UBC Faculty Research and Publications

Health policy and systems research collaboration pathways: lessons from a network science analysis English, Krista M; Pourbohloul, Babak Aug 28, 2017

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-12961_2017_Article_241.pdf [ 1.81MB ]
JSON: 52383-1.0354988.json
JSON-LD: 52383-1.0354988-ld.json
RDF/XML (Pretty): 52383-1.0354988-rdf.xml
RDF/JSON: 52383-1.0354988-rdf.json
Turtle: 52383-1.0354988-turtle.txt
N-Triples: 52383-1.0354988-rdf-ntriples.txt
Original Record: 52383-1.0354988-source.json
Full Text

Full Text

RESEARCH Open AccessHealth policy and systems researchcollaboration pathways: lessons from anetwork science analysisKrista M. English1,2 and Babak Pourbohloul1*AbstractBackground: The 2004 Mexico Declaration, and subsequent World Health Assembly resolutions, proposed a concertedsupport for the global development of health policy and systems research (HPSR). This included coordination acrosspartners and advocates for the field of HPSR to monitor the development of the field, while promoting decision-makingpower and implementing responsibilities in low- and middle-income countries (LMICs).Methods: We used a network science approach to examine the structural properties of the HPSR co-authorshipnetwork across country economic groups in the PubMed citation database from 1990 to 2015. This analysissummarises the evolution of the publication, co-authorship and citation networks within HPSR.Results: This method allows identification of several features otherwise not apparent. The co-authorship networkhas evolved steadily from 1990 to 2015 in terms of number of publications, but more importantly, in terms of co-authorship network connectedness. Our analysis suggests that, despite growth in the contribution from low-income countries to HPSR literature, co-authorship remains highly localised. Lower middle-income countries havemade progress toward global connectivity through diversified collaboration with various institutions and regions.Global connectivity of the upper middle-income countries (UpperMICs) are almost on par with high-income countries(HICs), indicating the transition of this group of countries toward becoming major contributors to the field.Conclusions: Network analysis allows examination of the connectedness among the HSPR community. Initially (early1990s), research groups operated almost exclusively independently and, despite the topic being specifically on healthpolicy in LMICs, HICs provided lead authorship. Since the early 1990s, the network has evolved significantly. In the full setanalysis (1990–2015), for the first time in HPSR history, more than half of the authors are connected and lead authorshipfrom UpperMICs is on par with that of HICs. This demonstrates the shift in participation and influence towardregions which HPSR primarily serves. Understanding these interactions can highlight the current strengths andfuture opportunities for identifying new strategies to enhance collaboration and support capacity-building effortsfor HPSR.Keywords: Health policy, Systems research, Low- and middle-income countries, Co-authorship networks,Capacity-building* Correspondence: babak.p@ubc.ca1Complexity Science Lab, School of Population & Public Health, University ofBritish Columbia, Vancouver, British Columbia, CanadaFull list of author information is available at the end of the article© The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.English and Pourbohloul Health Research Policy and Systems  (2017) 15:71 DOI 10.1186/s12961-017-0241-5BackgroundThe Mexico Ministerial Statement for the Promotion ofHealth (the Mexico Declaration) [1], and subsequent WorldHealth Assembly resolutions, proposed a concerted globalprogramme of work to support the development of healthpolicy and systems research (HPSR). This included coordin-ation across partners and advocates for the field of HPSR tomonitor the development of the field, while promotingdecision-making power and implementing responsibilitiesin low- and middle-income countries (LMICs) [1–3].Bibliometric analysis of HPSR provides a systematicand scientific means of monitoring this development.This task has been carried out by a number of groups inrecent years [4, 5], including the authors of this paper[6]. These results have demonstrated that great strideshave been made to support and ensure meaningful in-clusion of LMICs in HPSR. While lead authorship fromLMICs is increasing and outpacing the growth in leadauthorship in life and biomedical sciences (PubMed) ingeneral, LMIC authors are significantly under-representedin terms of absolute number of HPSR publications ontopics relevant to, and including, LMICs.Building on this understanding, questions remain re-garding the intricate collaborative interactions that shapethese trends. Understanding these interactions can high-light the current strengths and future opportunities foridentifying new strategies to enhance collaboration andsupport increased LMIC contribution to HPSR.To address this, a special framework is required. Thisframework must