Skip to main content
NCBI home page
The Journals of Gerontology Series B: Psychological Sciences and Social Sciences logoLink to The Journals of Gerontology Series B: Psychological Sciences and Social Sciences
. 2022 Oct 31;78(2):280–290. doi: 10.1093/geronb/gbac174

Social Isolation, Loneliness, and Depressive Symptoms: A Twelve-Year Population Study of Temporal Dynamics

Mengsha Luo 1,✉
Editor: Rodlescia S Sneed
PMCID: PMC9938924  PMID: 36315577

Abstract

Objectives

Social isolation and loneliness are two different aspects of social connections. Whether social isolation and loneliness precede depressive symptoms, or depressive symptoms precede feelings of loneliness and social isolation, or both, has not been fully established. This study aims to examine the possible reciprocity in the relationship between the two aspects of social connections and depressive symptoms among middle-aged and older adults.

Method

This study analyzed four waves of data (2008–2016) from the Health and Retirement Study (N = 5,393 individuals) and investigated within-person level cross-lagged associations of social isolation and loneliness with depressive symptoms using random intercept cross-lagged panel models.

Results

This study revealed a unidirectional relationship between social isolation and depressive symptoms and a bidirectional relationship between loneliness and depressive symptoms at the within-person level. Specifically, net of trait levels and prior states, earlier state depressive symptoms predicted future state social isolation. That is, when adults feel depressed more frequently than they usually do, they are more likely to be socially disconnected than usual at a later time. In the reverse direction, earlier state social isolation did not predict future state depressive symptoms. Within-person deviation in prior expected depressive symptoms predicted deviation in expected loneliness 4 years later and vice versa. Moreover, the strength of the two cross-lagged effects did not differ.

Discussion

Social isolation and loneliness are linked to depressive symptoms differently. Though depressive symptoms might be a potential antecedent of social isolation, they might be both a potential antecedent and an outcome of loneliness.

Keywords: Perceived isolation, Social disconnectedness, Social integration, Social relationship, Subjective isolation

Social relationships have been identified as a key predictor of human health (Berkman & Krishna, 2014; Holt-Lunstad et al., 2017). Research has consistently documented a positive effect of social relationships on a wide range of health outcomes (Cacioppo & Cacioppo, 2014; Holt-Lunstad, 2018; Umberson & Karas Montez, 2010; Umberson et al., 2010), with behavioral, psychosocial, and physiological pathways proposed to explain the observed associations. The relationship between social relationships and mental health has also been assessed extensively, with many studies demonstrating a positive association (Courtin & Knapp, 2017; Santini et al., 2015).

Social isolation and loneliness are conceptualized as two separate domains of social connections: the structural and functional aspects, respectively (National Academies of Sciences Engineering & Medicine, 2020). Despite much discussion about social isolation and loneliness in the literature, there remains a lack of clarity around the definitions and the relationship between the two concepts (Wigfield et al., 2022). Nonetheless, in line with recent systematic reviews and meta-analyses (Courtin & Knapp, 2017; Holt-Lunstad et al., 2015; Landeiro et al., 2017), this study used social isolation and loneliness to denote the objective and subjective aspects of the lack of social connections. While social isolation reflects the structural aspect of social connections and is characterized by a low quantity of social contact, loneliness reflects the functional aspect of social connections and is determined by the quality of social relationships. The coronavirus disease 2019 (COVID-19) has accelerated the significance of social connections due to the “lockdown” measures imposed by the government (van Tilburg et al., 2021). Prepandemic research reveals that more than one third of adults aged 45 and older in the United States report feeling lonely and that 24% of community-dwelling adults aged 65 and older are socially isolated (Anderson & Thayer, 2018; Cudjoe et al., 2020).

Social isolation and loneliness are closely related to depressive symptoms. Substantial research has documented a cross-sectional relationship: social isolation and loneliness are more common in adults with depressive symptoms than in their peers without depressive symptoms (Evans et al., 2019; Lim et al., 2016). Beyond the cross-sectional association, longitudinal studies and meta-analyses further indicate that social isolation and loneliness have a significant effect on depressive symptoms (e.g., Cacioppo et al., 2006; Erzen & Çikrikci, 2018; Huxhold et al., 2020; Lam et al., 2020). For instance, Segel-Karpas et al. (2022) analyzed data from three waves of the Health and Retirement Study (HRS) and found that loneliness at time 2 predicted depressive symptoms at time 3.

In line with the view of the convoy model that social relations have important health consequences (Antonucci et al., 2014), these studies assume that social isolation and loneliness lead to depressive symptoms and do not test for the reversed direction. However, it is possible that depressive symptoms may cause people to withdraw from social interactions or impair the quality of these social interactions, thereby elevating people’s feelings of loneliness. Indeed, cognitive discrepancy theory suggests that depressive symptoms affect cognitive processes and influence people’s judgments about the adequacy of social interaction (Perlman & Peplau, 1981). People with severe depressive symptoms are likely to recall negative information and experience social interactions as less rewarding (Lewis et al., 2017). Empirically, a small number of studies have examined the reversed effect and showed that depressive symptoms increase the risk of developing loneliness and isolation in late adulthood (Finlay & Kobayashi, 2018).

Though the majority of prior research has examined only one effect while overlooking the other effect, a handful of studies explicitly address the question of the temporal direction of the relationship between social isolation and depressive symptoms and between loneliness and depressive symptoms. Results from individual studies concerning the direction of both social isolation and loneliness and depressive symptoms are mixed, however. Cacioppo et al. (2010) used a 5-year study of a sample of 229 middle-aged and older adults and concluded that it is a loneliness that predicted depressive symptoms, rather than the other way around. Consistent with this unidirectional effect, a more recent study using a national sample also found that loneliness predicted subsequent changes in depressive symptomatology, but not vice versa (Domènech-Abella et al., 2021). In contrast, Domènech-Abella et al. (2019) analyzed data from The Irish Longitudinal Study on Ageing and showed a bidirectional relationship between loneliness and major depressive disorder.

The temporal direction between social isolation and depressive symptoms is uncertain as well (Berkman et al., 2000). The same study by Domènech-Abella et al. (2019) showed that the association between objective social isolation and major depressive disorder was unidirectional. Schwartz and Litwin (2019) used data from two waves of the Survey of Health, Aging and Retirement in Europe and found a reciprocal relationship between social networks and mental health, such that baseline social connectedness led to mental health improvement and a better initial mental state led to richer social networks. Using 10-year follow-up data from the National Social Life, Health, and Aging Project, Santini et al. (2020) also documented a reciprocal relationship between social isolation and depressive symptoms.

