Effects of sources of social support and resilience on the mental health of different age groups during the COVID-19 pandemic

Social support has been described as “support accessible to an individual through social ties to other individuals, groups, and the larger community” [3]. Although social support can be measured in many ways, perceived social support is the most commonly measured index of social support [4], given its ease of measurement and evidence that it is a better predictor of mental health than other measures [5, 6]. A wealth of research has demonstrated the protective effect of perceived social support on mental health in stressful situations [7‐9]. Besides, received social support (a receipt of supportive behaviors) is also an important sub-construct of social support [10]. Although the distinction between perceived social support and received social support exists, from the perspective of the stress and coping on social support [11], it is believed that the relationship between perceived and received social support should be relatively high, especially when the support demand matches the type of support provided [12]. Similarly, some authors suggest that perceived support can be assessed through the recall of supportive behaviors provided [13]. We believed that during the COVID-19 pandemic, an individual’s need for support matches the type of support provided, therefore, we measured social support individuals received to reflect the social support individual perceived.

It has been suggested that social support and social support resources should be viewed as related but distinct concepts [14]. Perceived support represents one’s subjective perceptions of the extent to which social network members are available to provide social support [15]. Several studies [16] have concluded that perceived quality of support is more strongly associated with mental health than with the actual structure of personal networks. Generally, perceived social support can come from a variety of sources, including, but not limited to, family, friends, romantic partners, pets, community ties, and coworkers. A wealth of research has demonstrated that different sources of social support have different influences on mental health in youth [17‐19]. Specifically, social support from family, but not from friends was related to posttraumatic stress disorder and depression symptoms [17]. And peer social support had a stronger protective effect against psychological distress than did family social support in another study [19]. Sources of support can be natural (e.g., family and friends) or more formal (e.g., mental health specialists or community organizations), and it is necessary to consider the different sources of social support when exploring the relationship between social support and mental health [20].

The different sources of social support will co-exist within persons, and combine to form different configural profiles. The traditional method, such as variable-centered methods, may overlook differences in social support among heterogeneous groups of individuals. Although there are many papers on sources of social support that have relied on the variable-centered methods. New methods (e.g. Latent profile analysis, LPA) have received growing interest in the typologies of social support networks in recent years (e.g. [21, 22]). What’s more, although based on slightly different concepts, some researchers have used cluster analysis to study social network typologies for a long time (e.g. [23]). More importantly, the existing papers demonstrated different support profiles would be differently associated with mental health outcomes [21, 22, 24]. Specifically, Burholt et al. [21] adopted LPA and found that the four social support profiles differ in their relationships to wellbeing in older migrants. The migrants with family-oriented profiles will experience lower levels of loneliness. Lau et al. [22] found that as compared with participants in the family-dependent social support profile, older adults in the locally integrated social support profile were negatively associated with dementia. McConnell et al. [24] adopted cluster analysis and explored how the combinations of different sources of support impact mental health among lesbian, gay, bisexual, and transgender youth. Researchers found that social support profiles showed different patterns of association with mental health outcomes. Individuals with the high support profile being the least lonely; youth low in all sources of social support are the most vulnerable for hopelessness and anxiety; social support from peers and significant others—even in the absence of family support—play an important protective role in hopelessness and anxiety [24].

Compared to the traditional clustering method, LPA is more reasonable to test the results of individual clustering by directly estimating the membership probability from the model ([25, 26]). Specifically, LPA is a person-centered method that identifies subgroups of individuals who share common social support patterns [27]. Besides, the outcome variables can be incorporated into the LPA model to build a regression mixture model to verify the relationship between profiles and outcomes. Hence, we conducted an LPA to explore the profiles of sources of social support and adopted the regression mixture model (BCH method) to explore the relationship between social support profiles and mental health in this study.

