Multisectoral nutrition planning in Nepal: Evidence from an organizational network analysis

Abstract Multisectoral approaches are central to the global Scaling Up Nutrition (SUN) movement and the Sustainable Development Goals. Nepal joined SUN in 2011 and approved the first 5‐year Multisectoral Nutrition Plan (MSNP) in 2012, covering 2013–2017. This mixed methods study draws on organizational network analysis (ONA) and qualitative interviews with a sample of 22 organizations to examine (1) levels of engagement and network dynamics among government sectors and development organizations and (2) milestones and processes in the development and implementation of Nepal's MSNP. Findings suggest that the development of the MSNP was related to the high density of organizational connections; the leadership role of the Nepal's National Planning Commission and the National Nutrition and Food Security Secretariat; and the bridging roles played by a few government ministries and UN agencies that linked organizations that did not have direct relationships with each other. Specialized roles were observed for the three types of working relationships: policy dialogue, strategic planning and implementation. Partners were less connected on MSNP implementation than for policy dialogue and strategic planning, which may have constrained collaborative scale‐up efforts. The Ministry of Agricultural Development, in particular, was the conduit for connecting non‐health sectors into the broader network. Our study offers insights into the structure and dynamics of multisectoral planning in Nepal. It also contributes to a small but growing literature that illustrates how ONA can be applied to measure and use network results to elucidate the processes for strengthening multisectoral planning and implementation of nutrition‐specific and nutrition‐sensitive interventions.

Food Security Secretariat; and the bridging roles played by a few government ministries and UN agencies that linked organizations that did not have direct relationships with each other. Specialized roles were observed for the three types of working relationships: policy dialogue, strategic planning and implementation. Partners were less connected on MSNP implementation than for policy dialogue and strategic planning, which may have constrained collaborative scale-up efforts. The Ministry of Agricultural Development, in particular, was the conduit for connecting non-health sectors into the broader network. Our study offers insights into the structure and dynamics of multisectoral planning in Nepal. It also contributes to a small but growing literature that illustrates how ONA can be applied to measure and use network results to elucidate the processes for strengthening multisectoral planning and implementation of nutrition-specific and nutrition-sensitive interventions.

K E Y W O R D S
case study, intersectoral collaboration, organizational network analysis, qualitative research, social networks, Nepal, undernutrition The opinions and statements in this article are those of the authors and may not reflect official UNICEF policies.

| INTRODUCTION
Multisectoral approaches for nutrition feature prominently in the strategies and plans of governments in middle-and low-income countries. Although such approaches were adopted as early as the 1970s, they often achieved short-lived results due to wavering political support, a lack of sustainable resources and inadequate institutional capacity (Levinson & Balarajan, 2013;Warren & Frongillo, 2017).
However, multisectoral approaches were revived in the last decade in response to growing evidence of the impacts of malnutrition on child survival, development and economic growth , the need for progress on nutrition to achieve multiple Millennium Development Goals (MDGs) (UNICEF, 2009) and Sustainable Development Goals (SDGs) (Development Initiatives, 2017) and the recognition that a combination of nutrition-specific and nutrition-sensitive interventions is needed to tackle the multiple determinants of malnutrition Black et al., 2013;Ruel, Quisumbing, & Balagamwala, 2018).
The global Scaling Up Nutrition (SUN) movement in 2010 brought greater attention to nutrition governance in achieving a sustainable reduction in malnutrition (Acosta & Fanzo, 2012). It mobilized multiple stakeholders-governments, donors, UN agencies, civil society and the private sector-in collective multisectoral country-led efforts to improve nutrition. Countries that join the SUN movement are encour- Although there is a strong case for acting across several sectors to improve nutrition, documented experience on how to design, plan and scale up multisectoral approaches is quite rare (The World Bank, 2013;World Health Organization, 2013), although some recent case studies have begun to explore this area (Cunningham, Headey, Singh, Karmacharya, & Rana, 2017;Garrett & Natalicchio, 2011;Levinson et al., 2013;Pelletier, Gervais, Hafeez-ur-Rehman, Sanou, & Tumwine, 2016). The shortage of empirical research on multisectoral collaboration is juxtaposed with a high level of global interest that underscores the importance and timeliness of advancing this knowledge base (Glandon et al., 2019).
Nepal is an active participant in many global initiatives and has a long history of implementing multisectoral nutrition plans (Pomeroy-Stevens et al., 2016). This began with the development of nutrition strategies (1978 and 1986), the National Plan of Action for Nutrition (NPAN) in 1998 and an integrated plan of action in 2006 (Shrimpton, Crum, Basnet, Mebrahtu, & Dahal, 2014). Building on these efforts and a commitment to meeting the MDG target to half As Nepal's leadership in multisectoral nutrition planning is gaining global recognition, it becomes more critical to systematically identify the drivers of success and weaknesses of the MSNP's development and implementation. We designed a mixed methods study to examine the structure and working relationships of the organizations that contributed to the development first MSNP with a focus on the role of partnerships in strengthening the multisectoral platform to improve nutrition. The main objectives were (1) to determine the levels of engagement and network dynamics among key organizations and sectors of government and (2) to understand critical milestones and processes in the development of Nepal's MSNP. Our study contributes to a small but growing literature illustrating the way organizational network analysis (ONA) can be adopted to measure and use network results as an input into the 'black box' of the processes and dynamics to promote multisectoral strengthening for more effective implementation.