simultaneously capture the contribu-tions of individuals (e.g. authors, policymakers, imple-menters, institutions) in the HPSR literature (micro-levelfactors), as well as the national, regional or global leveltrends (macro-level factors). Recent advances in networkscience have contributed to the development of a frame-work that allows us to analyse these micro- and macro-level trends as well as other dynamic complexities.The digitisation of publications and the databases thathouse them have propelled bibliometric studies to attemptto capture network structures from authors’ names, affilia-tions and geospatial distribution. In recent years, massivedatabases, at various levels of granularity, have becomereadily available for analysis. New methods for analysishave provided inspiration for identifying new metrics andfurthering our understanding of the significance and rela-tive contribution of authors, institutions, as well as re-gional and/or multidisciplinary collaborations. The coreconcept behind this network analysis approach is basedon developments in the physics and computer sciencecommunities over the past decade [7–9].MethodsWe explore a network representation of co-authorshipdata, hereafter referred to as a co-authorship network.This network is comprised of nodes and edges; eachnode represents an author who has co-authored at leastone HPSR publication, while each edge (link) is repre-sented by a line connecting two nodes and correspondsto publication(s) that were co-authored by those two au-thors (nodes) (right inset, Fig. 1). The co-authorship net-works provide compelling insights into the current stateof collaboration within the discipline, between regionsand over time.A co-authorship network can help identify efficientopportunities to strengthen the research capacity inLMICs through international collaborations. The net-works can also demonstrate both the gaps and emergingtopics within health policy and systems research, facili-tating oversight for regional planning to ‘stay ahead ofthe curve’ by building home-grown capacity relevant totomorrow’s needs. Similarly, researchers, may identifystrategies to maximise their scientific contribution and/or influence on policy decision-making.A co-authorship network captures collaboration pat-terns between authors. The type, frequency, distance andnumber of collaborations determine the pace at whichthe discipline advances. Co-authors are identified frombibliometric data that have been narrowed to the specificfield of study. Additional information contained withinthe database may enrich the networks and reveal otherinteresting features about the collaborations. Identifyingthese patterns over time facilitates our understanding ofthe dynamic interactions and provides an opportunity toidentify strengths and challenges in the HPSR co-authorship network.PubMed was used to study the network of contribu-tors to the HPSR literature. PubMed is a vast resourceof literature relevant to the life and biomedical sciences,including more than 26 million citations, as of August2016. It has twice as many health policy-relevant publi-cations as the next largest collection [6].Details of our data collection and processing approachwas reported in a previous publication (please see [6]).In summary, we used a high-level keyword search strat-egy to identify the literature relevant to HPSR and en-sure inclusivity. Additional terms and keywords can beadded to refine the search or learn more about sub-groups under the HPSR umbrella. The syntax of thehigh-level keyword search strategy used the logical Bool-ean operators “AND” and “OR”: (health AND policy)OR “health system*”. While the specific topic of thepaper may be related to any area within the scope ofHPSR, this strategy assumes that papers related to HPSRwould have the words ‘health’ and ‘policy’ or ‘health sys-tem(s)’ somewhere in the text. PubMed includes a pre-scribed set of filters to identify specific topics related toclinical queries and medical genetics [10]. The exclusioncriteria can be applied to the search strategy using theEnglish and Pourbohloul Health Research Policy and Systems  (2017) 15:71 Page 2 of 10Boolean operator, “NOT”, thereby removing the irrele-vant clinical literature [11]. The species filter was appliedto restrict the results to human studies [12], resulting inapproximately 85,000 HPSR publications.The HPSR literature was further refined to a cohort ofpublications that captured topics relevant to LMICs,resulting in a subset of approximately 7000 from theabove 85,000 HPSR publications. This subset serves asthe basis for much of the analysis that follows.To identify the collection of papers with its main topicfocused on an issue relevant to a LMICs, we first per-formed the keyword search strategy to identify the sub-set of publications relevant to health policy and systemsresearch. We then used the title and abstract sections,denoted by the tag “Title/Abstract [TIAB]”, as it isintended to most concisely describe the main focus andpurpose of a paper. Therefore, HPSR publications with amain focus relevant