Social isolation and mental well-being among Korean older adults: a focus on living arrangements

The datasets used in this study was from the Korea National Health and Nutrition Examination Survey (KNHANES), conducted by the Korea Centers for Disease Control and Prevention, an agency of the South Korean government. The survey participants were chosen through a comprehensive process involving questionnaire surveys and medical examinations. While the number of respondents may exhibit slight variations annually, the 2018 KNHANES employed in this study involved 13,000 individuals across 4,416 households. In the 2020 KNHANES, 14,000 individuals from 4,800 households were surveyed. The survey encompassed individuals aged 1 year and above, covering a broad age range of up to 100 years. The participants were categorized into children, adolescents, and adults. Each group underwent tailored survey inquiries based on their individual characteristics. In the case of adolescents, parents responded on their children's behalf. The survey content was organized into two modules, with one module consistently applied annually and the other subject to periodic replacement.

The sampling approach used in this study involved area probability sampling with multi-cluster sampling. Under this sampling design, once the survey area was determined, households were chosen through household member verification, facilitated by interviewers. The interviewer, in collaboration with local health and community centers, examined the survey area boundaries and determined the appropriate number of households. In accordance with the designated household count in the survey area, the interviewers further divided the area into zones and selected 25 households from each designated zone. Subsequently, the chosen households underwent a visitation process to confirm essential survey information and collect contact details for scheduling future phone interviews for participation in the survey. Selected households were notified of their participation through the "Notice of Household Selection for the Korea National Health and Nutrition Examination Survey." This notice, issued by the head of the local government, was distributed 1 month before the survey commenced. A week before the survey initiation, the interviewers proactively reached out to selected households to coordinate a preliminary appointment. This early engagement aimed to facilitate communication between the respondent and interviewer, allowing for a pre-determined mutual arrangement of interview times. Interviewers conducted the interviews using the Computer-Assisted Personal Interviewing (CAPI) method.

To conduct our analysis, we focused on a specific subset of our data consisting of older adults aged 65 years and above. This subset was extracted from a larger dataset, which enabled us to concentrate on this demographic group. We used a statistical technique known as Zero-inflated Poisson Regression Analysis to investigate the factors influencing the mental health of older adults. Poisson regression models are commonly used when analyzing count data, where the outcomes are represented by nonnegative integer values. In our case, we were interested in understanding mental health outcomes, which often involve counts of symptoms or indicators. However, standard Poisson regression may not adequately handle datasets with an excess of zero counts, which is common in mental health studies where many individuals may not exhibit any symptoms. To address this issue, we employed a specialized form of Poisson regression, known as the Zero-inflated Poisson Regression. This technique was specifically designed to handle datasets with excess zero counts such as those encountered in our study.

As shown in Figure 1, the distribution of mental health scores (measured using the PHQ-9) for respondents in 2018 and 2020 was not normal. Instead, it exhibited a highly skewed distribution with a preponderance of values clustered toward the lower end, resembling a Poisson distribution with excess zeros. This phenomenon is indicative of overdispersion, in which the variance of the data exceeds what would be expected under a standard Poisson distribution. We evaluated two possible models: the standard Poisson model and the Zero-inflated Poisson (ZIP) model. To determine the most suitable model, we conducted the Vuong test, which is utilized for comparing non-nested models (28, 29). Based on the Z-statistics (0.15, 0.01) from the Vuong test for models incorporating only main effects and those including both main and interaction effects, we were unable to reject the null hypothesis at the level of 0.05 that the standard Poisson model and the ZIP model fits the data equally well. Despite the possibility of choosing either model, we ultimately selected the ZIP model due to additional criteria such as the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC). The AIC and BIC indices were lower for the ZIP model compared to the standard Poisson model, indicating a better fit (see Supplementary Appendix). All statistical analyses were conducted using the SAS 9.4 software.

Table 1 outlines the sample characteristics of the two-time-point data used in our study. In the 2018 dataset, females comprise 57.65% of the sample, whereas males represent 42.35%. The age distribution indicates a concentration in the 65–74 age range, with 30.55% aged 65–69 and 50.51% aged 70–74. Households predominantly consist of two members (51.02%), followed by single-member households (23.81%). Marital status indicates a vast majority (99.09%) of married individuals. Educational attainment varies, with a significant proportion having no formal education (56.84%), while employment status indicates that 66.57% are unemployed. Income distribution skews toward lower categories, with 39.87% falling into the lowest-income bracket. For home ownership, the majority (73.71%) own their homes. Although there is a slight variation, the distribution of the 2019 sample conforms to this pattern.

Table 2 shows the disparities in mental health, as measured by the PHQ-9 scale instrument, across various demographic and socio-economic factors. Gender exhibits a significant difference, with females reporting a higher mean score (2.73), compared to males (1.68), indicating potentially poorer mental health among females (p < 0.001). Age does not show a consistent pattern, with slight variations across age categories, but these are not statistically significant differences. The number of household members presents a notable disparity in mental health scores. Individuals living in households with a single-member report the highest mean score (3.07), indicating that potentially poorer mental health compared to older adults in households with two or more members (p < 0.001). Older adults who are not married report significantly higher mean scores (4.62) compared to married ones (2.25), suggesting that marital status may influence mental health outcomes. Similarly, those with no formal education report higher mean score (2.70) compared to those with higher levels of education (p < 0.001).

