Prevalence of post-traumatic stress symptoms among people influenced by coronavirus disease 2019 outbreak: A meta-analysis

Background As one of the most widely researched consequence of traumatic events, the prevalence of post-traumatic stress symptoms among people exposed to the trauma resulting from coronavirus disease 2019 (COVID-19) outbreak varies greatly across studies. This review aimed at examining the pooled prevalence of post-traumatic stress symptoms among people exposed to the trauma resulting from COVID-19 outbreak. Methods Systematic searches of databases were conducted for literature published on PubMed, EMBASE, Web of Science, the Cochrane Library, PsycArticle, and Chinese National Knowledge Infrastructure until October 14, 2020. Statistical analyses were performed using R software (PROSPERO registration number: CRD42020180309). Results A total of 106,713 people exposed to the trauma resulting from the COVID-19 outbreak were identified in the 76 articles, of which 33,810 were reported with post-traumatic stress symptoms. The pooled prevalence of post-traumatic stress symptoms among people exposed to the trauma resulting from COVID-19 outbreak was 28.34%, with a 95% confidence interval of 23.03-34.32%. Subgroup analysis indicated that older age, male and bigger sample size were associated with higher prevalence of post-traumatic stress symptoms. After controlling for other factors, the results of meta-regression showed that the influence of gender and sample size on prevalence is no longer significant. Conclusions Symptoms of post-traumatic stress disorder (PTSD) were very common among people exposed to the trauma resulting from COVID-19 outbreak. Further research is needed to explore more possible risk factors for post-traumatic stress symptoms and identify effective strategies for preventing PTSD-related symptoms among people exposed to the trauma resulting from COVID-19 outbreak.

individuals may experience the fear of infection, social isolation, exclusion, and stigmatization, as patients, related caregivers, and workers, or even the general public [11,12]. It is said that there is a dose-response relationship between the degree of trauma and the mental health burden of disasters [13]. The prevalence of PTSD is higher among people who were directly exposed to the disaster, lower among related caregivers and rescue workers, and yet even lower in the general population. These different populations are likely to represent different levels of severity of trauma exposure, with direct victims having the highest exposure and associated PTSD prevalence while people in the general population having the lowest levels of exposure and associated PTSD prevalence [14]. Currently, the relationship between the degree of trauma and the mental health burden of COVID-19 outbreak is unclear.
As one of the most widely researched consequence of traumatic events [14], the prevalence of post-traumatic stress symptoms among people exposed to the trauma resulting from COVID-19 outbreak varies greatly across studies [15][16][17][18]. In addition, many factors have been reported to be associated with the prevalence of PTSD during COVID-19 outbreak, such as gender, age, and degree of trauma exposure [8,19,20], but the results are not consistent in different studies. The possible causes of the inconsistencies in the current estimates were unclear. For taking effective measures to reduce the psychological sequelae caused by COVID-19 across the world, it is necessary to determine a more accurate estimation of the prevalence of post-traumatic stress symptoms among people exposed to the trauma resulting from COVID-19 outbreak, and to explore the possible causes of the inconsistencies in the current estimates. This review aimed at examining the pooled prevalence of post-traumatic stress symptoms among people exposed to the trauma resulting from COVID-19 outbreak, summarizing possible vulnerability factors of post-traumatic stress symptoms and examining potentially vulnerable populations, try to provide a reference for COVID-19 and possible outbreak of infectious diseases in the future.

Methods
This review was reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA guideline) and Meta-analyses of Observational Studies in Epidemiology guidelines [21,22]. The protocol of this review is registered in the International Prospective Register of Systematic Reviews (registration number: CRD42020180309). See Supplementary Material for the details.

Study selection
Studies were included if they meet the following criteria: (a) the study was observational study; (b) the participants were adult aged ≥18; (c) information about prevalence of post-traumatic stress symptoms among people exposed to the trauma resulting from COVID-19 outbreak was provided; (d) the full article was written in English or Chinese. Studies were excluded if (a) the report was a review, comments, meta-analysis, or protocol and (b) the participants with comorbid symptoms or chronic disease (such as mental illness, cancer, etc.).