to LMICs can be efficiently identi-fied by limiting the search to the list of 135 LMICs andsynonyms for “developing country” that appear in thetitle and abstract [13]. Keywords (topics) may also be in-cluded here, but without mention of an LMIC, it wouldbe difficult to determine whether the topic is specificallyrelevant to LMICs or of a more general HPSR issue rele-vant to high-income countries (HICs).The networks were produced by Cytoscape, an open-source software platform for visualising complex net-works [14]. The input to this software comprised ofcompiled files downloaded from PubMed as describedabove. The visualisation techniques used to show thenetworks can include millions of nodes and edges. Thisscalability is advantageous when studying networks thatare increasing in size over time, such as the emergingand expanding discipline of HPSR.Interpretation of a co-authorship network structure re-quires careful consideration, illustrated through the insetin Fig. 1. Let us assume that six individuals co-author apaper. In this case, these individuals are represented bysix nodes in the network, and since they are all co-authors on the same paper, each pair of them must beconnected to one another with an edge, resulting in 6 ×(6–1)/2 = 15 edges between them (see left inset in Fig. 1).Similarly, if a paper is co-authored by 10 authors, thenthe 10 nodes representing these authors must be con-nected to one another by 10 × (10–1)/2 = 45 edges.Therefore, while each author is uniquely represented byFig. 1 Health policy and systems research co-authorship networks from 1990 to 1994 (left panel) and 1990 to 1999 (right panel). Node colourrepresents the economic classification of first-authors’ country, as per the World Bank. The left inset shows the small disjoint chains (SDC) prior to1994, where each chain is comprised of authors from the same economic region, and very often, from the same institution. The right inset showsthe authors becoming gradually more connected, yet still considered to have SDC structure. There were 378 nodes in 1990–1994 and 1119 nodes in1990–1999. Orange colour indicates low-income countries (LICs), green for lower middle-income countries (LowerMICs), pink for upper middle-incomecountries (UpperMICs) and blue for high-income countries (HICs)English and Pourbohloul Health Research Policy and Systems  (2017) 15:71 Page 3 of 10a node in the network, a paper may be represented bymultiple edges depending on the number of co-authorson that paper.On the other hand, let us assume that two authors co-authored just one paper. In this case, the two authorsare represented by two nodes, while the edge betweenthem represents the sole co-authored publication. Simi-larly, let us assume that two authors co-authored 15 pa-pers together. In this case, again, the two authors arerepresented by two nodes; however, they are connectedby a thicker edge representing all 15 publications co-authored by them. As such, the thickness of an edge de-pends on the number of papers co-authored betweentwo authors (nodes) within a given time interval; thehigher the number of co-authored papers, the thickerthe edge connecting these two nodes.Results and DiscussionContribution of different economic regions to the HPSRliteraturePrior to 2014, PubMed only required the first author ofa paper to provide their institutional affiliation as part ofauthor bibliographic data. The first author’s affiliationwas used as a proxy to represent author’s country ofresidence. Given that only one institution/country isassigned to each publication in PubMed, this affiliationwas attributed to the same paper, regardless of the sub-sequent authors’ affiliations. While this facilitates captur-ing the global connectivity of co-authors, it limits ourability to analyse all co-authors’ countries. Despite thislimitation on the secondary analysis of the database, im-portant observations can be summarised with regards toregional contribution to the HPSR literature, as it de-pends, by and large, on the first-authors’ affiliations.In addition to the global behaviour of the HPSR co-authorship network, the contribution of different economicregions may be examined. The World Bank’s 2016 fiscalyear country economic classification was applied retro-spectively to all previous years. This classification includeslow-income countries (LICs; with a gross national income(GNI) of US$1025 or less in 2015), lower middle-incomecountries (LowerMICs; with a GNI between US$1026 andUS$4035), upper middle-income countries (UpperMICs;with a GNI between US$4036 and US$12,475), and HICs(with a GNI greater than US$12,476) [15]. The specificcolour codes used in the following figures correspond todifferent World Bank economic regions.To analyse the HPSR publications, systematically, we di-vided the period from 1990 to 2015 into five consecutive5-year intervals; the last interval covers 6 years to include2015, the last year before conducting this study.Figure 1 (left panel) shows the HPSR co-authorshipnetwork for the first time interval between 1990 and1994, which represents 