Defining the global health system and systematically mapping its network of actors

Relevant literature from global health, international relations, law, political science, and public policy was reviewed to understand how scholars are conceptualizing the global health system and how cognate global systems have previously been defined. This literature was used to inform development of a clear, applicable definition of the global health system (see Additional file 1 for a more detailed account of this process).

The helpfulness of this definition was tested through application to a mapping exercise of the global health system, in which websites that represent global health actors were identified. Using a systematic search protocol, a network of global health actors was created, characterized, and its social network analyzed. This methodology took advantage of the Internet’s network structure in a novel way, mapping online interactions to probabilistically identify key offline relationships in the global health system.

Importantly, the methodology was designed to be internally valid, transparent, and replicable, identifying a minimally-biased network of actors operating within the global health system. It was not designed to create a comprehensive or complete list of all actors operating within the global health system or to identify a list of actors based on their power or influence within the system. Through this process, it is hoped that our study output will provide a snapshot of the current global health architecture, reflecting a true network that can be used to both inform our understanding of the global health system and allow for further application and development of our methodology.

The findings of this study were presented to senior leaders of seven prominent global health organizations (i.e., Gavi, the Vaccine Alliance, Global Fund to Fight AIDS, Tuberculosis and Malaria, Joint United Nations Programme on HIV/AIDS [UNAIDS], United Nations Population Fund, United Nations Children’s Fund [UNICEF], WHO and World Bank) at a workshop in Geneva, Switzerland in December 2014 to assess the analytic helpfulness and comprehensiveness of the findings.

The literature review identified an abundance of definitions for key terms related to the global health system (see Additional file 1 for detailed methodology and findings of the literature review). Terms were defined using a variety of approaches and with varying boundaries, highlighting the usefulness of conceptualizing the global health system in different ways to serve different purposes [3].

Building on the work of Slzezak et al. (2010), [4] Hoffman et al. (2012), [19] and Frenk & Moon (2013), [3] the following definition for the global health system is proposed:

The global health system includes the transnational actors that have a primary intent to improve health and the polylateral arrangements for governance, finance, and delivery within which these actors operate.

Under this definition, the interactions between global health actors are influenced by the actors themselves, the internal arrangements of the system, and external forces, such as actors and arrangements from other important global policy domains. Accordingly, a global health actor is defined as an individual or organization that operates transnationally with a primary intent to improve health (see Table 1 for further explanation of key terms found within this definition).

The definition developed is holistic in nature, whereby the system is viewed from a global perspective. As a result, emphasis is placed on transnational actors engaged in such matters as protecting health security, promoting human rights, responding to humanitarian crises, and facilitating international development, and not on national actors or any particular sub-system of actors. The definition is comprehensive in its ability to consider multiple aspects of the system. First, the definition is inclusive of actors that operate with a primary intent to improve health, not limiting the system to include only those actors that hold powerful or influential roles in the global health system. Additionally, the definition considers relations between actors, arrangements that influence the system’s functions, as well as the interaction of the system components with internal and external forces. Importantly, the definition also applies to developing our practical understanding of the global health system. The ability to effectively translate conceptual aspects of the definition into clear inclusion/exclusion criteria for a systematic review process, as outlined in this study, illustrates is value at the operational level (refer to Additional files 3 and 4 for detailed application of the definition to inclusion/exclusion screening criteria).

The key global health system stakeholders who were consulted on the study’s findings agreed that the definition developed for the global health system was analytically helpful. To their knowledge, the research was noted to be the first systematic attempt to map the global health system, and that having access to the basic list of actors was helpful. However, the participants agreed the resulting map of 203 actors did not comprehensively present the most important and influential actors in the global health system. The exercise revealed that senior leaders at major global health organizations value global representativeness, and that the results of future studies would be more helpful if their findings were generalizable at the global level.

