Background

Depression is a global health issue

Depression is a common mental disorder that affects about 280 million people globally and is the second leading cause of disability worldwide [1, 2]. Depression is a mood disorder characterized by a persistent feeling of sadness and loss of interest accompanied by somatic and cognitive changes that affect an individual’s ability to function [3]. The lifetime prevalence of depression varies by sex and region, ranging from 2.6% among males in the World Health Organization (WHO) Western Pacific Region to 5.9% among females in the WHO African Region [Sub-Saharan Africa (SSA)] [4]. The forty-eight African countries that lie south of the Sahara make up SSA. It is the poorest region in the world and contains twenty-four of the twenty-seven countries in the World Bank’s Low-income classification [5]. Although the estimated prevalence of depression in SSA is considerably high, the true prevalence is possibly higher due to underdiagnosis caused by stigma, paucity of mental health services, and inadequate research [6]. About 75–90% of people with depression and other mental disorders in low-and-middle-income countries (LMICs) do not receive treatment [7, 8]. This “treatment gap” is higher in many SSA countries. For instance, it is 99.8% in Sierra Leone [9]. Lack of treatment despite increasing prevalence results not only in disability but also productivity losses, which cost the global economy $1trillion annually [10].

Depression among young people

The term "young people" refers to adolescents and youths. Adolescents are young people aged 10 to 19, while youths are between ages 15 and 24 [11]. Depression is the second most prevalent mental disorder among young people [2] and warrants increased attention for the following reasons. First, depression usually starts during adolescence and persists into adulthood, especially when undiagnosed or inadequately treated. Research shows that 50% of Common Mental Disorders (CMDs) appear by age 14 and three-quarters by age 24 [12]. The prevalence of depression is low in children (1%), then rises substantially during adolescence, especially among females [13]. This observed increase is attributable to the sociobiological changes typical of the post-pubertal phase, such as increased social understanding and self-awareness, changes to the brain circuits involved in responses to reward and danger, and elevated stress [14, 15]. Secondly, the prevalence of adolescent depression is rising. A recent meta-analysis revealed that the global point prevalence of depression in young people (25.2%) had doubled from pre-pandemic estimates [16]. The pandemic exacerbated an already rising prevalence observed in both High Income Countries (HICs) and LMICs over the last two decades [17,18,19,20]. Currently, the point prevalence of adolescent depression in LMICs ranges from 18% in China to 51% in Zambia [21]. Thirdly, the negative consequences of depression in young people are enormous. Adolescent depression is a major risk factor for suicide, which is a leading cause of death among young people, particularly in LMICs [22]. Furthermore, depression is associated with self-harm, substance use, risky sexual behaviour, and poor educational attainment [23, 24]. Although prevention, early diagnosis, and treatment can reduce its high burden and negative outcomes, 80% of young people with CMDs in LMICs and almost 100% in many SSA countries do not get the care they need [25].

Depression among young people in Sub-Saharan Africa

Young people account for a third of SSA’s population [26]. One in ten young people in SSA suffers from CMDs, particularly anxiety and depression [27]. A recent systematic review estimated the point prevalence of depression among young people in SSA at 26.7% and shows that the prevalence in many SSA countries is higher than the global estimate [28]. This is likely due to contextual risk factors such as poverty, conflict, poor healthcare, HIV, and teenage pregnancy, in addition to those common in HICs (e.g., parental psychopathology) [27]. Despite this, only nine of the forty-eight countries in SSA have comprehensive policies for adolescent mental health, resulting in huge barriers to care [29]. It is on this premise that WHO, in its landmark World Mental Health Report, calls for contextually appropriate, cost-effective interventions for adolescent depression in SSA [22]. For every $1 invested in these interventions, Stelmach et al. expect $125 in health and economic benefits returned to the regional economy [25].

