Impact statement
The synthesis of 63 articles in this study gives a glimpse into the prevalence of five common psychiatric conditions: conduct disorder, depression, anxiety, attention-deficit hyperactivity disorder and post-traumatic stress disorder among sub-Saharan African youth. The high rates of depression, post-traumatic stress disorder and anxiety underscore the urgency for targeted interventions and policy reform. Our review compares the prevalence of these conditions among sub-Saharan African youth to global estimates for these conditions. It also calls attention to the pressing need for culturally sensitive and standardized assessments to measure mental health conditions.
Introduction
An estimated 13% of all adolescents have at least one diagnosed mental disorder (Kuehn, Reference Kuehn2021; UNICEF, 2021). In sub-Saharan Africa (SSA), 23% of the population (256 million people) is between the ages of 10–19 years (UNICEF, 2014; Agyepong et al., Reference Agyepong, Sewankambo, Binagwaho, Coll-Seck, Corrah, Ezeh, Fekadu, Kilonzo, Lamptey, Masiye, Mayosi, Mboup, Muyembe, Pate, Sidibe, Simons, Tlou, Gheorghe, Legido-Quigley and Piot2017; Sequeira et al., Reference Sequeira, Singh, Fernandes, Gaikwad, Gupta, Chibanda and Nadkarni2022) and SSA adolescents are the fastest-growing population in the world (Sequeira et al., Reference Sequeira, Singh, Fernandes, Gaikwad, Gupta, Chibanda and Nadkarni2022). Additionally, despite carrying a vast proportion of the global burden of mental disorders, the ratio of psychiatrists to the population of most SSA countries sits at less than 1 per 1,000,000 (WHO, 2021a), and the ratio of child psychiatrists is even lower (at 1 per 4,000,000 people) (Belfe and Saxena, Reference Belfe and Saxena2006).
A previous meta-analysis estimated that 26.9% of youths aged 10–19 years in SSA experienced depression, 29.8% reported anxiety, 40.8% emotional and behavioural problems, 21.5% for post-traumatic stress disorder (PTSD) and 20.8% suicidal ideation (Jörns-Presentati et al., Reference Jörns-Presentati, Napp, Dessauvagie, Stein, Jonker, Breet, Charles, Swart, Lahti, Suliman, Jansen, van den Heuvel, Seedat and Groen2021). The high prevalence of mental disorders suggests that research is needed to support the availability and accessibility of mental health interventions to promote well-being (Cortina et al., Reference Cortina, Sodha, Fazel and Ramchandani2012; Jörns-Presentati et al., Reference Jörns-Presentati, Napp, Dessauvagie, Stein, Jonker, Breet, Charles, Swart, Lahti, Suliman, Jansen, van den Heuvel, Seedat and Groen2021). Another meta-analysis estimated that 14.3% of children aged 0–16 years in SSA had at least one mental disorder (Cortina et al., Reference Cortina, Sodha, Fazel and Ramchandani2012). As these studies present the data differently, with some focusing on specific disorders (e.g., depression and anxiety) and some on psychopathology in general, it is difficult to draw overall prevalence rates of mental disorders among youths in SSA. While previous reviews have assessed the prevalence of mental disorders among SSA youth (Jörns-Presentati et al., Reference Jörns-Presentati, Napp, Dessauvagie, Stein, Jonker, Breet, Charles, Swart, Lahti, Suliman, Jansen, van den Heuvel, Seedat and Groen2021), to our knowledge, none so far have assessed examined factors that may moderate the prevalence of mental disorders among SSA youth (e.g., use of diagnostic vs. screening procedures, special vs. general population), which is an important step forward to address challenges in mental healthcare for SSA youth.
Consolidating prevalence data on mental disorders among youth in SSA may be useful not only for better understanding the epidemiology of mental disorders in SSA but also for considering what resources would be most beneficial in this setting. To add to previous work analyzing the overall prevalence of general psychopathology in SSA, our meta-analysis was conducted with the aim of determining the prevalence of multiple common mental disorders in SSA; this work may be a step towards developing targeted interventions and supporting appropriate mental health policies. We evaluated whether reported prevalence rates were affected by the disorder type, year of study, mean age of the sample, percentage of female participants, type of psychometric scale (i.e., diagnostic or screening), whether the scale was validated in context, special population status (e.g., former child soldiers, trauma survivors) and region (East, West, Central or Southern Africa) (African Development Bank, 2018).
Method
The study was pre-registered on PROSPERO CRD42022326574 (see Supplement 1) and conducted in accordance with PRISMA guidelines (see Supplement 2) (Page et al., Reference Page, McKenzie, Bossuyt, Boutron, Hoffmann, Mulrow, Shamseer, Tetzlaff, Akl, Brennan, Chou, Glanville, Grimshaw, Hróbjartsson, Lalu, Li, Loder, Mayo-Wilson, McDonald and Moher2021). The search strategy was developed by three researchers with support from a university librarian. The disorders selected are the five most identified mental health conditions among youth (i.e., depression, PTSD, anxiety, conduct disorder and attention-deficit hyperactivity disorder [ADHD]) (Weisz and Kazdin, Reference Weisz and Kazdin2017). To assess the effectiveness of the search strategy, the search terms were trialled in the various databases, and the researchers made adaptations to ensure that the search yielded relevant studies (see Appendix A). The three researchers also created and piloted the eligibility criteria to identify appropriate studies.
