Understanding the capacity of community-based groups to mobilise and engage in social action for health: Results from Avahan

ABSTRACT Community mobilisation improves outcomes from HIV to maternal and child health. Yet, little health research has explored why some community groups are better able to mobilise than others. We address this gap by considering the case of Avahan, the India AIDS Initiative, which sought to foster community mobilisation, including the creation of community-based groups serving men who have sex with men (MSM), female sex workers (FSWs), and injection drug users (IDUs). Using quantitative and qualitative data collected from 58 community-based groups from 2009–2012 across six Indian states, we analyse variation in groups’ action on behalf of their members. Based on a mixed effects logistic regression, we find that older groups and those with bank accounts, crisis committees, or strategic relationships were most likely to take action on behalf of members by demanding rights or confronting gatekeepers and opinion leaders. Analysis of qualitative data reveals the types of action organisations took on behalf of members (mediation, removal of community members from harm, and advocacy), but also that sometimes organisations refused to take action, or community members declined their assistance. These findings indicate that organisations formalising, creating structures for social action, and building networks are important strategies to foster community mobilisation.


Introduction
Community mobilisation strategiesexplicit efforts to 'engage and galvanise community members to take action towards achieving a common goal' (Lippman et al., 2013, p. 2)have been shown to improve a number of different health outcomes across settings, including HIV prevention (e.g., Beck et al., 2018;Cornish et al., 2014). The Avahan AIDS initiative, a Bill and Melinda Gates Foundation funded programme running from 2003-14 in six high-burden states in India, relied on community mobilisation as a a core strategy to reduce HIV transmission and develop organisations to take over programme implementation after the Foundation's exit. Community groups representing female sex workers (FSW), men who have sex with men (MSM), injection drug users (IDU), and transgender individuals were formed to lead community mobilisation, and carried out activities that included peer education, engagement with police, media, and other local power holders, and community building (Avahan, 2008;Biradavolu et al., 2015).
Research has shown community mobilisation, which frequently includes the creation and strengthening of support organisations, to be effective in HIV prevention efforts by altering the social environment and increasing social cohesion (Cornish et al., 2014;Kerrigan et al., 2013;Kerrigan et al., 2019;Lippman et al., 2012;Pettifor et al., 2018). Specific research has linked exposure to community groups with reduced sexually transmitted infections (Yadav et al., 2013). Despite the HIV-related benefits of community mobilisation, we know very little about why some organisations are better able than others to mobilise communities. In the analysis below, we use data from the experience of the Avahan Initiative to assess the ability of community-based groups created as part of community mobilisation efforts to take action on behalf of their FSW, MSM, and IDU members.

