key: cord-0912624-r3s7fya6
authors: Xu, R.; Zhang, X.; Liu, D.; Li, Q.; Wang, Y.; Jiao, R.; Gong, X.; Hou, X.; Xu, T.; Qing, X.; Song, K.; Kavcic, V.; Yan, S.; Gu, R.; Stratton, T.; Jiang, Y.
title: Coping with COVID-19 Stressors: Adverse and Protective Factors Responding to Emotions in a Chinese Sample
date: 2021-11-30
journal: nan
DOI: 10.1101/2021.11.30.21250895
sha: c67dd5eec6f9bc47b0e6742f3bbaf84c37c18f17
doc_id: 912624
cord_uid: r3s7fya6

Background: The potential roles of affective responses to environmental stressors in individuals' physical and mental health are complex and multi-faceted. This study, then, explores Chinese citizens' emotional responses to COVID-19-related stressors and influence factors which may boost or buffer such effects. Methods: From late March to early June (2020), a cross-sectional study was conducted using an anonymous online questionnaire included demographic characteristics, COVID-19-related stressors related to individuals' daily functioning, and the self-assessed impact of protective and adverse internal factors on emotions. Results: 1,662 questionnaires were received from residents in 32 Chinese provinces classified by prevalence level according to COVID-19 infections. Among the 17 positive and negative emotional responses, agglomerative hierarchical clustering revealed four subclassifications: (1) stress relations; (2) missing someone relations; (3) individual relations; and (4) social relations. Additionally, heightened regional prevalence levels positively corresponded to intensity of stress relations. Lowest intensity of social relations was found in the areas surrounding Wuhan and coastal areas. Specially, economic- and work-related stressors as well as negative self-perceptions (e.g., suppression, emotionally unstable, self-denial) implicated in negative emotions. While positive emotions were tied to demographic characteristics (e.g., high education, young age and male) and protective traits (e.g., creativity, sympathy, social responsibility), and inversely linked to relationships- and pandemic-related stressors, etc. Conclusion: Associations were clearly noted among Chinese residents' emotions to specific stressors during pandemic. Providing appropriate psychological resources/supports during future or extended public health crises may help offset the cognitive burden of individuals striving to regain an adequate level of normalcy and emotional well-being.

Amadin, & Omoregie, 2021; Gupta & Sengupta, 2020; Shah, Quint, Nwaru, & Sheikh, 1 2021). The negative impact of the pandemic on world economics is already evident, 2 as the risk of global economic crisis has been accumulating (Bank, 2021). Under these 3 influences, the mental health of the public is of great concern during the pandemic 4 (Alfawaz et al., 2021; Shanahan et al., 2020). 5 When individuals are exposed to traumatic events, their emotional responses would 6 be modulated, which may further lead to irrational and impulsive behavior (Ceschi, 7 Billieux, Hearn, Furst, & Van der Linden, 2014; Scott & Montgomery, 1984) . Indeed, 8 public fear-driven behaviors during the COVID-19 pandemic led to increases in doctor 9 visits, pressures on the healthcare system, hoarding of food and daily necessities, and 10 misuse of personal protective gear (PPG), etc. (Mahase, 2020; Oosterhoff & Palmer, 11 2020). Reports of irrational behaviors (e.g., delays in seeking necessary medical 12 attention) and violence against doctors and vulnerable groups increased significantly 13 during the restrictions (Gaballa, AlJaf, Patel, Lindsay, & Hlaing, 2020; Ghosh, 2018). 14 While much attention has been paid to complex emotional responses (e.g., anxiety 15 and depression) to COVID-19 (Barzilay et al., 2020), basic emotions (e.g., happiness, 16 sadness, fear, etc.) remain largely ignored. However, the spontaneous expression of 17 basic emotions is more typical in "routine" social life and, subsequently, vulnerable 18 during disruptive or anomic conditions, such as natural disaster (Y. Li et al., 2020). For 19 instance, citizens in Croatia frequently exhibited fear, discouragement, and sadness 20 during ten days of the COVID-19 lockdown (Dogas et al., 2020). Additionally, the 21 public's basic emotional expression has become a sensitive early warning indicator of 22 infectious outbreaks such as measles, H1N1, and Ebola(Ahmed, Bath, Sbaffi, & 23 during the first half of 2020 (Huang et al., 2020) -and infective prevalence and double-1 edged policies (e.g., lockdowns) subjected residents to complicated or contradictory 2 emotional states (Jin et al., 2020; Y. Li et al., 2020). In fact, residents were forced to 3 overcome difficulties prior to encountering social isolation and conflict (Singh & Subedi, 4 2020; Venkatesh, 2020; Williams, Armitage, Tampe, & Dienes, 2020). Throughout all 5 this, the relationship of basic emotions and environmental stressors was not fully 6 studied. Additionally, little information was available to assess perceived feelings 7 during pandemic, although cognition was recognized to play a key role between such 8 environmental insults and emotions. Thus, our study aimed to investigate the basic 9 emotions and the links with self-perceived stressors, all of which facilitate governments 10 to provide timely and effective responses to the pandemic. The cross-sectional study utilized a non-random "snowball" sample designed to elicit 15 maximum participation by consenting Chinese adults. We used an online platform 16 of "Wenjuanxing" (https://www.wjx.cn/) and advertised by WeChat, a popular social 17 networking platform. From late March to early June (2020), respondents to an 18 anonymous online questionnaire came from 32 Chinese provincial districts -which 19 were classified by prevalence level (1 = lowest, 5 = highest) according to cumulative 20 COVID-19 infections compiled prior to the data collection deadline (see Appendix 1). 21 and (3) protective and adverse aspects of well-being (e.g., creative, facing death from 1

