Abstract
Thailand has a higher depression rate among older adults than other Southeast Asian countries, and the proportion of older adults living alone is growing. However, evidence regarding the relationship between living alone and depression among older adults in Thailand is scarce. Thus, this study examined this issue, focusing on two research objectives: (1) to examine the relationship between living alone and self-assessed depression using the matched sample of data acquired by propensity score matching and (2) to explore sociodemographic patterns in self-assessed depression using the full sample of data. This study employed a cross-sectional design with individuals aged 65 or above from the 2019 Health and Welfare Survey dataset (N = 6,164, including a matched subsample of 2,702 individuals). Ordinal logistic regression analyses were performed using a single-item measurement of self-assessed depression. The results showed that older adults living alone reported higher self-assessed depression than those living with family, suggesting that living alone could independently increase the risk of depression among this group, regardless of sociodemographic factors (objective 1). We also found sociodemographic disparities in self-assessed depression. Specifically, older adults who were low-income, female, unemployed, chronically ill, or living in rural regions reported higher self-assessed depression than their counterparts (objective 2). These findings reveal a need to address the increased risk of depression among older adults living alone, which could be accomplished, for example, by establishing a social support network to supplement the support traditionally provided by older adults’ families in helping them overcome daily life challenges. Such a network could be created by integrating social work professionals into existing community health volunteer programs as a pragmatic approach. Additionally, reinforcing external resources, such as participation in social activities, could reduce sociodemographic disparities in mental health. Thus, the government should continue developing community-based social participation programs for older adults.
Keywords
Introduction
Thailand’s population is rapidly aging, and approximately 15% of its citizens were aged 65 or above in 2023. This percentage is the highest in Southeast Asia and is expected to reach approximately 30% by 2050 (United Nations, 2024). Meanwhile, the proportion of older adults who live alone is also growing. In 2002, 6.3% of the older population was living alone (National Statistical Office of Thailand, 2018), and this percentage had increased almost threefold to 16.2% by 2019 (Paek & Zhang, 2025). This increase is even more striking considering that the percentage recorded in 2002 included people aged 60 and above, whereas the 2019 percentage included those aged 65 and above.
Moreover, depression, which is a highly prevalent mental health issue in older adults, has become a public concern in Thailand (Charoensakulchai et al., 2019; Paek & Zhang, 2024). In 2021, people aged 70 or above who experienced depressive symptoms made up approximately 4.9% of Thailand’s total population. This percentage ranks third in Southeast Asia after Malaysia (5.6%) and the Philippines (5.1%; Institute of Health Metrics and Evaluation, 2024).
Older adults generally encounter many negative life changes, such as physical frailty and the death of a spouse or other significant people, as well as economic insecurity and decreased social roles after retirement. These physical, psychosocial, and economic stressors can elevate the risk of adverse mental health outcomes, such as depression (Kim et al., 2018; Lim et al., 2017; Lin & Wang, 2011).
These challenges can be especially serious for older adults living alone, as they often lack diverse family resources and support, making it difficult to cope with the stresses associated with aging. According to social convoy theory, families can serve as a “convoy” to assist their senior members in overcoming life’s challenges by providing material and nonmaterial support in the form of, for example, giving care when an older family member suffers an illness, assisting them in maintaining their independence in daily activities, and offering emotional support. According to this theory, families represent the most basic and essential social network for supporting older adults (Antonucci & Akiyama, 1987; Jia et al., 2023; Li et al., 2018). Therefore, older adults living alone may have a high likelihood of developing mental health conditions, including depression.
This theoretical perspective is especially evident in Asian countries, such as Thailand, where family solidarity and filial piety are valued, as well as in underdeveloped countries that provide inadequate public welfare support for older adults (Chinnappan, 2015; Kamiya & Hertog, 2020; Teerawichitchainan et al., 2015). Especially in Thailand, which has traditionally maintained an extended family system, families have served as the primary source of support that senior family members rely on to meet their everyday needs. This support helps reduce loneliness and social isolation while allowing older adults to maintain their emotional stability, which, in turn, helps prevent adverse mental health outcomes, such as depression (Teerawichitchainan et al., 2015).