The systematic search was conducted on 20 December 2021 using the following four databases: PsycInfo (n = 928), Pubmed (n = 2,858), AfriBib (n = 234) and Africa Journals Online (n = 100 articles could be accessed). The available articles from all four databases were exported to EndNote X9 (EndNote Team, 2013) and uploaded to Rayyan (Ouzzani et al., Reference Ouzzani, Hammady, Fedorowicz and Elmagarmid2016), where duplicate records were identified and removed.
Screening
The title, abstract and full text of each search result were independently double-screened by four authors using the piloted pre-specified inclusion criteria. The inclusion criteria included the following: (1) an empirical study, (2) published in English (as this was the only language that all authors spoke fluently), (3) involving participants from SSA with a mean age of less than 19 years, (Viner, Reference Viner2013) and (4) including a prevalence measure of one or more of the selected disorder types (i.e., anxiety problems, depression problems, conduct problems, ADHD and post-traumatic stress disorder [PTSD]). We defined a prevalence measure as a measurement taken with a tool used to identify a mental health diagnostic status or an established (i.e., psychometrically validated in some setting) measure of symptom levels. The studies could be of a general or special population (e.g., youth living with HIV). Studies were excluded if the participants were already selected or self-selected for the presence of mental disorders or symptoms (i.e., people already seeking or receiving mental health care). Furthermore, if several mental disorders were included, the studies needed to report a prevalence for each condition (i.e., not a general measure for distress or disorder). Additionally, we excluded studies using non-probabilistic search strategies to mitigate the risk of bias in prevalence estimates (see Appendix A).
Data Extraction
Data pertaining to the following were extracted from each included article: (a) study characteristics, (b) participant characteristics and (c) prevalence of included mental disorder(s). Four researchers completed the data extraction independently; however, any ambiguity in reporting was explored through weekly meetings. The characteristics extracted from each study included study location by country, objectives of the study and study design. The extracted participant and study characteristics included: sample size, age range, mean age, percent female, sampling method and, if applicable, the special population characteristics (e.g., juvenile offenders).
Regarding the prevalence of selected mental disorders, the following information was extracted: selected disorder, psychometric scale(s) used, informant (i.e., self-reported, teacher- or parent-reported) and prevalence measure. For studies that included participants from multiple regions or reported prevalence rates for more than one of the five psychiatric disorders, we extracted the sample size, mean age, percentage of female participants and psychometric scale type as reported for each disorder and/or region. We also investigated whether the psychometric scales were culturally validated in the study’s context. For manuscripts that reported that the chosen tools were validated, we assumed they were indeed validated. For manuscripts that did not include information about scale validation, we cross-checked the broader literature to determine whether the scales were validated at the time of their inclusion in the studies (indicated as “No” if the scale was not validated, “Yes” if the scale was validated, or “Yes†” if the scale had been validated after the study).
Quality Appraisal
Each study was evaluated by two independent authors using the Johanna Briggs Institute Tools for cohort and cross-sectional study designs (Moola et al., Reference Moola, Munn, Sears, Sfetcu, Currie, Lisy, Tufanaru, Qureshi, Mattis and Mu2015). Any discrepancies in the appraisals were resolved by consensus.
Data Analysis
Prevalence rates were obtained from each included study and organized by selected disorder, as presented in Table 2. When studies included more than one prevalence rate for the same disorder (e.g., multiple scales used to assess the same condition), a weighted average of all reported prevalence rates was calculated by two authors. For studies that did not provide a confidence interval (CI) around the prevalence estimate, the 95% CIs for all reported prevalence rates were calculated (Eberly College of Science, 2022).
Using the metafor package in R (Version 4.3.1 (2023-06-16)) (Viechtbauer, Reference Viechtbauer2022), logit-transformed proportions were calculated for each prevalence rate, and the inverse variance method was applied to estimate the pooled prevalence of each condition (Berkey et al., Reference Berkey, Hoaglin, Antczak-Bouckoms, Mosteller and Colditz1998; Harrer et al., Reference Harrer, Cuijpers, Furukawa and Ebert2021). We then used a mixed-effects meta-regression, with the proportion specified as a random effect and the sub-group variable specified as a fixed effect, and a logit-transformation applied to the proportion to test the following hypothesized moderators: year of study, location, psychometric scale type, disorder type, mean age of the sample, percentage of female participants, region or special population status.
To assess the heterogeneity of the studies included in the review, forest plots were created for each of the five specified disorders. Furthermore, sensitivity analyses were conducted to assess heterogeneity (I 2) after the removal of studies of a special population or those that used a non-validated tool. Sensitivity analyses were also employed to evaluate heterogeneity after the removal of outliers and influential cases that were identified through influence (Viechtbauer and Cheung, Reference Viechtbauer and Cheung2010) and Graphic Display of Heterogeneity (GOSH) plot (Olkin et al., Reference Olkin, Dahabreh and Trikalinos2012) diagnostics. Finally, subgroup analyses were conducted to investigate the variance in prevalence between general and special populations for culturally validated measures compared to non-validated measures and for studies that used a screening tool compared to those that used a diagnostic tool.
