The relationship between perceived stress and prolonged grief disorder among Chinese Shidu parents: effects of anxiety and social support

As a negative emotion, anxiety is triggered by perceived threats that may come from the environment (e. g. social occasions or new locations) or from within oneself (e. g. unusual physical sensations) [31]. Although perceived stress and anxiety are both emotional responses with many similarities between symptoms, perceived stress is caused by an existing external stress-causing factor or “stressor”, when the stressor disappears, so does perceived stress; whereas individuals with anxiety continue to feel pressure even after the stressor is gone. According to the biobehavioral attachment-based model of prolonged grief proposed by Shear and Shair [32], the attachment system established in infancy is activated under stress for regulating emotions to cope with stress and establish personal security. However, after the death of the attachment figure, great bereavement-related stress activates the attachment system, and individuals with insecure attachment style may experience anxiety symptoms due to their inability to update old attachment schemas and form new attachment relationships in time [32]. The anxiety symptoms are focused on the unease related to the loss of the important refuge and secure base provided by the attachment figure, as well as the confrontation with death. This painful emotion further activates the attachment system and inhibits engagement in social life, and then the distress escalates, interfering with and complicating the grieving process [32]. Some previous studies in other populations have explored the relationship between perceived stress and anxiety. The study among Chinese college students supported that perceived stress could significantly positively predict anxiety [33]. Li et al. [34] also reported higher levels of perceived stress was strongly connected with increased risk of anxiety in breast cancer patients. Regarding to anxiety and PGD, previous studies have reported anxiety was significantly correlated with PGD [18, 35]. The study in bereaved caregivers by Aoun and colleagues [16] displayed when participants had anxiety, the risk of PGD was 8 times higher than those without anxiety. Thus, we propose a hypothesis that anxiety mediate the relation between perceived stress and PGD.

Social support refers to emotional, informational and instrumental assistance provided to the individual by primary and secondary group members [36]. The buffering model indicates that when people face stress, social support plays a buffering role through people’s cognition to reduce the negative effects of stress on them and protect their physical and mental health [37]. Cao et al. found that both perceived social support and objective social support could reduce anxiety among Chinese Shidu parents [5]. Results of a study by Razurel et al. [38] showed during the prenatal period, parturient received esteem and information support from their mother as well as information support from their friends, which could moderate the association between stress and prenatal anxiety. Zou et al. [39] found the relation between perceived stress and anxiety was moderated by social support in Chinese frontline COVID-19 medical staff, specifically, among individuals with low social support, higher perceived stress was related to more anxiety. Furthermore, researchers have reported perceived social support moderated the link between attachment anxiety and health outcomes, and less anxious individuals’ health benefited from high support [40]. An empirical study also revealed social support could moderate the association between anxiety and depression in survivors of Wenchuan Earthquake [41]. Social support may also help individuals cope with PGD. Song et al. [42] found PGD symptoms of Shidu parents were negatively correlated with social support. The study in caregivers of people with motor neurone disease revealed social support was a critical predictive factor of PGD, and increasing social support could reduce psychological distress [16]. These findings coincide with the biobehavioral attachment-based model of prolonged grief emphasizing that social support from others and the social environment can compensate for the bereaved's temporary withdrawal from interpersonal interactions [32]. A good social support system can help the bereaved maintain a sense of control, stop avoidance, and gently encourage them to think about future plans and adapt to life without the deceased [32]. Therefore, we hypothesize social support can moderate the direct and indirect pathways of the mediating model.

To sum up, the current study pursued three goals: 1) to explore whether perceived stress significantly and positively associated with PGD, 2) to determine whether anxiety mediate the relationship between perceived stress and PGD, 3) to examine whether the direct and the indirect effects of perceived stress on PGD are moderated by social support (see Fig. 1). This is the first investigation to examine the mediating role of anxiety and the moderating role of social support in the relation between perceived stress and PGD among Chinese Shidu parents. We hope this study will help the government and primary health care institutions formulate targeted interventions to improve mental health of Chinese Shidu parents in the future.