Community mobilisation and health
A substantial body of literature demonstrates the positive effect of community mobilisation on health outcomes. Generally speaking, community mobilisation improves health outcomes through pathways of greater knowledge and collective efficacy that increase usage of condoms and health services, as well as demands for greater rights (Lippman et al., 2013). For example, the Avahan Initiative's reliance on community mobilisation facilitated reductions in HIV prevalence and other sexually transmitted infections (Kerrigan et al., 2015). Consistent with these findings, many studies of community mobilisation efforts among FSWs associated with the Avahan project have found that increased collectivisation enhances health awareness and leads to consistent condom use with clients (Blankenship et al., 2008;Glassman & Temin, 2016;Halli et al., 2006;Kuhlmann et al., 2014;Ramesh et al., 2010;Reza-Paul et al., 2008;Saggurti et al., 2013). Moreover, collectivisation through organisations can improve chances of identifying HIV due to higher rates of HIV testing (Beattie et al., 2014;Zhang et al., 2013). Other studies have examined the mechanisms linking mobilisation to improved health outcomes, finding that perceived community support coupled with increased knowledge resulting from sensitisation activities enhanced community members' ability to negotiate condom use, increased their sense of self-efficacy, and enabled them to pursue treatment in public and private health facilities (Beattie et al., 2014;Glassman & Temin, 2016;Gurnani et al., 2011;Reza-Paul et al., 2008;Saggurti et al., 2013;Yadav et al., 2013). Community mobilisation, or collectivisation, can take many shapes and forms and does not have to be institutionalised to reduce HIV incidence. For instance, Mantsios et al. (2018) found that female sex workers' participation in a community savings group, one form of non-institutionalised collectivisation, was associated with increased consistent condom use in Tanzania. Community mobilisation has other positive effects that may indirectly influence health outcomes through greater efficacy. Scholars have shown that greater exposure to community mobilisation was associated with increased self-esteem and ability to speak out against violence and stigma (Argento et al., 2011). With heightened perception of collective power against violence and stigma (Biradavolu et al., 2009;Dixon et al., 2012;Kuhlmann et al., 2014), community members have experienced less violence at the hands of police (Argento et al., 2011;Punyam et al., 2012). Community-based organisations can contribute to the process of community mobilisation by addressing crises as well as by carrying out advocacy. For example, Narayanan et al.'s (2015) work demonstrated how members of an organisation of transgender persons successfully leveraged their shared identity to negotiate with government officials for land for housing, despite opposition from the broader community.
Community mobilisation may also increase financial security. For instance, a study by Patel et al. (2018) found that FSWs with higher levels of collective efficacy and collective agency were significantly less likely to report financial vulnerabilities. Moreover, FSWs who were members of organisations had greater financial security and a better sense of social protection than those who did not belong to organisations (Patel et al., 2018). Community mobilisation also allows members to become less dependent on exploitative sources of income as it can create platforms for the initiation of micro-credit and savings schemes Blanchard et al., 2013).
The benefits of community mobilisation notwithstanding, implementation challenges abound. A systematic review of articles published on community-based interventions in eight projects across three countries, including Avahan in India, demonstrated the hindrances presented by lack of consistent funding sources, regressive legal frameworks, and the convergence of social stigma, marginalisation, and the exclusion of vulnerable groups (Kerrigan et al., 2015). George et al. (2014) found that the transfer of the Avahan Initiative to the Indian government reduced the emphasis on community mobilisation. The frequent expectation that organisation members work on a voluntary basis further challenges mobilisation efforts, as was observed in a community-based intervention in Tanzania (Boesten, 2011). If governments or donors designate community-based organisations as the sole way to access services, they may exclude non-members and/or tensions may emerge in the community, and moving away from advocacy leaves organisations at risk of government cooptation (Lange, 2008).

Capacity of community-based organisations
For organisations to form an effective part of collective mobilisation, they need to be capable of acting on behalf of their members. The broader literature on nonprofits, organisations, and nongovernmental organisations (NGOs) provides some clues as to characteristics likely to increase this capacity for social action, including age, resources, and leadership and leadership structures.
Although newer organisations tend to be more flexible with less rigid agendas than older organisations, they are more likely to fail due to limited resources (Chambré & Fatt, 2002). Limited resources often are the Achilles heel of nonprofit organisations. Not only are organisations with better access to resources more likely to survive, but smaller organisations that struggle to mobilise resources are likely to remain small (Burger & Owens, 2013). Donors are extremely important in the founding and sustenance of nonprofits in low-and middle-income countries (Appe & Pallas, 2018). On the flipside, intense competition for resources diminishes coalition-buildinga key catalyst for community mobilisation (Lippman et al., 2013).
Of course, increased reliance on donors may affect an organisation's policy actions (AbouAssi, 2014) and reliance on government funding in particular reduces the ability of organisations to hold government accountable. Professionalisation is key to the survival of nonprofits, but can come at a cost of disconnecting from the community (Lippman et al., 2013). The term 'NGO-isation' describes a process in which civil society actors de-politicise, professionalise, and disconnect from grassroots communities (Choudry & Kapoor, 2013). In this process, organisations stop looking like 'civil society' actors, inherently politicised and informal actors rooted in a 'grassroots community'. Oftentimes, this process is explained through an increased reliance on aid, or cooptation by government (Atia & Herrold, 2018). However, organisations can decide to 'de-NGOise' when it is deemed strategic for mobilising support from the community (Shrestha & Adhikari, 2010).
Strong and creative leaders with supplementary income that enables them to dedicate time to their organisations have been key to the success of community-led AIDS activism (Boesten, 2011). Inexperienced leadership may also explain the inability to mobilise resources (Boesten, 2011). Within the Avahan programme, studies have confirmed significant heterogeneity in the capacity of community-based organisations, with some performing better in advocacy and governance and others excelling in resource mobilisation and programme management (Chakravarthy et al., 2012;Narayanan et al., 2012). But there is as of yet sparse literature explaining variation in the capacity of community-based organisations to engage in the types of social action the literature has shown to improve the health of their members.
In the analysis that follows, we explore the drivers of variation in community-based groups' actions on behalf of their members. We define actions on behalf of members as demanding rights or confronting gatekeepers and opinion leaders. We hypothesise that organisations with greater capacity (older, more resources), as well as established mechanisms for social action (advocacy and crisis committees, partnerships with outside groups) will be more likely to take action on behalf of their members. Using qualitative data, we then identify features of cases where organisations did and did not take action on behalf of their members.