Results
The systematic search identified 4,120 search hits, with 3,783 studies included in screening after removing duplicates. After the title and abstract screening process, 3,639 studies were excluded and 140 studies underwent full-text screening. As a result, 77 more studies were excluded, and 63 studies were included in the final review (see Supplement 2).
As seen in Table 1, the studies included in this meta-analysis (n = 63) were conducted in 14 countries across SSA. The most common locations were Nigeria (n = 15), Kenya (n = 12) and South Africa (n = 10). The remaining studies were conducted in Uganda (n = 8), Ethiopia (n = 4), Tanzania (n = 3), Ghana (n = 3), Democratic Republic of Congo (n = 2), Rwanda (n = 1), Malawi (n = 1), Burundi (n = 1), Namibia (n = 1), Zambia (n = 1) and Zimbabwe (n = 1). Most studies were cross-sectional (n = 57), and six studies employed a cohort design. The studies included in this review varied in sample sizes, populations and sampling methods. Sample sizes ranged from 31 participants to 4,795 participants. The total sample size of all included studies was 55,071. The most common sampling method was cluster sampling (n = 24). Other studies utilized stratified random (n = 8), multistage (n = 7), random (n = 6), systematic (n = 3) and population-based (n = 2) sampling. Participant ages ranged from 0 to 28 years, and the mean age of the total sample was 13.63 (SD = 2.52). We included the mean or median age, as reported in the studies. Overall, the total sample comprised 46.65% females. As seen in Table 2, more than half of the studies reported on the prevalence of mental disorders in the general population, 30 articles studied a special population as follows: HIV-positive children/adolescents (n = 11), violence-affected youth (s = 3), refugees (n = 2), orphans (n = 2), juvenile offenders (n = 2), children seeking primary medical care (n = 2), rape survivors (n = 1), former child soldiers (n = 1), child labourers (n = 1), cancer patients (n = 1), pregnant adolescents (n = 1), youth in vocational training (n = 1) and overweight and obese children (n = 1). Most studies included in this review assessed depression (n = 34), followed by ADHD (n = 23), anxiety disorders (n = 22), PTSD (n = 19) and conduct problems (n = 12). The disorders were measured using different informants, including caregivers (n = 20), teachers (n = 11) and self-reports (n = 45) (see Table 2). Furthermore, in the 63 studies, six tools were not culturally validated, and nine scales were validated in the context after the study was published. Among the 63 included studies, 38 unique scales were used; of these, 13 were diagnostic, and 25 were screening tools. Most (n = 20) of the scales were only used in one study. The most frequently used psychometric tools were the Disruptive Behaviour Disorder Rating Scale (DBDRS) for ADHD (n = 6), Mini International Neuropsychiatric Interview for Children and Adolescents (MINI-KID) for depression (n = 6), Patient Health Questionnaire-9-item (PHQ-9) for depression (n = 8) and the Children’s Depression Inventory (CDI) for depression (n = 6). See Appendix C for additional details on the scales used.
Table 1. Study characteristics
Note:
† Weighted average, excluding studies that did not report these figures;
‡ Median age
Table 2. Prevalence data
Note: Above are the prevalence from each study, accompanied by the psychometric tools and their contextual validation indicated as “No” if the scale was not validated, “Yes” if the scale was validated, or “Yes†” if the scale had been validated after the study. Additional details of psychometric scales can be found in Appendix C.
Based on the prevalence reported in Table 2, we calculated the following pooled prevalence rates for youth mental health conditions: depression 15.27% [CI 9.92; 22.78], PTSD 12.53% [CI 7.59; 20.00], anxiety disorders 11.78% [CI 7.27; 18.54], conduct disorder 9.76% [CI 4.93; 18.41] and ADHD 6.55% [CI 4.61; 9.23]. Additionally, there were two studies, Kamaue et al. (Kamau et al., Reference Kamau, Kuria, Mathai, Atwoli and Kangethe2012) and Gureje et al. (Reference Gureje, Omigbodun, Gater, Acha, Ikuesan and Morris1994), that included estimates of several anxiety conditions in their studies. Gureje et al. (Reference Gureje, Omigbodun, Gater, Acha, Ikuesan and Morris1994) detailed the following additional prevalence: separation anxiety 1.7% [CI 1.0–2.7], overanxious disorder 0.7% [CI 0.3–1.5], and simple phobia 2.0% [CI 1.2–3.1]. Similarly, Kamaue et al. (Reference Kamau, Kuria, Mathai, Atwoli and Kangethe2012) listed panic disorder at 5.8%, agoraphobia at 2.6%, specific phobia at 7.1%, social phobia at 12.8%, panic disorders at 5.8% and separation anxiety disorder at 2.6%. The estimates of overall anxiety disorders from these studies were included in weighted average calculations, but we have listed these prevalence measures as additional information.