Unlike other crises or disasters (e.g., earthquake, economic crisis), a wider range of people will face a series of stressful events due to the pandemic itself and the related quarantine measures; therefore, the support provided both within the micro context (e.g., spouse, family, friends) and macro context (e.g., community, institutions/organizations, or even society as a whole) is vital to help ease psychological burdens. Accordingly, along with family support, it is also essential to consider the role of other sources of social support in mental health within a broader context. Furthermore, it is important to acknowledge individuals may experience different relative levels of these forms of support (e.g., high levels of support from community or even society as a whole, but low levels of family support) during the COVID-19 pandemic. However, little is known about how combinations of different sources of social support impact mental health during this period. Since individuals will face a series of stressful events during the COVID-19 pandemic, we infer that people with a high level of social support profile will experience a higher level of mental health than those with moderate or low social support profiles. Therefore, our first hypothesis is as follows:

We can often observe that, although many people may suffer from difficulties or crises, mental health outcomes still differ among individuals, and psychological resilience is assumed to play an important role in explaining such variance. Psychological resilience can be considered as either a trait or a process/outcome [28]. As a process, researchers referred to it as a dynamic process encompassing positive adaptation within the context of significant adversity [29]. As a trait, resilience represents a constellation of characteristics that enable individuals to adapt to the circumstances they encounter [30] and resilience has a positive impact on individual mental health [31]. A meta-analysis suggested that resilience may play a role in protecting mental health, accelerating recovery, and mitigating the negative effects of a crisis [32]. For instance, Tugade and Fredrickson [33] describe resilience as a quick and effective recovery after stress, while Patel and Goodman [34] conceptualize resilience as preserving mental health in the face of adversity. Therefore, as a trait, psychological resilience is a key resource in helping to prevent the negative impact of a crisis. And our second hypothesis is as follows:

As previously noted, sources of support and psychological resilience are two resources that protect individual mental health in stressful situations. Research shows that social support is key to resilience when it is considered to be a process/outcome (e.g., [35, 36]). However, when resilience is considered to be a trait, we cannot help asking, since there are individual differences in social support and psychological resilience, could abundance in perceived social support compensate for lack of psychological resilience? One study found that sources of social support (family, friends, and others) significantly moderated the relationship between resilience and subjective well-being in college students [8], while another study demonstrated that the relationship between resilience and psychological distress was not moderated by social support in patients diagnosed with cancer [37]. According to the stress-buffering hypothesis, researchers claimed that social support acted as a buffer to alleviate the negative influence of stress on well-being [38]. Scholars found that family support played a unique role in buffering the negative effects of stress by considering the different sources of support [39]. What’s more, Anschuetz [40] found that social support served to attenuate the negative effects of various trait vulnerabilities (such as neuroticism and introversion). Based on the stress-buffering hypothesis, we claim that social support from different sources can also buffer the negative effect of low levels of resilience (a trait vulnerability) on mental health, especially in a brand-new context, such as the COVID-19 pandemic. Since we adopt LPA to identify the profiles of the sources of social support, we extend the compensation hypothesis to social support profiles. Hence, our third hypothesis is as follows:

As a global public health emergency, the COVID-19 pandemic has caused great suffering to all people living in affected areas. When considering individual mental health as well as protective factors during the COVID-19 pandemic, age is a significant and unavoidable issue. First, many studies have shown that emotional well-being often increases across the entire adult life stage; thus, with age, people’s emotions and well-being tend to remain stable or even improve. Researchers refer to this phenomenon as the “paradox of aging” [41]. Moreover, according to the latest data reported in Nature Medicine, compared to those aged 30–59 years, those aged below 30 and above 59 years were 0.6 and 5.1 times more likely to die, respectively, after developing symptoms of COVID-19 [2] . Hence, older adults may experience more stress than other age groups. Therefore, the question remains as to whether the paradox of aging still exists within the context of the COVID-19 pandemic. However, according to the paradox of aging [41], we hypothesize that:

