Co-residence with children as a mediator between widowhood and loneliness in older adults

Abstract

Background

Loneliness in older adults, exacerbated by widowhood, is a significant public health concern. While widowhood can lead to changes in living arrangements, its impact on loneliness may vary across cultural contexts. In Western societies, widowhood often results in older adults living alone, which can intensify feelings of loneliness. However, in China, the cultural norm of filial piety and multigenerational households may lead to different outcomes. As few studies have explored this connection over time, this research seeks to bridge this gap using data from older Chinese adults.

Methods

Using 16 years of data from the Chinese Longitudinal Healthy Longevity Survey, which covers 21,986 individuals aged 65–104 years, we conducted causal mediation analysis to test if changes in living arrangements (i.e., living alone versus with children) serve as a mediator between widowhood and loneliness. The potential variation in this mediation effect by gender and age was also evaluated.

Results

Spousal loss was associated with an increase in loneliness. However, living with adult children post-loss reduced this emotional strain compared to living alone. The mediating influence of living arrangements was notably stronger for women than men and intensified with age in long term. In the short term, the mediating impact of living arrangements is significantly greater, particularly for older adults under 80 years old.

Conclusion

Alterations in living arrangements play a pivotal role in mediating the effects of widowhood on loneliness among China’s older adults. Encouraging co-residence with adult children post-spousal loss, especially for older women and the eldest age groups, might mitigate social isolation. These insights both deepen our theoretical understanding and suggest interventions to enhance the well-being of widowed older adults.

Introduction

Loneliness is increasingly recognized as a major public health issue, particularly among older adults. Research has demonstrated that loneliness is associated with increased mortality risk in older adults, with a 9% increase in risk overall [1]. A recent meta-analysis revealed that loneliness and social isolation significantly increase the risk for premature mortality, showing an effect size comparable to smoking and obesity [2]. Loneliness is also linked to grief and depressive symptoms in older adults [3]. These findings underscore the importance of addressing loneliness in older populations to enhance well-being and longevity. Consequently, investigating risk factors for loneliness in older adults is critically important and can inform potential interventions to improve older adults’ well-being. In particular, widowhood has been identified as a major risk factor for loneliness, as a substantial body of research demonstrates that bereaved spouses experience a pronounced increase in loneliness compared to their still-married peers [4,5,6,7,8].

Losing a spouse is a major life event that can increase feelings of loneliness. The stress mechanism theory posits that significant life events, such as the loss of a spouse, act as stressors that can have profound impacts on an individual’s mental and emotional well-being [9, 10]. Widowhood is considered a critical stressor due to the loss of social, emotional, and instrumental support from the partner, as well as daily interaction and companionship [11]. Women tend to exhibit higher levels of resilience in coping with losing her spouse [12, 13]. The stress mechanism theory highlights that the extent of this impact is influenced by an individual’s coping mechanisms, the availability of social support, and changes in living arrangements post-widowhood [14]. Increased support from children, relatives, and social participation after widowhood may help mitigate the adverse mental health effects of bereavement [15]. Nevertheless, the availability and efficacy of social support can vary based on factors such as living arrangements. There is a need to explore living arrangements as a potential mediating mechanism linking widowhood to loneliness in older adults. The transition to widowhood often necessitates changes in living arrangements, which can profoundly affect psychological well-being in older adults [16,17,18].

Losing a spouse frequently necessitates changes in living arrangements, as widowed individuals must adapt to living without their marital partner. The convoy model of social relations suggests that individuals are surrounded by a network of close relationships that provide support throughout life. In the context of widowhood, this model helps explain how the loss of a spouse can disrupt an individual’s social convoy, potentially leading to changes in living arrangements. According to the convoy model, when a key member of the convoy (the spouse) is lost, individuals may seek to restructure their living arrangements to maintain or rebuild their support network. This could involve moving in with adult children or other family members who are part of their social convoy. The quality and stability of these relationships within the convoy are crucial in buffering against the negative effects of stressors [19, 20]. The impact of widowhood on older adults’ living situations is complex and varies across contexts. Population data indicates most married older adults reside with a spouse, while the majority of widowed live alone across Western nations [21, 22]. In China, widowhood’s effect on living arrangements is especially pronounced due to the country’s social and economic context. Research has shown older Chinese widowhood often increases the likelihood of coresidential living, especially for women [1, 18].

