**4. Discussion**

Our analysis of micro and community-level stakeholder perceptions towards drowning interventions revealed opportunities for the implementation of drowning reduction programs in the Sundarbans.

The findings suggested that all recommended interventions must be introduced together in a comprehensive program for maximum e ffectiveness. According to the participants, barrier-based interventions were considered appropriate for 1–2-year-old children, childcare for 3–5-year-old children, and swim and rescue training for children over the age of 6 years. Participants were largely homogenous

in this view, given cultural norms around childrearing and care. In addition, first responder training was perceived as important to encourage appropriate post-drowning actions. The age-appropriateness identified by participants for each intervention was in line with WHO implementation guidelines [24].

While the core components of these interventions would remain the same, such as ensuring that childcare spaces are secure and are provided during at-risk hours, the delivery processes of a comprehensive program should be adapted to the Sundarbans context. These changeable program characteristics include the nature of the community delivery agents, the capacity and capability of available workforce, availability of infrastructure and resources, partnership opportunities, methods of communication, and cultural adaptations such as changes to language and messaging [25,26]. The design and development of the comprehensive drowning program should involve community groups and stakeholders to ensure sustainability.

In this study, some specific intervention adaptations were identified as appropriate to the Sundarbans. An essential finding for the barrier-based intervention was that the preferred type of barrier varied by household. Hence, a drowning program may seek to deliver customised barriers for each household. Participants identified that childcare services should have an adequate child to caretaker ratio to ensure child security and provide pick up and drop o ff services to encourage attendance. These provisions to ensure child safety and support for attendance were also identified in international guidelines on childcare provision [27,28]. Participants were similarly concerned with safety for swim and rescue training services.

Participants also noted the need for complementary awareness activities, such as to dispel improper beliefs around e ffective child rescue techniques. Common responses to child drowning incidents involve engaging local quack doctors to perform rituals and trying to remove water by spinning the child over an adult's head [7,29]. Changing problematic norms and beliefs is an important step in behaviour change, and Sundarbans communities must be informed that such actions do not save children [30–33]. However, awareness itself is not su fficient to change behaviour, and must be accompanied with the removal of obstacles to change and capacity building [31,34]. Hence, awareness and first responder training in the Sundarbans may also need to target local 'quack' doctors who have some authority over community responses to drowning and may override individuals advocating for the administration of proper first aid. Ensuring that these local doctors themselves promote and administer appropriate first response may be critical for sustainable impact.

The sustainability of programs improves when they leverage existing governmen<sup>t</sup> structures [9]. Our findings sugges<sup>t</sup> that the ICDS, ASHA and SHG programs may provide platforms through which a drowning reduction program may be promoted and implemented. ASHA workers may play a promotive and monitoring role for the program and may also be involved in first responder training. However, many ASHA workers are overburdened with their duties and their drowning program role may be more sustainable if it is incorporated into their existing activities, such as providing barrier monitoring support as part of their regular household visits [35]. SHG members also showed willingness to be involved with drowning reduction activities and provided a network through which program activities can be advertised. Members were also available to be recruited for program delivery. ASHA worker and SHG members' performance may also vary depending on the frequency of visits from governmen<sup>t</sup> supervisors, so independent program monitoring may be required [36].

Concerns were raised around the utilisation of Anganwadi centres for childcare services. The ICDS program su ffered from unsafe venues, lack of Anganwadi training and poor sanitary conditions. The local governmen<sup>t</sup> also had limited authority over the changing operations of centres to include more hours of childcare, requiring permissions from state-level bureaucrats, which may take time given decision makers are risk-averse to changes. NGO participants suggested that a parallel program was more feasible. The long-term goal of health program design, implementation and scale up is often the uptake of these programs by government, as this improves the likelihood of sustained funding and delivery [14,37]. A parallel program may be less likely to be picked up by the governmen<sup>t</sup> as the ICDS program already provides childcare services as per policy. In addition, optimising existing Anganwadi

centres may require fewer resources than opening new centres. Community and local-government engagemen<sup>t</sup> activities should seek to decide on which model has long-term feasibility: optimising Anganwadi centres or running a parallel program.

Key facilitating factors that will enable implementation were identified by participants. Consistent community engagemen<sup>t</sup> and buy in of local leaders were essential. This is well founded in other LMIC contexts [38]. However, participants also noted that local governmen<sup>t</sup> was a ffected by nepotistic practices that may a ffect program quality. In West Bengal, a study found that local governmen<sup>t</sup> members were allocating agricultural resources to communities with more power, land and connections [39]. Hence, strict protocols and oversight may be required to ensure the equitable distribution of program resources.

Community participants advocated for local individuals to be trained as childcare and swim training providers. Implementation analyses have shown that local community-based workers best operate when they have access to resources, training and monitoring. Additionally, the building of soft skills, such as communication and leadership, is vital [40]. Community worker engagemen<sup>t</sup> and managemen<sup>t</sup> should be carefully defined and involve incentive structures appropriate to the context and matching community expectations [41–43].

Participants also noted that community-level committees are e ffective mechanisms through which residents can own programs and monitor implementation. These committees can also be engaged in advocacy and engagemen<sup>t</sup> activities and be instrumental in ensuring that implementation responds to community needs [26]. Increased community ownership of health programs may lead to better adaptation to the context and a greater likelihood of sustainability and acceptability [12,44]. However, the underlying assumption of all participants was that an NGO with expertise in child programs, such as CINI, would take primary lead in implementing and supporting community-level committees and program delivery. CINI has over 46 years of experience in delivering child programs in rural regions of West Bengal and is a suitable lead agency.

The development of the intervention may also consider the incorporation of other ideas. Although there is limited evidence on the e ffectiveness (and on the ethics) of tying children indoors to the ends of rope, there is some evidence that walking school bus programs can prevent injury in children [45]. While these have previously been used to reduce road tra ffic injuries, in the context of the Sundarbans, this may help reduce drowning events during commutes to school [46]. However, this intervention does not target the age group with the largest burden—1–4-year-old children.

To ensure the community ownership and development of an acceptable and feasible program, the next step of program design should involve community participatory approaches [10,47]. The present study found a range of issues that may a ffect program delivery, such as unpredictable geography, poor connectivity, religion and the caste of program providers, remoteness, poverty, poor governmen<sup>t</sup> program monitoring structures, requirement for appropriate incentives, recruitment challenges and the availability of appropriate venues. Communities are the best informants for how context-specific issues can be addressed and managed [48]. Participatory approaches will also improve community buy-in and redistribute the power of change into the community's hands [49]. The range of stakeholders as identified in the stakeholder analysis and should be appropriately engaged, starting with Gram Panchayat and block-level o fficials and moving to individual community leaders and members. No lifesaving organisations were identified which conduct drowning prevention activities, which was unsurprising as lifesaving organisations have had limited contribution to drowning prevention capacity development and advocacy in remote regions of India.

### *Limitations of This Study*

Due to ethical constraints, we were not able to gather information on participants' caste or religion. It is unclear if the perspectives found are representative across a range of religious groups. In addition, some governmen<sup>t</sup> program workers were recruited with assistance from Gram Panchayats. These may have been the more active and well performing workers and may not be fully representative of typical programs. This was particularly mitigated by ensuring at least one poor performing worker of each type was purposively recruited.
