Suicide is one of the fastest-growing and least-understood causes of death, particularly in low and middle-income countries (WHO 2011). The institutions involved in the derivation of taxonomic categories used to define and report suicide deaths, importantly shape how the burden is understood, (un)validated, and responded to at varying levels of social authority (familial, community, national). Despite growing recognition of the high burden of suicide deaths in low and middle income countries (WHO 2008) and an alarming level of suicide among Nepali women specifically (Suvedi et al, 2009), Nepal’s Ministry of Health does not systematically collect or report on suicides (Khan, 2002; Pradhan et al, 2011). Suicide data are filtered through reporting systems shaped by social, cultural, legal, and medical institutions. This study seeks to analyze networks drawn by police, policy, and health officials in order to better understand vital surveillance in Nepal and investigate how institutional networks affect how suicide deaths might be (un)documented and (un)reported within varying institutions.

Social network analysis can be used to not only show the ties between actors in a system, but the impact the relationships between actors can have on decision-making, flow of information and the overall structure of a society (Borgatti et al 2009). Particularly in developing contexts, such as Nepal, it is important to understand how increasingly complex information is being collected, assessed and shared. Four months of ethnographic fieldwork were conducted in Kathmandu, Nepal. 30 semi-structured interviews were conducted across health system levels from the macro (WHO, World Bank, DfID) to national bodies (Ministry of Health, Ministry of Police) and key organizations (government hospitals, police departments). Interviewees were asked to hand draw how deaths due to suicide were differentiated and documented. Subjects included data pathways and repositories so that comparisons could be made between the ‘official’ death surveillance system with how it is understood and followed by those implementing it. Textual data was coded and content analysis was performed in MaxQDA 10.1. Pictoral networks were coded and quantified and social network analysis (SNA) was performed in R Studio Version 0.98.1062.

Overall, there was large variation across the participant perceived networks, where some networks were linear pathways dominated by a single institution (police or community) with few nodes involved in data transmission, while others were complex and communicative. Such disagreement suggests disconnection amongst institutions. SNA demonstrated that police institutions controlled the majority of suicide information collection and reporting, while health and community institutions were peripherally involved. According to international ‘gold standard’ surveillance networks (WHO, 1999), macro level institutions like WHO should always be the ‘terminal’ point of data deposits. However, no informants indicated that information arrived at the WHO headquarters within Nepal. Network structures varied dramatically across institutional informants and surveillance was reported as incomplete and fragmented. Suicide surveillance remains a fragmented and disconnected process. Findings indicate a need for better communication of health systems frameworks across sectors (police, health, policy). Improved health information systems will further enhance the success of subsequent health improvement programs.