Psychosocial interventions for depression in young people

Psychosocial Interventions for mental disorders are interpersonal or informational activities/techniques that influence outcome through changes in mediating biopsychosocial factors. They include psychological therapies like Cognitive Behavioural Therapy (CBT), Interpersonal Psychotherapy (IPT) and Psychodynamic therapy, and social interventions like peer support services and skill building [30]. Psychosocial interventions are the first-line approach for depression in young people and antidepressants should be used only in cases unresponsive to psychological therapy [31, 32]. Research over the years have established the efficacy of psychosocial interventions in the treatment of adult depression [33,34,35]. Given this evidence, different interventions have been adapted for adolescent and youth populations. CBT and IPT are the most extensively tested in young people and reviews have shown that they reduce depressive symptoms [36,37,38]. Attachment-based family therapy, though less extensively researched, has also shown some positive effect [39]. These interventions have proven effective when delivered in individual and group formats [40, 41] and via bibliotherapy or technology-assisted methods, although to varying degrees [42, 43]. They have also been delivered in different settings, such as schools and communities [38].

Majority of the studies that established these interventions as evidence-based were conducted in HICs and less is known about their effectiveness in LMICs. In recent years, these interventions have increasingly been tested in LMICs. Findings show that interventions developed in HICs might not be acceptable, feasible, or effective in LMICs due to contextual differences such as dissimilar cultural perceptions of depression, and barriers to care (e.g., low awareness, insufficient mental health workers, stigma, and poverty) [44]. Innovative solutions like task-shifting (use of non-mental health professionals), cultural adaptation, and the use of digital technologies have been tested with mixed results [45, 46]. There is a need to understand how these interventions are adapted to fit different contexts and how these modifications affect their effectiveness. Systematic reviews of LMICs involve only a few SSA countries, thus limiting their applicability to the region [44, 47, 48]. Though it remains unclear which psychosocial interventions are most effective in the region, no systematic review has been conducted on this topic. This review aims to identify and describe psychosocial interventions for depression among young people in SSA, determine their efficacy and explore factors that affect their efficacy. As most SSA countries do not have policies for young people’s mental health, findings from this study will contribute to future research and policy development.

Methods

This systematic review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [49] [Additional Files 1, 2].

Search strategy

The search strategy was developed with the assistance of a research librarian and the systematic database search was first conducted in July 2022 (updated May 2024) with keywords identified using the PICO framework [50] (Table 1). The keywords with their MeSH terms and synonyms were combined with Boolean operators (“AND” and “OR”) and wildcards (*, ?) to run a comprehensive search on Medline (OVID). This search strategy was then adapted to Web of Science, PsycInfo and Cochrane Central Register of Controlled Trials (CENTRAL). The detailed search strategy for each database is shown in Additional File 3. The reference lists of included articles were also searched to identify other relevant papers.

Table 1 Identification of search terms using PICO framework
Full size table

Eligibility criteria

The inclusion and exclusion criteria were generated using the PICOS framework [50], as shown in Table 2. Studies that reported the effect of various psychosocial interventions on depressive symptoms in any adolescent or youth population in SSA were included. To best capture the state and quality of research, papers were not included or excluded based on study design or quality assessment. Due to limited resources for translation, only papers in English language were included.

Table 2 Inclusion and exclusion criteria
Full size table

Data management and extraction

All records captured by the search terms were exported to EndNote 20 Library. After de-duplication, titles and abstracts were independently screened by two reviewers (LO and EU). Papers that did not meet the eligibility criteria were excluded. The full texts of the remaining papers were screened by same reviewers against the eligibility criteria. A data extraction form was developed to extract relevant information from the papers such as country, study design, intervention setting, screening instrument, intervention characteristics, and outcome (depressive symptoms pre- and post-intervention). Three tables were developed from this form and are presented in the results section.

Risk of bias assessment

The Cochrane Risk of Bias tools for randomized control trials (RCTs) version 2 (RoB2) and Risk of Bias in Non-randomized Studies of Interventions (ROBINS-I) were used to assess the risk of bias (RoB) for RCTs and Non-Randomized Studies of Interventions (NRSI) respectively [50]. These specific RoB tools were used due to the methodological differences between study designs, particularly randomization which is an important consideration in judging bias. For RCTs, six RoB domains (randomization process, deviation from intended intervention, missing outcome data, measurement of outcome, and selective reporting) were assessed and studies judged to have high, “some concerns” or low RoB. For NRSI, six domains (confounding, participant selection, deviation from intervention, missing data, outcome measurement, and selective reporting) were assessed. Studies were judged as having a low, moderate, serious, or critical risk of bias [50].