In particular, the instrument gauged respondents' experiences with seventeen basic frightened -whereas anger could be directed toward oneself or others (Praill, 10 Gonzalez-Prendes, & Kernsmith, 2015). Additionally, thinking (si) might encompass 11 annoyance and distracted thinking (Du, 2000) . Conversely, in Chinese culture, joy (xi) 12 represents positive emotions such as happiness/joy, comfort/relaxation, a sense of 13 accomplishment, and safety (Dictionary Editing Room, 2011). All emotions could result 14 in response to a pandemic, which were assessed on a 4-point Likert-type scale ranging 15 from 0 ("Not at all") to 3 ("A great deal"). 16 Environmental stressors assessed during the COVID- 19 Berger, 2001). The correlation matrix is then recalculated and the maximum coefficient 4 again identified between the new and uncombined clusters. This is repeated until all 5 items are grouped together into a single cluster or the procedure exceeds the specified 6 number of iterations. 7

A major challenge to classifying multiple objects within a singular construct is a large 9 intra-class variation that typically exists. The LASSO approach to parameter estimation 10 alters the model-fitting process to select only a subset of covariates which produce an 11 optimally predictive and interpretable model (Tibshirani, 2011) .This goal is achieved by 12 utilizing a regularization process which, when applied across all predictors in a 13 multivariate model, allows the contributions of some coefficients to be reduced to zero 14 and, hence, completely removed from the model. This advantage, among others, 15 distinguishes LASSO from similar techniques (e.g., ridge regression) designed to avoid 16 model over-fitting by reducing residual "noise" 17 (http://wavedatalab.github.io/machinelearningwithr/post4.html). 18 As a parallel to designating measures as dependent or independent variables, LASSO 19 uses machine learning (ML) vernacular to reference inputs ("features") and outputs 20 ("labels"). More specifically, "features" are properties of the data used to "train" or 21 develop the model-fitting algorithm, while "labels" are the resulting output returned after 22 computing. 23 certain emotion cluster. Normalizing the feature variables to obtain a standard sample 1 set, each standard sample had a feature vector (X) and a label variable (Y) upon which 2 the LASSO regression model was built. The number of features ultimately included in 3 the final model is selected by a cross-validation method and the parameters obtained. 4 We chose 20% of the overall data as the test set used to gauge performance of the 5 selected model. 6 A linear regression that models the response variable y using a set of features, With multiple potential predictors, a smaller, more parsimonious subset of "features" is 22 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted November 30, 2021. ; https://doi.org/10.1101/2021.11.30.21250895 doi: medRxiv preprint sought which exhibits the strongest effects. Toward this end, OLS estimates are 1 suboptimal to regularization approaches which impose "penalties" on certain 2 coefficients to avoid model overfitting. As mentioned above, LASSO regression shrinks 3 some coefficients and sets others to 0 -obtaining the optimal solution of the following where || || = ∑ | =1 |, the first part represents the goodness of the model fitting, 7 and the second part represents the parameter penalty. Here, λ ≥ 0 is a complexity 8 parameter that controls the amount of shrinkage as coefficients are reduced or set to 9 zero: the larger the value of λ, the greater the amount of shrinkage. The smaller the 10 regularization parameter , the less punitive the model and the more features it retains. 11

