Linking communities and health facilities to improve child health in low-resource settings: a systematic review

Abstract Community–facility linkage interventions are gaining popularity as a way to improve community health in low-income settings. Their aim is to create/strengthen a relationship between community members and local healthcare providers. Representatives from both groups can address health issues together, overcome trust problems, potentially leading to participants’ empowerment to be responsible for their own health. This can be achieved via different approaches. We conducted a systematic literature review to explore how this type of intervention has been implemented in rural and low or lower-middle-income countries, its various features and how/if it has helped to improve child health in these settings. Publications from three electronic databases (Web of Science, PubMed and Embase) up to 03 February 2022 were screened, with 14 papers meeting the inclusion criteria (rural setting in low/lower-middle-income countries, presence of a community–facility linkage component, outcomes of interest related to under-5 children’s health, peer-reviewed articles containing original data written in English). We used Rosato’s integrated conceptual framework for community participation to assess the transformative and community-empowering capacities of the interventions, and realist principles to synthesize the outcomes. The results of this analysis highlight which conditions can lead to the success of this type of intervention: active inclusion of hard-to-reach groups, involvement of community members in implementation’s decisions, activities tailored to the actual needs of interventions’ contexts and usage of mixed methods for a comprehensive evaluation. These lessons informed the design of a community–facility linkage intervention and offer a framework to inform the development of monitoring and evaluation plans for future implementations.


INTRODUCTION
With 5 million under-5 deaths in 2021, childhood mortality still presents a major health challenge globally (UNICEF, 2023).Local, national and international bodies, including local government and NGOs, have implemented a wide variety of interventions focused on health systems' improvement to try to reduce these avoidable deaths.This is especially true in resource-constrained settings where child mortality is higher (Cha and Jin, 2020;World Health Organization, 2023).Causes of this burden include poor healthcare provision at local health centres (Wanjau et al., 2012;Kruk et al., 2018), and poor health knowledge and trust towards the healthcare system at community levels (Woskie and Fallah, 2019;Moucheraud et al., 2021).Therefore, these improvement interventions have focused on all system levels, from facility improvements to community health education.Often, development programmes work at the community and facility level at the same time, trying to tackle health challenges from multiple perspectives (Metwally et al., 2020).
Frequently, these types of programmes do not focus on creating a connection between the two different locations of action, i.e. facility-based and community-based interventions remain confined to their settings.This therefore hinders the possibility of obtaining sustainable and long-term improvements (Oguntunde et al., 2018).However, whenever community and facility interventions are implemented, a communityfacility linkage component could be included in the delivery plan.Community-facility linkage is defined as a 'formalized connection between a health facility and the communities it serves to support improved health outcomes' (Gulaid, 2015, p.7).There is no definitive evidence in the literature regarding the effectiveness of interventions aimed at strengthening such connections, but numerous approaches could foster accountability improvement for health in local communities, reduce missed opportunities for adequate care, and improve appropriate and timely care-seeking behaviours and overall health outcomes (World Health Organization, 2008;Diaconu et al., 2020).
One feature that this type of 'whole system' implementation could benefit from is the more active participation of local communities in the intervention design and delivery.Their knowledge of the health challenges peculiar to the intervention setting could help to tailor interventions to the population's actual needs (World Health Organization, 2016).Also, empowering communities to take ownership and lead development programmes is considered a powerful tool to obtain a long-lasting transformation, sustainable over time, maintaining positive outcomes after the intervention is concluded (Bigdeli et al., 2020).
As these interventions can vary a lot in how they are implemented and the evidence of effect is mixed, we conducted a systematic literature review to identify the components that make them successful, or not.Our research focused on answering this research question: 'what key features (contexts, mechanisms, strengths, weaknesses, limitations) shape outcomes of previous community-facility linkage interventions to improve child health in rural areas of low or lower-middle income countries (similar to the socio-economic and geographical conditions in Jigawa, Nigeria)?'.Our aim was to understand which attributes contribute to better child health outcomes and to illuminate which characteristics are linked with improved community participation when implemented in specific settings.While a prior literature review explored community participation and health facility committees as a mean to connect communities with health facilities (McCoy et al., 2012), there is a gap in the existing research.To the best of our knowledge, there has been no comprehensive review specifically addressing various types of communityfacility linkage interventions and their impact on child health.This review was conducted with the purpose of informing the theory and design of a whole systems intervention being trialled in Jigawa State, Nigeria, with the specific aim of supporting the design of the monitoring and evaluation plan for the community-facility linkage component of it [INSPIR-ING Project: (ISRCTN39213655, 2019;King et al., 2021)].This intervention aims to reduce childhood mortality due to pneumonia and other infectious diseases in a setting where the under-5 mortality rate is 192 deaths per 1000 live births (National Population Commission-NPC and ICF, 2019), and there is low-quality healthcare system, poor community health education and lack of protective and preventive factors (Iuliano et al., 2020;King et al., 2020;Shittu et al., 2020).