| Study design
Our primary method was ONA, supported by open-ended questions to understand the catalytic events, milestones and organizational partnerships leading to the development of the MSNP. Reports and grey literature were also reviewed to supplement the data on the key events and triggers that enabled multisectoral collaboration.
ONA is a social network method that is grounded in a sociological paradigm based on the premise that political, social and economic actors can be perceived as social networks of relations made up of

Key messages
• The development of Nepal's Multisectoral Nutrition Plan led to the establishment of strong governance structures at the national level and fostered a high level of organizational connectivity between sectors and stakeholders.
• The agriculture sector can play a larger bridging role in mobilizing non-health sectors to participate in broader multisectoral networks for nutrition. interrelated units (i.e. actors or organizations) (Faust & Wasserman, 1994;Scott & Carrington, 2011). ONA addresses the limitations of standard frameworks, tools and methods in studying the complexity of multisectoral collaboration (Bennett, Glandon, & Rasanathan, 2018;Glandon, Meghani, Jessani, Qiu, & Bennett, 2018) and offers a novel relational approach to identify the network structure and the roles of key organizational actors in multisectoral collaborations. The results go beyond organizational charts to detect underlying dynamics that may not always correspond to reported structures and official roles.

| Identification of sample
A review of organizations engaged in the development and implementation of the MSNP and consultations with the UNICEF regional and Nepal offices was used to create a bounded list of 22 organizations in line with ONA processes (Dozier et al., 2014). For each selected organization, up to three key informants who had been involved in one or more stages in the development of the MSNP were identified and listed in order of priority for data collection. Most of the interviews were conducted with one organizational representative; however, other individuals from the same organization were also interviewed if the primary respondent did not feel equipped to answer some questions.
Respondents included eight government organizations; eight UN agencies or programme sections; two donors; and four organizations representing civil society and national and international NGOs (

| Data analysis
ONA data were entered in Excel files and constructed into matrices for each ONA measure. UCINET software Version 6 (Borgatti, Everett, & Freeman, 2002) was used to analyse the data, and NetDraw (Borgatti, 2002)  The intensity of relationships was measured by asking key informants whether the nature of their relationship amounted to communication (lowest level of intensity), coordination, collaboration or integration (highest level of intensity). These measures are derived from a review of organizational partnership frameworks that define a continuum of integration from low to high (Gajda, 2004;Ruducha & Jadhav, 2018). A minimum confirmation process was used to verify the relationship intensity. For example, if Organization A indicated that they coordinated with Organization B on the development of the MSNP, whereas Organization B indicated they collaborated with Organization A, then the confirmed intensity of the relationship was set as coordination. See Table 2 for a complete list and definitions of all ONA measures.
The colours of the nodes in the plots represent the types of organizations in the network, and the size of the nodes was adjusted for T A B L E 2 Definitions of network measures (Freeman, 1978;Hanneman & Riddle, 2005) Node or individual organizational ties Degree centrality is calculated by counting the number of adjacent links to or from an organization. Based solely on direct connections, it reflects the potential power of having direct relationships. These direct links reduce the reliance on intermediaries to access information or resources. The assumption is that more connections are better than fewer connections.
Betweenness centrality measures the extent to which organizations fall between pairs of other organizations or individuals on the shortest paths (geodesics) connecting them. This measure represents potential mediation or flow of information or resources between organizations in the network. It is used to assess the power in networks, as an organization may control the flow of information and potential resources, thereby increasing dependence of others who are not directly connected in the network.

