Turning to Religion During COVID-19 (Part II): A Systematic Review, Meta-analysis and Meta-regression of Studies on the Relationship between Religious Coping and Mental Health throughout COVID-19

The COVID-19 pandemic and the many associated socio-economic changes constitute a stressful event that required adaptation to new, dynamic, and often threatening conditions. According to the literature, coping strategies are one of the factors that determine a person’s degree of adaptation to stressful situations. A systematic review and meta-analysis was performed on the relationship between religious coping and selected indicators of mental health. Due to the large amount of data, this work has been divided into two parts: Part I discussed the positive mental health indicators (Pankowski & Wytrychiewicz-Pankowska, 2023), while this Part II discusses negative mental health indicators. A systematic review of the databases of Science Direct, EBSCO, Cochrane, PubMed, and Google Scholar identified 33 articles related to the severity of depressive symptoms: 30 to anxiety, 23 to stress, 1 related to PTSD symptoms and peritraumatic stress, and 5 related to general negative mental health. The limitations of the research as well as further directions for exploration are discussed. Clinical trial registration This Review was pre-registered at OSF: osf.io/54ygr (https://doi.org/10.17605/OSF.IO/GMNFV).


Introduction
Mental health (MH) issues are one of the greatest difficulties facing the healthcare system today (McCartan et al., 2021;Pfefferbaum & North, 2020). More and more studies and analyses are indicating the negative impact of mental health issues on, inter alia, work (Kessler et al., 2006), relationships (Sharabi et al., 2016), and other areas of life (Lépine & Briley, 2011). The COVID-19 pandemic has also led to the emergence and intensification of a variety of types of mental difficulties, through socio-economic factors (Agberotimi et al., 2020;Kourti et al., 2021;Lindau et al., 2021) as well as through psychological mechanisms (Cénat & Dalexis, 2020;Coelho et al., 2020).
Research conducted in various periods of the pandemic around the world has clearly shown the threat that it poses to mental well-being. Data from studies and reviews indicated a very high prevalence of, inter alia, depressive symptoms (Necho et al., 2021), anxiety (Kan et al., 2021), the presence of PTSD symptoms (Zhang et al., 2021), and peritraumatic stress (Jiménez et al., 2021). The above difficulties were the subject of research both in specific populations who were particularly exposed to stress resulting from, for example, the nature of their work (healthcare workers, in particular; Salari et al., 2020a), the risk of a more severe course of COVID-19 (Yan et al., 2022), but also the general population (Salari et al., 2020b). In addition to determining the scale of the problem, numerous analyses have focused on trying to identify factors that could buffer such negative responses to pandemic stress.
According to the theoretical models used in research on stress (Lazarus & Folkman, 1984; a more detailed description is provided in the first part of the review), coping strategies may be very important for MH in stressful situations. Earlier studies conducted in various populations confirmed the importance of coping strategies for MH, including the severity of depressive symptoms, anxiety, and stress. It should also be noted that coping strategies are a modifiable factor that can be shaped by interventions.
This part of the review focuses on negative mental health indicators: the relationships of religious coping (RC) with the severity of depressive symptoms, anxiety, stress levels, the severity of peri-/post-traumatic stress symptoms, and general MH indicators was analysed in more detail. The aim was to synthesize information from both cross-sectional and longitudinal studies and to try to determine the strengths of the relationships and the factors that may be responsible for variability in this area.

Methods
Detailed data on search strategy, selection criteria, data extraction, quality assessment, and statistical analysis can be found in the first part of the review: Turning to Religion During COVID-19: A Systematic Review, Meta-analysis and Meta-regression of Studies on the Relationship between Religious Coping and Mental Health throughout COVID-19 (Part I) (Pankowski & Wytrychiewicz-Pankowska, 2023). Information on the quality of studies included in the review can be found in Appendix 1, funnel plots in Appendix 2 and number of studies and participants conducted per country in Appendix 3.

Results
Detailed data on the review, including general information (number of people who participated in the research, broken down by country, etc.), can be found in the first part of the review; this part focuses on the description of studies describing the relationship between RC and negative MH indicators. Effect sizes obtained in metaanalyses were also transformed into Cohen's d, CLES (Common Language Effect Size) and Odds ratio (Appendix 4).

