3.1 Study Characteristics
A total of 168 countries were included in the search process when looking for articles covering the various forms of SWA. However, the final studies included as part of the systematic review were derived from 27 countries (Philippines, Myanmar, Malaysia, Cambodia, Indonesia, China, Thailand, Vietnam, Hong Kong, South Africa, Nigeria, Ethiopia, Ghana, UAE, Turkey, Brazil, Bangladesh, India, Egypt, Israel, Lebanon, Iraq, Iran, Mexico, Saudi Arabia, Oman, South Korea). A total of 42 articles were finally included in a systematic review with 11 articles from India, five articles from Brazil, four articles each from Ethiopia, and China, and three each from Nigeria, Saudi Arabia, and Israel. While Indonesia, Thailand, Vietnam, South Africa, Lebanon, Mexico, and Oman produced two articles each, the Philippines, Myanmar, Malaysia, Cambodia, Hong Kong, Ghana, Turkey, Egypt, Iraq, and South Korea each accounted for one article each.
A total of 31 articles were considered for the meta-analysis. Although the ‘net’ to obtain the required articles was spread throughout the databases and the systematic review did yield conditions beyond social anxiety disorder (including social withdrawal, loneliness, hikikomori, and shyness), the meta-analysis itself left us with only social anxiety disorder because we wanted to provide the most meaningful results. The following paragraphs will elaborate on the pooled prevalence of social anxiety disorders in the Global South, categorized according to the types of tools used to obtain the required data.
3.2 The pooled prevalence of SAD assessed by the SPIN
The meta-analysis (Fig. 2a) displays the pooled prevalence of SAD assessed by the SPIN in 14 studies (49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62). The pooled data included 7669 samples, of which 1727 had SAD, and its heterogeneity was high (I2 = 98.64%) and significant (Q = 956.56, p < 0.001). Therefore, the random-effects model was used, showing that the pooled prevalence rate of SAD was 28.76% (95% CI: 20.19%-38.17%). Publication bias was observed from the funnel plot (Fig. 2b). A sensitivity analysis was performed and suggested that seven studies (49, 51, 54, 55, 58; 59, 61) be excluded after examining by Egger’s test (Intercept=-2.28, p = 0.690), Begg’s test (Kendall’s Tau=-0.429, p = 0.177) and the funnel plot (Fig. 2c). The pooled prevalence rates changed to 24.79% (95% CI 20.46%-29.39%) with 858 SAD cases out of 3371 samples for seven articles (50, 52, 53, 56, 57, 60, 62).
3.3 The pooled prevalence of SAD assessed by the SIAS
The meta-analysis (Fig. 3a) displays the pooled prevalence of SAD assessed by the SIAS in three studies (63, 64, 65). The pooled data included 7352 samples, of which 1883 had SAD, and its heterogeneity was high (I2 = 93.29) and significant (Q = 29.822, p < 0.001). Therefore, the random-effects model was used, showing that the pooled prevalence rate of SAD was 24.60% (95% CI: 16.65%-33.52%). The funnel plot was not affected by publication bias (Fig. 3b) after being examined by Egger’s test (Intercept = 0.063, p = 0.992) and Begg’s test (Kendall’s Tau = 0.333, p = 0.602), hence no sensitivity analysis was performed.
3.4 The pooled prevalence of SAD assessed by SCARED
The meta-analysis (Fig. 4a) displays the pooled prevalence of SAD assessed by a sub-scale of SCARED in two studies (64, 66). The pooled data included 1178 samples, with 198 having SAD; its heterogeneity was high (I2 = 85.83) and significant (Q = 7.058, p = 0.008). Therefore, the random-effects model was used, showing that the pooled prevalence rate of SAD was 17.15% (95% CI: 11.78%-23.29%). The funnel plot did not show publication bias (Fig. 4b) after examining Egger’s test (Intercept = 24.836, p = 0.103) and Begg’s test (Kendall’s Tau=-0.763, p = 0.317), hence no sensitivity analysis was performed.
3.5 The pooled prevalence of SAD assessed by the LSAS
The meta-analysis (Fig. 5a) shows the pooled prevalence of SAD assessed by LSAS in 11 studies (67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77). The pooled data included 7777 samples, with 4123 having SAD, and its heterogeneity was high (I2 = 99.83%) and significant (Q = 5978.507, p < 0.001). Therefore, the random-effects model was used, showing that the pooled prevalence rate of SAD was 44.96% (95% CI: 19.32%-72.13%). Publication bias was observed from the funnel plot (Fig. 5b). A sensitivity analysis was then performed and suggested that 7 studies (67, 69, 70, 72, 73, 74, 76) should be excluded after examining by Egger’s test (Intercept = 14.56, p = 0.228), Begg’s test (Kendall Tau = -0.013, p = 0.747) and funnel plot (Fig. 5c). The pooled prevalence rates changed to 10.77% (95% CI: 5.31%-17.87%) with 224 SAD cases out of 2316 samples for 4 articles (68, 71, 75, 77).
3.6 The pooled prevalence of SAD assessed by semi-structured interview
The meta-analysis (Fig. 6a) shows the pooled prevalence of SAD assessed by semi-structure interviews in two studies (78, 79). The pooled data included 3951 samples, with 184 having SAD; its heterogeneity was high (I2 = 98.32) and significant (Q = 59.65, p<. 001). Therefore, the random-effects model was used, showing that the pooled prevalence rate of SAD was 5.84% (95% CI: 1.12%-13.91%). The funnel plot did not show publication bias (Fig. 6b) after examining Egger’s test (Intercept = 21.469, p = 0.106) and Begg’s test (Kendall’s Tau = 0.991, p = 0.317), hence no sensitivity analysis was performed.