Mental health interventions by lay counsellors: a systematic review and meta-analysis

Abstract Objective To investigate the effectiveness of community-based mental health interventions by professionally trained, lay counsellors in low- and middle-income countries. Methods We searched PubMed®, Cochrane Central Register of Controlled Trials, PROSPERO and EBSCO databases and professional section publications of the United States National Center for PTSD for randomized controlled trials of mental health interventions by professionally trained, lay counsellors in low- and middle-income countries published between 2000 and 2019. Studies of interventions by professional mental health workers, medical professionals or community health workers were excluded because there are shortages of these personnel in the study countries. Additional data were obtained from study authors. The primary outcomes were measures of post-traumatic stress disorder, depression, anxiety and alcohol use. To estimate effect size, we used a random-effects meta-analysis model. Findings We identified 1072 studies, of which 19 (involving 20 trials and 5612 participants in total) met the inclusion criteria. Hedges' g for the aggregate effect size of the interventions by professionally trained, lay counsellors compared with mostly either no intervention or usual care was −0.616 (95% confidence interval: −0.866 to −0.366). This result indicates a significant, medium-sized effect. There was no evidence of publication bias or any other form of bias across the studies and there were no extreme outliers among the study results. Conclusion The use of professionally trained, lay counsellors to provide mental health interventions in low- and middle-income countries was associated with significant improvements in mental health symptoms across a range of settings.


Introduction
A recent reappraisal of the global burden of mental illness using a broad definition of mental illness as a disease concluded that it accounted for a greater percentage of the global burden of disease, in terms of years lost to disability, than any other disease category. 1 Moreover, according to the World Health Organization (WHO), in 2011 between 76% and 85% of people with mental illnesses in low-and middle-income countries went untreated. 2 This gap was partly due to a shortage of mental health professionals and to resources being concentrated in large, centrally located institutions rather than in community settings. 3 In many places in the world, there may be only one psychiatrist for every 500 000 people and most professional mental health resources are taken up by patients with severe mental illnesses. 4 This problem has been exacerbated by the coronavirus disease 2019 pandemic, which has further challenged people's psychological and physical resilience. 5,6 The use of community health workers has been examined as a partial solution to the shortage of mental health workers. However, in many settings, there are few community health workers, they are overburdened and little research has been performed into their cost-effectiveness in providing mental health interventions. 7 In 2016, WHO predicted a shortage of 18 million health workers (including medical doctors, nurses and community health workers) in low-and middle-income countries by 2030. 8 One proposed solution is to use professionally trained, lay counsellors to provide mental health interventions. In 2003, WHO's Department of Mental Health and Substance Dependence held a meeting to address the gap in mental health provision in low-and middle-income countries after large-scale disasters and conflicts. 9 Meeting participants highlighted the importance of maximizing community resources. Subsequently, guidelines for emergency relief efforts created by the Inter-Agency Standing Committee proposed the use of tiered care and the committee recommended that mental health interventions could be delivered by trained, nonprofessional, community members. 10 People who need additional help could be referred to mental health professionals.
The Grand Challenges in Global Mental Health initiative, launched by the United States National Institutes of Health and several global organizations, met in 2010 and created a list of 40 grand challenges in response to the shortage of mental health services in low-and middle-income countries. 11 Recommendations included the development of sustainable models of training for, and increasing the number of, ethnically diverse lay and specialist mental health service providers. In addition, the Lancet Commission on Global Mental Health and Sustainable Development noted in 2018 that there had recently been a shift from reliance on a single group of experts for providing mental health services towards the use of nonspecialist providers such as teachers, community health workers, law enforcement officers and people with lived experience. 12 Yet, there have been few studies of mental health interventions facilitated by professionally trained, lay community members, particularly in situations where professional resources are scarce. Nevertheless, there is evidence suggesting that lay community workers can effectively provide mental health interventions in low-resource communities. 13 For this review, we investigated the potential for expanding health resources rather than taxing already overburdened health workers. As we could find no universally accepted definition of the term, we defined a lay counsellor as a professionally trained member of the community who had no specific mental health training before being trained in the use of one or more mental health interventions.
The aim of our literature review was to investigate the effectiveness of community-based mental health interventions facilitated by professionally trained, lay counsellors in low-and middle-income countries at the level of the community. Involving the community as agent of change is the least-used category of community-based interventions in public health, and this strategy can promote healthier communities and strengthen the community's capacity to address health issues. 14 Although previous reviews of treatment for mental disorders in lowand middle-income countries have sometimes included the use of lay counsellors, 15-21 our review examines exclusively the effectiveness of professionally trained, lay counsellors from the community in treating common mental disorders in low-and middle-income countries.

Methods
This review is registered with PROSPE-RO (CRD42019118999). Our literature review included studies that involved the provision of mental health interventions by lay counsellors living in the local community in low-and middle-income countries. In eligible studies, counsellors were drawn from a broad cross section of the community and were exclusively individuals who were not employed as medical or mental health professionals, teachers or community health workers and who did not work for nongovernmental organizations or government institutions.
Study inclusion criteria were selected by two authors and a literature search found that no prior review used the same criteria. Eligible studies had evaluated the use of professionally trained, lay counsellors to facilitate mental health interventions in low-and middle-income countries in a randomized controlled trial published between 2000 and 2019. We also included preventive studies or studies not in English. We did exclude studies that used closely related interventions, such as professionally led self-help groups, media-distributed interventions and interventions involving peer support (i.e. involving individuals who derived their knowledge from personal experience rather than formal training). 22 We also excluded studies of interventions by professional mental health workers, medical professionals or community health workers.
We searched the PubMed®, Cochrane Central Register of Controlled Trials, PROSPERO and EBSCO (EBSCO Information Services, Ipswich, United States of America, USA) databases using the search terms lay counsellors, mental health interventions in low and middle income countries, mental health interventions after disasters, lay counsellors mental health, community member facilitated mental health, lay counsellor mental health Africa randomised controlled trials, community mental health interventions in LMICs, community-member facilitated mental health randomised controlled trials, cognitive behavioral therapy based intervention by community health workers and taskshifting. In addition, we searched the professional section publications of the United States National Center for PTSD (United States Department of Veterans Affairs, Washington DC, USA) using the terms posttraumatic stress disorder research and PTSD research. We also examined reference sections of the relevant studies identified.
The titles and abstracts of studies identified by the search were examined by two authors according to preferred reporting items for systematic reviews and meta-analyses guidelines using a flow diagram for data extraction designed for this review: 23 Details are available from the data repository. 24 Disagreements about which studies to include were discussed until agreement was reached, sometimes in consultation with a third author. One author extracted data from the selected studies and data accuracy was cross-checked by another. Study authors were contacted, where relevant, to obtain data that were not included in the study but were necessary for the meta-analysis or for assessing the risk of bias. Where several papers reported the same data, the data were included only once. An assessment of inter-rater reliability of coding found that 98.7% of data entries were in agreement, with only 2 of 148 data points having to be changed (data repository). 24

Data analysis
We performed the meta-analysis using Comprehensive Meta-Analysis v. 2.2.050 (Biostat, Englewood, USA). Sample sizes and the means and standard deviations of outcome measures needed for the meta-analysis were obtained from the study publications. Effect sizes were derived from differences in the means of outcome measures at the first assessment following the intervention between individuals who received mental health interventions from lay counsellors and those in control groups. Both these effect sizes and an aggregate effect size are expressed using Hedges' g coefficient. For this meta-analysis we used direct measures with negative numbers reflecting a reduction in symptoms after treatment. Effect sizes were calculated using the pooled standard deviation. 25 , categorized them as previously suggested: 26 (i) Hedges' g between 0.0 and ±0.2 indicates no effect; (ii) from > ±0.2 to < ±0.5 indicates a small effect; (iii) from > ±0.5 to < 0.8 indicates a medium effect; and (iv) ≥ ±0.8 indicates a large effect.
As the variation in outcome measures was likely to exceed the sampling error, we decided to use a randomeffects meta-analysis model. With a random-effects model, there are likely to be fewer issues with type-I statistical errors (i.e. false-positive findings) and confidence intervals are generally more precise. Further, the calculated confidence intervals are less likely to overstate the degree of precision of the meta-analysis. 25 We examined the homogeneity of the variance in, and the distribution of, effect sizes between the studies using a normal Q-Q plot (SPSS, IBM, Chicago, USA). We performed an outlier analysis to better understand the Systematic reviews Lay mental health counsellors Suzanne M Connolly et al.
contribution of individual effect sizes to the aggregate effect size and to assess their overall impact. We used standardized Cochrane procedures to assess the potential risk of bias over seven domains: (i) random sequence generation; (ii) concealment of group allocation; (iii) quality of blinding of participants and personnel; (iv) blinding of outcome assessments; (v) reporting of incomplete outcome data; (vi) outcome reporting; and (vii) other sources of bias. 27 Each domain was judged as having a high, unclear or low potential risk of bias. Six authors independently assessed the risk of bias in six or seven studies each and reached a consensus with another evaluator. Disagreements were resolved by involving another author. Authors did not participate in the assessment or approval of a study if they were involved in the study or personally knew one of the study's authors.

Results
We identified 1072 studies and, in addition, we examined the reference sections of three systematic literature reviews and meta-analyses and one other study ( Fig. 1). 18,20,21,28 After removing 110 duplicate records, the remaining 962 studies were screened by reading the abstracts or quickly reviewing the data. Subsequently, we assessed the full texts of 247 studies to determine whether they met the inclusion criteria and 217 were excluded. The characteristics of the studies excluded on the basis of abstract or full text reviews are available from the data repository. 24 Finally, 19 studies met the inclusion criteria. [29][30][31][32][33][34][35][36][37][38][39][40][41][42][43][44][45][46][47] In addition, we identified 11 supporting papers that provided additional data on studies reported in these 19 papers. [48][49][50][51][52][53][54][55][56][57][58] Where more than one paper was associated with a particular study, we included the paper that contained the most information pertinent to the inclusion criteria or analyses. As one study reported two trials, the meta-analysis involved data from a total of 20 trials. One author was involved in two of the studies included and two authors personally knew an author of the study by Robson et al.; 37 consequently, they did not participate in the assessment or approval of the studies concerned. Researchers from six studies were contacted to provide data needed for the meta-analysis that had not been included in the published articles. 29,[38][39][40]43,44

Meta-analysis
Of the 20 trials, 14 found that the intervention by professionally trained, lay counsellors had a significant effect: in five it was a large effect, in six a medium effect and in three a small effect ( Table 2 and Fig. 2). Only six trials found no significant effect. Overall, the interventions had a medium effect (i.e. Hedges' g: −0.616; 95% confidence interval: −0.866 to −0.366) as calculated using the random-effects model (Fig. 2). Although the aggregate effect size calculated using the fixed-effects model was smaller (Hedges' g: −0.407; 95% confidence interval: −0.460 to −0.353), it was still significant. Nevertheless, given the initial rationale, use of the randomeffects model was warranted. Moreover, when the distribution of the effect sizes observed in the 20 trials was examined using the Q test to identify outliers, it was found that some differences were   35 (i) Nine refugees (five women and four men; mean age: 27 years) from the community were trained as counsellors. Skills required to be accepted for the training included literacy in English and literacy in their mother tongue, as well as the ability to empathize with their clients and a strong motivation to carry out this work; and (ii) their educational level varied from primary school to university education 6 weeks of education in counselling for alcohol problems, a psychoeducational and social skills intervention, and in general counselling skills Tol et al. (2008) 40 (i) An unspecified number of lay counsellors, or interventionists, who had to be older than 17 years and have at least a high school education, were selected from local target communities based on a selection procedure assessing social skills through role plays; (ii) they were generally people with no formal mental health training but had some experience as volunteers in humanitarian programmes; and (iii) a study author stated in email correspondence that "the interventionists were newly hired community members, not currently employed as community health workers, etc. "  31 Fourteen adults (seven women and seven men) who were community-based lay therapists without a mental health or medical background Intensively trained local lay counsellors underwent narrative exposure therapy training for an unspecified length of time Tol et al. (2012) 39 (i) An unspecified number of lay counsellors, or interventionists, who had to be older than 17 years and have at least a high school education, were selected from local target communities based on a selection procedure assessing social skills through role plays; (ii) they were generally people with no formal mental health training but had some experience as volunteers in humanitarian programmes; and (iii) a study author stated in email correspondence that the interventionists "were newly hired community members, not currently employed as community health workers, etc. " Counsellors were trained in a cognitive behavioural therapybased intervention and supervised in implementing the intervention for 1 year before the study unlikely to be due to sampling variation (i.e. there was heterogeneity). This finding provides further support for using the random-effects model, which assumes that the trials are interchangeable and that not all trials drew participants from the same population. We found that the data in Murray et al. 's study were not normally distributed (P < 0.001). 34 In addition, data values from Robson et al. 's study appeared to be extreme. 37 Consequently, we performed an additional outlier analysis, which showed that the values from Murray et al. 's study were within acceptable boundaries. Moreover, this study examined post-traumatic stress disorder, which was consistent with most other studies in the meta-analysis, and it involved a substantial number of participants: 131 in the treatment group and 136 in the control group. An additional normality test performed for Robson et al.'s study failed to show a significant result. Consequently, following commonly accepted practice, 59 we decided not to remove any studies in conducting our meta-analyses, as there were no extreme outliers.
Details of our assessment of the risk of bias in the 19 studies across all seven bias domains are available from the data repository. 24 In summary, we found that: (i) the risk of bias due to incomplete outcome data reporting was low in 95% (18/19) of studies; (ii) the risk of bias due to random sequence generation was low in 79% (15/19); (iii) the risk of selective reporting bias was low in 68% (13/19); (iv) the risk of other sources of bias was low in 63% (12/19); (v) the risk of bias due to outcome assessment was low in 58% (11/19); (vi) the risk of allocation concealment selection bias was low in 53% (10/19); and (vii) the risk of performance bias associated with blinding participants and personnel was low in 42% (8/19).
We also assessed the 19 studies for publication bias by creating a funnel plot of Hedges' g against the standard error (data repository). 24 There was no clear pattern and there was no general shift in effect sizes to either side of the estimated aggregate mean. Although the individual study effect sizes did not appear to be randomly distributed, we determined using Duval and Tweedie's trim-and-fill method that no study needed to be excluded, 60 despite appearing to be an outlier. In addition, this analysis did not suggest that any individual value needed to be adjusted.

Discussion
Our meta-analysis of 20 randomized controlled trials suggests that professionally trained, lay counsellors can

Study Lay counsellor selection Lay counsellor training
Tol et al. (2014) 38 (i) Lay counsellors comprised an unspecified number of locally identified nonspecialized facilitators trained and supervised in implementing the intervention for 1 year before the study. Facilitators had at least a high school diploma and were selected for their affinity and capacity to work with children as demonstrated in role plays and interviews; and (ii) a study author stated in email correspondence that the lay counsellors "were newly hired community members, not currently employed as community health workers, etc. " Counsellors were trained in a cognitive behavioural therapybased intervention and supervised in implementing the intervention for 1 year before the study Murray et al. 34 Twenty-three adult counsellors (11 from study sites and 12 external); their backgrounds varied but all counsellors had at least a high school education and basic communication skills (i) Counsellors were adults with no prior professional training or qualification in the field of mental health, they had completed at least secondary school education, and were fluent in the vernacular language used in the study setting; (ii) 11 counsellors participated in the trial; and (iii) a study author stated in email correspondence that "The lay counsellors were not employed when we recruited them for our programme"

Systematic reviews
Lay mental health counsellors Suzanne M Connolly et al.
provide effective mental health interventions in low-and middle-income countries. In particular, 14 of these trials found that outcomes improved significantly more in the intervention than the control group. Moreover, Hedges' g for the aggregate effect size of the lay counsellor interventions on symptoms indicated a medium effect size, when calculated using a random-effects metaanalysis model. Previous reviews have often focused on or included programmes that involve training professional or paraprofessional health workers (who are already overburdened) to deliver mental health services. 7 We found no previous metaanalyses of therapy given exclusively by lay counsellors with which to compare our results.
Our findings may be important for health authorities, policy-makers and other stakeholders planning psychiatric health care. In particular, the use of lay counsellors could provide valuable, first-tier, mental health services for people in underserved communities. The heterogeneity of the interventions used in the studies we identified is both a strength and a weakness. On the one hand, the diversity in the type and length of treatment and training and in the supervision and support provided for lay counsellors makes it difficult to draw firm conclusions. On the other, this diversity is a strength as it demonstrates the reality of the treatment provided by lay therapists in low-and middle-income countries.
The use of lay counsellors for mental health interventions is only a partial solution to the gap in mental health provision in low-and middleincome countries. The treatment gap in these countries actually reflects deeper systemic global problems, such as the unequal distribution of resources and vast disparities in income. 13 The use of lay counsellors could also be problematic. A qualitative investigation in South Africa found that, in several studies, the lack of formal supervision, standardized training and a clear definition of the lay counsellor's role led to poor treatment fidelity. 61 Other problems were inconsistent remuneration and health-care managers who did not appreciate the importance of counselling. 61 Further research is needed on mental health interventions facilitated by lay counsellors in places where mental health needs outstrip professional resources. Although the interventions examined in this systematic review and meta-analysis show promise for reducing the mental health burden globally, they will need to be tested using more stringent methods.

Study
The data we obtained from randomized controlled trials have several limitations. Some authors did not adequately report outcome data or provide a satisfactory description of the lay counsellors or their training. In addition, some studies involved few participants and appeared to lack adequate statistical power, whereas others did not adequately describe blinding or masking procedures. Also, the cost-effectiveness of using lay counsellors to provide different interventions will need to be evaluated. Finally, there are several mental health interventions that can be administered by professionally trained community members that have not yet been examined. They will need to be studied if the goal of closing the gap in mental health provision globally by harnessing community resources as agents of change is to be pursued seriously.
Although randomized controlled trials of mental health interventions by community members that require minimal professional therapist involvement are scarce, we identified 20 such trials. Together, these trials demonstrate that professionally trained, lay counsellors have a promising role to play in helping close the mental health treatment gap in low-and middle-income countries. ■

Systematic reviews
Lay mental health counsellors Suzanne M Connolly et al.

Resumen
Intervenciones de salud mental realizadas por consejeros no profesionales: una revisión sistemática y un metanálisis Objetivo Investigar la eficacia de las intervenciones de salud mental basadas en la comunidad y llevadas a cabo por asesores no especializados con formación profesional en países con ingresos bajos y medios. Métodos Se realizaron búsquedas en las bases de datos PubMed®, el Registro Cochrane Central de Ensayos Controlados (Cochrane Central Register of Controlled Trials), PROSPERO y EBSCO y en las publicaciones de la sección profesional del National Center para TEPT de los Estados Unidos para obtener ensayos controlados aleatorios de intervenciones de salud mental realizadas por consejeros legos con formación profesional en países con ingresos bajos y medios, publicados entre Lay mental health counsellors Suzanne M Connolly et al.