Relationship-level ties
Multiplexity is a measure that describes multiple relationships among the same set of organizations. The measures of the relations can be directed or not, and the relations can be recorded as binary, multivalued nominal or valued (ordinal or interval). In this study, three types of relationships were specified: advocacy and policy, strategic planning and scale-up. So the multiplexity score was either 1 (if only one type of linkage existed), 2 (any combination of two linkages) or 3 (all three relationships were confirmed).
Intensity describes the level of interaction between different organizations or nodes. The levels of interactions were classified as communication, coordination, collaboration or integration, in order of increasing intensity: communication (interaction as necessary to inform others or to check on specific issues); coordination (moderate-intensity interaction to share new ideas, ensure that duplication/overlap is minimized, etc.); collaboration (a close, ongoing, reciprocal, working relationship); and integration (organizations are part of a unified entity as they share decisionmaking, financial and/or human resources and have common policies and procedures).

Network-level ties
Centralization is an expression of how tightly the network structure is organized around its most central point. The general procedure involved in any measure of graph centralization is to look at the differences between the centrality scores of the most central point and those of all other points. Centralization, then, is the ratio of the actual sum of differences to the maximum possible sum of differences.
Density is defined as the sum of the ties divided by the number of possible ties (i.e. the ratio of all tie strength present to the number of possible ties). The density of a network may give us insights into phenomena such as the speed at which information diffuses among the nodes and the extent to which actors have high levels of social capital and/or social constraint. betweenness centrality. The acronyms and organizational categories listed in Table 1 were used in constructing the visual plots.
Qualitative data were abstracted into a predesigned matrix that was organized by key themes: advocacy and policies; government and partners; initiation of policy change; process timeline for key events; achievements and successes; and lessons learned. Additional themes were also documented and included information about the monitoring and evaluation of MSNP, key challenges and governance processes.
The information from interviews as well as the literature/document review was used to construct a timeline for the development of the MSNP (Figure 1) to document the sequence of events that led to the development of the MSNP.
All information from the qualitative interviews and ONA instruments was password protected, and personal identifiers were not used. Data were cleaned, and any potential inconsistencies were double-checked with paper-based instruments and audio recordings.

| Ethical considerations
UNICEF and BGH reviewed the study, and it met with standards of research for their respective institutions that were set by US mandate for IRBs. Our study posed minimal risk to participants; they were not a vulnerable group; and we obtained informed consent that included information about the study purpose, processes, contact information (if questions arose) and how the data would be used and protected to maintain respondent confidentiality.  Respondents also discussed how the key ministries coordinated to ensure effective programme implementation of an integrated set of essential nutrition interventions (both nutrition specific and nutrition sensitive), which were monitored through a joint management information system. However, a few respondents also described some tensions between sectors working together as organizational priorities

| Catalytic events: Organizational actors and milestones
were not always aligned as each sector absorbed the MSNP goals uniquely into their own plans and activities. Finding the right balance between the sector-specific demands and multisectoral coordination was cited as challenging by some. In addition, budgetary constraints, particularly for the non-nutrition sectors, were also challenging to the MSNP. respondents also expressed concern about the number of committees at the local level and the long-term sustainability of these structures.
Concerns were also raised about local capacity in recognition of the importance of a district-led approach given the diversity of contexts 3.2 | ONA: Network linkages and dynamics

| Overall MSNP network
To explore the overall MSNP network connectivity, we assessed the confirmed network plot (Figure 2), along with density (extent that all potential connections are realized) and degree of centralization (potential power of direct relationships). These latter two measures of network closure are important features affecting the internal process of organizing the performance of policymaking networks (Burt, 2000). with only a few organizations not reciprocating their relationships.
Due to a high density (53.3%), the centralization score was 51.4%, indicating that the MSNP network is not dependent on one or a few organizations to act as intermediaries in conveying the plans and activities of the network (Table 3).  (Table 4).
Organizations also clustered together, such as government bodies on

| Working relationships
The MSNP planning process started with advocacy to make the case for multisectoral collaboration to address child stunting with a package of nutrition-specific and nutrition-sensitive interventions. This led to a policy dialogue to develop consensus on the major goals and objectives followed by strategic planning activities to develop the MSNP and the implementation and scale-up of multisectoral activities (Figure 3a,c). The density of relationships and centralization are the highest for policy dialogue, followed by strategic planning and implementation (Table 3). The central broker and highest number of direct connections in all three types of linkages is the UNICEF Nutrition Section (Table 4). This illustrates UNICEF's supportive role to the government through the creation of bridges between organizations that enabled transmission of ideas and work to foster a unified understanding and potential for joint action.
Although the structure of the networks appears similar, there are distinct ways in which the organizations were positioned to fulfil different roles and responsibilities. For policy dialogue and development  (Table 4).

