The use of multilevel emotion regulation strategies in the context of critical public events: the more the better?

Recent years have witnessed several critical public events, which always significantly impact individuals' emotional and mental health. COVID-19, as a typical public health emergency, has brought people an increase in negative experiences including fear, anxiety, and loneliness, and a decrease in positive emotions and well-being during the pandemic. Considering that similar critical events may arise in the future, COVID-19 may provide a good opportunity for us to investigate the protective factors that can enhance mental health in these challenging circumstances. Prior studies have demonstrated that emotion regulation is crucial for maintaining psychological well-being. However, whether emotion regulation promotes psychological adjustment to critical public events depends on flexible strategy use. To better adjust to critical public events, individuals should consider contextual demands and select appropriate strategies. The more the chosen strategy adapts to the situation, the more effectively it can help individuals achieve their emotion regulation goals, resulting in higher utility. Moreover, possessing the capability to select and adjust strategies according to the circumstances is indicative of one's psychological well-being. Research indicates that psychological flexibility is a fundamental aspect of promoting healthy personal and social functioning, while inflexibility is associated with an increased risk for psychopathology such as depression and anxiety. Therefore, the relationships between mental health and ER strategies are likely to be bidirectional. The first aim of this study is to re-examine the relationship between ER strategies and mental health during the pandemic. One component of ER flexibility is an individual's ER repertoire, the range of different ER strategies used across situations. Emotions are experienced and regulated at multiple levels: intrapersonal (e.g., cognitive reappraisal, experiential avoidance), interpersonal (e.g., enhancing positive affect, perspective-taking, soothing, social modeling), and hyper-personal (digital/social media-based regulation). Each level has advantages and disadvantages, and social media has introduced new hyper-personal strategies (e.g., humor via memes) that can be protective but also risky due to emotional contagion. Few studies have simultaneously examined all three levels. Therefore, we collected two waves of longitudinal data, conducted cross-lagged analysis to examine relationships between strategies and mental health, and used cluster analysis to identify profiles of emotion regulation strategies. Mental health was measured by emotions (positive/negative), loneliness, and life satisfaction.

The introduction situates emotion regulation (ER) as central to psychological well-being (John and Gross, 2004; Aldao et al., 2010) and emphasizes regulatory flexibility (Bonanno and Burton, 2013; Chen and Bonanno, 2021). Intrapersonal ER during uncontrollable situations often involves cognitive reappraisal (adaptive) and experiential avoidance (risk-prone) (Tamir et al., 2007; Troy et al., 2013; Akbari et al., 2022). Interpersonal ER provides social support via enhancing positive affect, perspective taking, soothing, and social modeling (Hofmann et al., 2016), though overreliance may compromise autonomy. Hyper-personal ER via digital technologies includes posting/retweeting emotional content (Smith et al., 2022), with potential for anxiety via online emotional contagion (Prikhidko et al., 2020). Humor and memes on social media spread rapidly during COVID-19 and can upregulate positive affect (Skorka et al., 2022; Berger et al., 2021; Rime et al., 2020). Prior work notes people use multiple ER strategies in daily life, but few studies integrate hyper-personal strategies with intra- and interpersonal strategies simultaneously (Altan-Atalay and Ray-Yol, 2021; Shao et al., 2021; Rottweiler et al., 2023).