Meta-analysis

The meta-analysis included clinical trials wherein participants were randomly allocated to either receive a psychosocial intervention or be placed in control conditions, with depression scores reported as an outcome. Statistical analysis was performed using R software (version 4.3.1) and the metafor package (version 4.4.0). To ensure uniformity and reproducibility of results, standardized mean differences (SMD) alongside 95% Confidence Intervals (CI) were calculated for each study using extracted data (mean and standard deviation). SMDs were employed because studies used different screening instruments to evaluate depression scores [50]. Pre- to post-intervention changes were analysed but follow-up impacts were not considered due to a lack of information in some studies and variations in follow-up periods. Effect sizes (SMD) were calculated using Hedges’ g because it corrects for small sample bias. This was pertinent given the relatively small sample sizes in some included studies [51]. Effect size magnitudes were categorized as small (0.20–0.50), moderate (0.50–0.80), and large (> 0.80) based on Cohen's rule of thumb [52]. A random-effects model was used a priori to account for expected heterogeneity among studies, including variations in intervention types, delivery modalities, participant characteristics, and screening instruments. Statistical heterogeneity was assessed using Cochran’s Q test and I2. An I2 value of 0 to 25% can be considered as low, 50% as moderate and 75% and above as a high level of heterogeneity. Subgroup analysis was pre-determined to explore variations in psychosocial interventions [52]. Sensitivity analysis was conducted by excluding studies with high risk of bias to assess the robustness of findings. Publication bias was evaluated using a funnel plot and Egger’s regression test [53].

Results

The search strategy identified 1,638 papers across the four databases. After de-duplication and title and abstract screening, 67 papers were sought for full-text screening. One could not be retrieved despite attempts to contact the author. Forty-five papers were excluded after full-text screening and one paper was identified via citation searching. Reasons for exclusion are listed in Additional File 4. Hence, a total of 22 studies were included in the review. The PRISMA Diagram illustrates the selection process (Fig. 1).

Fig. 1
figure 1

PRISMA flow diagram illustrating steps to paper selection

Full size image

Description of included studies

The twenty-two included studies were conducted across ten different SSA countries; nine in Nigeria [54,55,56,57,58,59,60,61,62], four in Kenya [63,64,65,66], two each in the Democratic Republic of Congo (DRC) [67, 68] and South Africa [69, 70] and one each in Botswana [71], Burundi [72], Mauritius [73], Rwanda [74], and Uganda [75]. Ninteen studies were RCTs, while three were NRSIs. Among the NRSI, one was a controlled clinical trial (CCT) [55] while two were pre-post intervention studies [54, 61]. Seventeen studies included only adolescents while the other six included youths. Table 3 shows the important characteristics of all included studies. The studies were carried out in different settings and targeted different populations. Most studies (fourteen) were conducted in schools; nine in secondary schools, and six in tertiary institutions. While the majority of interventions included the general student population, others targeted specific populations with increased risk of depression such as students with sickle cell disease in Nigeria (SCD) [54], internally displaced [63] and students from low-income families in Kenya [64,65,66], and orphaned students in Rwanda [74]. Among the non-school-based interventions, three were delivered in clinics to adolescents living with HIV in Botswana [71], trauma-exposed adolescents in South Africa [70] and depressed adolescents attending a psychiatric clinic in Nigeria [61]. The others were delivered in communities to war-affected adolescents in DRC [67, 68, 75] and orphans of the HIV epidemic in South Africa [69]. Some studies excluded participants with co-morbid psychiatric disorders, substance use, intellectual difficulties, and suicidality [55, 56, 60, 62, 68, 74]. Others also excluded participants based on depression screening. For example, while Are et al. [55] excluded people with severe depression, Eseadi et al. [58] included only individuals with moderate-to-severe depression. In contrast, three studies did not apply these exclusion criteria [64, 65, 67]. McMullen et al. noted that this was done to “keep the trial as naturalistic as possible” [67]. Overall, eleven studies had relatively small sample sizes (less than 60).

Table 3 Description of included studies
Full size table

Nine different screening instruments were used to assess depressive symptoms, with Beck’s Depression Inventory (BDI) being the most used. All the instruments, except the Acholi Psychosocial Assessment Instrument (APAI), were developed in western countries. APAI was developed and used in Uganda by Bolton et al. [75]. It was modified into the AYPAI and used in the two DRC studies. The reliability of APAI was 84%, similar to BDI, while AYPAI was 74% [67, 68].

