Six RCTs [25–30] and four Cluster RCTs [31–35] were reviewed. Six trials focused on CHW interventions for promoting exclusive breast feeding [27, 28, 32, 29, 30, 33]. One trial each focused on malaria , diarrhoea  and nutrition [34, 35]. One examined CHW's effectiveness for child health interventions in general .
In all the RCTs, the intervention involved service delivery at homes of beneficiaries. In all the RCTs, the intervention actually consisted of a bundle of interventions. We identified three broad categories of interventions in the RCTs we reviewed. Training: which involved the provision of knowledge and skills based training to CHW and included practice sessions. Systemic Interventions: This involved establishment of clear roles and specific responsibilities (roles made clear to CHW, other health workers and beneficiary communities); it also included rigorous supervision and mentoring of CHWs by formal health service professionals and good referral support from the formal health service. Interventions involving the positioning of the CHW within the beneficiary community: through the CHW being explicitly selected by beneficiary community and the CHW being positioned as a role model within the beneficiary community.
The ten trials reported a variety of primary and secondary outcomes. We inductively categorised these outcomes from the point of view of assessing the CHW's performance, as being positive - when the intervention worked or negative. In eight trials, the intervention had a positive outcome and the CHW's performance was good. In one trial the intervention had a negative outcome, but the CHW's performance was good [34, 35]. In one trial the intervention had a negative outcome and the CHW's performance was not good .
Additional File 1 summarises the RCTs, the context, the intervention, the mechanisms triggered and the reported outcomes. Additional File 1 shows that in all the trials, more than one type of intervention was applied to improve CHWs performance. It also shows that the outcomes are reported not in terms of CHW performance, but rather in terms of the consequences of their performance on specific health outcomes.
In six of the trials, the CHW interventions were implemented in urban areas amongst beneficiary communities who were poor and had an unmet need. The interventions were embedded in or closely linked to local public health care services, and were implemented by locally trusted agencies. In such a context when the intervention involved selection or election of CHW from within the beneficiary community, particularly of such individuals who were trusted and seen as role models by the community - it triggered a sense of relatedness between the CHW and beneficiaries, a sense of responsiveness and responsibility amongst CHWs; it also led to a feeling of being valued by peers for fulfilling the needs of their community. When this was so, the interventions had positive outcomes [27–32].
In the same context, when the intervention involved training CHWs on specific tasks targeted at specific situations, and the training was supplemented by practice sessions and on-job mentoring - it triggered a sense of self efficacy amongst CHWs; the skills building and practice sessions helped CHWs gain enactive mastery on the tasks and triggered a sense of confidence in being able to solve problems.
Similarly, when the intervention involved CHWs being supervised and mentored by local formal health services,- it triggered a sense of credibility of being a part of the system and a sense of assurance both for themselves and for the community, that there was a system of back-up in times of need. When this was so, it contributed to the intervention having positive outcomes.
In Kidane et al's study , the CHW interventions were implemented in a rural area amongst beneficiary communities who were poor, had an unmet need and in addition had strong community solidarity in view of past collective adverse experiences. In this case too, the CHW intervention was embedded within local health services and the intervention was composed of three key components similar to the trials implemented in urban areas. In spite of the context being slightly different (rural compared to urban settings), the intervention triggered similar mechanisms as in trials set in urban settings.
In Bhandari et al's study , the CHW intervention was implemented in a rural area amongst beneficiary communities who were not so poor and probably had only some unmet needs. Here too, the CHW intervention was embedded within local health services, the intervention was composed of similar components, but only the training component triggered similar mechanisms as mentioned above . Unlike other trials, in Bhandari et al , an expectation of appreciation by authority (study team during study period) and possibility of being rewarded was explicitly reported as a factor that may have motivated CHW's to perform better.
In another rural context, with the CHW intervention targeting poor beneficiaries with unmet needs , with a similar training intervention, but when the CHW's were not explicitly chosen by beneficiaries and were appointed by the political establishment - it led to an absence of relatedness and responsibility amongst CHWs. It compromised the CHW's motivation to perform, undermining intervention outcomes .
In the same context, because the CHW's were not working together with, nor supervised by local health services, it led to poor sense of legitimacy of the CHW amongst beneficiaries and formal health services. When this was so, the CHW's performance and intervention outcomes were compromised.
