Applying systems thinking to task shifting for mental health using lay providers: a review of the evidence

Objective. This paper seeks to review the available evidence to determine whether a systems approach is employed in the implementation and evaluation of task shifting for mental health using lay providers in low- and middle-income countries, and to highlight system-wide effects of task-shifting strategies in order to better inform efforts to strength community mental health systems. Methods. Pubmed, CINAHL, and Cochrane Library databases were searched. Articles were screened by two independent reviewers with a third reviewer resolving discrepancies. Two stages of screens were done to ensure sensitivity. Studies were analysed using the World Health Organization's building blocks framework with the addition of a community building block, and systems thinking characteristics to determine the extent to which system-wide effects had been considered. Results. Thirty studies were included. Almost all studies displayed positive findings on mental health using task shifting. One study showed no effect. No studies explicitly employed systems thinking tools, but some demonstrated systems thinking characteristics, such as exploring various stakeholder perspectives, capturing unintended consequences, and looking across sectors for system-wide impact. Twenty-five of the 30 studies captured elements other than the most directly relevant building blocks of service delivery and health workforce. Conclusions. There is a lack of systematic approaches to exploring complexity in the evaluation of task-shifting interventions. Systems thinking tools should support evidence-informed decision making for a more complete understanding of community-based systems strengthening interventions for mental health.


Introduction
Globally, mental health accounts for a large and growing burden of disease (Whiteford et al. 2013). Recent estimates from the WHO Mental Health Surveys indicate an interquartile range of lifetime DSM-IV disorder prevalence between 18.1% and 36.1% (Kessler et al. 2009). According to the Global Burden of Disease Study, between 2005 and 2013, disability-adjusted lifeyears attributed to mental and neurological disorders increased by 9.7% and 16.1%, respectively (Murray et al. 2015a). Despite this burden, a study across 17 countries demonstrated that only 20% of persons with common mental disorders (CMDs) received treatment in the year prior to the survey, with only 10% receiving minimally adequate treatment (Wang et al. 2007). Availability and scale-up of essential health services to achieve health system goals is often impeded by health workforce shortages (WHO, 2006). This is especially true of mental health in low-and middleincome countries (LMICs), where availability of services is not matched to population needs (Weinmann & Koesters, 2016). The World Health Organization (WHO) estimates that there is a need for 1.18 million mental health workers to move towards closing the mental health treatment gap (Fulton et al. 2011).
In LMICs, not only are there limited mental health services available, utilization of existing services is also poor for a multitude of reasons, including geographic, cultural, and financial access barriers (Murthy, 2011;van Ginneken et al. 2013;Chowdhary et al. 2014;Joshi et al. 2014;Chibanda et al. 2015;Weinmann & Koesters, 2016). Mental health service delivery is highly context-specific with culturally defined interpretations of stigma, trust, and utility affecting success and impact (Murthy, 2011;van Ginneken et al. 2013;Weinmann & Koesters, 2016). Integration of mental health services into primary care presents a strategic opportunity to overcome some of these access barriers and reach the largest number of people while minimizing stigma and discrimination (WONCA, 2008). Integration of mental health services is also in line with the essential public health function of early diagnosis and prevention; it requires primary care providers to be trained in identifying poor mental health and taking action towards treatment (WONCA, 2008). Mental and physical health are interconnected, and early detection can lead to improved health outcomes and increase costeffectiveness for the health system (WONCA, 2008;Levin & Chisholm, 2015). However, effective integration requires strengthened primary care systems. The value of lay provider programmes in strengthening these systems towards universal health coverage, which includes provisions for mental health, has been recognized globally (Schneider & Lehmann, 2016). For the purposes of this paper, lay providers are defined broadly as individuals who may or may not have basic literacy skills or some form of formal post-secondary education with subsequent informal or formal pre-service training (Olaniran et al. 2017). They are often unpaid or may receive an allowance based on the programme (Olaniran et al. 2017).

Recognizing task shifting as a system intervention for mental health
In resource-poor settings, task shifting has been an effective approach to addressing health workforce challenges and strengthening systems for mental health (Eaton et al. 2011;Kakuma et al. 2011). Several systematic reviews have supported the use of task shifting for mental health focused on specific populations, such as with people living with HIV/AIDS or mothers with postpartum depression (Rahman et al. 2013;van Ginneken et al. 2013;Chowdhary et al. 2014;Atif et al. 2015;Chibanda et al. 2015). Task shifting includes shifting service delivery of specific tasks from professionals with higher qualifications to those with fewer qualifications or creating a new cadre with specific training (WHO, 2007b). It is meant to alleviate the heavy workload of specialists and to ensure that those with no access to specialists have a means of accessing some level of mental health services (Patel et al. 2007). By shifting service delivery for less complex cases to lay providers, the system allows mental health specialists to focus on more complex cases with the hope that quality of care delivery will also improve (Weinmann & Koesters, 2016).