To conclude, results from prior longitudinal research concerning the directionality of the two aspects of social connections and depressive symptoms are largely inconsistent. These heterogeneities can be attributable to differences in statistical approaches, populations, and measures. The underlying direction of the associations is thus still not well understood. Overall, prior research has assessed longitudinal associations of social connections and depressive symptoms but was limited without assessing the bidirectionality; social connection has been considered as a risk factor rather than a consequence of depressive symptoms. Additionally, the few studies considering the possibility of reciprocal effects have been limited by analyzing only one aspect of social connections; few studies have considered both aspects of social connections and loneliness jointly in one sample.

The Present Study

The aim of this study is to further untangle the temporal dynamics underlying the relationship between social connections and depressive symptoms. It distinguished two aspects of social connections: social isolation and loneliness. Examining both social isolation and loneliness is important because, for some individuals, the perceived isolation can be unrelated to objective social relationships (Cornwell & Waite, 2009). Moreover, this study investigated social isolation and loneliness not only as potential antecedents of depressive symptoms but also as outcomes of depressive symptoms in middle and later life. To pursue this aim, this study used data from a nationally representative sample of American adults from the HRS that were followed up for a maximum of 12 years. Informed by the convoy model of social relations and cognitive discrepancy theory, this study hypothesized that the relationship between social isolation and depressive symptoms would be bidirectional; that is, depressive symptoms and the two aspects of social connections reciprocally influence each other over time. Additionally, it examined which effect was larger when a reciprocal effect was observed. Understanding the roles of social isolation and perceived isolation in the etiology of depressive symptoms could have implications for public health interventions aimed at preventing mental disorders and social disconnectedness.

Method

Data

Data from the HRS were used to explore the longitudinal dynamics between social connections and depressive symptoms. The HRS is a biennial longitudinal study involving a nationally representative sample of individuals aged 51 and older in the United States. The initial biennial core HRS interview was conducted in 1992. In 2006, in addition to the core interview, the HRS team initiated what is referred to as an Enhanced Face-to-Face (EFTF) Interview that included a set of social relationship measures to a random half sample. This random half sample was followed every 4 years. HRS waves 8 (2006), 10 (2010), 12 (2014), and 14 (2018) half-samples thus comprised the longitudinal sample for the present study. The study sample was restricted to those aged over 50 and had reported information on social connections and depressive symptoms. As required by the analytic method (Hamaker et al., 2015), participants were further restricted to those who attended at least three times of the EFTF. The final sample consisted of 5,393 individuals with a total of 18,777 observations.

Social Isolation

In line with recent research (Read et al., 2020), social isolation was captured by five binary items: whether the respondent (a) lived alone; had less than monthly contact, including face-to-face, telephone, or written/e-mail contact with (b) child(ren), (c) other family members, or (d) friends, and (e) did not attend meetings of nonreligious organizations, such as political, community, or other interest groups in the past month. This index incorporates living arrangements, social network size, contact frequency with network numbers, and levels of social engagement. Social isolation scores ranged from 0 to 5, with higher scores indicating higher isolation.

Loneliness

The HRS provides researchers with a three-item and an 11-item measure of loneliness. Given that the 11-item loneliness measure is available only after wave 9 (2008), the three-item measure that is available consistently beginning in wave 8 was used. This three-item measure has been widely used and shown to be a valid assessment of loneliness in the HRS sample (Hughes et al., 2004). Respondents were asked how much of the time they feel lack of companionship, left out, and isolated from others (1 = hardly ever or never, 2 = some of the time, and 3 = often). The three items were summed, with higher scores indicating more loneliness. Cronbach’s alpha of the loneliness scale for each wave was above 0.80.

Depressive Symptoms

Depressive symptoms were measured by the eight-item CES-D scale (Radloff, 1977; Steffick, 2000). Respondents were asked to respond either “yes” or “no” to statements about their feelings in the week before the interview. The eight questions covered depression, anxiety, sleep disturbance, loneliness, and sadness. Given the study objectives, one item that directly measured loneliness was excluded. The “yes” responses to the remaining seven items were summed. Cronbach’s alpha ranged from 0.76 to 0.78 across waves.

Covariates

The analyses adjusted for a set of potential confounding variables, including respondents’ sociodemographic characteristics and health status. These characteristics were measured at baseline. Sociodemographic characteristics included age (range = 51–96, in years), gender (0 = male, 1 = female), race (0 = non-Hispanic white, 1 = racial/ethnic minorities), education (0 = less than college, 1 = college and above), and employment status (0 = not employed/unemployed/retired/disabled, 1 = full-time/part-time employed). Health conditions were indicated by physical disabilities, functional limitations, chronic diseases, and self-rated health. Physical disabilities were assessed by five activities of daily living and five instrumental activities of daily living (Katz et al., 1963; Lawton & Brody, 1969). Functional limitations indicated difficulties in mobility and were measured by a count of 12 tasks that respondents had difficulty performing (Nagi, 1976; Rosow & Breslau, 1966). Chronic diseases indicated individuals’ physiological health and were measured by a count of eight chronic conditions (e.g., stroke) that respondents were diagnosed with. Self-rated health was assessed by a single question ranked on a 5-point scale ranging from 1 = excellent to 5 = poor. In addition, given the positive relationship between the two dimensions of social connections, baseline social isolation and loneliness were each controlled for when fitting models for loneliness and social isolation, respectively.

Analytic Approach

Before examining the longitudinal relationships between social connections and depressive symptoms, measurement invariance tests were performed to ensure that observed indicators were measuring the same constructs. This was done with a set of four increasingly stringent tests (Meredith, 1993): configural measurement invariance, weak measurement invariance, strong measurement invariance, and strict measurement invariance. A baseline model in which the number of factors and pattern of loadings were constrained to be identical across measurement occasions (configural invariance model) was first tested. This model was compared to a more constrained model of equal factor loadings (weak invariance model), followed by a model of equal observed variables intercepts (strong invariance model), and a model of equal unique variances (strict invariance model).