Living arrangements can significantly influence access to social support, and older adults’ choice of living situation after widowhood—either alone or with others—can have important implications for loneliness. The socioemotional selectivity theory posits that as people age, they prioritize emotionally meaningful relationships and interactions [23]. Transitioning to solo living is difficult for many widowed elders and may reduce opportunities for social interaction, potentially heightening feelings of isolation and loneliness [24,25,26,27]. Conversely, if older adults choose to co-reside with family, especially adult children, after widowhood, they may have more opportunities to maintain meaningful social relationships and participation. Consequently, living with others may provide increased access to social support, potentially mitigating widowhood’s effect on loneliness. Co-residence with children provides more opportunities for intergenerational interaction, which may satisfy older adults’ need for meaningful social relationships. This aligns with the emphasis on emotional experiences in socioemotional selectivity theory [28]. For example, one study found older adults with children living nearby were less lonely than those without proximate children [29]. Similarly, another study revealed adult children can buffer against feelings of loneliness for older individuals by providing emotional and instrumental support [30].

Past studies indicate the effect of spousal loss on loneliness may vary based on factors such as gender and age [31, 32]. However, few studies have formally examined the moderating effects of gender and age on living arrangements’ mediating role. Some findings indirectly suggest moderating effects of gender and age. The stress mechanism theory helps explain why widowed men are more likely to experience loneliness due to losing their spouse, a consistent finding across studies [33]. This heightened loneliness risk is especially pronounced for widowed men living alone, indicating widowers often struggle adjusting to solo living. However, men’s distant relationships with children throughout their lives may limit the impact of co-residence. This can be understood through the lens of the convoy model of social relations, which emphasizes the importance of lifelong social networks [19]. Moreover, men frequently rely more heavily on their spouses as their primary source of emotional support. The loss of a spouse, therefore, creates a significant void that proves challenging to fill, even in the presence of co-residing children. Women, benefiting from their broader social networks, often have multiple sources of emotional support. This diversity in emotional resources makes the transition to co-residing with children more impactful in mitigating loneliness for women [34, 35].

In addition to gender, age may also moderate the mediating effect of living arrangements on the relationship between widowhood and loneliness. On one hand, widowhood’s impact on co-residing with children appears to vary by the bereaved parent’s age. Several studies found younger widowed parents are more likely to move in with children than those widowed at older ages [36]. One study revealed being widowed before age 75 had a greater effect on co-residence compared to being widowed after 75 [37]. This could be interpreted through the stress mechanism theory, as younger widowed individuals might experience more stress adjusting to their new situation and thus seek support through co-residence [37]. On the other hand, living arrangements’ effect on loneliness may differ among widowed older adults of varying ages. Research suggests living arrangements may have a more pronounced impact on loneliness for adults aged 80 and above [38]. This aligns with the socioemotional selectivity theory, which posits that as people perceive time as more limited, they focus more on emotionally satisfying relationships, potentially making co-residence with children more impactful for the oldest old in combating loneliness [39].

Fig. 1

Conceptual Model of the Relationship between Widowhood, Living Arrangements, and Loneliness

Figure 1 presents our conceptual model of the relationship between widowhood, living arrangements, and loneliness. The model proposes that widowhood directly influences loneliness (H1) and that this relationship is mediated by changes in living arrangements (H2). Additionally, the model suggests that the mediating effect of living arrangements is moderated by gender (H3) and age (H4). This model is informed by several theoretical frameworks. The stress mechanism theory helps explain the direct link between widowhood and loneliness, as well as how this might differ by gender. The socioemotional selectivity theory informs our understanding of how living arrangements might impact loneliness, particularly in older age groups. The convoy model of social relations provides context for understanding how widowhood might lead to changes in living arrangements.

Our study seeks to bridge the existing knowledge gaps by conducting a longitudinal moderated mediation analysis using a comprehensive dataset representative of Chinese older individuals. By leveraging 16 years of panel data from the Chinese Longitudinal Healthy Longevity Survey, we aim to scrutinize how changes in living arrangements mediate the correlation between widowhood and loneliness over time. This method ensures a meticulous mediation test while confirming the temporal sequence of predictors, mediators, and outcomes specific to the Chinese demographic.

This study provides unique insights into the causal relationships between widowhood, living arrangements, and loneliness through a 16-year longitudinal analysis, illuminating how these factors interact within the Chinese cultural context. Focusing on the older population in China is significant due to the country’s rapidly aging population and rising widowhood rates [40]. Moreover, distinctive cultural factors in China, such as filial piety and traditional family structures, profoundly influence living arrangements and social support for older adults [41, 42]. The research finds that co-residing with adult children can alleviate loneliness among widowed older adults [15], suggesting these cultural practices may have protective effects on the emotional well-being of older adults. However, as society evolves and traditional values gradually weaken, these protective effects may be changing. Therefore, this study not only reveals the complex relationships between widowhood, living arrangements, and loneliness but also offers valuable insights into how cultural factors shape the well-being of older adults in China’s rapidly changing society. Additionally, examining whether changes in cohabitation status after spousal loss differentially impact loneliness for men versus women and young-old versus oldest-old adults would delineate vulnerability factors and critical intervention points in China.To our knowledge, this will be the inaugural study delineating whether the progression from spousal loss, through altered living situations, to amplified loneliness is more pronounced for certain subgroups of China’s widowed senior population. Our findings will highlight potential intervention points and identify segments most susceptible to widowhood-induced social isolation, thus guiding public health strategies.