However, primarily due to urbanization and economic growth, the traditional family structure has become less common as nuclear families have become the norm. As a result, the number of older adults living alone is increasing. This shift has weakened family support, which can negatively affect people’s mental health, with older adults living alone being particularly at risk.
Literature Review and Research Objectives
While the relationship between living alone and depression in older adults has been examined in various countries, research on this issue in Thailand is limited, as, to our knowledge, only one study has specifically examined this matter. Overall, a positive association has been found, suggesting that older adults who live alone are more likely than other groups to experience depression (Chou et al., 2006; Fang et al., 2024; Fukunaga et al., 2012; Huang et al., 2023; Jeon, Jang, et al., 2007; Jeon, Choi, et al., 2017; Joutsenniemi et al., 2006; Lee & Kim, 2024; Nguyen et al., 2024; Noguchi et al., 2023; Osborn et al., 2003; Paek & Zhang, 2024; Park et al., 2017; Pei et al., 2022; Russell & Taylor, 2009; Stahl et al., 2017; Waite & Hughes, 1999; Zheng et al., 2023).
However, the strength of this relationship varies depending on individual sociodemographic and contextual factors. Several studies have found that economic well-being or community social support, among other factors, can modify this relationship. Specifically, these factors could mitigate the positive association (Chou et al., 2006; Fukunaga et al., 2012; Jeon, Jang, et al., 2007; Jeon, Choi, et al., 2017; Joutsenniemi et al., 2006; Nguyen et al., 2024; Noguchi et al., 2023; Osborn et al., 2003; Park et al., 2017; Russell & Taylor, 2009; Waite, & Hughes, 1999; Zheng et al., 2023). In some cases, these factors weakened the relationship between living alone and depression such that it became statistically insignificant (Chou et al., 2006; Fukunaga et al., 2012; Nguyen et al., 2024; Osborn et al., 2003; Park et al., 2017; Russell & Taylor, 2009; Zheng et al., 2023) or even negative (Jeon, Jang, et al., 2007; Jeon, Choi, et al., 2017; Noguchi et al., 2023).
While these previous studies have contributed significantly to our understanding of this relationship, they may have an important limitation. Namely, they may not have adequately controlled for confounding effects, and it remains unclear whether living alone directly causes depression or if this association is an artifact of other sociodemographic or contextual factors.
For example, the positive relationship between living alone and depression may be confounded by income. It is plausible that income is associated with lower depression and that older adults living alone have lower incomes than those living with others. In this case, older adults living alone would appear to be more depressed than those living with others, not because they live alone but because they have a low income.
The abovementioned scenario is likely to be accurate in Thailand, where sociodemographic status is substantially lower among older adults living alone than those living with family (Meemon & Paek, 2020; Paek & Zhang, 2025). A previous study analyzing nationwide data reported that older adults living alone, compared to those living with family, had higher percentages of older, female, low-income, less-educated, and chronically ill individuals. Especially regarding income, the national poverty rate for older adults living alone was approximately 40%, nearly three times higher than the 14% of those living with family (Meemon & Paek, 2020). This highlights the need for a more rigorous analysis that can distinguish between these effects.
In addition, other studies in Thailand have shown sociodemographic disparities in depression (Aung et al., 2016; Haseen & Prasartkul, 2011; Paek & Zhang, 2024; Ratanasiripong et al., 2022; Suttajit et al., 2010). While findings vary across studies, income and education were commonly identified as protective factors against depression. However, since nearly all of these studies were conducted in specific geographical areas, evidence from nationwide data is still needed to validate the generalizability of their findings (Aung et al., 2016; Haseen & Prasartkul, 2011; Ratanasiripong et al., 2022; Suttajit et al., 2010).
Therefore, given these observed disparities and the potential for confounding, this study uses a nationwide data source to accomplish two research objectives: (1) to examine the relationship between living alone and self-assessed depression by analyzing a matched sample of data acquired by a propensity score matching (PSM) method to equally balance sociodemographic status between older adults living alone and those living with family and (2) to explore overall sociodemographic patterns in self-assessed depression by analyzing the full sample of data.
Based on the literature review and the research objectives, this study established and tested the two hypotheses as follows.