All but five studies [54, 55, 65, 74, 75] reported mean depressive symptoms after a follow-up period (Table 3). The majority of studies adopted a three-month follow-up period [57, 59, 67, 68, 72]. The longest follow-up period was twelve months [69] while the shortest was 1 week [63]. One study had no follow-up [65] while another [60] did not report the length of follow-up.

Intervention characteristics

The majority of studies (13) tested the effect of CBT-based interventions on depressive symptoms [54,55,56,57,58,59,60,61,62, 67, 68, 70, 72]. Two studies each exammined the effects of Interpersonal Psychotherapy (IPT) [69, 75] and Wise Interventions (WI) [64, 65]. Three studies tested Creative Psychological Interventions (CPI) such as Expressive Writing (EW) [63, 74] and arts-based therapy [66]. The other interventions combined two psychotherapautic approaches. The Resourceful Adolescent Program (RAP), a universal preventive programme, combined both IPT and CBT techniques [73] while the intervention by Olashore et al. [71] combined psychoeducation and problem-solving. Table 4 shows the important intervention characteristics of each study.

Table 4 Intervention characteristics
Full size table

CBT-based interventions

All but two CBT-based studies tested manualized interventions delivered in group face-to-face format. One delivered CBT in an individual format [70] while the other was an online guided self-help intervention [62]. The two group Trauma-Focused CBT (TF-CBT) also included a few individual sessions for trauma narration to “prevent vicarious traumatization” [67, 68]. The different manuals used across studies included core CBT elements like psychoeducation, cognitive restructuring, activity scheduling, problem-solving, and relaxation techniques, as shown in Table 4. The manual used by Are et al. [55] was developed in Nigeria by one of the co-authors and used in two other Nigerian studies [56, 61]. This is in contrast with other studies which used manuals developed in Western countries. Two interventions added other elements to CBT. The Guided Internet Assisted Intervention (GIAI) combined CBT techniques with interactive peer support [62], while another study [72] added creative expressive elements to CBT.

In terms of intensity, all the interventions can be considered Low-intensity as they were either delivered in high volume (group format) and/or by non-mental health professionals, or as self-help. However, they varied in duration, with seven interventions lasting 5 weeks or less and six lasting 8 weeks or more (Table 4). Although some interventions lasted 5 weeks, they had multiple sessions per week (15 sessions overall) in contrast with others, which had one session per week (five sessions). Eight interventions were delivered by mental health professionals, while two were delivered by non-professionals (e.g., teachers and lay healthworkers). In the guided self-help, guidance was provided by therapists [62].

Interpersonal psychotherapy

Like most CBT-based studies, the two IPT studies were manualized interventions delivered in group face-to-face format. One intervention was delivered to Ugandan adolescents displaced by war [75] while the other was for HIV orphaned adolescents in South Africa [69]. In both studies, 1.5-2 hr weekly sessions were delivered by lay facilitators for 16 weeks, using the same IPT manual developed by a humanitarian organization. Though one study [75] randomized participants to intervention (IPT), and two control groups (Creative Play and Waitlist), they only analyzed the IPT group against the waitlist group.

Other intervention types: wise interventions and creative psychological interventions

Wise Interventions are a novel class of ordinary, briefer, and precise positive psychological interventions aimed at altering a specific way in which people think or feel [76]. The WIs (Shamiri) are the first positive psychology intervention to combine three WIs (Growth Mindset, Value Affirmation, and Gratitude). Shamiri means “thrive” in Kiswahili, which reflects the intervention’s focus on positive psychology rather than mental illness [64, 65]. The two WIs were conducted among low-income students in an urban slum in Kenya. One intervention [64] was delivered in a group format over 4 weeks by former high school graduates, while the other (Shamiri Digital) [65] was a single-session digital self-help intervention.

Three studies examined the effect of Creative Psychological Interventions on depressive symptoms. CPIs encompass a variety of psychotherapautic techniques that utilize creative and expressive forms of communication and expression to address psychological and emotional issues [77]. Two of these studies tested Expressive Writing. Writing for Recovery (WFR) [63] adopted a structured testimonial/narrative approach in communicating an emotional experience to normalize distressing reactions while Emotional Writing was an unstructured writing intervention for HIV-orphaned adolescents and involved participants writing about their deepest emotions concerning their loss [74]. The other CPI was conducted among high school students and employed art-based psychotherapeutic approaches to facilitate psychological change [66].