Data extraction
Data extraction was conducted independently in pairs by trained researchers who used standardized data extraction forms. Two reviewers (D.Q. and Y.L.L.) checked the titles, abstracts, and fulltexts of the initial search results independently. Data were extracted on first author, country or area, survey period, target population, study design, sample size, response rate, percentage of male participants, mean age of participants, instruments used to identify posttraumatic stress symptoms, prevalence of post-traumatic stress symptoms, and quality score of the included studies. Any discrepancies that emerged in these procedures were discussed and resolved by involving a third reviewer (S.Y.X.).

Quality assessment
Two independent reviewers (J.H. and F.Y.O.Y.) used the established guidelines, the Loney criteria, to evaluate the methodological quality of the included studies, which has been widely used to evaluate observational studies [23,24]. The included papers were scored according to eight criteria, such as study design, sample size, sampling method, response rate, definition of participants, appropriateness of measurement and analysis. The scores range from 0 to 8, with a score of 0-3 as low quality, 4-6 as moderate, and 7-8 as high [25]. See Table S3 for details on the quality assessment.

Statistical analyses
When data were available for three or more papers, prevalence was combined [26]. When there were four or more papers, quantitative subgroup analysis was conducted [27]. All the statistical analyses were performed using the "meta" (4.12-0) and "metafor" package (2.4-0) of R version 4.0.0. Between-study heterogeneity was evaluated by Cochran's Q test and quantified by the I 2 statistic, with values >50% indicating moderate heterogeneity [28,29]. As we expected considerable heterogeneity, we calculated pooled prevalence with the random effects model. The pooled prevalence of post-traumatic stress symptoms among people exposed to the trauma resulting from COVID-19 outbreak was combined using Logit transformation method by a random effects model [29,30]. If more than one dataset was reported for the same group of participants, the outcomes that were assessed at the baseline were used. In order to compare the prevalence from different papers, we conducted subgroup metaanalysis. Because subgroup analyses should be interpreted with caution [31], we planned a priori to limit our subgroup analyses to a small number of baseline characteristics including gender, age, area, population, survey time after the outbreak, sample size, assessment tool, and quality score. The difference between subgroups was examined using the Cochran's Q chi-square tests [30]. Mixed-model metaregression analyses were performed by using Freeman-Tukey double arcsine method to explore potential moderators on the heterogeneity [32]. Publication bias was investigated by funnel plot and Egger's test [30,33]. To evaluate the consistency of the results, sensitivity analysis was performed. In this study, sensitivity analyses were planned a priori for the primary analyses set by: (a) excluding studies one by one and (b) excluding studies with extreme outcomes [30,31]. All the statistical tests were two-sided, with a significance threshold of p < 0.05.

Literature search
As shown in Figure 1, a total of 7,032 references were identified. Among them, 3,897 duplicates were removed. By screening titles and abstracts, 2,964 irrelevant articles were excluded. A total of 171 potentially relevant full-text articles were independently assessed based on the selection criteria. Further, 95 studies were excluded because of the following reasons: duplicate articles or results (n = 6); review or conference abstract (n = 2); did not provide data on PTSD (n = 75); unable to locate full text (n = 10); not in English or Chinese (n = 1); and not for participants aged ≥18 (n = 1). Finally, 76 eligible studies were included in this review [7][8][9]19,20,. The reliability for the full-text review between the two reviewers (D.Q. and Y.L.L.) was rated as good (Kappa = 0.78) [105]. See Figure 1 for the details. Table 1 presents the main characteristics of the 76 included studies. Among them, 66 were in English and 10 were in Chinese. Most of the included studies were from Asia, such as China, Indian, and Singapore. See Table 1 for the details. From the 76 papers, 1 (1.31%) study was rated as high quality, 70 (92.11%) were rated as moderate, and 5 (6.58%) were rated as low quality. The reliability for the quality assessment between the two reviewers (J.H. and F.Y.O.Y.) was rated as good (Kappa = 0.73) [105]. For data extraction, all the criteria received a kappa value >0.85. Details of the methodological quality assessments of all 76 studies are showed in Tables S2 and S3.