Background
India's HIV epidemic is concentrated in specific key populations, with overall prevalence approximately 0.2% in 2017 (UNAIDS, 2020). Despite the low prevalence, India's large population means there are close to two million HIV-infected people (UNAIDS, 2020). As of 2019, the Indian government funded the HIV response almost entirely (UNAIDS, 2019), but only slightly more than half of HIV-infected individuals received treatment (UNAIDS, 2020). As in all contexts, HIV rates among key populations are higher than among the general population. Sex work is technically legal in India but stigmatised, and national HIV prevalence among sex workers is estimated to be 1.6% (UNAIDS, 2019). Same-sex sex is illegal, and HIV prevalence is estimated to be 2.7% among MSM (UNAIDS, 2019). Among those who inject drugs, estimated HIV prevalence is 6.3% (UNAIDS, 2019).
Avahan's goal was to reduce HIV transmission through addressing the proximate and distal determinants of HIV. Proximate determinants included the presence of sexually transmitted infections, condom use, and type of partner (Avahan, 2008). The initiative addressed distal determinants such as stigma, legal environment, medical infrastructure, and gender rolesthrough structural interventions, including community mobilisation (Avahan, 2008). Community mobilisation goals included: (1) developing cohesion and shared identity among marginalised individuals; (2) building these individuals' capacity to manage HIV prevention programming; and (3) empowering them to define and implement their own agendas for change as a community (Narayanan et al., 2012). Community mobilisation began with paid peer educators, then evolved to community based groups, intended to be responsible for advocacy, addressing violence, and access to entitlements following the Indian government takeover of Avahan at the end of the Gates Foundation's involvement (Narayanan et al., 2012). A large NGO managed implementation of the programme in each state. 1 India has a long history of social organising, including many self-help groups and savings and credit organisations (Gugerty et al., 2019) that grew out of the Gandhian philosophy of self-reliance popularised during the independence struggles. India is also a democracy, but government efforts to control the activities of NGOs have increased over time, especially through increasing restrictions on receipt of foreign funding (Bornstein & Sharma, 2016).

Sample and data description
The data for this analysis come from Praxis-Institute for Participatory Practices (Praxis for short), an Indian nonprofit organisation focused on participatory research and capacity building. The Avahan Initiative contracted Praxis to design and implement a framework to monitor and 'nurture' Avahan's community mobilisation efforts across all six states taking part in the initiative. To do so, Praxis developed the Self-Administrable Tools to Assess Community Preparedness for Vulnerability Reduction (PRAXIS Institute for Participatory Practices, 2013). The tool simultaneously assesses/monitors and facilitates progress towards critical consciousness and independence of community based groups. Details on the tool, broadly applicable to a number of different settings, are available in an online publication (PRAXIS Institute for Participatory Practices, 2013) 2 and the data produced by application of the tool are available through Harvard's Dataverse.
Across four years (2009)(2010)(2011)(2012), there is information on a total of 58 organisations serving FSW (24), MSM (14), and IDU (20) across all six states: 14 in Andhra Pradesh, 12 each in Nagaland and Maharashtra, 10 in Karnataka, 6 in Tamil Nadu, and 4 in Manipur. 3 The organisations in Manipur and Nagaland almost all serve IDUs; the other states have organisations serving more than one key population.
The qualitative analysis relied on two types of data. First, data on challenges faced during election processes as well as general management issues collected from office bearers 4 provided information on both connectedness to the community and organisational capacity. Second, data on crises faced by community members provided insights on when organisations did, and did not, take action on behalf of their members and came from community members' and organisational committees' reports of three distinct crises faced by community members. Across all four years of data collection, 41 community-based groups provided information on 236 crises and on 94 challenges in management or in election processes. The analysis includes data from the 120 crisis descriptions that mention the community-based group's role.
When examining social action outcomes, we use both quantitative and qualitative data from all four years.

