1. Introduction
In the United States, there are over 17.3 million adults (7.1% of the adult population) who have experienced a major depressive episode in the past 12 months [
1], making depression one of today’s most prevalent chronic adverse health outcomes [
2]. Using almost any metric, depression is both debilitating and costly. More than a psychological burden, depression has numerous long-term health consequences, including heart disease, injuries, inflammation, and lowered immunity functioning, and has been linked to suicide and suicide ideation [
3,
4]. The economic burden of depression resulting from disability, morbidity, and mortality is estimated to be
$USD 210 billion annually [
5].Further, the CDC reports that one-year after COVID-19, the percentage of adults reporting recent symptoms of anxiety and depressive disorder increased from 36.4% to 41.55% [
6].
Therefore, after a time of isolation and quarantine, it is especially important to understand what (e.g., which features of the social environment) might help mitigate depression and, therefore, should be targets of future intervention trials. For clinicians, depression is typically defined as a persistent intra-psychic disorder in which individuals feel sad and hopeless and derive little pleasure or have little interest in activities, but depression is often influenced by the social contexts in which an individual is embedded [
7,
8]. To study these social networks, researchers often measure personal (or ego) networks: these are the self-identified set of social relationships immediately surrounding a focal individual who provide support and protect that focal individual from becoming depressed or help him or her more effectively cope with the depression symptoms [
9]. Although the positive relationship between social support and protection from depression is well-documented [
9,
10,
11,
12,
13,
14,
15], there is a significant research gap: it is unclear how pathways of support are organized (i.e., the structure of personal support networks) for depressed and non-depressed individuals and how those structural features impact an individual’s engagement in critical depression mitigation behaviors. In one of the few published findings in this area, Hall and colleagues [
12] found that among seniors (i.e., 62 and older) the size of the ego’s self-reported network and its mean degree (number of observed links divided by potential links among network members) promoted open communication, which in turn can mitigate depression. These are promising findings but have three limitations. First, these networks are not specific to the types of topics (e.g., important or intimate matters) expected to be central in mitigating depression (ref). Second, it was unclear if these findings would generalize to other age groups. Third, the structural features examined of these networks were limited.
Indeed, social support networks and their features can be understood in terms of the role of a specific type of personal support network called a confidant network. A confidant network consists of the ego and the ego’s strong-tie relationships (i.e., confidants) to which the ego turns to talk about important matters [
12,
16,
17]. Through their companionship and support, confidants are integral to an individual’s ability to avoid and/or attenuate feelings of distress and depression. What, however, are the network structural features of social support (or confidant) networks that may effectively mitigate depression? The answer to this question remains unclear.
To fill this gap by drawing on cross-sectional confidant network data collected from 1002 adults, we investigate: (1) the degree to which the confidant networks of depressed and non-depressed individuals are structurally and compositionally different, and (2) the degree to which three structural features of confidant networks—i.e., the number of confidants (network size), average number of ties of a network member (mean degree), and degree to which confidants are tied to other confidants (constraint)—are associated with engagement in depression mitigation behaviors (e.g., activities that promote personal mental health benefit). Answering these questions has implications not only for how we theorize the relational infrastructure that underpins social support in both depressed and non-depressed populations but also for depression mitigation interventions.
4. Results
4.1. Participant Characteristics
A summary of sociodemographic characteristics of the sample, stratified by depression status, is shown in
Table 2. Results show that depressed and non-depressed participants were comparable in demographics, with majorities in both subsamples being White/Non-Hispanic, earning less than
$USD 60,000 per year, and female. Participants in both samples were on average 37.4 years of age (SD = 12.79). There are also no significant differences between the two samples regarding age, years of education, and race/ethnicity. In preliminary analyses (not shown here), we also entered age, education, race/ethnicity, income, religious/political affiliation as covariates; none of them were significantly related to depression mitigation behaviors. Therefore, none of those variables were included in the featured analysis.
Given the gender bias in our samples (82% of depressed participants and 79% of non-depressed participants were female) and known associations between gender and depression, we performed a series of t-tests (not shown) to compare network measures between females and males. There were no significant differences by gender regarding the network structure measures, however, female participants reported greater gender diversity in their personal confidant networks (p < 0.01). As such, interpretation of our findings should be made with this caveat in mind.
