No data on community planning or initiatives at all.

Twenty-one of the 25 studies did not make any reference to the community as driving or motivating the CLHW program. Twenty-two articles provide no information on the issue of community participation in program planning. In some cases this seems to be because partnerships at national or global levels drive the study initiative, and in others—such as the Pakistani “Lady health workers program”—despite emphasis on the notion that the program is decentralized (see also a related review article detailing the same program). In these articles there is no mention that these programs may have emerged from the communities themselves, or with community input. This includes six studies of CLHW programs that were devolved to local NGOs. A few studies test effectiveness of specific techniques or instruments (e.g. rapid diagnostic tests) and are typically initiated by research-oriented partnerships without consideration of community inputs or initiatives [37,45–48] (with one exception: the evaluation of a CLHW programme in Siaya district, Kenya by Rowe et al., who assessed the influence of health committees, soliciting the opinions of women in villages on guideline adherence [38]).

Too few details to be useful.

Some studies mention the role of the community more explicitly, but provide too little detail to be useful for further analysis. Matthews et al. note the use of a participatory evaluation method in South Africa that motivated CLHWs to participate in the design of the evaluation of their work [49]. Gazi et al. note that in order to perform their tasks effectively, depot-holders (described as women from the community who promote good health practice and use of clinics and who keep a stock of contraceptives and oral rehydration salts) require the support of their families and of the community in general, and they report that most (over 80%) receive such support [43]. However, although they report that in the preliminary stages the NGOs informed local landlords, politicians, club members and local leaders about their activities to overcome problems with local gangsters, they do not say how this impacted the program. Colvin et al. report on the contribution of traditional healers to a rural tuberculosis control program in Hlabisa South Africa [50]. Although initiated from the outside, this pilot program builds explicitly on the possibility of using an indigenous network of traditional healers, offering the option of traditional healers in one sub-district for directly observed treatment (DOT) supervision. However, the article does not report that traditional healers had any influence on how this program was organized; they were “selected and trained” without much detail on how well the traditional healers themselves were integrated in their respective communities.

Community input acknowledged and (to some extent) described.

Five studies stand out as acknowledging, in greater detail, that the programs in question were developed as part of community initiatives [31,39–42]. Rennert describes a program in Honduras, which was initiated through a privately funded U.S. hospital-based group—The Brigade—and is, as such, a top-down initiative [39]. However, the small support structure of this internationally driven initiative became, through necessity, dependent on local organization, forcing community involvement during the setting up of the program. The authors describe how a community health committee was required by the project in both of the target communities, to oversee them locally, with CLHWs in leadership roles. In each community, the community health committees were given autonomy in the selection process. Finally, Debpuur et al. report on the Navrongo Project’s impact on contraceptive use and knowledge, fertility and reproductive preferences [42]. Funded through the Ghana Ministry of Health to address the crisis in rural human resources for health, the study uses general demographic survey data to compare the impact of a traditional, indigenous social cooperation, termed the zurugelu approach, with the mobilization of support for community health planning using clinic-based, community health nurses and community health aides. The authors note that the zurugelu approach involves health care action committees that include elders, traditional peer networks, and linkages between supervision and traditional self-help schemes. The health aides, or yezura zenna, are chosen by the community.

Little information on who the CLHW actually were.

Who are the community health workers in the studies? In about half of the studies testing for the impact of CLHW programs, we find little information on who the CLHW actually were, including their engagement with previous health roles and their ties to the communities. Fifteen studies provide little to no information about where CLHWs were recruited from or where they had been before their recruitment [37,42,43,45–49,51–58]. Where mentions are made, this is restricted to comments such as: “girls from the community having studied up to secondary school” [51]. Yansaneh et al. [59] and Kelly et al. [48] note in passing that CLHW volunteers were “selected by their communities”, but provide no further detail on how. Harvey et al. explain that most of the recruited CLHWs had preexisting experience with malaria treatment, but say nothing in the study about the background and previous community roles of the CLHWs other than a locality criterion: “All participating CLHWs lived in Chongwe or Chibombo District” [37]. This lack of information also characterizes the study by Debpuur et al. using the zurugelu approach [42]. Although the program includes indigenous networks, the article only notes that the CLHWs were recruited “using the traditional community system” and are called "yezura zenna".

Brief demographic descriptions.

Five papers note that CLHW were recruited in new roles and provide brief demographic characteristics [22,60–62]. For example, Baqui et al. discuss how recruits are women with at least a 10th grade education, a mean age at time of recruitment of 23 years, and more than 60% of whom are single, divorced or separated [61]. Hadi et al. note that volunteers were selected from among the local area, most had five years of schooling, and were generally middle-aged and poor women [22]. The authors argue that because the volunteers were a homogeneous group of less educated, married women, they did not expect age variation to have a bearing on performance. In this set of studies, it seems that the two defining characteristics consistently mentioned are place-based—the recruits are from the community—and demographic. The study by Colvin et al implicitly provides information about the history of the recruits, as they were all traditional healers, although the authors do not go to any length explaining anything else about these people’s community histories [50].

Acknowledgment of importance of previous experience, but no testing.

However, four studies do add information about community selection processes and inputs, and even to some extent about whether recruits have worked in community health roles before [38–40,50]. Teela et al. describe how all CLHWs had completed a minimum of 4 months training prior to the program, while making a commitment to work three years in their communities [40]. The CLHWs’ experience varied from reproductive to primary health care and had in all cases been at least two years. But none of these differences were tested. Adam et al. and Rennert et al. describe how the CLHWs recruited were respectively farmers [44], or either nuns or farmers, brought forward by a local committee organizing the program [39]. Although not much detail is given about the community relationship of the recruits, this is more detail than that provided in other studies. Another small group of studies focuses specifically on peer health workers, or people with previous experience of a specific disease as CLHW [41,63,64]. Here, previous experience is explicitly acknowledged, yet like most studies, very little additional information is available on these recruits and on their relationship with the community. In general, all of these studies restrict themselves to identification of the importance of recruitment in collaboration with communities, and acknowledgment that previous experience in health is desired, but with no testing of its impact on effectiveness. “Previous experience” remains largely unqualified but appears ideally to be clinical.

Previous experience focused on, with limited testing.

Two studies are exceptions to the general pattern [31,38]. Scott et al. focus specifically on community relationships, also using a qualitative methodology, including focus groups, observation and interviews. They note how policies in India mandate CLHW recruitment through community mobilization processes and community-based accountability, yet the CLHWs were often selected without community consultation and are seen as entirely accountable to the primary health care unit, where they also received their remuneration [31]. Another exception is, once more, the evaluation study by Rowe et al. in Siaya district, Kenya, which focuses particularly on commonly made errors in managing childhood illness. The authors note that some villages established health committees to select CLHWs, while obtaining the opinions of women in the village [38]. The authors explain that they requested women’s opinions because most of the patients’ caretakers, with whom CLHWs interact, are mothers. They argue that women would contribute ideas about the characteristics of CLHWs important for providing good care, although the authors do not explain what characteristics surfaced in these discussions. They do, however, test for the impact of this selection process empirically (see 3.2).

Little to no indication of community input in training.

Nearly all studies give very little indication, or none at all, of community input in the CLHW training process and curriculum. Jacob et al describe an assessment of diagnostic capacity for dementia that includes community participation [60]. They asked CLHWs to nominate people with dementia based on their knowledge of the local community, and asked them to obtain information from health workers and key informants living in the village in order to reach a conclusion on people who might be suffering from the condition. It is not clear how much this training included community relationships or input. Five studies explicitly describe a community relations element in CLHW training, designed to improve community engagement. This included mostly capacity building in interpersonal communication skills, behavioral change communication, community entry and diplomacy [40,42,61,65].

Input from the community taken into account qualitatively.

Only three studies took into account input from the community in training itself. Rennert et al. [39] mention that the Brigade developed a week-long training course specific to community needs, including a needs-based assessment for curriculum design, but no further detail is given. Adam et al. describe a community-based participatory process that was used to refine the spacing of the training schedule to accommodate the needs of the small-scale farmers volunteering to be CLHWs, and of their supportive community [44]. Finally, as mentioned above, there is the qualitative evaluation study of the Indian ASHA programme by Scott et al. [31]. Here, specific focus on community relationships illustrates the lack of community input, despite government policies mandating otherwise, eventually leading to mistrust between villagers and CLHWs.

Little to no information about ongoing relationships and connections.

Nearly all studies provide little to no information about the extent to which the CLHW program forged ongoing relationships and connections to community processes and dynamics to sustain motivation or create opportunities for input on program implementation by the community. Relationships are described in slightly more detail in eight studies [22,37,43,45,52,53,61,62,64]. For example, Gazi et al. dedicate a special section in their article to the interaction between depot holders and the community. They note that depot holders were generally valued for being community members, but leave out details on how the program intentionally engaged with the community. Ratimbasoa et al. note how in exchange for their participation, CLHWs received an annual allocation of rice as compensation for their work and help from the villagers to maintain their fields [45]. Overall, most of these studies lack detail on the way engagement proceeded, or how it contributed to decisions made during program implementation.

Structural constraints (e.g. distance, security) motivating engagement.

Three studies appear to be closely related due to constraints in the program environment [39,40,66]. The qualitative study by Rennert et al. has already been described. Here, program health committees were set up to implement the US-funded initiative in two distant communities without any major national organization overseeing the work. The committees included mayors, a member from a local water board, the director of the local orphanage, two community elders, the principal and teachers from local schools, and a local church leader. Because the Brigade chose not to intervene in the health worker selection process beyond the basic criteria outlined above, the health committees developed autonomy during the program, although few details are given as to what this meant. Similarly, distance plays a role in promoting community engagement in the program studied by Teela et al., who document community-based delivery of maternal care in conflict-affected areas of eastern Burma [40]. The multi-ethnic collaboration of local community-based maternal and child health care organizations seeks explicitly to build community trust and confidence because of the special challenges which the negative security environment poses to reaching programmatic goals.

Inherent programmatic focus on community engagement.

Only two programs studied appear to be motivated to develop community engagement to achieve programmatic gains. Chang et al. [41] detail a community-based program, including peer health educators (people living with HIV) that comprise about two thirds of the project staff. As a church initiative, the program involved extensive pre-existing community relationships which, according to the authors, helped to promote buy-in, follow-up, adherence, and dissemination of HIV care and prevention knowledge within the community. Agboatwalla & Akram report community support in the case of the Pakistani Health Education and Literacy Project (HELP) managed by a local NGO in the form of maintenance of activities, such as paying for CLHW services, and by monitoring of the project with little top-down supervision [51].

Of course mentioning community engagement does not in itself provide evidence that it is proceeding as envisaged. For example, Scott et al. observe that monthly Village Health Days or health planning sessions in which local people including CLHWs (ASHAs) are expected to participate in project planning do not occur as planned [31]. They document that if Village Health Days happen, these events feature nurses or local politicians lecturing the people about health issues and upcoming health events, such as immunization camps. The authors describe how the health system hierarchy limits opportunities for CLWHs to communicate up the chain of status and income.

Integration of community health resources does not matter so much.

With respect to effectiveness, most studies provide anecdotal evidence that the community relationship matters to program outcomes. Just three studies indicate indirectly that the quality of integration of community health resources in themselves do not matter much to outcomes [54,55,64]. Torpey et al. find that CLHWs are effective in adherence counseling despite only limited community interaction because of the mere practice of “following up”, regardless of the content of the interaction or the relationships built [64]. Similarly, White & Speizer find that mere CLHW outreach visits increase adherence to family planning methods [54]. Wilkingson & Davis find CLHW treatment to be effective without attention to community factors and attribute this success to decreased workload in health clinics [55].

Integration of community health resources matters.

A few studies stand out for indicating more explicitly that attention to traditional roles and networks improves program effectiveness. Chang et al. find CLHWs effective in providing AIDS care [41]. While its authors do not distinguish between the roles of peer health workers and nurse clinicians in assessment of effectiveness, the study was initiated out of a church-based community network with community-based planning and finds very strong results of treatment adherence in a low-income setting, equal to those in high-income countries [41]. The qualitative assessment by Rennert et al. [39] concludes that the CLHW program was effective, with recruits coming from documented pre-existing roles within the community, including farming and nursing. The authors point out that while CLHWs donated their time to the project, the communities would support them in return during periods of need, such as harvesting. Similarly, Ratsimbasoa et al. [45] report on a successful program where in exchange for their participation, CLHWs received food staples (oil, rice), or help from villagers to maintain fields along with an annual allocation of rice. Teela et al. [40] report that the MOM program overcame delays in care seeking; the authors attribute this to strong focus on building community trust. Using qualitative data, the authors explicitly address topics such as the project introduction to the community, relationships with community members and leaders, and collaboration with health workers and traditional birth attendants, suggesting that these issues are central to the success of the intervention. The authors argue that a more refined framework for achieving improved access within a community-based program should consider factors such as social norms surrounding care-seeking, perceptions of the seriousness of obstetric emergencies, gender and power relations, household-decision-making, and traditional practices. Despite this, substantial obstacles and challenges remain in the context of a militarily insecure environment.

Combining community health resources with biomedical health outreach.

In two other cases, the article by Debpuur et al. describing the zurugela approach [42] and the study by Colvin et al. [50], the contribution of traditional networks and roles is more specifically evaluated and even controlled for. Colvin et al. report that the outcomes of indigenous healers trained in the DOT program do not differ from other groups with respect to effectiveness, a positive result as expectations had been that they would be lower. They show that of 1,816 patients in Hlabisa District, there was no significant difference in treatment outcome comparing intervention and control areas (77% vs. 75%), or treatment completion (88% vs. 75%), while patients of traditional healers who had completed treatment revealed high levels of satisfaction with the care received. While Colvin et al. note distrust between some of the traditional healers and medically trained treatment staff, Debpuur et al. find that medical (nurse) and community-based indigenous health coalitions together provide the most effective results, a conclusion also reached by Dudley et al. [53]. Debpuur et al. find impact on fertility most pronounced when program outreach combines the involvement of nurses, traditional leaders and male volunteers. However, the authors note a lack of insight into the relative contribution of chiefs, elders, and social networks in reproductive health action mobilization, and suggest that more research is needed on this. While this is somewhat puzzling, considering that the study is focused on evaluating the effectiveness of an indigenous approach, this focus seems to have been more on male volunteers in family planning, and in this context, indigenous, traditional networks are defined as those of male dominance. Thus, the zurugela approach seems to try to use existing roles and community structures to strengthen community health by changing existing roles and structures

Relevance of remuneration observed, but no testing for other motivators.

Nine studies provide information about motivation, but do not connect this to retention or attrition [31,38,39,45,50,59,63,65,67]. With regard to attrition, the discourse in some studies suggests a particularly high relevance of remuneration [31,62,64,67]. Rowe et al. find that consultations performed by CLHWs who thought that they received four or more benefits had higher levels of overall patient adherence than consultations performed by CLHWs who thought that they received fewer than four benefits [38]. Dudley et al. note that although treatment supporters received R30 (US$4) per month per patient, the funding was limited and the sustainability of such projects was of serious concern to health managers and communities [53]. While strong views such as this are often expressed regarding the influence of remuneration in retention, this notion exists without insights into the complementary relevance of other non-monetary benefits and relationships. Three studies clearly note the importance of community recognition and social standing [31,50,59]. Aside from these factors, Jennings et al. explain significantly higher performance for communication on general prenatal care by lay nurse aides relative to midwives in a program in Benin [65]. They argue that this arises from recognition in the clinical field (by superiors), from the opportunity to be more involved in patient care, and from satisfaction with an expansion of professional competencies through capacity building.

Attrition data exist, but community relationships not investigated.

Eight studies provide actual data on attrition [22,40,43,44,52,61,62,64], which ranges from 10% to 50% or “high”. High attrition—when discussed—is related to life events, a mismatch of expectations, or lack of supervision and monitoring. Baqui et al. find a 50% attrition rate but do not further discuss any of the causes for this, while Hadi finds “high” attrition in the ARI program and attributes this to a lack of adequate training, monitoring and supervision. Neither study describes community-based causes for attrition. In their study of depot-holders, Gazi et al. also find a 53% attrition rate, and attribute this to CLHWs finding other jobs, marriage, out-migration, sickness, greater volume of work than expected, and unsatisfactory performance in the eyes of the supervising institute. They note that many depot-holders considered the earnings insufficient and that some reported that they were embarrassed that their earnings were lower than those of housemaids. They detail additional problems in the community of Dhaka, which they attribute to alternative work opportunities and the general population’s greater involvement in cash economies. The authors do not further relate these identified causes for attrition to the noted importance of community dynamics, though they do describe the influence of "local conditions” and cite a number of factors influencing retention, including support from CLHWs’ families for their work (which seems high), a welcoming attitude towards CLHW activities, and the pride and social status expressed as a benefit by many CLHWs. At the low end of attrition, two studies ignore community relationships as possible causes for high retention rates. Frazao et al. find 13% attrition in a program promoting oral health [52], while Torpey et al describe a 9% attrition rate [64]. Both studies do note some community engagement during implementation, suggesting the importance of social interactions at family and community levels. Tulchinsky et al. find 10% attrition during ten years of operation which, as already noted, is related to a high stipend [62]. However, the authors also note that those who left did so for reasons “unrelated to the operations of the program” but do not specify what this means. Finally, the studies by Teela et al. [40] and Adams et al. [44] both find low attrition rates of around 20%, and both studies suggest a relatively high level of attention to community relationships during input and training/planning, and program implementation respectively. In both cases, causes for attrition are not specifically explained. Teela et al. do note negative security and logistical factors (distance, topography, weather) as severe program obstacles [40].

No longitudinal data, or descriptive indications.

Finally, with respect to sustainability, Debpuur et al. note how continuation of contraceptive use is vulnerable to even minor or temporary lapses in program intensity, as women readily abandon contraception if program support is disrupted [42]. From this perspective it is striking that at least sixteen studies provide no longitudinal data or descriptive indications of sustainability [37,45,46,48,49,52,54,55,58–61,63–65]. Soares et al. note that drug and gang violence driven by external factors jeopardized sustainability of the favela program [57]. Agboatwalle et al. claim that their program has become self-sustaining as a result of the integration of social (e.g. literacy) and health activities, indicating some relevance of community relationships at macro-level, yet they document only one year between pre- and post-intervention [51].

Importance of trust between community and health system.

Some studies indicate the importance to sustainability of the relationship between supervisors and CLHWs, where CLHWs are positioned as members of communities with local connections. For example, Gazi et al. note that retention (and performance) of depot-holders depends on the extent to which supervisors value how much CLHWs know their locality and how they see this linked to increases in the number of users [43]. Colvin et al. note anecdotally how this very relationship, as it was to some extent characterized by distrust between traditional healers and medically trained professionals, could become an impediment to program sustainability [50]. A few studies stand out by virtue of the sustainability illustrated [31,40,41,44,53,62]. Chang et al. note that the alternative AIDS care program in Uganda was effective over at least a 2-year period, and attribute this sustainability explicitly to extensive community relationships, including community-based program planning. Teela et al. and Tulchinsky et al. make similar claims [40,62]. Scott & Shanker argue that it is precisely the lack of community participation, which, together with outcome-based remuneration structure, poor institutional support, and a rigid hierarchical structure of the health system, challenged the long-term sustainability of the ASHA program [31]. Considering the complex influences of, for example, political patronage in resource constrained settings, or patriarchal values influencing the care burden of the often female CLHW, it seems that accurate evaluations of CHLW programs in low-income countries still have a long way to go.