Quantitative measures
We consider three sets of variables along a continuum towards action on behalf of members. Table 1 shows descriptive statistics for these variables, including how they vary by population served (FSW, MSM, or IDU). The first set of variables is characteristics of organisations the literature associates with survival and capacity. The second set of variables refers to factors that should explicitly facilitate organisational action on behalf of members. The third set of variables, our main outcome variables, relates specifically to actions taken by organisations on behalf of their members.
Capacity for social action. We define capacity for social action as characteristics of organisations associated with survival and capacity. Indicators include (1) age of the organisation (whether founded before 2006, yes/no);(2) financial resources in 2008 (whether organisation's resources in rupees fall above 75th percentile for all organisations, yes/no); (3) the number of members in the organisation (continuous variable); and (4) whether the organisation has a bank account that they manage (yes/no).
Mechanisms for social action. We define mechanisms for social action as the organisational structures and connections which enable social action. Indicators include (1) whether the organisation has an 'advocacy committee' (yes/no); (2) whether the organisation has a 'crisis committee' (yes/ no); (3) whether the organisation has any partnerships with advocates or unions (yes/no); (4) whether the organisation includes community members in electing or nominating managers, key staff, or member representatives and/or whether community members participate in regional/ state level meetings (yes/no). Social action on behalf of members. These are the main outcome variables. Indicators include: (1) whether the organisation demanded rights of members in cases of arrest, violation of property rights, and discrimination (yes/no); (2) whether the organisation pressured opinionmakers like faith leaders, the neighbourhood community, politicians, and the media on an issue related to the interest or health of a community member (yes/no); (3) whether the organisation pressured gatekeepers, such as goondas, 5 pimps, brothel owners, and police, on an issue related to the interest or health of a community member (yes/no).

Qualitative measures
Within the qualitative data, we strove to identify the management challenges experienced by organisations, the crises faced by community members, and the actions taken by community-based groups in response to those crises. We coded management challenges based on whether they stemmed from weak community connectedness and low trust or weak teachnical capacity. We coded descriptions of organisational responses to crises faced by community members across five broad categories: (1) the organisation took action; (2) the organisation refused to act; (3) community members did not want organisational involvement; (4) there was no suitable role for the organisation; (5) the organisation did nothing.

Analysis
The quantitative analysis consists first of bivariate comparisons between organisational capacity and population served (FSW, MSM, or IDU) and the second and third groups of variables: mechanisms for social action and actual social action. We then present a mixed effects multivariate panel regression explaining variation in social action on behalf of members over the four years of available data. Using a random intercept model, we clustered panel data by year and state to account for nonindependence of observations across time and geographic context. Given the relatively small sample size (∼135 organisation years), in addition to population served, we parsimoniously selected variables for the regression based on significant bivariate relationships and sufficient sample size. For organisational capacity we include age, bank account status, and resources. For mechanisms for social action, we include existence of crisis committees, as well as strategic relations with advocates or unions.
The qualitative analysis is primarily descriptive and identifies some of the many factors driving the results observed in the quantitative analysis. These include management challenges facing organisations as well as the reasons why organisations did not support community members, or why community members refused organisational help. Quotes are referenced by the population served, location, and an organisational identification number.

Organisational capacity
As shown in Table 1, about half of organisations served FSWs, 20% served MSM, and a third served IDUs. Most organisations formed through the Avahan process, or in the context of it: 80% were founded in 2006 or later. On average, groups serving MSM were oldest, followed by those serving FSWs. The two oldest groups formed in the 1990sone for FSW in 1996, and another for MSM in 1997but the oldest IDU group was not founded until 2006.
Organisations varied in their membership size and resources. FSW groups had the most members, followed by MSM, and then IDU groups. These figures at least partially reflect differences in the size of each key population (certainly for FSWs compared to the other two groups), and membership size was associated with resource availability. In general, FSW and older organisations were better resourced: FSW groups had three to four times the volume of resources as MSM groups, and older organisations had three to nine times more resources as newer organisations. In 2011, only 40% of the groups had a bank account they managed. More MSM and older organisations managed their own bank accounts.
These different measures paint a picture of moderate organisational capacity, as defined by age, membership, and resources, with a good deal of variability among organisations.

Mechanisms for social action
Whereas less than half of organisations (43%) in 2011 had advocacy committees, two-thirds (67%) had crisis committees. FSW organisations were most likely to have advocacy committees (50%), while MSM organisations were most likely to have crisis committees (79%). A greater percentage of older organisations had crisis committees whereas more newer organisations had advocacy committees. MSM-serving groups were part of more strategic relationships with either advocates or unions than groups serving other populations. More FSW-serving groups involved their community in selecting leadership/staff (85%), followed by MSM (71%) and then IDU groups (50%).
Taken together, these findings indicate that about half of groups had mechanisms in place for social action, with important differences by key population served.

Action on behalf of members
As shown in the bivariate results in Table 2, about three quarters of groups demanded rights on behalf of their members in 2011, whereas closer to half played a lead role with gatekeepers or with opinion makers. While FSW-serving groups were most likely to demand rights and take a lead role with opinion makers, MSM-serving groups were most likely to take a lead role with gatekeepers. IDU-serving groups were the least likely to take action across all three measures. Organisational capacity had a relationship with social action. Older organisations were significantly more likely to take a lead role with gatekeepers (100% of organsations formed before 2006 vs. 35% formed later). On average, organisations that played a lead role with opinion makers had around twice as many members compared to those that did not. They also had far more resources in 2008. Organisations that managed their own bank accounts were more likely to play a lead role with gatekeepers.
In terms of mechanisms for social action, organisations with crisis committees were more likely to both demand rights on behalf of their members as well as take lead roles with gatekeepers, and organisations that had strategic relations with advocates or trade unions were more likely to take a lead role with both gatekeepers and opinion makers. Organisations that involved the community in these processes were more likely to take a lead role with gatekeepers As the multivariate regressions in Table 3 show, older organisations (founded before 2006) had over 12 times greater odds of taking a lead role with gatekeepers, and nearly four times greater odds of taking a lead role with opinion makers, as compared to newer organisations. IDU-and MSMserving groups had 84% and 66% lower odds, respectively, of demanding rights on behalf of their members compared to FSW-serving groups. Organisations which managed their own bank accounts were substantially more likely to demand rights on behalf of members and interact with gatekeepers. Well-resourced organisations had over six times greater odds of interacting with opinion makers. Organisations with crisis committees had nearly five times greater odds of interacting with gatekeepers. Finally, organisations that had strategic relations with advocates and/or unions had significantly higher odds of demanding rights on behalf of members as well as taking a lead role with gatekeepers.
To better understand the factors related to staff and organisational structure that might limit organisations' ability for social action on behalf of their members, we turned to the qualitative data. Many staff shared the concern that organisations were becoming disconnected from the community. A member of a FSW group in Karnataka commented that community members ''are unhappy with non-community members being recruited as staff and feel that they do not have the space to recruit members or express what they feel' (FSW, Karnataka, #62153113). Members of another group shared a similar feeling of disconnect when they reflected that ''pantshirt' (professionalised) and 'Sariwala' (community) MSM did not trust one other and that pantshirt MSM would seek their own interests above those of community members if in leadership positions (MSM, Maharashtra, #61152613). In other cases, professionalisation meant increased hiring standards (educational qualifications and professional experience), rendering some staff positions unattainable to the average community member.