Cultural/contextual adaptation

As shown in Table 4, sixteen studies reported some form of adaptation to suit the local culture and context, while six did not. The most common form of adaptation was the delivery of the intervention in the local language. This involved group discussions in the local language and/or translation of the screening instrument and manual. Three studies [67, 68, 75] used locally developed screening instruments and three used locally developed manuals [55, 56, 61]. Eseadi et al. adapted the intervention manual to incoporate the religious philosophies and traditions of the participants. Another common adaptation was the use of local metaphors, and culturally applicable analogies and exemplars. The two Trauma-Focused CBT in the DRC also used cultural games and songs [67, 68]. Are et al. encouraged the use of helpful cultural and religious coping mechanisms while Osborn et al. encouraged the use of local arts, languages and traditions [55, 66]. Three interventions ensured rigorous cultural appropriateness by involving community stakeholders in the development and implementation of the interventions [63,64,65].

Risk of bias assessment

Eleven of the ninteen RCTs were judged as having “some concerns”. Five studies were judged high risk, while three had a low RoB (Fig. 3). Bias arose mainly from the ‘Randomization Process’ and ‘Measurement of the Outcome’ domains (Fig. 2).

Fig. 2
figure 2

Risk of bias summary graph

Full size image

Figure 3 presents the RoB for each RCT. In the Randomization Process, studies were judged high or had “some concerns” due to lack of information on allocation concealment [62, 64, 66, 69,70,71,72,73] and/or significant differences in prognostic variables between intervention and control groups [72, 74]. For instance, in one study, participants differed significantly in baseline depressive symptoms, social support, and experience of traumatic events [72]. In the Measurement of the Outcome domain, five studies [56, 59, 60, 62, 72] were judged to have “some concerns” due to unblinded outcome assessors. Two studies [63, 73] were rated high because unblinded intervention facilitators supervised the post-intervention completion of screening instruments, making outcome assessment more likely to be influenced by knowledge of allocation. For Deviation from Intended Intervention, one study [62] was rated high because it neither reported information on intervention fidelity nor whether participants were analyzed in the groups they were randomized to (e.g. using Intention-To-Treat(ITT) analysis). Furthermore, 41% of participants dropped out and were excluded from the outcome analysis. Five studies had “Some Concerns” due to lack of information on intervention fidelity and or analysis methods [58, 63, 73] or reported protocol deviations [69, 75]. In the Missing Outcome Data domain, most studies had low RoB. However, one study [67] was rated high due to potential outcome-related missing data. One study [61] had “some concerns” due to lack of missingness accounting and sensitivity analysis. Only one study [63] was judged as having a high RoB in the Selection of Reported Results Domain because it used multiple methods to assess treatment effects but reported only one set of results. For the Non-randomized studies, the two pre-post intervention studies [54, 61] had serious RoB while the controlled trial [55] had a low RoB (Table 5).

Fig. 3
figure 3

Risk of bias judgment for included RCTs

Full size image
Table 5 Risk of Bias Judgement for Non-Randomized Studies of Interventions (NRSI)
Full size table

Effect of psychosocial interventions

Table 6 shows the results of each study, including baseline and post-intervention mean depression scores, treatment effect and factors affecting efficacy. Overall, 19 studies, of which 12 were CBT-based, reported statistically significant reductions in depressive symptoms in the intervention compared to control groups. Although the follow-up durations varied between studies, majority of the CBT interventions (nine) maintained their effects at follow-up. Although both IPT interventions were similar (same manual, duration, and delivery personnel), results were mixed. Bolton et al. [75] reported a significant decrease from baseline depressive symptoms (P = 0.02) only amongst girls, while Thurman et al. [69] revealed no significant decrease (P = 0.145). Despite the differences in their delivery formats, both Wise Interventions reported significant decreases from baseline mean depression scores in intervention groups compared to control [64, 65]. Interestingly, the single-session digital version (Shamiri Digital) [65] showed a more significant reduction and larger effect size (P = 0.028, Cohen’s d = 0.5) than the four-session group face-to-face version (Shamiri Group)(64) (P = 0.038, d = 0.32). Two of the three creative psychological interventions showed significant effects. “Pre-Texts”, the arts-based therapy showed a significant reduction in depression scores (P = 0.001, d = 0.52). The structured writing intervention, Writing for Recovery (WFR) [63] resulted in a significant decrease in depressive symptoms (P = 0.0001) with a large effect size (np2 = 0.338). In contrast, the unstructured writing intervention showed no significant change from baseline (P = 0.518) [74]. The two other interventions had small effect sizes. The universal preventive intervention, Resourceful Adolescent Programme (RAP) [73], resulted in a significant decrease from baseline depressive symptoms (P < 0.001), with a small effect size (d = 0.32). However, the effects were not maintained at 6-months follow-up. The intervention which combined psychoeducation and problem-solving also yielded a small effect (ƞp2 = 0.20, p = 0.001). Overall, both culturally adapted and non-adapted interventions showed positive effects.