In the same study, where the CHW intervention was embedded within the local public health services, but the intervention did not clearly articulate the CHW's roles (to CHWs, to beneficiaries and to the health services), the uncertainty because of ill defined roles led to lowered motivation, less involvement (of CHWs, of cadres of formal health services and beneficiaries), and ultimately to poor performance and outcomes . The lack of clarity of roles in the intervention also led to an environment of confusion (for CHWs, cadres of public health services and beneficiaries alike) and compromised the outcomes (performance of CHWs and the intervention outcomes).
When the same intervention elements were applied in a context  where the intervention did not address an unmet need of the beneficiaries, it had no value for the beneficiary community. In such a situation, the intervention outcome, in spite of good performance by CHWs, was compromised .
Discussion and Conclusions
The aim of this review was to explore if RCTs could yield insight into the working of interventions involving CHWs for improving child health, when examined from a realist perspective. We found that RCTs did yield some insight, but this insight was very general.
In the context of CHW programs targeting the poor with an unmet need, and embedded in or closely linked to local health care services, we can conclude that:
Training interventions in the form of knowledge and skills based training complemented by ongoing in-field mentoring, can improve the CHW's performance when they are able to trigger the following mechanisms:
a sense of self efficacy and enactive mastery of the tasks,
an increase in self esteem,
assurance that there is a system for back-up support.
Health system related interventions in the form of setting clear roles and specific responsibilities for CHWs, ensuring mentoring for CHWs by health workers from local public health services, ensuring good referral support for CHWs from local public health services, can improve the CHW's performance when they are able to trigger the following mechanisms:
a sense of relatedness with the local public health services, and thus accountability towards the system,
a sense of credibility and legitimacy of being part of the local public health services,
an anticipation of being valued by the local public health services and the community,
a perception of improvement in social status,
an assurance that there is a system for back-up support.
Interventions involving better positioning of the CHW within communities (Eg: Selection of the CHWs in consultation with beneficiary communities; the CHWs being members of the beneficiary community, and perceived by them as role models) can improve the CHW's performance when they are able to trigger the following mechanisms:
an anticipation of being valued by the community,
a perception of improvement in social status, and having a valuable social role
a sense of relatedness with and accountability to the beneficiaries
This inventory of CMOC, derived from the RCTs under review, is limited, but a beginning nevertheless. This inventory should be seen as a set of generic hypotheses derived from the best existing evidence. The inventory is by no means complete and as stated earlier it is very generic. It is generic because the context has not been sufficiently reported in the RCTs. For example, similar interventions were implemented in urban and rural settings, and they had similar outcomes, but the context (rural vs urban) was not sufficiently disaggregated and described to allow sufficient understanding whether or not different aspects of this broader context triggered different mechanisms. However, we found that many of these generic hypotheses are corroborated by findings from many earlier reviews [4, 10, 11] giving an inkling of the external validity of these CMOC. For instance, Lehmann and Sanders  found, that to be effective, the CHW must not only be from the beneficiary communities, they must conform to the norms and customs of the community they serve. They also found that selection, training, supervision and availability of logistic support are important factors for the CHW's performance; they also found that the CHWs's success is contingent to their embedment in beneficiary communities and support from the government and political establishment. Haines et al  point out that CHW interventions can be undermined by a corrupt, partisan and patronage based political establishment. They also point out that CHW intervention outcomes depend on whether power equations in the relationship between CHWs and professionals lead to creation of trust and harmony or rivalry and distrust. Though not labelled as such by the authors, they are in fact referring to mechanisms being triggered by the interaction of the intervention and the context. Haines et al  also found that CHWs performed best when they had limited responsibilities and focused tasks. They also found that supportive supervision and support from formal health services is critical to CHW's success. Lewin et al  found evidence that lay health workers (CHW) can deliver certain specific services well and not a wide range of services. Bhattacharya et al  and Haines et al  conclude that to be successful, CHW interventions need to have multiple incentives, simultaneously, at multiple levels (individual, community, and health system levels), tailored to local context. While these reviews do point to the importance of context, we found that most of these reviews gave only limited insight into the context in which various interventions were applied; this was probably due to the focus of the reviews.
During our literature search, we came across a vast array of material on the deployment of the CHWs for child health in LMIC. As a follow up to this review, further realist review of this literature needs be undertaken. This can not only yield a more complete set of CMOC, it can help test and refine these to develop a better understanding of the working of the CHWs in LMIC.