Severity of Depressive Symptoms and Religious Coping
First, the relationship between the severity of depressive symptoms and RC was analysed. A total of 33 articles analysing this relationship were identified. The research was conducted from November 2019 to August 2021 and a total of 24,644 people participated. Various methods were used to assess RC, including the Brief Coping Orientation to Problems Experienced Inventory (Brief-COPE), Brief Religious Coping Orientation to Problems Experienced Inventory (Brief-RCOPE), Spiritual/ Religious Coping Scale , and others. In the case of severity of depressive symptoms, various methods were also used: the Patient Health Questionnaire-9 (PHQ-9), Depression, Anxiety, and Stress Scale , and the Beck Depression Inventory (BDI). Almost half of the studies conducted indicated no relationship between RC and the severity of depressive symptoms. A summary of the results is shown in Table 1.
For a more detailed analysis of the relationship between these two variables, we performed a meta-analysis. For inclusion of a measure in the analysis, we required that it appears in at least 3 surveys. A thorough analysis of the tools used allowed for the performance of a meta-analysis of the relationship between depressive symptoms assessed with DASS-21 and RC assessed with Brief-COPE (4 studies).

Meta-analysis
The meta-analysis conducted for the relationship between severity of depressive symptoms (DASS-21) and RC (Brief-COPE) included three studies. Studies identified in the literature search as meeting the inclusion criteria were pooled to give a correlation − 0.15 [− 0.23; − 0.06] which was statistically significant (Z = − 3.18; p < 0.01) (Fig. 1). Statistically significant heterogeneity was observed between studies (Q = 24.01; p < 0.001). The estimated amount of total heterogeneity was Tau2 = 0.007 and I 2 = 87.5%. (2023)  Due to the high heterogeneity of the results, the percentage of women was analysed as potential moderator. Unfortunately, due to deficiencies in the reported data, it was not possible to include more moderators. The tests for moderators showed that the percentages of women in the study (QM (1) = 0.55 non-significant) were not statistically significant moderator in the performed studies.

Journal of Religion and Health
Next, the relationship between nRC and the severity of depressive symptoms (HADS) was analysed. The meta-analysis conducted for the relationship between severity of depressive symptoms (HADS) and nRC (Brief-RCOPE) included three studies. Studies identified in the literature search as meeting the inclusion criteria were pooled to give a correlation 0.17 [0.13; 0.22] which was statistically significant (Z = 7.44; p < 0.001) (Fig. 2). Statistically significant heterogeneity was not observed between studies (Q = 1.35; p > 0.05). The estimated amount of total heterogeneity was Tau2 = 0 and I 2 = 0%.
Then, the relationship between pRC and the severity of depressive symptoms (HADS) was analysed. The meta-analysis conducted for the relationship between severity of depressive symptoms (HADS) and pRC (Brief-RCOPE) included three studies. Studies identified in the literature search as meeting the inclusion criteria were pooled to give a correlation − 0.06 [− 0.12; 0.00] which was not statistically significant (Z = − 1.81; p > 0.05) (Fig. 3). Statistically significant heterogeneity was not observed between studies (Q = 2.94; p > 0.05). The estimated amount of total heterogeneity was Tau2 = 0.0009 and I 2 = 32.1%.

Severity of Anxiety Symptoms and Religious Coping
Next, the relationship between the severity of anxiety symptoms and RC was analysed. A total of 30 studies analysing the relationship between anxiety and RC were identified. The research was conducted from February 2020 to August 2021 and a total of 21,368 people participated in it. Furthermore, a variety of methods were used to assess both anxiety and RC, including the Brief-COPE, Brief-RCOPE, SRCOPE-14, and others. Similarly for anxiety, a variety of methods were used, including DASS-21, HADS, and BAI. Almost half of the studies conducted indicated no relationship between RC and the severity of anxiety symptoms. A summary of the results is shown in Table 2.