The next step was to examine the temporal dynamics between social connections and depressive symptoms. The typical modeling approach is the cross-lagged panel model (CLPM) method. Most prior research on the temporal order of social connections and depressive symptoms also used this method (Cacioppo et al., 2010; Domènech-Abella et al., 2021). However, recent research suggests that this method might conflate within-person and between-person effects due to that it assumes no stable between-person differences in the variables of interest (Mund & Nestler, 2019). To address this limitation, the random intercept cross-lagged panel models (RI-CLPM) method was proposed. The RI-CLPM decomposes observed variables into two parts: trait-like, time-constant, or between-person effect and state-like, time-varying, or within-person effect (Hamaker et al., 2015). By including a random intercept in the model, the RI-CLPM controls for trait-like, time-invariant stability. Controlling for trait-like components is important because these components do not change over time and do not easily fit into notions of causality.

Figure 1 shows the illustration of the RI-CLPM for the estimation of the longitudinal relationship between social connections and depressive symptoms. Two sets of models examining the temporal directions of social connections and depressive symptoms were fitted. These were done separately for the two dimensions of social connections. In the first set, a standard RI-CLPM was fitted. In the second set, another RI-CLPM constraining the grand means of social connections and depressive symptoms was fitted. Additionally, the CLPM was fitted to the data to produce results that could be compared with prior studies. The goodness of fit was evaluated based on chi-square (χ 2), comparative fit index (CFI), Tucker–Lewis Index (TLI), the root mean square error of approximation (RMSEA), and the standardized root mean square residual (SRMR). CFI and TLI values of 0.95 or higher, RMSEA values of 0.06 or lower, and SRMR values of 0.08 or lower are considered a relatively good model fit (Hu & Bentler, 1999).

Figure 1.

Figure 1.

Open in a new tab

The random intercept cross-lagged panel model to investigate the association between lack of social connections (indicated by social isolation and loneliness) and depressive symptoms. SC = social connections; DS = depressive symptoms; ri = random intercept.

Missing data might bias the results. Two things were done to reduce the potential bias. First, the nearest-neighbor interpolation method was used to fill in missing data by using the within-person information across waves. Second, full information maximum likelihood (FIML) estimation was used to further handle the incomplete data. This method utilizes all the information of the observed data and maximizes the likelihood of the model given the observed data and thus is superior to traditional approaches (e.g., listwise deletion) in handling missing data (Enders & Bandalos, 2001). Additionally, I compared the characteristics of the study sample and the excluded sample (Supplementary Table S2). The maximum likelihood (MLR) estimator with robust standard errors was used to take into account any nonnormality in the sample. In addition, all continuous covariates were grand-mean centered for meaningful interpretation.

Results

Table 1 presents weighted descriptive statistics of covariates and longitudinal information on social isolation, loneliness, and depressive symptoms. These descriptive statistics of all variables were weighted by cross-sectional weights designed specifically for the sample to account for the sampling structure and nonresponse.

Table 1.

Weighted Baseline Characteristics

% Mean SE
Age 61.425 0.066
Gender (female = 1) 54.041
Race (non-White = 1) 19.949
Education (college and above = 1) 29.022
Employment (employed = 1) 48.42
Physical disabilities (range: 0–10) 0.302 0.010
Functional limitations (range: 0–10) 2.315 0.023
Chronic diseases (range: 0–8) 1.521 0.011
Self-rated health (range: 1–5) 2.549 0.010
Longitudinal summaries across wave
Wave SI LN DS
 2006 1.722 4.460 1.120
 2010 1.918 4.401 1.095
 2014 2.113 4.386 1.129
 2018 2.126 4.330 1.057
Open in a new tab

Notes: SE = standard error; SI = social isolation; LN = loneliness; DS = depressive symptoms.

Supplementary Table S3 compares the characteristics of the excluded and included samples. Compared to individuals who were included in the present study, these excluded were more likely to be older (mean = 66.43 vs. 63.99, t = 19.94), male (percentages = 0.45 vs. 0.41, t = 5.93), non-White (percentages = 0.38 vs. 0.24, t = 24.30), with low education (percentages = 0.56 vs. 0.50, t = 9.31), nonemployed (percentages = 0.68 vs. 0.62, t = 8.99), physically disabled (mean = 0.71 vs. 0.28, t = 27.81), functionally limited (mean = 3.12 vs. 2.46, t = 19.30), and have more diagnosed chronic diseases (mean = 2.08 vs. 1.67, t = 23.39) and poor self-rated health (mean = 3.03 vs. 2.59, t = 32.64). Moreover, they also reported higher levels of social isolation (mean = 2.22 vs. 2.01, t = 12.56), loneliness (mean = 4.59 vs. 4.34, t = 11.49), and depressive symptoms (mean = 1.49 vs. 1.10, t = 17.68).

Table 2 shows the MLR-estimated means and correlations among the two forms of social connections and depressive symptoms. Over the 12-year follow-up, middle-aged and older adults became more socially disconnected; the social isolation score increased from 1.72 to 2.33, a 35% jump. However, they did not become lonelier or more depressed over time; the temporal trends of loneliness and depressive symptoms remained relatively stable. Table 2 also displays the correlations between the key study variables within each wave and across waves. Social isolation and depressive symptoms were correlated cross-sectionally (rs = 0.17–0.20) and longitudinally (rs = 0.13–0.19). This is even more true for the interplay between loneliness and depressive symptoms: cross-sectional relationship ranged from 0.36 to 0.41, and the longitudinal relationship ranged from 0.29 to 0.33.

Table 2.

Maximum Likelihood With Robust Standard Errors (MLR)-Estimated Means and Correlations Among Two Forms of Social Connections and Depressive Symptoms

Variables Mean SIT1 SIT2 SIT3 SIT4 DST1 DST2 DST3 DST4 LNT1 LNT2 LNT3 LNT4
SIT1 1.716 1.000
SIT2 1.961 0.616 1.000
SIT3 2.167 0.548 0.629 1.000
SIT4 2.326 0.498 0.591 0.622 1.000
DST1 1.143 0.200 0.192 0.182 0.186 1.000
DST2 1.102 0.162 0.174 0.183 0.168 0.538 1.000
DST3 1.158 0.134 0.154 0.181 0.184 0.481 0.541 1.000
DST4 1.132 0.157 0.159 0.160 0.200 0.465 0.505 0.552 1.000
LNT1 4.438 0.406 0.334 0.310 0.284 1.000
LNT2 4.354 0.330 0.371 0.298 0.289 0.594 1.000
LNT3 4.342 0.312 0.335 0.362 0.331 0.509 0.561 1.000
LNT4 4.358 0.289 0.313 0.332 0.413 0.464 0.518 0.590 1.000
Open in a new tab

Notes: SI = social isolation; LN = loneliness; DS = depressive symptoms; T1 = Time 1 (2006); T2 = Time 2 (2010); T3 = Time 3 (2014); T4 = Time 4 (2018).