Study characteristics
Pooled prevalence of post-traumatic stress symptoms among people influenced by the COVID-19 outbreak There were 76 studies reported prevalence of post-traumatic stress symptoms among people exposed to the trauma resulting from the COVID-19 outbreak. The forest plot in Figure 2 depicts the details. A total of 106,713 people exposed to the trauma resulting from the COVID-19 outbreak were identified in the 76 articles, of which 33,810 were reported with post-traumatic stress symptoms. The random effects model was used to determine the pooled prevalence (Q = 14,854.51, I 2 = 99.70%, p < 0.001), the pooled prevalence of post-traumatic stress symptoms among people exposed to the trauma resulting from COVID-19 outbreak was 28.34%, with a 95% confidence interval (CI) of 23.03-34.32%.

Subgroup analysis for the included studies
The details of subgroup analyses are presented in Table 2. Significant differences in the prevalence of post-traumatic stress symptoms between different ages were found (Q = 221.97, p < 0.001). The results indicated that older participants (with a mean age ≥51) showed higher prevalence of post-traumatic stress symptoms (62.16%), younger participants (with a mean age between 18 Table 3 presents the results of meta-regression analyses. Due to too much missing data (only 57.89% of studies reported data) on the mean age of participants, we were unable to include this variable in the meta-regression model. Bivariate meta-regression suggested that higher prevalence estimates reported in studies which used IES as assessment tool (β = À0.11, p = 0.061). Specifically, assessment tool accounted for 3.16% of the heterogeneity across studies, but the difference between different groups was not significant.

Publication bias and sensitivity analysis
Funnel plot of publication bias is presented in Figure 3. The funnel plot of publication bias is basically symmetric, but publication bias cannot be ruled out, so Egger's test was conducted. The results of the Egger's test showed that publication bias was not found in this study (t = À0.971, p = 0.334). When each study was excluded one-by-one, the recalculated combined results did not change significantly. The pooled prevalence of post-traumatic stress symptoms ranged from 27.18% (95% CI: 22.40-32.55%) to 28.96% (95% CI: 23.72-34.81%), and the I 2 statistic has remained at 99.7%. The results in the current study indicate that no individual study significantly influenced the overall results. In order to examine the influence of extreme cases on the current results, we excluded all the extreme results to conduct a sensitivity analysis (with a prevalence <5% or >70%). After removing eight studies [9,45,53,56,74,77,82,97], the pooled prevalence of post-traumatic stress symptoms was 26.20% (95% CI: 22.17-30.67%), and the I 2 statistic was 99.5%. Compared with the initial results, no significant changes were found. See Figure S1 for the details of sensitivity analysis.

Key findings
This review has highlighted the importance of considering the psychological impacts of people exposed to the trauma resulting from COVID-19 outbreak. A total of 106,713 people exposed to the trauma resulting from the COVID-19 outbreak were identified in the 76 articles, of which 33,810 were reported with post-traumatic stress symptoms. The pooled prevalence of post-traumatic stress symptoms among people exposed to the trauma resulting from COVID-19 outbreak was 28.34%, with a 95% CI of 23.03-34.32%.

