INTRODUCTION
1. Background
The population aged 65 years and older in South Korea in 2025 was approximately 10.51 million, accounting for more than 20% of the total population and indicating that the country is now super-aged [1]. This aging trend is expected to continue, with the share of the population represented by older adults expected to exceed 30% by 2036 and 40% by 2050 [1]. Along with this aging, family structures have become more nuclear; the proportion of older adults living alone reached 32.8% as of 2023, an increase of approximately 13% compared with 2020, and that figure continues to rise [2]. Given the relatively vulnerable circumstances of older adults living alone, this increase is becoming a major issue that must be addressed in a super-aged society [3].
Older adults living alone face heightened vulnerability in both mental and physical health domains due to economic hardships, loneliness, and social isolation [4]. The lack of cohabiting family members limits access to emotional, practical, and informational support, which are essential for maintaining health and well-being [5,6]. These challenges are reflected in consistently poorer self-rated health (SRH) outcomes among this population, including higher rates of chronic illness, functional limitations, depression, and poor nutritional status compared with those cohabiting with others [2,7].
The health of older adults is a complex outcome that reflects accumulated life experiences and disparities in socio-economic resources throughout life. A proper understanding requires a comprehensive approach that considers more than the mere presence or absence of disease and one that includes functional status [8]. SRH is a multidimensional indicator that reflects not only physical health but also psychological well-being, social functioning, and the ability to cope with daily life [9]. It captures older adults’ overall perceptions and satisfaction with their health and has therefore been widely adopted in aging research [9,10]. Despite its subjective nature, SRH has been shown to correlate with objective health outcomes and can be a practical and valid tool for assessing the health status of older adults [10]. SRH has also been widely used to measure health outcomes and predict healthcare utilization. Prior studies have shown that SRH can be influenced by pain, mobility, self-care ability, and psychological factors [9,11]. These health dimensions have a significant effect on both inpatient and outpatient healthcare expenditures [12], making SRH a valuable indicator for identifying vulnerable older adults, informing local health policies, and targeting clinical interventions.
Older adults living alone tend to lack the economic, psychological, emotional, and social resources that are typically obtained through family relationships, and are often living in circumstances where these needs cannot be adequately met [13]. For this reason, community-level environmental factors can be understood as influential elements affecting their self-rated physical health [13]. To more accurately assess SRH among older adults living alone, an integrated approach is required, one that considers not only individual characteristics but structural components of the physical, social, and service environments of the local community; in other words, policy-related factors. This study distinguishes between individual-level characteristics and local government–level policy factors related to SRH among older adults living alone and analyzes those factors using a multilevel model. It was designed to promote a more precise and comprehensive understanding of SRH and provide foundational data that can be used to develop interventions and policies that can effectively enhance health equity.
2. Research Purpose
This study identifies multidimensional factors that can relate to SRH among older adults living alone. Our analysis includes individual-level factors, such as sociodemographic characteristics, health status, and healthcare utilization, and local government–level structural factors, including the proportion of older adults by region, resource allocation, and accessibility of healthcare.
METHODS
Ethics statement: This study was conducted after receiving an exemption from ethical review from the Institutional Review Board of the National Cancer Center (IRB No. NCC2025-0183).
1. Research Design
1) Research Hypotheses and Conceptual Model
This study is a secondary analysis of data from the Community Health Survey (CHS) and Statistics Korea and is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology guidelines (https://www.strobe-statement.org). It uses a multilevel cross-sectional design to examine the effects of individual- and regional-level factors on SRH among community-dwelling older adults living alone. The following research hypotheses were established. First, in the individual-level model, we investigated whether significant regional differences in the average level of SRH among older adults living alone persist after accounting for influential individual-level variables.
Second, in the regional-level model, we examined the extent to which regional variation is explained by welfare budget expenditures for older adults at the metropolitan government level. Another key analytic focus was whether including regional-level predictors, such as local welfare budget allocation, significantly reduced the variance in SRH among older adults across regions.
