In line with Millennium Development Goal (MDG) 4, many countries are on track to reduce under-five mortality by two thirds from 1990 levels by 2015 [1]. This progress has not been uniform for all under-five age groups, however. Neonatal mortality has been relatively resistant to change and the 3.7 million neonatal deaths that occur annually worldwide account for an increasing proportion of all under-five deaths [2–5]. In Bangladesh, around 85% of births occur at home, 57% of all under-five deaths are in the first month of life and the neonatal mortality rate remains high at 37 per 1000 live births [6].

Tackling the burden of neonatal deaths, particularly deaths in the first twenty-four hours of life, requires community-based interventions to improve the supply and demand for maternal and neonatal health care and the use of safe home-delivery and newborn care practices that can prevent neonatal deaths [7–9]. Several studies provide encouraging evidence on how home visits or community mobilisation with concurrent health services strengthening can improve maternal and neonatal health in South Asia [10, 11]. Sustaining and scaling-up interventions to increase coverage remain critical challenges.

Low-cost, participatory, community-based approaches involving women's groups are effective at improving home delivery practices and birth outcomes in a range of settings. The women's group method significantly reduced neonatal mortality by 30% and 45% in Nepal and India, respectively, and improved hygienic home delivery practices and newborn care in Bangladesh, though it did not have an impact on neonatal mortality overall [7, 12, 13]. Intervention coverage was one women's group per 756 population and one per 468 population in Nepal and India, respectively [7]. The percentage of women who gave birth and reported attending women's groups was around 30% to 45% in India and 50% in Nepal [12, 13]. Coverage in Bangladesh was much lower at one women's group per 1414 population and the proportion of women who delivered and reported attending women's groups was 3%, with just 9% of women of reproductive age becoming women's group members [7].

Program coverage has been observed to be an independent determinant of neonatal mortality, even when adjusted for type of intervention and baseline mortality levels [14]. For community-based interventions to have a substantial impact on birth outcomes, therefore, it is necessary to have a large enough population coverage over a sustained period [14–16]. We hypothesise that the women's group intervention in Bangladesh did not show an impact on neonatal mortality because of its relatively low coverage relative to that used in India and Nepal. For this reason, we increased the coverage in the same geographical areas and the percentage of women in reproductive age and pregnant women exposed to the intervention. The impact of this scaled-up delivery of the intervention is the subject of an ongoing cluster randomised controlled trial detailed elsewhere [17].

Scaling-up is frequently discussed but seldom analysed or rigorously studied [18] yet examination and documentation of how scaling-up experiences are initiated, led, managed and monitored provide a deeper knowledge base and valuable lessons [19]. This paper details experiences in our expansion of women's groups in Bangladesh. We emphasise the importance of monitoring and evaluating the success of scale-up in relation to specific targets, with practical examples of how this may be done in resource-poor settings.

Recruitment and training were considered successful and facilitators' skills in leading women's group meetings generally improved over time. There were some difficulties in meeting education and marital status criteria for recruitment due in part to household responsibilities, and the lack of availability of educated women, who were often engaged in better-paid jobs or positions offering greater benefits. Staff turnover was reasonably high, with 18 out of the 36 recruited facilitators resigning. Turnover mainly occurred in the first three months following recruitment and was primarily due to issues of family constraints on freedom of movement or offers of more competitive employment opportunities. This led to a revision of recruitment strategies, in that shortlisted candidates worked in the field on a voluntary basis for seven days before officially being appointed as facilitators.

In a population of 243,341 people in the nine intervention unions, 648 new women's groups were added to the pre-scale-up 162 groups. All 810 groups continue to run to date. The population coverage is estimated at one women's group per 300 population (Table 1), an approximately five-fold increase in coverage relative to pre-scale-up levels. Following scale up, there is approximately one women's group per 57 ever-married women of reproductive age, compared to one group per 283 prior to scale-up and 23% of the 45,820 ever-married women in reproductive age living in the intervention areas are women's group members. Approximately 29% of women who gave birth and were interviewed as part of the community-surveillance system during the period January 2009 to June 2010 reported attending women's groups, compared to 3% prior to scale-up. Average attendance at each women's group meeting in the newly-formed groups is shown in Figure 2.

