Factors Associated with Acute and Persistent Psychological Disturbances During the COVID-19 Pandemic

Background: The emergence of Corona virus disease (COVID-19) and wide-spread counter-measures, such as quarantine and social distancing, can have a signicant impact on mental health of individuals. Methods: This online study anonymously screened 13,332 individuals worldwide for acute psychological symptoms related to Corona virus disease 2019 (COVID-19) pandemic from March 29 th to April 14 th , 2020. A total of n=12,817 responses were considered valid. n=1077 participants from Europe were screened a second time during May 15 th to May 30 th to longitudinally ascertain the persistence of psychological effects. Results: Female gender, pre-existing psychiatric condition, and prior exposure to trauma were identied as notable factors associated with increased acute psychological symptoms during COVID-19. The same factors, in addition to, being related to someone who demised due to COVID-19 and using social media more than usual predicted persistence of psychological disturbances. Optimism, ability to share concerns with family and friends like usual, positive prediction about COVID-19, and daily exercise predicted fewer acute and persistent psychological symptoms. Conclusions: Females, psychiatric patients, and individuals with previous exposure to trauma are at increased risk of being persistently psychologically affected by COVID-19.


Background
The emergence of novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in December 2019 and the global spread of corona virus disease 2019 (COVID- 19) has transpired as the most severe and publicized human crisis in recent history. As of August 1, 2020, the global burden of COVID-19 has exceeded 17 million cases worldwide. 1 Quarantine, isolation, and social distancing have been recommended by the World Health Organization (WHO), Center for Disease Control (CDC), and health o cials worldwide to combat the spread of COVID-19. 2,3 To adequately enforce implementation of these measures, one-third to half of the world remained in complete lock-down for several weeks. 4 As a result of these extensive lock-down measures, economic markets have demonstrated alarming instability, with little indication of a timely recovery. The International Labour Organization projected that up to 25 million jobs could be affected overall. 5 The impact of COVID-19 on mental health of the masses has emerged as a matter of enormous concern. 6 A number of factors related to COVID-19 can adversely affect the mental health of individuals, with an even higher risk in those predisposed to psychological conditions. 7 Being in quarantine or isolation for extended periods of time has been associated with depression, anger, anxiety, and suicide as reported following the SARS epidemic of the early 2000s. 7 Similarly, uncertainty of economic recovery and loss of job security are important factors previously associated with neuropsychiatric perturbations. [8][9][10] Concerns have also been raised about increase in incidents of domestic violence and 'screen time' of individuals during the COVID-19 pandemic, [11][12][13] which are known risk factors for the development or worsening of psychological conditions. 14 Furthermore, fear and paranoia of being infected with SARS-CoV-2 and the stigma associated with manifesting symptoms such as cough or sneezing could negatively impact mental well-being. 15 The fear of losing a loved one and the grief following loss are other potential disturbances to mental health accompanying disease outbreaks. 16,17 Finally, it remains a consideration that SARS-CoV-2 may itself have neuropsychiatric manifestations as its effects on the nervous system are increasingly reported in patients who do not exhibit prominent respiratory tract symptoms. 18 A number of studies from China have reported signi cant increases in symptoms of anxiety, distress, and risk of PTSD in students and health professionals assessed during the COVID-19 pandemic. 13,19−25 A timely assessment on a global scale is paramount to display the mental health impact of the COVID-19 pandemic. With this data, health systems can strive to improve mental health services to reduce the longterm morbidity and mortality related to the COVID-19 crisis. Furthermore, this information could aid policymakers in improving the compliance of masses to the lock-down measures. 7 To address this, we assembled a team of health professionals (neuroscientists, psychiatrists, psychologists, data scientists, and medical students) across all continents to develop a global study on the mental health impact of COVID-19. Our primary assessment employed a fully anonymous online survey screening individuals in multiple countries for indicators and/or risk of general psychological disturbance, post-traumatic stress disorder (PTSD), depression, and suicidal ideation. The prevalence of these conditions was then cross-analyzed with participants' demographics, opinions/outlooks, personality traits, current house-hold conditions, previous psychiatric disease history, and factors associated with COVID-19 to identify speci c risk and resilience factors. We found alarming global trends for general psychological disturbances, risk for PTSD and depression, and suicidal ideation that were speci cally predicted by participant demographics, personality traits, house-hold conditions, previous psychiatric disease and/or risk factor history and prediction about COVID-19 resolution. A follow-up assessment of the European participants one month later showed persistence of these effects.

