Introduction

Mental healthcare in low- and middle-income countries (LMICS) has undergone a number of changes in recent years, both with regard to the conceptualization of mental illness, but also in relation to the envisaged design and modes of service delivery. Specifically, the advent of the recovery movement has spurred a growing interest in psychosocial rehabilitation and prompted clinicians to redefine their understanding of treatment from one that sets as its primary goal symptom reduction to a much broader conceptualization (Kramers-Olen 2014). This revised conceptualization includes an acknowledgement of individuals’ strengths, values, and attitudes towards themselves and their illnesses. This renewed vision of recovery seeks the establishment of a life that is fulfilled, satisfying, and meaningful, despite the limitations imposed by a debilitating illness (Parker 2012).

The changes indicated have occurred often within the context of amendments to legislature that have sought to directly impact upon the delivery of mental health services. In South Africa (SA), for example, one of the first changes to healthcare policy following the fall of apartheid was the introduction of the White Paper for the Transformation of the Health System (Department of Health 1997a) as well as the accompanying National Health Policy Guidelines for improved Mental Health in SA (Department of Health 1997b), which advocated for the delivery of mental health services based on primary healthcare principles. Thereafter, the Mental Healthcare Act was promulgated, which explicitly advocated for the rehabilitation of individuals with severe mental illness, defining this as “a process that facilitates an individual attaining an optimal level of independent functioning” (Republic of South Africa 2002). It made provision for the integration of mental health services with general health services at the primary healthcare level, and advocated for the deinstitutionalization of participants in long-term psychiatric care with commensurate development of community mental health services. These intended improvements, however, have not been without challenges. Deinstitutionalization policies were adopted but without the necessary development of community health services, and, consequently, “revolving door” readmissions continue to burden the tertiary psychiatric hospital system (Vally and Cader 2012) where a substantial proportion of psychiatrically ill individuals still continue to receive primary treatment (Lund et al. 2010). More recently, SA promulgated a new mental health policy, which adopts a recovery based approach to mental health and more directly highlights the role of individuals in their own recovery (Department of Health 2012).

This devolvement of mental health services in developing countries from the tertiary to community level necessitates exploration of the applicability of psychosocial rehabilitation interventions that reflect the ethos of recovery – those that are person-oriented and consider the individual beyond the symptom presentation, encourage the development of adaptive and social skills needed for success at daily life, are cognizant of the individual’s context and the demands it places on the individual, and promotes self-determination and independence as far as this is possible. Guided by a human rights paradigm and multi-system approach, potential intervention strategies for these individuals may include psychoeducation, skills training, vocational rehabilitation, family interventions, cognitive remediation, and peer support (Kleintjes et al. 2013; Kramers-Olen 2014). Peer-delivered services (PDS), in particular, are now regarded as an important component in the recovery process for those struggling with mental health difficulties. Those who provide PDS, sometimes referred to as peer specialists or consumer providers, occupy a unique position in that typically they are individuals who themselves have struggled with mental health difficulties but who have successfully engaged in their own recovery process and have a desire to, in turn, support others.

The growing and widespread interest in the recovery movement has resulted in a dramatic proliferation of PDS, despite preliminary evaluations of its efficacy yielding inconclusive results. Peer workers have been enlisted to deliver interventions in a variety of contexts and for a great number of purposes, such as reducing risky sexual behaviour (Lyons et al. 2006), harm reduction in intravenous drug users (Jain et al. 2014; Walsh et al. 2008), and for the self-management of chronic physical diseases (Tang et al. 2011). But its application to mental health outcomes appears particularly promising. For example, PDS has shown promise in supporting those with post-partum depression (Montgomery et al. 2012), bereaved children (Metel and Barnes 2011), and adolescents who have a parent with a mental illness (Hargreaves et al. 2008). Empirical evaluations of the effectiveness of these sorts of interventions have, however, been equivocal. Some have found PDS to result in immense benefits to consumers, including improvements with symptomology, quality of life (Rivera et al. 2007), self-empowerment, hope and self-esteem (Tse et al. 2014), and perceived stigmatization (Verhaeghe et al. 2008). Others though have found the benefits of having received peer support to be no better or worse than treatment delivered by a health professional (Chinman et al. 2014; Fuhr et al. 2014; Lloyd-Evans et al. 2014; Wright-Berryman et al. 2011).

It is important to conduct a review of the evidence in support of PDS, especially at this time. Mental illness accounts for a substantial proportion of the global burden of disease and contributes greatly to disability in society at large (World Health Organization 2001). It is predicted that rates of mental illness will continue to increase exponentially, and so too will the experienced disability and economic costs associated with it. These detrimental effects will be especially disabling in LMICS where the need for mental health services is substantial but the concomitant scarcity of trained professionals in these locales results in an ever-widening treatment gap (Patel et al. 2007). The feasibility and cost-effectiveness of delivering some forms of mental health intervention (e.g., culturally adapted cognitive-behavioural psychotherapy) in resource-poor contexts have been demonstrated and there is now preliminary evidence supporting their use in LMICS (Chibanda et al. 2015; Murray et al. 2013; Vally and Maggott 2015). But their widespread uptake is hampered by the inequitable distribution of resources, both human and financial. In response, some (e.g., Chatterjee et al. 2009; Cook 2011) have advocated for the adoption of task-shifting as a feasible strategy for reducing the treatment gap. This would require tasking the delivery of some mental health interventions that would normally be delivered by trained and certified practitioners to lay non-specialists or peers. But before this can be done with confidence, and certainly prior to the adoption of policies that would facilitate and promote this, an interrogation of the evidence supporting the use of PDS in impoverished environments is required, given that the great majority of investigations and the resulting evidence in support of the adoption of PDS originate in high-income countries (HICS).

Aims of the Study

This study sought to ascertain, within the context of a systematic review and meta-analysis, the extent to which mental health interventions delivered by peers implemented in the resource-constrained contexts of LMICS, where these sorts of interventions are arguably most needed, are effective, and to quantify the magnitude of this effect. The following types of intervention were included in this review:

  1. a.

    Mutual support groups, where the relationships between clients and providers are viewed as being equal and reciprocal, despite the fact that some individuals may be more experienced than others. A prime example of this model of support is the ubiquitous 12-step programs (e.g., Alcoholics Anonymous).

  2. b.

    Individually delivered peer support services delivered by peer mental health service providers. These individuals typically are users of mental health services, in recovery, who have been employed to provide discrete components of standard mental healthcare within the context of a formal mental healthcare service. This support is separate from or may be in addition to standard treatment provided by traditional healthcare practitioners. The peer provider may have volunteered to provide these services or may be receiving some form of payment to do so.

  3. c.

    For the purpose of this review, we have extended our definition of “peer” to also include community mental health workers, some of whom may not necessarily be consumers of services themselves, but rather lay, non-specialist individuals recruited from the community within which the respective intervention was implemented. Some studies refer to these types of lay community workers as peers or mentors, to indicate that these individuals are resident in the same context as those receiving the intervention.

Methodology

We evaluated the impact of interventions delivered by non-specialist peers on specific objective outcomes (i.e., depression and post-traumatic stress), as these were the most frequently examined outcome measures in PDS studies.

Search Strategy

A comprehensive literature search was undertaken of the following databases to systematically identify literature relating to the implementation and evaluation of mental health interventions delivered by peers in LMICS: MEDLINE, Web of Science, PsycINFO, Sabinet, and Academic Search Premier. Databases were searched from inception until September 2015 and then repeated in December 2015. Variations of search terms were used in the following combinations: “peer support”; “peer-led”; “peer-delivered”; “peer mentor”; “peer specialist”; “consumer provider”; “lay mental health worker”; “community mental health worker”; and “mental health”; “psychological health”; “psychology”; “psychiatry” or “psychiatric”. Every search within each database used the same search terms to ensure a consistent search strategy was followed. Searches were limited to English-language articles only. Additional search parameters were set to include only studies conducted with samples from low-income, lower-middle-income, and upper-middle-income countries, as defined by the World Bank criteria using the following additional key search terms, “developing country”; “poor country”; “low-income”; “LMIC”; or “LAMIC”.

Eligibility Criteria

For inclusion, studies were required to have implemented an intervention delivered by an individual with either lived experience of the same psychological diagnosis, or resident in the same locale as the treated group of individuals. This criterion, therefore, excluded programs implemented by individuals performing an extension to their usual duties, such as teachers, school officials, graduate university students, or practitioners in training.

All possible variations of “peer” were included in the search strategy and consequently we defined a ‘peer worker’ as ‘any lay individual recruited from the same community within which the intervention program was conducted, who may have been a consumer of mental health services themselves and was recruited following successful engagement in their own recovery process, was explicitly trained in the conduct of the respective intervention but had no formal professional or paraprofessional certification in mental health prior to commencement of the study’. Both individual and group-administered support interventions were included. We also included studies of peer support where interventions may have been multicomponent in nature, but only if the effect of peer support alone was either reported separately or could be derived from the data provided. Where peer support studies were conducted in which the instituted intervention focused primarily on areas other than mental health (e.g., physical health programs such as HIV adherence trials, drug, or alcohol abuse) these were deemed acceptable for inclusion only when mental health outcomes were also reported, in which case, only data for the relevant outcome(s) were extracted.

We expected the number of eligible studies to be limited and, thus, both randomized controlled trials and quasi-experimental studies (pre- and post-evaluation of a single group without a comparison condition or uncontrolled, non-randomized groups) were included. Unpublished dissertations and case studies, however, were excluded. We did not place restrictions on whether studies included child or adult samples.

Procedures

The titles and abstracts of all resulting citations were screened according to the eligibility criteria by two individuals who worked independently. Where the two researchers produced a different decision about the potential inclusion of a particular study, the respective citation was discussed between both researchers and the first author until consensus could be reached. Where consensus could not be reached, the article’s full-text version was retrieved for closer inspection. As an additional search strategy, the bibliographies of all retrieved articles were manually searched to identify potentially eligible studies. Prominent authors known to publish in the field of PDS studies were contacted to determine whether there were potentially eligible studies that were currently in press and had not yet been published.

Data Extraction

The full-text articles of all studies deemed to be eligible for inclusion were assessed for quality and the following information procured: authors, country of implementation, sample sizes, outcome measures assessed, type of comparison condition(s) if any, type of peer worker, length of training received by peer worker, regularity of supervision received by peer worker, contents of treatment programs, treatment formats, length of treatment programs, number of assessment points (post-intervention and/ or follow-up), reported effect sizes (or when these were not reported, data were obtained that would facilitate their computation).

Statistical Analysis

The software program, Comprehensive Meta-Analysis (Borenstein et al. 2009), was used to conduct all statistical analyses. The standardized mean difference (a Cohen’s d effect size) and associated standard error were calculated for each respective study, for each outcome measure where studies reported multiple outcomes, and for each assessment point where studies collected data at both post-intervention and follow-up.

For studies that included comparison conditions, we calculated Cohen’s d for differences in treatment effects between the peer-delivered intervention and the control condition(s). For quasi-experimental studies, a Cohen’s d was also calculated, but in these cases to denote the magnitude of within-group treatment effect between assessment points. When means and standard deviations were not reported, the available statistics (e.g., the sample size and p-value from an appropriate between-groups t- or F-test) were used to calculate effect sizes. Where the statistics reported were not sufficient to conduct an effect size computation, the respective authors were contacted to provide data that would facilitate this. If attempts at contacting authors were unsuccessful, the article was excluded.

Effect sizes were interpreted as follows: effect sizes of 0.2 and less were regarded as small, from 0.3 to 0.7 deemed to be moderate in magnitude, and those from 0.8 and above were considered large (Cohen 1977). Some studies reported the results from more than one instrument for the same outcome variable (e.g., depression may have been assessed using both the Edinburgh Postnatal Depression Scale [EPDS] and the General Health Questionnaire [GHQ]). Where this was the case, we calculated the average standardized mean difference of the two combined inventories. Effect size estimates were combined across studies to obtain a summary statistic for pre- to post-intervention change for each respective outcome measure.

The Q statistic was calculated as a measure of the potential heterogeneity between studies. A significant Q value indicates that the null hypothesis of homogeneity can be rejected and that the included studies were most likely not functionally equivalent (Field and Gillett 2010). The pooled mean effect size for each outcome measure was calculated using the random effects model. Random effects models are deemed to be most suitable when using data from a number of studies where one expects the effect size to vary between studies, and where it is likely that the included studies are not functionally equivalent. Random effects models also allow statistical inferences to be made about studies beyond those included in the present meta-analysis (Field and Gillett 2010).

When conducting a meta-analysis, it is recommended that the possibility of publication bias be investigated. This refers to the possibility that overall effect sizes may be overestimated due to a bias in the published literature that favors the publication of significant findings. A number of analyses - a fail-safe N; Duval and Tweedie’s (2000) trim and fill procedure; Egger’s (Egger et al. 1997) test of the intercept; and Begg and Mazumdar’s (1994) rank correlation test - were derived to determine the presence of likely publication bias. These are also reported separately for each set of analyses (i.e., per outcome measure at each assessment time point).

Results

Study Characteristics

A total of 354 potentially relevant citations were identified using the pre-defined criteria. A further two additional studies were procured following direct contact with a prominent author. The articles were screened for inclusion by first examining their titles, and duplicates and those that were clearly irrelevant were excluded. Then, the abstracts of those that remained were examined and those that did not meet the inclusion criteria were excluded. Where a decision regarding inclusion could not be made on the basis of the information contained in the abstract, the full-text article was retrieved for closer inspection.

After this process, 30 studies remained, all of whose full-text articles were retrieved and examined in detail. We then excluded a total of 16 studies for the following reasons: (a) employed professional mental health providers or paraprofessionals rather than peers or lay individuals (n = 4), (b) studies were not conducted in a LMIC (n = 6), (c) studies delivered a multi-component intervention program using peers in addition to professionals and the input of peers could not be isolated in analyses (n = 2), (d) there was no mental health outcome reported (n = 1), and (e) the work involved was not an intervention study (n = 3). Fourteen (n = 14) eligible studies, which included a total of 18,411 subjects, were included in the final sample of studies. Figure 1 illustrates the procedure followed in the selection and inclusion of studies at each stage of this process.

Fig. 1
figure 1

Flow diagram illustrating search strategy

Full size image

The findings from all of these studies were used for the meta-analysis. South Africa contributed six studies (Cooper et al. 2002, 2009; le Roux et al. 2013; Richter et al. 2014; Rotheram-Borus et al. 2014a, b), while India contributed three (Dias et al. 2008; Tripathy et al. 2010; Vijayakumar and Kumar 2008); there were two from Pakistan (Ali et al. 2010; Rahman et al. 2008), and one each from Jamaica (Baker-Henningham et al. 2005), Uganda (Neuner et al. 2008), and Libya (Stanford et al. 2014). The core characteristics of the included studies are summarized in Table 1.

Table 1 Characteristics of studies included in the meta-analysis
Full size table

Meta-Analysis

Meta-analytic procedures were conducted separately for post-intervention and follow-up data.

Post-Intervention

At post-intervention, twelve studies assessed the outcome of depressive symptomology following the conduct of a peer-delivered intervention. The individual pre- to post-treatment effect sizes ranged from 0.04 to 0.87. The pooled effect size of the 12 studies was 0.24 (95 %; CI 0.09–0.38, p < 0.05), which, according to Cohen’s (1977) criteria, represents an effect of small magnitude. A computation of heterogeneity between studies was significant but moderate in magnitude (Q = 51.82, p < 0.001). The computed effect sizes corresponded to a z value of 5.17 (p < 0.001) and a fail-safe N of 72, indicating that a further 35 studies with null results would be required for the 2-tailed p value to exceed 0.05. Duval and Tweedie’s (2000) trim and fill procedure was employed to determine whether adjustment for potential publication bias was required, and, following a random effects model estimation, it was revealed that no adjustment was necessary (d = 0.23, 95 % CI 0.09–0.38). Furthermore, additional tests of potential publication bias were computed: Egger’s (Egger et al. 1997) test of the intercept was significant (t = 3.01, df = 10, p < 0.05), but Begg and Mazumdar’s (1994) test was not (τ = 0.02, p > 0.05).

Three studies provided data for post-traumatic stress (PTS) at post-intervention (Stanford et al. 2014; Vijayakumar and Kumar 2008; and a comparison between two active experimental conditions in Neuner et al. 2008). The individual pre- to post-treatment effect sizes ranged from 0.15 to 0.97. The pooled effect size for the 3 studies was 0.61 (95 % CI 0.04–1.18, p < 0.05), which sits within the moderate range. Heterogeneity between the sampled studies was significant but small (Q = 12.86, p = 0.002). These effects corresponded to a significant z value of 5.02 (p < 0.001) and a fail-safe N of 17 null studies.

Following additional investigation using Duval and Tweedie’s (2000) trim and fill procedure, no adjustment for potential publication bias was required (d = 0.61, 95 % CI 0.04–1.18). Both Egger’s (Egger et al. 1997) test of the intercept (t = 1.94, df = 1, p > 0.05) as well as Begg and Mazumdar’s (1994) test (τ = 0.02, p > 0.05) were not significant, confirming the likelihood of publication bias to be low. Table 2 displays the pooled effect size results for each of these outcome measures at post-intervention.

Table 2 Peer-delivered interventions in LMICS, pooled effect sizes for pre- to post-intervention change
Full size table

Long-Term Follow-up

Five studies provided data at follow-up for depression. Individual pre- to post-treatment effect sizes ranged from 0.07 to 0.81. The pooled effect size for the 5 studies was 0.5 (95 % CI 0.07–0.92, p < 0.05), an effect of moderate magnitude. Heterogeneity was minimal (Q = 7.96, p = 0.05). Additional analyses produced a fail-safe N of 37 null studies and a significant z value of 5.63 (p < 0.001). Analyses of potential publication bias suggested that this was unlikely: Duval and Tweedie’s (2000) trim and fill procedure produced an effect size of equivalent magnitude to the original computation (d = 0.49, 95 % CI 0.07–0.92), Egger’s (Egger et al. 1997) test of the intercept was not significant (t = 0.62, df = 3, p > 0.05) and so too was Begg and Mazumdar’s (1994) test (τ = 0.01, p > 0.05).

With regard to PTS, a single study, Neuner et al. (2008), provided data for two follow-up comparisons; two separate experimental conditions (peer-delivered narrative exposure therapy [NET] and trauma counseling [TC]) were each compared to a monitoring group control (MCG) condition. Data for these two comparisons were pooled. Individual effect sizes were large, 1.19 for the NET group when compared to MCG and 1.47 for the TC group when compared to the MCG. The pooled effect for the two comparisons was similarly large (d = 1.35). Table 3 displays individual as well as pooled effect sizes for all studies that reported follow-up data.

Table 3 Peer-delivered interventions in LMICS, effect sizes for pre- to follow-up change
Full size table

Discussion

Mental illness continues to contribute, significantly, to the global burden of disability. Its detrimental effects appear especially pronounced for those resident in poor, resource-deprived contexts where a significant proportion of the people who would benefit from the receipt of mental health services do not receive this care. In SA, for example, it is estimated that approximately 75 % of individuals with a serious mental illness who require treatment, do not receive such treatment (Seedat et al. 2009). This may be the result of a scarcity of skilled mental health practitioners in these locales as well as the prioritization of financial resources to avenues that are likely deemed more pressing, such as management of the human immunodeficiency virus epidemic and other infectious diseases (e.g., tuberculosis) that are especially prevalent in LMICS. Recently, a number of authors have fervently advocated for the adoption of task-shifting of mental health interventions to peers and non-specialists as a key strategy for closing the treatment gap and meeting the immense need for mental health interventions in poor communities (Buttorff et al. 2012; Chatterjee et al. 2009; Mendenhall et al. 2014; Patel et al. 2007; Petersen et al. 2012; Saraceno et al. 2007; Saxena et al. 2007). This review summarized, following a systematic review and meta-analysis, the preliminary evidence for the effectiveness of utilizing peers in the treatment and management of mental health conditions in poor, developing world countries.

The two mental health conditions most frequently examined in relation to peer-delivered mental health services are depression and PTS. The pooled overall effect size of accumulated studies that assessed treatment outcome directly following receipt of an intervention targeted at depression was small (d = 0.24), indicating that peer-delivered programs are associated with a minimal reduction in depressive symptomology in LMIC samples. For studies that again assessed outcome of depressive symptomology at some follow-up point, the pooled overall effect size was of moderate magnitude (d = 0.5). For PTS, the overall effect size was of moderate magnitude (d = 0.61) at post-intervention, having pooled the results of a trio of relatively homogenous studies, while at follow-up, the pooled effects were substantial (d = 1.35). However, this latter computation was based on the combined results of only two comparison conditions, within the context of the same study, and is, therefore, limited in its utility. While moderate heterogeneity was present across all analyses, all investigations of potential publication bias suggested this was unlikely. It can, therefore, be assumed that all studies included in each set of analyses were functionally equivalent.

This study’s results concur with previous investigations of the effectiveness of PDS in the developed world. While some studies from HICS have found the implementation of PDS to be no more beneficial than treatment-as-usual delivered by experts (e.g., Chinman et al. 2015; Fuhr et al. 2014; Lloyd-Evans et al. 2014), where programs implemented by peers have produced statistically beneficial results, both in comparison to specialist-delivered treatment (SDT) and as an adjunct to SDT when compared to usual care, the magnitude of benefit, in concurrence with the present study, has tended to be small (e.g., Chinman et al. 2014; Druss et al. 2010; Kim and Free 2008).