Furthermore, employment status and household income category also demonstrate significant differences, with employed individuals and those in higher income categories reporting lower mean scores (p < 0.001) as compared with those who are unemployed and belong to lower income categories. House ownership is also associated with mental health disparities, with homeowners reporting lower mean scores (2.04) than non-homeowners (2.93). Finally, the year factor indicates no significant difference in mental health scores between 2018 (before COVID-19) and 2020 (during COVID-19), suggesting that while the pandemic may have influenced overall mental health trends, it did not manifest a significant difference within this specific age population. The proportion of the low-educated females living alone without ownership was found to be 7.8 percent in 2018 and 5.5 percent in 2020.

Table 3 presents the results of the Zero-inflated Poisson Regression analysis that examined the association between various factors and depression severity. Model 1 demonstrates that older adults living in larger households exhibited lower rates of depression than those living alone. Specifically, older adults residing in two-person households have an incidence rate ratio (IRR) of 0.85 (95% CI: 0.75–0.90), indicating a 15% reduction in depression severity. Similarly, older adults in three-person households show a greater reduction (IRR: 0.81; 95% CI: 0.73–0.89), while those in households of four or more persons have a modest reduction (IRR: 0.87; 95% CI: 0.78–0.90).

We hypothesized that the level of depression severity among older adults would be higher in the era of COVID-19 than the period before it; however, our study did not find a significant change in depression severity in the year 2020, marked by the COVID-19 pandemic, compared to the reference year of 2018 (IRR: 0.98; 95% CI: 0.93–1.04), suggesting that the pandemic's onset did not have a uniform impact on depression rates across the study population.

For gender, females exhibit a slightly higher depression severity (IRR: 1.06; 95% CI: 1.01–1.13) compared to males. Participants aged 75–79 show a slight increase in depression severity (IRR: 1.08; 95% CI: 0.98–1.16), but it did not show statistical significance. Other age groups did not show significant differences from the 65–69 reference group. For income levels, compared to the middle income group, the lowest income group is more likely to experience depression severity (IRR: 1.11; 95% CI: 1.02. 1.22). Married individuals and those with higher household income, higher levels of education, employment, and house ownership are associated with lower depression severity, highlighting the protective effects of socio-economic stability and social support.

Notably, older adults reporting higher levels of stress reveal significantly increased depression severity, with those feeling very stressed showing nearly three times the severity (IRR: 2.67; 95% CI: 2.40–2.96) compared to those who almost never feel stressed.

To test our hypothesis that older adults living alone are more likely to experience increased depression severity during the pandemic compared to those living with family members, we included the year as a moderating variable in Model 2. Model 2 reveals that the three interaction effects of year and living arrangement on mental illness is not statistically significant. This suggests that the impact of living arrangements on mental health remains consistent, regardless of the COVID-19 period.

One of the strengths of ZIP model is that it allows us to analyze the likelihood of having excess zeros (or a PHQ-9 score of zero) for depressive symptom. Although this analytical property is not our main interest, we have additionally provided these analytical results. Table 4 presents the results of a logistic regression model from the ZIP model, focusing on the probability of an excess zero in PHQ-9 depression scores. The analysis indicates that all living arrangement categories, compared to the single household reference group, do not have significant effects on the likelihood of an excess of zero. Gender is found to be a significant predictor, with females showing lower log-odds of an excess zero outcome compare to males (B: −0.39, p < 0.001). Advancing age is associated with a higher likelihood of an excess zero, particularity evident in the 70–74 ae bracket (B:0.29, p < 0.01). Socioeconomic factors such as lower household income categories exhibit lower log-odds of an excess zero (B:-0.30 for the lowest income and − 0.36 for low income, p < 0.05), suggesting a possible greater acknowledgment of depressive symptoms among economically disadvantaged groups. Higher education beyond 13 years also correlates with a higher likelihood of an excess zero (B: 0.35, p < 0.05), as does being employed (B: 0.25, p < 0.01). Stress levels markedly affect the likelihood, with those feeling stressed or very stressed showing significantly lower log-odds of an excess zero (B: −1.78 and − 1.5, p < 0.001), emphasizing the influence of perceived stress on the non-reporting of depressive symptoms. These results consistent in Model 2, which contained the interaction effect, indicating that the COVID-19 variable had no effect.

Our study investigated the impact of living arrangements on the mental health of older adults, with a particular focus on the severity of depression among older adults in South Korea. We found compelling evidence that older adults living with others experienced lower depression severity than those living alone. Furthermore, we identified a notable trend wherein the incidence of depression severity decreased as the number of household members increased: older adults living with one person fare better than those living alone, those with two family members fare better than those with one, and those with three or more members did not exhibit further improvement compared with those with two or more members. The result indicates several factors inherent in living with others, such as increased social interaction, emotional support, and a sense of belonging collectively act to buffer against the negative mental health impacts of loneliness and social isolation. These findings underscore the significance of living arrangements in influencing the mental health of older adults, which aligns with prior studies conducted across diverse national contexts (7, 30, 31).