The cross-sectional study was conducted in Shenyang, a city in northeast China, from March 2017 to September 2017. We selected five urban districts in Shenyang and randomly collected 15 communities in each district. All Shidu parents who met the following inclusion criteria in these communities were recruited as participants: 1) lived in the community for over half a year; 2) lost the only child and had no biological or adopted another child at the time of the survey; 3) could communicate in Chinese clearly. The exclusion criteria was having severe mental illness (e.g. mental retardation, Alzheimer’s disease). We used the following formula to calculate the sample size: \(n=\frac{{Z}_{1-a/2}^{2}p\left(1-p\right)}{{d}^{2}}\). The prevalence of PGD in bereaved parents ranged from 10% to 30% around the world [20‐24], we chose 20% as the expected prevalence. The sampling error allowed in this study was d = 0.05, and 95% CIs was adopted, so \(Z_{1-a/2}\) = 1.96. We calculated the minimum sample size: n = [1.96² × 20% × (1-20%)]/(0.05)² ≈ 246. Due to the particularity of the loss, we assumed the response rate of the subjects was 70%, which required the sample size: n = 246/70% ≈ 352. Community family planning commissioners, who had access to all Shidu families in the community, registered participants who met the inclusion criteria. The commissioners explained the purpose of this survey and invited Shidu parents who were willing to join in the survey to the community office to fill out self-report questionnaires. Participants provided written informed content before the investigation, and interviewers explained the purpose and significance of the investigation. Community family planning commissioners could assist in providing clarification if participants had any questions. When subjects had trouble in completing questionnaires by themselves, face-to-face interviews were used to obtain the answers. Before data collection, in order to make sure a standardized process, all community family planning commissioners were trained for two days. After the investigation, subjects were paid 100RMB as compensation. In total, a sample of 595 subjects was surveyed. Of these, 79 refused to join in. Among the 516 subjects, 11 were not included in the analysis because more than 30% of the data were missing. As a result, 505 parents remained for analysis, and effective response rate was 84.87%. GPower was used to conduct power analysis of the sample size, when α = 0.05, effect size = 0.3, the power of the sample size was 0.99. There were no significant demographic differences between the 505 included participants and the nonparticipants. The missing data were lost completely at random.

The study was approved by the Ethics Committee of China Medical University, and all the research procedures were conducted in accordance with ethical standards.

Demographic characteristics consisted of parents’ gender, age, marital status, education level, employment, chronic disease, annual household income, debt, religious belief, marital satisfaction. Bereavement-related information included gender and age of the deceased child, time since the child’s death, cause of death and whether the participant had a grandchild. If participants had been diagnosed with any of chronic diseases (e. g. hypertension, diabetes, cardiovascular disease, chronic hepatitis, ulcer, gout and arthritis), chronic disease was recorded as ‘yes’. Debt, religious belief and whether the participant had a grandchild were recorded as ‘yes’ or ‘no’. Cause of death was divided as ‘accident’, ‘acute disease’, ‘chronic disease’ and ‘mental disease’.

PGD symptoms were measured by the Prolonged Grief Questionnair-13 (PG-13), a screening instrument to diagnose PGD according to the criteria defined by Prigerson [10]. It consists of 13 items addressing separation distress (1-2); cognitive, emotional and behavioral symptoms (3-11); duration criteria (12); and impairment criteria (13). Item 1 to 11 are rated on a 5-point Likert scale (1 = never, 5 = several times a day/overwhelmingly), and items 12 and 13 are yes/no questions. The diagnostic criteria for PGD include the following: 1) the individual has lost a significant other; 2) separation distress: at least one item with a rating of 4 or higher; 3) cognitive, emotional and behavioral symptoms: at least five items with a rating of 4 or higher; 4) duration criteria: item 12 must receive ‘yes’ response that means separation distress should last for more than 6 months after the death; 5) impairment criteria: item 13 must be answered as ‘yes’ that means being defective in social, occupational and other crucial aspects [10]. The Chinese version of PG-13 has been proved to have satisfactory psychometric properties [9]. In the investigation, the internal consistency coefficient for the scale was 0.91.

Perceived stress was assessed by the Perceived Stress Scale-10 (PSS-10), which evaluated the degree of stress that a participant experienced over the past month [43]. It consists of four positive items (4, 5, 7 and 8) and six negative items (1, 2, 3, 6, 9 and 10). Each item is rated on a 5-point Likert scale, ranging from 0 (never) to 4 (all of the time). The total scores range from 0 to 40, and higher scores indicate more stress. This scale has shown adequate validity and reliability in Chinese population [44]. Cronbach’s alpha was 0.86 in this study.

Anxiety symptoms were assessed by the 20-item Self-Rating Anxiety Scale (SAS) [45]. Items are scored on a 4-point Likert scale ranging from 1 (rarely or none of the time) to 4 (most or all of the time), according to the frequency of anxiety symptoms during the past week. Five items (5, 9, 13, 17 and 20) are reverse scored. The total scores of this scale range from 20 to 80, and scores of 50 or more is considered as the presence of anxiety symptoms. The higher the total scores, the more severe anxiety symptoms. The Chinese version has shown good validity and reliability [46]. Cronbach’s alpha was 0.91 in the current study.

We used the Duke-UNC Functional Social Support Questionnaire (FSSQ) to evaluate social support, which is an 8-item questionnaire [47]. Each item is rated on a 5-point Likert scale, ranging from 1 (much less than I would like) to 5 (as much as I would like). The scale provides total scores between 8 and 40 points, and the higher the total scores, the greater social support. The scale has shown sufficient validity and reliability in Chinese population [48]. The internal consistency of FSSQ in the study was 0.87.