The first step is to test for the measurement invariance. Supplementary Table S1 reports the results. The strict invariance model showed an adequate model fit with a CFI of 0.94 and a TLI of 0.93 for social isolation and a CFI of 0.99 and a TLI of 0.99 for loneliness, suggesting that the strict invariance models adequately captured the observed data. Thus, social isolation and loneliness were not interpreted differently over time.

Testing the Temporal Dynamics Between Social Connections and Depressive Symptoms

The next step is to examine the temporal order of social connections and depressive symptoms. Table 3 reports model fit indices. The model fit of RI-CLPM for the relationship between social isolation and depressive symptoms was acceptable, with Akaike information criterion (AIC) = 118,236, Bayesian information criterion (BIC) = 118,572, CFI = 0.98, TLI = 0.96, SRMR = 0.02, RMSEA = 0.03, χ 2 (73) = 419, p < .001 (M1a). The model fit of the RI-CLPM that constrains the grand means over time was not satisfactory (M1b). Additionally, the chi-square difference test with Satorra–Bentler correction that takes the maximum likelihood estimation into account (Satorra & Bentler, 2010) showed that the model fit of M1a was better than that of M1b (∆χ 2 = 893, p < .001). Likewise, model comparison tests showed that the RI-CLPM for the interplay between loneliness and depressive symptoms without constraining grand means fitted the data better (∆χ 2 = 21, p < .001). Therefore, later analyses were based on M1a and M2a.

Table 3.

Model Fit Indices of Different Cross-Lagged Panel Models

df χ 2 AIC BIC CFI TLI SRMR RMSEA
SI ↔ DS
M1a 73 419 118,236 118,572 0.975 0.964 0.022 0.030
M1ba 79 1,312 119,218 119,514 0.912 0.880 0.033 0.054
LN ↔ DS
M2a 73 296 127,087 127,424 0.982 0.974 0.019 0.024
M2ba 79 317 127,097 127,394 0.981 0.974 0.018 0.024
Open in a new tab

Notes: Covariates include age, gender, race/ethnicity, education level, employment status, physical disabilities, functional limitations, chronic diseases, self-rated health, and baseline social connections. SI = social isolation; LN = loneliness; DS = depressive symptoms; AIC = Akaike information criterion; BIC = Bayesian information criterion; CFI = comparative fit index; TLI = Tucker–Lewis index; SRMR = standardized root mean square residual; RMSEA = root mean square error of approximation.

aM1b and M2b constrain the grand means of social isolation and depressive symptoms to be equal across different waves.

Table 4 shows the estimates of the RI-CLPMs for the interplay between social isolation and depressive symptoms. Autoregressive and cross-lagged parameters obtained from the RI-CLPM reflect pure within-person estimates. Summarizing the within-person autoregressive effects, when an individual’s social isolation (depressive symptoms) was above his or her average level of social isolation (depressive symptoms) at a specific time point, he or she was expected to score above his or her usual level of social isolation (depressive symptoms) 4 years later. Summarizing the within-person cross-lagged effects, the positive cross-lagged path from depressive symptoms to social isolation suggested that a score above the person-specific mean in depressive symptoms at a given point in time was associated with an isolation score above the person-specific mean in social isolation at a later point in time, net of previous deviations from the person-specific mean in isolation. The path from social isolation to depressive symptoms was not significant (β = 0.001, p = .97).

Table 4.

Standardized Estimates of the RI-CLPMs for the Relationship Between Social Connections and Depressive Symptoms

SI ↔ DS LN ↔ DS
Parameter Estimate SE Estimate SE
Autoregressive pathways
 SC2006 → SC2010 (α2) 0.166*** 0.020 0.182*** 0.022
 SC2010 → SC2014 (α3) 0.153*** 0.020 0.175*** 0.022
 SC2014 → SC2018 (α4) 0.16*** 0.022 0.187*** 0.025
 DS2006 → DS2010 (δ2) 0.155*** 0.022 0.139*** 0.022
 DS2010 → DS2014 (δ3) 0.153*** 0.024 0.134*** 0.023
 DS2014 → DS2018 (δ4) 0.178*** 0.028 0.155*** 0.027
Cross-lagged pathways
 DS2006 → SC2010 (γ2) 0.037** 0.015 0.054*** 0.017
 DS2010 → SC2014 (γ3) 0.036** 0.015 0.054*** 0.017
 DS2014 → SC2018 (γ4) 0.038** 0.016 0.058*** 0.018
 SC2006 → DS2010 (β2) 0.001 0.016 0.038* 0.018
 SC2010 → DS2014 (β3) 0.001 0.015 0.036* 0.017
 SC2014 → DS2018 (β4) 0.001 0.017 0.041* 0.019
Correlations among state variables
 SC2006 ↔ DS2006 0.063** 0.022 0.185*** 0.024
 SC2010 ↔ DS2010 0.032* 0.020 0.150*** 0.024
 SC2014 ↔ DS2014 0.050** 0.020 0.149*** 0.023
 SC2018 ↔ DS2018 0.063** 0.025 0.221*** 0.026
Open in a new tab

Notes: Social connections (SC) were captured by two forms: social isolation (SI) and loneliness (LN). Models controlled for age, gender, race, education, employment status, physical disabilities, functional limitations, chronic diseases, functional limitations, self-rated health, and baseline social connections. SE = standard error; DS = depressive symptoms; RI-CLPM = random intercept cross-lagged panel models.

***p < .001; **p < .01; *p < .05.

For the longitudinal relationship between loneliness and depressive symptoms, the autoregressive pathways were also significant for state loneliness (β s = 0.17–0.19, p < .001), suggesting that state-like aspects of loneliness at an earlier time predicted loneliness years later. In addition, earlier state depressive symptoms predicted future state loneliness (β s = 0.06, p < .001). The path from loneliness to depressive symptoms was likewise positive (β s = 0.04, p = .03). Given the reciprocal relationship between loneliness and depressive symptoms, the next step is to determine which of the two cross-lagged effects was larger. The Wald test of parameter constraints indicated that ∆χ 2 = 0.51, df = 108, p = .48. Thus, the cross-lagged effect of loneliness on depressive symptoms did not differ significantly from the effect of depressive symptoms on loneliness.