Methods

Data

This study draws from the Chinese Longitudinal Healthy Longevity Survey (CLHLS). Initiated in 1998 by researchers from Peking University and Duke University, the CLHLS is a continuous survey targeting Chinese older adults aged 65 and above [43]. The survey encompasses randomly selected half of the counties/cities in 22 out of China’s 31 provinces, covering an estimated 85% of the nation’s population. Follow-up surveys were administered in 2000, 2002, 2005, 2008–2009, 2011–2012, 2014, and 2018. Our analysis specifically employs data from the 2002 to 2018 waves, integrating both the original samples from 2002 and samples added in the ensuing follow-ups. In total, data from 21,986 respondents, aged between 65 and 104, were utilized, culminating in 42,389 observations. For participants who passed away between waves, mortality data were obtained by interviewing a family member. The comprehensive nature of this longitudinal dataset, complemented by its mortality follow-up, offers a distinctive platform to meticulously examine the causal and moderated mediation effects of widowhood on loneliness using sophisticated statistical methods.

In addition, this study further restricts the sample to individuals who had a spouse and did not live with children at the baseline, with 11,031 observations from 4,398 respondents. This allows us to more accurately compare the short-term and long-term effects of widowhood on living arrangements and loneliness. By focusing on this subset of the sample, we can better distinguish between those who transitioned to living with children after experiencing spousal loss and those who were already living with children before becoming widowed. In our study, such cases who were widowed multiple times or who remarried during the study period are very rare, as the proportion of older adults experiencing multiple marital transitions is extremely low. To ensure the consistency of our treatment status, we have excluded individuals who remarried during the follow-up period from our analysis. By doing so, we aim to maintain a clear and consistent treatment status for all participants throughout the study period.

Measures

Table 1 delineates the measurement of variables and their descriptive statistics. The study’s dependent variable is self-rated loneliness, gauged using a single question: “How often do you feel lonely?” Respondents chose from five options: never, rarely, sometimes, often, and always. For the purposes of causal inference analysis, this self-rated loneliness was recoded into a binary variable: higher levels of loneliness, encompassing responses “always”, “often”, and “sometimes”, were coded as 1; whereas lower levels, comprising “rarely” and “never”, were coded as 0. Such a bifurcation is standard and accepted in gerontological loneliness research, aligning with the everyday understanding of the term [44].

Our primary exposure variable is widowhood, defined via a time-varying binary indicator reflecting whether a respondent lost their spouse. The mediator in focus is living arrangement, bifurcated as living with children or living alone. This variable was updated at each survey wave. In the baseline sample, a mere 2.99% of respondents resided in nursing homes, with a mere 0.88% cohabitating with others. These samples were excluded from the study as it focuses on exploring the influence of residing with children in a community setting.

The analysis accounted for several covariates: age (ranging from 65 to 104 years), gender (with males coded as 0 and females as 1), place of residence (urban coded as 1, town as 2, and rural as 3), years of education, pension status (yes coded as 1; no as 0), social activities and health status. Health status was assessed using multiple indicators: cognitive function, basic activities of daily living (ADLs), instrumental ADLs (IADLs), and chronic diseases. An aggregated health index was then constructed from these indicators using the Item Response Theory model [45]. To streamline model estimation and convergence, this health status was segmented into four tiers based on the index: very bad, bad, good, and excellent. Social activities is assessed using various indicators: household activities, card games or mahjong, social engagements, and travel. Based on these indicators, a comprehensive index is derived through item response theory. To facilitate model estimation, the social activities index is categorized into four levels: never, rarely, generally, and frequently. Table 1 presents the baseline sample characteristics and includes an examination of the sample distribution by marital status.

