily linear relationship between age and social isolation, with higher risk among older adults. No consistent association was found between age and loneliness. These findings are consistent with a large-scale cross-sectional study conducted in Korea in 2021, reporting a linear relationship of age with social isolation and loneliness42. Pre-pandemic evidence is mixed, reporting U-shaped patterns (where adolescents and older adults are the most vulnerable)43, inverted U-shaped patterns44, or a linear increase in loneliness with advancing age45,46. However, such discrepancies may reflect differences in sample composition, cultural context, and measurement timing.

The MLM revealed that social isolation increased significantly over time. However, the rate of increase did not differ significantly between age groups, indicating a broadly uniform increase in social isolation across the adult lifespan. This extends earlier research by demonstrating the persistence of increased social isolation47,48.

At baseline, greater social isolation was associated with older age (relative to teens), the absence of COVID-19 infection, higher depression severity, the male sex, lower education or income, and residence near the capital. Anxiety was a nonsignificant predictor, suggesting the central role of depressive symptoms, as supported by previous research49.

By contrast, loneliness remained stable overall, with significant declines only among those in their 40 s, 50 s, and 60 s. Previous studies found that loneliness increased sharply during strict pandemic restrictions but returned to pre-pandemic levels after measures were eased50,51. These findings suggest that the resilience of social bonds and the mitigating effects of increased social support and motivation to restore social connections52.

The decline among those in their 40 s, 50 s, and 60 s suggests age-specific vulnerabilities and adaptive capacities. Older adults may have developed more effective mechanisms for emotional regulation and social adaptation over time52. These patterns are consistent with the socioemotional selectivity theory53, which posits that social goals shift with age. Adolescents and individuals in early adulthood who are motivated to expand their social networks may have been more sensitive to pandemic‐related restrictions on social exploration. Conversely, middle‐aged and older adults tend to prioritize emotionally meaningful relationships, which may help them regulate emotions more effectively and maintain social well‐being under such constraints. Digital engagement may also play a different role by age. A multi-country study found that social media use reduced loneliness among older adults (60 +) but increased emotional loneliness among younger adults (18–29 years)54.

After adjusting for covariates, social isolation and loneliness were the most pronounced among middle-aged adults, underscoring their heightened psychosocial vulnerability. During the pandemic, individuals in midlife (aged 41–65) reported the highest levels of loneliness and weakest social networks8. The highest prevalence of social isolation and loneliness was noted in the 50–59 age group, potentially linked to life transitions like children leaving home or employment changes7,55.

Depression consistently emerged as a strong risk factor for both outcomes, while anxiety was more closely associated with loneliness56,57. This is consistent with previous network analysis research indicating that depression has a more direct impact on social isolation, while the influence of anxiety on loneliness is significantly mediated by depression49. Anxiety is primarily linked to loneliness and is more closely related to the perceived quality of relationships rather than the number of social contacts56.

Several sociodemographic factors were associated with greater vulnerability to social isolation and loneliness at baseline. Individuals with lower educational attainment and income and those residing in the capital region were more likely to experience higher social isolation and loneliness, consistent with previous research58,59. Sex differences were observed only for social isolation, with males reporting higher levels than females, aligning with previous findings that men are more prone to social isolation across the lifespan60. This is attributable to men’s generally smaller and less frequent social networks outside the family, making them more susceptible to isolation61. Conversely, sex differences in loneliness were nonsignificant, despite fewer social contacts. Women’s broader and more diverse social networks may buffer against loneliness, even during restrictive periods62. However, some studies have found significant sex differences in loneliness63,64, limiting the generalizability of these findings. Additionally, COVID-19 infection was negatively associated with social isolation, possibly reflecting more frequent social contact among those infected, aligning with previous research on in-person activity and lower isolation7.

Clinical and policy implications

The present findings revealed that social isolation has significantly increased over time in Korea. Moreover, individuals with lower income were particularly vulnerable to social isolation. However, evidence-based interventions are not covered by insurance in Korea, making them less affordable, especially for socially isolated individuals with generally lower incomes65.

Furthermore, while loneliness significantly decreased over time among adults in their 40 s, 50 s, and 60 s, this improvement was not observed among adolescents and young adults. Therefore, early intervention is critical, as adolescents who experience social isolation and loneliness may continue to face these challenges into adulthood, increasing the risk of chronic isolation and loneliness later in life. To address these challenges, it is essential to implement age-specific strategies that recognize social isolation and loneliness as distinct yet interconnected issues.

For adolescents, ongoing monitoring and early prevention programs are needed to reduce long-term psychosocial risks. Social isolation is increasing among young adults in their 20 s and 30 s even as loneliness remains relatively stable. Interventions should focus on expanding access to social spaces and opportunities to develop new networks, with promising digital platforms and cognitive behavioral therapy (CBT)-based approaches66, especially given this group’s high digital literacy67,68.

Middle-aged adults, identified as the most vulnerable, may benefit from integrated interventions that support social connectedness and psychological well-being, such as group counseling, behavioral activation, and community engagement. Internet-based CBT is effective for reducing loneliness by targeting cognitive and behavioral patterns69,70.

For older adults, maintaining existing social ties and minimizing participation barriers caused by physical limitations are crucial. Remote interventions, such as video-call based group CBT, have been effective in reducing loneliness and depressive symptoms among older adults71. Meta-analytic evidence further supports the benefits of programs focused on social support and maladaptive social cognition72.

Although most interventions focus on loneliness, recent findings indicate that online approaches can help reduce social isolation73. Future research should develop comprehensive strategies addressing objective social networks and subjective loneliness.

Strengths and limitations

This study’s longitudinal design allowed for a comprehensive understanding of changes in social isolation and loneliness, overcoming the limitations of cross-sectional or short-term research. By assessing these constructs concurrently, we elucidated the trajectories of objective and subjective aspects of social connection over time. To our knowledge, this was the first study in South Korea to examine age-related effects on these outcomes using a wide age range (15–79 years) and a structured age classification. Rigorous MLM enabled robust covariate adjustment and revealed that middle-aged adults were the most affected after adjustment. The use of a large, nationally representative sample and in-person survey administration further minimized selection bias and ensured the inclusion of individuals with limited access to technology, such as older adults.

Despite these strengths, this study has several limitations. First, unmeasured factors such as life events might have influenced the observed patterns. Second, differences in measurement tools (e.g., the LSIS) limit the comparability of prevalence rates across international studies. The LSIS measures social isolation as comprising structural (network size and frequency) and functional (social support) dimensions74. The functional items reflect perceived isolation, which should be considered when interpreting the results. Additionally, because data were collected only during the COVID-19 pandemic, it is difficult to conclude that social isolation and loneliness increased compared with the pre-pandemic levels. Comparisons between LSIS-based and structurally focused estimates warrant careful consideration. Moreover, self-reported assessments may be subject to biases such as social desirability or recall errors. Finally, cultural differences in social structures or policies may affect trajectories of social isolation and loneliness. Therefore, cross-cultural studies are needed to generalize these results.

Conclusion

This study demonstrated that, during and after the COVID-19 pandemic in Korea, social isolation increased across most age groups, whereas loneliness remained stable or declined among middle-aged and older adults. Both outcomes were the most pronounced among middle-aged adults after adjusting for relevant factors, highlighting their vulnerability. Depression was a key risk factor for both, while anxiety was mainly related to loneliness. Sex differences appeared only for social isolation, with males being the most affected. These findings stress the need for targeted, multifaceted interventions and further research on contextual and generational differences.