In addition, there were within-time correlations: within-person increases in social isolation above his or her mean level were accompanied by his or her increases in depressive symptoms compared to his or her average (β s = 0.03–0.06). So was the relationship between loneliness and depressive symptoms (β s = 0.15–0.22). The random intercepts of social isolation and depressive symptoms that indicate trait-like, stable between-person differences were also positively related, suggesting that those who were more socially disconnected than others were also more depressed than others (β = 0.08, p = .02). Similarly, the random intercept of loneliness was closely linked to that of depressive symptoms, such that those who perceived more isolation than other individuals reported higher CES-D scores than others (β = 0.51, p < .001).

Given that prior research on the directionality between social connections and depressive symptoms has used the CLPM method, to compare with prior research, the CLPM was fitted to the data. Fit indices of the CLPMs were satisfactory. For instance, indices for the interplay between social isolation and depressive symptoms were acceptable (M1a), with AIC = 118,950, BIC = 119,267, CFI = 0.93, TLI = 0.90, SRMR = 0.03, RMSEA = 0.05, χ 2 (76) = 1,077, and p < .001. However, compared to the fit of RI-CLPM, it had higher values of AIC, BIC, SRMR, and RMSEA and lower CFI and TLI values. The chi-square difference test with Satorra–Bentler correction also suggested that the RI-CLPM provided a better representation of the data than the CLPM (∆χ 2 = 658, p < .001). Likewise, the fit of RI-CLPM for the relationship between loneliness and depressive symptoms was also better than that of the CLPM (∆χ 2 = 514, p < .001).

Supplementary Table S3 reports standardized estimates for both autoregressive and cross-lagged effects obtained from the CLPMs. In terms of the relationship between social isolation and depressive symptoms, both the autoregressive paths amounted to 0.16 for depressive symptoms and for social isolation, indicating that the rank-order (i.e., the stability of between-person differences) of the two were similar. The cross-lagged path from depressive symptoms to social isolation reached statistical significance, suggesting that prior CES-D scores were associated with subsequent social isolation. The cross-lagged path from social isolation to depressive symptoms did not reach statistical significance (p = .27). The within-time correlations indicated a codevelopment of social isolation and depressive symptoms over time such that more increases in isolation were associated with more increases in depressive symptoms. In terms of the relationship between loneliness and depressive symptoms, depressive symptoms and loneliness reciprocally influenced each other over time (Supplementary Table S3). Thus, the results replicated findings reported in the main text of a unidirectional effect of depressive symptoms on social isolation and a bidirectional relationship between loneliness and depressive symptoms, although the estimates derived from CLPMs were much larger than those derived from the RI-CLPM, which was expected given the conflation of between-person differences and within-person effect.

Discussion

Whether social isolation and loneliness predict depressive symptoms, or depressive symptoms increase feelings of loneliness and social isolation, or both, has not been fully established. This study provides further evidence on the temporal dynamics between social connections and depressive symptoms. Specifically, this study used data from nationally representative middle-aged and older adults in the United States to examine whether a person’s deviation from their average level of social connections predicted future deviation from their average level of depressive symptoms, and vice versa. Results showed that, at the within-person level, there was a unidirectional relationship between social isolation and depressive symptoms and a reciprocal relationship between loneliness and depressive symptoms such that higher levels of depressive symptoms at an earlier time predicted higher levels of loneliness at a later time and vice versa. Moreover, the strength of the two effects was not dissimilar.

First, this study did not reveal a reciprocal relationship between social isolation and depressive symptoms but a unidirectional relationship. On the one hand, earlier state social isolation did not predict future state depressive symptoms. In other words, smaller social networks or poorer social integration might not affect later state depressive symptoms. Socioemotional selectivity theory provides some insights into understanding this relationship. This theory posits that individuals are selective in terms of whom they choose to keep in their network, particularly in old age, when social networks begin to serve the function of emotion regulation (Carstensen, 1995). Thus, as individuals grow older, they tend to intensify close relationships and let go of less close relationships. Empirical research lends support to this idea. For instance, one study found that older Chinese adults who live alone are more likely to be involved in social activities, even though they are more likely to feel lonely (Gu et al., 2019).

On the other hand, earlier state depressive symptoms predicted future social isolation. When adults experience social isolation more frequently than they usually do, they are also more likely to report higher CES-D scores than usual at a later time. The finding of depressive symptoms as a predictor but not an outcome of changes in social isolation is in contrast with several prior studies (e.g., Domènech-Abella et al., 2019, 2021; Huxhold et al., 2020). This is because most prior studies tend to assume an a priori path from social isolation to depressive symptoms rather than the opposite direction (e.g., Chang et al., 2016; Lam et al., 2020). However, this finding is in line with qualitative research. One recent study using a convergent mixed-methods design reveals that poor mental health often results in social isolation; people use solitude as a coping mechanism to manage their mental health (Finlay & Kobayashi, 2018).

Second, this study revealed a bidirectional relationship between loneliness and depressive symptoms. Earlier state depressive symptoms predicted future state loneliness; at the same time, earlier state loneliness predicted future state depressive symptoms. In terms of the former association, the aforementioned cognitive discrepancy theory offers one explanation for the loneliness–depressive symptoms link. Cognitive discrepancy theory suggests that perceived isolation is unpleasant and distressing, resulting from a discrepancy between one’s desired and achieved levels of social relations (Perlman & Peplau, 1981). Depressive symptoms affect this discrepancy process: people with depressive symptoms are more likely to recall less positive information and appraise their social interactions more negatively (Burholt & Scharf, 2014). In terms of the cross-lagged effect of loneliness on depressive symptoms, Hawkley and Cacioppo (2010) employed the loneliness regulatory loop to explain the health consequences of loneliness: loneliness is accompanied by impairments in attention, behavior, cognition, and affect, which can have health consequence through their impact on genetic, neural, and hormonal mechanisms.

Moreover, the strength of the cross-lagged effect of depressive symptoms on loneliness was not significantly different from that of the cross-lagged effect of loneliness on depressive symptoms. Indeed, loneliness is strongly linked to depressive symptoms to the extent that prior research claims that loneliness might be an associated symptom of depression (Radloff, 1977). Of note, the equal strength of the two cross-lagged effects does not necessarily mean that loneliness and depressive symptoms are identical. The two constructs are empirically and theoretically distinct (Hawkley et al., 2008). Loneliness is about how people feel about their social connections in particular and depressive symptoms are about how people feel generally (Weiss, 1973). Nonetheless, the reciprocal association between the two constructs suggests that middle-aged and older persons may face a “vicious cycle”: people with high perceived isolation will experience poor mental health, and such poor mental health outcomes will further cause more feelings of loneliness by reducing their ability to maintain their usual social interactions. However, on a positive note, reducing one symptom may mean alleviation of another symptom.