Table 1 Demographic characteristics of older adults by Marital Status at Baseline and during Follow-Up

Analysis

Causal mediation analysis was conducted based on the counterfactual framework developed by Imai et al. [46]. Let $\:{T}_{i}$ denote a binary treatment (widowhood), $\:{M}_{i}\left(t\right)$ the potential mediator value (living arrangement) under treatment $\:t$, and $\:{Y}_{i}\left(t,\:m\right)$ the potential outcome (loneliness) under treatment $\:t$ and mediator value $\:m$. The causal mediation effect for unit $\:i$ is defined as:

$$\:{\delta\:}_{i}\left(t\right)={Y}_{i}\left(t,\:{M}_{i}\left(1\right)\right)-{Y}_{i}\left(t,{M}_{i}\left(0\right)\right)$$

This captures the indirect effect of the treatment on the outcome through the mediator. The average causal mediation effect (ACME) is $\:\stackrel{-}{\delta\:}\left(t\right)=E\left[{\delta\:}_{i}\left(t\right)\right]$. Key assumptions like sequential ignorability clarify the conditions under which $\:\stackrel{-}{\delta\:}\left(t\right)\:$has a causal interpretation. This counterfactual framework provides a general definition of causal mediation effects and illuminates identification assumptions. It also allows flexible modeling strategies that permit nonlinear relationships and nonparametric models.

Analyses were conducted using the mediation package in R, as proposed by Tingley et al. [47]. This package employs parametric regression models to calculate the Average Causal Mediation Effects (ACME). For both the binary mediator and the outcome, multilevel logit models with random intercepts were used. Additionally, the results presented in Tables 2 and 3 are derived from these multilevel logit models, which account for the hierarchical structure of the data by allowing for variation at the group level. Survey weights were applied in all analyses to ensure that the results are representative of the broader population. These weights also adjust for the overrepresentation of centenarian samples, thereby mitigating any potential bias in the estimation results due to the unique characteristics of this age group. The ACME estimations adjusted for potential confounders, including age, gender, place of residence, years of education, pension status, and health status, and social participation. Subsequent to determining mediated effects, we assessed moderated mediation by introducing interaction terms between the exposure, mediator variable, and both gender and age categories. This phase of the analysis aimed to discern whether the mediating influence of living arrangements exhibited variability contingent upon the levels of these moderators. Specifically, for gender, the analysis sought to ascertain if males and females exhibited distinct mediation effects. In terms of age, the goal was to determine whether the mediation effect manifested differently between older and younger segments of our sample. A notable interaction would signal variations in mediation effects across these defined subgroups. To bolster the robustness of our findings, bootstrap resampling techniques—with a total of 1,000 iterations—were applied to generate confidence intervals surrounding both the ACME and the index of moderated mediation.

Results

To discern the causal mediation effect, we began by evaluating the mediation models delineated in Table 2. A review of the Akaike Information Criterion (AIC) across the models highlighted disparities in data fit. Notably, among the sample comprising married and widowed individuals at baseline, Model 3, with an AIC of 41390.063, was identified as the most fitting model, superseding Model 2 (AIC = 41460.018) and Model 1 (AIC = 41473.409). Conversely, for the sample consisting of married older adults who did not live with their children at the baseline, Model 1 with an AIC of 9489.007, was found to be the most appropriate model. Across all three models, the coefficient for “Widowhood” was both positive and statistically significant (p < 0.001). When controlling for other variables, our findings suggest that widowed individuals exhibit a greater propensity to co-reside with their children compared to their married counterparts.

The interaction term “Widowhood×Female” was statistically significant in both Model 2 and Model 3 among the sample comprising married and widowed individuals at baseline. The affirmative coefficients (0.499 and 0.566, respectively) underscore a heightened effect of widowhood on living with children among females in contrast to males. To be precise, female widows are more inclined to live with their offspring compared to male widows, a trend particularly accentuated in Model 3. Furthermore, in Model 3 among the sample comprising married and widowed individuals at baseline, the interaction term “Widowhood×Age” bore a positive coefficient of 0.716 (p < 0.001), signifying that the impact of widowhood on co-residing with children intensifies with age. This implies that as widowed individuals age, they are increasingly inclined to share a living space with their children. In contrast, for the sample consisting of married older adults who did not live with their children at the baseline, both interaction terms “Widowhood ×Female” and “Widowhood×Age” are statistically insignificant. This finding suggests that the effect of widowhood on living arrangements during the follow-up period is not moderated by gender and age.

Table 2 Mediation Model Estimation for Predicting Co-residence with children among widowed individuals for different samples

We subsequently estimated the outcome models, with results presented in Table 3. Using a multilevel logit model, we assessed the influence of widowhood, co-residence with children, their respective interactions, and additional control variables on loneliness. In both the sample of individuals who were married or widowed at baseline and the sample of married older adults not living with their children at baseline, Model 3 demonstrated superior data fit, evidenced by the lowest AIC value compared to Model 1 and Model 2. A consistent finding across all models was the pronounced loneliness experienced by widowed individuals, denoted by the positive and significant coefficients for “Widowhood” (p < 0.001), thereby supporting our H1 hypothesis. Conversely, co-residence with children correlated with diminished feelings of loneliness, as evidenced by the negative and statistically significant coefficients for “Co-residence with children children” (p < 0.001).

