key: cord-0924819-qf7mvot6
authors: Choi, Incheol; Kim, Joo Hyun; Kim, Namhee; Choi, Eunsoo; Choi, Jongan; Suk, Hye Won; Na, Jinkyung
title: How COVID-19 affected mental well-being: An 11- week trajectories of daily well-being of Koreans amidst COVID-19 by age, gender and region
date: 2021-04-23
journal: PLoS One
DOI: 10.1371/journal.pone.0250252
sha: be68c0057fdf398751e853ad4cbb1e8ead60ef22
doc_id: 924819
cord_uid: qf7mvot6

The present study examined the daily well-being of Koreans (n = 353,340) for 11 weeks during the COVID-19 pandemic (January 20 –April 7). We analyzed whether and how life satisfaction, positive affect, negative affect, and life meaning changed during the outbreak. First, we found that the well-being of Koreans changed daily in a cubic fashion, such that it declined and recovered during the early phase but declined substantially during the later phase (after COVID- 19 was declared world pandemic by WHO). Second, unlike other emotions, boredom displayed a distinctive pattern of linear increase, especially for younger people, suggesting that boredom might be, in part, responsible for their inability to comply with social distancing recommendations. Third, the well-being of older people and males changed less compared to younger people and females. Finally, daily well-being dropped significantly more in the hard-hit regions than in other regions. Implications and limitations are discussed.

The beginning of the year 2020 was marked by the global pandemic of COVID-19. Since the first pneumonia case of unknown cause was reported in Wuhan, China, on November 17, 2019 [1] , the virus has been rapidly spreading worldwide with tragic effects. Within six months, more than 10 million people around the world were infected, and about 500,000 people have died [2] . No cure or vaccine for the virus was invented at the time of writing this article. The influence of the COVID-19 pandemic is not limited to physical health. Nearly every aspect of human life has been affected, including the economy, travel, politics, religion, education, and international relations. Individuals' social lives have also been substantially compromised by the outbreak, as social contacts have been strongly discouraged or even prohibited. All these unexpected and unprecedented life changes and uncertainties lead to a growing alert right after the patient 31 was confirmed (February 23). Since March, the number of daily confirmed cases in Korea decreased dramatically, which led many health experts and media over the world to claim that Korea successfully flattened the curve [17] . In stark contrast, the global situation got worse, forcing WHO to declare COVID-19 as a pandemic on March 11. In response to the worsening global situation, the Korean government ordered a nationwide intensive social distancing on March 22 . Thus, the current analysis could show not only people's initial psychological responses to the COVID-19 epidemic but also their reaction and (or) adaptation to the recovery of the local situation and the continuing/worsening global situation.

Individuals ranging in age from 14 to 71 years (N = 353,340, M age = 30.70, SD age = 12.34) completed an online well-being survey from Kakao company during the study period (January 20 to April 07). Overall, 79% of respondents were female, and 36.8% were in their 20s ( Table 1) . The survey was performed on an online survey platform launched by Kakao cooperation, one of the largest Internet companies in South Korea. Any Kakao app users can voluntarily participate in the survey via smartphone or computers (http://together.kakao.com/hello), and they can respond to each survey at any time and across multiple occasions. During the study period, on average, 4,472 respondents participated in the survey per day, and a total of 490,014 responses were used in the analysis. The proportion of respondents across different regions reflected the regional distribution of the South Korean population. The privacy policy of the company restricted the collection of participants' demographic information to age, gender, and region of residence. Conducting secondary data analysis of the Kakao survey was approved by the Kangwon National University Institutional Review Board (#201910009002). Measures Well-being. A 10-item questionnaire was used to measure well-being. The items were adapted from well-established measures of subjective well-being (life satisfaction, positive affects and negative affects) and eudaimonic well-being (life meaning) [18] . Previous literature well addressed that the items selected are valid to represent various aspects of well-being [11] [12] [13] . In addition, the 10-item index was also validated among Korean populations [19] .