4.2. Confidant Network Characteristics
Irrespective of depression status, participants reported on average M = 4.17 confidants (SD = 1.68) with a range of 0 to 6. Both samples named a similar proportion of family members in their network (30% in the depressed sample and 29% in the non-depressed sample). As expected, depressed participants reported more connections with therapists, 46 out of 390 depressed participants reported having therapists in their network, while only 17 out of 686 non-depressed participants named a therapist as a confidant.
It is worth noting that 37 participants (n = 29 in the depressed sample and n = 8 in the non-depressed sample) reported no confidants. Some participants nominated pets or gods (n = 7 in the depressed sample, n = 20 in the non-depressed sample) as one of their confidants. Although we recognize that pets and spiritual figures play important roles in the lives of these participants, we opted to exclude these participants from the analysis.
To understand how this exclusion decision might affect our analysis, we compared differences in depression mitigation behavior between those who did not nominate a confidant (or did not name a human confidant) and those who nominated at least one eligible confidant. Not assuming equal variances, independent t-tests showed significantly lower depression mitigation engagement among participants who did not have a confidant (M = 3.84, SD = 0.17) than those who did have at least one confidant (M = 4.35, SD = 0.04) (p < 0.01, t (1052) = 3.21, 95% CI:(−0.68, −1.65), Cohen’s d = 4.13).
4.3. Differences in Network Structure: Depressed Versus Non-Depressed Samples
Table 3 provides the raw scores of each network measure separately for the clinically depressed and non-clinically depressed samples. Results indicate that while depressed participants nominated fewer confidants on average (
M = 4.15, SD = 1.71) than non-depressed participants (
M = 4.19, SD = 1.63), this difference is not significant (
t = −0.31,
p > 0.50). There were also no significant differences in network constraint between depressed (
M = 0.61, SD = 0.15) and non-depressed (
M = 0.61, SD = 0.14) participants. Finally, there were also no significant differences in the family/kin composition of depressed (
M = 0.29, SD = 0.38) and non-depressed (
M = 0.29, SD = 0.34) participants. We only found significant differences between the confidant networks of depressed and non-depressed participants in terms of mean degree, the confidants of depressed participants had significantly fewer ties with the other confidants in their networks (
M = 0.74, SD = 1.12) than the confidants of non-depressed participants (
M = 1.95, SD = 1.12) (
p < 0.01). Other measures (i.e., network size, constraint) were not significantly different.
4.4. Relationships between Network Structure and Depression Mitigation Behaviors
Among the 1002 participants, 214 (21%) reported fewer than three confidants. As some of the network structure variables cannot be computed on personal networks with fewer than three alters (e.g., personal network constraint), we opted to restrict our analytic sample for the second portion of the analysis to only participants who nominated three or more confidants. In a pre-test (not shown here), we compared the differences in depression mitigation engagement between those who nominated fewer than three confidants and those who named three confidants or more. Not assuming equal variances, independent t-tests showed no significant differences.
Results of the hierarchical linear regression models are shown in
Table 4,
Table 5 and
Table 6. Network size predicted depression mitigation engagement (
B = 0.11, SE = 0.04,
p < 0.01), such that larger network size was associated with more depression mitigating behaviors. This supports our H1, which predicts the positive associations between network size and depression mitigation behaviors. While mean degree was not significantly related to actively engaging in depression mitigation behaviors (
B = −0.05, SE = 0.07,
p > 0.05), network constraint is related to depression mitigation behaviors (
B = −1.15, SE = 0.48,
p < 0.05), such that having a more constrained confidant network was negatively associated with more depression mitigation behaviors, providing answers to our research question on the role of constraint in mitigating depression behaviors (see detailed discussion of this result in the discussion section). Furthermore, all models were adjusted for depression status, and in all cases, having experienced clinical depression did not significantly predict depression mitigation engagement. That is, clinically depressed individuals neither uniformly differed from non-depressed individuals in the degree to which they engaged in behaviors to mitigate their depression when network structure factors were accounted for.