Despite the present analyses having been based on a relatively small sample size, this study provides important preliminary evidence in support of the development and implementation of PDS in resource-poor and financially deprived contexts. However, further investigation is required. Some studies in HICS have included comprehensive assessments of additional outcome measures beyond the reduction of clinical symptoms and found these to be beneficially impacted by peer support; for example, quality of life (Fuhr et al. 2014); number of admission events (Davidson et al. 2012); number of days during an admission (Chien et al. 2006); quality of interpersonal relationships, self-assessments of recovery (Chinman et al. 2015); hope and empowerment (Lloyd-Evans et al. 2014); and self-care and sense of well-being (Sledge et al. 2011). These constructs are precisely those that are reflective of the ethos embodied in the concept of recovery (Parker 2012). Future studies of peer-delivered support in LMICS should follow suit and include measurement of these recovery-oriented domains.

Meta-analytic studies of intervention programs facilitated by “non-specialist health workers” (NSHWS), defined as individuals who are already employed in some health-related role, either professional (doctors, nurses, or social workers) or lay, but who have not had any specialist mental health training, provide an additional point of comparison for the present study. In exploring the effectiveness of task-shifting to these types of individuals working in LMICS, van Ginneken et al. (2013) reported that moderate benefits are likely in relation to depression and anxiety (d = 0.66), perinatal depression (d = 0.42), and PTS (d = 0.36). These computed effects are somewhat higher than those of the present study. This difference may be reflective of a core set of skills possessed by those already employed in helping professions that may favorably benefit the delivery of interventions. However, while task-shifting to NSHWS may be a prudent course of action in some middle-income countries, in more deprived low-income countries where the numbers of individuals employed in most health-related roles are limited, the use of lay individuals from the community may be a more necessary route.

We believe that an integral strength of many of the studies included in this meta-analysis, lies in the meticulous selection and comprehensive training of suitable individuals to be employed as peers, the supervision plan, as well as the manner in which the intervention was delivered. For example, in all studies, peers were required to be contextually knowledgeable; speak the local language; have passed basic secondary schooling; exhibit both motivation to fulfill the role and an understanding of the concept of recovery; while some studies paid additional attention to selecting individuals with particular personality traits and capacities (e.g., empathy, social skills, emotional vocabulary). While theoretical training that was specific to the nature of the particular intervention was always provided (e.g., child development or parenting skills), this was often supplemented with multiple didactic methods (e.g., lectures, case studies, and role-plays) along with regular supervision by an expert.

Prior to enacting large scale roll-out of peer-led interventions, two factors require further consideration. First, decisions about large-scale adoption of any intervention program should be informed by sound, contextually-specific empirical evidence in support of its adoption. The present study contributes to this evidence base, but further study is required. Second, the success of any task-shifting initiative is greatly dependent on ensuring that those who deliver the respective intervention are both accepted by the community they serve and suitably skilled to deliver a program of high quality. As such, the selection and training of peer practitioners is paramount. Clients are most accepting of practitioners who are local (from the same community), middle-aged, with some education, and who share similar experiences. Individuals who match this profile are regarded as suitable role-models and appear most successful at fulfilling the role of a peer practitioner (Singla et al. 2014a). Moreover, there is some evidence to suggest that peer-led supervision is feasible, acceptable, and as reliable as expert-delivered supervision (Singla et al. 2014b). This is encouraging, particularly when considering the scalability of PDS; however, clearly defined and tested protocols are needed to inform the training and monitoring of peer supervisors.

On the basis of our preliminary investigation, it can be concluded that mental health interventions delivered by peers, can be potentially effective mechanisms for managing common mental health conditions in deprived LMIC contexts, with these locations being ones that suffer from an invariably large treatment gap and a commensurate dearth of suitably qualified mental health practitioners. However, while PDS may represent an important evolution in the provision of mental health services, particularly in deprived contexts, a number of factors require consideration prior to scaling up and implementation. These relate specifically to the selection, training, and supervision of peers engaged in mental health work to ensure that interventions delivered by peers retain the necessary quality and fidelity for success. As procedures and protocols related to these issues are refined, the work that follows should interrogate its comparative efficacy for different types of services (e.g., tertiary healthcare versus primary healthcare centers versus community-based work or home visiting type programs) and for varying mental health conditions, beyond depression and PTS.