Task shifting requires various parts of the system to be working in harmony in order to be successful (GHWA, 2007;Yaya Bocoum et al. 2013). Conditions such as regular supervision, availability of resources and tools, access to medicines, quality training, and exposure to technological updates through in-service training are enabling factors in ensuring successful redistribution of tasks among health workforce teams (Yaya Bocoum et al. 2013;Agyapong et al. 2016). Buy-in and acceptance of task-shifting interventions across a wide range of stakeholders is also important in their success (Yaya Bocoum et al. 2013). For example, perceptions of a loss of hierarchal structures, shift in earnings, and burden of supervision are examples of barriers that higher professional cadres may experience regarding task shifting (Niekerk, 2008;Philips et al. 2008;Zachariah et al. 2009). Therefore, task shifting is a system-wide intervention that has implications beyond the players and programmes directly involved in its implementation; it reallocates resources across the health system to trigger change.
Using systems thinking to evaluate the impact of task shifting With the recognition that task shifting for mental health is a system-wide intervention, understanding its potentially far-reaching implications and impact across the system becomes valuable for appropriate decision making, health system planning, and implementation of interventions. System-wide effects can be captured using the suite of tools available in systems thinking to collect information across a multitude of stakeholders and mechanisms involved in a given context (AHPSR, 2009;Peters, 2014). The six building blocks of the health systemservice delivery, health workforce, information technology, medical products, financing, and leadershipare made dynamic, adaptive, and interactive through a systems thinking lens as it is designed to explore how different elements are connected in a system and the impact and implications of these connections (Table 1) (Adam et al. 2012). Systems thinking also incorporates another key health systems element: communities and people (Adam et al. 2012). Therefore, in our application of the building blocks approach, we have added a seventh to account for communities and people. By enhancing understanding of different perspectives across multiple interacting agents, the changing context in which they interact, and the changes resulting from new patterns created over time, systems thinking can serve as an important policy toolkit (Adam et al. 2012;Peters, 2014).
Applying evidence for success in capacity building for mental health care The growing burden of disease attributed to mental health calls for approaches that strengthen the capacity of the health system to equitably and appropriately address the wide range of mental and neurological disorders (Whiteford et al. 2013). The mhGAP (Global Mental Health Gap Action Programme) was launched in 2008 to provide technical guidance, tools and training to help address the challenges of availability in resource-poor settings (WHO, 2008). Global mental health has seen attention in academic circles through special series in The Lancet and PLoS, which highlight integration of mental health into primary care as a key strategy (Patel et al. 2007;Patel & Thornicroft, 2009).
Integration of mental health into health systems, especially in primary care systems, is not without its challenges, particularly in resource-poor settings (Patel et al. 2010a;Eaton et al. 2011;Weinmann & Koesters, 2016). Poor policy implementation, inadequate human resource allocation to support the process, poor community engagement, and low access to medicines are among the challenges of integration (Patel et al. 2007;Eaton et al. 2011;van Ginneken et al. 2013). Systems thinking contributes to documenting the system-wide impact of a given intervention, as well as enhancing the ability to predict both intended and unintended consequences of the intervention, critical in designing successful large-scale reform.
Few studies focus on the wider impact of task shifting across the health system and, likewise, the scale-up of mental health strategies (Eaton et al. 2011;Yaya Bocoum et al. 2013). This weakness in the literature undermines the complexity of interactions and changes that take place in health systems, and stymies the potential scale-up and sustainable integration of promising task-shifting strategies (Adam et al. 2012;Yaya Bocoum et al. 2013). To ensure that LMICs can expand large-scale mental health strategies and achieve integration into primary care, a system-wide approach can be an effective tool in understanding, evaluating, and implementing bespoke strategies (WHO, 2007b;AHPSR, 2009). This paper reviews the available evidence to determine whether a systems approach is employed in the implementation and evaluation of task shifting for mental health using lay providers in LMICs. It seeks to highlight system-wide effects of task-shifting strategies in order to better inform efforts to strengthen community mental health systems.