Study Design
The study comprised a primary and a follow-up assessment. The primary assessment comprised a crosssectional electronic survey-based assessment of individuals above the age of 18 years willing to participate in the study. The anonymous survey was conducted among participants from diverse demographic groups across continents using standardized self-report scales to screen for general psychological disturbance, risk for PTSD, and symptoms of depression. Speci c responses were also independently assessed to screen for suicidal ideation. The survey was available online for a period of 15 consecutive days starting 18:00 Central European Time (CET) on March 29, 2020 and concluding on 18:00 CET on April 14, 2020. The secondary assessment was performed one month after completion of the primary assessment for a period of 15 consecutive days starting 18:00 CET on May 15, 2020 and concluding 18:00 CET on May 30,2020. The secondary assessment was limited to European participants who had lled the primary survey.

Questionnaire development
The questionnaire was developed via close consultation between a neuroscientist, a neuropsychologist, a psychiatrist, a data scientist, and a psychiatry clinic manager. The questionnaire included closed-ended questions that assessed participant characteristics and opinions, and screened for neuropsychiatric conditions through standardized and validated self-report scales. The questionnaire prototype was prepared in English (Appendix 1) and translated into 10 additional languages (Arabic, Bosnian, French, German, Greek, Italian, Persian, Polish, Spanish, and Turkish; Appendix 2). The translation was performed by bilingual native speakers and vetted by volunteers native to those countries. The feasibility of each questionnaire was con rmed using pilot studies comprised of 10 participants each. These responses were excluded from the nal analysis.
The questionnaires (Appendix 1) included a section on participant demographics (age, gender, country, residential setting, educational status, current employment status) house-hold conditions (working/studying from home, home isolation conditions, pet ownership, level of social contact, social media usage, time spent exercising), COVID-19 related factors (knowing a co-worker, friend, or family member who tested positive for or demised due to COVID-19, prediction about pandemic resolution), personality traits (level of optimism, level of extroversion), previous history of psychiatric disease and/or trauma, previous exposure to human crisis, and level of satisfaction with actions of the state and employer during the current crisis. All questionnaires were rated on binary (yes/no) responses or Likerttype scales.
The other sections contained general health assessment based on WHO Self-Reporting Questionnaire-20 (SRQ), Impact of Event Scale (IES), and Beck's Depression Inventory II (BDI). 24,26,27 These scales were chosen based on their common usage and e cacy in previously employed works studying the psychological impact of human crises, including the SARS epidemic. [28][29][30][31][32][33][34][35][36] IES was purposefully adjusted to assess the impact of an ongoing event rather than a past event. For this purpose, the past tense was converted into the present tense in each question without changing the subject matter. This adjustment was performed in consultation with an independent neuropsychologist not involved in the study. For all scales, participants were prompted to think of and report their physical and psychological state during the preceding week. The secondary assessment was only limited to SRQ.

Data Collection
Primary assessment Using a non-randomized referral sampling (snowball sampling) method, participants were contacted by a team of 70 globally diverse members (study authors and volunteers that have been acknowledged in the acknowledgement section) using electronic communication channels including posts on social media platforms, direct digital messaging, and personal and professional email lists. For primary assessment, the data collection procedures were repeated at least thrice during the data collection period (March 29-April 14, 2020). The data was collected exclusively online for participants under 60 years of age. For participants who were 60 or above, a special provision was allowed for assistance in recording their responses online as older adults are often not comfortable with virtual platforms. 37 Our data collection strategy resulted in a total of 13,332 responses during the primary assessment. Surveys completed by participants who were younger than 18 (n = 34), those with missing responses for all dependent variables (n = 112), lled the second time (n = 325), missing geographic location (n = 20), and from WHO AFRO region (n = 24) were excluded from the nal analysis. When the responses were missing for individual items, the missing data were considered null and excluded from the analysis for that particular variable. The number of participants for top 12 countries and the regions encompassing the other countries is represented in the Supplementary item S1.