Search strategy
We conducted a systematic literature review examining three online databases (Web of Science, PubMed and Embase).By utilizing these databases, we aimed to ensure a thorough and multidisciplinary approach to our research topic, allowing us to gather a comprehensive overview of the existing literature, capturing a wide range of relevant studies and ensuring the validity and reliability of our findings.We used these search terms: health facility/hospital, communitybased organizations/community health workers/community development organizations, linkage/referral, child/paediatrics (specific search strategy for each database is detailed in Appendix 1).We did not apply any additional filter in our search at this stage.The search strategy was peer-reviewed by the INSPIRING research team with the technical support of the librarian team of the authors' institute.After searching the online databases, we obtained records of the articles to screen for studies published by 3 February 2022.

Inclusion/exclusion criteria
The variables included in the selection criteria were chosen to find interventions with similarities to the ones being implemented in the INSPIRING Jigawa trial (King et al., 2021).As Jigawa is predominantly non-urban, and Nigeria is a lower-middle-income country (WorldBank, 2021), two of the inclusion criteria imposed that the studies had to be set in rural areas (thereby excluding those set in urban environments) and in low-or lower-middle income (LMIC) countries.The intervention being evaluated had to have a community-facility linkage component-which was defined as: any form of intermediation or connection between the health facility and the community it served.This linkage could be established through mechanisms such as community health workers, health development committees, or other formalized bodies or activities aimed at bridging the gap between communities and health centres, specifically with the goal of enhancing under-5 child health.We specifically sought interventions that went beyond mere focus on health facility strengthening or community health promotion; studies emphasizing these aspects exclusively were excluded.While these activities could coexist, a crucial criterion was the presence of a connecting factor that facilitated collaboration between health facilities and community initiatives.
Apart from an interest in the characteristics of the interventions assessed in the articles, the outcomes of interest had to be related to under-5 children's health, and included: under-5 mortality, care-seeking behaviour, healthcare attendance rates and relationship between communities and healthcare providers.
Other inclusion criteria were: peer-reviewed articles (to ensure high quality, reliability and rigour in our systematic review) containing original data and written in English (due to limitation in translation capabilities).We did not apply any date restriction or restriction on study design used for intervention evaluation.

Selection of studies, data extraction and risk of bias assessment
Following PRISMA reporting guidelines (Page et al., 2021), the first step involved conducting an analysis of all the found articles' titles to discover those to potentially include.In the second step, all the abstracts of the articles selected based on the title were scrutinized.Third, a full-text analysis of all the abstract-based potentially relevant studies was performed.The final choice of the articles to include was made based on whether the intervention methodology and outcome data were deemed fitting the predefined inclusion/exclusion criteria, and whether the papers met quality standards (any limitations were noted and are reported in the Results table).The first and last author independently conducted the same screening process.Once they each reached the final stage of selecting papers for inclusion, the two authors met and resolved in discussion any disagreement on eligibility.We did not register the review, and we decided not to use any bias assessment tools.The variety in methodologies of the included articles would have required the usage of multiple checklists, and we deemed that a critique of each paper's quality by the authors, who have methodological expertise to assess the articles, suffices, along with the fact that the included papers have already been peer reviewed.We carefully evaluated each study's methodology and potential of bias against the limitations reported by the papers and our own assessment.

Analysis
In order to analyse previous interventions based not only on their characteristics and successes, but also on their transformative and community-empowering capacities, we used Rosato's integrated conceptual framework for community interventions (Rosato, 2015).Any intervention can be assessed based on nine practice variables.Each of these variables is classified using roman numerals from I (1) to V ( 5), the more community-driven and centred, the higher the ranking (with the following variable descriptions not being mutually exclusive, but rather each new classification adding to the previous ones): individuals to changes in health knowledge and behaviour in groups in the community to increases in community capacities to change the political system).• Target group (who the intervention is targeting for improvements in health: from the entire community to the same community but within that some specific marginalized subparts, to functional subparts of the community, defined by shared needs, that are the most marginalized and suffer at the hands of the wider community).• Existing strengths and weaknesses and • Role of external agent (how much the intervention recognizes and builds on the existing capabilities of the community to address health issues and what the role of the external agent in the intervention is: from a community considered weak and lacking capacities to help itself, therefore needing the external agent to take all responsibilities, to a community lacking some skills but in possession of some other capacities-an external agent needed to marshal the existing ones-to a community considered to be capable of solving its health problems, with the external agent only reinforcing these capacities).• Participation (involvement of the community in decisionmaking about the intervention: from being simply informed to being consulted tokenistically, to joint decision-making with external agents, to the community leading the decision-making process and being accountable for the outcomes).• Role of community (from being the passive setting/target of the intervention, to being the active source of solutions for the intervention, to being the active agent of change, defining its own health problems, setting its own agenda and being the source of solutions).• Tools and methods (what the intervention employs, from clinical health methods in a short timeframe to behaviourchange ones, to social capacity-building methods in a longer period).• Resources (responsibility of mobilizing resources from laying on the external agent, to being equally split between the external agent and the community, to being entirely in the hands of the community, with the external agent only supporting them and brokering access to the resources).
Five statements representing different levels on the intervention continuum were created for each of the nine practice variables by Rosato.We identified the most fitting statement that reflected each paper's practice concerning each variable and assigned scores.We then plotted these scores on spider diagrams, facilitating a visual representation of the results.
We also applied realist principles to synthesize the included papers.According to realist methodology (Pawson et al., 2005;Pawson, 2013), there is a causal path to explain how an intervention works, what makes it successful or why it fails.Starting from a programme theory to justify the intervention plan, then Context (the physical and social environment) and Mechanisms [resources offered by the programme and how they are received and resonate among the participantsreasoning (Dalkin et al., 2015)] lead to specific Outcomes (CMO configuration).Through this iterative process, a refined understanding of the intervention's underlying mechanisms and contextual factors emerges.This leads to the development of a Middle Range Theory (MRT) (Pawson, 2017), which serves as a specific and contextually grounded framework.
While the original programme theory remains the foundation, the insights gained through the realist analysis process enhance the theoretical framework and the MRT becomes a new starting point for future implementations.
We aimed to build a monitoring and evaluation tool to assess the community-facility linkage component of future projects.So, we gathered the points in common that the MRTs of the interventions included in this review showed and used them to build a new framework to inform future analysis plans.
Furthermore, our study has an interest in foregrounding transformative principles in our research and implementation practices.Mertens (1999Mertens ( , 2007) ) has argued research and evaluation praxis is more effective when decision making, planning and implementation include direct involvement of communities-instead of leaving them as mere targets of interventions-making change more plausible and sustainable.As such, our assessment of the papers included in this review also sought to interrogate the extent to which the existing literature speaks to these principles.
Studies included a wide range of interventions to link communities and facilities, groupable into two main categories: development committees and community health workers/volunteers.Development committees, defined as health facility workers and representatives of the community regular meetings to discuss issues and ways to improve healthcare provision and attendance (in one case specific drugs uptake), were the main component of six studies (Björkman and Svensson,  Olayo et al., 2014;Cannon et al., 2017;Pol et al., 2017;Oguntunde et al., 2018;Ndaba et al., 2020), two of which included also the involvement of community volunteers (Olayo et al., 2014;Cannon et al., 2017).The main protagonists of the other eight studies, of which five included a specific health facility strengthening component (e.g.staff training, equipment and drugs provision) (Bari et al., 2006;Waiswa et al., 2015;Gupta et al., 2017;Awasthi et al., 2019;Var et al., 2020), were health volunteers, who create a bridge between the local community and facility by conducting health education, promotion, quality improvement processes (at community and/or household level) and providing Primary Health Care (Kamugisha et al., 2018;Tancred et al., 2018;Kushitor et al., 2019).Only one intervention included economic incentives for participants-cash benefit for mothers, with free delivery and free treatments for sick neonates (Gupta et al., 2017).

Intervention effects on child health
All the papers highlighted positive effects resulting from the interventions.While a statistical meta-analysis was not conducted, the individual studies consistently reported favourable outcomes (Table 2).

Healthcare access
Eleven of the 14 papers measured changes in community members' healthcare access (Bari et al., 2006 et al., 2020).Improvements in healthcare access were measured through healthcare facility and emergency transport utilization, use of clinic cards, but mostly via an increase in immunization rates.In one study, a reduction of gender disparity in vaccine uptake to the point of having more girls vaccinated than boys was reported.However, the complementary qualitative analysis revealed mistrust towards this practice despite the intervention, with girls being vaccinated because they were considered more expendable than boys by their parents, fearing vaccines' negative events (Gupta et al., 2017).

Community mobilization
Ten studies assessed community mobilization (Bari et al., 2006;Björkman and Svensson, 2009;Olayo et al., 2014;Waiswa et al., 2015;Cannon et al., 2017;Pol et al., 2017;Oguntunde et al., 2018;Tancred et al., 2018;Ndaba et al., 2020;Var et al., 2020).This outcome refers to improvements in health knowledge and practices (including vaccination uptake), changes in hygiene measures and water and sanitation practices in households, participation in health activities and self-referral of sick children.Improvements were reported by all studies, with only one assessed outcomehygiene measures, not significantly increased in one study (Var et al., 2020), and another study reporting socio-cultural barriers, despite a successful intervention (Cannon et al., 2017).While most of the studies focused exclusively on mothers, one reported a rise in involvement of men in child health (Tancred et al., 2018).

Healthcare facilities' services provision
Changes in facility services provision were mentioned by nine papers (Björkman and Svensson, 2009;Waiswa et al., 2015;Cannon et al., 2017;Gupta et al., 2017;Pol et al., 2017;Oguntunde et al., 2018;Tancred et al., 2018;Kushitor et al., 2019).Improvements in quality of care, equipment, waiting times, drug availability and referral systems were reported by most of them.They were often matched, though, with reports of inadequate funding, which would lead to compromises and only partial ameliorations.For example, Oguntunde et al.'s study reported committee members paying out-of-pocket for facility equipment (Oguntunde et al., 2018), Gupta et al. described how no maintenance plan/funds for the new referral system impeded broken ambulances repair (Gupta et al., 2017), and Waiswa et al.'s paper recounted that more than 60% of the facilities reported at least one drug stock-out (Waiswa et al., 2015).Moreover, one study pointed out poor equipment and supplies, with services provided not tailored to local community needs (Kushitor et al., 2019), and another reported that the increased demand gained was not met with an increase in health services capacity, resulting in inadequate patient assistance (Tancred et al., 2018).

Healthcare providers' practices and relationships with community members
Changes related to healthcare providers were presented in six papers (Björkman and Svensson, 2009;Gupta et al., 2017;Pol et al., 2017;Oguntunde et al., 2018;Var et al., 2020), with overall positive outcomes.Improvements ranged from practices (hygiene) to expertise (knowledge and ability to recognize new-born danger signs), from availability and attitude towards the patients to a reduction of absenteeism from the facilities to overall care provision.Gupta et al.'s study reported an inadequate staff increase to meet all the new activities scheduled for implementation (Gupta et al., 2017).Pol et al.'s study, instead, explored the relationship with communities as another aspect of provider change (Pol et al., 2017).Healthcare workers recounted in their feedback reflections a better understanding of community needs (possible thanks to the establishment of a communication channel with the linkage intervention).The relationship between community members and healthcare providers was qualitatively assessed by two other papers.While Ndaba et al. described improved communication and an overall stronger connection (Ndaba et al., 2020), Kushitor et al. reported a poor relationship, with community members complaining about the rudeness of the healthcare workers during Focus Group Discussions.In addition to this, men from the same study felt excluded from service provision, due to the focus exclusively on mothers and children (Kushitor et al., 2019).

Under-five mortality
Information on changes in mortality rates among children under the age of five was provided by three papers (Björkman and Svensson, 2009;Cannon et al., 2017;Kamugisha et al., 2018), with Waiswa et al. reporting that their study was not powered to detect differences in mortality between intervention and control (Waiswa et al., 2015).Cannon et al.'s qualitative case study reported initial estimates on undisclosed

Community participation assessment
The included papers were assessed using Rosato's framework, and a graphic representation of this classification can be found in Figure 2. We grouped the spider diagrams according to the type of intervention implemented, using this framework to offer a perspective to explain the outcomes of each of them.Overall, the variables' scores provide an insight into which characteristics of the included studies carry the biggest transformative and community-empowering capacities and which areas need strengthening, to inform future interventions.
Of the nine variables, 'Conceptualization of health', 'Goal' and 'Tools and Methods' were the domains with overall good scores, while 'Resources', where applicable (five papers did not specify how funding was mobilized), was assigned low scores with the exception of one study (Pol et al., 2017).'Participation' was the variable with the biggest variation, being the one with the highest number of papers scoring I (n = 6), as well as the highest number of papers scoring V (n = 6).
Five of the six studies that used community groups (two of which also involved community volunteers), had the overall highest scores, with 'Participation' scoring V in all of them.This is not surprising for Pol et al.'s paper, as community empowerment was the main intervention focus (Pol et al., 2017), nor for Olayo et al. who, by reporting participation as the most improved outcome, confirmed the importance of involving communities to obtain strong commitment to an intervention (Olayo et al., 2014).The spider diagrams of Oguntunde et al. and Ndaba et al.'s studies, instead, highlight that the lack of involvement of marginalized/at risk groups ('Target' score II), even with a high 'Participation' score, can lead to low awareness of the intervention in the community (Oguntunde et al., 2018) or to limited ability to reach isolated parties (Ndaba et al., 2020).Some of Cannon et al. case-study's low scores ('Participation', 'Community Role' and 'Tools' I), then, explain why the intervention did not succeed in tackling local socio-economical barriers, changes in which were reported as something needed to strengthen the effectiveness of the implementation (Cannon et al., 2017).
Community volunteers/healthcare workers were the core of most interventions, with five of them including a health facility strengthening component.The scores overall were lower, with a few exceptions, like Tancred et al.'s study: the high general scores confirmed the importance of empowering the community for a successful implementation, but at the same time low scores in variables like 'Resources' highlight that such empowerment needs to go hand in hand with a financial plan to improve and sustain the changes, explaining the inadequate service provision reported in the article (Tancred et al., 2018).Among these papers, 'Participation', 'Community role' and 'External agent role' scores were generally low, which could help to explain the poor caregivers' care seeking behaviour change outcomes of Var et al.'s paper (probably due to a lack of engagement of community members in the intervention design and implementation) (Var et al., 2020), or the mistrust towards immunization reported by Gupta et al. (Gupta et al., 2017), confirming once more how interventions should focus on empowering community members and make them protagonists, not just recipients of health implementations.

Realist assessment
Following realist principles, we extracted the information on the programme theories behind the interventions and developed CMO configurations to deduce the MRTs from each paper (Table 3).By doing so, we noticed that the included studies had many similarities.
The programme theories behind each intervention, for example, followed the same general principles: community participation/education, communication with healthcare providers (via meetings, or the establishment of intermediate bodies like committees), and health services ameliorations, can all lead to improvements in quality of care, health outcomes and/or relationship between communities and healthcare providers.
Context and Resources varied based on the different specifics of the setting and the intervention, despite all being set in rural LMIC areas.However, the various participants were in general very responsive to these types of interventions (Reasoning), even when issues were reported with the actual implementation (compared to the original idea) and the exclusion of some populations such as men (Kushitor et al., 2019), or when participants were not able to act upon what they learned during the intervention due to unchanged external matters (Var et al., 2020).
The MRT that we obtained from each paper showed some key factors in common.As MRTs can function as starting points for future intervention designs, we summarized the shared elements that the MRTs showed and built a new framework to inform future implementation and analysis plans (Figure 3).
Trust, collaborative spirit and active participation emerged as fundamental aspects for an intervention like communityfacility linkage to succeed (which are also among the attributes in Rosato's framework).While one could consider the importance of these attributes only among community lessons learned from these papers show how valuable it is for these same characteristics to be present in healthcare providers as well.Other important factors for successful interventions pertain to the intervention setting in which the community of interest lives and acts: socio-economic dynamics, including gender disparities, power distribution/political hierarchies, external factors like transportation means/road security to materially link the community and its local health facility, any of these factors can influence these intervention's outcomes and must be considered in future implementations.In light of these reflections, it becomes clear how tailoring the implementation to the actual characteristics of the context and the needs of the recipient community is vital to obtain equality and sustainable transformation over time-echoing the calls from the literature for inclusion of local communities' voices in every step of development interventions as a way to achieve longlasting improvements (Campbell and Jovchelovitch, 2000;Andersson, 2011;Petiwala et al., 2021).

Discussion
In this review, we assessed past interventions linking communities and local facilities to understand which features can influence their successes.Overall, all the studies reported positive results, from improved care-seeking behaviour and healthcare access for community members, to increased expertise and better practices for healthcare providers.Community mobilization, measured in terms of discrete health practices rather than wider shifts in norms, showed positive changes but proved that some of these interventions' designs were incapable of tackling socio-economic barriers.Health facilities' service provision improved generally, but some issues were flagged, from inadequate funding to sustain such changes to new services being offered that did not adapt to the local needs.Finally, one peculiar finding was that impacts on child mortality rates were not properly assessed by any of the papers.
In terms of design and delivery conditioning aspects, lack of awareness of the intervention among local community members often led to only partial achievement of changes.For this type of intervention to be effective, active participation and engagement should be put in place, as without high community involvement long-lasting and effective improvements are unlikely to be obtained.Moreover, hard-to-reach households need to be taken into account more carefully.The authors of papers in which the exclusion (even if involuntary) of hard-toreach communities led to only partial success of the projects stressed the need to increase efforts to include entire populations more actively in future implementations, to obtain sustainable and equitable changes.This point feeds into a bigger discussion on the need to include equity impact routinely in any study design and implementation; only when it becomes the norm, the risks of unintended effects like the ones reported here start to reduce (Morrison et al., 2019;Spencer et al., 2019).When considering the need to be as inclusive as possible, then, involvement of fathers in a type of intervention generally delivered to the children's mothers becomes another key point (Tully et al., 2017).As they are usually the heads of the family, responsible for any financial and health-related decision, and there is evidence that their presence in such studies can improve child health, it would be advisable to routinely include them in any implementation.When fathers have been included, their interest and involvement in child health have increased (Tancred et al., 2018).Moreover, their exclusion could potentially harm the attendance to care of women and children, given the aforementioned decision-making role of men in the family (Kushitor et al., 2019).
Another point of discussion is the importance to match health interventions with financial plans, as a lack of funding can hinder the sustainability of the changes, and one could incur in the risk to be unable to meet a demand increase (i.e.vaccines requests after an education programme) with an adequate supply provision (sufficient number of vaccine doses in the local facilities).Since a failure to satisfy these standards would undermine the whole purpose of a linkage intervention, it becomes clear the need to work on all sides of an implementation.
A further issue highlighted by our analysis of Kushitor et al. (Kushitor et al., 2019) is the challenge of maintaining the same structure for the intervention over time.A programme that was supposed to be carried over by the community became   For what concerns outcomes of interest, then, it was fascinating to observe that changes in mortality rates are practically not used to measure changes obtained by this type of intervention.Determining differences in such variable, though, could be extremely impactful, especially when considering implementations in areas with high child mortality rates.It might be worth exploring the feasibility of including it among the indicators of future implementations, whilst being conscious of the challenges it would entail-i.e.larger samples needed.
In terms of study methodology, we believe that to better understand the real effect of this type of intervention, mixed methods approaches are desirable.Outcomes like the increase in immunization rates for girls in Gupta et al.'s paper would have just appeared positive with an exclusive quantitative approach (Gupta et al., 2017).The qualitative data collected allowed the authors to understand that the health education component was not as successful, with the local population still afraid of vaccines.Plus, it helped to unveil gender disparity issues, awareness of which could inform future implementations.A thorough formative work with communities could have probably detected (and better addressed in the implementation phase) this issue, confirming once more the importance of carefully studying and including local populations' voices in the design and delivery of an intervention (Trickett et al., 2011).Some limitations must be acknowledged in relation to the studies included in this review.In regard to the data collection process, a few studies acknowledged the possibility of recall and selection bias (Waiswa et al., 2015;Cannon et al., 2017;Pol et al., 2017;Var et al., 2020), some potential issues with the quality of data (i.e.self-reported) (Cannon et al., 2017;Kamugisha et al., 2018;Kushitor et al., 2019;Var et al., 2020) and with confounding factors (other development projects undergoing in the same study area) (Olayo et al., 2014;Gupta et al., 2017).For what concerns the data analysis, some papers recognized the lack of control groups (Gupta et al., 2017;Kamugisha et al., 2018;Oguntunde et al., 2018), small sample size (Waiswa et al., 2015;Tancred et al., 2018;Awasthi et al., 2019;Ndaba et al., 2020;Var et al., 2020) and early stage analysis (i.e.before study completion) (Björkman and Svensson, 2009;Pol et al., 2017;Ndaba et al., 2020) as limitations to their studies.
With respect to our application of Rosato's framework, it is not surprising that the variables of Health, Goal and Tools have generally good scores in this type of intervention.'Conceptualization of health' as more than the absence of disease and disorders (score I: Medical model); 'Goal' including health knowledge and behaviour change as well as the eradication of specific health problems; and 'Tools' including behaviourchange and social capacity-building methods rather than simply clinical ones.All these are prerequisites of a communityfacility linkage intervention.For variables like 'Participation', 'External agent role' and 'Community role', it was in general more challenging to attribute the highest scores.Leaving the intervention decision-making process entirely in the hands of an auto-sufficient community functioning as the main active agent of change, with the external representatives acting only as an additional support resource, is the most challenging part to achieve.Analysing the papers via this framework, though, has given more evidence that when this effort is made (and community is really empowered and is the active protagonist of the intervention), better outcomes are achieved.That said, it must be noted that scoring the interventions based only on the condensed study description sections-and on the reviewers' subjective interpretations-could not render justice to them.One way to improve the reliability of the attributed scores would be to contact the study authors and have them grade their own work according to the integrated framework's guidelines-and compare the results.In any case, these scores are not to be interpreted in absolute terms, but just as a useful tool to guide the analysis of the community participation component of these studies, and to better understand what is needed for an intervention to become truly empowering and sustainable.
We used realist principles for a review of the literature, but we did not follow the meticulous structure of Realist synthesis (Rycroft-Malone et al., 2012).We appreciate that this could be perceived as unconventional.We agreed on this hybrid structure because we wanted to incorporate the explanatory aims of realist principles into the systematic review's evaluative nature, as we believed it could help to answer our research question in a more complete way.During the inception phase of the study, we decided to limit the search to academic publications, aiming to uphold the quality, reliability and rigour of our systematic review.However, we recognize that this choice does come with restrictions, and we acknowledge that there is a limitation in our choice to restrict our focus solely to peer-reviewed articles from academic databases.This is not advisable when performing a realist synthesis, which encourages the exploration of a broader range of sources, including reports and various forms of literature.
In addition to this, we note that our decision not to use checklists for risk of bias assessment could present a limitation to our study as it may have resulted in us not checking all aspects that may have led to bias in the included studies.However, we are confident in our ability to critically assess the papers included in the review, given our extensive experience with systematic and peer reviews.
One further limitation is associated with our decision to include only studies written in English, primarily due to restricted translation capabilities, which could potentially introduce language bias into our findings.However, it is worth highlighting that we did not encounter any non-English publications during our search.
Our analysis of factors that contributed to the success or limitations of these studies has highlighted the following key lessons: the need for more active inclusion of hard-toreach populations, preference for mixed-methods methodology, involvement of community members in programmatic decisions and a focus on the actual context need.This illuminates the importance of transformative considerations within evaluation plans, and Figure 3 suggests a potential evaluation framework that can be used to assess the impact of community-based and facility link interventions.This review offered an opportunity to reflect on this type of intervention, which, despite its potential being recognized at the international level, is still not particularly diffused.We provided an angle to analyse it, including a community empowerment integrated framework and application of realist principles, all trying to shed light on key features for the success of community-facility linkage interventions.We hope this will be helpful to influence other implementations and serve as a guide for future evaluation plans.

Figure 2 .
Figure 2. Spider diagrams to describe community participation

Figure 3 .
Figure 3. New framework based on studies MRTs

Table 2 .
Studies' key results