| Relationship strength
In examining the strength of relationships, the concept of multiplexity (participation across the three types of working relationships) and   We also uncovered challenges in the network structure and relationship patterns. The organizations in the network tended to cluster together by their own type such as the UN agencies and government ministries. This finding fits with the homophily principle of networks in which people and organizations with similar characteristics tend to stick together (Yuan & Gay, 2006). It is more difficult to achieve multi- Prior research has recognized that the structure of relations among actors and the location of individual actors in the network have important behavioural and attitudinal consequences for the individual units and for the system as a whole (Knoke, 1994) and can close 'structural holes' or gaps in a network (Yuan & Gay, 2006). The structural holes theory posits that organizations (or people) that bridge the holes in direct linkages possess more social capital because they have access to a more diversified group of organizations that can lead to more opportunities and better performance (Burt, 2000;Varda, 2011). The highest bridging roles or 'boundary spanning actors' were NNFSS and UNICEF Nutrition Section. Social capital is enhanced by sharing information, facilitating common understanding and generating trust and commitment (Pelletier et al., 2018). However, brokers who become the conduits of knowledge exchange must balance their positions and not become overwhelmed by their brokering role (Long, Cunningham, & Braithwaite, 2013).
Collective action is often considered a success criterion in networks (Kickert, Klijn, & Koppenjan, 1997). Whereas the MSNP network of organizations had strong connectivity in the policy dialogue and the strategy development of the MSNP, the implementation and scale-up displayed the weakest linkages. This finding was aligned with an earlier study conducted in Nepal that cited low nutrition capacity at district and community levels as a challenge to scaling up the MSNP interventions (Shrimpton et al., 2014). Globally, studies have also identified common difficulties such as the lack of incentives for non-health sectors to coordinate for nutrition, poor or non-existent mechanisms for accountability and limited functional vertical and horizontal coordination at district and community levels (Beyero, Hodge, & Lewis, 2014;Warren & Frongillo, 2017). Our study has several limitations. Using qualitative methods and ONA cannot ascribe causality and be generalized to other settings.
The use of ONA to assess multisectoral collaboration is a new area of work with no established standards of what constitutes a strong or a weak network. Although there is growing evidence that concepts such as density may influence the performance of partnership and networks, their exact mechanisms are not well known (Mays & Scutchfield, 2010). Some network measures such as density are affected by network size and may not be directly comparable across networks with substantially different structures.
Our selection of organizations and respondents was based on generating participants from a few key organizations and may have potentially excluded the less connected groups. However, we asked organizational respondents to include other organizations that they worked with that were not on a generated list of respondents. This process yielded very few others. As the study did not involve organizational respondents at the subnational level, this may have affected the perceptions about the subnational implementation networks in Nepal. Organizational interviews required the selection of an individual that was deemed knowledgeable about their organization's relationships with others. It is possible that some respondents did not have full knowledge of these organizational interactions. Lastly, the study lead/interviewer was employed by UNICEF, and controls were put into place to reduce potential bias. This included an informed consent process, no UNICEF staff were present during the interviews and the data were not accessible to UNICEF.

| CONCLUSIONS
Multisectoral collaboration is a central strategy for the achievement of the SDGs and of the global SUN movement but has remained an elusive concept for many countries. Despite the development of multisectoral nutrition strategies and plans across the globe, there is a shortage of empirical research on multisectoral collaboration to improve nutrition outcomes. We find that the development and implementation of the first MNSP in Nepal was characterized by a high degree of multisectoral and multi-stakeholder collaboration. Our study contributes to the small but growing literature that illustrates how ONA can be applied to measure, assess and use network results to understand the processes and dynamics to strengthen multisectoral planning and implementation.

CONFLICTS OF INTEREST
The authors have no conflict of interest.

CONTRIBUTIONS
JR designed the research study and conducted all interviews. AB analysed qualitative data, and CM performed ONA and developed visual plots in accordance with an analytical plan developed by JR, CM and AB. JR, HT, AB and CM wrote sections of the paper. HT provided guidance on the framing of this paper. All authors edited the paper.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.