Table 6 Results
Full size table

Meta-analysis

The meta-analysis incorporated data from 18 RCTs involving 2338 participants. One study, identified as an extreme outlier due its exceptionally large effect size (Hedges’ g = −33) and high risk of bias was excluded from the analysis [62]. The random effects model revealed a significantly large effect of psychosocial interventions compared to the control groups (Hedges’ g = −1.55, 95% CI −2.48, −0.63). Figure 4 presents the forest plot, displaying the effect sizes (Hedges’ g) for each study, the pooled effect size and 95% confidence intervals. Negative effect sizes indicate a more favorable outcome (symptom reduction) for the psychosocial intervention groups relative to the control groups. The analysis revealed high heterogeneity among the included studies (I2 = 98.8%, Q = 574, p < 0.0001), indicating significant variability. Sensitivity analyses, excluding studies with a high risk of bias amplified the effect size (Hedges’ g = −1.96, 95% CI −3.5 to −0.86), although the heterogeneity remained unchanged. However, removing the four outliers whose effect sizes were much larger than the others (g > 4) resulted in a reducted effect size (g = −0.59, 95% CI −0.98, −0.2) and decreased heterogeneity (I2 = 93.8%). Interestingly, these outliers were the group CBT interventions with the longest duration (12 weeks) and were all conducted among undergraduate students in Nigeria [57,58,59,60]

Fig. 4
figure 4

Forest plot – effects of Psychosocial Interventions of depressive symptoms. Negative effect sizes signify a more favorable outcome (symptom reduction) for the psychosocial intervention groups relative to the control groups

Full size image

Subgroup analysis, depicted in Fig. 5, indicated significant differences in effect size by intervention type (Test for subgroup differences: p = 0.027). CBT-based interventions represented in nine studies, had the strongest effect (Hedges’ g = −2.84, 95% CI −4.29; −1.38), albeit with high within-group heterogeneity (I2 = 99.7%). Other subgroups, with fewer studies, exhibited smaller average effects. Wise Interventions had a moderate effect (g = −0.46, 95% CI −0.53, −0.39) while Interpersonal Psychotherapy (g = −0.08, 95% CI −1.05, 0.88) and Creative Psychological Interventions (g = −0.29, 95% CI −1.38, 0.79) showed small non-significant effects. Within-group heterogeneity was high for IPT and CPI (99.9% and 99%, respectively), while WI showed no heterogeneity (I2 = 0%).

Fig. 5
figure 5

Forest plot for sub-group analysis

Full size image

Publication bias was assessed using funnel plots and Egger’s test, which indicated significant funnel plot asymmetry (p < 0.0001), raising concerns about potential publication or small study bias (Fig. 6). However, Duval & Tweedie’s trim and fill analysis estimated no missing studies. The pooled effect size was maintained after the small-study bias adjustment, suggesting the robustness of the findings [78].

Fig. 6
figure 6

Funnel plot for assessing publication bias. The observed funnel plot asymmetry should be interpreted cautiously, considering the high heterogeneity among the included studies

Full size image

Factors affecting intervention efficacy: predictors, moderators and mediators

As shown in Table 5, majority of studies neither assessed potential predictors nor conducted moderator and mediation analysis. Three studies conducted statistical analysis to determine what baseline variables would predict outcomes [55, 64, 75]. Gender predicted the outcome in one study [75], as the intervention was only effective for girls, but it was not an outcome predictor in the others [55, 64]. Age did not predict outcome in any of the three studies.

