Methodology
Eligibility Criteria
All cross-sectional studies published since 1990 where the prevalence of any type of anxiety disorder was estimated were included in the study. We included all the studies where the age group of the sample population belonged to 10-19 years. If more than 50% of the sample belonged to the 10-19 years of age group, then those studies were also included. The studies that reported any type of anxiety disorder such as generalized anxiety disorder, OCD, PTSD, panic disorder, and social phobia (or social anxiety disorder) were included. We excluded all other studies that did not fulfill the inclusion criteria.
Search Strategy
We searched Medline and ProQuest databases for peer-reviewed articles. The search strategy was developed using combined terms related to anxiety, general anxiety, mental health, anxiety disorder, phobia, stress, obsession, panic, India, prevalence, cross-sectional, and burden. From ProQuest, only thesis and dissertations were chosen using the appropriate filter [8]. A detailed search strategy specific to both databases is mentioned in Supplementary 1.
Risk of Bias Assessment
The two dimensions of the Quality in Prognosis Studies (QUIPS) tool that are relevant to observational studies, (1) study participation and (2) study outcome, were used to assess the likelihood of bias in the articles included in the study [9]. Each domain’s evaluation yields a subjective estimate of bias risk (low, moderate, or high). The supplementary document provides the tool for risk of bias assessment (Supplementary 2).
Data Extraction
A data extraction sheet was used to extract the data regarding the authors’ name, study area, study participants, sampling strategy, age group, and prevalence. Simultaneously, the confidence interval (CI) was calculated and mentioned in the sheet. For most of the studies, the CI value was not mentioned in the original study, and therefore it has been calculated using a formula such as (p̂ +/- z* (p̂(1 – p̂)/n)0.5), where p̂ is prevalence, z value is 1.96, and n is the sample size. The risk of bias was also mentioned in the data extraction sheet.
Reliability
Two reviewers (D.P. and S.M.) checked the articles for the title and abstracted for selection of the studies in a blinded way. Rayyan web-based platform was used for this purpose. In case of any dispute regarding the inclusion of the study, the senior researcher (D.P.S.) took the final decision. All data extracted were checked by all three reviewers.
Analysis
We have provided a descriptive analysis of all the studies included in the meta-analysis. The I2 statistic, for the variance not due to sampling error across studies, was used to analyze heterogeneity between estimates. High heterogeneity is indicated by an I2 value of more than 75%. We included those papers in the meta-analysis where any form of diagnostic tool was used for detecting any type of anxiety illness in teenagers aged 10 to 19 years, as well as studies with more than half of the participants aged 10 to 19. The meta-analysis was carried out using the R program and a random-effects model (to account for heterogeneity). A 95% C) was derived for a pooled prevalence number. When the estimate for a study went toward either below 20% or above 80% in a meta-analysis of prevalence, log transformation was required for normalization of the distribution of prevalence of all studies. After log transformation, the final pooled result and 95% CIs were back-transformed for the final result. We used the Baujat test to find the study resulting in heterogeneity, and the outlier was removed once to find out the effect of the study in heterogeneity and pooled estimate. We used subgroup analysis on the basis of risk of bias, where we classified the studies having a high and moderate risk of bias and studies having a low risk of bias. We used Meta-Essentials for subgroup analysis.
Ethical issues
As this study analyzed data from studies available in the public domain, no ethical clearance was sought. This systematic review and meta-analysis was registered in PROSPERO before the initiation of the review (reference number: CRD42022345574).
Results
The search results returned a total of 2,296 articles from the two databases, and after exclusion of duplicates, 2,270 articles were considered for screening by titles. After screening for the titles, 72 articles were selected for screening by abstract. Among full-text screening for 20 articles, finally, 13 articles were selected for quantitative analysis (Figure 1). Two of the articles were excluded for being part of the same study, and five articles were excluded for being review articles.