Meta-analysis
The analysis of the studies included in the review allowed us to perform three metaanalyses concerning the relationship between RC assessed with Brief-COPE and Brief-RCOPE and the severity of anxiety assessed with HADS.
The meta-analysis conducted for the relationship between severity of anxiety (DASS-21) and RC (Brief-COPE) included four studies. Studies identified in the literature search as meeting the inclusion criteria were pooled to give a correlation 0.00 [− 0.08; 0.07] which was not statistically significant (Z = − 0.04; p > 0.05) (Fig. 4). Statistically significant heterogeneity was observed between studies (Q = 17.58; p < 0.01). The estimated amount of total heterogeneity was Tau2 = 0.04 and I 2 = 83%.
Due to the high heterogeneity of the results, the percentage of women was analysed as potential moderator. Unfortunately, due to deficiencies in the reported data, it was not possible to include more moderators. The tests for moderators showed that the percentage of women in the study (QM (1) = 0.12; non-significant) was not statistically significant moderator.
Subsequently, the relationship between nRC and the severity of anxiety assessed with HADS was analysed. The meta-analysis conducted for the relationship between severity of anxiety (HADS) and nRC (Brief-RCOPE) included three studies. Studies identified in the literature search as meeting the inclusion criteria were pooled     (1) and end (30) of the month were used as the starting and ending points to give a correlation 0.26 [0.18; 0.33] which was statistically significant (Z = 6.66; p < 0.001) (Fig. 5). Statistically significant heterogeneity was not observed between studies (Q = 3.50; p > 0.05). The estimated amount of total heterogeneity was Tau2 = 0.002 and I 2 = 43%. Potential moderators were also analysed: the percentage of women and relationship status (the percentage of married people). Unfortunately, due to deficiencies in the reported data, it was not possible to include more moderators. The tests for moderators showed that the percentage of women in the study (QM (1) = 2.88; p = 0.09) and the percentage of married people (QM (1) = 2.98; p = 0.08) were at the statistical trend level.
Finally, we performed a meta-analysis concerning the relationship between pRC and levels of anxiety. The meta-analysis conducted for the relationship between severity of anxiety (HADS) and pRC (Brief-RCOPE) included three studies. Studies identified in the literature search as meeting the inclusion criteria were pooled to give a correlation − 0.02 [− 0.21; 0.17] which was not statistically significant (Z = − 0.21; p > 0.05) (Fig. 6). Statistically significant heterogeneity was observed between studies (Q = 17.53; p < 0.01). The estimated amount of total heterogeneity was Tau2 = 0.03 and I 2 = 87%.
Due to the high heterogeneity of the results, potential moderators were also analysed in more detail: the percentage of women and relationship status (the percentage of married people). The tests for moderators showed that the percentage

Stress
In the next step, the relationship between stress level and RC was analysed. A total of 23 studies meeting the criteria were identified. The research was conducted from March 2020 to October 2020. A total of 16,557 people participated in the research. In the vast majority of studies, the Brief-COPE was used to assess RC, and stress levels were measured with the Perceived Stress Scale (PSS) and DASS-21. In over half of the studies (12), no significant statistical relationship was found between RC and the level of stress. A summary of the results is shown in Table 3.

Meta-analysis
Analysis of the research allowed us to conduct one meta-analysis regarding the relationship between RC assessed with Brief-COPE and the intensity of stress assessed with DASS-21. Unfortunately, despite the large number of studies assessing the relationship between stress measured with PSS and RC, it was not possible to obtain the data needed to perform the calculations (correlation coefficients).
The meta-analysis conducted for the relationship between level of stress (DASS-21) and RC (Brief-COPE) included four studies. Studies identified in the literature search as meeting the inclusion criteria were pooled to give a correlation − 0.06 [− 0.09; − 0.03] which was not statistically significant (Z = − 1.80; p > 0.05) (Fig. 7). Statistically significant heterogeneity was observed between studies (Q = 14.86; p < 0.01). The estimated amount of total heterogeneity was Tau2 = 0.04 and I 2 = 80%.
Due to the high heterogeneity of the results, the percentage of women was analysed as potential moderator. The tests for moderators showed that the percentage of women in the study (QM (1) = 0.16; non-significant) was not statistically significant moderator of the conducted research.    *In studies that did not specify the exact dates on which data were collected (months only), the beginning (1) and end (30) of the month were used as the starting and ending points