Second, according to socioemotional selectivity theory (SST), as individuals become more aware of time limitations in life as they age, they are more likely to spend time with emotionally close relationships than young people are [42‐45]. SST also suggests that, with age, individuals narrow the size of their social network and focus on a smaller circle of friends and relatives [42‐45]. Researchers found that older adults have a greater proportion of close social partners and fewer peripheral partners in their networks compared to younger adults [46, 47]. Within social support research, Potts [48] found that the older adults received different levels of social support from within and outside the retirement community because of the different levels of social network quantity between the within and outside the retirement community. Thus, for individuals of different ages, in terms of differences in their social networks, it is expected that the levels of sources of social support they received are different. Furthermore, a meta-analysis that systematically reviewed the characteristics of social support (types and sources) associated with protection from depression across life stages (e.g., childhood and adolescence, adulthood, older age), found that sources of support varied across life stages, with parental support being the most important among children and adolescents, whereas adults and older adults relied more on spouses, followed by family and then friends [49]. Hence, it is worth exploring whether when everyone is under a situation in which people cannot avoid being involved, such as the COVID-19 pandemic, if there are age differences in perceived sources of social support. Therefore, according to SST, we hypothesize that:

Third, research suggests that psychological resilience may vary according to age [50]. A survey based on a large cross-sectional Japanese sample demonstrated that psychological resilience could increase from the young adult stage to the older adult stage [51]. Furthermore, research showed that cancer patients with higher resilience, particularly older patients, experienced lower psychological distress, indicating that the relationship between resilience and psychological distress is moderated by age [37]. Research also demonstrated that age moderated the relationship between sources of social support and mental health, indicating that family social support is associated with lower symptoms of posttraumatic stress disorder and depression in participants aged 16–19 years, while friend social support is associated with lower symptoms for participants over 20 years old [17]. The possible role of age was examined in these analyses, especially the age differences in associations between resilience and mental health [37], and the moderating role of age between sources of support and mental health [17]. As previously noted, according to the stress-buffering hypothesis [38], people with a higher level of social support profile may buffer the negative effect of low levels of resilience on mental health. Few studies have examined whether the buffering effects of social support profiles vary by age, but the role of age has been associated with the variables of interest. What’s more, according to SST, older groups will have a greater proportion of high family-based support profile than the other age groups. Therefore, we hypothesize that the buffering effect between the high family-based support profile and resilience is greater in older adults:

The study protocol was approved by the institutional review board: Ethics Committee of Institute of Psychology, Chinese Academy of Sciences (reference number: H20016) and all respondents signed informed consent before participation. Data were collected from 23,192 participants via the First Affiliated Hospital of the University of Science and Technology of China online data platform, mostly from the Anhui Province (98.97%), from March 25 to April 1, 2020. Respondents’ ages ranged from 18 to 85 (M = 41.58, SD = 14.63), and 7437 (32.1%) were male. Given that the data reported in Nature Medicine, the fatality rate is different among different age groups (below 30, 30–59, above 60) [2], the population was stratified into different age groups in this study. Furthermore, emerging adult, which a focus on ages 18–25, differs from adolescence and young adult according to Arnett’s study [52]. We used age as a categorical variable rather than a continuous variable and the three groups were divided by 25 and 60 years (18–25, 25–59, above 60). The three age groups were emerging adults (2045 participants, 31.10% male, range = 18–25), adults (18,159 participants, 29.50% male, range = 26–59), and older adults (2988 participants, 48.5% male, range = 60–85).

SPSS version 21.0 was used for descriptive statistics, one-way analysis of variance (ANOVA), and chi-square tests.