Included Studies
All of the included studies had a cross-sectional design. Three of the studies used the Screen for Child Anxiety Related Disorders (SCARD) tool [10-12]. DSM-5 and DSM-5 Text Revision (DSM-5 TR) were used in five studies [13-17]. The Depression, Anxiety and Stress Scale – 21 (DASS-21), Westside Test Anxiety Scale, and Test Anxiety inventory were the other tools used in the studies [18-21] (Table 1). In one study, one pre-tested questionnaire was used for diagnosing anxiety disorder [22].
Risk of Bias
All the studies were classified as high, moderate, and low risk on the basis of subjective assessment of studies using the QUIPS tool [23]. Bias in selecting participants and bias in outcome measurement were assessed for all included studies. One study was found to have a moderate risk of bias, and three studies had a high risk of bias. All of the other studies had a low risk of bias (Table 1).
Table
1: Description of the studies along with risk of bias assessment
DASS, Depression, Anxiety and Stress Scale; DSM, Diagnostic and Statistical Manual of Mental Disorders; MINI, Mini International Neuropsychiatric Interview
Meta-Analysis
The pooled prevalence was found to be 0.23 with a CI of 0.11-0.41 (Figure 2). The I2 statistics was found to be significant, with a heterogeneity of 99.67%. As the variability was high, random effect model was used to calculate the pooled estimate. During subgroup analysis on the basis of risk bias, the pooled prevalence was found to be 0.41 (CI 0.14-0.96) for studies having more than low risk. The pooled estimate for the studies with low risk of bias is found to be 0.29 (CI 0.11-0.46). Table 2 shows the weightage of different studies with respect to pooled estimates (Table 2). The Baujat test has detected a study conducted by Pillai et al. as an outlier. After removing this study from the analysis, no significant change is detected in heterogeneity and pooled prevalence.
Table
2: Weightage of different studies in respect to pooled prevalence using random effect model
CI, confidence interval
Publication Bias
The Begg and Mazumdar rank correlation test found that the publication bias is not present in this meta-analysis (p=0.085). Figure 3 shows the funnel plot having a symmetrical distribution of studies with respect to standard error and effect size (Figure 3).
Discussion
Out of the 13 studies, nine studies had a low risk of bias and rest of the studies had either moderate or high risk of bias. The pooled estimate for the studies with a low risk of bias was found to be 0.29 (CI: 0.11-0.46) and that for other studies it was 0.41 (CI: 0.14-0.96). The random effect model was used to find out the pooled prevalence as high level of heterogeneity was present among studies. No tool exists for the objective assessment of the quality of bias of cross-sectional studies. Two domains of the QUIPS tool relevant to cross-sectional studies were used here for subjective assessment of bias. This tool was piloted by other authors for the same purpose and was previously used in one meta-analysis [24]. This tool also followed the guidelines of Cochrane collaboration [25]. The prevalence value in different studies can be attributed to different reasons such as type of study population, type of study tool, and type of sampling strategy. Meta-regression analysis could have been conducted to find out those factors. The prevalence of anxiety among adolescents varies in a wide range in different countries. In the USA, approximately 30% of adolescents suffer from some type of anxiety disorder [26]. Among the south-east Asian countries, the prevalence of anxiety in adolescents varies from 21.4% in Pakistan to 9% in Bhutan [27,28]. In the USA, unemployment and substance abuse are found to be significant risk factors for anxiety in adolescents [29]. Poverty and social instability play a crucial role in Pakistan [30]. In Bhutan, the prevalence of substance abuse is found to be lower than that in the USA or Pakistan [29-31]. Those risk factors are prevalent in India also, which lead to similar kind of result in comparison with the USA or Pakistan [32]. This study would help find out the burden of anxiety disorders In India in the pre-COVID-19 era, which has been grossly aggravated due to the COVID-19 pandemic. The COVID-19 pandemic has been found to be a significant risk factor for causing anxiety disorder [33,34].
Strengths
Our study helps get an overview of the burden of anxiety disorders in India, as studies from almost every part of India were included in the analysis. Both types of population such as school students and non-school going children were included in those studies.
Limitations
We did not have access to some databases such as OVID, Embase, Web of Science, and Scopus due to financial constraints. Though we have included two databases as per the requirement prescribed by the Cochrane collaboration group, other databases were not screened.