2. Data Collection and Study Population
We conducted a cross-sectional secondary analysis of data from the 2023 CHS to examine SRH among older adults living alone at the individual level. The CHS is a nationwide survey conducted annually by the Korean Disease Control and Prevention Agency since 2008 to support the planning and evaluation of regional healthcare programs. The target population is adults aged 19 years and older in cities, counties, and districts (si/gun/gu) across South Korea.
The data collection period for the 2023 CHS was May 16 to July 31, 2023. Samples were selected using a multistage stratified probability sampling method that considered region and housing type. Trained surveyors visited the selected households and collected data through face-to-face interviews [14].
Among the 231,752 respondents to the 2023 CHS, 81,898 were adults aged 65 years or older. Response data from 21,677 individuals who reported living alone were selected for this analysis. This group accounted for 9.35% of all respondents (weighted %=10.29).
To estimate the multilevel model, individual-level data were obtained from the CHS, while regional-level data were derived from the Korean Statistical Information Service and other governmental sources. Because each CHS does not include contextual information on regional-level characteristics, such as welfare budget allocation, it was necessary to merge two distinct data sources. This approach is consistent with the rationale for multilevel analysis, which simultaneously considers individual and contextual determinants of health outcomes [15], thereby allowing for a more comprehensive understanding of factors associated with SRH among older adults living alone.
3. Instruments
1) Individual-Level Variables
To identify the determinants of SRH among older adults living alone, we used variables based on the social determinants of the health framework, and selected them in accordance with previous domestic studies [7,8].
(1) General characteristics
The general characteristics included age, sex, monthly income, educational attainment, and employment status. Educational attainment was the highest level of education completed (no formal education, elementary school, middle school, high school, or ≥university). “Other” included responses that could not be classified into standard categories (e.g., no formal education, traditional studies such as Korean classics). Employment status was classified as either “Employed” or “Unemployed.”
(2) Health behaviors and psychosocial characteristics
Variables in this category included current smoking and alcohol use, physical activity level, depressive symptoms, level of stress, perception of the social-physical environment, frequency of social contact, and social activity participation.
Current smoking and alcohol use were both dichotomized into “Yes” or “No.”
Physical activity was measured using the Korean version of the short-form International Physical Activity Questionnaire (IPAQ) [16]. Respondents reported the frequency (days per week) and duration (minutes per day) of vigorous activity, moderate activity, and walking during the previous 7 days. Following the IPAQ scoring protocol, the total physical activity level was calculated by multiplying the assigned metabolic equivalent of task (MET) value for each activity by the minutes performed and the number of days. Specifically, vigorous activity was assigned 8.0 METs, moderate activity 4.0 METs, and walking 3.3 METs. The calculation formula was as follows:
Total MET-min/wk=(8.0×vigorous min/day×day/wk)+(4.0×moderate min/day×day/wk)+(3.3×walking min/day×day/wk)
A higher total MET-min/wk score indicated a higher level of physical activity.
Depressive symptoms were assessed using the Patient Health Questionnaire-9 (PHQ-9) [17]. The PHQ-9 consists of nine items rated on a 4-point Likert scale ranging from 0 (“not at all”) to 3 (“nearly every day”). Total scores range from 0 to 27, with higher scores indicating greater severity of depressive symptoms.
Socio-physical environment was measured using seven standard items from the CHS: trust, mutual support, safety, natural environment, living environment, transportation, and medical services. Responses were coded as “No”=1 and “Yes”=2, and total scores were summed, with higher scores indicating more positive perceptions of the local environment. Stress was measured using a single CHS item: “How much stress do you usually feel in your daily life?” Responses were rated on a 4-point Likert scale (1=“almost none” to 4=“very much”), with higher scores reflecting greater stress. This measure was derived from the CHS and reflects perceived stress frequency rather than a standardized multi-item scale.