Our definition of the global health system was used effectively in a novel search methodology that took advantage of one of the world’s most powerful Internet crawlers and related search algorithms (i.e., Google “related:URL” search). The methodology proved useful and efficient, systematically generating the largest network of global health actors to date. Notably, all identified actors were organizations. However, the network did include organizations founded by individuals, such as the Bill & Melinda Gates Foundation, Elizabeth Glaser Pediatric AIDS Foundation, and Elton John AIDS Foundation. This reflects the tendency for websites to be created by organizations, whereas individuals tend to use social media or web pages on organizations’ websites for their online presence (e.g., Facebook pages and Twitter accounts representing celebrities or university webpages featuring faculty members).

Basic analysis of the network structure reveals interesting findings about the online network of global health actors that may shed light on the offline global health system structure. The 26 global health actors identified in the search “related:www.who.int”, which subsequent searches were based upon, are almost identical to the top 26 global health actors by degree centrality (after WHO, which ranks first). The only exception was the International Federation of Medical Students’ Associations, which was identified in the first related search but had one too few connections to rank in the top 27 actors by degree. Interestingly, Population Services International and UNAIDS – two organizations that were not identified in the first related search and therefore whose placements in the network were not dictated by the search methodology – tied for the 27th ranking by degree. This indicates that the related search function found them to be highly related to many other websites. They are placed in relatively central position in the online network in terms of activity. Offline, this may suggest Population Services International and UNAIDS are relatively well-connected and active players in the global health system.

Within the group of 26 actors identified through the first related search, there were eight organizations that consistently ranked in the top ten global health actors after WHO by degree, betweenness centrality, and closeness centrality: FHI360, Global Fund to Fight AIDS, Tuberculosis & Malaria, Global Health Council, Médecins Sans Frontières, Stop TB Partnership, U.S. Agency for International Development, U.S. Centers for Disease Control and Prevention, and UNICEF. This result indicates the websites of these actors are central to the network: they are related to a relatively large number of other actors’ websites, probably important in facilitating connections between websites, and allow for efficient connections to other actors in the system.

Through basic characterization of global health actors, various conclusions can be drawn. Categorization of global health actors by type indicates the overwhelming presence of global civil society organizations and non-governmental organizations in the online network of global health actors. The emergence of global health public-private partnerships is seen through an online presence that makes up just over 9% of identified global health actors [3]. Despite overwhelming influence of actors like the Bill & Melinda Gates Foundation, [21] philanthropic organizations represented only 0.5% of identified global health actors. Visualization of the network by type of actor shows actors of the same type in distinct groupings, such as those of public-private partnerships, UN entities and intergovernmental organizations, and private industry respectively, suggesting that websites of the same type of global health actors are related to one another. Interestingly, the pre-chosen focal point of the global health system – the WHO – is directly connected to all actor types with the exception of philanthropic organizations and multilateral development banks, for which only one actor was identified in each category.

Geographical distribution of actors’ international headquarters, spanning 73 cities across 16 countries, points to the global nature of the system. However, the overwhelming presence of actors’ headquarters in high-income countries – more than half in the U.S. – clearly suggests an uneven distribution of actors’ leadership globally. Significantly, no actor websites were located in the BRIC countries (i.e., Brazil, Russia, India and China). Several major centres of activity were identified, notably Washington, D.C., New York City, and Geneva. This reflects the location of various influential actors – global health and non-global health alike – around which global health actors have decided to co-locate. For example, the U.S. government and World Bank are headquartered in Washington, D.C., the UN is based in New York City, and WHO is in Geneva.

The inception of new global health actors over time indicates three distinct phases of rapid growth. The rate of inception of global health actors remained low until 1945, when a period of increased growth began, continuing until 1952. This growth coincides with the creation of the UN system in 1945 [22]. New global health actors may have been created as part of the nascent UN system itself, or in tandem with its development and also with that of other important multilateral organizations. A second phase of expansion began in 1970 and lasts through the decade, with a peak occurring in 1979. This expansion may reflect increased interest in international economic development at that time. For example, in 1973 the World Bank announced its pledge to increase financing for development by 40% over the following 5 years [23]. Last, an increase in actors is seen beginning in 1986 and continuing until 2006, coinciding with a quintupling of global health financing [24]. Each subsequent surge in the creation of new actors occurs at a higher rate than the previous one: an average of 2.00 actors were established per year from 1945 to 1952, 2.55 per year from 1970 to 1979, and 4.91 per year from 1986 to 2006). This reflects an increase in the rate of growth of the global health system over time. The period of analysis ends with a lull in inception of new actors, due in part to the decade being incomplete, but also probably due to how emerging actors’ websites may need time to gain relevance and popularity online and the 2008 global financial crisis.