Age moderated the efficacy of Shamiri Digital [65], as younger adolescents reported a larger decline in depressive symptoms. Contrastingly, neither age nor gender were moderators in the study by Thurman et al. [69]. Results for level of adherence as a moderator were also mixed, as increased adherence was associated with symptom reduction in one study [75] but not in another [59]. The severity of depression was an effect moderator in Shamiri Digital as there was better response among adolescents with moderate-to-severe depression scores [65].

Increase in coping, self-esteem, hope, and knowledge of depression were identified as possible mediators [54, 55, 61].

Discussion

This systematic review is the first to assess the efficacy of psychosocial interventions for depression among young people in Sub-Saharan Africa (SSA). The meta-analysis showed that these interventions, particularly CBT, significantly reduce depressive symptoms, although substantial heterogeneity necessitates cautious interpretation of pooled estimates. Subgroup analysis indicated significant variation in efficacy by intervention type. CBT was shown to be the most effective intervention, corroborating findings from systematic reviews in other contexts [21, 38]. Wise Interventions (WI) showed moderate effects, while IPT and CPI had small effects. However, limited number of studies per group may affect the reliability of the subgroup estimates. The WIs particularly the single session intervention (SSI) had effect sizes comparable to some well-established psychological treatments, which is interesting because WIs were not originally developed for depression [65, 76]. A similar study in the US found a digital single-session WI to be moderately effective in reducing adolescent depressive symptoms [79]. As positive psychology interventions, WIs could help address stigma, which is a major barrier to treatment in SSA. However, more studies are needed to ascertain the durability and reproducibility of their effects. A review by Weersing et al. [38] found IPT to be effective among young people but less so in group format, which could explain the lower efficacy of Group-IPT compared to Group CBT in our review. Furthermore, the effective Group-IPT studies from their review were conducted among the general population, as opposed to war-affected adolescents in this review. In contrast, Group TF-CBT studies showed large effects, suggesting that they might be better suited than IPT for trauma-affected adolescents in SSA.

Exploring sources of heterogeneity

While the sensitivity analysis and publication bias adjustment suggest robust findings, the high overall and subgroup heterogeneity necessitate cautious interpretation of pooled effect sizes. Thus, it might be more informative to focus on understanding the sources of variation across studies rather than relying solely on the pooled estimate. Subgroup analysis revealed high within-group heterogeneity for all subgroups except WI, indicating that variability extends beyond intervention types. Differences in study population and design, and intervention delivery may contribute to this variability. For instance, within CBT studies, heterogeneity likely arises from differences in participant characteristics, and intervention delivery methods. Studies with homogenous characteristics and delivery modalities, such as the culturally adapted group trauma-focused CBT (TF-CBT) for war-affected adolescents in DRC [67, 68], showed consistent effect sizes (g = −2.10, −1.95). In contrast, the individual TF-CBT for South African adolescents with moderate and severe PTSD had a smaller effect size (g = −0.51). The comparatively smaller effect size is attributable to the study’s inclusion of more clinically severe cases and its use of a treatment-as-usual control compared to the non-clinical population and waitlist control in the DRC studies. The group CBT intervention delivered to war-affected secondary school students in Burundi exhibited a much lower effect size (g = −0.06) than the other studies involving trauma-affected youths, likely due to its non-trauma-focused approach [72].

The other CBT interventions targeted general populations of high school and undergraduate students and also exhibited varying effects likely due to dissimilar population, inclusion criteria and intervention duration. For example, the extensive group CBT studies (12 week duration) delivered to a general population of undergraduate students in Southeast Nigeria [57,58,59,60] yielded much larger effect sizes (g =−4.08 to −6.25) compared to the shorter interventions (5 weeks) delivered to clinical populations of secondary school students in Southwest Nigeria (g = −1.04, −1.26) [55, 56].

The two group IPT trials showed disparate effects potentially due to differences in cultural adaptation and trauma exposure. Culturally adapted IPT for war-displaced adolescents had a moderate effect size [75] while non-adapted IPT delivered to HIV-orphaned adolescents showed no effect [69]. The CPI subgroup also had significant heterogeneity, possibly stemming from variations in participant and intervention characteristics. For instance, the structured EW intervention was effective for displaced secondary school students in Kenya [63], while unstructured EW showed no effect for orphaned adolescents in Rwanda [74]. Further studies are needed to ascertain the efficacy of EW as an intervention for depression and whether structured EW is more effective than unstructured. Unlike other subgroups, WIs showed no heterogeneity, likely because both studies were conducted by the same authors with similar participants and settings.