Search strategy
The electronic databases of PubMed, CINAHL, and Cochrane were searched between 5 September 2016 and 30 October 2016 (Annexure 1). The search strategy consisted of three concepts: (1) lay providers, including community health workers, health aides, local references to community health workers such as accredited social health activists, non-physician health workers, community-based practitioners, and other associated terms; (2) mental health, including the standard set of disorders under the definition of CMDs such as anxiety, depression, dementia, schizophrenia, and substance abuse, as well as strategies for treatment such as supportive counselling, cognitive behavioural therapy, and others; and (3) LMIC setting, as this study is focused on alternatives for delivery of mental health services in resource-poor settings. These concepts were expanded to include similar terms and combined using 'and' to build the search. Further, the references of included articles were searched to identify additional citations that were not captured in the search as a means of ensuring the robustness of the study. These were included when the full text satisfied the inclusion criteria of being set in an LMIC, focusing on mental health and evaluating a task-shifting strategy of service delivery from providers with higher or more specialized qualifications to those with lower qualification. However, all eligible references were already captured in the search. Search was limited to publications between January 1996 and September 2016.
After completing the electronic search, the titles and abstracts of all identified articles were independently reviewed by two authors, who assessed whether the article should be included or excluded according to pre-defined criteria. These criteria are included in Table 2. Articles that met any of the criteria for exclusion were eliminated. In the first round of screening, articles meeting at least four of five criteria for inclusion based on titles and abstract review, were included. In the case of inter-rater disagreement, a third reviewer was consulted on the inclusion or exclusion of the article in question. The third reviewer was blinded and has expertise in health systems research. Articles intended for inclusion were combined in a Microsoft Excel spreadsheet and any duplicates were removed.
Full-text versions of identified articles were examined in order to reassess inclusion based on articles meeting all five criteria before establishing the final set to be included in the study. A two-stage approach to inclusion was employed to ensure sensitivity. For inclusion in this review, the study must have: (1) evaluated the implementation and/or impact of an intervention; (2) had significant focus on mental health; (3) been set in an LMIC; (4) employed task-shifting strategies where service delivery was transformed from a professional cadre with higher qualification to lay providers with lower qualifications and minimal mental health training; and (5) involved training of lay providers was limited to fewer than 3 years. The training criteria was articulated with input from a health workforce specialist in order to keep the focus on task shifting to providers with fewer qualifications without excluding task shifting to qualified providers who lack specialized mental health training as we considered this relevant to our study. Where length of training was not specified, we used our collective judgment to determine whether task shifting was towards a provider with minimal mental health training. An expanded interpretation of evaluation was used to include both quantitative and qualitative studies that reported on randomized control trials, cohort studies with before and after measures, survey and/ or observational assessments of stakeholder perceptions, acceptability and satisfaction, case studies, and analysis of qualitative data.

Data extraction and analysis
Two study authors read all included full texts and extracted the following data: setting, year of publication, aim of study, type of intervention, sample size, outcomes measured, results, health system implication(s), and barriers and facilitators of implementation. Critical Appraisal Skills Programme (CASP) tools were used to assess the quality of the studies (CASP, 2016). The initial screening questions (see Table 3) were used to ensure that included studies met minimal quality standards. Risk of bias and limitations of included studies were then assessed using more detailed items found on CASP checklists for different types of studies (see Table 4). (1) The research article evaluated an intervention/implementation strategy (2) Mental health was a significant component of the intervention/implementation strategy (3) The intervention/implementation strategy was introduced in a LMIC (4) The intervention/implementation strategy involved task-shifting to lay providers (5) Lay providers had fewer than 3 years of training (1) The peer-reviewed publication was not a research article (2) The peer-reviewed publication was not in English (3) Pilot study To determine whether a system-wide approach was taken in the evaluation of the intervention and to identify system-wide effects when available, authors identified features of interventions relevant to the WHO building blocks framework as well as systems thinking characteristics used in the study (AHPSR, 2009). Systems thinking characteristics considered included: capturing perceptions and interactions of multiple interacting agents, network analysis, mapping of contextual factors, process mapping, describing feedback mechanisms, and other approaches that could inform system dynamics modelling (Peters, 2014). Manuscripts were coded for identification of barriers, facilitators, and outcomes that were relevant to each of the six building blocks: (1) Service Delivery, (2) Health Workforce, (3) Information Technology, (4) Medicines & Medical Devices, (5) Financing, and (6) Leadership and Governance (WHO, 2007a). A seventh building block for communities and people was also included in data abstraction. Authors also made note of the range of stakeholders consulted in the study. The building blocks model allowed for a systematic way to determine whether the impact of the intervention was assessed beyond the specific building blocks in which they were implemented (i.e. health workforce and service delivery in the case of task shifting for mental health). By looking at the level and range of stakeholder engagement, we were able to identify instances were roles and interactions of stakeholders not directly involved in the intervention were explored, as is customary in systems thinking. Use of system dynamics theory, causal loop diagrams, and other system modelling techniques were also included in the extraction criteria, but none were found.

Results
From the 1357 papers identified, 817 were found through PubMed, 271 from Cochrane Library, and 269 from CINAHL. Removing 249 duplicates, 1108 papers were screened based on titles and abstracts. Of these, 147 met the criteria for the first stage of inclusion (four out of five criteria for inclusion met). Upon reviewing full texts, a final set of 30 papers were Economic evaluation Was a comprehensive description of the competing alternatives given? Does the paper provide evidence that the programme would be effective? Were all effects of the intervention identified, measured and valued appropriately? Were all important and relevant resources required and health outcome costs for each alternative identified, measured in appropriate units and valued credibly? Were costs and consequences adjusted for different times at which they occurred (discounting)?
Was an incremental analysis of the consequences and cost of alternatives performed? Was an adequate sensitivity analysis performed?
All studies showed that task shifting for mental health was feasible and acceptable in the given contexts; however, perceptions of quality of care provided by lay providers remain uncertain (Patel et al. 2011;Petersen et al. 2011;Rotheram-Borus et al. 2015;Agyapong et al. 2016). A meta-analysis of outcome measures was not done as the interventions were diverse, conducted at multiple scales, and included qualitative evaluations of stakeholder perceptions. See Table 5 for characteristics of included studies.
Studies were conducted in India, Ghana, Zimbabwe, Pakistan, Malawi, South Africa, Uganda, Indonesia, Iraq, Argentina, Botswana Jamaica, Ethiopia, Zambia, and Thailand, primarily at the district (includes village) level (see Table 5). Across these different contexts, community mental health programmes were variable in nature with some being more integrated into existing health systems (Patel et al. 2010b(Patel et al. , 2011Petersen et al. 2011;Mendenhall et al. 2014;Agyapong et al. 2015a, b;Nimgaonkar & Menon, 2015;Agyapong et al. 2016;Wright & Chiwandira, 2016). Others were more programme-specific in nature and targeted specific at-risk populations, such as mothers suffering from depression, people living with HIV/AIDS,    Prevalence of ICD-10 CMDs and the severity of symptoms of depression and anxiety in individuals attending public primary healthcare facilities with a CMD and in the subgroup of individuals with depression, over a 12-month period, was reduced using the MANAS collaborative stepped-care intervention led by lay health counsellors. Reduction in the risk of suicidal behaviours (plans or attempts) and disability days (days of no work or reduced work) and weaker effects on overall disability scores were also seen (Continued) global mental health

Intervention effects across the building blocks
Of the 30 studies, 25 (83%) included mention of the six WHO health system building blocks other than service delivery and health workforce ( Table 6). All 30 studies included some aspect of the seventh additional building block (communities and people) through community engagement and/or efforts to understand community needs in order to best integrate lay providers.
Sixteen studies of the 25 (80%), considered the role of information and technology. This building block was often mentioned in terms of use of technology for screening of mental illness (Hung et al. 2014;Padilla et al. 2015), use of mobile technology for supervision of lay providers (Tomlinson et al. 2011;Magidson et al. 2015;Agyapong et al. 2016), and need for improved data management tools to ensure adequate follow-up patients at-risk of poor mental health (Agyapong et al. 2015b;Abas et al. 2016). Facilitators identified to support this need were the use of step sheets for enhanced fidelity to interventions and training on documentation of patient visits on mobile phones Rotheram-Borus et al. 2015;Murray et al. 2015b).
Eleven studies (55%) considered the implications of the medicines and medical devices. The discrepancies between training and service delivery in prescribing practices were a challenge in task shifting for mental health (Agyapong et al. 2016). That is, lay providers, not trained in prescription of psychotropic medicines, found themselves prescribing them due to community needs (Agyapong et al. 2015a). Shortage of medicines and the resulting limitations placed on lay providers were impediments in achieving improved health outcomes and demoralized providers who were unable to provide adequate care (Petersen et al. 2011; risk of harm to self or others, (2) providing client-focused psychoeducation, (3) providing clients with psychological and emotional support, and (4) promoting psychosocial support through families and the wider community 224 people in distress HSAs' approach to mental health care delivery was found to be both credible and practical in meeting the needs of the population. Sustained scale-up and integration of the delivery model was observed. Increased case detection was seen by HSAs. No changes were observed in visits to psychiatric hospitals SD, HRH, HIS, C Step sheets used to ensure fidelity and follow-up To better serve the psychosocial needs of the population, 'the apprenticeship model included feedback loops encouraging local counsellors and supervisors to modify delivery of components to increase the fit with the culture and local setting, based on their ongoing experiences' The trial is a collaboration across NGOs: Burma Border Projects (an international NGO), and three local service organizations -Assistance Association for Political Prisoners-Burma (AAPP), Mae Tao Clinic (MTC), and Social Action for Women (SAW), funded by US Agency for International Development Victims of Torture Fund Buttorff et al.
SD, HRH, FS, C Determining cost to households of mental illness is difficult due to the variable ways households cope with illness Scale-up found to be cost effective based on model proposed (Continued) global mental health SD, HRH, FS, C Gender differences in treatment response; unable to provide care to all (psychosis patients) due to financial limitations Lorenzo et al.

Lack of horizontal coordination across different sectors involved in disability management
Referral management systems; financial advice to patients Identification of lack of coordination across sectors working on disability and associated feedback mechanism Education, Social Development, Transport sectors involved; lack of coordination was a challenge Magidson et al.
SD, HRH, IT, C Telemedicine for supervision Mendenhall et al. (2014) SD, HRH, IT, MD, FS, LG, C Lack of infrastructure, overburdening workload, community preferences around who should work as lay providers, lack of recognition for taking on new roles, unclearly defined roles, lack of private spaces for mental health consultation, and confidentiality; social and educational factors posed challenges to acceptability (i.e. perceived inability to provide sufficient care); lack of transport to a health facility, inadequate compensation, and limited availability of specialists for training and supervision of lay providers; failure to prioritize psychotherapy and behavioural interventions alongside a bias toward medication Where high-risk cases were identified, the Child Protection Unit was informed, initiating an investigation for child abuse and neglect Murray et al.
SD, HRH, IT, C Transportation, personnel problems, culture and climate, and buy-in Apprenticeship model using step sheets and detailed information allowed the project to work Barriers and facilitators identified during the implementation of the project were fed back into implementation design, adjusting for cultural and contextual needs (e.g., addition of alcohol use) Neuner et al.
SD, HRH, FS, LG, C Forced repatriation in settlement camps forced refugees into hiding and a resettlement programme caused loss to follow-up; basic package of health services did not include mental health Access to food, economic situation and educational background were captured to provide sociodemographic background. The Ugandan government, the red cross, and the United Nations High Commissioner for refugees provided basic package of health services, and food packages, respectively  Nimgaonkar & Menon, 2015). The bias towards medication as treatment also created challenges for prioritization of psychotherapy and behavioural interventions, affecting demand-side acceptability (Mendenhall et al. 2014). Fifteen studies (70%) raised financing issues in task shifting for mental health with most referring to lack of funds as a limitation to scale-up, pointing to the need to prove cost-effectiveness as a means of ensuring investment by policy and decision makers (Agyapong et al. 2015a(Agyapong et al. , 2016Murray et al. 2015b). Lack of financial incentives for lay providers and their supervisors was another challenge raised (Mendenhall et al. 2014;Abas et al. 2016). Some studies mentioned demandside financing as a barrier to improved mental health delivery, citing the ability to pay for basic mental health services from the patient's perspective (Neuner et al. 2008;Nimgaonkar & Menon, 2015;Abas et al. 2016). Ensuring that referrals were made to services covered by social protection mechanisms, was raised as an important element of providing sustainable and effective mental health service delivery by lay providers (Lorenzo et al. 2015;Padilla et al. 2015). Supporting patients through advice for socioeconomic well-being and links to income-generating projects was a means through which lay providers tried to address demand-side financial barriers (Baker-Henningham et al. 2005;Petersen et al. 2011;Lorenzo et al. 2015).