Methods
Data Source and Study Sample
This study employed a cross-sectional design using the 2019 Health and Welfare Survey (HWS) dataset, a nationwide survey dataset that contains a representative sample of the Thai population. Samples in the dataset are drawn from all 76 Thai provinces in equal proportions to the population size of each province using a two-step stratified sampling procedure. The dataset is composed of four sections: (1) sociodemographic status, (2) illnesses and health care utilization, (3) health-risk behaviors, and (4) housing conditions. The Thai National Statistical Office collects and releases the dataset biannually.
The 2019 dataset was used in this study since it was the most recent report with a complete set of information. The 2021 and 2023 datasets were also collected, but the 2021 dataset was not released, as it was collected during the COVID-19 pandemic. Meanwhile, the 2023 dataset omitted some important variables of interest in this study, such as income.
The study sample for analysis included individuals aged 65 and above. Since this study included two research objectives, we created two distinct samples. The first is the matched sample of data acquired by a PSM for the first research objective (to evaluate the relationship between living alone and self-assessed depression after adjusting for confounders). The second is the full sample of data for the second research objective (to evaluate overall sociodemographic patterns in self-assessed depression).
The PSM used was one-to-one caliper matching with a caliper width of .005 and no replacement. The .005 caliper width was chosen based on the statistical equality of sociodemographic status between the matched cases and the number of matched cases, as suggested by previous studies (Connolly & Gagne, 2016; Lunt, 2014).
Specifically, propensity scores were estimated based on a binomial logistic regression analysis in which living alone (yes vs. no) served as the dependent variable and all other sociodemographic factors were independent variables. Particularly regarding the categorization of living alone, the HWS dataset provided information on the total number of family members in the household (including immediate family members and other relatives). The “yes” group (living alone) was defined as having one family member in the household, and the “no” group (living with family) was defined as having two or more family members.
The PSM method was repeated, with the caliper width progressively reduced by .005 from the optimal value of .2, as proposed by Austin (2011). When the caliper width exceeded .005, certain variables, such as income and education, showed statistically significant differences between the living-alone and living-with-family groups. Conversely, when the caliper width was below .005, the number of matched cases substantially decreased, indicating that only certain groups of older adults (e.g., high-sociodemographic groups) were matched. The caliper width of .005 could ensure the number of matched pairs and the level of equality between the two groups.
The original 2019 HWS dataset included 63,594 individuals, 9,927 (15.6%) of whom were aged 65 or above. We identified that 3,759 of these individuals had missing values concerning their self-assessed depression, while four individuals had unidentified values regarding their education. After excluding these missing and undefined values, 6,164 individuals (1,355 living alone) were used as the full sample of data for the second research objective. After caliper matching, 1,351 matched individuals were used as the matched sample of data for the first research objective, accounting for 99.7% of the full sample of data from individuals living alone. Figure 1 is a flowchart illustrating the study’s sample selection process.
Process for selecting the study sample.
Variables and Measurement
The outcome variable, self-assessed depression, was treated as an ordinal variable with three levels (none, slight, and moderate or higher), measured using a single-item question since the dataset did not include a multiple-item depression scale, such as the “Geriatric Depression Scale.” Specifically, participants were asked, “How depressed do you feel nowadays?” The original responses were based on a 5-point scale (none, slight, moderate, severe, and extreme). However, because very few participants responded with the “severe” or “extreme” options, they were combined with the “moderate” category to create the new “moderate or higher” category in this study.
In addition, after considering previous studies and the availability of variables in the dataset, we considered eight sociodemographic factors as independent variables: living alone, income, age, gender, education, working status, chronic condition, and region. Age and income were assessed on a continuous scale. In particular, income was assessed using the household equivalence income scale (Organization for Economic Co-operation and Development, 2009).
The remaining variables were measured on a dichotomous scale: living alone (yes vs. no), gender (male vs. female), education (high vs. low), working status (yes vs. no), chronic conditions (yes vs. no), and region (urban vs. rural). Regarding education, the “low” category indicated primary education or lower, while the “high” category indicated secondary education or higher. All other variables, including the “living alone” categories previously defined, followed the standard yes versus no or categorical labels.
Statistical Analysis
Descriptive analysis was first conducted to encapsulate the study samples and variables. In the analysis, sociodemographic status was compared between the living-alone and the living-with-family groups using a Chi-square test and t-test to evaluate the performance of the caliper matching used.
The outcome variable (self-assessed depression) was an ordinal variable including three levels (none, slight, and moderate or higher); accordingly, a proportional odds model (POM) was used to investigate the relationship between the outcome and independent variables (Hosmer & Lemeshow, 2000). POM, known as an ordinal logistic regression model, is a widely used statistical method for analyzing multiple factors related to outcomes with an intrinsic order, such as “low,”“middle,” and “high.” Thus, POM is a suitable analytical method to accomplish this study’s research objectives.
Specifically, two separate POMs were established according to each research objective. One POM used the matched sample of data for the first objective (to evaluate the relationship between living alone and self-assessed depression after adjusting for confounders). The other POM used the full sample of data for the second research objective (to evaluate overall sociodemographic patterns in self-assessed depression).
The parallel lines assumption and the lack of fit of the POMs were tested using the score test and Pearson’s goodness-of-fit test, respectively. Moreover, to ensure reliable regression coefficients, we inspected multicollinearity in the POMs according to the variance inflation factor and crosschecked the crude odds ratio (COR) and adjusted odds ratio (AOR). The scores of the variance inflation factor ranged between <1.00 and 1.43. The crosscheck showed no substantial regression coefficient changes between COR and AOR. Neither finding suggested multicollinearity in the POMs (Hosmer & Lemeshow, 2000; Menard, 2002). Data processing and analysis were carried out using SAS version 9.2 and IBM SPSS Statistics version 20.
Results
Descriptive Analysis
Table 1 shows the findings of the descriptive analysis. The findings from the full sample of data (N = 6,164) show that aspects of sociodemographic status differed significantly between older adults living alone and those living with family. Specifically, compared to older adults living with family, those living alone had a lower mean income (6,047.84 baht/$178 USD versus 8,584.76 baht/$253 USD), a higher mean age (74.08 years vs. 72.39 years), and higher percentages of females (67.75% vs. 57.58%) and unemployed individuals (79.04% vs. 68.12%).
Descriptive Statistics of the Study Samples and Variables.
Note. SD = standard deviation; Education Low and High = primary education or lower and secondary education or higher.
Meanwhile, the findings from the matched sample of data (n = 2,702) show that the caliper matching used equally balanced sociodemographic status between the living-alone and living-with-family groups (p-values > .05 for the Chi-square and t-tests). Figure 2 shows the propensity score distributions for the living-alone and living-with-family groups before and after PSM. After matching, the distributions are relatively identical, suggesting that the PSM method effectively matched older adults living alone with those living with family.
Propensity score distributions before and after PSM.
The findings from the matched sample of data show that self-assessed depression was higher in older adults living alone than in those living with family. Specifically, the percentage of self-assessed depression in the “none” category was 79.35% for older adults living alone, while it was 82.98% for those living with family. Meanwhile, the percentages in the “slight” and “moderate or higher” categories were 17.47% and 3.18% for older adults living alone, respectively, which were higher than 15.03% and 2.00% for those living with family. These findings suggest that living alone could increase depression among older Thais, irrespective of variations in their sociodemographic status.
Proportional Odds Model
Tables 2 and 3 present the findings of the POMs using the matched and full samples of data, respectively. According to the score test (p-value = .063 and .260) and goodness-of-fit test (p-value = .976 and .297), the parallel lines assumption and goodness-of-fit were satisfied in the POMs.
Proportional Odds Model Using the Matched Sample of Data (n = 2,702).
Note. COR and AOR = crude and adjusted odds ratio; 95% CI = 95% confidence interval; DF = degrees of freedom; Education Low and High = primary education or lower and secondary education or higher.
Statistically significant at .05.
Proportional Odds Model Using the Full Sample of Data (n = 6,164).
Note. COR and AOR = crude and adjusted odds ratio; 95% CI = 95% confidence interval; DF = degrees of freedom; Education Low and High = primary education or lower and secondary education or higher.
Statistically significant at .05.
Table 2 reveals the findings of the POM using the matched sample of data for the first research objective (to evaluate the relationship between living alone and self-assessed depression after adjusting for confounders). The findings reveal that living alone was positively related to self-assessed depression, with an AOR of 1.28 and a 95% confidence interval (CI) [1.06, 1.56]. Specifically, the AOR value indicates that for older adults living alone, the odds of having a feeling of slight or moderate or higher depression versus a feeling of no depression were 1.28 times that of those living with family. Rejecting the null hypothesis 1 (i.e., no relationship between living alone and self-assessed depression), this result shows that older adults living alone had higher levels of self-assessed depression than those living with family. This suggests that living alone could enhance the risk of depression for older Thais, irrespective of their sociodemographic status.
Table 3 reveals the findings of the POM using the full sample of data for the second objective (to evaluate overall sociodemographic patterns in depression). The findings reveal a positive association between living alone and self-assessed depression, with an AOR of 1.25 and a 95% CI [1.07, 1.46].
In addition, the findings show sociodemographic disparities in self-assessed depression. Specifically, low-income groups tended to have higher self-assessed depression than high-income groups, with an AOR of .77 (95% CI [.67, .89]). The older groups also tended to have a slightly higher self-assessed depression than younger groups, with an AOR of 1.02 (95% CI [1.00, 1.03]). Furthermore, self-assessed depression was elevated among females (AOR = 1.18, 95% CI [1.02, 1.36]), as well as individuals who were unemployed (AOR = 1.31, 95% CI [1.11, 1.56]), suffering from chronic conditions (AOR = 2.02, 95% CI [1.74, 2.35]), and living in rural areas (AOR = 1.15, 95% CI [1.01, 1.32]). Rejecting the null hypothesis 2 (i.e., no sociodemographic disparities in self-assessed depression), these findings consistently indicate that older adults with low sociodemographic status had higher levels of self-assessed depression than those with high sociodemographic status.
Discussion
Objective 1: Living Alone and Self-Assessed Depression
Regarding the first research objective, the use of caliper matching to control for possible confounding variables revealed that older adults living alone tended to have higher levels of self-assessed depression than those living with family. This means that living alone could independently increase the risk of depression among older Thais, regardless of sociodemographic factors. This finding corresponds to previous findings (Chou et al., 2006; Fang et al., 2024; Fukunaga et al., 2012; Huang et al., 2023; Jeon, Jang, et al., 2007; Jeon, Choi, et al., 2017; Joutsenniemi et al., 2006; Lee et al., 2024; Nguyen et al., 2024; Noguchi et al., 2023; Osborn et al., 2003; Paek & Zhang, 2024; Park et al., 2017; Pei et al., 2022; Russell & Taylor, 2009; Stahl et al., 2017; Waite, & Hughes, 1999; Zheng et al., 2023). Furthermore, this finding indicates that the observed relationship is unlikely to be driven solely by sociodemographic confounders, as mentioned in the introduction.
In addition, this finding partly supports social convoy theory, as it implies that families can act as a social support network and protect their older relatives against mental health issues. However, due to limitations related to the use of secondary data and a lack of relevant indicators, this study could not clarify the theory’s sequential mechanisms (i.e., family support helps reduce loneliness and social isolation and maintain emotional stability, which can eventually prevent adverse mental health outcomes). Future studies should fill this methodological gap by utilizing mediation analysis that considers emotional states, such as loneliness and social isolation.
Furthermore, this study’s quantitative approach provided a limited understanding of the specific effects of family support on depression (e.g., material versus emotional support or support for daily life versus support for occasional serious incidents). This approach also failed to explain the relationship observed in this study from the perspective of Thailand’s traditional family system. Future studies could resolve this limitation by utilizing a qualitative research approach, such as in-depth interviews and case studies.
Based on our findings, we recommend establishing a social support network that can supplement the traditional family role in assisting older adults to overcome daily life challenges, as this would help reduce the risk of depression among older adults living alone. One approach to establishing such a network could be to integrate social work professionals into existing community health volunteer programs.
Thailand has implemented community health volunteer programs to address the problem of low access to health care due to a lack of resources and infrastructure. Each volunteer monitors 5–15 households in a village or community and connects individuals with limited mobility (e.g., those with disabilities) to health care providers through free vehicle support and other arrangements (Meemon & Paek, 2020). The government should consider employing additional social work professionals in these programs. The duties of these social work professionals could include providing home care services for older adults living alone, especially those who have difficulty with independent living and often find themselves disconnected from their families and communities. As a starting point, a pilot program could be carried out in a few select communities before efforts are gradually expanded according to evaluations and available resources.
Objective 2: Sociodemographic Disparities in Self-Assessed Depression
Concerning the second research objective, the findings from the full sample of data indicate sociodemographic disparities in self-assessed depression. Specifically, low-income, older-aged, female, unemployed, chronically ill, and rural groups tended to have elevated levels of self-assessed depression. These sociodemographic disparities, particularly those associated with income and education, support previous Thai studies conducted in specific areas (Aung et al., 2016; Haseen & Prasartkul, 2011; Ratanasiripong et al., 2022; Suttajit et al., 2010). Based on this, the identified disparities may exist in the older population throughout Thailand.
Sociodemographic disparities can be understood through social stress theory, which posits that individuals with a disadvantaged social status are likely to experience poor mental health because of their greater likelihood of facing stressful circumstances and their lack of coping resources. The theory suggests that enhancing external resources (e.g., community-based social participation programs) can effectively reduce the sociodemographic disparities associated with poor mental health (Lincoln et al., 2005; Meyer et al., 2008). Consistent with this perspective, previous studies have demonstrated that social (e.g., sports clubs and recreational activities) and productive engagement (e.g., paid working and unpaid volunteering) are protective against depression in older adults (Fang et al., 2024; Huang et al., 2023; Paek & Zhang, 2024; Zheng et al., 2023).
Thailand has emphasized the significance of social participation for the mental health of older adults. Accordingly, local governments and relevant social and religious organizations have been developing and providing older adults with various social participation programs, such as sports and recreational programs and lifelong education programs, to promote their psychological well-being and prevent social isolation (Department of Older Persons, 2023). Studies have found that social participation programs benefit mental health of older adults (Aung et al., 2016; Apinunmahakul, 2012; Sasiwongsaroj et al., 2015).
However, to our knowledge, no studies have evaluated these programs’ overall accessibility and participation rates. Some studies have revealed that older adults’ accessibility to health care was substantially poorer than younger adults’ and was disproportionate to sociodemographic status among older adults. The unavailability of public transportation (especially for those in rural regions) and companions (especially for those with limited mobility) are frequently mentioned as reasons for low accessibility (Meemon & Paek, 2020; World Bank, 2016). We expect the accessibility of social participation programs for older adults to reflect such findings. Thus, the government needs to assess the overall participation rate and accessibility of social participation programs and develop policies to improve them.
Study Limitations
This study has some noteworthy limitations that future studies could address. First, future studies should consider subdividing the simple dichotomous category of housing type used in this study (i.e., living alone versus living with family). Specifically, regarding the living alone category, we did not consider that some older adults living alone may have children who live nearby and visit them frequently, as observed in a previous study in Thailand (Teerawichitchainan et al., 2015). The level of depression among older adults in this category may differ from that of older adults living alone with no children living nearby. Likewise, the category of living with family may contain various housing types (e.g., living with a spouse only, living with children only, or both), and depression might differ among people living in these housing types (Noguchi et al., 2023; Park et al., 2017). We initially aimed to subdivide and analyze housing types, but the sample size of older adults living alone (n = 1,355) was insufficient, and the caliper matching process did not provide enough matched cases to ensure unbiased findings.
Second, although this study’s target population comprised older adults, the unavailability of related information in the HWS dataset prevented us from considering essential aging-related variables, such as the index of activities of daily living. Moreover, since the 2019 HWS dataset only provided cross-sectional information about participants, this study could not apply specific inclusion and exclusion criteria, such as the status of pre-existing mental health problems (e.g., dementia) or negative life events (e.g., bereavement), which can directly correlate with self-assessed depression in older adults (Osborn et al., 2003; Ross-Adelman et al., 2025). The omission of these criteria might have led to the over- or underestimation of the causal relationship between living alone and self-assessed depression in this study. This limitation could be addressed by considering the HWS dataset in combination with additional data sources containing aging-related information, such as the Survey of Older Persons in Thailand.
Third, since there is limited evidence regarding the link between living alone and depression among older adults in Thailand, this study was conducted for exploratory purposes. Thus, we made limited efforts to perform in-depth statistical analysis, such as subgroup analysis with potential interaction effects (e.g., gender differences by income level). Future efforts to address this limitation would allow for a more comprehensive assessment of this subject.
Fourth, this study assessed self-assessed depression using a single-item question. While such single-item measures of self-reported mental health are extensively accepted as a reasonable substitute for more valid multiple-item measures, such as the “Geriatric Depression Scale” (Ahmad et al., 2014; McKenzie & Marks, 1999), future research should utilize multiple-item measurements to validate our findings. To our knowledge, such multiple-item measures are not available in recent nationwide datasets in Thailand, which the National Statistical Office should consider.
Fifth, since the 2019 HWS dataset in this study is outdated, this study’s findings may not reflect present patterns of self-assessed depression and its influencing factors. Although regular nationwide survey datasets are generally consistent across years, future research should investigate the reliability and validity of this study’s findings using a more recent HWS dataset.
Sixth, although the study analysis was performed using a nationally representative sample (HWS dataset), the application of survey weights warrants discussion. Specifically, we initially performed two separate POMs using the entire sample—one with survey weights and one without. Both analysis results were nearly identical. However, in the weighted POM, the weight scores exhibited an extremely large range (from a minimum of 25.97 to a maximum of 19,786.33). This large range substantially increased the total variance of the regression estimates and accordingly caused a significant violation of model performance tests, specifically the parallel lines assumption and the lack-of-fit test. Due to this issue, and the fact that the HWS dataset already constitutes a nationally representative sample, we ultimately decided to proceed with the unweighted POM. Future study may consider other analytical methods, such as multinomial logistic regression analysis, to address this issue, thereby ensuring the representativeness of estimates.
Lastly, our cross-sectional analysis did not allow us to evaluate the causality of sociodemographic factors’ relationships with depression. Future studies should adopt longitudinal approaches to assess long-term trends in depression and its associated factors among older adults.
Conclusion
Concerning the first research objective, the findings indicate that older adults living alone tend to have higher self-assessed depression than those living with family. In other words, living alone could independently elevate the risk of depression among older Thais, irrespective of sociodemographic factors. Furthermore, this finding indicates that the observed positive relationship between living alone and depression is unlikely to be driven solely by sociodemographic confounders. Therefore, a social support network that can supplement the traditional family role in providing support for older adults to overcome daily life challenges should be established to address the increased risk of depression among older adults living alone. This could be done, for example, by integrating social workers into existing community health volunteer programs.
Regarding the second research objective, the findings show substantial sociodemographic disparities in self-assessed depression. Specifically, older adults who were low-income, female, unemployed, chronically ill, or living in rural regions tended to have higher levels of self-assessed depression than their counterparts. Social stress theory and relevant empirical research propose that reinforcing external resources (e.g., social participation programs) can effectively reduce sociodemographic disparities associated with poor mental health. Accordingly, the government should continue developing community-based social participation programs for older adults. Moreover, the overall accessibility and participation rates of these programs should be monitored.
Although this study offers valuable insights, future studies are needed to validate our findings. Specifically, multiple-item measurements of depression (e.g., the Geriatric Depression Scale), various types of co-residence (e.g., living with a partner only, living with children only, or both), and longitudinal designs should be considered to gain a more comprehensive understanding of depression and its related factors—especially the role of living alone—among older adults. Addressing the mental health issues of older adults, especially those who are living alone and socially disadvantaged, is crucial for enhancing healthy aging and the quality of life of this population.
Footnotes
Acknowledgements
We thank Dr. Punchada Sirivunnabood, Dean of the Faculty of Social Sciences and Humanities at Mahidol University, for her research assistance, and the National Statistical Office of Thailand for the data they offered.
Ethical Considerations
This study was granted “Exemptions from IRB review” by the Office of The Committee for Research Ethics of the Faculty of Social Sciences and Humanities, Mahidol University (Certificate of Exemption No.: 2025/003.2002).
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
The 2019 HWS dataset that supports the findings of this study is available from the National Statistical Office of Thailand. However, restrictions apply regarding the availability of the dataset, which was used under license for the present study and, thus, is not publicly available. However, the dataset is available from the authors upon reasonable request and with the permission of the National Statistical Office of Thailand.