Peri-and Post-traumatic Stress Disorder Symptoms and General Mental Health Indicators
Lastly, the strength of the relationship between RC and the severity of peri-(n = 1) and post-traumatic (n = 1) stress disorder and general mental health indicators (n = 5) was analysed. The research was conducted from March 2020 to June 2020. In the RC assessment studies, Brief-COPE and Brief-RCOPE were mainly used. A variety of tools were used to assess negative mental health indicators: HADS, DASS, and GHQ-12 overall scores. A summary of the results is shown in Table 4. Due to the large variety of tools used, it was not possible to perform a meta-analysis.

Discussion
This second part of the systematic review focused on the relationship between Religious Coping (RC) and negative mental health (MH) indicators: severity of depressive symptoms, anxiety, stress, symptoms of peri-and post-traumatic stress disorder, and general negative indicators of MH. The vast majority of studies were cross-sectional and approximately half of them indicated no relationship between RC and the analysed variables. Despite the large number of studies included in the review, few could be analysed further. The meta-analyses performed included only 3-4 studies and therefore should be interpreted with caution. For meta-analysis of studies using Brief-COPE, in which RC is operationalized in a neutral manner, the results were characterized by statistically significant heterogeneity, which could not be explained with the use of possible moderators. Conversely, meta-analyses of the relationships of positive RC (pRC) and negative RC (nRC) with negative MH indexes.
The vast majority of studies on the relationship between RC and the severity of depressive symptoms were cross-sectional. Data were collected in different countries, at different stages of the pandemic, and in diverse populations. Our attempt to synthesize the results shows that they are very inconclusive: about half of the studies indicate no relationship, which is also confirmed by the results of two longitudinal studies; many analyses also indicate that this relationship is negative.  *In studies that did not specify the exact dates on which data were collected (months only), the beginning (1) and end (30) of the month were used as the starting and ending points We further explored this relationship with the use of meta-analyses, which indicated negative correlation (Brief-COPE), positive correlation (nRC) and no relationship (pRC) between the analysed variables. The first of them, concerning the neutral RC (Brief-COPE), was characterized by a very high heterogeneity of the results, while none of the possible moderators turned out to be statistically significant. As in the first part of the systematic review, attention should be paid to both the method of reporting the results (in the form of a correlation matrix) and the description of the studied sample, which were very different between the analysed studies. Further calculations with nRC and pRC showed a positive and lack of relationship, respectively, with MH. In the case of these analyses, the results were homogeneous, but they were based on very little data (n = 3), which significantly limits the possibility of generalizing these results. To sum up, it is worth considering a similar direction of dependence as in the case of QoL (see first part of the review), which suggested a greater influence of mood on the strategy chosen than the influence of a given strategy on the severity of depressive symptoms. At this point, it is worth emphasizing that the content of the items contained in the Brief-RCOPE (p/nRC) may reflect the effects of the coping process to a greater extent than the respondents' approaches to stressful situations (see also first part of review). Regarding levels of anxiety and RC, the results of the studies are very similar to those obtained for intensification of depressive symptoms, which may be due to the strong correlation between these two variables. The studies included in the review were conducted on a wide variety of populations, in many countries, and using a variety of methodologies. Longitudinal studies showed no correlation between RC and anxiety, as did a large proportion of cross-sectional studies. The meta-analysis of studies assessing the relationship between Brief-COPE and the level of anxiety found no relationship, but it was characterized by high heterogeneity of the results which could not be explained by moderators. In turn, nRC was characterized by a positive relationship with anxiety levels, and the result was homogeneous. This result should be interpreted with caution due to the small number of studies included. In the case of pRC, large heterogeneity of the results was noted: some results indicated a positive relationship, others a negative relationship. Again, the moderators that could be considered did not explain this variation between surveys. It should be noted that the meta-analyses of the relationship between RC and anxiety included the same studies as in the case of the intensification of depressive symptoms.
Next, the relationship between stress levels and RC was analysed. As before, the studies included in the review were conducted in many countries, populations, and throughout the pandemic period. The different methodology used to assess RC and stress made it difficult to more accurately analyse the relationship between the variables. All studies included in the review were cross-sectional, and half of them indicated no relationship between the analysed variables. Due to the ambiguous results of our investigation, it was decided to carry out a metaanalysis, which also indicated a very large diversity in the results. The statistically significant heterogeneity was not explained by the moderators used in the meta-regression. Summing up, the analysis of studies identified by the review indicated no relationship or a very weak relationship between RC and levels of stress, which perhaps suggests that RC may have a different function, not necessarily related only to stress reduction.
Lastly, studies on peri-and post-traumatic stress and general negative MH indicators were analysed. In the case of the first two variables, we found only one study each; it is therefore not possible to generalize these results. No relationship was found with peritraumatic stress disorder, and a study of post-traumatic stress disorder symptoms showed a positive relationship. On the other hand, studies on the relationship between RC and general negative MH indicators partially indicated a positive relationship in studies that used DASS-21, but no relationship for studies that used MHI-5 and HADS. In the case of the above studies, it was not possible to analyse the results in more depth, and the small number of studies makes it difficult to generalize. In conclusion, studies on the relationship between RC and negative mental health indicators suggest that this strategy has little or no protective effect. Moreover, some studies even indicate negative mental health effects associated with RC. In view of the positive relationship between RC and PTSD, it should also be considered to what extent the relationship between RC and PTG suggested in first part of review may be due to spiritual bypassing (see: Cashwell et al., 2010). It should be emphasized that only one study on the relationship between PTSD and RC could be identified; therefore, the relationships of RC with PTG, PTSD, and spiritual bypassing require further in-depth analysis. As indicated in the first part of the review, most analyses focused on the role of a given strategy in the variable-centred approach; it is possible that RC may be effective in a specific group of people or in a specific configuration of the strategies used, but it would require more in-depth analysis to confirm this.