Follow-up assessment
For the follow-up assessments, data collection was limited to European participants only. The data collection team from Europe contacted the potential participants using the same electronic communication channels that were used for data collection during primary assessment. The data collection procedures were repeated three times during the data collection period. A total of 1,077 responses were collected from Europe during the follow-up study. Against the 6,207 responses from Europe collected during the primary assessment, this established a response rate of 17.35%.
Non-adjusted analysis for SRQ, IES, and BDI scores Mean scores with standard deviations were calculated for SRQ, IES and BDI scores from all valid responses (n = 12,817) and compared across all of the following categorical predictors via Kruskal-Wallis tests with the Chi-square function. The categorical predictors included gender, residential status, education level, employment status, being a medical professional, working remotely from home, satisfaction with employer, satisfaction with the state (government), home-isolation status, interaction with family and friends, social media usage, ability to share concerns with a mental health professional, ability to share concerns with family and friends, prior exposure to a human crisis situation, previous exposure to trauma, level of extroversion, prediction about COVID-19 resolution and one's self-determined role in the pandemic.
Multiple Regression Models for SRQ, IES, and BDI Multiple linear and logistic regression models were built for SRQ, IES, and BDI using mean scores and cutoffs for respective categorical classi cation.
For linear regression, generalized linear models with the glm function were devised using the lme4 package (Bates et al., 2015). The three univariate linear regression models, one each for SRQ, IES, and BDI, were tted and corrected for multiple comparisons followed by glm function analyses. Following the Bonferroni correction for multiple comparisons, the p-value threshold was set to 0.017. For each linear regression model, 'age' was entered as a continuous independent predictor whereas all aforementioned predictors were entered as categorical xed effects. Poisson family and log link function were used to model BDI and SRQ factors. In order to choose the best model (based on Akaike information criterion; AIC or Bayesian information criterion; BIC) from the set of predictors, stepwise model selection was performed from the MASS package (Venables et al., 2002).
Logistic regression was performed to generate odds ratios (ORs) for SRQ, IES, and BDI using the following categorization scheme; SRQ: 0 = normal (0-7 points), 1 = concern for general psychological disturbance (8-20 points); IES: 0 = normal (0-23 points), 1 = PTSD is a clinical concern (24-32 points), 2 = threshold for a probable PTSD diagnosis (33-36 points), 3 = Severe condition (high enough to induce immunosuppression) (≥37 points). For generating ORs, the variables were regrouped as 0 = no concern versus any type of concern (1/2/3); BDI: 0 = These ups and downs are considered normal (1-10 points). . For generating ORs, the variables were regrouped as 0 = no concern versus any type of concern (levels 1/2/3/4/5). Cut-offs for SRQ, IES, and BDI were de ned using least stringent thresholds for each of these measures from previous literature to ensure high sensitivity of the screening. 24-28 Furthermore, separate OR analysis was performed with reference level set to 0 = absence of symptom that was compared to presence of symptom (varying severity levels of the symptom regrouped into one category). Correlations between SRQ, IES, and BDI were performed through Pearson's correlation test and illustrated as x ~ y plots.
For the follow-up study, a generalized linear model with the glm function was tted using the lme4 package (Bates et al., 2015). All predictors were entered as categorical xed effects. Poisson family and log link function were used to model the SRQ factor. An interaction effect was introduced to inspect whether the follow-up assessment and working from home, satisfaction with the employer, having a preexisting psychiatric condition, closely knowing someone who died of COVID-19, and residence (urban or rural) had a signi cant effect on SRQ score progression during the primary and follow-up assessments.
All statistical analyses were performed by the analysis team comprising MP, SG, PR, and AJ in consultation with ZB. Unadjusted analyses of SRQ, IES, and BDI scores between different participant demographics/characteristics showed signi cantly (p < 0.017) greater prevalence of psychological symptoms in participants who were female, unemployed, working remotely from home, dissatis ed with the response of their employer/state to COVID-19, home-isolated alone, with a pet, interacting with friends/famiy less than usual, using social media more than usual, and in those with less-than-usual ability to share concerns with friends/family. Signi cantly (p < 0.017) higher scores on SRQ, IES, and BDI were also seen in participants who self-reported as being pessimist or introvert, not feeling in control during COVID-19, and having an overall negative prediction about COVID-19 resolution.

Follow-up Study
The demographic distribution of the participants included in the follow-up study of European participants was similar to the primary assessment with higher numbers from those participants who were female (74.57%), working/studying remotely from home (56.95%), and currently under home-isolation with a partner/family (65.30%). A majority of participants also reported increased social media usage (61.94%), less-than-usual or minimal interaction with family and friends (65.30%), and feeling a sense of control in protecting themselves and others during the COVID-19 pandemic (61.85%).
Unadjusted analyses of SRQ scores between different participant demographics/characteristics showed signi cantly (p < 0.05) higher prevalence of psychological symptoms in participants who were female, medical or healthcare professionals, dissatis ed with the response of their employer/state to COVID-19, interacting with friends/family less than usual, using social media more than usual, and in those with less-than-usual ability to share concerns with friends/family. Signi cantly (p < 0.05), higher scores on SRQ were also seen in participants with pre-existing psychiatric conditions, previous exposure to traumatic experiences, and who self-reported as being pessimist or introvert. Means and standard deviations for all comparisons are presented in the Supplementary item S6.
Adjusted analysis using generalized linear model for the SRQ questionnaire is reported in Main Item 5.
The following factors were independently associated with increased SRQ scores on the follow-up assessment: psychiatric condition that worsened during the COVID-19 pandemic (SRQ mean-coe cient:

Discussion
This study highlights a signi cant impact of COVID-19 pandemic on mental health worldwide and provides unprecedented evidence for persistence of these effects in a large population sub-set of European participants.
In addition to reporting prevalence, a major aim of this study was to identify speci c risk and resilience factors for psychological perturbations during the current COVID-19 crisis. Worsening of a pre-existing psychiatric condition, female gender, exposure to trauma before age 17, and working remotely predicted higher risk of general psychological disturbance, PTSD, depression, and increased concerns about physical health and appearance. Additionally, considering oneself an introvert was associated with heightened risk of general psychological disturbance and depression; being unemployed, living alone, and limited interaction with family and friends also increased the risk for depression. An overall protective effect against all major psychological perturbations was observed for the following factors; increasing age, considering oneself an optimist, positive prediction about COVID-19 outcome, ability to share concerns with family and friends like usual, daily physical exercise/sport for 15 minutes or more, and being satis ed with the actions of employer/state in response to COVID-19.
To ensure that the psychological symptoms assessed in this study are related to COVID-19 pandemic, the participants were repeatedly prompted to consider COVID-19 and their feelings during the preceding one week while lling in the survey. Furthermore, the phrase 'this crisis' was present in all the screening questions, for example, 'I am unable to sleep well during this crisis'. We would like to further highlight the difference in the proportion (22%) of participants who reported pre-existing psychiatric conditions versus the ones who report general psychological disturbance (43%) assessed through SRQ indicating an impact of COVID-19 pandemic (Supplementary Item S8). Furthermore, we compared the prevalence of depression in all of our featured countries based on different BDI cut offs for depression versus the recent-most available statistics from WHO (2017) and notice a remarkable difference (Supplementary Item S9).
To the best of our knowledge, this study is the rst worldwide assessment of the mental health effects of COVID-19. Previous studies on the psychological impact of COVID-19 have been mostly from China 13,[19][20][21][22][23][24][25]39 The largest of these studies (n = 52,730) that surveyed voluntary public participants, reported symptoms of psychological distress in almost one-third of the participants according to the peritraumatic distress index. 40 Another notable study, on health professionals (n = 1,255), revealed depression, anxiety, and symptoms of general distress in almost half of participants, and sleep disturbances in almost 8%. 13 One-third of the participants in a Chinese study on college students (n = 7,143) in the Hubei province reported symptoms of anxiety. 25 Some of our observations are supportive of ndings in these studies, such as female gender, living alone, and negative prediction about COVID-19 outcome arising as risk factors for psychological perturbations. However, our study identi es several unique risk and resilience factors that were not investigated previously.
Parallels can also be drawn between our study and existing research on the psychological effects of the SARS and other previous epidemics. These studies reported PTSD, anxiety, distress, anger, and confusion as major sequelae of the epidemic and quarantine measures. 35,41−43 It has previously been reported, however, that very few studies investigated speci c risk or protective factors 7 for these mental health disturbances. One notable study showed longer quarantine duration, boredom, nancial instability, stigma, inadequate resources and information de cit to exacerbate the negative psychological impact from the SARS outbreak. In noteworthy contrast to our work, the study was performed several months after the epidemic had occurred. 44 Identi cation of speci c risk and resilience factors is an essential rst step for developing strategies to mitigate the negative psychological impact of COVID-19 at a regional and global level. For example, selective vulnerability of females indicated in this study warrants further investigation for both the contributing factors and the resulting implications of such increased risk. These include social factors such as increased reporting of domestic violence in relation to COVID-19 45 , possible caregiver stress, and the impact of changes in roles and responsibilities secondary to the current health emergency.
Furthermore, increased risk of psychological perturbations in individuals with pre-existing psychiatric conditions and/or trauma exposure necessitates the initiation and/or expansion of mental health support systems available remotely. 40 Emerging evidence now supports the e cacy of web-and social-media based interventions in promoting mental health of masses focusing on paradigms based on mindfulness, positive psychology, and exercise. [46][47][48] Such interventions could be developed at the governmental and institutional levels and delivered to the masses via main-stream and social media.
Indeed, media outlets could also play a major role in promoting optimism and a positive attitude towards COVID-19 resolution, both of which were identi ed in our study as important resilience factors.
Furthermore, the association between remote working with increased psychological symptoms calls for optimization of the work-from-home settings and a greater emphasis on the general well-being of employees. This is further corroborated by the observation that participant satisfaction with the employer-response to the COVID-19 pandemic is associated with reduced psychological symptoms in this study.
This research has several strengths. This study employed the 2nd largest sample size to date in examining the mental health impact of COVID-19, and the number of participants well exceeds previous studies on the SARS epidemic. The only study with a larger sample size 40 employed a single scale for screening psychological disturbances. The administered measures in our study allowed for simultaneous screening of multiple psychiatric co-morbidities and the ndings can provide invaluable insight to global health systems. The availability of the questionnaire in 11 different languages is a notable and unprecedented effort to provide the study as much generalizability as possible. Furthermore, the timing of this study is an important strength. The primary assessment was performed from March 29-April 14, 2020. This timing coincides with the peak of COVID-19 pandemic in North America and Europe-a time period when almost one-half of the world remained in complete lock-down. 4 The follow-up assessment was performed with a targeted approach in European after a one month interval when the situation had improved considerably in Europe. This longitudinal assessment is a considerable strength of the study.
Finally, while cultural and linguistic factors are known to possibly impact psychological outcome measures when translations are utilized, the signi cant correlation between SRQ, IES, and BDI scores in this study cross-validates the assessment of psychological symptoms and con rms that COVID-19 pandemic is globally affecting the overall mental health of individuals.
The study also has potential limitations that warrant consideration when interpreting the results. First, the study employed a non-randomized sampling strategy. While this method has certain disadvantages, we hope that our results will catalyze the development of more studies on this essential topic that could be conducted by global outlets such as WHO and the European Union (EU) on a world-or continent-wide scale. Second, the data collection was exclusively done in an online format that may exclude those lessversed in web-usage, such as illiterate, disadvantaged, underdeveloped, or rural populations. We tried to reduce this bias by translating the study questionnaire into native/o cial languages for each of the featured countries. The third considerable limitation is the use of self-reporting scales rather than clinical veri cation. However, the anonymous nature of the survey and widespread social distancing measures preclude such veri cation. Additionally, it is not possible to adjust for the confounding effect of non-COVID-19-related individual crisis situations on participant responses. We tried to reduce this effect by formatting survey questions in such a way that would prompt participants to consider their mental state over the preceding week, rather than current mood.
Utilizing the 'feedback' feature in our online questionnaire, several participants expressed that participation in the survey helped them focus on their mental health. Furthermore, a number of participants reported eating more than usual for comfort or out of boredom. This feedback could aid in efforts to develop mental health screens speci c to COVID-19 pandemic. 49