5. Discussion
This study, for the first time, investigated associations between confidant network structures and depression mitigation behaviors in an online general sample of clinically depressed and non-depressed adults across the age continuum. We also compared the confidant networks of clinically depressed individuals and non-depressed individuals. Although simply having a confidant network was a significant predictor of one’s engagement in depression mitigation behaviors, as our preliminary analysis showed, we also learned that, irrespective of depression status, specific network structures—namely, network size and constraint—played what could be meaningful roles in enabling depression mitigation engagement, at least among those who had a confidant network. These results remind us of the importance of considering an individual’s social embeddedness when accounting for the factors that facilitate or impede their ability to cope with life stresses. Therefore, the featured analysis adds to existing empirical research on the role of interpersonal factors as main predictors of mental health outcomes. Specifically, we showed that in addition to our existing knowledge about the role of network size (i.e., the number of confidants that a person has access to) in mitigating depression, it is important to also consider the relationships among confidants (e.g., the possible value of low constraint) in order to more fully understand the effects of network structure on depression mitigation. These findings suggest a possible intervention target, namely, adding a network member (e.g., a therapist), who does not have a relationship with the ego’s other social network members, might enhance ego’s depression mitigation behaviors. We discuss the findings for constraint in this regard in more detail below.
Our findings have implications for how we think about depression and its mitigation. First, we did not find a statistically significant difference between depressed and non-depressed individuals regarding network size in our sample. However, this result is not inconsistent with prior work that has identified associations between social isolation (lack of social connections) and depression [
51]. Keep in mind that in this analysis, we only included participants who nominated at least three confidants—those are participants that may already have sufficient levels of social support. Also, from those participants, we found network size played a significant role in mitigating depression, which is consistent with previous studies that focus on the size of networks in protecting one against depression [
2,
52]. Our results on network size, therefore, have positive implications: when in a time of need, clinically depressed people may have a sufficient confidant network upon which to rely, at least at the onset of their depression. The number of confidants one has in one’s social network matters in helping one seek behaviors that can contribute to the improvement of one’s depression. Indeed, it is necessary for researchers to look into ways that can help clinically depressed individuals in maintaining or expanding their confidant networks for the sake of improving their depressive symptoms.
Second, in our comparison of the network structure of depressed and non-depressed participants, we found significant differences in mean degree such that the confidants of non-clinically depressed individuals were, on average, more connected than those who are clinically depressed (shown through mean degree). This is not unexpected when evaluating a confidant network. Additionally, as reported in the previous section, only 2% (17 out of 686) of non-depressed participants reported having therapists/mental health professionals outside of their regular confidant networks. Indeed, non-depressed individuals’ networks can be more “natural” given that non-depressed people reported highly connected clusters of individuals as their confidants. However, our results showed that this type of cohesive structure—indicated by the average of mutual ties shared among confidants—does not necessarily enable depression mitigation behaviors. This is also not unexpected because having more confidants who know one another does not necessarily motivate behaviors to manage or mitigate depression. Having inter-connected clusters and groups of individuals (e.g., a sister, mom, and boyfriend) may lead one to actively engage with those in a cluster in pursuit of a common goal or activity, in which case any given member (such as the ego) might more passively be drawn in, rather than actively initiating or planning the activity. Indeed, if such a network initiates more group activity with a depressed ego, then it is likely that there would be less need for egos to initiate depression mitigating behaviors by themselves, therefore, suggesting an unclear correlational pattern between the connectedness dimension of cohesiveness and depression mitigation.
Third, in our study, networks for depressed and non-depressed individuals did not differ in constraint. However, our results also showed that this structure—a more radial rather than interconnected confidant network—can be very important in motivating depression mitigation behaviors. For depressed individuals, a radial confidant network can very well represent participants’ actively seeking information and privacy management. Indeed, we found that networks with more constraints (i.e., when more alters are connected) can hinder depression mitigating behaviors. This result is consistent with previous studies that showed low constraint can be beneficial for individuals under other information-seeking circumstances (where access to unique types of information is desired [
27].
Figure 1A is a relatively less constrained confidant network, the ego seems to be the bridge between the “therapist (as nick-named by the participant)” and the family members. This structure indicates that the depressed participants are more in control of his/her conditions. Compare this to the network shown in
Figure 1B, where the two “shrinks (also as nick-named)” of the ego know each other, and also know the ego’s closest confidants, such as her boyfriend. The information flow between those connected alters may “constrain” the ego’s active management of his/her mental wellbeing. Indeed, within a more constrained confidant network where everyone knows ego’s state of depression, egos may see less reward value from taking active control of his/her condition. That is, in a “less cohesive network” with more structural holes, depressed egos may have more needs, since they may have less contact with alters and, therefore, have more motivation to initiate depression mitigating behaviors. Further studies are needed to investigate the mechanisms that may enable “less constrained” networks to play a positive role. For example, are such networks effective because they encourage those who are depressed to actively seek control of their mental health conditions?