Conversely, increases in results in reduced numbers of features. 12

Additionally, Kruskal-Wallis ANOVA were adopted to explore the differences of each 13 emotional cluster (i.e., positive, negative; stress relations, missing sb. relations, 14 individual relations, social relations) among five regional prevalence levels of COVID- 15 19 infective (1 = lowest, 5 = highest). Post hoc comparisons were also conducted. 16 A critical p value of < .05 was set for all inferential tests, and analyses were conducted 17 using Python (https://www.python.org/) and R software platforms 18 (https://www.python.org/). 19

Nineteen (19) of 1,681 the questionnaires were excluded due to incomplete responses or participants not living in China -yielding a final, analyzable sample of 1,662. The 23 "typical" respondent was female (60.5%, n = 1,006) and 30.7 years of age (SD = 10.3). 24 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted November 30, 2021. ; https://doi.org/10.1101/2021.11.30.21250895 doi: medRxiv preprint accounted for virtually one-half of the sample. Most respondents (63.2%) had 2 bachelor's degrees or above, and the prevalence rate of COVID-19 infection across 3 the five regions ranged from ~13.4 to 28.0%. Sample demographics are summarized 4

in Table 1 -along with reported emotional reactions and influential factors (e.g., 5 situational stressors, protective/adverse internal factors). Descriptive statistics of the 6 latter are presented below. 7

Seventeen (17) emotional responses to COVID-19 were divided into two categories 9 reflecting positive and negative emotions and, subsequently, four nested 10 subclassifications: For negative emotions, the emotion "missing someone who is close 11 to you" was distanced from those related to stress (e.g., sadness, fear/fright, anger 12 toward oneself or others, worry, annoyance and distracted thinking due to anger). 13 Positive emotions, in contrast, were classified as individual-(e.g., happiness, 14 relaxation, comfort, safety and accomplishment) or social-relations (e.g., emotionally 

A cross-validation process intended to find the optimal value ( ) -designating the 25 most parsimonious model with a minimum average MSE (see Figure 2A ). As 26 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted November 30, 2021. ; https://doi.org/10.1101/2021.11.30.21250895 doi: medRxiv preprint increases, more coefficients are set to zero -leading to a sparser model (see Figure  1 2B). 2 Corresponding coefficients for the emotional models with all relevant predictors are 3 reported in Figure 3 . Results showed that the predictors of negative emotions included 4 a tendency to suppress emotions, emotional instability, self-denial, vulnerability to 5 impact, economic problems, fatigue/sleepiness, work-related problems, unequal 6 treatment, feeling limited/constrained, lacking confidence etc. In contrast, positive 7 emotions were driven by creativity, sympathy, social responsibility, high education etc., 8 while factors, such as tendency to suppress emotions, age, intimate partner 9 relationship, social isolation and facing death from COVID-19, had the opposite effect. 10

Significant differences in emotional clusters among residents from regions with varying 12 levels of COVID-19 infective prevalence were found (see Table 2 ), with pairwise, post 13 hoc comparisons (see Table 3 ) showing positive emotions in the highest prevalence 14 region were lower than the medium or lowest prevalence regions. Significant study also showed that Chinese residents' emotional expression via Weibo (a Chinese 6 version of Twitter) was strongest before the actual peak of COVID-19 outbreak and 7 declined thereafter (Y. Li et al., 2020). Additionally, our study revealed strength of 8 negative emotional response was sensitive to gradient regional prevalence levels by 9 investigated data. Consistently, significant correlations were found between regional 10 cumulative COVID-19 cases and Web searches by Italian netizens including the 11 generic terms "fear" and "anxiety" (Rovetta & Castaldo, 2020). However, missing sb. The oldest conjectures was confirmed that happiness depends not just on absolute 24 things but inherently on comparisons with other people (Clark & Oswald, 2002) . 25

Evidence of comparison effects was also revealed in our results, that was high intensity 26 of positive emotions related with experiencing fewer and not serious negative incidents, 27 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in Positive emotional state and self-perceptions played protective roles for residents in 10 encounter of difficulties. As previous studies confirmed, positive emotions serving as 11 the ultimate target or a moderator, helped people to cope with negative events (Waugh, 12 2020). Similarity, positive self-perceptions associated with high self-esteem could 13 minimize influence of negative information (Showers, 1992) . It was worth to notice that 14 creativity made a great contribution to positive emotional response to the COVID- 19. 15 To tackle the unanticipated difficulties and intricate problems during pandemic, 16 interaction between creativity and positive feelings may play an important role to factors also highlighted some self-assessed characteristics which were related with 20 social members (e.g., sympathy and socially responsible). Consistently, 21 conscientiousness and openness were found as a protective factor for sadness, 22 depression, stress and tension caused by stress (Schlee et al., 2020). Higher 23 sympathy associated with higher prosocial behavior, which was also found to be the 24 positive link between sadness regulation and prosocial behavior and to mediate by 25 higher sympathy and trust (Song, Colasante, & Malti, 2018) . Lastly, high education, 26 young age and male contributed to positive emotional state, which was proved to 27 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The limitation of the study is that most responders in our sample haven't experience 3 traumatic events, such as confirmed COVID-19 infection. Thus, they reported lots of 4 maladjustment for changes of routine life. Those who and whose family members have 5 fallen illness due to virus infections, may show more intensity of negative emotions and 6 different stressors and influence factors. In addition, we did not ask participants to 7 report their quality of life, thus the potential influence of emotion on subjective well-8 being has not been fully understood. 9