We observed high heterogeneity between studies for all conditions (see forest plots in Supplement 3). Sensitivity analyses indicated that heterogeneity did not substantially decrease after removing studies that included a special population, those that used a screening tool or tool that was not contextually validated. Similarly, we conducted a sensitivity analysis after removing outliers, studies with special population and non-validated tools and GOSH plot diagnostics did not substantially reduce heterogeneity (see Appendices D and F). Furthermore, we performed a subgroup analysis of population types, which revealed no significant differences between the pooled prevalence for studies that included a general vs. special population (see Appendix E). Subgroup analyses of validated vs. non-validated tools indicated a significant difference in the pooled prevalence of PTSD and anxiety. However, in both cases, only one study used a non-validated tool: Bukenya et al. (Reference Bukenya, Kasirye, Lunkuse, Kinobi, Vargas, Legha, Tang and Miranda2022), which was conducted in Uganda using the PC-PTSD questionnaire, and Adewuya and Famuyiwa (Reference Adewuya and Famuyiwa2007), which was conducted in Nigeria using the VADTRS and VADPRS questionnaires. Because of the limited number of non-validated tools, these results must be interpreted cautiously. Finally, we carried out a subgroup analysis of psychometric tools, which demonstrated a significantly higher prevalence for depression and conduct disorder when a screening tool was used to estimate prevalence compared to a diagnostic tool (see Appendix G).
In the full model, the disorder type (i.e., ADHD, Anxiety, Depression, Conduct, PTSD) was found to be a significant predictor of prevalence, with depression and PTSD having a higher prevalence. Other than disorder type, none of the hypothesized moderators were significantly associated with an increased reported prevalence rate in the full model. However, given that previous studies have consistently reported differences in psychiatric morbidities across genders (Seedat et al., Reference Seedat, Scott, Angermeyer, Berglund, Bromet, Brugha, Demyttenaere, de Girolamo, Haro, Jin, Karam, Kovess-Masfety, Levinson, Medina Mora, Ono, Ormel, Pennell, Posada-Villa, Sampson and Kessler2009; Remes et al., Reference Remes, Brayne, van der Linde and Lafortune2016; Van Droogenbroeck et al., Reference Van Droogenbroeck, Spruyt and Keppens2018; Miranda-Mendizabal et al., Reference Miranda-Mendizabal, Castellví, Parés-Badell, Alayo, Almenara, Alonso, Blasco, Cebrià, Gabilondo, Gili, Lagares, Piqueras, Rodríguez-Jiménez, Rodríguez-Marín, Roca, Soto-Sanz, Vilagut and Alonso2019) and age (Park et al., Reference Park, Bang and Kim2014), we conducted an exploratory analysis to probe whether these variables were associated with increased reported prevalence rates when tested alone as individual moderators. When modelled separately, we found that the mean age of the participants significantly moderated the reported prevalence rates (B = 0.02, p < 0.001, k = 69). In contrast, the percentage of female participants did not significantly moderate the reported prevalence (B = 0.0009, p = 0.386, k = 100).
Quality Appraisal
The studies ranged in the level of detail provided for the inclusion criteria and description of the study setting and participants, which may have impacted our evaluation of confounding variables. The average appraisal rating across the 57 cross-sectional studies included in this review was 6.74/8, indicating a moderate quality of the included cross-sectional studies. Regarding the six cohort studies included in this review, some did not detail their evaluation of confounding variables and loss to follow-up. As this information was uncertain, it was difficult to assess the overall quality of the studies, as it might have indicated possible bias. This resulted in a lower-quality rating for cohort studies included in this review, with an average rating of 6.83/11.
Discussion
This meta-analysis evaluated the prevalence of five mental disorders among youth in SSA. The prevalence of depression was found to be the highest, at 15.27%, with PTSD having the second highest prevalence, at 12.53%. Anxiety had a prevalence rate of 11.78%, with conduct disorder observed among 9.76% and ADHD among 6.55% of SSA youth. This prevalence drew from cohort and cross-sectional findings across 14 countries in SSA involving a total sample of 55,07 participants with a mean age of 13.63 (SD = 2.52) (see Tables 1 and 2). The global estimates for depression and conduct disorder were greater than those found in this review (UNICEF, 2021; Shorey et al., Reference Shorey, Ng and Wong2022). The global prevalence of ADHD and anxiety were comparable to those calculated in this review (Merikangas et al., Reference Merikangas, Nakamura and Kessler2009; Thomas et al., Reference Thomas, Sanders, Doust, Beller and Glasziou2015). Although there has not been a global estimate for PTSD among a general population, a previous meta-analysis of PTSD among trauma-affected youth found a similar prevalence to the one found by this review (Alisic et al., Reference Alisic, Zalta, van Wesel, Larsen, Hafstad, Hassanpour and Smid2014). Additionally, a systematic review of PTSD prevalence in LMICs found a widely ranging prevalence of PTSD, similar to what has been found in this review (Yatham et al., Reference Yatham, Sivathasan, Yoon, da Silva and Ravindran2018).