Durability of effect and factors affecting efficacy

The intervention effects were generally maintained at follow-up, but many studies had short or no follow-up periods. Considering the high relapse and recurrence rates among adolescents [83], future research is needed to determine the durability of effects [80]. Additional studies are also required to uncover factors influencing intervention efficacy. While age, gender and adherence level were assessed in few studies, their predictive or moderating roles varied. A systematic review had similarly mixed findings on predictors and moderators of efficacy [38]. This gap is a key area of focus for future research as identifying these factors can lead to the development of tailored interventions. The role adaptation plays in efficacy is also worth exploring. Generally, evidence from this review is mixed as with other reviews. In their review, Cuijpers et al. did not find any indication that a specific contextual adaptation was associated with better outcomes [47]. Contrarily, another systematic review found culturally adapted interventions more effective than non-adapted ones but it was unclear what specific adaptations were important [81]. Since cultural adaptation can increase acceptability and adherence which are both associated with increased effectiveness [82], future studies should examine what adaptations are beneficial in different SSA countries.

Intervention delivery modalities

All the interventions can be classified as Low-Intensity Psychological Interventions (LIPIs) due to reduced usage of therapist’s time (fewer sessions and/or high-volume delivery in group format), delivery by non-professionals, or as digital self-help interventions [83]. Compared to HICs and LMICs, this review found more interventions delivered in group format [21, 38]. Though group interventions are resource-effective hence better-suited for SSA, they are not suitable for everyone. For example, people with social phobia, interpersonal problems, recent traumatic events, and actively suicidal patients might be better served by individual therapy [84]. More studies are required to determine which people are better served by group vs individual interventions. Innovative approaches like embedding individual sessions in group interventions as done in the TFT-CBT studies [67, 68] can be further explored as they could prove more cost-effective. Interventions delivered by both mental health professionals and non-professionals were found to be effective. This concurs with a systematic review which found psychosocial interventions delivered by lay facilitators to be effective [85]. This finding is important for mental health policy in SSA as the region has the lowest ratio of mental health workers per population in the world [22]. Other novel innovative LIPIs were digital self-help and single session interventions (SSIs). Both self-help Digital Mental Health Interventions (DMHI) proved effective despite different durations (10-weeks vs single-session) and approaches (CBT-based vs WI/positive psychology). The evidence base for DMHI is growing and reviews in other contexts have found DMHI like computerized CBT to be effective in reducing adolescent depression [86]. An added advantage of digital SSIs is their ability to circumvent the high attrition often seen in multisession interventions [65]. Digital self-help interventions represent an opportunity to increase treatment access to young people in SSA due to increasing access to digital technology [87]. DMHI are also resource-effective and can help combat stigma thus warrant further exploration in the region.

Limitations

The high heterogeneity observed in this study introduces challenges to the certainty and generalizability of the pooled effect size estimate. This variability among included studies suggests that combining their effect sizes may not be ideal, as they might measure different effect sizes based on study population. However, interventions between studies varied substantially (e.g., content, frequency, duration, adaptation and delivery personnel), suggesting that high heterogeneity may be inevitable.

Publication bias remains possible even though no missing studies were estimated. Notably, the lack of change in effect size after adjustment raises the possibility that funnel plot asymmetry may be attributable to between-study heterogeneity rather than small-study bias. Funnel plots, by assuming that the dispersion of effect sizes is due to sampling error, do not control for the fact that studies may be estimators of different true effects, further underscoring the importance of interpreting the pooled effect size in light of heterogeneity [50].

Another important limitation is the inability to assess the durability of effects due to inconsistent reporting of follow-up time across included studies.

Conclusion

This study provides evidence supporting the efficacy of psychosocial interventions, particularly CBT, in alleviating depressive symptoms among young people in SSA. However, the observed heterogeneity highlights the importance of considering intervention types, delivery modalities, participant populations and factors affecting efficacy. Thus, while psychosocial interventions show promise in reducing youth depressive symptoms in SSA populations, further research to identify components that work best for specific subgroups is imperative. Tailored interventions for specific populations may be more effective than a one-size-fits-all approach.