In line with the information related to the literature, coping is increasingly recognized as a necessary antecedent factor in the development of PTG and does not lead to PTG directly [32]. The “pathway” and “outcomes” of PC should be adequately understood in the context of the COVID-19 aftermath. According to the functional descriptive model of PTG [14], an essential element of the PTG process is the management of distressing emotions evoked by traumatic events, which may be produced by the adoption of adequate coping strategies [33].

Emotion regulation (ER) is a cognitive-behavioral process whereby individuals consciously and/or non-consciously adjust their internal affective states (e.g., magnitude, duration, and expression) to meet environmental demands appropriately, thereby generating adaptive responses [34]. Controlling efforts and the capacity for self-guide without consideration of incentives or encouragement are necessary for ER. To alter or sustain emotional state, ER includes a variety of cognitive and an attentional processes. John and Gross [35] categorized ER strategies into cognitive reappraisal (CR) and expressive suppression (ES). CR is a form of cognitive change by which individuals alter their focus from the negative aspects of trauma to the positive aspects. This process is an antecedent-focused strategy to induce ER through subjective attempts to reinterpret the plot, which happens at the preceding stage of emotion generation. On the contrary, as a response-focused strategy that takes place towards the end of the emotion-generating process, ES is mostly manifested in the conduct of the person who suppresses the expression of emotion to regulate emotion consciously [36]. Considering that these two commonly used ER strategies represent fundamentally distinct disengagement and engagement routes for managing distressing emotions, recent ER and PTG research suggests that only the CR regulatory preference profile has important functions in terms of enhancing individuals’ positivity levels and has downstream effects on trauma recovery [37,38]. The theoretical model of PTG proposes that as individuals reintegrate their disrupted core beliefs, they attempt to regulate their emotions in a way that facilitates constructive thinking and allows them to willingly engage with trauma-related memories and emotions [14]. To date, the nature of the possible relationship between the habitual use of PC and CR strategies has not yet been clearly established, while far less attention has been paid to the potential mediating role of CR between PC and PTG. Therefore, we propose the following hypothesis:

Another proximal factor that strongly affects the psychological growth outcomes of trauma-exposed individuals is psychological resilience (PR)—a process of good adaptation in the face of adversity; trauma; tragedy; threats; or other potentially negative repercussions of stressors such as family and relationship problems, serious health problems, and financial problems [39]. Generally, PR refers to a context-dependent “reconfiguration” after trauma exposure, which represents a multi-dimensional human ability to quickly return to pre-crisis status [40] and demonstrate improved physical fitness, reaction and coping skills despite going through traumatic events [41]. From a behavioral perspective, PR can be defined as a stable trajectory of healthy functioning following adverse events [42]. Similarly, it can also be viewed as a dynamic mechanism that mobilizes internal and external resources to strengthen the capacity to withstand setbacks and bounce back from adversity [43]. As proposed by the Rutter Model of Development, mental resilience functions via mitigating the devastating impact of risk factors, relieving collateral reactions to hazards in the environment, fostering self-esteem and self-efficacy, and offering opportunities and resources for individuals [44,45].

Within the limited literature on the resources that may contribute to combating the detrimental consequences and improving positive personal progress during the COVID-19 aftermath, coping strategies and PR are undoubtedly among the most studied [46]. Coping and resilience overlap but have different constructs concerning their effects on behavioral changes. The former refers to cognitive and behavioral strategies to handle the negative psychological outcomes arising from adversity [22], whereas the latter refers to the reduction of vulnerability to challenges and difficulties [47] and the improvement of adaptive capacity to recover [48]. Accordingly, utilizing effective coping strategies such as positive reinterpretation appears to be a means of resilience-building intervention. Such interventions could provide possibilities for developing adaptive responses with appropriate scaffolding and guidance, which is integral to realizing one’s growth [49]. Considering the recent focus of resilience research, some studies have observed beneficial relationships between PR and PTG [50,51]; however, the mediating effects of college students’ PR on the relationship between PC and PTG have not been thoroughly revealed. Based on previous findings, it can be inferred that PC is indirectly associated with PTG through PR. Thus, we propose the following hypothesis:

Evidence also seems to be mounting regarding the association of ER and PR, which concluded that a high-level ER is positively related to PR, and the abovementioned discrete ER strategies (CR and ES) are likewise related to PR [52,53]. In particular, the positive correlation between CR and PR has been supported in several studies, even after controlling for other explanatory variables [54–57]. CR is regarded as an adaptive strategy of positively reinterpreting a stressor to mitigate or resist its emotional impact [58], which may further facilitate accommodative psychosocial processes and redefine the meaning of experiences after stressful situations. Even so, there remains a dearth of in-depth studies on the serial mediating mechanism of “PC-PTG” concerning the essential roles of CR and PR. Therefore, we formulated the following hypothesis:

The cognitive process of attempting to reconcile conflicting assessments of a traumatic experience and fundamental beliefs is known as rumination, which is a steady, ongoing, recurrent metacognitive process [45]. According to the Response Styles Theory, rumination was originally conceptualized as a way of reacting to distress and is characterized by repetitive and passive thought about the cause and consequences, and significance of distress and related clues [59]. This notion may be comparable to the rearrangement of values and priorities to find a positive meaning in stressful events, and attempts to integrate the experiences into life narrative. Typically, rumination can be classified into two main types: intrusive rumination (IR) and deliberate rumination (DR) [60]. Each has been assigned a different role in the PTG process. Automatic IR is predominant and affects the persistence of tension and stress in the initial stage of post-traumatic re-adaptation [60]. For many people, IR is gradually replaced by a more reflective and controlled DR with time. DR is conceptualized as repetitive cognitive engagement that motivates individuals to have purposeful thoughts, consciously thinking about the traumatic experience and even about solutions and consequences—what happened and what it can mean [32]. As a kind of adaptive type of rumination, DR about events is engaged in voluntarily and consists of more intentional and controlled consideration focusing on making sense of the experience, problem-solving, reminiscence and anticipation even in the midst of the uncertainty [61]. Thus, the DR process arises from efforts to incorporate individuals’ traumatic experiences into their own cognitive model and reconstruct their understanding of themselves to a greater extent, which may trigger personal growth thereafter [62]. With this in mind, the intensity and persistence of DR have been witnessed as crucial determinants of PTG; that is, the more rumination participants tended to report experiencing soon after the event, the greater the degree of PTG [63,64]. Furthermore, coping is theoretically embedded between IR and DR [65,66]. The active utilization of self-sufficient coping strategies, such as positive reinterpretation, seems to be conducive to more positive cognitive engagement and behavioral adjustments, which may pave the way for fostering PTG. Thus, we can distill from the above evidence that DR generally acts as a mediator between PC and PTG, and propose the following hypothesis:

Considering the abovementioned classification, CR is an antecedent-focused strategy in which individuals reframe their belief of a situation with the objective of modifying its emotional influence [67]. The successful CR of negative experiences may help individuals apply psychological leverage to adapt to [68,69] and alleviate the harmful effects of stress on maladaptive consequences [70]. Previous research has proposed that rumination accounts for the relationship between multiple aspects of ER difficulties and post-traumatic stress disorder severity [71]. In this case, with a view to the positive psychological factors associated with recovery, it is possible that DR may also be a critical cognitive process linking specific aspects of ER to PTG. In addition to Hypothesis 5, the following hypothesis is proposed:

By verifying the variables that prompt the development of PTG within the context of the COVID-19 aftermath, public health efforts may be tailored accordingly to constructively foster positive psychological changes among college students. Although previous studies have focused on either the bivariate relationships between PC, ER, PR, and DR, or whether these processes alone are associated with or predict PTG, an integrated framework that includes multiple constructs can provide a platform to better depict broad research and theoretical synthesis. To fill this gap, the foremost initiative of this study was to explore the interconnections between the abovementioned factors and the mediating roles of CR, PR, and DR specifically and serially in the relationship between PC and PTG. According to a thorough review of the existing literature, the current study is the first to address these five concepts holistically, providing in-depth insight into the positive mechanisms underlying college students’ PTG in light of the COVID-19 pandemic. Fig. 1 shows a diagram of the mediation model proposed for the six hypotheses, which depicts the relationships between the independent (PC), mediator (ER, PR and DR), and dependent (PTG) variables.

The studies involving human participants were reviewed and approved by the Ethics Committee of Liaoning Normal University (No. LL2023005) on May 01, 2021. A non-probability convenience sampling method was used to recruit participants between May 18 and July 22, 2021. This kind of sampling means that researchers choose the sample as opposed to randomly selecting, which is commonly used in the field of mental health during COVID-19 [72–74]. It allows researchers to generate large samples in short time periods and requires minimal cost. Restricted by isolation policy and movement restrictions, the questionnaire hyperlink was sent to familiar teachers from several universities in mainland China via WeChat and email. The online survey was carried out with the assistance of course teachers by distributing an electronic QR code from the Wenjuanxing platform to full-time undergraduate students before or after their class. The questionnaire can only be submitted after all the items have been completed. Given that the participants were recruited by their teachers and the data was obtained through student self-reports, the socially desirable response bias should be carefully considered. With this in mind, all participants remained anonymous and voluntary. They were informed of the academic purpose of this scientific investigation prior to data collection, and their personal information was kept confidential to protect privacy during the study process. They were also told that their responses were neither good nor bad and unrelated to their academic performance or personal impression.