Several other findings are worth noting. Controlling for trait-like tendencies, earlier state depressive symptoms predicted future state depressive symptoms. Likewise, earlier state isolation and loneliness predicted future state isolation and loneliness, respectively. In other words, when adults report more social isolation, loneliness, and depressive symptoms than they usually do at a specific time point, 4 years later, they will also experience more isolation, loneliness, and depressive symptoms than usual, respectively. Such results are consistent across the three 4-year intervals. Additionally, positive within-time correlations existed: within-person increases in social isolation or loneliness above person-specific mean level were accompanied by person-specific increases in depressive symptoms compared to his or her average. Beyond these within-person level associations, significant between-person level associations existed as well. The two aspects of social connections were positively associated with depressive symptoms, indicating that individuals with higher social isolation or loneliness than other individuals tended to report higher depressive symptoms than other individuals. Thus, the temporal dynamics among the trait- and state-like components of social isolation, loneliness, and depressive symptoms are nuanced and must be carefully disentangled to shed light on the underlying processes.

The findings of this research have theoretical and practical implications. First, this study further suggests the necessity of disentangling the roles of social isolation from loneliness. The two are different aspects of social connections and are linked to depressive symptoms differently. With or without controlling for trait-like aspects, social isolation was predicted by depressive symptoms and not the other way around; loneliness both predicted and was predicted by depressive symptoms. Second, that prior state social isolation did not lead to future depressive symptoms suggests that public health interventions aimed at preventing or alleviating mental disorders shall not target promoting social connections first. However, state loneliness led to future depressive symptoms such that individuals who felt lonelier than they usually do were also more likely to display higher depressive symptoms than usual years later. Thus, efforts to address loneliness could aid in the prevention of mental disorders. Meanwhile, clinicians, care providers, or social workers should also be aware that mid- and later-life depressive symptoms can precipitate loneliness; adults with depressive symptoms might need intervention to combat loneliness. In short, it is essential to address both the structural and functional aspects of social connections in a complementary way to promote the psychological health of the middle-aged and older population.

This study has several strengths. It contributes to the field by examining the associations of both social isolation and loneliness with depressive symptoms in the same sample. Prior studies have focused on only one aspect of social connections and rarely have considered the three constructs together. Additionally, this study uses four measurement occasions that span 12 years from a large, nationally representative prospective study to investigate the temporal dynamics underlying the association between social isolation, loneliness, and depressive symptoms. Prior research on the direction of social connections and depressive symptoms only collected data for a short time period or used a small sample. Moreover, this study uses the RI-CLPM method to distinguish the trait- and state-like components of social connections and depressive symptoms, which can better assess the temporal precedence and causality.

This study entails several limitations. First, the scales for depressive symptoms were based on the self-report of symptoms. The results might differ if depressive symptoms have been assessed by clinical evaluation and diagnosis. Second, the excluded sample was older, more likely to be men, non-White, poorly educated, nonemployed, and had worse physical and mental health as well as more social disconnectedness. Thus, even though the FIML method was used to reduce selective attrition bias, attrition might still affect the results. Additionally, social isolation and loneliness were only two aspects of social connections and might be insufficient to provide a full understanding of social relationships and their well-being consequences.

To conclude, within-person autoregressive effects reveal that state social isolation, loneliness, and depressive symptoms predict themselves over a 12-year observation period. Within-person cross-lagged effects reveal a unidirectional effect between social isolation and depressive symptoms and a bidirectional effect between loneliness and depressive symptoms. Earlier state depressive symptoms predict future state social isolation and not vice versa; thus, when middle-aged and older adults feel depressed more frequently than they usually do, they become more socially isolated than usual at a later time. Earlier state depressive symptoms predict future state loneliness and earlier state loneliness predicts future state depressive symptoms; moreover, the strength of the two bidirectional associations is not different. In addition to within-person level cross-lagged effects, there are consistent between-person level cross-lagged effects such that individuals with higher levels of social isolation or loneliness than other individuals tend to have more depressive symptoms than other individuals.

Supplementary Material

gbac174_suppl_Supplementary_Materials
Click here for additional data file. (100.8KB, pdf)

Acknowledgments

The conducted research was not preregistered with an analysis plan in an independent, institutional registry. The current study made use of publicly available deidentified data from the Health and Retirement Study (HRS) website: https://hrs.isr.umich.edu/.

Funding

Data collection for the Health and Retirement Study (HRS) was supported by the National Institute on Aging (NIA U01AG009740) and conducted by the University of Michigan. This study was supported by the National Social Science Fund of China (Grant No. 20CRK007).

Conflict of Interest

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References

  1. Anderson, G. O., & Thayer, C. E. (2018). Loneliness and social connections: A national survey of adults 45 and older. Washington, DC: AARP Foundation. doi: 10.26419/res.00246.001 [DOI] [Google Scholar]
  2. Antonucci, T. C., Ajrouch, K. J., & Birditt, K. S. (2014). The convoy model: Explaining social relations from a multidisciplinary perspective. Gerontologist, 54, 82–92. doi: 10.1093/geront/gnt118 [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Berkman, L., & Krishna, A. (2014). Social network epidemiology. In Berkman L., Kawachi I., & Glymour M. (Eds.), Social epidemiology(pp. 234–289). New York, NY: Oxford University Press. [Google Scholar]
  4. Berkman, L. F., Glass, T., Brissette, I., & Seeman, T. E. (2000). From social integration to health: Durkheim in the new millennium. Social Science and Medicine, 51, 843–857. doi: 10.1016/S0277-9536(00)00065-4 [DOI] [PubMed] [Google Scholar]
  5. Burholt, V., & Scharf, T. (2014). Poor health and loneliness in later life: The role of depressive symptoms, social resources, and rural environments. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 69, 311–324. doi: 10.1093/geronb/gbt121 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Cacioppo, J. T., & Cacioppo, S. (2014). Social relationships and health: The toxic effects of perceived social isolation. Social and Personality Psychology Compass, 8, 58–72. doi: 10.1111/spc3.12087 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Cacioppo, J. T., Hawkley, L. C., & Thisted, R. A. (2010). Perceived social isolation makes me sad: 5-year cross-lagged analyses of loneliness and depressive symptomatology in the Chicago Health, Aging, and Social Relations Study. Psychology and Aging, 25, 453–463. doi: 10.1037/a0017216 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Cacioppo, J. T., Hughes, M. E., Waite, L. J., Hawkley, L. C., & Thisted, R. A. (2006). Loneliness as a specific risk factor for depressive symptoms: Cross-sectional and longitudinal analyses. Psychology and Aging, 21, 140–151. doi: 10.1037/0882-7974.21.1.140 [DOI] [PubMed] [Google Scholar]
  9. Carstensen, L. L. (1995). Evidence for a life-span theory of socioemotional selectivity. Current Directions in Psychological Science, 4, 151–156. doi: 10.1111/1467-8721.ep11512261 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Chang, S.-C., Pan, A., Kawachi, I., & Okereke, O. I. (2016). Risk factors for late-life depression: A prospective cohort study among older women. Preventive Medicine, 91, 144–151. doi: 10.1016/j.ypmed.2016.08.014 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Cornwell, E. Y., & Waite, L. J. (2009). Social disconnectedness, perceived isolation, and health among older adults. Journal of Health and Social Behavior, 50, 31–48. doi: 10.1177/002214650905000103 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Courtin, E., & Knapp, M. (2017). Social isolation, loneliness and health in old age: A scoping review. Health and Social Care in the Community, 25, 799–812. doi: 10.1111/hsc.12311 [DOI] [PubMed] [Google Scholar]
  13. Cudjoe, T. K. M., Roth, D. L., Szanton, S. L., Wolff, J. L., Boyd, C. M., & Thorpe, R. J.Jr (2020). The epidemiology of social isolation: National Health and Aging Trends Study. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 75, 107–113. doi: 10.1093/geronb/gby037 [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Domènech-Abella, J., Mundó, J., Haro, J. M., & Rubio-Valera, M. (2019). Anxiety, depression, loneliness and social network in the elderly: Longitudinal associations from The Irish Longitudinal Study on Ageing (TILDA). Journal of Affective Disorders, 246, 82–88. doi: 10.1016/j.jad.2018.12.043 [DOI] [PubMed] [Google Scholar]
  15. Domènech-Abella, J., Mundó, J., Switsers, L., van Tilburg, T., Fernández, D., & Aznar-Lou, I. (2021). Social network size, loneliness, physical functioning and depressive symptoms among older adults: Examining reciprocal associations in four waves of the Longitudinal Aging Study Amsterdam (LASA). International Journal of Geriatric Psychiatry, 36, 1541–1549. doi: 10.1002/gps.5560 [DOI] [PubMed] [Google Scholar]
  16. Enders, C. K., & Bandalos, D. L. (2001). The relative performance of full information maximum likelihood estimation for missing data in Structural Equation Models. Structural Equation Modeling, 8, 430–457. doi: 10.1207/S15328007SEM0803_5 [DOI] [Google Scholar]
  17. Erzen, E., & Çikrikci, O. (2018). The effect of loneliness on depression: A meta-analysis. International Journal of Social Psychiatry, 64, 427–435. doi: 10.1177/0020764018776349 [DOI] [PubMed] [Google Scholar]
  18. Evans, I. E. M., Llewellyn, D. J., Matthews, F. E., Woods, R. T., Brayne, C., & Clare, L. (2019). Social isolation, cognitive reserve, and cognition in older people with depression and anxiety. Aging and Mental Health, 23, 1691–1700. doi: 10.1080/13607863.2018.1506742 [DOI] [PubMed] [Google Scholar]
  19. Finlay, J. M., & Kobayashi, L. C. (2018). Social isolation and loneliness in later life: A parallel convergent mixed-methods case study of older adults and their residential contexts in the Minneapolis metropolitan area, USA. Social Science and Medicine, 208, 25–33. doi: 10.1016/j.socscimed.2018.05.010 [DOI] [PubMed] [Google Scholar]
  20. Gu, D., Feng, Q., & Yeung, W. J. (2019). Reciprocal dynamics of solo-living and health among older adults in contemporary China. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 74, 1441–1452. doi: 10.1093/geronb/gby140 [DOI] [PubMed] [Google Scholar]
  21. Hamaker, E. L., Kuiper, R. M., & Grasman, R. P. P. P. (2015). A critique of the cross-lagged panel model. Psychological Methods, 20, 102–116. doi: 10.1037/a0038889 [DOI] [PubMed] [Google Scholar]
  22. Hawkley, L. C., & Cacioppo, J. T. (2010). Loneliness matters: A theoretical and empirical review of consequences and mechanisms. Annals of Behavioral Medicine, 40, 218–227. doi: 10.1007/s12160-010-9210-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Hawkley, L. C., Hughes, M. E., Waite, L. J., Masi, C. M., Thisted, R. A., & Cacioppo, J. T. (2008). From social structural factors to perceptions of relationship quality and loneliness: The Chicago Health, Aging, and Social Relations Study. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 63, S375–S384. doi: 10.1093/geronb/63.6.s375 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Holt-Lunstad, J. (2018). Why social relationships are important for physical health: A systems approach to understanding and modifying risk and protection. Annual Review of Psychology, 69, 437–458. doi: 10.1146/annurev-psych-122216-011902 [DOI] [PubMed] [Google Scholar]
  25. Holt-Lunstad, J., Robles, T. F., & Sbarra, D. A. (2017). Advancing social connection as a public health priority in the United States. American Psychologist, 72, 517–530. doi: 10.1037/amp0000103 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Holt-Lunstad, J., Smith, T. B., Baker, M., Harris, T., & Stephenson, D. (2015). Loneliness and social isolation as risk factors for mortality: A meta-analytic review. Perspectives on Psychological Science, 10, 227–237. doi: 10.1177/1745691614568352 [DOI] [PubMed] [Google Scholar]
  27. Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55. doi: 10.1080/10705519909540118 [DOI] [Google Scholar]