High between-study heterogeneity was observed, suggesting considerable variation in prevalence. Interestingly, heterogeneity did not substantially decrease despite removing outliers and influential cases. To further explore potential reasons for observed differences in reported prevalence rates, we assessed the significance of our hypothesized moderators as predictors in a meta-regression. We found that the disorder type significantly moderated the prevalence in the full model, and the mean age of the study sample significantly moderated reported prevalence rates when modelled separately, with increased age correlating to an increase in reported prevalence. The direction of this effect is consistent with other studies that indicated adolescence as a time of increased incidence of psychiatric morbidities, including mood disorders (Merikangas et al., Reference Merikangas, Nakamura and Kessler2009; WHO, 2021b; Sulley et al., Reference Sulley, Ndanga and Mensah2022). However, these results are subject to ecological and aggregation biases and should be interpreted with caution. Our pooled prevalence combined a range of samples, with some being from a special population. When we conducted a subgroup analysis to determine if this impacted the resulting pooled prevalence rates, we found that for PTSD, ADHD and conduct disorder, the prevalence rate was lower among special populations (e.g., youth living with HIV, refugees, violence-affected youths, juvenile offenders) and for anxiety and depression, the rate was higher. However, these differences were not statistically significant (see Appendix E). Additionally, in a sensitivity analysis, when special populations were removed, heterogeneity did not decrease (see Appendix D).
SSA is a vast geographic region characterized by important cultural and contextual differences that may have impacted the diversity of results. Prior research suggests that mental disorder burdens are comparatively higher in East Africa (Cataldi, Reference Cataldi2021; Ferrari et al., Reference Ferrari, Santomauro and Herrera2022), with less expenditure noted for neurological disorders in this region (Etindele Sosso and Kabore, Reference Etindele Sosso and Kabore2016). However, while disorder type was significantly associated with a difference in reported prevalence rates, we found no significant effect for region nor the interaction between region and disorder type in our review.
Some of the included studies used several scales, including screening and diagnostic tools; however, typically, in the manuscripts of the included studies, only a single percentage was reported using results obtained via the screening tool. During data extraction, we indicated which type of scale (e.g., screening vs. diagnostic) was used to determine the included prevalence rate. We found screening tools were often used to indicate prevalence, which may be misleading as this approach cannot always differentiate between those at risk of a mental disorder and those who could be diagnosed with that disorder. Unfortunately, even in high-income countries with relatively well-staffed health systems, issues such as long waiting times for a formal diagnosis from a qualified practitioner often hinder the accurate tracking of diagnosed mental disorders (NHS, 2022). The subgroup analysis for diagnostic vs. screening tools found significant differences for depression and conduct disorder, where the pooled prevalence of studies using a diagnostic tool was lower than studies using a screening tool.
We also noted that most studies used self-report scales. However, the studies that included parent and teacher ratings often found higher rates of the given disorder than the self-reported scores, which aligns with previous studies that have shown discrepancies in inter-rater reports (Brown et al., Reference Brown, Wissow, Gadomski, Zachary, Bartlett and Horn2006; Papageorgiou et al., Reference Papageorgiou, Kalyva, Dafoulis and Vostanis2008; Boman et al., Reference Boman, Stafström, Lundin, Moghadassi, Törnhage and Östergren2016). Additionally, some authors used scales that had not been validated in the context of SSA. We found that when the scale was not validated in context, the rates of PTSD and anxiety were significantly higher (see Appendix F).
Overall, the studies were evaluated to be of moderate quality. There was a high level of uncertainty, with gaps in reporting such as consideration of confounding variables, reporting study/participant characteristics, and losing participants to follow-up, which impacted the level of quality in the assessed studies. The lack of reporting of some study quality indicators might suggest a need for more adequate reporting methods and analysis in these studies.
One limitation of this review is the lack of standardization in prevalence measurements. Depression, for example, was defined differently across studies, including depression problems, elevated depression symptoms, or major depressive disorder. Similarly, anxiety-related conditions were referred to as emotional problems, affective problems or anxiety disorders. Previous research has demonstrated the great clinical and diagnostic heterogeneity (Fried, Reference Fried2017; Dennis-Tiwary et al., Reference Dennis-Tiwary, Roy, Denefrio and Myruski2019; Athira et al., Reference Athira, Bandopadhyay, Samudrala, Naidu, Lahkar and Chakravarty2020; Drzewiecki and Fox, Reference Drzewiecki and Fox2024) observed for both anxiety and depression. Thus, in the screening process, the authors of this meta-analysis considered the core symptoms described in each paper as well as the psychometric scales to ensure that these were screening or diagnostic tools for anxiety and depression. Still, the variety of measures used likely contributed to variation in the reported prevalence rates, thus increasing the confidence interval of the pooled prevalence. Relatedly, several scales were used to assess the prevalence of each disorder, which may have also contributed to the significant variation in the reported prevalence across studies. Furthermore, our review included 63 studies, six tools were not culturally validated, and nine scales were validated in the context after the study was published. Of the included studies, 38 unique scales were used; of these, 13 were diagnostic, and 25 were screening tools. Most (n = 20) of the scales were only used in one study (see Table 2).