Negative emotional response was sensitive to regional prevalence levels of COVID-11 19. Economic and work-related problems largely contributed to negative emotions and 12 relationship-and epidemic-related stressors largely affected positive emotions. 13 Negative self-images (i.e., emotional expression, self-denial, emotionally unstable, 14 lacking confidence) were adverse factors for emotions. Protective factors for emotions 15 included creativity, social-orient traits, as well as high education, young age and male. 16 Future study should investigate emotional supports/measures based on interactions 17 among particular stressors and cognitive burden, as well as protective factors. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted November 30, 2021. maximum coefficient method. showed stressors whose coefficients weren't zero. Adverse internal factors 3.54 (3.5) Note: SD, Standard Deviation. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted November 30, 2021. ; https://doi.org/10.1101/2021.11.30.21250895 doi: medRxiv preprint All rights reserved. No reuse allowed without permission.

perpetuity.

preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in

The copyright holder for this this version posted November 30, 2021. ; https://doi.org/10.1101/2021.11.30.21250895 doi: medRxiv preprint

Relationship quality and mental health during 1 COVID-19 lockdown

An Exploration of the Relationships 3 between Attitudes Towards Anger Expression and Personal Style of Anger Expression in

Women in the USA and Canada

Different Crises, Different Patterns of Trauma. The 8 Impact of a Social Crisis and the COVID-19 Health Pandemic on a High Violence Area

The Impact of COVID-19 on Italian Web Users: A Quantitative 11 Analysis of Regional Hygiene Interest and Emotional Response

The 14 Effect of Environmental Stressors on Tinnitus: A Prospective Longitudinal Study on the 15 Impact of the COVID-19 Pandemic

Urban violence: pandemic of trauma

Impact of COVID-19 national lockdown 19 on asthma exacerbations: interrupted time-series analysis of English primary care data. Emotional distress in young adults during the COVID-19 pandemic: evidence of risk and 1 resilience from a longitudinal cohort study

Psychological Protection Handbook for Primary and 4

Secondary School Students and Parents During the Anti-epidemic Period

Compartmentalization of positive and negative self-knowledge: keeping bad 7 apples out of the bunch

COVID-19 and stigma: Social discrimination towards frontline 10 healthcare providers and COVID-19 recovered patients in Nepal

Helping yourself helps others: Linking children's 13 emotion regulation to prosocial behavior through sympathy and trust

The Use of Public Health Indicators to Assess Individual Happiness

Global Economic Prospects

Regression shrinkage and selection via the lasso: a retrospective

Working in a pandemic: Exploring the 2 impact of COVID-19 health anxiety on work, family, and health outcomes

Impacts of COVID-19: A research agenda to support people in their fight

A creative-bonding intervention and a 7 friendly visit approach to promote nursing students' self-transcendence and positive 8 attitudes toward elders: a pilot study

Keep it to yourself? Parent 11 emotion suppression influences physiological linkage and interaction behavior

The roles of positive emotion in the regulation of emotional responses to 14 negative events

Coronavirus (COVID-19) Dashboard

Public perceptions and experiences 17 of social distancing and social isolation during the COVID-19 pandemic: a UK-based 18 focus group study

Job Insecurity 20 and Financial Concern During the COVID-19 Pandemic Are Associated With Worse

Handbook of Public Psychological Protection for COVID-19

An East Meets West Approach to the 4 Understanding of Emotion Dysregulation in Depression: From Perspective to Scientific

The Neural Basis of Fear Promotes Anger 7 and Sadness Counteracts Anger

Geographical distance to the epicenter of Covid-19 10 predicts the burnout of the working population: Ripple effect or typhoon eye effect? 11

The Neural Basis of Fear Promotes Anger 13 and Sadness Counteracts Anger

Anti-epidemic and Peace of Mind--A Comprehensive Reader for 15 Psychological Self-help Rescue of the Epidemic