The present survey started with sociodemographic details of the participants, including sex, grade, major (medical, non-medical), residence, and only-child status. Further questions about their COVID-19 awareness were conducted through “How much do you care about your physical/mental health during this outbreak?”, “Recognition of the effectiveness of pandemic prevention and control measures in China” and “Assessment of the future pandemic situation in China”, based on a self-assessment 5-point Likert scale ranging from 1 (none) to 5 (very significant). Thereafter, participants were asked to complete a battery of psychological assessments. The psychometric instruments (CR, PR, DR, PTG, and PC) were developed using standard methods of translation, test, revision, and back-translation and were applicable to the Chinese group. All these scales were administered in the stated order. The progress bar provides an overview of the percentage of completed questions. The questionnaire was submitted only after completing all the items. Otherwise, the assessment system automatically recorded the data as null data. To ensure participants’ involvement, attention-stability-test items were intermittently inserted into the questions. Questionnaires completed in less than four minutes (decided by the pretest) were deemed invalid and deleted. The final sample comprised 463 students, predominantly females (n = 365; 78.83%). Most of the respondents were from 2nd and 3rd grade (78.2%).

Data processing was performed with IBM SPSS Statistics for Windows, Version 28.0 (SPSS; IBM, Armonk, NY, USA). The sociodemographic characteristics were all categorical variables and were reported as numbers and percentages. A common method variance test was performed using Harman’s single-factor test according to the cross-sectional nature of our research design, following the criteria of Podsakoff et al. [81]. Before starting the analysis, the assumptions about the normal distribution of the dataset were checked. The outputs of the Shapiro-Wilk test revealed that variables were skewed distributed (p < 0.05), except for PR. Descriptive statistics and bivariate correlation (r) analyses were conducted using SPSS 28.0. Considering the normal distribution results, the main numerical variables were represented by median (interquartile range), and relationships among these variables were computed by Spearman analyses. Meanwhile, non-parametric Mann-Whitney-Wilcoxon (MWW) test and t test were used to compare the differences in PC, CR, PR, DR and PTG among students in different groups of categorical variables (MWW for PC, CR, DR and PTG; T for PR). Finally, the serial multiple mediation hypothesis was tested using Preacher and Hayes’s method in PROCESS Macro using SPSS 3.4. Four models were fitted: regressing the first mediator (CR) on the predictor (PC), regressing the second mediator (PR) on the first mediator and predictor, regressing the third mediator (DR) on the first mediator and predictor, and regressing the outcome on all mediators and predictors. A non-parametric bootstrap approach was applied to test the total, direct, and indirect effects and calculate bias-corrected percentile confidence intervals (CIs) by constructing 5000 resamples with an original sample size of 463; each resample was constructed by sampling cases from data with replacements. The absence of zero in the 95% CI of the path coefficient indicates that the mediation effect was significant. All analytical tests were two-sided, and the alpha level was set at p < 0.05 to determine statistical significance.

Sample characteristics regarding the sociodemographic data are provided in Table 1. Excluding invalid questionnaires of 24 participants, the final sample size was 463 (effective recovery rate was 95.1%). To ensure an appropriate statistical power and effect size, a post hoc statistical power calculator (G*Power, version 3.1.9.7; Heinrich Heine University Düsseldorf) [82,83] was applied with a medium effect size (f² = 0.15) as well as a significance level of 0.05 [84,85]. Given the final sample size (n = 463) and number of predictors (4), we achieved an observed power (1-β) over 0.99, which is greater than the conventional adequacy standard of 0.80 [74].

Among the enrolled 463 respondents, 365 (78.83%) were female. The majority (78.2%) of the sample were 2nd and 3rd year undergraduates; 18.8% were first year undergraduates. Over all, approximately one in five of the sample (20.5%) were medical students. Almost half of the sample (44.5%) reported that they were from multiple-child families and over half (59.4%) lived in urban areas. Furthermore, regarding the awareness of the COVID-19, participants considered more about their physical and mental health during the epidemic (3.13 ± 1.05) and showed a positive attitude toward the situation (4.41 ± 0.85). Most of them approved of the local government’s prevention measures and were not worried about the future epidemic in China (1.70 ± 0.65).

The comparative analyses were conducted to explore the differences in PC, CR, PR, DR and PTG among students in different groups of the above categorical variables. No statistically significant difference was found in Grade, Major, and Residence. Only the students grouped by Gender and Family had a significant difference in PR. The PR of male students was significantly higher than that of female students (2.56 ± 0.51 vs. 2.42 ± 0.49, t = 2.52, p < 0.05). Also, the PR of students coming from only child families was significantly higher than that of students coming from multiple-child families (2.51 ± 0.48 vs. 2.38 ± 0.49, t = 2.79, p < 0.01).

First, an explanatory factor analysis (EFA) including all items using unrotated principal components factor analysis to reduce dimensionality was conducted to statistically measure the common method’s degree of variation. The results revealed that there are 13 factors with eigenvalues greater than 1, and the largest variance of any one factor was 24.36%, which did not exceed the critical value of 40% [81,86]. Thus, there was no evidence of serious CMV in the present study. That is to say, the variation between the dependent and independent variables was caused more by differences in the nature of the variables than by the data collection and measurement methods. Second, a full collinearity test was also performed in order to test the CMV among the present research instruments. Calculated based on the formula VIF_j = 1/(1-R²_j) (R²_j is the coefficient of determination, the measurement coefficient obtained by regression with other independent variables; variance inflation factor, VIF). As Kock [87] proposed, the occurrence of a VIF greater than 3.3 indicates pathological collinearity, which also means that CMV might contaminate the model. However, it has been found that all VIF values resulting from the full collinearity test turned out to be lower than 3.3 (ranging from 1.05 to 1.68). Altogether, the tests above concluded that this study is considered free of CMV.

Descriptive statistics and a correlation matrix of the study variables are displayed in Table 2. According to the normal distribution of the present data, the scores of PC, CR, PR, DR and PTG were described by first quartile (Q1), Median and third quartile (Q3). Moreover, non-parametric Spearman correlation analysis was undertaken to calculate relationships among the research variables. The bivariate correlation results were also given in Table 2 and showed that PC, CR, PR, DR, and PTG were positively correlated with each other (p < 0.05). These interrelationships result above were generally in line with our expectations, which met the prerequisites for further research hypotheses and conducting mediation testing.

For further testing the mediation effects, a serial multiple mediation model was performed in PROCESS (Model 81) for SPSS [88], using PC as a predictor, three mediators (CR, PR and DR), and PTG as the outcome. The conceptual model (see Fig. 1) is based on four linear regression analyses. The first regression analysis tests the effects of PC on CR (path a1). The second regression model tests the combined predictive effects of PC and CR on PR (paths a2 and c1). The third regression model tests the combined predictive effects of PC and CR on DR (paths a3 and c2). The fourth regression predicts the PTG by the independent variable PC and the three mediators (paths b1, b2, b3 and d). Here, path d depicts the direct effect of PC on the PTG controlled for the effects of the three mediators. In contrast, path d’ indicates the total effect of PC on PTG without considering the mediators.

Model indices are given in Table 3. In the path of a1 → b1, PC positively predicted CR (β = 0.72, p < 0.001). However, CR was not a significant predictor of PTG (β = 0.01, p > 0.05). In the path of a2 → b2, PC positively predicted PR (β = 0.51, p < 0.001), while PR positively predicted PTG (β = 0.60, p < 0.001). In the path of a3 → b3, PC positively predicted DR (β = 0.15, p < 0.05), while DR positively predicted PTG (β = 0.34, p < 0.001). In the path of a1 → c1 → b2, CR had a significant positive effect on PR (β = 0.12, p < 0.001), meanwhile, in the path of a1 → c2 → b3, CR had a significant positive effect on DR (β = 0.10, p < 0.01) (Fig. 2).

For the prediction of PTG (Table 4), PC was a statistically significant and positive predictor (b = 0.93, p < 0.001; 95% CI [0.76–1.09]) in the total effect model without consideration of the mediators (d’). However, the explained variance increased by ΔR² =0.1 when the mediators CR, PR and DR were included in the model. All possible indirect effects were significant (a2 → b2: b = 0.31, 95% CI [0.19–0.42]; a3 → b3: b = 0.05, 95% CI [0.01–0.11]; a1 → c1 → b2: b = 0.05, 95% CI [0.02–0.08]; a1 → c2 → b3: b = 0.03, 95% CI [0.01–0.05]) except path a1 → b1 (b = 0.01, 95% CI [−0.07–0.09]). Correspondingly, both the total indirect effect (b = 0.44, 95% CI [0.30–0.58]) and the direct effect of PC on PTG (d) were significant (b = 0.48, p < 0.001; 95% CI [0.28–0.69]). The effect size of the intermediary pathways was calculated using the formula (ab)/c, showed that path a2 → b2, a3 → b3, a1 → c1 → b2 and a1 → c2 → b3 accounted for 33.33, 5.37, 5.37 and 3.23% of the total effect, respectively. According to the above analysis results, the research hypotheses H1-H6 passed the test except H2. These indicated that the PC was closely related to students’ CR, PR, DR and PTG. PC was positively predictive of PTG in an indirect manner with PR and DR as parallel mediators. More importantly, PC could also enhance PR and DR by improving the adoption of CR, which finally facilitated students’ PTG.