378 authors (nodes). Thenetwork is comprised of small groups of authors, ran-ging from 2 to 10, and who collaborate in clusters thatare separate from one another, referred to as small dis-joint chains or small disjoint components (SDCs). Giventhe very low number of co-publications between authorsduring this interval, almost every SDC in this figure islimited to one economic region, i.e. all nodes withineach SDC have the same colour. This corresponds to theearly stage of the formation of the HPSR literature, whenmany groups and individuals work in isolation. This timeinterval also experienced a low number of publications(five or fewer) per person.Collaboration and co-authorship between individuals isnot an isolated activity; it spans across their professionalcareers. As such, it is important to view and analyse theircollective behaviour, in a cumulative manner, over time.To achieve this objective, we present the cumulative net-works for the subsequent intervals after 1994. In otherwords, we investigate the network behaviour for the inter-vals of 1990–1999, 1990–2004, 1990–2009 and finally,1990–2015, by incrementally adding new nodes and edgesto the existing network from previous interval(s).Figure 1 (right panel) shows the network for the inter-val 1990–1999, with 1119 authors contributing to theHPSR literature. An increase in the number of publica-tions and participation of more authors during this ex-tended interval marks the beginning of formation ofclusters that are composed of authors from different re-gions (see right inset in Fig. 1). Despite this evolution,the global structure of the network remained, by and large,disconnected and only comprised of SDCs. In addition,while only papers that focus on a topic relevant to LMICshave been included, the majority of first authors are fromHICs, while very few are from LICs. Furthermore, duringthe initial stage of HPSR development, HIC nodes play aprominent role in binding the network together.The cumulative interval between 1990 and 2004 marksan important transition in overall (global) connectivityof the HPSR co-authorship network. For the first time,the volume and diversity of collaboration grew to 2887authors. This network size allowed for the formation ofthe largest connected component (LCC). This compo-nent is magnified within a dashed ellipse in Fig. 2. Theformation of LCCs is indicative of the ability of co-authors to work collaboratively beyond their previously-isolated SDC and establish new ties with authors inother SDCs over time. A closer look at the LCC revealsthat, at this initial phase, the dendritic structure of LCCremains fragile and the connectivity of the componentdepends on a few critical edges (co-authored papers).While 606 (21%) nodes belong to the LCC in this inter-val, the majority of nodes (2281 or 79%) are still SDCs.The next cumulative interval between 1990 and 2009captures the evolution of a more robust LCC, which is aEnglish and Pourbohloul Health Research Policy and Systems  (2017) 15:71 Page 4 of 10result of expanding collaboration among a larger groupof authors (2394 of 6769 nodes). The robustness of thenetwork (Fig. 3, left panel) reached a level wherebythe overall connectivity was not dependent on a fewedges. However, despite the formation of a stableLCC, the majority of nodes (~65% or 4375 nodes) re-main within SDCs.During the last cumulative interval between 1990 and2015 (Fig. 3, right panel), for the first time the numberof nodes within the LCC (9623 or 61%) exceeds that cor-responding to SDCs (6078 or 39%). The robustness ofthe network is indicative of the existence of multiplepathways between different groups and individuals, lead-ing to cross-fertilisation of ideas and contribution of abroader group of experts from different disciplines tothe HPSR literature. Stratification by region (Fig. 4)shows improvement in all economic regions.An important global feature of the 1990 to 2015 net-work is the emergence of a strongly connected clusterinfluenced by the UpperMICs (Fig. 5). This emergentpattern, predominantly driven by Brazil, China, SouthAfrica, Iran and Thailand, has helped the UpperMICs toshape the global structure of the HPSR co-authorshipnetwork on par with HICs. More importantly, this emer-gent cluster also acts as a hub to connect authors fromall economic regions (see the lower panel in Fig. 5).Among LowerMICs, the global spread is predominantlydriven by India, Pakistan, Kenya and Nigeria.Facilitating the growth of similar hubs in the years tocome may considerably strengthen the global structureand robustness of the network, especially if it integrates,more profoundly, authorship from LICs and LowerMICs.HPSR literature by the numbers: co-authors, publications,citationsThe co-authorship network may also be examined interms of the authors’ collaborative reach, by consideringtheir ‘degree’. A node’s degree is the number of edgesemanating from it. In the context of a co-authorshipnetwork, a node’s degree is the overall number of otherindividuals with whom they co-authored. An authormight have one or few publications co-authored withmany people, thus a high degree. Alternatively, an au-thor may have many publications co-authored with fewFig. 2 Health policy and systems research co-authorship network structure from 1990 to 2004. In contrast to previous intervals, a large connectedcomponent (LCC) is formed during this interval (upper part of the left panel). The right panel shows an enlarged view of this LCC, which showsvarious sub-structures, may suggest the beginning of a broad and heterogeneous pattern of collaboration among co-authors. Colour codes arethe same as Fig. 1. Of a total of 2887 nodes in this interval, 2281 contribute to the formation of SDCs and 606 belong to the LCC. Orange colourindicates low-income countries (LICs), green for lower middle-income countries (LowerMICs), pink for upper middle-income countries (UpperMICs)and blue for high-income countries (HICs)English and Pourbohloul Health Research Policy and Systems  (2017) 15:71 Page 5 of 10individuals overall, then the node has a lower degree. Itis also possible that an author has several publicationsco-authored with several people overall (high degree), orone has few publications with few people (low degree).The frequency distribution of degrees for all nodesacross the network is called the ‘degree distribution’ ofthat network. It is important to highlight that the degreeonly corresponds to the papers that satisfy our searchcriteria; thus, an author might have produced more pa-pers in any given interval than shown, but these wouldbe outside the HPSR scope of this analysis. Figure 6shows the degree distributions of number of publicationsfor the LCC of three networks introduced earlier (blackdots). In these figures, both horizontal and vertical axesare in logarithmic scale, which allows values with differ-ent orders of magnitude to appear in one figure. Alsoshown in each panel is a fitted (red) line to data points.Such line on a logarithmic (log–log) plot is indicative ofscale-free (or power law) distribution. In networks withscale-free distribution, a small fraction of nodes has verymany contacts (right hand side of data points in eachpanel), while the majority of nodes have very few con-tacts (left hand side of data points in each panel).Progressive examination of the three panels revealsthat, generally, the same group of authors contribute tothe right-hand of the distribution tails shown in Fig. 6.This conforms with the notion that ‘the rich get richer’,which is a generic feature of scale-free networks, andhave been observed in a wide array of network struc-tures representing natural and socio-technological sys-tems. In the context of co-authorship networks, thisimplies that few groups/authors could establish them-selves as key players by increasingly attracting relevantfunds and human resources over time, to sustain theirFig. 3 The structure of health policy and systems research co-authorship network from 1990 to 2009 (left panel) and 1990 to 2015 (right panel).Node colour represents first-authors’ economic region. Compared with the previous figures, the size, connectivity and robustness of the largeconnected component (LCC) grows over time. From 1990 to 2009 (left panel) 4375 nodes (65%) contribute to the formation of SDCs and 2394nodes (35%) belong to the LCC, while during 1990–2015 (right panel) these are 6078 (39%) and 9623 (61%), respectively. Orange colour indicateslow-income countries (LICs), green for lower middle-income countries (LowerMICs), pink for upper middle-income countries (UpperMICs) and bluefor high-income countries (HICs)English and Pourbohloul Health Research Policy and Systems  (2017) 15:71 Page 6 of 10HPSR publication. While the establishment of strong hubsis generally viewed positively, at the global level, there is arisk of inadequate distribution of resources in regionswhere they are needed most. As such, it would be import-ant to iteratively examine the future potential for newhubs to emerge in different socioeconomic regions.In a co-authorship network, nodes may also representthe number of HPSR publications per author. In additionto the number of publications, it is also important toexamine to what extent an individual’s work has had animpact on the scientific community. A measure used toevaluate this impact or influence is the number of timesan author’s paper is cited. Since a network structure en-capsulates information about all papers published by aperson, a more appropriate measure is the total numberof times that an author's papers are collectively cited up tothe end date in each interval.To examine number of publications and times citedmore closely, we extract the most prolific HPSR au-thors (to the end of 2015) who published 15 HPSR pa-pers or more, along with their first neighbours. Thefirst neighbours of a node are the other nodes directlyconnected to the original node by an edge, regardless oftheir number of publications. This subset of 21 mostprolific authors and their first neighbours leads to anetwork of 1026 nodes, which is shown in Fig. 7. In thisfigure, the node’s inner colour corresponds