Coverage Indicator		TOTAL
Number of Women's Groups	Pre-scale-up	162
	Newly formed	648
	Total	810
Population in 2009		243,341
Ever-married women in reproductive age in 2009		45,820
Population Coverage (Total population/number of women's groups) (2009 pre-scale-up value)		300 (1502)
Coverage of reproductive-aged women (Total population of ever-married women in reproductive age/number of women's groups) (2009 pre-scale-up value)		57 (283)
% of reproductive aged-women who are women's group members (2009 pre-scale-up value)		23 (9)
% deliveries to women attending women's groups (2009 pre-scale-up value)		29 (3)

Intensity of the exposure to the intervention among women's group members in 2009 was estimated at 70%, meaning that, on average, women's group members were exposed to 70% of the intended dosage for 2009, i.e. an average of 8 out of 12 meetings were attended.

For both women's groups and community health worker interventions, coverage is thought to be a critical factor for a positive impact on neonatal mortality rates [14]. Simple quantitative data indicate considerable success in the scale-up of the women's groups in rural Bangladesh. We showed an approximately five-fold increase in population coverage, a 2.5-fold increase in the women's group membership among women of reproductive age and a 10-fold increase in the proportion of women who gave birth and reported attending a women's group. Scale-up was achieved without financial incentives for women and without any increase in managerial staff. Nevertheless, enhancement of already strong operational capabilities and institutional knowledge within our Perinatal Care Project (PCP) was crucial for the success of scale-up. Furthermore, PCP works under the auspices of BADAS, whose positive reputation as the largest non-governmental healthcare provider in Bangladesh facilitated acceptance of the intervention.

The specific impact of the scale-up on mortality rates is yet to be determined through an ongoing cluster-randomised trial [17], which includes detailed analysis of process indicators of delivery and receipt of the intervention. Nevertheless, documenting the process and success of scale-up is itself important if development initiatives are to foster participation and remain accountable to communities, and to avoid cumbersome organisational structures that are detached from their grassroots bases [23]. This paper does not presume to provide the ultimate solution to such challenges, but rather shares experiences of a single intervention in a resource-poor setting. Though limited to just nine unions in rural Bangladesh, these experiences and the described principles and processes of scale-up and its measurement are likely to be relevant to the wider development community, notwithstanding the need for intervention- and context-specific alterations and for programme flexibility.

Experience of delivery of the intervention on a smaller scale enabled PCP to meet the majority of requirements for successful scale-up as described by Simmons et al. (2006)[24] and summarised by Gilson and Schneider (2010)[19]. For example, clear messages about the objectives and advantages of women's groups were communicated to the community, while mapping exercises, household visits and community orientation meetings ensured early involvement through personal contact. Similarly, recognition of the importance of locally recruited and trained facilitators to manage groups, the participatory nature of the intervention, and active diffusion of key messages through household visits and community, governmental and non-governmental networks were central to the scale-up process. Concurrent systematic monitoring of the process and outcomes of scaling up further met Simmons et al's (2006)[24] and Gilson and Schneider's (2010)[19] recommendations to use evidence to guide and evaluate scale-up.

The scaling-up process described is a logical sequence and represents more or less the order in which various stages were initiated. In practice, however, the linear presentation is artificial and it is important to recognise that scaling-up is a dynamic, non-linear, iterative process whereby the various phases may occur simultaneously, in different orders, have feedback loops and may have to be repeated or revisited [18]. Feedback systems that allow for periodic programme corrections and continued innovation are central to successful scale-up and require managerial flexibility and strategic flair [19, 25].

Training and capacity building of staff is crucial for a thorough understanding of the group process and a sense of ownership at the grassroots level. Recruitment and selection of appropriate staff is critical, as are planned strategies to cope with staff turnover. Local recruitment was vital in maintaining effective relationships with the local community through familiarity with local customs, knowledge and beliefs. Recruitment criteria, however, cannot be rigid in areas of low literacy, low education and where there are gender barriers to employment. The resource-poor, rural context of our study area and an initial lack of appreciation by project managers for the need for family understanding of facilitators' roles increased vulnerability to staff turnover at the beginning of scale-up. This necessitated greater flexibility in recruitment criteria and processes, including a probationary period of field exposure, which ultimately minimised disruption caused by staff turnover. Effective systems of supervision, review and refresher training of project staff are likely to enhance delivery and sustainability of any intervention and therefore the success of any scale-up initiatives.