Lastly, our results showed that when accounting for the effects of network structures (i.e., size, network composition), depression diagnosis (depressed or not) is not a significant predictor of depression mitigation behaviors for those who have a confidant network. Those results speak to the importance of having a confidant network. That is, depression mitigation behaviors can be activated, whether one is clinically depressed or not, via a confidant network. Indeed, as reported in previous sections, those who do not have a confidant network, also despite their depression diagnosis, showed significantly less depression mitigation behaviors. Those results indicate that depression may be less likely to occur if one has the right number of people to whom one can confide (represented by size), especially with this group of people organized in a way that can help the depressed to manage their symptoms (represented in constraint). Indeed, having the right type of confidant networks matters a lot in helping one combat depression. Perhaps subsequent interventions can use this result to develop more effective interventions in developing interpersonal connections.
6. Implications of Network Interventions
The implications of these results could be profound in designing depression interventions from a network perspective. The current work shows the importance of situating the clinically depressed individuals within a network of members who know and care about the ego’s situation (i.e., size), and meanwhile where the ego has the capacity to actively coordinate with each of these others to help him or her combat depression (e.g., ego actively managing depression condition). Especially, the development of social media platforms and mobile technologies can enable such interventions. For example, a just-in-the-moment, adaptive intervention (JITAI) approach [
53], is a means of delivering interventions such that participants can receive in-the-moment feedback on their social network platforms or mobile phones; it has already been shown to be quite advantageous in improving mental health outcomes [
54]. Adopting the JITAI framework, participants can get in-the-moment feedback both on support-seeking behaviors and on maintaining privacy. Therefore, JITAI interventions can have the capacity to (1) identify the social support network of participants on social network sites, (2) create messages/intervention materials in helping participants to better communicate while maintaining personal privacy when participants need help, and (3) monitor the evolution of social network structures among the participants. Such interventions may be able to elicit a change of social support structure for the depressed participants (especially altering constraint structures).
Indeed, the challenge might be to provide the ego with optimal support while navigating privacy and communication issues throughout this care network. However, if those challenges are addressed (as in our JITAI example), changing the social network structures among the depressed (e.g., encouraging the patient to actively bridge information between health providers and family members) could facilitate information flow regarding the patient [
55]. It could provide caretakers of depressed individuals with greater levels of social support and generate synergetic effects in helping the depressed individuals [
56,
57]. Indeed, one’s social networks could be another promising targeted intervention component that matters in combating depression.
7. Limitations and Future Directions
Even though this study has shown the importance of confidant networks in depression, it has some limitations. Specifically, it is still limited in its scope in terms of the studied problem. First, because of the limits of the design, we were not able to collect changes in members of confidant networks over time. Therefore, results of this study can only be interpreted as associations rather than as providing evidence of a causal relationship between depression mitigating behaviors and changes in one’s social relationships. In a subsequent study, it would be even more meaningful to examine changes in network structures over time from the onset of depression to the mitigation of depression-related behaviors and recovery. Second, because of the voluntary nature of this study, most of our participants were female, therefore, the results of this study may not generalize to the social networks of men who are clinically depressed (or not). Subsequent analyses could take a further look at the role of gender by recruiting more male participants. Moreover, because ResearchMatch.org operates on a voluntary basis, our sample may be subjected to the same self-selection bias. There is a possibility that the results of the study cannot be generalized to those who are severely depressed or socially isolated. Indeed, in our analysis, we found that depressed participants reported greater engagement in depression mitigation behaviors than those who were not depressed. While this is a positive sign, it can also indicate that our depressed participants from researchmatch.org could be more active than other depressed individuals. Furthermore, as one astute reviewer noted, researchers have not included a measure of the quality (as well as the quantity) of support from social network members and this should also be considered in future work. Research on individuals with depression may require more rigorous observational studies than the self-reported methods we have used in this study. Third, due to the scope of the current paper, we were not able to examine other variables that can play a role in mitigating depression behaviors. For example, our data did not enable us to further look into how economic backgrounds, religion, and ethnicity affect depression mitigation. There is a need for subsequent research to examine further the role of those variables at different levels of scale.