Model 3’s interaction term “Widowhood×Female” among the sample of individuals who were married or widowed at baseline was significant (-0.229, p < 0.05), suggesting a differential impact of widowhood on loneliness between genders. Precisely, widowhood’s influence on loneliness appears less marked for females than males. However, the interaction term “Widowhood × Female” in Model 3 was insignificant among the sample of married older adults not living with their children at baseline. This suggests that the effect of widowhood on loneliness during the follow-up period is not moderated by gender. In both the sample of individuals who were married or widowed at baseline and the sample of married older adults not living with their children at baseline, the “Widowhood×Age” interaction was significant and negative (-0.244, p < 0.001) in Model 3, hinting at a diminishing effect of widowhood on loneliness as age progresses. Conversely, the “Co-residence with children×Female” interaction wasn’t significant in the sample of individuals who were married or widowed at baseline, but it is significant in the sample of married older adults not living with their children at baseline. The “Co-residence with children×Age” interaction in the sample of married older adults not living with their children at baseline was significant (-0.064, p < 0.01). Across all models, the consistent negative and significant interaction between “Widowhood” and “Co-residence with children” indicates a marked difference in the effect of living arrangement between the widowed and married older populations.

Table 3 Outcome models estimating the effects of Widowhood and co-residence on loneliness for different samples

For both the mediation and outcome models in the sample of individuals who were married or widowed at baseline, Model 3 offers a superior fit. As such, we utilized Model 3 for our causal mediation effect estimation. In contrast, for the sample of married older adults not living with their children at baseline, we calculated the causal mediation effects using Model 1 for the mediation model and Model 3 for the outcome model. As detailed in Table 4, the magnitude of the mediation effect and its portion of the overall effect are showcased. This analysis aimed to illuminate the role of living arrangements in mediating the connection between widowhood and perceived loneliness.The Average Causal Mediation Effect (ACME) quantifies how living arrangements influence the relationship between widowhood and loneliness. The Average Direct Effect (ADE) sheds light on the inherent effect of widowhood on loneliness without accounting for mediators. The proportion mediated (average) indicates the share of the total effect that is accounted for by the mediating variable.

In the first sample, consisting of individuals who were married or widowed at baseline, the total effect of widowhood on loneliness is estimated at 0.248 (95% CI: 0.234 to 0.262). The ACME in this group is -0.023 (95% CI: -0.026 to -0.020), and the ADE is 0.271 (95% CI: 0.257 to 0.286). The proportion mediated is -0.092 (95% CI: -0.105 to -0.080), indicating that approximately 9.2% of the total effect is mediated by living arrangements. In the sample, consisting of married older adults not living with their children at baseline, the total effect of widowhood on loneliness is higher, estimated at 0.330 (95% CI: 0.293 to 0.362). The ACME in this group is -0.034 (95% CI: -0.043 to -0.026), and the ADE is 0.364 (95% CI: 0.327 to 0.395). The proportion mediated in this sample is -0.104 (95% CI: -0.134 to -0.077), suggesting that about 10.4% of the total effect is mediated by living arrangements. Thus, addressing our research hypotheses, our findings provide strong support for H2 in both samples.

Comparing the results across the two samples reveals notable differences. In the sample of individuals who were married or widowed at baseline, the mediation effect (ACME) is slightly smaller in absolute terms compared to the sample of married older adults not living with their children at baseline. The proportion mediated is significantly lower in the first sample (9.2%) compared to the second sample (10.4%), with a formal statistical test confirming that this difference is statistically significant (p < 2.2e-16). These differences suggest that the role of living arrangements as a mediator between widowhood and loneliness is more pronounced among older adults who were not living with their children at baseline. This highlights the critical impact of co-residence with children on mitigating loneliness in the aftermath of widowhood, particularly for those who initially did not live with their children.

Table 4 Total effect, average causal mediation effect (ACME), average direct effect (ADE), and proportion mediated for different samples

In our quest to unravel the complex interplay between widowhood, living arrangements, and the ensuing feelings of loneliness, it’s crucial to account for possible gender-based variations in mediation effects. Table 5 delineates the observed mediation effects for both male and female older cohorts. For the sample of individuals who were married or widowed at baseline, The ACME for males is -0.017 (95% CI: -0.021 to -0.014), the proportion mediated for males is -0.064 (95% CI: -0.079 to -0.050), suggesting that approximately 6.4% of the total effect is mediated by living arrangements. In contrast, for females in the same sample, the total effect is slightly lower at 0.228 (95% CI: 0.212 to 0.244). The ACME for females is -0.014 (95% CI: -0.027 to -0.023), and the proportion mediated for females is -0.118 (95% CI: -0.140 to -0.098), indicating a higher mediation effect compared to males, with 11.8% of the total effect being mediated by living arrangements. In the sample of married older adults not living with their children at baseline, the total effect for males is estimated at 0.334 (95% CI: 0.285 to 0.383). The ACME for males in this group is -0.021 (95% CI: -0.033 to -0.010), and the proportion mediated for males is -0.064 (95% CI: -0.102 to -0.031). For females, the ACME for females is -0.048 (95% CI: -0.058 to -0.038), and the proportion mediated for females is -0.148 (95% CI: -0.186 to -0.115), indicating a substantial mediation effect, with 14.8% of the total effect being mediated by living arrangements. Thus, addressing our research hypotheses, our findings provide strong support for H3 in both samples.