Comparison with the literature
Based on the results of the World Health Organization World Mental Health Surveys in 2017, the global lifetime prevalence of PTSD-related symptoms among the general population was 3.90-5.60% [106]. The pooled prevalence of post-traumatic stress symptoms among people exposed to the trauma resulting from COVID-19 outbreak in this study was 28.34%, which was much higher than the general population. Between December 2019 and May 2020, a few reviews related to mental health and infectious disease outbreak reported limited data on pooled prevalence of PTSD-related symptoms during the COVID-19 outbreak. In Salehi et al.'s [15] study, the pooled prevalence of PTSD-related symptoms during the coronavirus outbreaks (SARS, MERS, and COVID-19) was 18%, which was much lower than our results. In Krishnamoorthy et al.'s [16] study, the pooled prevalence of PTSD-related symptoms during the COVID-19 outbreak ranged from 21.94 to 27.00% [17,18], were lower than our results. Those researches have explored a variety of mental health problems (such as depression, anxiety, and insomnia), PTSD is only one of the outcomes. Thus, we believe their search strategies on PTSD are inaccurate enough and the data they included were limited, which may affect the results. Despite the high heterogeneity in their studies, no subgroup analysis was conducted to explore the source of the heterogeneity on the pooled prevalence of PTSD in most of them [15,16,18], which we think is another limitation. In Cooke et al.'s [17] study, although the source of the heterogeneity was explored, they only included age and gender as moderators. Moreover, we found that the pooled prevalence of post-traumatic stress symptoms among people exposed to the trauma resulting from COVID-19 outbreak was higher than flood survivors (15.74%) and hurricane survivors (17.81%) [107,108], but equal to earthquake survivors (23.66%) [109]. Compared with COVID-19 outbreak, some natural disasters such as flood and hurricane can be predicted, while earthquakes, infectious disease like COVID-19 were often happened suddenly and without a warning, pose a huge threat to people's health and property in a short period of time [109]. Therefore, earthquakes and infectious disease outbreak might have caused more damage to people's mental health than flood and hurricane.
The prevalence of post-traumatic stress symptoms in older adults is significantly higher than the younger people in the subgroup analysis, which was consistent with other studies [32]. Due to too much amount of missing data (only 57.89% of studies reported data on mean age of participants), we were unable to include this variable in the meta-regression model. Thus, this observation requires further clarification. Based on the current results, healthcare providers should pay more attention to the assessment of early trauma responses among older COVID-19 patients in the clinical practice and implement early psychological interventions accordingly. It is said that females were more likely to develop PTSD [109]. In the current study, however, males were associated with higher prevalence of post-traumatic stress symptoms. The possible reason is that males are more likely to be affected by COVID-19 and reported a higher fatality rate [110], they may experience higher level of severity of trauma exposure. In addition, although the epidemic situation is more serious in Europe and the Americas, no significant difference was found between different regions. It might be related to issues with numbers of studies, we think more prevalence studies in low-income countries are needed to understand the panorama of PTSD among people influenced by COVID-19. Besides, Previous research has shown that patients of infectious disease often directly suffering from the symptoms and traumatic treatment. After being cured, they were more vulnerable to social discrimination than other groups [14]. These experiences may result in higher prevalence of PTSD among them when compared with other populations. Although the prevalence of PTSD among COVID-19 patients is higher than that of other populations in this study, the difference is not significant, which need further exploration. Also, post-traumatic stress symptoms among people exposed to the trauma resulting from COVID-19 outbreak were higher in the immediate aftermath of the outbreak (0-4 weeks), but the difference was not significant, which was inconsistent with other studies [109,111,112]. The possible reason is that the epidemic has not abated over time and has been spreading, people have been exposed to the trauma resulting from COVID-19 outbreak. Thus, we think ongoing surveillance is essential. Furthermore, we found that the pooled prevalence of post-traumatic stress symptoms among people exposed to the trauma resulting from COVID-19 outbreak identified by different assessment tools was not significant. Studies have indicated that prevalence identified by screening tools were higher than prevalence identified by diagnostic tools [113]. However, all the included studies in this study were used screening tools, we were unable to explore the difference. It is note of worthy that some instruments derived from different conceptualization of the disorder and they may encompass different symptoms (e.g., Impact of Event Scale-Revised derived from the Diagnostic and Statistical Manual-Fourth Edition, Text Revision (DSM-IV-TR) while Post-Traumatic Stress Disorder Checklist-5 from DSM-5, including more symptoms than the former) [114]. Given that the included articles use many different screening tools, we think further research is needed to explore the influence of different screening tools on the prevalence of PTSD among people influenced by COVID-19 outbreak. It is reported that studies with poor methodological quality generally yielded more extreme prevalence estimates [115], the current study showed similar results. After controlling for other factors, however, the results of meta-regression showed that the influence of methodological quality on prevalence is no longer significant. Hence, this observation requires further clarification.