Another limitation of this study is that the systemic search was conducted 2 years ago, and there have been several studies published since on the prevalence of youth mental health conditions. For example, Woolgar et al., Reference Woolgar, Garfield, Dalgleish and Meiser-Stedman2022 found that preschool children were at risk of developing PTSD following exposure to trauma. They reported a pooled prevalence of 21.5% in their review and noted heterogeneity across the included studies as well as a lack of representation from LMICs. Another study (Yang et al., Reference Yang, Wen, Huang, Riem, Lodder and Guo2022) found an estimated global youth PTSD prevalence of 28.15% following the outbreak of coronavirus. Specific to youth in SSA, Jörns-Presentati et al., Reference Jörns-Presentati, Napp, Dessauvagie, Stein, Jonker, Breet, Charles, Swart, Lahti, Suliman, Jansen, van den Heuvel, Seedat and Groen2021 found a prevalence of 26.9% for depression, 29.8% for anxiety disorders, 40.8% for emotional and behavioural problems, 21.5% for PTSD and 20.8% for suicidal ideation. Another study on SSA youth (Hunduma et al., Reference Hunduma, Dessie, Geda, Yadeta and Deyessa2023) reported the following prevalence: 19% for depression, 20% for anxiety, 5% for ADHD and 15% for conduct disorders. Our meta-analysis included a larger sample of studies conducted in SSA and thus added to these previous studies. Although the estimates reported by recent reviewers are slightly higher than our prevalence estimates, these reviews together highlight the prevalence and importance of youth mental health difficulties in SSA.
The comparability of measures was not restricted to between studies only as, in some studies, more than one measure was used to assess the same disorder. Moreover, as reported in a previous meta-analysis (Cortina et al., Reference Cortina, Sodha, Fazel and Ramchandani2012), there is a clear dearth of locally derived tools for measuring psychopathology. In this review, many researchers used measurement tools derived from Western populations to assess mental health symptoms. Although some studies reported on the psychometric properties of Western-derived tools when used in the SSA population, many did not. Furthermore, although psychometrically validated, some tools constructed in higher-income settings may not accurately capture culturally salient features of these disorders that are unique to the African population (Osborn et al., Reference Osborn, Kleinman and Weisz2021). This indicates a need for more culture-sensitive, psychometrically validated tools tailored for use in the African population. It also indicates a need for standardized methods of assessing psychopathology. Relatedly, an important limitation of our study is that the search was conducted in English, and the inclusion of studies was limited to those published in English. Future meta-analyses and systematic reviews may include more languages to increase the representation of studies published in LMICs and SSA.
Our findings suggest a need to explore further reasons for the varying prevalence rates of studies across SSA. As our study demonstrated, there is currently an extensive range of psychometric tools used to assess mental disorders in this setting, which might lead to increased variation in prevalence estimates. Further research should find a way to reconcile multiple tools used to screen for and diagnose a single mental disorder to consolidate and compare rates of mental conditions across subgroups and populations. To scale up mental health interventions in SSA, we believe a closer investigation into the cultural and contextual factors—including socioeconomic variables, language and culture around mental health, and religious or spiritual beliefs—that may affect the prevalence of mental disorders would support the implementation process. Finally, although our review did not find a significant moderating effect of special populations, we believe that it would be necessary to evaluate these samples further to explore potential risk and protective factors for developing mental disorders within SSA. For example, there might be support systems that target at-risk youth, such as community-based organizations or non-governmental services, which may influence the prevalence rates.
Our meta-analysis revealed the differences in prevalence between types of mental disorders, which may implicate the clinical prioritization of certain conditions, such as PTSD and depression, in SSA. Despite the high prevalence rates of mental disorders in SSA, service availability remains limited. Previous research demonstrated that critical barriers to implementing youth mental health interventions include stigma, negative beliefs and having few delivery platforms outside of school-based settings, which may exclude individuals not attending formal education (Heflinger and Hinshaw, Reference Heflinger and Hinshaw2010; Jenkins et al., Reference Jenkins, Kiima, Njenga, Okonji, Kingora, Kathuku and Lock2010; Ndetei et al., Reference Ndetei, Mutiso, Maraj, Anderson, Musyimi and McKenzie2016a). The facilitators for implementing mental health interventions include positive experiences and mental health literacy; thus, positive psychology and psychoeducation may be priority mental health intervention research areas (Aguirre Velasco et al., Reference Aguirre Velasco, Cruz, Billings, Jimenez and Rowe2020). Additionally, a recent meta-analysis has found that youth psychotherapies are particularly effective in LMICs compared to non-LMICs; thus, given the rates of mental disorders in LMICs combined with the promising effects of youth psychotherapies, researchers, funders and policy-makers may emphasize scale up of youth mental health interventions in SSA (Venturo-Conerly et al., Reference Venturo-Conerly, Eisenman, Wasil, Singla and Weisz2023). Due to the high rates of mental disorders, our findings suggest that mental health care should be integrated into primary youth care, as early detection and intervention are critical for reducing the chronicity and severity of mental disorders. However, to support the successful integration of mental health into primary youth care, more research is needed to validate psychometric tools in local contexts. Furthermore, as recommended by Sequeira et al., Reference Sequeira, Singh, Fernandes, Gaikwad, Gupta, Chibanda and Nadkarni2022, Reference Sequeira, Singh, Fernandes, Gaikwad, Gupta, Chibanda and Nadkarnia greater understanding of common mental health conditions and social determinates is needed to bridge the gap stigma creates.
Open peer review
To view the open peer review materials for this article, please visit http://doi.org/10.1017/gmh.2024.82.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/gmh.2024.82.
Data availability statement
All articles included in this review are available in the described databases, and the sample characteristics and prevalence details are included in this article.