Visualization of the network by year of inception shows newer global health actors in more central positions, and older organizations on the periphery of the network, indicating that the most influential actors in the global health system may not necessarily be the most established ones.

The proportion of actors whose primary intent is to improve health (61.6%), versus those who describe improving health as one of their primary intents (38.4%), illustrates the complex interaction between health and other global policy domains such as economic development and environmental protection. Network visualization shows some trends according to primary intent, such as in the high number of organizations for which improving health was not the primary intent that are connected to the U.S. Agency for International Development and the International Life Sciences Institute, respectively – two organizations for which improving health was not the sole primary intent.

We developed a clear, comprehensive and practical definition for the global health system that has the potential to facilitate consistent and in-depth research about this system in the future. In our study, the definition proved operational to map the structure of this system and analytically helpful for facilitating further understanding of the group of actors and their interconnections that people have come to naturally think of as the ‘global health system’. While this study was designed to yield a map of actors, it did not select actors based on their power or influence; this means the results are limited in terms of what features are analyzed. Results may also overlook some of the most important entities. The validation exercise with key global health stakeholders in December 2014 confirmed these findings; stakeholders agreed the definition was analytically helpful and the consultation revealed that the derived mapping does not exactly match with prevailing expert views of the system. However, this article presents a first systematic mapping of the global health system that can be improved upon in future exercises. Importantly, the use of our definition combined with Internet searches allowed for the systematic compilation of the largest list of global health actors to date. While the algorithm underlying the Google query refinement “related:URL” is not publicly available, the search engine is publicly available which makes it possible to replicate the methodology. Notably, any replications of the methodology will yield results that reflect the relationship between web pages on the date it is carried out.

Interestingly, despite defining the global health system from a holistic, global perspective, results may suggest a global health system characterized predominantly by the development agenda. This may reflect the nature of the global health system as it currently stands, defined by years of responding to global health challenges through a development lens. Alternatively, the methodology may have captured a portion of the global health system that is disproportionately focused on development assistance, perhaps as a result of using WHO as the search’s starting point. This could have occurred due to a tendency for WHO to provide links to its funders on its website. In the latter case, the post-2008 emergence of global health actors may have been underestimated, as these actors may be more likely to focus on global public goods and threats rather than development assistance to low- or middle-income countries. In this case, the search algorithm may not have found them to be related to actors identified earlier in the search.

By placing WHO at the centre of the system, this study used only one point of entry to probe the global health system. Using WHO’s website as the starting point of the search may have biased the search in favour of those types of actors to which WHO’s internal policies allow the organization to hyperlink. This may have increased the number of traditional and non-controversial actors identified, such as other UN entities and major WHO funders. However, our decision to start with WHO is justified by the widely acknowledged prominence and important role of this UN entity within the global health system. Other points of entry in approaching this exercise have their own advantages and disadvantages; for example, we could have begun the search with several global health actors – perhaps those with the largest global health budgets – but any starting list would have created its own biases.

Two other sources of potential bias have been identified. The use of www.google.com as well as a proxy server located in the U.S. may have biased results towards actors located in the U.S. However, the U.S.-based version of Google was selected because this is the domain of the original search engine launched by Google and is thus considered its standard product on which other location-based Google search engines are based [25]. Also, the default language of the WHO’s website is English. Since related search functions consider text-based analysis of websites’ content as one input when determining if web pages are related, the initial search may have biased results against actors whose default websites are not in English. This would include many important bilateral development agencies (e.g., Norway’s Norad) and global health actors in developing countries (e.g., Brazil’s Fiocruz).