Various definitions of scaling-up exist, relating to complexity, impact and interactions with other organisations, but the concept of scaling-up as expansion of coverage is the most common [15, 23]. Geographical expansion is associated with particular challenges such as increasing distance from project headquarters, and larger areas and organisational structures to manage. Notwithstanding the introduction of groups into new villages in the existing unions, we did not expand geographically. Rather, our experience was an increase in the intensity of intervention in the same areas to increase population coverage. Some of the issues around effective geographical expansion and increased intensification are the same, however. In each case, scale-up does not mean exact duplication of pre-existing strategies. In our scale-up, lessons were learned from the previous, smaller-scale implementation of women's groups but existing structures and procedures were not always copied. For example, the schedule of group meetings and content were revised before scale-up to emphasise participation of women of reproductive age, and especially pregnant women. As a flexible replication of similar women's groups interventions in Nepal and India we endorse the recommendation that replication as a path for scaling up is only likely to work if done flexibly [23].

Strategies to measure the success of scale-up depend on specific objectives and anticipated outcomes. In relation to population coverage or intervention expansion, specific indicators of success, and how they are measured, are essential. There is no one-size-fits-all strategy for monitoring and evaluation and different methods have differing degrees of complexity and resource demands [26, 27]. Data from three different sources were used to measure the success of our scale-up, a particular strength of this initiative.

The capacity needed for such monitoring and evaluation systems should not be underestimated, however. Complex systems add to the cost of interventions considerably and are vulnerable to the limitations of population survey methods, such as recall and reporting biases [18, 27]. Furthermore, direct measurements of certain phenomena are not always straightforward or possible. For example, intervention coverage among pregnant women in the women's group intervention could not be measured directly as it was not possible to accurately measure the pregnancy status of all reproductive-aged women. The proportion of births that occur to women who report attendance at a women's group was therefore used as a proxy measure of coverage among pregnant women. This estimation depends on accurate reporting of women's group attendance and on all pregnancies ending in delivery, which is unlikely to be true. Nevertheless, provided that limitations are acknowledged and measurement methods are consistent pre- and post-scale-up, utilising proxy measures and actual data are more informative than entirely modelled estimates. Alternative methods for estimating coverage of an intervention among pregnant women with varying degrees of complexity, cost and data demands are described and discussed elsewhere [28].

Although the sustainability of women's groups is yet to be formally evaluated, unpublished data suggest that a high proportion of groups in Nepal and Malawi are running long after withdrawal of project financial support. Successful delivery of the intervention as planned during the first 18 women's group meetings may be considered indicative of institutional sustainability in terms of ongoing capacity to lead training, maintain the infrastructure, equipment and supplies and provide the necessary inputs and environment for implementation. Similarly, fairly consistent attendance levels at the women's groups (Figure 2) and reasonable population exposure estimates indicate programmatic sustainability from a community participatory perspective. Continued capacity building of women's group members, perhaps with a focus on facilitation methods for example, may obviate the need for paid facilitators and thus enhance the long-term financial sustainability of the intervention, whilst political sustainability can only be achieved through enhanced communication and advocacy efforts at local, national and global levels.

The ability to influence national policy from local initiatives is therefore a final determinant and indicator of successful scale-up [23, 25]. Before the women's groups became active, considerable time was spent identifying available services and other governmental and non-governmental organisations in the selected unions to communicate and build understanding and trust within the local community. These efforts enhance the participatory nature of scale-up as opposed to being purely driven by the implementing institutions or experts. In the women's groups, establishing links with other key players was beneficial in a number of ways, not least by involving traditional birth attendants and community health workers in raising awareness of the intervention among pregnant women encountered during their routine activities. We cannot objectively assess the success of wider stakeholder engagement in terms of bridging gaps and influencing policies, but these efforts are undoubtedly important for the acceptability and long-term sustainability of scale-up.

Strong operational capabilities, clarity of scale-up and intervention objectives and wide stakeholder engagement were critical to the successful scaling-up of the women's group intervention in rural Bangladesh. It was possible to increase community engagement with the intervention without financial incentives and without an increase in managerial staff. Monitoring and feedback systems that allowed periodic programme corrections and continued innovation were central to successful scale-up and required programmatic and operational flexibility.

ⁱ The definition of ever-married used in this paper includes married or divorced but not widowed women.

ⁱⁱ Community nutrition workers have a minimum of 8 years of basic schooling and have received 21 days training on nutrition and counselling.

ⁱⁱⁱ A Good Practice Guide [21] to delivering the women's group intervention, which may facilitate further replication and scale-up, is available at: http://www.wcf-uk.org/knowledge/wcf-publications/45-wcf-publications/363-good-practice-guide.