  28. Hughes, M. E., Waite, L. J., Hawkley, L. C., & Cacioppo, J. T. (2004). A short scale for measuring loneliness in large surveys: Results from two population-based studies. Research on Aging, 26, 655–672. doi: 10.1177/0164027504268574 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Huxhold, O., Fiori, K. L., Webster, N. J., & Antonucci, T. C. (2020). The strength of weaker ties: An underexplored resource for maintaining emotional well-being in later life. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 75, 1433–1442. doi: 10.1093/geronb/gbaa019 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Katz, S., Ford, A. B., Moskowitz, R. W., Jackson, B. A., & Jaffee, M. W. (1963). Studies of illness in the aged. The index of ADL: A standardized measure of biologic and psychologic function. JAMA, 185, 914–919. doi: 10.1001/jama.1963.03060120024016 [DOI] [PubMed] [Google Scholar]
  31. Lam, B. C. P., Haslam, C., Steffens, N. K., Yang, J., Haslam, S. A., Cruwys, T., & Pachana, N. A. (2020). Longitudinal evidence for the effects of social group engagement on the cognitive and mental health of Chinese retirees. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 75, 2142–2151. doi: 10.1093/geronb/gbz134 [DOI] [PubMed] [Google Scholar]
  32. Landeiro, F., Barrows, P., Nuttall Musson, E., Gray, A. M., & Leal, J. (2017). Reducing social isolation and loneliness in older people: A systematic review protocol. BMJ Open, 7, e013778. doi: 10.1136/bmjopen-2016-013778 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Lawton, M. P., & Brody, E. M. (1969). Assessment of older people: Self-maintaining and instrumental activities of daily living. Gerontologist, 9, 179–186. doi: 10.1093/geront/9.3_Part_1.179 [DOI] [PubMed] [Google Scholar]
  34. Lewis, G., Kounali, D. Z., Button, K. S., Duffy, L., Wiles, N. J., Munafò, M. R., Harmer, C. J., & Lewis, G. (2017). Variation in the recall of socially rewarding information and depressive symptom severity: A prospective cohort study. Acta Psychiatrica Scandinavica, 135, 489–498. doi: 10.1111/acps.12729 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Lim, M. H., Rodebaugh, T. L., Zyphur, M. J., & Gleeson, J. F. (2016). Loneliness over time: The crucial role of social anxiety. Journal of Abnormal Psychology, 125, 620–630. doi: 10.1037/abn0000162 [DOI] [PubMed] [Google Scholar]
  36. Meredith, W. (1993). Measurement invariance, factor analysis and factorial invariance. Psychometrika, 58, 525–543. doi: 10.1007/bf02294825 [DOI] [Google Scholar]
  37. Mund, M., & Nestler, S. (2019). Beyond the cross-lagged panel model: Next-generation statistical tools for analyzing interdependencies across the life course. Advances in Life Course Research, 41, 100249. doi: 10.1016/j.alcr.2018.10.002 [DOI] [PubMed] [Google Scholar]
  38. Nagi, S. Z. (1976). An epidemiology of disability among adults in the United States. Milbank Memorial Fund Quarterly. Health and Society, 54, 439–467. doi: 10.2307/3349677 [DOI] [PubMed] [Google Scholar]
  39. National Academies of Sciences Engineering and Medicine. (2020). Social isolation and loneliness in older adults: Opportunities for the health care system. Washington, DC: National Academies Press. [PubMed] [Google Scholar]
  40. Perlman, D., & Peplau, L. A. (1981). Toward a social psychology of loneliness. In Duck S. & Gilmour R. (Eds.), Personal relationships: Personal relationships in disorder (Vol. 3, pp. 31–56). London: Academic Press. [Google Scholar]
  41. Radloff, L. S. (1977). The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385–401. doi: 10.1177/014662167700100306 [DOI] [Google Scholar]
  42. Read, S., Comas-Herrera, A., & Grundy, E. (2020). Social isolation and memory decline in later-life. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 75, 367–376. doi: 10.1093/geronb/gbz152 [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Rosow, I., & Breslau, N. (1966). A Guttman health scale for the aged. Journal of Gerontology, 21, 556–559. doi: 10.1093/geronj/21.4.556 [DOI] [PubMed] [Google Scholar]
  44. Santini, Z. I., Jose, P. E., York Cornwell, E., Koyanagi, A., Nielsen, L., Hinrichsen, C., Meilstrup, C., Madsen, K. R., & Koushede, V. (2020). Social disconnectedness, perceived isolation, and symptoms of depression and anxiety among older Americans (NSHAP): A longitudinal mediation analysis. Lancet Public Health, 5, e62–e70. doi: 10.1016/S2468-2667(19)30230-0 [DOI] [PubMed] [Google Scholar]
  45. Santini, Z. I., Koyanagi, A., Tyrovolas, S., Mason, C., & Haro, J. M. (2015). The association between social relationships and depression: A systematic review. Journal of Affective Disorders, 175, 53–65. doi: 10.1016/j.jad.2014.12.049 [DOI] [PubMed] [Google Scholar]
  46. Satorra, A., & Bentler, P. M. (2010). Ensuring positiveness of the scaled difference Chi-square test statistic. Psychometrika, 75, 243–248. doi: 10.1007/s11336-009-9135-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Schwartz, E., & Litwin, H. (2019). The reciprocal relationship between social connectedness and mental health among older European adults: A SHARE-based analysis. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 74, 694–702. doi: 10.1093/geronb/gbx131 [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Segel-Karpas, D., Cohn-Schwartz, E., & Ayalon, L. (2022). Self-perceptions of aging and depressive symptoms: The mediating role of loneliness. Aging and Mental Health, 26, 1495–1501. doi: 10.1080/13607863.2021.1991275 [DOI] [PubMed] [Google Scholar]
  49. Steffick, D. E. (2000). Documentation of affective functioning measures in the Health and Retirement Study. HRS/AHEAD Documentation Report DR-005. 2000. Ann Arbor, MI: Survey Research Center, University of Michigan. [Google Scholar]
  50. Umberson, D., Crosnoe, R., & Reczek, C. (2010). Social relationships and health behavior across the life course. Annual Review of Sociology, 36, 139–157. doi: 10.1146/annurev-soc-070308-120011 [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Umberson, D., & Karas Montez, J. (2010). Social relationships and health: A flashpoint for health policy. Journal of Health and Social Behavior, 51, S54–S66. doi: 10.1177/0022146510383501 [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. van Tilburg, T. G., Steinmetz, S., Stolte, E., van der Roest, H., & de Vries, D. H. (2021). Loneliness and mental health during the COVID-19 pandemic: A study among dutch older adults. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 76, e249–e255. doi: 10.1093/geronb/gbaa111 [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Weiss, R. S. (1973). Loneliness: The experience of emotional and social isolation. Cambridge, MA, USA: The MIT Press. [Google Scholar]
  54. Wigfield, A., Turner, R., Alden, S., Green, M., & Karania, V. K. (2022). Developing a new conceptual framework of meaningful interaction for understanding social isolation and loneliness. Social Policy and Society, 21, 172–193. doi: 10.1017/S147474642000055X [DOI] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

gbac174_suppl_Supplementary_Materials
Click here for additional data file. (100.8KB, pdf)

Articles from The Journals of Gerontology Series B: Psychological Sciences and Social Sciences are provided here courtesy of Oxford University Press

RESOURCES