Cases where community-based groups took action on behalf of community members
The qualitative data also provided insights into the exact types of social action community-based groups took on behalf of their members. These included: (1) mediation with family, community, police, or hospitals; (2) removal of key population members from harm or heightened risk; or (3) advocacy and protest. The most common type of intervention was mediation, in which groups supported key population members in crisis through both formal and informal mechanisms. Formal mechanisms mainly included reporting cases to the police. For instance, a woman approached a community-based group for help when her HIV-positive husband insisted on having unprotected sex. The group took the husband to the police station and filed a complaint (FSW, Karnataka, 351514211). More generally, community-based groups went to the police in cases where community members were subject to domestic violence, or harassment from broader community members or family. Organisations also helped bail out community members when arrested without just cause by the police. In some cases, organisations resorted to legal action against police officers when police refused to drop charges or when there was unjust use of force. In contrast, informal mechanisms included mediation with family, other loved ones, or broader community members. In one case, family members threw a man out of his home after they discovered his sexual orientation. A community-based group intervened by hosting the man in their office for two months, and group members paid several visits to the member's father to convince him to allow his son back into the house, until he eventually agreed (MSM, Maharashtra, 522132011).
Community-based groups also helped create safe environments for members. For instance, one woman approached a group for support in finding an alternative source of income. The group linked her to a government micro-loan programme, and she was later able to open and manage a petty shop (FSW, Tamil Nadu, 38121314). In another case, the same community-based group provided financial support to an HIV-positive woman and her children after her parents and husband abandoned them (FSW, Tamil Nadu, 38121314). In more rare scenarios, the group's response to crises faced by community members was to engage in advocacy or protest. For instance, following the killing of a patient by staff in a drug rehabilitation centre, a community-based group organised a protest rally, and in response, the rehabilitation centre compensated the family (IDU, Manipur, 691175213).
Cases where community-based groups did not take action on behalf of community members In a little less than half of cases, however, organisations did not take action to respond to crises faced by community members. Reasons for organisational non-response included refusal (roughly half of cases with no action), community member request for non-involvement (approximately one quarter of cases with no action), or lack of viable action to be taken (the remaining quarter of cases with no action). Cases where the community-based groups chose not to get involved were mostly when the organisation felt that the community members were 'immoral', or perceived that they were at fault. For instance, an organisation refused to support a woman who involved young girls in her home-based sex work (FSW, Tamil Nadu, 38121312). Another group refused to support a man who requested more money from a client than originally agreed upon (MSM, Maharashtra, 52132014). In some cases, the organisation perceived that the crisis was irrelevant to their mandate. Cases deemed irrelevant included disputes between family members or problems with government officials/police that did not relate to community members' identities as FSWs, MSMs, or IDUs.
Community-based groups also did not take action in cases where community members did not request their involvement, usually because the key population member or their family did not want involvement from the group or was simply unaware of what the group could do to help. These cases were most frequent among MSMs and FSWs. For example, a woman, whose work as a sex worker was unknown to her husband, refused organisational intervention following blackmail and repeated harassment by a client so as not to reveal her identity (FSW, Maharashtra, 651133614). In another case, a man approached a group for help after facing trouble with a client but soon requested that the group stop their involvement after the client threatened to disclose his identity if a report was filed with the police (MSM, Karnataka, 62157311).