Life satisfaction and life meaning were measured using a single item each, asking about the extent to which respondents were satisfied with their lives and felt that it was meaningful. For affective well-being, participants responded to 8 questions, each question assessing "How much are you feeling each emotion at this moment?" anchored by three positive affects (PA: happy, joyful, and relaxed) and five negative affects (NA: bored, annoyed, anxious, depressed, stress) [19, 20] . Participants responded to each of the well-being indicators on an 11-point scale ranging from 0 to 10 (e.g., 0 = not satisfied, 10 = very satisfied; see S1 Table for the actual questionnaire). The well-being index was created by averaging scores from 10 items (α = .913). Composite scores for PA and NA were also computed by averaging across three or five emotions, respectively, with high internal consistency (α = .875 for positive affect, α = .886 for negative affect).

Demographic variables. Gender, birth year, and residential place were self-reported. The age was estimated from the reported birth year, and individuals were divided into three age groups, young (10s-20s), middle (30s-40s), and old (50s-70s), in the main analysis.

Considering the large sample size, we adopted a more conservative approach by performing tests at a significance level of α = .001, as in Stone et al. [21] .

Polynomial models were used to examine the well-being trajectories over 79 days. More specifically, we fitted polynomial models in the multilevel modeling framework using the lme4 package in R [22] , considering the nested structure of the data. About 26 percent of participants answered the survey more than once during the study period. The demographic characteristics, such as age, gender, and region of residence, were included in the model to examine individual differences in the trajectories.

Ten-fold cross-validation [23] was used to select the optimal degree of a polynomial for each well-being measure. For the 10-fold cross-validation, we split the data into 10 roughly equal-sized subsets. For each measure, we fitted polynomial models with varying degrees (from 0 to 6) to nine subsets of the data. We obtained the prediction error of the estimated model by calculating the squared difference between the observed score and the predicted score for the fresh data that was set aside in estimation. This process was repeated 10 times by using each subset of the data that was set aside to calculate the prediction error. Finally, the mean predictor error was calculated for the entire data. For each well-being measure, we examined the mean predictor errors across different degrees of polynomials. Due to the large sample size, the prediction error kept decreasing as the degree of polynomial increased. To avoid overfitting, we chose the degree at an elbow point, where the prediction errors started to stabilize, as the optimal one for further analyses. Based on the cross-validation results, we chose a linear model for boredom and joy and a cubic model for all the other well-being measures (S1 Fig) . Using the respective model, we analyzed not only the overall well-being index but also each subcomponent of the index. In the following section, we first report the results based on the well-being index and then describe additional results from subcomponents we thought were interesting as well as informative. We also note that detailed results regarding all subcomponents of the well-being index are reported in the Supporting information.

Our first question was whether/how the daily well-being of Koreans changed during the COVID-19 outbreak. As can be seen in Fig 1, overall well-being displayed the decline-recovery-decline pattern, hence a cubic pattern (S2 Table) . A close look at Fig 1 shows that the overall well-being declined up to February 10, then gradually ascended to the first week of March, and steeply declined after that. Recall that WHO declared COVID-19 a pandemic on March 11 and the Korean governement implemented the intensive social distancing policy on March 22. Interestingly, the initial decline in well-being occurred well before the situation in Korea worsened, consistent with the past research showing that well-being starts to decline before the critical event, such as divorce and unemployment [24] . It is even more intriguing that the wellbeing of Koreans declined very steeply after the first week of March, during which the situation in Korea started to improve. We discuss later why this might have been so. The same cubic pattern of well-being change emerged for life satisfaction, PA, NA, and life meaning (Fig 1 and S2  Table) .

To examine whether the change patterns varied by different emotions, we analyzed each emotion measure separately. As Fig 2 shows, all emotions displayed the identical cubic pattern of change except for boredom and joy (see S3 Table) . Specifically, boredom continued to increase, whereas joy continued to decrease during the study period. All the other emotions showed a brief period of recovery in the middle, yet boredom and joy did not. The distinctive pattern of boredom and joy seems very important because it may, at least partially, account for the difficulty people feel in observing social distancing [25] . Social distancing has been promoted as one of the most effective preventive measures against the spread of COVID-19 by health experts. Yet, many people do not seem to follow the advice, potentially costing the lives of others and their own [26] . We speculate that the high level of boredom and lack of joy may make social distancing difficult. Past research has shown that people who are ill-equipped to tolerate boredom willingly inflict electric shock to themselves to fight boredom [27] .

The unique pattern of boredom and joy raises the question of who might be suffering from boredom and joylessness most. The frequent media reports [28] , as well as our observations of young people failing at social distancing, imply that young people suffered from boredom and joylessness most. To answer the question, we tested whether the model with interactions between the polynomials and the age groups yielded a significantly better fit compared to the baseline model without the interactions. The results showed that the model fit significantly increased after adding the interaction terms for boredom and joy, χ 2 (2) = 197.75, p < .001 for boredom, χ 2 (2) = 135.65, p < .001 for joy (S4 Table) . As can be seen in 

As stated earlier, a comparison between the city of Daegu (and its adjacent areas, Gyeongbuk province) with the rest of Korea offers a rare opportunity for a natural experiment. Note that about two-thirds of patients in Korea during this period were residents of Daegu/Gyeongbuk. First, we examined the mean differences in well-being between Daegu/Gyeongbuk and other regions during the study period controlling for age and gender. As can be seen in S7 Table, Daegu/Gyeongbuk scored lower compared to the other regions in most well-being measures. For example, the estimated regional trajectories in Fig 4 show that the Daegu/Gyeongbuk residents tended to manifest a lower level of overall well-being throughout the study period, except for the first week. Second, we further examined whether the changing pattern differed between the two regional groups by comparing an interaction model with a baseline model without interaction. For the overall well-being index, the interaction model did not significantly improve the model fit compared to the baseline model, χ 2 (3) = 7.318, p = .062, suggesting that the shape of the trajectory did not differ between the two regions (S8 and S9 Tables). However, in the case of boredom, a model with the interaction was better compared to the baseline 

model without the interaction, χ 2 (1) = 11.464, p = .001 (S8 and S10 Tables). As can be seen in Fig 4, the boredom level in Daegu/Gyeongbuk region (b = .818, SE = .046, p < .001) increased more steeply compared to the other regions (b = .654, SE = .014, p < .001). This is not surprising given that social distancing was most strongly enforced in Daegu/Gyeongbuk. Our regional contrast suggests that the changes in the daily well-being of Koreans during the study period were partially, if not entirely, due to COVID-19.

As virus, COVID-19 is more fatal to people older than 60 years of age compared to younger people [29, 30] . Accordingly, are psychological perils also greater for older than for younger people? Alternatively, since older people are more emotionally stable and better equipped to deal with existential threats [31, 32] , do older people suffer less mentally?

To answer these questions, we compared three age groups: Young (10s & 20s), Middle (30s & 40s), and Older (over 50s). First, as shown in S7 Table, the well-being index showed a Ushape. Overall, it was the highest among older adults and lowest among middle-aged adults, which is consistent with the baseline well-being differences among Koreans across different age groups [9] . Next, we tested whether the trajectories of well-being measures varied across different age groups by comparing model fit, as in the previous section. The results showed that the model fit significantly increased after adding the interaction term for the well-being index, χ 2 (6) = 195.35, p < .001 (S4 Table) . As Fig 5 indicates , overall, old people showed the highest level of well-being, which was followed by young people, and then middle-aged people. All the age groups displayed a pattern of decline-recovery-decline. However, toward the end of the study period, young and middle-aged groups reached a decreased well-being of similar levels, yielding an increased gap from old people. The same pattern emerged for life satisfaction, NA, and life meaning (Fig 6 and S12 Table) , except that middle-aged people showed a higher level of life meaning than young people. For PA, the pattern appeared to be similar as other well-being indices, but it was not statistically significant.

As reported earlier, age differences were also observed for specific emotions. In general, the gap between old people and middle-aged/young people tended to increase toward the end of the study period as illustrated in the stress trajectories. (Fig 7 and S5 Table) . 

How COVID-19 affected mental well-being It is intriguing that despite greater biological vulnerability to COVID-19, older people were more resilient than younger people in terms of emotional well-being. The socioemotional selectivity theory [33] posits that with aging, people voluntarily narrow their social network and limit their activities to those they enjoy; hence, social distancing might be a more natural process for them. Therefore, the social lives of older people might have been less disrupted by COVID-19. Alternatively, older people might simply be better able to cope with the fear of death compared to their younger counterparts. Future research should uncover the exact mechanism(s) why older people's mental well-being is less affected by COVID-19 than that of younger people.

Whose well-being was affected more by COVID-19, males and females? First, as shown in S7 Table, the well-being of women was lower compared to that of men during the study period across all the measures. However, the gender differences were consistently observed before COVID-19 [9, 10] , suggesting that they may reflect the baseline gender differences in wellbeing among Koreans, rather than differences related to COVID-19. Hence, to further explore COVID-19 related gender differences, we tested whether changes in daily well-being would show different patterns for men and women. As in previous sections, we compared an interaction model with a baseline model without interactions. Gender difference in change patterns was significant for all measures (S13 Table) . As can be seen in Fig 8, women tended to experience a greater amount of overall change than men despite some fluctuations in the middle. This was true for most of the specific emotions (see Fig 8 for annoyance and stress as examples). Interestingly, however, the slope for boredom and joy was steeper for men than for women (Boredom: b = .769, SE = .030, p < .001 for male vs. b = .644, SE = .015, p < .001 for In other words, on average, men experienced a larger change in boredom and joy compared to women, unlike other well-being indices (see S15 and S16 Tables for details).

Our dataset did not distinguish homemakers from working moms, singles from married individuals, and families with children from those without children. Hence, more fine-grained analyses concerning gender differences in well-being during the outbreak should be conducted in future studies.

This study is one of the first large-scale observational studies that examined the real-time selfreports of different aspects of well-being in response to living during an unprecedented epidemic of COVID-19. We examined trajectories of well-being during the first 11 weeks of the COVID-19 outbreak in South Korea. Enough data points recorded each day during the pandemic allowed us to capture the change patterns of various aspects of well-being as the pathogen threat unfolded over time.

Our data revealed several interesting as well as important observations. First, COVID-19 significantly affected the well-being of Koreans. The levels of life satisfaction, life meaning, positive affect, and negative affect systematically changed during the study period. Well-being steadily decreased since the first national health alert. Interestingly, for most well-being measures, the decrease was already evident right after the 1 st health alert (i.e., when the first confirmed case was announced), reaching the lowest point before the surge of the widespread COVID-19 in Korea. Although well-being bounced back early in March, the recovery did not last but dropped even more substantially from around the time when WHO declared a pandemic and the social distancing was implemented, showing cubic patterns overall. Notably, well-being was significantly lower in Daegu/Gyeongbuk area, the hardest-hit region in Korea, compared to other regions, suggesting that the changes in well-being during the study period were due, at least partially, to COVID-19. Second, the decline in well-being was much steeper at the end of March. This is very intriguing because in late March and into April, the situation in Korea improved dramatically. The number of daily new patients dropped below 10 during this period. Several reasons could have contributed to this surprising decline. It may reflect psychological reactions to the social consequences of COVID-19, specifically social distancing. The Korean government announced the first social distancing campaign on March 2 and advised the highest level of social distancing on March 22. A major decline in social life during this period might have contributed to a significant decrease in well-being. Or the continuous exposure to news covering COVID-19 pandemic in other countries may have increased fear and uncertainty regarding COVID-19, which in turn may have aggravated Koreans' well-being. Or reduced income and fear of unemployment started to affect Koreans' well-being. We speculated that the decline was due to such multiple causes.

Third, patterns of trajectories varied depending on emotions. Unlike other emotions, boredom and joy linearly increased and decreased, respectively. Interestingly, the linear changes in boredom and joy were most evident among younger people (vs. middle-aged and older adults) and males (vs. females). Similar to our findings, preliminary research conducted in other countries revealed that younger adults in Italy suffered from boredom more compared to other age groups [5] . Furthermore, Chinese males, compared to females, were more likely to visit crowded places during the COVID-19 pandemic [34] . Indeed, research on the individual difference in boredom has reported that age and being female are negatively associated with boredom proneness [35, 36] .

Taken together, these observations could have important implications not only for researchers in the relevant fields but also for policy makers. First, two different patterns were identified across different emotions, a linear pattern for boredom/joy and a cubic pattern for the other emotions (e.g., anxiety). The finding may imply that a pandemic such as COVID-19 may affect our mental well-being through two pathways, the biological and the social pathway. Specifically, the former may be associated with fear and anxiety of infection and death, while the latter with boredom from a monotonous life. The continuously increasing trend of boredom during the outbreak suggests that one serious obstacle to implementing social distancing or home confinement may be the experience of boredom. Second, we observed significant differences as a function of age and gender in boredom. Boredom, then, from a policy perspective, deserves special attention as a potential threat to pandemic containment measures. In particular, it may be important for policy makers to target these specific groups and develop interventions that help them manage boredom while complying with social distancing. Finally, the present analysis revealed that women showed greater disturbances in well-being than men, which is consistent with the mounting evidence of gender inequality during the COVID-19 pandemic worldwide [37, 38] . As school and daycare closes, the responsibility of taking care of children and unpaid domestic work fell disproportionately on women, resulting in greater strain in their daily lives [39] . Thus, it would be crucial for policy makers to strive to mitigate the negative impacts of COVID-19 on women, especially mothers with young children.

The present research is one of the first nationwide reports to describe the well-being consequences of COVID-19. The various features of the present research (e.g., large sample size, diverse well-being measures, daily well-being tracking over 11 weeks, and multilevel analysis) make the present research unique. Prior observational studies used Twitter or Google Trends as a data source to track well-being or emotional responses to external natural events [40] [41] [42] . However, such data are only a proxy for well-being as they often contain irrelevant information (i.e., chatter). In addition, the online data collection used in the present study is superior to most prior studies on collective responses to major life events that relied on retrospective reports [43] [44] [45] .

Despite these merits, the present research also has some (unavoidable) limitations that make us hesitant to generalize the current findings to other nations. First, COVID-19 is far from over yet. Hence, the current findings might be true only for an earlier phase of the outbreak. Continuing efforts are hence needed to monitor the well-being of citizens until the COVID-19 outbreak finally ends. Second, countries differ vastly in terms of various aspects of the COVID-19 outbreak (e.g., numbers of patients and death, social distancing policies, etc.). Every country should conduct its analysis. We cannot generalize our findings to all other nations. Rather, the present study wishes to share some lessons from Korea's experience so that we may collectively understand the psychological aftermath of this unprecedented threat and better prepare ourselves for similar future outbreaks. Third, our survey was mainly crosssectional, although some participants (25.9%) reported their daily well-being more than once during the outbreak, leading us to use multilevel modeling. Robustness checks were conducted to test whether our main findings would hold for different subsamples who provided more than two responses. Results revealed that the main findings of decline-recovery-decline pattern held across these subsamples (see S2 Fig and S17 Table) . Hence, it did not capture much within-person variation in well-being. Similarly, the present research adopted a quantitative approach based on a self-reported scale and thus, it would be a worthy endeavor for future research to explore the precise nature of one's experiences during the pandemic with a qualitative approach. Finally, despite the large sample size, the participants were predominantly young and female, thus, the moderating role of age or gender should be interpreted with caution. We are well aware of this limitation, but we would like to point out that it would be unrealistic to design a true panel study beforehand in cases like COVID-19 and other unexpected disasters. 

China's first confirmed Covid-19 case traced back to November 17 South China Morning Post2020

Coronavirus disease (COVID-2019) situation reports

Managing mental health challenges faced by healthcare workers during covid-19 pandemic

Multidisciplinary research priorities for the COVID-19 pandemic: a call for action for mental health science. The Lancet Psychiatry

Evaluating COVID-19 Public Health Messaging in Italy: Self-Reported Compliance and Growing Mental Health Concerns. medRxiv

The psychological impact of the COVID-19 epidemic on college students in China

The consequences of the COVID-19 pandemic on mental health and implications for clinical practice

A longitudinal study on the mental health of general population during the COVID-19 epidemic in China

The satisfaction with life scale

Subjective well-being

The Contours of Positive Human Health

Organization for Economic Cooperation and Development (OECD)

The Geography of World Happiness

Diseases Control and Prevention. The updates of COVID-19 in Republic of Korea

South Korea Reports No New Domestic COVID-19 Cases: Statista; 2020 [updated April 30

On Happiness and Human Potentials: A Review of Research on Hedonic and Eudaimonic Well-Being

Within-person day-of-week effects on affective and evaluative/ cognitive well-being among Koreans. Emotion

The Role of Coping Strategies in Maintaining Well-Being During the COVID-19 Outbreak in South Korea

Day-of-week mood patterns in the United States: On the existence of 'Blue Monday', 'Thank God it's Friday' and weekend effects

Package 'lme4'. Convergence

The elements of statistical learning: data mining, inference, and prediction

Subjective well-being and adaptation to life events: A metaanalysis

Experience with Social Distancing Early in the COVID-19 Pandemic in the United States: Implications for Public Health Messaging

A group of young adults held a coronavirus party in Kentucky to defy orders to socially distance. Now one of them has coronavirus: CNN health

Just think: The challenges of the disengaged mind

California to 'young people' ignoring social distancing: 'Wake up, don't be selfish': KRON 4; 2020

Analysis on 54 Mortality Cases of Coronavirus Disease 2019 in the Republic of Korea from

Demographic science aids in understanding the spread and fatality rates of COVID-19

Emotion dynamics across adulthood in everyday life: Older adults are more emotionally stable and better at regulating desires. Emotion. 2020:No Pagination Specified-No Pagination Specified

Intraindividual variability in positive and negative affect over 45 days: Do older adults fluctuate less than young adults? Psychology and aging

Social and emotional patterns in adulthood: support for socioemotional selectivity theory

Knowledge, attitudes, and practices towards COVID-19 among Chinese residents during the rapid rise period of the COVID-19 outbreak: a quick online cross-sectional survey

Cognitive and affective predictors of boredom proneness

Age and Gender Differences in Boredom Proneness

The indirect impact of COVID-19 on women. The Lancet Infectious Diseases

Gender inequality during the COVID-19 pandemic: Income, expenditure, savings, and job loss

From Insights to Action: Gender equality in the wake of COVID-19

Assessing the impact of the coronavirus lockdown on unhappiness, loneliness, and boredom using Google Trends

National and Local Influenza Surveillance through Twitter: An Analysis of the 2012-2013 Influenza Epidemic

Sadness Shifts to Anxiety Over Time and Distance From the National Tragedy in

Reexamining adaptation and the set point model of happiness: reactions to changes in marital status

Does A Major Earthquake Change Attitudes and Well-Being Judgments? A Natural Experiment

Psychological adaptation to the Great Hanshin-Awaji Earthquake of 1995: 16 years later victims still report lower levels of subjective wellbeing