Design: Two-wave longitudinal online survey during the COVID-19 epidemic with a 1-week interval (control measures remained the same). Informed consent obtained; confidentiality assured; participation voluntary. Matching across waves used personal identifiers (name, phone, social media accounts, Alipay for participation fees). Ethical approval: Sichuan Normal University. Sample: Initial N=1,189; longitudinal completers N=859. At Time 1 (T1), 71% in low-risk areas (45% male); at Time 2 (T2), 78% in low-risk areas (46% male). Participants who only completed T1 did not differ significantly from longitudinal participants on demographics or study variables. Measures: Emotions via Chinese PANAS (18 items; 9 positive, 9 negative), 5-point Likert; reliability: Positive Affect α_T1=0.937; α_T2=0.952; Negative Affect α_T1=0.897; α_T2=0.916. Life satisfaction via Chinese SWLS (5 items, 7-point Likert; α_T1=0.92; α_T2=0.92). Loneliness via Chinese short UCLA Loneliness Scale (6 items, 4-point Likert; α_T1=0.876; α_T2=0.877). Intrapersonal ER: Cognitive reappraisal via Chinese ER Scale (6 items, 7-point Likert; α_T1=0.822; α_T2=0.839). Experiential avoidance via AAQ-II (7 items, 7-point Likert; α_T1=0.914; α_T2=0.924). Interpersonal ER: Chinese IERQ (20 items, 5-point Likert; α_T1=0.921; α_T2=0.934) with subscales enhancing positive affect, perspective taking, soothing, social modeling. Hyper-personal ER: willingness to retweet humorous memes and saving behavior. Meme selection: 104 memes sourced online; 170 participants rated pandemic relevance and fun; 10 memes selected for formal study. Willingness to retweet measured on 1–5 scale (α_T1=0.884; α_T2=0.918). Saving behavior measured as yes/no (α_T1=0.865; α_T2=0.893). Measurement model: CFA in Mplus 8.3 indicated good fit for all scales across waves (χ²/df < 5.5, CFI > 0.90, RMSEA < 0.08, SRMR < 0.08). Analytic approach: Cross-lagged structural equation modeling in Mplus 8.3 (ML estimation), with fit indices (χ²/df, RMSEA, CFI, TLI, SRMR); controlled for gender, age, and risk level. Model fit: χ²/df=4.79, p<0.001, TLI=0.911, CFI=0.972, RMSEA=0.065 (95% CI [0.059, 0.072]), SRMR=0.069. Common method bias: Harman’s single-factor test indicated first factor explained 22.16% (T1) and 26.6% (T2) variance (<40%). Cluster analysis: Model-based clustering using Mclust (Gaussian finite mixture, orientation matrix shared), selecting number of components and covariance parameterization via BIC; control variable effects regressed out before clustering. Nine clusters identified (BIC=-30418.75; ICL=-30559.66); standardized deviations from strategy means computed per profile to aid interpretation.

- Descriptive change over time (paired t-tests): Positive emotion increased from M_T1=2.95 (SD=1.03) to M_T2=3.18 (SD=1.06), t=-8.616, p<0.001. Life satisfaction increased from M_T1=22.19 (SD=7.30) to M_T2=22.94 (SD=7.00), t=-4.655, p<0.001. Negative emotion decreased from M_T1=2.00 (SD=0.82) to M_T2=1.93 (SD=0.83), t=3.286, p=0.001. Loneliness decreased from M_T1=11.81 (SD=3.95) to M_T2=11.59 (SD=3.79), t=2.269, p=0.024. Experiential avoidance and retweeting willingness decreased; perspective taking, social modeling, saving behavior increased; cognitive reappraisal, enhancing positive affect, soothing showed no significant change. - Correlations: All strategies except experiential avoidance were positively correlated with positive emotion and life satisfaction, and negatively with negative emotion and loneliness. Experiential avoidance showed the opposite pattern. - Cross-lagged relationships (controlling gender, age, risk level): Cognitive reappraisal (T1) -> Positive emotion (T2): β=0.098, SE=0.025, z=3.865, p<0.001. Experiential avoidance (T1) -> Negative emotion (T2): β=0.162, SE=0.029, z=5.558, p<0.001; -> Loneliness (T2): β=0.141, SE=0.028, z=5.082, p<0.001; -> Positive emotion (T2): β=-0.095, SE=0.025, z=-3.798, p<0.001; -> Life satisfaction (T2): β=-0.153, SE=0.023, z=-6.779, p<0.001. Perspective-taking (T1) -> Life satisfaction (T2): β=0.054, SE=0.027, z=1.993, p=0.046. Saving humorous memes (T1) -> Positive emotion (T2): β=0.067, SE=0.027, z=2.474, p=0.013; -> Loneliness (T2): β=-0.092, SE=0.035, z=-2.606, p=0.009. Psychological states predicting strategy use: Negative emotion (T1) -> Experiential avoidance (T2): β=0.114, p<0.001; -> Social modeling (T2): β=0.073, p=0.008. Positive emotion (T1) -> Experiential avoidance (T2): β=-0.047, p=0.044; -> Perspective-taking (T2): β=0.071, p=0.048; -> Soothing (T2): β=0.114, p=0.001; -> Willingness to retweet (T2): β=0.068, p=0.021; -> Saving behavior (T2): β=0.109, p<0.001. Life satisfaction (T1) -> Enhancing positive affect (T2): β=0.120, p<0.001; -> Perspective-taking (T2): β=0.106, p=0.005; -> Social modeling (T2): β=0.094, p=0.010; -> Willingness to retweet (T2): β=0.063, p=0.034. Loneliness (T1) -> Enhancing positive affect (T2): β=-0.056, p=0.043; -> Soothing (T2): β=-0.060, p=0.037; -> Social modeling (T2): β=-0.110, p<0.001. - Cluster profiles (9 groups): Low user (22.5%, n=201): below-mean use of all strategies except experiential avoidance. Low hyper (11.4%, n=102): below-mean humorous meme saving; others near mean. Average user (17.4%, n=156): all strategies near mean. High hyper (5.9%, n=53): hyper-personal strategies above mean; others near mean. Independent (2.3%, n=21): intrapersonal above mean; others below mean except social modeling near mean. Low inter (2.9%, n=26): interpersonal strategies below mean; others near mean. Three Multi-user groups: Multi-user low (18.4%, n=165), medium (13%, n=116), high (6.1%, n=55): all strategies except experiential avoidance above mean, to varying degrees. - Mental health differences by profile: Significant group effects on emotions and mental health at both T1 and T2 (ps < 0.001). Multi-user-high exhibited highest positive emotion and life satisfaction and lowest negative emotion and loneliness; Low inter had highest negative emotion; Low user and Low hyper showed poorer psychological states. Multi-user-high differed from other multi-user groups by lower experiential avoidance and higher soothing and humorous meme saving. - Overall: Experiential avoidance is a longitudinal risk factor for negative affect and loneliness; perspective-taking and humorous meme saving are protective, associated with higher life satisfaction and positive emotion. Using multilevel strategies, especially high use across levels with low experiential avoidance, is linked to better mental health during critical events.