Frequency of social contact was measured with three items from the CHS: “How often do you meet or contact your closest relatives (including family), neighbors, or friends (excluding neighbors)?” Responses were coded as follows: “Less than once a month”=1, “Once a month”=2, “2~3 times a month”=3, “Once a week”=4, “2~3 times a week”=5, and “4 or more times a week”=6. Higher scores indicated greater frequency of contact.
Social activities participation was measured with four dichotomous (Yes/No) items: “Do you regularly participate at least once a month in religious, social (e.g., alumni, seniors’ centers, hometown or kinship groups), leisure, or charity activities?” Responses were coded as “No”=1 and “Yes”=2, with higher scores reflecting higher levels of social engagement. Because social interactions are multidimensional, encompassing both the frequency of personal contact and the extent of formal or informal group participation, these two variables were analyzed together to provide a comprehensive assessment of social contact. Furthermore, as the CHS does not provide a validated instrument for assessing social contact, frequency-based measures derived from these items were used to address this limitation.
2) Regional-Level Variables
Based on the recommendations of prior studies [7,8], regional-level variables included the following:
• The welfare budget for older adults per metropolitan municipality (including support for senior welfare centers, senior employment programs, senior centers, housing and medical care for older adults, community-based care facilities, basic old-age pensions, care services for older adults, and both national and local-level programs)
• Population aged 65 years or older
• Number of healthcare resources (e.g., public health centers, clinics, and hospitals)
• Number of senior leisure and welfare facilities
The distributions of several regional-level variables (e.g., welfare budgets for older adults, number of healthcare resources, number of leisure and welfare facilities for older adults) were highly skewed. To address non-normality and improve model stability, these variables were natural log-transformed prior to multilevel analysis. For clarity, descriptive statistics and tables present the original, untransformed units, while log-transformed values were used in the regression models.
4. Statistical Analysis
Multilevel analysis was conducted using the HLM program (Scientific Software International, Inc.) to account for the hierarchical structure of the data, with individuals (level 1) nested within metropolitan regions (level 2). A four-step modeling strategy was applied. First, a null model without predictors was estimated to calculate the intra-class correlation coefficient (ICC). Second, an individual-level model was constructed including sociodemographic, behavioral, and psychosocial variables. Third, a regional-level model was tested to examine contextual factors, such as welfare budgets for older adults, healthcare resources, and the number of senior welfare facilities. Finally, a full model incorporating both individual- and regional-level predictors was estimated. Cross-level interaction terms were added to determine whether the effects of regional-level welfare resources varied by individual characteristics such as age, income, and physical activity. A random-intercept-only model was specified, assuming intercepts vary across regions but slopes are fixed. Parameters were estimated using Maximum Likelihood for all models to allow for comparison using likelihood ratio tests. The final full model parameters were re-estimated using Restricted Maximum Likelihood for the most accurate variance component estimation. Fixed effects were presented as regression coefficients (β) with associated p-values, and random effects were reported as variance components at each level. All continuous predictors were mean-centered prior to analysis to facilitate the interpretation of cross-level interaction terms. Model fit was evaluated using likelihood ratio tests and pseudo R2 statistics. Statistical significance was set at p<.05 (two-tailed).
5. Ethical Considerations
This study was conducted after receiving an exemption from ethical review from the Institutional Review Board of the National Cancer Center (NCC2025-0183). The CHS is a publicly available dataset, and the researchers obtained approval to access and use the original data in accordance with the data disclosure and management policies of the Korea Disease Control and Prevention Agency. All data used in the analysis were anonymized prior to use.