Acknowledgements
The authors would like to acknowledge Moti Heda and Huong Le for supporting the piloting and development of the review search strategy.
Author contribution
CJ developed and piloted the search terms, conducted the search and contributed to the screening, analysis and write-up of the manuscript. NJ and BO contributed to the manuscript’s screening, data extraction, analysis and write-up. RB contributed to the screening, data extraction and write-up process. EV contributed to the data extraction, analysis and write-up of the manuscript. CM and DN contributed as supervisors to the write-up and editing of the draft manuscript. KVC supervised the project, including the development of the inclusion criteria and search strategy, the provision of comments on the analysis plan, and the offering of manuscript edits.
Financial support
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
Competing interest
KVC is a co-founder and scientific director at the Shamiri Institute, a non-profit organization that aims to provide accessible mental healthcare for youth in the global South. BO and RB are employees of Shamiri Institute. The other authors have no conflicts of interest to disclose.
Appendices
Appendix A: Search Strategies
PubMed
(((((("mental disorders"[MeSH Major Topic]) AND ((prevalence[Title/Abstract]) OR epidemiology OR incidence) AND ((("africa"[All Fields]) OR (sub-sahara*) OR (Algeria) OR (Angola) OR (Benin) OR (Botswana) OR (Burkina Faso) OR (Burundi) OR (Cabo Verde) OR (Cameroon) OR (Central African Republic) OR (Chad) OR (Comoros) OR (Congo) OR (Cote d’Ivoire) OR (Djibouti) OR (Egypt) OR (Equatorial Guinea) OR (Eritrea) OR (Eswatini) OR (Ethiopia) OR (Gabon) OR (Gambia) OR (Ghana) OR (Guinea) OR (Kenya) OR (Lesotho) OR (Liberia) OR (Libya) OR (Madagascar) OR (Malawi) OR (Mali) OR (Mauritania) OR (Mauritius) OR (Morocco) OR (Mozambique) OR (Namibia) OR (Niger) OR (Nigeria) OR (Rwanda) OR (Sao Tome and Principe) OR (Senegal) OR (Seychelles) OR (Sierra Leone) OR (Somalia) OR (South Africa) OR (South Sudan) OR (Sudan) OR (Tanzania) OR (Togo) OR (Tunisia) OR (Uganda) OR (Zambia) OR (Zimbabwe)))
Limits: Humans, Child (birth-18), Adolescent, Young-adult, abstract, full text
Africa Journal Online
mental (health OR illness OR disease OR disorder) AND (prevalence OR incidence) AND (young adult OR adolescent OR child)
AfriBib
Psychiatry
PsycInfo
(mental health OR mental illness OR mental disorder OR psychiatric illness OR behavioral disorders OR conduct disorders OR attention deficit hyperactivity disorder OR oppositional defiant disorder OR autism spectrum disorder OR anxiety OR depression OR Obsessive-Compulsive Disorder OR eating disorder OR bipolar OR schizophrenia? OR Post Traumatic Stress Disorder) AND (youth OR adolescent OR child* OR young adult) AND (Africa OR sub saharan OR (Algeria) OR (Angola) OR (Benin) OR (Botswana) OR (Burkina Faso) OR (Burundi) OR (Cabo Verde) OR (Cameroon) OR (Central African Republic) OR (Chad) OR (Comoros) OR (Congo) OR (Cote d’Ivoire) OR (Djibouti) OR (Egypt) OR (Equatorial Guinea) OR (Eritrea) OR (Eswatini) OR (Ethiopia) OR (Gabon) OR (Gambia) OR (Ghana) OR (Guinea) OR (Kenya) OR (Lesotho) OR (Liberia) OR (Libya) OR (Madagascar) OR (Malawi) OR (Mali) OR (Mauritania) OR (Mauritius) OR (Morocco) OR (Mozambique) OR (Namibia) OR (Niger) OR (Nigeria) OR (Rwanda) OR (Sao Tome and Principe) OR (Senegal) OR (Seychelles) OR (Sierra Leone) OR (Somalia) OR (South Africa) OR (South Sudan) OR (Sudan) OR (Tanzania) OR (Togo) OR (Tunisia) OR (Uganda) OR (Zambia) OR (Zimbabwe)) AND (prevalence OR epidemiology OR incidence)
Appendix B. Eligibility Criteria for Included Studies
Pre-screening criteria:
-
1. Youths: The mean age is 0–19 years old. It can include studies that report a different mean age if the study reports results broken down by age and includes data on 0–18 years old.
-
2. English: The full article is available and published in English.
-
3. SSA: The study occurred in one or more countries in sub-Saharan Africa.
-
4. Empirical article: The article is an original empirical study (i.e., presents data collected for this study). Exclude meta-analyses, systematic reviews and narrative reviews.
Screening criteria:
-
1. Prevalence measure: Reports mental disorder prevalence using a measure of mental health diagnostic status or an established (i.e., psychometrically validated in some setting) measure of symptom levels.
-
a. Include studies that report the percentage of youth with a diagnosis or meeting a cutoff, and those that report raw numbers of those above a cutoff score on a scale.