Comparing the results across different samples and genders reveals notable differences. In both samples, the total effect of widowhood on loneliness is generally higher for males compared to females. Additionally, the mediation effect (ACME) and the proportion mediated are more pronounced for females, suggesting that living arrangements play a more significant role in mediating the relationship between widowhood and loneliness for women. This highlights the critical impact of living arrangements on mitigating loneliness, particularly for widowed women who may rely more on social and familial support. In terms of differences in total effects across different samples, the proportion of the mediation effect relative to the total effect for male older adults is similar in both samples. However, for female older adults, the proportion of the mediation effect is notably larger in the sample of married older adults not living with their children at baseline. This suggests that the role of living arrangements in mediating loneliness is more pronounced for women in this specific sample, indicating that living with children may be a more critical factor in reducing loneliness for widowed women who were not initially co-residing with their children.

Table 5 Total effect, average causal mediation effect (ACME), average direct effect (ADE), and proportion mediated by gender for different samples

In our exploration to decipher the interplay between widowhood, living arrangements, and feelings of loneliness, the role of age-specific variations in mediating effects warrants attention. A detailed analysis of Table 6 reveals an unmistakable trajectory in both the Average Causal Mediation Effect (ACME) and the proportion of the effect mediated by living arrangements with increasing age. In the sample of individuals who were married or widowed at baseline, the total effect of widowhood on loneliness and the proportion of the mediation effect exhibits notable variations across different age groups. At age 65, living arrangements mediate an average of 5.6% of the total effect. This mediation effect consistently rises: 7.5% for those aged 70, 12.8% for 80-year-olds, surging to 22.1% for nonagenarians, and a remarkable 45.2% for centenarians. This pattern underscores that as individuals advance in age, the total effect of widowhood on loneliness generally decreases. However, the proportion of the mediation effect through living arrangements increases with age, providing strong support for hypothesis H4. This finding indicates that the role of living arrangements in mitigating loneliness becomes more significant for older age groups. This trend highlights the growing importance of social and familial support structures in reducing loneliness among the oldest old, particularly those who experience widowhood.

Table 6 Total effect, average causal mediation effect (ACME), average direct effect (ADE), and proportion mediated by age for different samples

As shown in Table 6, as age increases, the total effect of widowhood on loneliness generally decreases among married older adults not living with their children at baseline. By the time people are over 90 years old, the total effect is no longer significant, which indicates that being widowed has no significant impact on the loneliness of the elderly during the follow-up period. The proportion of the mediation effect through living arrangements varies, showing a slight increase until age 80, but decreases and exhibits insignificant effect at older ages (90 and 100). This trend highlights the changing dynamics of social and familial support structures in mitigating loneliness among the oldest old, particularly those who experience widowhood and were not initially co-residing with their children. Other mechanisms could also be at play in explaining these discrepancies. For instance, the initial living arrangement (living with or without children) might interact with age and widowhood in complex ways to influence loneliness. Those who were not living with their children at baseline might have developed different coping mechanisms or social networks that become more relevant in very old age. Additionally, survivorship bias could be a factor, where those reaching very advanced ages might have unique characteristics that influence how widowhood and living arrangements affect their experience of loneliness. Further research into these potential mechanisms could provide valuable insights into the complex dynamics of loneliness among the oldest old.

Discussion

The present research delved into the intricate dynamics of widowhood, living arrangements, and loneliness among the older adults, concentrating on the Chinese milieu. In consonance with existing scholarship, our results underscore widowhood’s significant bearing on heightened loneliness [4, 8]. This investigation further solidifies the understanding of widowhood as a pivotal factor influencing living arrangements and delineates the mediating influence of these arrangements on widowhood-induced loneliness.