Finally, 10 studies (50%) mentioned leadership and governance issues with reference to task shifting for mental health. Programme-level supervision of lay providers, which was raised as a challenge across most of the studies included, was not captured as an overarching leadership and governance issue in this review as it is not sufficiently addressing system-level leadership and governance (Schneider & Lehmann, 2016). Perception surveys in Ghana directly involved policy directors, which provided an improved understanding of the gap between perceptions of lay provider programmes by policy directors and realities in the field (Agyapong et al. 2015a(Agyapong et al. , b, 2016. Other studies referenced the need for policy support to integrate mental health services by lay providers into existing practice, citing governance structures as facilitators in scale-up and integration of mental health services through decentralization of these services (Petersen et al. 2011;Nimgaonkar & Menon, 2015). Leadership and governance structures were also barriers to integration. In larger, multi-country studies, lack of clarity in lay provider roles and confidentiality issues undermined integration of programmes from a supply-side perspective (Mendenhall et al. 2014). Communitylevel acceptability of programmes and perceptions on who can be a lay provider were cited as demand-side challenges that need mitigation through improved  transparency, accountability, and leadership that listens to the needs of the population, such as the need for transportation (Mendenhall et al. 2014). One study highlighted the siloed effect of multiple vertical programmes addressing disability across different sectors with no oversight or horizontal coordination (Lorenzo et al. 2015). In programmes targeted at vulnerable populations, such as refugees and orphans, continuity was a challenge as these populations are mobile. Leadership and governance issues beyond the health sector played a heavy role in the ability of lay providers to provide necessary mental health services; therefore, collaboration with other officials was raised as being important to the intervention (Neuner et al. 2008;Murray et al. 2015b).

The use of systems thinking tools in evaluation of interventions
System dynamics theory or modelling tools were not directly used in any of the included studies; however, six studies took a more comprehensive approach in capturing system implications of the intervention being studied. An important element of systems thinking is understanding roles, characteristics, and interactions of the players involved. The perceptions surveys conducted in Ghana, the phenomenological approach across South Africa, Botswana and Malawi, the multicountry stakeholder perspective mapping, and the cross-country comparison of South Africa and Uganda through interviews and focus groups captured such perspectives and allowed for improved understanding of gaps to ensure successful scale-up and integration into the health system (Petersen et al. 2011;Mendenhall et al. 2014;Agyapong et al. 2015aAgyapong et al. , 2016Lorenzo et al. 2015). These studies demonstrated the range of actors necessary for successful integration and showed that actors may have different interpretations of challenges, and different strengths in mitigating these challenges. Systems thinking also should allow for a non-linear process of change, whereby study findings are fed back into the design of the intervention; implementation research methods facilitate this, making adjustments for context and cultural needs (Mendenhall et al. 2014;Murray et al. 2014;Nimgaonkar & Menon, 2015). In community mental health, robust referral pathways are an important piece of integration and working across stakeholders is necessary to ensure appropriate follow-up and service delivery for patients, not just within the health system, but also across other social sectors (Petersen et al. 2011;Lorenzo et al. 2015). Intersectoral components of included studies were captured in this review where available. Intersectoral collaboration here is based on the WHO concept of intersectoral action for health, defined as 'a recognised relationship between part or parts of the health sector with parts of another sector which has been formed to take action on an issue to achieve health outcomes (or intermediate health outcomes) in a way that is more effective, efficient or sustainable than could be achieved by the health sector acting alone' (WHO, 1997). Eight studies touched on efforts made beyond the health sector. These interventions focused on the education sector, where peer group support for AIDS counselling (Kumakech et al. 2009) or support for disability management (Lorenzo et al. 2015) would take place; across non-governmental organizations for vulnerable populations ; and with the criminal and social services sectors (Murray et al. 2015b). In addressing disability challenges, social development and transport sectors were involved to make the lived environment more supportive of those living with both physical and mental disabilities (Lorenzo et al. 2015). Collaboration with the judicial system was also important in cases where abuse and neglect were part of the diagnosis (Murray et al. 2015b).
Several studies raised social determinants of mental health, such as socioeconomic status, employment, lack of education, and violence as risk factors that needed to be addressed in order to enhance the positive effect of task shifting for mental health (Petersen et al. 2011;Mendenhall et al. 2014;Thurman et al. 2014;Lorenzo et al. 2015;Nimgaonkar & Menon, 2015;Wright & Chiwandira, 2016). One study highlighted health promotion activities through working with community resources, such as schools and churches, as an enabling factor (Wright & Chiwandira, 2016). Another mentioned the lack of such collaboration with other sectors as a barrier in seeing improved treatment outcomes (Thurman et al. 2014). Four studies had formal arrangements for embedding intersectoral practice in the task shifting (Petersen et al. 2011;Lorenzo et al. 2015). The intersectoral fora created to support these programmes strengthened their ability to integrate into existing systems and provided a wider range of community referral pathways for lay providers to use in linking their patients to the resources necessary for thriving, thereby indirectly enhancing mental health (Petersen et al. 2011;Chatterjee et al. 2014;Lorenzo et al. 2015). One such example is the referral of patients to income-generating programmes within the agricultural sector (Petersen et al. 2011).

Discussion
Despite the global call to action to improve scale-up and integration of lay provider programmes, the evidence base around implementation, scale-up, and integration of task-shifting strategies for mental health remains limited in both quantity and breadth (GHWA, 2013;Weinmann & Koesters, 2016). Moving from fragmentation to integration requires a move beyond issues specific to lay provider programmes, such as remuneration, training, and supervision (Schneider & Lehmann, 2016). It needs an understanding of large-scale public sector involvement, interactions across key actors, mobilization of these actors, and monitoring and evaluation tools that capture the complex adaptive parts within the system as they shift and respond to scale-up toward a true community system Schneider & Lehmann, 2016). A systems thinking approach can help capture these complexities and understand how to optimize community mental health systems (Peters, 2014).
This review demonstrates that there is space for more systematic approaches to studying health systems elements that affect and/or are impacted upon by task-shifting interventions for mental health. None of the included studies systematically studied system elements; however, many touched upon the WHO building blocks of the health system other than those directly related to task shifting (i.e. service delivery and health workforce). These studies included qualitative methods that allowed them to capture some of the interactions within the system and highlight barriers, facilitators and effects that fell outside the limited scope of the task-shifting intervention (Petersen et al. 2011;Lorenzo et al. 2015;Nimgaonkar & Menon, 2015;Padilla et al. 2015;Agyapong et al. 2016).

Barriers and facilitators of scaling up mental health care by the building blocks
Barriers to scaling up mental health services identified across studies included: stigma around mental health in the community (Ali et al. 2010;Nimgaonkar & Menon, 2015;Padilla et al. 2015); poor documentation and loss of follow-up due to lack of robust data management and patient management tools (Agyapong et al. 2015b); lack of access to psychotropic medicines and/or lack of sufficient training for rational prescribing practice (Patel et al. 2011;Agyapong et al. 2015a;Nimgaonkar & Menon, 2015); geographic and financial demand-side barriers to access of mental health services (Baker-Henningham et al. 2005;Petersen et al. 2011;Mendenhall et al. 2014;Agyapong et al. 2015a); poor collaboration with spiritual and traditional healers (Agyapong et al. 2015a); disconnect between providers and decision makers (Agyapong et al. 2015a(Agyapong et al. , 2016Rotheram-Borus et al. 2015); existing heavy workload of lay providers (Petersen et al. 2011;Hung et al. 2014); gender differences in responding to treatment (Larson-Stoa et al. 2015); and lack of access to community resources to support social determinants of mental health (Tomlinson et al. 2011;Thurman et al. 2014;Rotheram-Borus et al. 2015).
Facilitators to scaling up mental health services identified across studies included: suitability of lay providers due to their ability to relate to the community and their patients (Baker-Henningham et al. 2005;Padilla et al. 2015;Abas et al. 2016); support from specialized mental health professionals (Agyapong et al. 2015a, b); use of technology and telemedicine to support supervisory practice (Patel et al. 2011;Tomlinson et al. 2011;Magidson et al. 2015;Agyapong et al. 2016); integrated interventions that include life skill building for sustainable livelihood practice, social interaction, and self-care (Petersen et al. 2011;Chatterjee et al. 2014); and integration into existing networks with robust service delivery models that support lay providers (Petersen et al. 2011;Nimgaonkar & Menon, 2015).
While information and technology tools appear to be facilitators for optimizing service delivery by lay providers, care must be taken in the selection of technology solutions. It is critical to understand how providers use technology as a part of the system. Some tools require the interpretation and training of health professionals to be appropriately and efficiently used, suggesting that not all technologies are readily transferable across health workforce cadres (Jotheeswaran et al. 2015;Robbins et al. 2015). Inefficiencies in the system can also be found in poor access to medicines (WHO, 2009). Financial and procurement barriers impede access to essential psychotropic medicines, impeding the delivery of appropriate mental health care to those who require pharmacotherapy (Agyapong et al. 2015a;Nimgaonkar & Menon, 2015). Scaling up mental health treatment by lay providers without addressing access to medicines in parallel will prove unsuccessful by undermining the quality and impact of additional service provision (WHO, 2009).