LPA with a continuous dependent variable was performed using Mplus version 8.0, and the algorithm adopted was the BCH method [58]. First, LPA was used to identify the number and nature of classes of COVID-19-related support in all samples, with the scores of the five items that the support and assistance participants received from different sources as extrinsic variables, respectively. A series of models with progressively increasing numbers of classes from 2 to 6 were estimated and compared to determine the optimal model solution based on lower Akaike information criterion (AIC) values, lower Bayesian information criterion (BIC) values, significant Lo–Mendell–Rubin likelihood ratio test (LMR-LRT), bootstrap likelihood ratio test (BLRT) p-values, considerable class sizes (at least 5% of the sample), and higher entropy values [59]. To be specific, we based on the rules that the model with the lowest BIC and AIC index offered the best-fit indicator, and the significant test of LMR and BLRT showed that the k + 1 profile was superior to the k-profile. Moreover, we take the entropy into account, the higher the entropy, the more accurate the classification. More importantly, we followed the advice that if the additional profile adds a qualitatively new profile to the prior solution, the new profile might be retained. On the contrary, if an additional profile is only minor level differences in all profile variables compared to the prior solution, the new profile might not be retained due to reasons of parsimony [25, 60]. Second, for the outcomes, we used the BCH command, in which analyses reflect whether one profile is statistically significantly different from other profiles per criterion. Meanwhile, testing mean differences across the profile groups to mental health outcomes is also a validation of the classification results according to Daniel [26]. Third, the interaction between social support and resilience on mental health across different groups during the COVID-19 pandemic was assessed using the PROCESS version 3.4 macro of SPSS version 21.0, developed by Hayes [61]. In this approach, moderating effects were tested when 95% bias-corrected bootstrap confidence intervals based on 5000 samples did not contain zero.

This study has several important theoretical implications. First, to our knowledge, this is the first study to use LPA to examine patterns of sources of social support among three different age groups during the COVID-19 pandemic. Moreover, we identified five latent profiles of sources of social support: low social support (class 1), predominantly proximal support (class 2), predominantly remote support (class 3), moderate social support (class 4), and high social support (class 5). Second, the distribution of emerging adults and adults in class 2 was significantly lower than that of older adults, and there was a decline in perceived social support across age groups, except in supports from family. Therefore, these findings provide quantitative evidence for SST [42‐45]. Additionally, our results also offer nuanced evidence on how social support interacts with psychological resilience and further affects mental health. Thus, high levels of social support can buffer against the negative effects of low levels of resilience on mental health.

The current findings carry essential practical implications. First, according to our results, individuals should build a social support system for coping with the COVID-19 pandemic. People should keep social distancing, but not be socially isolated. During the pandemic, special attention should be paid to maintaining social connections and increasing perceived social support. Furthermore, family support is vital for all age groups, and especially for adults. The more difficulties there are to face, the more we need to provide increased social support to our family members. Individuals need to make full use of various social support resources to counteract the negative impact of the pandemic on mental health. Second, although the latent profile of sources of social support in different age groups was basically the same, there were differences in distribution, and sources of social support affected the mental health of different age groups in different patterns. Older adults perceived relatively lower social support, and only a moderate to a high level of perceived social support was found to be necessary to protect their mental health, thus reminding us that we should pay special attention to providing more social support to older adults during the pandemic. Finally, our findings suggested that high levels of resilience are necessary for young, middle-aged, and older adults to “bounce back” from the pandemic because the buffering effect of social support for different age groups was the same. When resilience is lacking, only high levels of social support resources from all aspects can compensate for the negative effects of low resilience on mental health. Therefore, it is an effective way for individuals to improve their psychological resilience against mental health risks brought on by the pandemic.

There are several limitations to this study. First, although our data were collected during the COVID-19 pandemic (data collection took place from March 25 to April 1, 2020), a month after the response to the major public health emergency had been lowered from level 1 to level 2 in Anhui Province. Therefore, the results may not reflect the most severe state of the sample location. Second, to reduce the difficulty of completing the questionnaires, we simplified our measurements, with each dimension of social support reflected in a single item. For instance, family support did not distinguish between support from a spouse and support from other family members, which has been found to have different effects on individual mental health in previous studies [68]. Also, we asked about support in general in the social support measurements, and a specific type of support (e.g. instrumental, emotional) should be examined in future studies to provide a more comprehensive picture of how different social support profiles affect mental health [69]. Finally, as this study had a cross-sectional design, we considered psychological resilience to be a trait indicator. However, existing studies have found that resilience may be influenced by social support as a process indicator [70]. Therefore, based on the cross-sectional data, this study was not able to reveal this relationship.