Through our longitudinal mediation assessment, we discerned that post-bereavement shifts in living circumstances play a crucial mediating role in the widowhood-loneliness nexus over time. Explicitly, the journey from cohabitation with a spouse to solitary living post-widowhood exacerbated loneliness. However, a mitigation in this surge was observed among widowed seniors who transitioned to living with their adult progeny, as opposed to those dwelling alone. Moreover, this study found that older adults who initially lived with only their spouse experienced a more pronounced reduction in loneliness when they moved in with their children, rather than living alone, after becoming widowed during the follow-up period. This finding underscores the critical role of social and familial support in alleviating the negative mental health impacts of widowhood. This observation resonates with earlier studies which posited that sustained familial co-residence and support can serve as bulwarks against the daunting void of social seclusion post-spousal loss [15]. The support from children after spousal loss can help alleviate adverse mental health effects like loneliness. This buttresses the argument that sharing a living space with offspring can counteract the detrimental psychosocial repercussions of spousal loss, a sentiment mirrored in diverse cultures like China [48], Japan [49], and India [50]. Sociocultural imperatives such as filial responsibilities influence the choices of the bereaved older adults regarding cohabitation with relatives and the ensuing implications for their emotional health [48]. Our study, set against the backdrop of China’s familial ethos, accentuates the protective role of intergenerational cohabitation in insulating against the isolating aftermath of widowhood. However, the changing societal dynamics, such as urbanization and increased internal migration, have led to an erosion of traditional values of filial obligations [51, 52]. This shift has implications for the emotional well-being of the bereaved older adults, as co-residence may become a source of daily friction, reducing freedom and increasing stress levels [35].

Our research sheds light on factors that heighten vulnerability by analyzing conditional indirect effects. We observed a more accentuated mediating impact of living arrangements for women compared to men. This is congruent with existing literature indicating that widowed men tend to live less frequently with family members and face greater challenges adapting to solitary living [53]. This gender disparity can be attributed to the distinct ramifications of social networks of men versus women. Widowed men typically maintain more distant relationships with their children and relatives compared to widowed women [54]. This disparity often results in less effective social support when co-residing with children. In contrast, women tend to nurture more intimate connections with their children throughout their lives. Consequently, upon widowhood and subsequent co-residence with children, senior women often find it easier to access extensive emotional support and experience a stronger sense of belonging.

Additionally, our data reveal an intensifying mediating influence of living arrangements as age progresses. For individuals aged 65 who are married or widowed at baseline, the mediating impact of co-residing with children accounts for a mere 5.6% of the total effect of widowhood on loneliness. Remarkably, this figure rises to over 22.1% for individuals aged 90. Such findings dovetail with studies indicating that older widowed individuals are particularly inclined to cohabit with their offspring, especially in societies with strong cultural norms around filial piety, such as China [35, 55], and the grave implications of social isolation for the oldest segment of the population. As individuals age, their social networks tend to shrink due to factors such as retirement, the formation of smaller offspring nuclear families, the progression towards widowhood, worsening health conditions, limitations on physical mobility, and natural selection of mortality among peers [56, 57].

With advancing age, there’s a palpable shift towards valuing relationships that offer emotional depth and a sense of belonging. The dissolution of the intimate spouse-child bond in the twilight years can profoundly disrupt socioemotional equilibrium, amplifying the allure of multigenerational living as a panacea for loneliness. Cultural ethos, encapsulated in the principle of filial piety, accentuates this trend. As elders age and grapple with the void left by a departed spouse, societal norms expect and encourage heightened caregiving roles for their children [12]. Furthermore, the escalating physical and instrumental needs characteristic of the “oldest-old” segment underscore the salience of familial co-residence over solitary existence [13]. Spanning the physical, social, and emotional realms, our results champion the pivotal role of multigenerational co-residence in alleviating loneliness among the older adults, especially in the aftermath of spousal bereavement in a society steeped in filial traditions like China. Advocating for policies that bolster family caregiving models could serve as a linchpin in enhancing the well-being of the eldest, widowed segment of the population.

However, this study indicates that for younger elderly individuals (under 80 years old) who live with their spouse but not with their children at baseline, the reduction in loneliness after becoming widowed and moving to live with their children during the follow-up period was more pronounced. Conversely, for older elderly individuals (over 80 years old), the reduction in loneliness under similar circumstances was diminished or even negligible. This suggests that for younger elderly individuals, the reduction in loneliness due to co-residence with their children is more pronounced, indicating that the short-term support provided by living together is significant. In contrast, for older elderly individuals, the reduction in loneliness is attributed to long-term accumulation, indicating that long-term support from co-residence has a greater impact. This further demonstrates that the mitigating effect of living with children on loneliness results from both short-term benefits and long-term accumulation.