RESULTS
1. General Characteristics of the Study Population
The general characteristics of the individual- and regional-level variables are presented in Table 1. The mean age of the participants was 76.80±7.32 years. Of the total sample, 5,301 individuals (24.5%) were male and 16,376 (75.5%) were female. The average monthly income was 945,400±885,500 KRW. With respect to educational attainment, 52.9% of participants had completed elementary school or less (24.1% had no formal education and 28.8% were elementary school graduates), representing the largest subgroup. Overall, 8,045 participants (37.1%) were currently employed. Regarding health behavior, 5,376 participants (24.8%) reported a current smoking habit, and 13,476 (62.2%) reported consuming alcohol. The average physical activity level was 1,382.04±2,525.83 METs. The mean score for SRH was 2.61±0.95. The average perceived stress level was 1.77±0.78, and the mean depression score was 3.20±3.93. The average score for the perceived social and physical environments was 8.34±1.44. The mean frequency of social contact was 12.01±3.98, and the average social activity participation score was 7.14±0.91.
For regional-level variables, the average number of older adults per metropolitan area was 540,052.48±520,600.00. The mean welfare budget for older adults was 254,975.35±134,742.30 million KRW. The average number of healthcare facilities was 5,986.00±6,860.48, and the average number of senior leisure and welfare facilities was 4,144.41±3,308.48.
2. Model Estimation
1) Null Model (Random Intercept Model)
In the unconditional (null) model, the variance at the regional level was estimated to be 0.030 (p<.001), and the residual variance at the individual level was 0.245 (p=.010). The ICC was 0.109, indicating that 10.9% of the total variance in SRH among older adults living alone was attributable to differences between regions. This finding supports the suitability of multilevel modeling for the data.
2) Random Coefficient Model (Unconditional Slope Model)
In the model including only individual-level predictors, several factors were found to significantly influence SRH. As age increased, SRH decreased by a factor of 0.216 (β=-0.216, p<.001). Females reported significantly lower SRH than males (β=-1.419, p=.021). Current smokers reported significantly lower SRH compared with non-smokers (β=-0.982, p<.001). Current alcohol use was significantly associated with better SRH (β=1.299, p=.010). Depressive symptoms were also significantly associated with lower SRH (β=-0.308, p=.013). Higher levels of physical activity (β=0.506, p=.001) was associated with a higher SRH. However, the effects of educational attainment, employment status, level of stress, frequency of social contact, and social activity participation were not significant in this model.
3) Conditional Model (Full Model With Regional-Level Variables)
When both individual-level and regional-level variables were included in the model, several individual-level predictors remained significant. Age was negatively associated with SRH (β=-0.215, p=.022), and females reported poorer SRH than males (β=-1.618, p=.024). Higher monthly income was positively associated with SRH (β=0.088, p<.001), and lower educational attainment was associated with poorer SRH (β=-0.147, p<.001). Compared with non-smokers (reference), smokers had lower SRH (β=-0.974, p<.001). Physical activity levels (β=0.402, p=.001)) and perceptions of the social and physical environment (β= 0.621, p=.256) were also positively associated with SRH. Depressive symptoms (β=-0.330, p=.027) and level of stress (β=-0.605, p=.017) were both significantly associated with lower SRH. Social activities participation was not significant factor in this model. These findings are summarized in Table 2, which presents the estimation results for models 1, 2, and 3.
At the regional level, higher welfare budgets for older adults (β=2.748, p=.004), more healthcare facilities (β=1.137, p=.005), and more senior leisure and welfare facilities (β=0.572, p=.002) were significantly associated with higher SRH.
4) Cross-Level Interaction Model
When we tested for cross-level interaction effects between individual- and regional-level variables, significant interactions were evident. The interaction effect on SRH between the welfare budget for older adults and age was negative and significant (β=-0.029, p=.003), as was the interaction with monthly income (β=-0.012, p<.001). In contrast, the interaction effect between the welfare budget for older adults and physical activity level on SRH was positive and significant (β=0.017, p=.001). These findings suggest that the effects of the welfare budget for older adults on SRH varies by individual characteristics such as age, income, and physical activity. Other interaction terms (sex, education, and current smoking) were tested but were not statistically significant.
DISCUSSION
This multilevel analysis examined the multilevel associations with SRH among older adults living alone, distinguishing between individual-level factors (level 1) and local government–level factors (level 2). The analysis revealed that individual-level factors had the greatest influence on SRH among older adults living alone, although certain policy-related local government–level factors also had a significant effect. Interaction effects between individual-level and local government–level variables were also identified.
At the individual level (level 1), age, sex, educational attainment, monthly income, smoking status, level of physical activity, current alcohol use, depressive symptoms, and level of stress all had significant effects on SRH. An older age was associated with lower SRH, which is consistent with previous findings indicating that SRH tends to decline as age increases [11]. Older individuals tend to have a higher prevalence of chronic diseases and greater physical functional decline, which increased the likelihood that they will perceive themselves to be in poor health [18]. Females showed lower SRH than males, which also aligns with prior studies [11]. Older females living alone had overlapping vulnerabilities and tended to receive relatively less economic and emotional support from family and society than men, which can negatively affect their SRH [19].
Monthly income and educational attainment, both indicators of socioeconomic status, were positively associated with SRH. This suggests that higher economic resources provide better access to medical services and resources necessary for health maintenance, which is a key factor in improving SRH among older adults living alone [18]. Conversely, higher depressive symptoms and stress levels were significantly associated with lower SRH. This highlights the critical need for mental health support and stress management programs tailored for this vulnerable population [11].
Higher levels of physical activity among older adults living alone were associated with a higher SRH. This may indicate that physical activity not only provides improvements in physical functioning but also has psychological and social benefits, such as enhanced self-esteem and life satisfaction [20]. The development of positive emotions can help individuals perceive their health more positively. In particular, physical activity in older adults has been shown to positively influence not only SRH but also other health indicators, including improvements in health-related quality of life, prevention of functional decline, and reduced risk of premature mortality [21]. Accordingly, sustainable public health policies and programs that promote physical activity are essential.
Current alcohol use was significantly associated with higher SRH. This counterintuitive finding requires careful interpretation, as excessive alcohol use is generally detrimental to health. It may suggest that, in this specific population of older adults living alone, current, moderate alcohol consumption is a proxy for better social engagement or fewer chronic illnesses, contributing to a more positive self-perception of health [22,23].
Interestingly, smokers reported lower SRH compared with non-smokers, which appears counterintuitive, given the established health risks of smoking. Prior evidence consistently shows that smoking is associated with increased poor SRH, emphasizing the need for health professionals to strengthen preventive communication about its harmful consequences [24]. However, the picture is more complex when living arrangements are taken into account. Henning-Smith and Gonzales [25] reported that older adults living alone did not necessarily rate their health more poorly than did those living with others, suggesting that the relationship between living alone and SRH varies by age and may be moderated by other health behaviors. This also raises the possibility that the association between smoking and SRH may differ among older adults living alone, compared with cohabiting peers. Further investigation into how smoking behavior intersects with living arrangements to influence SRH in older adults is warranted, particularly to determine whether unique psychosocial dynamics in single-person households mitigate the negative health perceptions typically associated with smoking.
At the local government level (level 2), the welfare budget, the number of healthcare facilities, the number of leisure welfare facilities for older adults were all found to influence the SRH of older adults living alone. In addition, reported SRH was higher in areas with a more leisure welfare facilities for older adults, presumably because an abundance of leisure resources within a region improves service accessibility for older adults living alone, promoting their participation in leisure activities [26]. This finding resonates with previous research in a similar context [27,28] which demonstrated that participation in leisure activities by older adults living alone positively contributed not only to their physical and mental health but also to a reduction in loneliness, depression, and suicidal ideation. To promote health and improve quality of life among older adults living alone, public leisure welfare services that can encourage engagement in such activities should be expanded.
In addition, regions with more healthcare facilities showed higher levels of SRH, likely because improved healthcare accessibility facilitates the management and prevention of chronic diseases. According to a report by the Korea Institute for Health and Social Affairs [29], 45.6% of older adults living alone reported that they were unable to receive appropriate nursing services when they experienced health problems, a response rate higher than those for other difficulties, such as financial instability, social isolation, or safety concerns. In this context, a shortage of healthcare institutions within a region is likely to lead to unmet healthcare needs, which limits opportunities for early detection and preventive intervention and can lead to negative health outcomes, such as worsening of disease, increased complications, and high mortality rates [30]. For older adults living alone, adequate provision of local healthcare resources and improved accessibility is likely to be key to enhancing health status, highlighting the need for policy interventions based on this foundation.
A key contribution of this study lies in its analysis of cross-level interactions. The results demonstrated that the welfare budget for older adults was negatively associated with age and income, but a positive interaction was seen with physical activity. These findings suggest that increases in welfare budgets may yield greater improvements in SRH among younger or lower-income subgroups, while older adults with higher levels of physical activity may experience amplified benefits. This reveals the importance of integrating individual behavioral factors with structural policy support. To effectively enhance the health of older adults living alone, local governments should account for heterogeneity within this population and design tailored strategies rather than adopting uniform policy measures.
Another important dimension is the difference between older adults living alone and those living with others. Prior studies have indicated that co-residing older adults generally report better SRH, largely due to greater access to instrumental and emotional support from family members [8]. In contrast, those living alone are more likely to experience social isolation, limited access to informal care, and heightened psychological vulnerability [13]. The present findings further highlight this disparity: the health of older adults living alone was particularly sensitive to community resources such as welfare budgets and facility availability, whereas studies of co-residing older adults found weaker associations. This suggests that living alone amplifies dependency on external resources, while family support may buffer certain health risks for those living with others. Accordingly, interventions to reduce health inequalities should treat living arrangement as a structural determinant and prioritize older adults living alone, who are at greater risk of experiencing compounded vulnerabilities.
The limitations of this study are four-fold. First, although we used a nationally representative dataset, selection bias due to the exclusion of institutionalized older adults and reporting bias in SRH and behavioral variables is possible. However, the use of standardized survey procedures and face-to-face interviews likely mitigated some measurement errors. Second, as this was a cross-sectional study, causal inferences cannot be made. Third, due to data limitations of the CHS, major variables known to influence subjective health and quality of life, such as specific chronic diseases, nutritional status, and cognitive function, were not included in the analysis. Additionally, constructs such as stress and social activity participation were assessed using single items, which may not fully capture the complexity of these concepts. Future research should consider using validated multi-item instruments to provide a more comprehensive assessment. Finally, while this study used a nationally representative complex sampling of data, we did not apply weights in the multilevel analysis. Although weighted descriptive statistics were closer to the population average, our preliminary analysis found no substantial differences in the fixed and random effects estimates between the weighted and unweighted multilevel models. This was a methodological choice to maintain model stability and interpretability. Future studies using similar datasets could explore the impact of weighting on multilevel model results.
CONCLUSION
This study conducted an integrated analysis of multilevel factors associated with SRH among older adults living alone by considering both individual-level and local government–level factors. The findings suggest that improving the health status of this population requires not only behavioral changes at the individual level, but redistribution of structural and institutional support at the community level. Two policy and practical recommendations to enhance SRH among older adults living alone follow from our findings. First, nurses should assess the health needs of individuals while considering the characteristics and limitations of local resources and tailor interventions to promote physical activity, provide emotional support, and strengthen community connections. Second, efforts should be made when formulating public nursing policies to reallocate resources and improve equity in regions with limited access to medical and leisure resources. However, such policies must incorporate context-sensitive strategies, recognizing that the benefits of regional welfare are attenuated for older individuals and those with higher income, and are amplified for those engaged in physical activity. Structural, context-sensitive interventions at the community level are essential to improving health equity and well-being in this population.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation 
Print 