Implications for the future
Epidemiological studies have demonstrated a rather high prevalence of mental health problems among different population after an epidemic of infectious disease [116][117][118]. While most of these mental health problems will fade out after the epidemic, symptoms of PTSD may last for a prolonged time and result in severe distress and disability [119]. In terms of applicability to COVID-19, we think ongoing surveillance is essential and healthcare policies need to take into account both short-term and long-term preventive strategy of PTSD. In addition, all the included studies identified PTSD by selfreporting questionnaires rather than clinical interviews by professional psychiatrists, as a consequence of which, the pooled prevalence of PTSD may have been overestimated. Only 6.39% reported data on COVID-19 patients. Thus, we think a large multicenter prospective study using a single validated measure of PTSD and measuring possible confounding factors in randomly selected COVID-19 patients is needed in the future, which would provide a more accurate estimate of PTSD among people influenced by COVID-19 outbreak, especially for COVID-19 patients. Although, there is little doubt that there is a dose-response relationship between the degree of trauma and the mental health burden of disasters [14], this relation may not necessarily mean that the principal mental health burden of people exposed to the trauma resulting from COVID-19 outbreak is among those who were most directly affected by the disease [13]. It will be important to establish whether indirect exposure to a trauma during a COVID-19 pandemic was correlated with higher risk of PTSD. Also, it is necessary to assess the relation between exposure to multiple traumas and risk of PTSD in the future. Additionally, subgroup analyses and the meta-regression analysis did not identify major sources of the heterogeneity although a high degree of heterogeneity between studies was observed. Therefore, there might be a considerable amount of uncertainty regarding the pooled prevalence of PTSD among people influenced by COVID-19 outbreak. Future research should, therefore, explore more potential risk factors for PTSD among people influenced by COVID-19 outbreak, especially genetic background as well as social support, previous traumatic events or concomitant psychiatric disorder [109,120].

Limitations
Firstly, we excluded studies were not written in English or Chinese. Besides, although subgroup analyses and meta-regression analyses were conducted to control many moderating factors for the pooled prevalence of post-traumatic stress symptoms, heterogeneity was still remained in this review. It is reported that heterogeneity is difficult to avoid in meta-analysis of epidemiological surveys, [121] which suggesting the need for caution when drawing inferences about estimates of PTSD in post-disaster research. Additionally, although our study included relevant studies across 30 countries, more than half of the eligible studies were from upper-high income countries. Prevalence studies were scarce for many countries, especially for low-income countries. Considering the inconsistency of the healthcare environment and socioeconomic status across the world, more prevalence studies in low-income countries are needed to understand the panorama of PTSD among people influenced by COVID-19 outbreak. Also, we noticed that all the included studies were used screening tools to assess post-traumatic stress symptoms, no studies included were used diagnostic tools. It is possible that the pooled prevalence of post-traumatic stress symptoms caused by COVID-19 outbreak was overestimated in this review. Thus, we think ongoing surveillance is essential. Lastly, some included studies were investigated the prevalence before the time threshold from the first event (usually 30 days), we were unable to check this possible bias between studies. Although we explored the influence of survey time on the pooled prevalence, no significant result was found, which need further clarification.

Conclusion
This review has highlighted the importance of considering the psychological impacts of people exposed to the trauma resulting from COVID-19 outbreak. A total of 106,713 people exposed to the trauma resulting from the COVID-19 outbreak were identified in the 76 articles, of which 33,810 were reported with post-traumatic stress symptoms. The pooled prevalence of post-traumatic stress symptoms among people exposed to the trauma resulting from COVID-19 outbreak was 28.34%, with a 95% CI of 23.03-34.32%. Further research is needed to explore more possible risk factors for post-traumatic stress symptoms and identify effective strategies for preventing and treating PTSD-related symptoms among people exposed to the trauma resulting from COVID-19 outbreak.
Financial Support. This research was supported by the Ministry of Science and Technology of China (Grant no.: 2016YFC0900802). The funding agency did not take part in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript.