The Association between Resilience and Psychological Distress during the COVID-19 Pandemic: A Systematic Review and Meta-Analysis

This study examined the association between resilience and psychological distress in healthcare workers, the general population, and patients during the COVID-19 pandemic. We searched the PubMed, Web of Science, PsycInfo, Science Direct, and Nursing and Allied Health databases. Included articles examined healthcare workers (e.g., physicians and nurses), the general population, and patients during the COVID-19 pandemic. Studies of exposure to other infectious diseases related to epidemics or pandemics (e.g., SARS and MERS) were excluded. This study was performed following the Cooper matrix review method and PRISMA guidelines, followed by a meta-analysis of study results using R version 4.1.2. A random effect model was used for the pooled analysis. This study was registered with PROSPERO (registration No. CRD42021261429). Based on the meta-analysis, we found a moderate negative relationship between overall resilience and psychological distress (r = −0.42, 95% confidence interval [CI]: −0.45 to −0.38, p < 0.001). For the subgroup analysis, a moderately significant negative relationship between overall resilience and psychological distress was found among healthcare workers (r = −0.39, 95% CI: −0.44 to −0.33, p < 0.001), which was weaker than in the general population (r = −0.45, 95% CI: −0.50 to −0.39, p < 0.001) and in patients (r = −0.43; 95% CI: −0.52 to −0.33; p < 0.001). This association was robust, although the heterogeneity among individual effect sizes was substantial (I2 = 94%, 99%, and 74%, respectively). This study revealed a moderate negative relationship between resilience and psychological distress in healthcare workers, the general population, and patients. For all these populations, interventions and resources are needed to improve individuals’ resilience and ability to cope with psychological distress during the COVID-19 pandemic and in future disease outbreaks.


Introduction
The rapid worldwide spread of coronavirus disease 2019 (COVID-19) has dramatically impacted various aspects of global public health. This disease, which affects the respiratory system and is easily transmitted from one person to another, can be fatal in vulnerable populations. On 20 June 2021, the World Health Organization (WHO) reported that the first wave of COVID-19 infections had reached approximately 177 million people and that 3.8 million people had died from the disease both in the general population and among healthcare providers [1].
In addition to its physiological impacts, the COVID-19 pandemic has adversely affected mental health, leading to psychological distress in populations across the globe [2,3]. Psychological distress is defined as the reaction of an individual to external and internal stresses and as a mixture of psychological symptoms encompassing stress, depression, and

Search Strategy
This study was registered in PROSPERO (registration No. CRD42021261429) to avoid duplication of effort and minimize the chance of reporting bias. The systematic review was conducted following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [38]. PubMed, Web of Science, PsycInfo, Science Direct, and the Nursing and Allied Health Database were used to search for studies for this review on 2 June 2021; the search strategy, including keywords and index terms, was adapted as necessary for each database. In addition, the reference list of each included source of evidence was screened to identify potential additional studies. Because COVID-19 first emerged in Wuhan, China, in December 2019, and because this study was limited to the first wave of the pandemic, only studies published from 1 December 2019 to 1 June 2021 were included in the review. Population-intervention-comparison-outcome (PICO) keywords applied during the database search are shown in Supplementary Table S1.

Selection Criteria
The inclusion criteria for the study were as follows: (1) a full-text journal article published in English; (2) original quantitative research focusing on the relationship between resilience and psychological distress, including depression, stress, anxiety, and PTSD, using correlational coefficients; and (3) use of self-reported measurement of resilience with the Connor-Davidson Resilience Scale (CD-RISC), which is the scale most widely used to assess psychological resilience. This instrument focuses on resources that can help individuals to recover from and adapt to disruptions or stressful events such as the COVID-19 crisis. We did not include studies using the Brief Resilience Scale (BRS), which directly measures one's ability to bounce back or be resilient but does not consider external resources [39]. In addition, although other resilience instruments have been applied in resilience studies, the various theoretical constructs and frameworks of resilience underpinning these instruments were not suitable for conducting our meta-analysis due to heterogeneity issues. Furthermore, we excluded sources if they (1) reported studies of other infectious diseases related to epidemics or pandemics (e.g., SARS or MERS); (2) were review or interventional studies; or (3) were gray literature, books, abstracts, or study protocols.

Study Selection
Following the literature search, all identified studies (N = 2106) found in PubMed (n = 496), Web of Science (n = 216), PsycINFO (n = 165), Science Direct (n = 111), and the Nursing and Allied Health Database (n = 891) were exported into EndNote X9 reference management software [40]. After duplicates were removed (n = 227), the 1879 remaining studies were exported into the Joanna Briggs Institute System for the Unified Management, Assessment, and Review of Information (JBI SUMARI) [41]. The researchers (TJ and WM) then independently screened their titles and abstracts in accordance with JBI SUMARI procedures based on selection criteria such as the publication language, participants, study design, and use of the CD-RISC self-report measure. After screening, 1626 ineligible records were excluded. The full texts of the 253 remaining articles were then retrieved for eligibility screening. During both title/abstract and full-text screening, discrepancies between two authors' (TJ and WM) independent assessments were generally resolved through discussion. Finally, a total of 33 studies meeting the eligibility criteria were included in the review. The study selection process and reasons for excluding particular studies are shown in Figure 1.

Quality Appraisal
The Joanna Briggs Institute (JBI) critical appraisal checklist for analytical observational studies [42] was used to assess the methodological quality of the included studies. This tool is specifically designed for the assessment of cross-sectional studies. The checklist includes eight items for appraisal of the following: clarity of inclusion criteria, an adequate description of the study subject and setting, validity and reliability of measurement, whether measurements of conditions were objective and standardized, identification of confounding factors, strategies to account for confounders, validity of outcome measurement, and appropriate use of statistics. The JBI checklist for cohort studies consists of eleven items addressing the representativeness of included participants, the representativeness and validity of exposure measurement, whether and how confounding factors were adjusted for the validity of outcome measurement, whether participants were outcome-free at the start of the study, the adequacy and completeness of follow-up, the strategies used to address incomplete follow-up, and the appropriate use of statistics. In both checklists, the following four options are provided for each item: "Yes", "No", "Unclear" and "Not applicable". The total checklist score ranges from 0 to 8 for cross-sectional studies and from 0 to 11 for cohort studies, with a higher score indicating higher quality; however, no cutoff point is provided to definitively identify the quality of studies [43].

Quality Appraisal
The Joanna Briggs Institute (JBI) critical appraisal checklist for analytical observational studies [42] was used to assess the methodological quality of the included studies. This tool is specifically designed for the assessment of cross-sectional studies. The checklist includes eight items for appraisal of the following: clarity of inclusion criteria, an adequate description of the study subject and setting, validity and reliability of measurement, whether measurements of conditions were objective and standardized, identification of confounding factors, strategies to account for confounders, validity of outcome measurement, and appropriate use of statistics. The JBI checklist for cohort studies consists of eleven items addressing the representativeness of included participants, the representativeness and validity of exposure measurement, whether and how confounding factors were adjusted for the validity of outcome measurement, whether participants were outcome-free at the start of the study, the adequacy and completeness of follow-up, the strategies used to address incomplete follow-up, and the appropriate use of statistics. In both checklists, the following four options are provided for each item: "Yes", "No", "Unclear" and "Not applicable". The total checklist score ranges from 0 to 8 for cross-sectional studies and from 0 to 11 for cohort studies, with a higher score indicating higher quality; however, no cutoff point is provided to definitively identify the quality of studies [43].

Data Extraction
Data were independently extracted from the studies by TJ and WM. Cooper's review matrix method was employed to assist in the analysis of the data [44]. We summarized the following data from the studies: author(s)/year, sample, country, the CD-RISC version (with reliability or internal consistency information expressed as Cronbach's alpha), number of participants, and correlation coefficients. In addition, study data regarding measures of psychological distress (depression, stress, anxiety, and PTSD), their reliability range, and the number of associations identified (K) were summarized.

Meta-Analysis
R software version 4.1.2 was used for all analyses. The pooled correlation coefficient between resilience and psychological distress was calculated using the values of correlation coefficients obtained in each study and employing the "metacor" package [45]. Correlation coefficient values were generated along with 95% confidence intervals (CI). A randomeffects model was used for the pooled analysis to account for unmeasured heterogeneity between studies. Correlations were classified as poor (correlation coefficient r < 0.20), average (r = 0.20-0.39), moderate (r = 0.40-0.59), strong (r = 0.60-0.79), and very strong (r ≥ 0.80) [46]. Publication bias was visually assessed using Begg's funnel plots generated by the "metabias" package [47] and statistically assessed using Egger's test in the "funnel.meta" package. The heterogeneity of r values between studies was tested by estimating a Cochran's Q statistic and an inconsistency index (I 2 statistic), with I 2 > 50% indicating substantial heterogeneity. Where heterogeneity was substantial, a subgroup analysis, which is superior to meta-regression [48], was performed for all the studies to further investigate the heterogeneity issue.

Quality Appraisal
Two authors (TJ and WM) independently appraised the quality of the 33 included studies using two versions of the JBI critical appraisal checklist based on the study design. For the 31 cross-sectional studies, scores ranged from 4 to the maximum of 8 (median score: 8/8, interquartile range: 7/8-8/8; Supplementary Table S2). The two cohort studies scored 7 and 11 (Supplementary Table S3). For 10 studies, the two authors (TJ and WM) assigned inconsistent scores; consequently, another author (CD) reviewed the scores and assisted them in reaching a consensus. The results of the quality appraisal are detailed in Supplementary Tables S2 and S3.

Resilience
The included studies measured resilience, an assessment of stress coping ability employed to target treatment for anxiety, depression, and stress reactions, using various versions of CD-RISC [81]. Specifically, studies employed  (Table 1).

Psychological Distress
The scales used to measure the four psychological distress variables, their range of reliability, and the number of associations with resilience (k) identified using each scale is shown in Supplementary Table S4. Most of the studies measured psychological distress using surveys containing scales for mental illness (k = 4), depression (k = 18), anxiety (k = 22), stress (k = 14), and PTSD (k = 2).

Relationship between Resilience and Psychological Distress
Based on this meta-analysis, a moderate negative relationship was detected between resilience and psychological distress (r = −0.42; 95% CI: −0.45 to −0.38; p < 0.001). Studies assessing this relationship showed high heterogeneity in their outcomes (I 2 = 97.7%). Findings regarding associations between resilience and psychological distress are summarized in Table 1.

Healthcare Workers
The forest plot for meta-analysis of the relationship between resilience and psychological distress among healthcare workers is provided in Figure 2. For the healthcare worker subgroup, a moderately significant negative relationship was found between overall resilience and psychological distress (r = −0.39; 95% CI: −0.44 to −0.33; p < 0.001). The heterogeneity of the effect sizes was high (I 2 = 94%).

General Population
The forest plot for meta-analysis of the relationship between resilience and psychological distress in the general population is shown in Figure 3. For the general population subgroup, a moderately significant negative relationship was found between overall resilience and psychological distress (r = −0.45; 95% CI: −0.52 to −0.38; p < 0.001). The heterogeneity of the effect sizes was high (I 2 = 99%).

General Population
The forest plot for meta-analysis of the relationship between resilience and psychological distress in the general population is shown in Figure 3. For the general population subgroup, a moderately significant negative relationship was found between overall resilience and psychological distress (r = −0.45; 95% CI: −0.52 to −0.38; p < 0.001). The heterogeneity of the effect sizes was high (I 2 = 99%).

General Population
The forest plot for meta-analysis of the relationship between resilience and psychological distress in the general population is shown in Figure 3. For the general population subgroup, a moderately significant negative relationship was found between overall resilience and psychological distress (r = −0.45; 95% CI: −0.52 to −0.38; p < 0.001). The heterogeneity of the effect sizes was high (I 2 = 99%).

Patients
The forest plot for meta-analysis of the relationship between resilience and psychological distress among patients is shown in Figure 4. For the patient subgroup, a moderately significant negative relationship was found between overall resilience and psychological distress (r = −0.43; 95% CI: −0.52 to −0.33; p < 0.001). The heterogeneity of the effect sizes was high (I 2 = 74%) but lower than those of the other subgroups.

Patients
The forest plot for meta-analysis of the relationship between resilience and psychological distress among patients is shown in Figure 4. For the patient subgroup, a moderately significant negative relationship was found between overall resilience and psychological distress (r = −0.43; 95% CI: −0.52 to −0.33; p < 0.001). The heterogeneity of the effect sizes was high (I 2 = 74%) but lower than those of the other subgroups.

Publication Bias
The funnel plot for publication bias is provided in Figure 5. Based on publication bias analysis, visual evaluation of the funnel plot revealed that the distribution of the studies deviated from the funnel, which one would normally expect in the absence of publication bias. Therefore, this figure provides no visual indication of skewedness of the effect sizes observed. The left-sided test for funnel plot asymmetry using Egger's regression test was not significant (p = 0.179), supporting the conclusion that no significant publication bias was present.

Publication Bias
The funnel plot for publication bias is provided in Figure 5. Based on publication bias analysis, visual evaluation of the funnel plot revealed that the distribution of the studies deviated from the funnel, which one would normally expect in the absence of publication bias. Therefore, this figure provides no visual indication of skewedness of the effect sizes observed. The left-sided test for funnel plot asymmetry using Egger's regression test was not significant (p = 0.179), supporting the conclusion that no significant publication bias was present.
analysis, visual evaluation of the funnel plot revealed that the distribution of the studies deviated from the funnel, which one would normally expect in the absence of publication bias. Therefore, this figure provides no visual indication of skewedness of the effect sizes observed. The left-sided test for funnel plot asymmetry using Egger's regression test was not significant (p = 0.179), supporting the conclusion that no significant publication bias was present.

Resilience and Psychological Distress
Based on the meta-analysis, we found a moderate negative relationship between resilience and psychological distress across populations during the COVID-19 pandemic (pooled r = −0.42; 95% CI: −0.45 to −0.38; p < 0.001). In other words, during the pandemic, the higher an individual's resilience, the lower the psychological distress. The results indicate that resilience is essential in promoting a person's positive mental health and reducing negative consequences. Our results align with a model of resilience hypothesizing that resilience supported mental health through risk reduction, protection, and promotion before the COVID-19 pandemic [82] and during the SARS pandemic [83]. More specifically, resilience reduces the depression, stress, anxiety, and PTSD associated with exposure to the COVID-19 pandemic. In addition, resilience appears to be a protective factor against adverse events and promotes a person's ability to cope with COVID-19. Individuals with high resilience may have good tolerance of negative feelings, a strong capacity for selfreflection, and a high sense of responsibility, all characteristics that can promote better coping with psychological distress [84].
This relationship is similar to that found in previous systematic reviews and metaanalysis studies conducted before COVID-19. Färber and Rosendahl [29] reported a negative correlation of −0.43 (95% CI: −0.39 to −0.48; p < 0.001) between resilience and mental health problems in patients with a somatic illness or health condition. In addition, in the general population, Hu et al. [85] found that trait resilience was negatively correlated with negative indicators of mental health (mean effect size: −0.36, 95% CI: −0.37 to −0.35) and was positively correlated with positive indicators of mental health (mean effect size 0.53, 95% CI: 0.49 to 0.51). Furthermore, Joyce et al. [86] observed that a high level of resilience was associated with lower levels of anxiety, psychological distress, and depression.
Our results showed that the negative relationship between resilience and psychological distress was weaker among healthcare workers than in the general population and patients. Previous studies have shown that resilience among healthcare workers was lower than in the general population [87,88]. One explanation for these findings is that healthcare workers frequently experience stress and burnout and thus are more likely to have low resilience [89,90]. Based on the results, more attention should be focused on healthcare workers showing lower levels of resilience to help them better cope with the negative effects of mental health issues.

Healthcare Providers
Pooled analysis revealed that overall resilience was significantly negatively correlated with psychological distress among healthcare providers (pooled r = −0.39; 95% CI: −0.44 to −0.33; p < 0.001); however, the correlation was lower than pre-pandemic (r = −0.61) [91]. Studies reported that resilience was significantly negatively related to psychological distress among female nurses [92] and rescue workers [93]. A possible explanation for this association is that resilience can emerge as the ability to take full advantage of their positive personal characteristics despite stressful occupational circumstances. Moreover, healthcare providers' experiences in providing care for COVID-19 patients and their family members, their access to expert colleagues and resources, their psychological knowledge, and their relatively high level of education [28] may help them to cope with psychological distress.

General Population
For the general population subgroup, a moderately significant negative relationship was found between overall resilience and psychological distress (pooled r = −0.45; 95% CI: −0.52 to −0.38; p < 0.001), which was somewhat weaker than the association found in a pre-pandemic study by Ghanei Gheshlagh et al. [30]. Those researchers found a moderate but stronger significant negative correlation between resilience and mental health issues in general populations (r = −0.54; 95% CI: 0.49 to 0.59; p = 0.0001). Other studies' results also showed that the negative correlation between resilience and mental health in the elderly population and among business professionals pre-pandemic (r = −0.63 and −0.55, respectively) was greater than during the pandemic [94,95]. The weaker negative correlation between resilience and psychological distress observed in the general population during the pandemic may be attributable to their having less social support than previously. One attribute of resilience is social support [96], and given the preventive measures and isolation experienced during the pandemic, members of the general population may well have been denied contact with a supportive community of family members, friends, and coworkers.

Patients
We found a moderately significant negative relationship between overall resilience and psychological distress in the patient subgroup (pooled r = −0.43; 95% CI: −0.52 to −0.33; p < 0.001). Similarly, a previous pre-pandemic study of somatically ill patients [29] reported that resilience was significantly negatively associated with psychological distress (r = −0.43; 95% CI: −0.48 to −0.39; p < 0.001). In addition, Cal et al.'s pre-pandemic study identified a significant negative relationship between resilience and mental health problems among patients with chronic illness [97]. In chronic illness patients, resilience may be a capacity that is developed over time in response to the stressors and hardships of contending with chronic disease. That is, compared to patients with acute illness, chronic illness patients tend to have higher resilience because in coping with their illness over the long term, they have time to adapt to their disease both physically and mentally [97].

Study Limitations
This study has limitations that should be noted. First, only research studies reporting correlational coefficients and employing the CD-RISC self-report measure were included, and only English-language publications were used; consequently, other relevant studies may have been inadvertently excluded. Second, only two of the included studies were cohort studies; the rest were cross-sectional and thus supplied only a snapshot of the existing situation with little or no longitudinal data. Third, the included studies measured psychological distress using various self-report measures and scales; given the heterogeneity of these measures, the pooled estimates should be interpreted with caution. Fourth, our search for relevant articles focused on publications from 1 December 2019 to 1 June 2021, but the first wave of the COVID-19 pandemic began and ended at various times, depending on the specific country and healthcare system involved. Fifth, we were unable to meaningfully compare the association between psychological distress and resilience before and after the COVID-19 pandemic, although based on previous systematic reviews, the association does not appear to have changed significantly. Future studies should examine the impact of COVID-19 on this association. Finally, this study focused on adult and older adult populations, and thus its findings may not be generalizable to child and adolescent populations.

Research and Clinical Implications Format
This study's findings shed light on the need to develop interventions for enhancing resilience among healthcare providers, the general population, and patients to decrease the long-term impacts of psychological distress. In clinical practice, these populations should receive psychosocial support during health emergencies such as COVID-19 and other infectious disease outbreaks. As an example, they could be provided with consultations with a psychologist to promote their resilience and reduce their psychological burden. Where healthcare providers are concerned, this approach might also reduce turnover rates and thus benefit the overall healthcare system.

Conclusions
This systematic review and meta-analysis identifies a moderate negative relationship between resilience and psychological distress among healthcare workers, the general population, and patients in the COVID-19 context, although this association seems weaker than that found in the pre-pandemic period. In addition, this negative relationship was somewhat weaker among healthcare workers than was observed in the general population and patients. For all three populations, psychosocial support is needed to improve resilience and the ability to cope with psychological distress during the COVID-19 pandemic and in future disease outbreaks. On the whole, this study's findings emphasize the need to develop specific interventions to enhance resilience in these populations.

Supplementary Materials:
The following supporting information can be downloaded at https: //www.mdpi.com/article/10.3390/ijerph192214854/s1, Table S1: Search strategies; Table S2: JBI critical appraisal checklist for analytical cross-sectional studies; Table S3: The quality appraisal for cohort studies; Table S4: Instruments to measure psychological distress.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.