Study Limitations
Similarly, as noted in the first part of the review, the cross-sectional nature of most of the analysed studies prevents conclusions about the impact of RC on MH. The results show the frequency of using a specific strategy with the simultaneous level of MH indicators and not the effect of the coping process. The conducted review also has several limitations. The most important is the small number of studies included in the meta-analyses, especially when the same studies were analysed for different MH indicators. In addition, as noted earlier, inconsistent descriptions of sample structures and the absence of correlation coefficients between the analysed variables significantly limited possibilities for further data analysis. Thus, we appeal to authors to consider the variables describing the studied population as broadly as possible: not only age or gender, but also others that may be relevant, because the obtained results may differ radically in studies conducted on different populations. Furthermore, reporting correlation coefficients between the variables examined would facilitate further meta-analysis of results. Unfortunately, due to missing data, it was not possible to perform a network meta-analysis.

Conclusions
The collected longitudinal data suggest that religious coping is not an effective method of coping with stress caused by the pandemic, especially in terms of negative mental health indicators. Data from cross-sectional studies suggest that this strategy, in particular negative coping, is used in connection with increased depressive or anxiety symptoms. A large proportion of studies failed to find any relationship between these two variables, as did individual longitudinal studies. It is possible that RC may play a different role that was not considered in the review, such as reducing the fear of death (Freh & Cheung Chung, 2021 Lopes and Nihei (2021)

Appendix 2: Funnel Plots of Analysed Studies
The results should be interpreted with caution due to the small number of studies (Figs. 8a, b, 9a-g).

Appendix 3
See Table 5. Fig. 9 a Religious coping and severity of depressive symptoms: funnel plot. b Relationship between negative religious coping and severity of depressive symptoms: funnel plot. c Relationship between positive religious coping and severity of depressive symptoms: funnel plot. d Religious coping assessed with Brief-COPE and anxiety: funnel plot. e Negative religious coping and anxiety: forest plot. f Positive religious coping and anxiety: forest plot. g religious coping and level of stress: forest plot
Author Contributions DP involved in conceptualization, methodology, formal analysis, investigation, resources, data curation, writing-original draft, writing-review and editing, visualization, supervision, project administration, and funding acquisition. KW-P involved in investigation, writing-review and editing, and funding acquisition.
Funding Proofreading of this work was supported by authors own funds.