Findings support the central hypothesis that multilevel emotion regulation, and especially regulatory flexibility reflected in diverse repertoires, buffers mental health during critical public events like COVID-19. Individuals who balance intrapersonal, interpersonal, and hyper-personal strategies show higher positive affect and life satisfaction and lower negative affect and loneliness, with the multi-user-high profile performing best. Intrapersonal-heavy users (independent profile) showed lower positive emotion and life satisfaction and higher loneliness, suggesting that relying predominantly on intrapersonal strategies may down-regulate negative affect but coincides with social disconnection. Hyper-personal strategies, particularly humor via meme saving and retweeting, appear protective for positive affect and social connection, despite potential risks of negative contagion. Cross-lagged results clarify bidirectional links: experiential avoidance both predicts and is predicted by negative states, reinforcing its role as a maladaptive strategy; perspective-taking enhances life satisfaction, likely via social connection and cognitive reframing; humorous-meme-saving boosts positive emotion and reduces loneliness, highlighting the utility of positive digital content. Together, these results underscore that different levels of ER offer complementary resources; balanced multilevel use expands regulatory options and improves overall effectiveness.

Emotion regulation strategies such as perspective-taking and humorous-meme-saving act as effective buffers for mental health during critical public events, increasing positive emotions and life satisfaction. Experiential avoidance functions as a risk factor, being positively associated with negative emotions and later loneliness. Profiles characterized by high use across intrapersonal, interpersonal, and hyper-personal strategies combined with low experiential avoidance show the most favorable psychological outcomes. These results identify protective multilevel ER profiles and confirm that ER flexibility is vital for maintaining well-being in the face of societal crises.

Participants were from China during the dynamic zero-COVID policy, with many residing in low-risk areas, potentially limiting generalizability. Despite the two-wave design, causal inferences remain limited and strict measurement invariance was challenging due to model complexity. All measures were self-report; hyper-personal ER strategies are not yet standardized in emotion research. Emotion regulation is dynamic; the study’s timeframe (1-week interval) may not fully capture temporal processes. Future work should employ longitudinal designs with more waves, experience sampling methods, and objective measures, and consider technologies like virtual reality to enhance ecological validity.