Discussion
Using mixed methods and a comprehensive dataset on organisations in India, we found that older, richer organisations with bank accounts, crisis committees, and strategic relations with advocates or unions were more likely to take action on behalf of their members, including demanding rights and confronting gatekeepers and opionionmakers. These findings are in line with other literature on how addressing crises can foster resilience and lead to collective capacity to bring about change (Argento et al., 2011;Chakravarthy et al., 2012;Narayanan et al., 2015). Our study is different from other studies in that we used unique data on organisations to identify why some are better able to stand up on behalf of their members. Understanding more about the organisational contribution to community mobilisation points to concrete steps organisations and their funders can take to foster characteristics that facilitate mobilisation.
We found that FSW and MSM organisations had more capacity and were better resourced for social action than IDU organisations. FSW organisation's greater resources and members may well result from sex workers having income to pay fees and a large population of potential members. There is also a history of organising among sex workers in India, as exemplified by the Sonagachi Project (Jana et al., 2004). In contrast, MSM organisations' strengthsself-managed bank accounts and strategic relationshipsmay stem from advantages related to gender, as men in India have higher rates of financial inclusion (Ghosh & Vinod, 2017). Organisations supporting IDUs are likely at a relative disadvantage because drug use is illegal, and highly stigmatised (Sabri et al., 2017).
That older organisations with bank accounts are better able to interact with gatekeepers and opinionmakers confirms findings in the literature which point to an association between organisational age and general capacity (Chambré & Fatt, 2002). Moreover, the longer the organisation has been operating in a community increases exposure to community mobilisation, which has been shown to increase communities' willingness and ability to speak out against violence and stigma (Argento et al., 2011).
Our findings highlight that organisations which set up crisis committees and pursue relations with advocates or unions are better positioned to demand rights on behalf of members and interact with gatekeepers. The importance of crisis committees, as well as the effect of having strategic relations with advocates or unions, on demanding rights and interacting with gatekeepers sheds light on the importance of outreach and networking with key stakeholders for succesful social mobilisation (Biradavolu et al., 2009;Narayanan et al., 2015;Punyam et al., 2012).
This study has limitations. The data used in the analysis were not collected with the intention of answering the main research question, thereby limiting the scope of our analysis. In particular, we lack a great deal of information about the exact social, economic, and political context in which each community-based group operated. Some variables had missing observations, which may have introduced selection bias as community-based groups with more complete information may also have had greater capacity, thus overestimating the capacity of these groups overall. The community involvement variables in particular had too much missing data for extensive analysis. Further, the quality of qualitative data varied across organisations, precluding a comprehensive analysis of action on behalf of members parallell to that used with the quantitative data.
Our study also had several strengths. Our mixed methods allowed us to identify both the organisational characteristics associated with social action, as well as the nature of those actions, and also helped clarify some of the reasons why organisations might justifiably not take action on behalf of members. Because the data came from within the broader Avahan Initiative, the findings can be interpreted relative to what is, and is not, possible for a large-scale intervention to achieve vis-àvis organisational capacity.

Conclusion
Our findings indicate steps that organisations representing marginalised populations and the funders that support them can take to increase their capacity for social action, and in turn to increase collective efficacy, demands for greater rights, and ultimately the health of their members. While a new organisation cannot suddenly gain the benefits of age, implementing partner organisations and funders can help newer organisations overcome deficits by assisting them in setting up structures to collect nominal membership fees and independently manage their finances. Groups could also be encouraged to invest in mechanisms which facilitate social action, such as crisis committees and relations with advocates or unions.
Suggesting that community-based groups set up bank accounts and build up resources is equivalent to suggesting they professionalise, which comes with some risks, as both the literature review and qualitative analysis indicate. Specifically, the shift from 'sariwala' to 'pantwala' may decrease the ability and motivation of groups to engage in social action on behalf of their members as leadership becomes more distant from the community. Community organisations around the world must balance these two poles, and perhaps the Avahan-related organisations could benefit from participation in networks with more established organisations in India and beyond that could help them manage these tensions.
This analysis points to a number of areas warranting further research. Greater inclusion of context-specific variables referring to the neighbourhoods, cities and states in which the organisations are locatedsuch as the socioeconomic status, extent of civic space, and policies influencing key populationswould further enhance our understanding of what allows community mobilisation to improve health outcomes. Moreover, while we know that organisations with greater capacity and that have mechanisms for social action are more likely to act on behalf of members, more can be learned aboutthe process shaping these enabling characteristics, such as the role played by partner NGOs or by government officials in creating an enabling environment for community mobilisation. Comparing the programming of organisations more likely to take action to those that did not can inform donors and partner NGOs on types of programming they could support to facilitate mobilisation. These types of research will continue to strengthen our understanding of how organisations can best facilitate community mobilisation, and in turn improve health outcomes.
Notes 1. The implementing partners were: Family Health International and Pathfinder International (Maharashtra); Karnataka Health Promotion Trust (Maharashtra and Karnataka); the International HIV/AIDS Alliance and Hindustan Latex Family Planning Promotion Trust (Andhra Pradesh); the Tamil Nadu AIDS Initiative (Tamil Nadu); and Orchid (Manipur and Nagaland). 2. https://www.participatorymethods.org/resource/beneficiaries-agents-change-self-administrable-tools-assesscommunity-preparedness. 3. Although Avahan also targeted transgender people, the data from the Praxis tool on the Harvard Dataverse include no organisations speficially targeting this population, although qualitative data from the tool mention transgender people. 4. Office bearers include the president, vice president, treasurer, secretary, and other staff responsible for day-today operations of the organisation. 5. A term used to refer to hired thugs.