-
b. If a study reports an odds ratio or prevalence ratio, but NO prevalence measure (i.e., % or raw number), do not include it. If a study reports an odds ratio and a prevalence measure of some kind, include it.
-
c. Do not include articles that only provide a mean score on a scale.
-
d. Do not include articles that only use one unvalidated item, or an unestablished collection of items, to measure the presence or absence of a disorder.
-
-
2. Selected disorder: Prevalence measure is of one or more of the following disorder types: anxiety problems (of all kinds, including OCD and specific phobias), depression problems, conduct problems, ADHD problems and post-traumatic stress disorder.
-
3. Population: The study is of the general population or a special population (e.g., youth with HIV, refugees) that is not already selected or self-selected for the presence of mental health problems or symptoms (i.e., the study cannot be of a group of people already seeking mental health services).
-
a. Example exclusion: Prevalence of anxiety and depression amongst a sample of youth seeking treatment at a mental health clinic.
-
b. Example exclusion: Prevalence of anxiety and depression amongst a sample of patients reporting post-traumatic disorder.
-
c. Example inclusion: Prevalence of post-traumatic stress disorder amongst children affected by conflict.
-
4. Sampling strategy: Researchers employed a probabilistic sampling strategy to recruit participants to their study.
Appendix C. Psychometric Scale Used
Appendix D: Results of Sensitivity Analysis after Removal of Studies with Special Populations and Non-Validated Tools
Note: This table displays the results of sensitivity analyses that were conducted to assess heterogeneity after the removal of studies that included special population(s) or non-validated tools, outliers and influential cases and screening tools, respectively. Heterogeneity remains high (I 2 > 80%) despite the removal of the above-mentioned studies across all conditions.
Appendix E: Results of Subgroup Analysis of General vs Special Populations
Note: No significant differences were observed between general and special populations for all conditions.
Appendix F: Results of Subgroup Analysis of Validated vs Non-Validated Tools
Note: Significant differences were observed between studies, including validated compared to non-validated tools for PTSD and anxiety. In both cases, studies using non-validated tools were disproportionately less and had a higher reported pooled prevalence than those using validated tools.
Appendix G: Results of Subgroup Analysis of Screening vs Diagnostic Tools
Note: Significant differences were observed for depression and conduct disorder, where the pooled prevalence of studies using a diagnostic tool was lower than studies using a screening tool.
Open access
Comments
Shamiri Institute
CMS-Africa, 13th Floor,
Pioneer Point, Chania Ave,
Nairobi, Kenya
20 December 2023
Dear Professor Judy Bass and Professor Dixon Chibanda,
My co-authors and I have enclosed for your review a copy of our manuscript entitled “Meta-Analysis: Prevalence of Youth Mental Disorders in sub-Saharan Africa” for your consideration Cambridge Prisms: Global Mental Health. This manuscript has not been published elsewhere and is not currently under consideration for publication in another journal. Our manuscript was prepared according to the submission guidelines of Cambridge Prisms: Global Mental Health. This meta-analysis has been pre-registered on the PROSPERO registry (CRD42022326574). This study was supported by the Shamiri Institute, a non-profit organization in Kenya and the United States that strives to provide accessible mental healthcare for youth in the global South.
We believe that our manuscript and our organization’s aim align well with the Cambridge Prisms: Global Mental Health journal’s scope, particularly in addressing treatment gaps and disparities in care, access, and capacity. In this manuscript, we review peer-reviewed articles pertaining to psychiatric conditions among sub-Saharan African youth. The 63 articles in this study give a never-before-seen glimpse into the prevalence of five common psychiatric conditions: conduct disorder, depression, anxiety, attention deficit hyperactivity disorder, and post-traumatic stress disorder among sub-Saharan African youth. Our manuscript also compares the prevalence of these conditions in sub-Saharan Africa to global estimates. While previous meta-analyses have explored therapies for youth in sub-Saharan Africa, to our knowledge, a study to classify the burden of common mental disorders among this population has not yet been published. We believe this is a key gap in the literature, which our manuscript fills.
We believe this manuscript will interest Cambridge Prisms: Global Mental Health’s readership, particularly because of its distinct analysis of five conditions, the breadth of studies included, and its international focus. All authors approved the final submitted manuscript and agreed to submit it.
Below are the authors’ full names and affiliations at the time of the review,
Cecilia E. Jakobsson, King’s College London, England & Shamiri Institute, Kenya
Natalie E. Johnson, Shamiri Institute, Kenya & Division of Clinical Epidemiology, University Hospital Basel, Switzerland
Brenda Ochuku, Shamiri Institute, Kenya
Rosine Baseke, Shamiri Institute, Kenya
Evelyn Wong, Shamiri Institute, Kenya & Stanford University School of Medicine
Christine W. Musyimi, Africa Mental Health Research and Training Foundation, Kenya
David M. Ndetei Africa Mental Health Research and Training Foundation, Kenya & Department of Psychiatry, University of Nairobi, Kenya & World Psychiatric Association Collaborating Centre for Research and Training, Kenya
Katherine E. Venturo-Conerly, Shamiri Institute, Kenya & Department of Psychology, Harvard University, USA
We thank you for your time and consideration and look forward to hearing from you.
Sincerely,
Cecilia Jakobsson,
Trainee Clinical Psychologist