From a theoretical standpoint, this study posits that the relationship between household environment and mental health is mediated by loneliness stemming from social isolation. While existing studies have predominantly focused on the impact of loneliness on mental health outcomes such as depression (32–34), our study offers a novel perspective by examining the household environment as a precursor to loneliness, which affects the mental well-being of older adults (35–37). Recent studies highlights the role of social support from household members in moderating the effects of loneliness on depression. These findings suggest that the nature of household dynamics directly correlates with the mental well-being of older adults, emphasizing the necessity of a supportive home environment (38, 39). We hypothesized that loneliness is not an inherent characteristic of older adults, but rather a consequence of their environment, with the household context playing a pivotal role. Older adults living alone are susceptible to social isolation and receive limited social and emotional support compared with those residing with family members, which can exacerbate feelings of loneliness and contribute to mental health decline (7).

Older adults who reside alone often experience compromised physical health, primarily due to limited access to healthcare without assistance. However, the correlation between older adults living alone and the experience of loneliness is a pressing concern, which subsequently contributes to a decline in their mental well-being (40, 41). Research indicates that loneliness primarily stems from social isolation characterized by a lack of meaningful social relationships (42). Consequently, the solitary living arrangements of older adults can foster an environment of social isolation, exacerbating feelings of loneliness, and thereby negatively impacting mental health. Studies have consistently identified loneliness as a significant factor that contributes to depressive symptoms in older adults (43, 44). The absence of daily social interactions and the lack of emotional and practical support mechanisms within the household can lead to heightened feelings of loneliness, thereby increasing the risk of depressive symptoms. The increased contact frequency with family and frequent participation in social activities, such as attending adult daycare centers, were associated with reduced loneliness among older adults (45). The importance of family social support in reducing loneliness levels among the older adult suggests that interventions designed to enhance social support systems could help mitigate loneliness and its detrimental effects on mental health (46).

While living alone has become increasingly prevalent in contemporary society, it is imperative to recognize the substantial economic costs associated with loneliness resulting from social isolation in many countries (47). Moreover, mental illness not only incurs an economic burden, but also poses significant social costs, including severe outcomes such as suicide, which warrant serious attention. It is essential to acknowledge that, while living alone may offer convenience and independence to younger generations, it represents a potentially hazardous social environment for older adults, posing risks to their mental and physical health.

South Korea is a high-risk society, with a high number of suicides among older adults. Given that the relationship between mental illnesses such as depressive symptoms and suicide (or suicidal ideation) is straightforward (48), depression due to loneliness may be one of the leading causes of suicide among older adults in South Korea. Our findings have several policy implications.

First, there is a pressing need to revamp the support programs for older adults living alone. Currently, South Korea offers various services for solitary older adults, such as safety verification, life skills education, and in-home assistance (49). However, this study indicates that older adults living with others tend to experience lower levels of depression. This underscores the importance of promoting co-living arrangements through policies and programs tailored to facilitate shared living among older adults, thereby mitigating loneliness and enhancing their mental well-being. Given the increased vulnerability of older adults living alone to social isolation and loneliness, it is imperative that these initiatives target this demographic. Additionally, housing policies should be re-evaluated to prioritize social connectedness and overall well-being, potentially exploring alternative housing options such as co-housing communities or age-friendly developments.

For example, the Naturally Occurring Retirement Community (NORC) program serves as a relevant model (50–53). Furthermore, NORCs residential communities for older adults fosters companionship and support among residents along with opportunities for collective activities and collaboration on common goals. Implementing similar collective residential programs for older adults living alone could effectively reduce social isolation and mitigate the mental health outcomes stemming from loneliness.

Second, our observation of reduced depression severity with increased household size up to a certain threshold underscores the potential benefits of larger households. Living with family members or others offers older adults both companionship and essential practical and emotional support, including assistance with health care needs which can be challenging for those living alone. This support system within the household helps alleviate loneliness and can significantly lessen the severity of depression (38, 39, 45). Policies encouraging multigenerational living arrangements or support systems for older adults cohabiting with family members can contribute to improved mental health outcomes. South Korea is typically categorized as a collectivistic culture (54). The country has changed from a large family system with multiple generations living together centered on the patriarch to a nuclear family system with increased individualism, leading to an increase in single-person households. It is important to emphasize the importance of family support and care for older adults.

Finally, it is crucial to acknowledge the socio-economic costs associated with loneliness and mental illness among older adults. Policies addressing social isolation not only enhance individual well-being but also alleviate economic burdens and prevent severe outcomes such as suicide. Given the correlation between living alone and social isolation among older adults, proactive measures are required to prevent and mitigate social isolation. This may entail community-based interventions, leveraging technology for social connections and fostering intergenerational interactions.