Our research offers several significant contributions to existing literature. Firstly, it presents one of the pioneering rigorous longitudinal examinations of the mediation between widowhood, living arrangements, and loneliness over a span of 16 years. By employing counterfactual causal inference methodologies, we have clarified both the temporal sequence and the causal dynamics. Specifically: 1,We established a clear temporal order: widowhood precedes changes in living arrangements, which then affect loneliness levels caused by widowhood. 2,Our longitudinal approach allowed us to track these changes, providing insight into how these relationships evolve. 3,The counterfactual methods enabled us to estimate causal effects more accurately by comparing observed outcomes to hypothetical scenarios. 4,We quantified both the direct effect of widowhood on loneliness and the indirect effect mediated through living arrangements. This approach not only addresses the constraints of cross-sectional analyses but also sets the chronological relationships among predictors, mediators, and outcomes. Secondly, our study delves into the relatively unexplored Chinese cultural milieu. Considering China’s historical propensity for co-residence and the evolving household dynamics, our findings underscore the enduring significance of familial support in determining the well-being of older adults in the face of societal transitions in China. Thirdly, our exploration of conditional indirect effects sheds light on the variations in the mediating role of living arrangements according to gender and age. By pinpointing the most susceptible demographics, we can fine-tune interventions to cater to those at heightened risk of social isolation following the loss of a spouse. Collectively, our meticulous analysis provides a holistic understanding of vulnerability across diverse population segments.

While this research offers significant insights, there are acknowledged limitations. Firstly, our study relied on a single-item, dichotomized measure of loneliness. Though this streamlined approach aids in analysis, employing multi-item scales to gauge the depth of loneliness might offer a more nuanced understanding. Secondly, our sample was confined to individuals aged 65 and above, potentially limiting the generalizability to younger widowed populations. Investigating living arrangements and loneliness post-bereavement in early life would be a valuable extension. Thirdly, our focus was narrowly set on co-residing with children as the primary mediator, leaving other post-widowhood living scenarios, such as independent senior housing or institutional settings, unexplored. A broader comparative study encompassing various living arrangements post-widowhood would be instructive. Lastly, incorporating quantitative data on co-residence preferences, relationship quality, and expectations would further enrich the narrative. Moreover, leveraging qualitative insights from detailed interviews or focus groups might shed light on individual motivations, personal interpretations, and familial dynamics that drive the observed mediation effects. A mixed-methods strategy, amalgamating our mediation models with qualitative findings, would provide a more holistic understanding. In sum, this research serves as a pivotal step in deciphering the pathways connecting widowhood and intensified loneliness. By addressing the aforementioned gaps in subsequent studies, a more comprehensive comprehension of the intricate dynamics faced by grieving older adults can be attained. This foundational work sets the stage for both future research and policy initiatives tailored to this demographic’s needs.

Conclusion

In summation, our study presents valuable longitudinal insights into the pronounced effects of living arrangement transitions on loneliness exacerbated by widowhood among China’s older adults. The findings underscore that the bereavement of a spouse augments feelings of loneliness over time. Yet, the shift to co-residing with adult children post-loss serves as a buffer against these adverse effects in contrast to solitary living. Notably, through conditional process analyses, we discerned that this mediating role of living arrangements intensifies more for women than men and amplifies with age.

This research contributes to the broader theoretical comprehension of the implications of severed close familial connections on psychological well-being and the restorative role of intergenerational co-habitation. On a practical level, our findings emphasize the potential merits of policies and initiatives fostering co-residence with adult offspring. Such interventions can be instrumental in curtailing social isolation among China’s widowed older population, especially focusing on women and the significantly aged. The promotion of awareness among families regarding the psychosocial advantages of co-habitation with grieving parents might also bolster the adoption of this traditional housing model during times of evolving societal norms.

This study’s merits lie in its pioneering empirical and theoretical perspectives, emphasis on the relatively unexplored Chinese context, and the spotlight on vulnerable demographic segments. Nonetheless, despite using longitudinal data and causal mediation analysis methods, our study design is limited in its ability to fully disentangle the temporal effects of widowhood, changes in living arrangements, and subsequent loneliness. Specifically, our model does not account for potential lagged effects where changes in living arrangements following widowhood may occur at different times, influencing loneliness in a delayed manner. As a result, our findings, while informative, may not represent truly causal effects. Unobserved confounding variables and the timing of changes could still influence the observed relationships. Future research could enhance the understanding of these dynamics by incorporating lagged mediators at subsequent time points (e.g., time t + 1) to better capture post-bereavement shifts in living circumstances. Additionally, experimental designs or more advanced causal inference techniques may further establish the causal nature of these relationships. Other certain areas of limitation in our research also suggest intriguing avenues for future exploration, such as the consideration of diverse living arrangements beyond the nuclear family, integration of qualitative insights, and assessment of younger age cohorts. Collectively, our research provides a robust foundation to guide academic pursuits and policy-making, aiming to uplift the well-being of grieving older citizens in China, a country grappling with escalating widowhood rates, demographic aging, and metamorphosing familial structures.