All the analyses were performed with IBM SPSS Statistics Version 23.0. Descriptive data were presented as frequency, percentages, mean scores and standard deviations. Independent t-test or one-way ANOVA was performed to compare the PGD scores across demographic characteristics and bereavement-related information. Pearson’s correlation coefficient was used to explore the associations between perceived stress, anxiety, social support and PGD. Multiple imputation was used to replace the missing data. Then we employed the SPSS PROCESS macro to test the mediation model and the moderated mediation model [49]. First of all, Model 4 was used to determine the mediation role of anxiety in the relation between perceived stress and PGD. Next, Model 59 was adopted to explore whether social support moderated the direct and the indirect effects of perceived stress on PGD. Finally, the significance of the moderation effect was examined by a simple-slope test which used the PROCESS macro. We kept chronic disease, annual household income, debt, religious belief, marital satisfaction and cause of death as potential confounders in the whole analysis, and categorical variables were converted into dummy variables. All the predictors were standardized to avoid multi-collinearity effects. In the study, on the basis of 5000 bootstrap resamples of these data, 95% bias-corrected confidence intervals (CIs) of the effect were computed. If 95% CIs don’t contain 0 that indicate significant effects. A two-sided p < 0.05 was regarded as statistically significant.

Since the present study collected data through self-reporting measurements, we employed Harman’s single-factor test to examine common method biases [50]. The results showed that eight factors’ eigenvalues were greater than one and the first factor could explain 34.30% of the variability in the data, which was less than the critical threshold of 40%, indicating that there was no serious common method bias.

Table 1 shows the demographic characteristics and bereavement-related information and results of univariate analysis. The mean age of the subjects was 61.06 ± 6.72 years (ranging from 25 to 84). The sample comprised 204 (40.40%) males and 301 (59.60%) females. The mean total score of PG-13 was 26.26 ± 9.67 (ranging from 11-55). Of the 505 participants, 9.11% (n = 46) met the criteria for PGD symptoms. In this study, results from the univariate analysis presented that chronic disease, annual household income, debt, religious belief, marital satisfaction and cause of death were significantly associated with PGD symptoms (p < 0.05), and these variables were kept as potential confounders in the mediation and the moderated mediation analyses.

Correlations among PGD, perceived stress, anxiety and social support are presented in Table 2. Perceived stress and anxiety were positively related to PGD symptoms (r = 0.62, p < 0.01; r = 0.64, p < 0.01). Social support was negatively associated with perceived stress, anxiety and PGD symptom (r = -0.47, p < 0.01; r = -0.43, p < 0.01; r = -0.44, p < 0.01). In addition, perceived stress had a significant positive relation to anxiety (r = 0.62, p < 0.01).

As shown in Table 3, after controlling for chronic disease, annual household income, debt, religious belief, marital satisfactory and cause of death, the direct and the total effects of perceived stress on PGD were significant (β = 0.62, p < 0.01; β = 1.02, p < 0.01), which proved perceived stress significantly and positively correlated with PGD. The indirect effect of perceived stress on PGD via anxiety was also significant (β = 0.40, p < 0.01), and the Bootstrap 95% CI for the indirect effect of anxiety was [0.29, 0.52] that didn’t include 0. The mediation effect accounted for 39.22% of the total effect of the relationship between perceived stress and PGD. Thus, anxiety was proved to partially mediate the relationship between perceived stress and PGD symptoms.

The PROCESS macro (model 59) by Hayes [49] was used to test the moderated mediation model. After controlling for chronic disease, annual household income, debt, religious belief, marital satisfactory and cause of death, Table 4 shows the interaction effect of anxiety and social support was negatively correlated with PGD (β = -0.01, p = 0.04), indicating social support moderated the pathway of “anxiety-PGD”. Nevertheless, the interaction effect of perceived stress and social support was insignificantly correlated with anxiety (β = 0.01, p = 0.15) and PGD (β < 0.01, p = 0.72), indicating social support did not moderate the pathway of “perceived stress-anxiety” and the direct pathway of “perceived stress-PGD”. To better explain the moderating effect of social support, we divided social support into three different levels (low, 1 SD below mean; moderate, mean; high, 1 SD above mean). Anxiety significantly mediated the link between perceived stress and PGD when social support was low (β = 0.36, 95%CI: 0.24, 0.49), moderate (β = 0.32, 95%CI: 0.22, 0.45) and high (β = 0.27, 95%CI: 0.13, 0.47). Figure 2 illustrates the plot of the link between anxiety and PGD at two levels of social support (1 SD below mean, 1SD above mean). The results indicated that anxiety was strongly associated with PGD for participants with low social support (β = 0.36, SE = 0.04, t = 8.52, p < 0.01), while the association was weaker for participants with high social support (β = 0.23, SE = 0.05, t = 4.57, p < 0.01).