The barriers and facilitators outlined here showcase the complexity involved in task shifting for mental health and the need for a broader systems approach to mitigating barriers and leveraging facilitators. By being community-based and having a deep understanding of community needs and assets, lay providers have an enhanced ability to identify social determinants of mental health within a given context (Richters et al. 2013;Padilla et al. 2015). This rich knowledge, combined with appropriate training, puts them in the optimal position to refer patients to relevant social services (Paudel et al. 2014). Mental health is often a comorbidity in chronic disease management; training programmes should also prepare lay providers with the knowledge and skills necessary to understand such linkages and to refer appropriately (Rotheram-Borus et al. 2015).
Establishing networks and intersectoral linkages is not easy. Despite policy support, implementation and scale-up of integrated approaches to strengthening community mental health remains a challenge ). Even where formal mechanisms are in place for intersectoral collaboration (i.e. where formal engagement of health, education and development sectors are embedded in programme design), participation was erratic and uncertain without senior officials present (Petersen et al. 2011). Existing models and formal arrangements of intersectoral collaboration require additional incentives and governmental support. In this way, partnerships move beyond platitudes and truly work as collaborative fora that support lay providers in assessing patient needs and selecting appropriate referral pathways.

Implications for future research
Mental health is rarely an isolated problem. It sometimes stems from physical, environmental, or sociocultural challenges and creates positive feedback loops that become difficult to break (Tomlinson et al. 2011;Thurman et al. 2014). Taking a systems thinking approach to unpacking task-shifting interventions for mental health will unveil extant opportunities and threats in the current system. A system-level understanding of interventions will allow for improved integration and effective engagement of important actors overlooked in the traditional model of implementation and evaluation. Such actors include caregivers, nongovernmental entities that support social determinants of mental health, employers, spiritual leaders, and other social sectors (e.g., education, agriculture, transport, social development, etc.) (Schneider & Lehmann, 2016). Representing the system overall and opportunities for improvement in the implementation and evaluation of such programmes can advocate for further investment in community mental health systems strengthening.
With appropriate evidence describing the roles and contributions of diverse sectors to mental health outcomes, there is potential to facilitate strategic intersectoral investment for optimal health impact as well as cost-effectiveness. Stigma, for instance, is an often-cited barrier in mental health seeking behaviours and even in provision of mental health care (van Ginneken et al. 2013;Nimgaonkar & Menon, 2015;Iheanacho et al. 2016;Weinmann & Koesters, 2016). A study of church-based lay providers showed that higher education was correlated with improved bio-psychosocial perspectives on mental health and fewer displays of stigma-based behaviour (Iheanacho et al. 2016).
Overcoming stigma is therefore not necessarily limited to the role of the health sector; the education sector can play an important part in addressing stigma by talking about mental health and raising awareness.
Due to the nature of the search strategy, this review highlighted interventions that were conducted primarily in the health sector. Few included studies employed task-shifting strategies across other sectors to enhance mental health promotion. More examples of such collaboration exist, especially in education and social services. Therefore, it would be worth conducting a more targeted review of the evidence on interventions happening in other sectors that have impacts on mental health. Skill-mixing interventions also warrant more emphasis as they highlight the need for a range of skills beyond the health sector to address mental health challenges. Comparing the effectiveness of mental health-related interventions housed in the health sector v. those in others would be valuable in identifying cost-effective opportunities for intersectoral collaboration and cohesive strategies for mental health.
Limitations of this review may include the wide variety of mental health interventions, populations studied, and outcome measures included, making it potentially difficult or inappropriate to apply this review's broader conclusions to unique mental health conditions. While the majority of studies did not explicitly use a systems thinking approach, some studies indicated implicit consideration of systems thinking characteristics. It is possible that studies neglecting to mention system-wide effects in final manuscripts did in fact acknowledge these effects in the design and implementation of interventions to some degree, but this data was subsequently not available for this review.

Conclusions
Task shifting for mental health has been demonstrated as an acceptable and effective approach to addressing the mental health gap in LMICs. This review shows the complexity of task-shifting interventions by exploring interactions of intervention elements and actors across the six WHO building blocks. There is a lack of systematic approaches to exploring this complexity in the evaluation of task-shifting interventions. Systems thinking tools should support evidenceinformed decision making for a more complete understanding of community-based systems strengthening interventions for mental health.