Conclusions
In conclusion, this effort highlights a signi cant impact of the COVID-19 pandemic at a regional and worldwide level on the mental health of individuals and elucidates prominent associations with their demographics, history of psychiatric disease risk factors, house-hold conditions, personality traits, and attitude towards COVID-19. These results could serve to inform health professionals and policymakers across the globe, aiding in dynamic optimization of mental health services during and following the COVID-19 pandemic, and reducing its long-term morbidity and mortality. Written informed consent was obtained from each participant to allow anonymous recording and analysis of their answers. The data was collected in a completely anonymous fashion without recording any personal identi ers. This strategy ensured that the con dentiality of the participants was maintained throughout all phases of the study. The study procedures were reviewed and approved by University of    Risk and resilience factors for general psychological disturbance (SRQ), risk for PTSD (IES), and depression (BDI). These foster-plots show the mean estimates and the 95% con dence intervals (CI) for adjusted coe cients signi cantly affecting SRQ, IES and BDI scores respectively generated through multiple regression models. Only predictors that survived Bonferroni correction for multiple comparisons  Factors associated with persistent general psychological disturbance on follow-up assessment. These foster-plots show the mean estimates and the 95% con dence intervals (CI) for adjusted coe cients affecting SRQ generated through multiple regression. Panel a shows xed predictor for SRQ scores during the follow-up assessment. Panel b indicates interaction terms included in our regression model, indicating a signi cant difference between the xed effects and SRQ scores during the rst phase of the data collection and the follow-up assessment. Factors increasing SRQ score are shown in red and factors decreasing SRQ scores are shown in blue.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.