to theauthor’s number of publications (see figure legend), sizecorresponds to the number of times cited and bordercolour represents the first author’s economic region.One important feature observed from this network isthat the number of publications does not necessarilycorrelate with number of times cited for an author. An-other feature is that, by and large, highly-cited authorsare from HICs or UpperMICs. Only a handful of toppublishers and/or highly-cited individuals come fromLowerMICs. Representation of LICs in this subset re-mains marginal.In general, bibliometric analysis examines the fre-quency of publications over time. Co-authorship and cit-ation analysis are an extension of this and are bestunderstood using network analysis.In this study, we used PubMed as the main databasedue to its comprehensiveness. This came at a limitationthat only the affiliation of the first author of a paper wasrequired for this dataset prior to 2014. Starting in 2014,PubMed has added subsequent authors’ affiliations tothe database.Availability of more refined data and resources in thefuture to include more country- and institution-specificinformation will allow us to capture more delicate pat-terns from the co-authorship. We did not include a listof most frequently published authors so as to avoid sin-gling out individuals.Fig. 4 The large connected component (LCC) in the middle corresponds to the interval 1990–2015. The four surrounding networks (grey background)are identical to the one in the middle, but stratified by the economic classification of the first authors’ country affiliation. For description on the areamarked by a dashed black circle, please see the next figure. Orange colour indicates low-income countries (LICs), green for lower middle-income countries (LowerMICs), pink for upper middle-income countries (UpperMICs) and blue for high-income countries (HICs)English and Pourbohloul Health Research Policy and Systems  (2017) 15:71 Page 7 of 10Fig. 6 Degree distributions (black dots) of the large connected components corresponding to three cumulative time intervals. The best logarithmic fitto the data set in each panel is depicted by the red lineFig. 5 Successive magnification of a segment of the HPSR co-authorship network from 1990 to 2015 (starting from top right panel, to left, to bottom rightpanel) reveals a more intricate collaborative relationship between authors from different economic classifications. While in the previous cumulative intervals,authors from high-income countries used to play a dominant role in the overall connectivity of the network, the 1990–2015 cumulative interval shows thatupper-middle-income countries (UpperMICs) are catching up in establishing their influence on the network. Orange colour indicates low-income countries(LICs), green for lower middle-income countries (LowerMICs), pink for UpperMICs and blue for high-income countries (HICs)English and Pourbohloul Health Research Policy and Systems  (2017) 15:71 Page 8 of 10ConclusionComplexity science and network analysis add tremen-dous value to our understanding of the growth in HPSR.This analysis shows patterns of knowledge production(publication), collaboration (co-authorship) and potentialpolicy influence (citation volume) over time and betweencountries. We consider that the bulk of citations maynot necessarily be restricted to purely academic studies,as many indexed publications indeed stem from pro-ceedings, reports, policy meetings, working groups, etc.This approach can identify and encourage support forregions with fewer publications and/or citations to in-crease participation and influence, as well as facilitat-ing opportunities for collaboration across economicclassifications to ensure LMICs meaningfully partici-pate in HPSR.This analysis summarised the evolution of the publica-tion, co-authorship and citation networks within HPSR.Initially (early 1990s), groups operated almost exclusivelyindependently and despite the topic being specifically onhealth policy in LMICs, HICs provided lead authorship.Since the early 1990s, the network has slowly but signifi-cantly evolved given the relatively short time period. Inthe full set analysis (1990–2015), for the first time inHPSR history, more than half of the authors are con-nected and lead authorship from UpperMICs is on parwith that of HICs. This demonstrates the shift in partici-pation and influence toward regions which HPSR pri-marily serves.Enhancing support for participation by the LMIC thatthe discipline is meant to serve is imperative for success,and in particular LICs, since publications in these coun-tries are increasing at a greater pace than any other eco-nomic region, but the absolute number is quite low.Thus, while capacity is expanding, additional supportwill greatly enhance this growth until they are more ad-equately represented within the discipline.This study provides an unprecedented perspective andsheds light on the regional heterogeneity in contributionto HPSR, necessitating elevated investment for HPSRcapacity-building in LICs and LowerMICs, facilitatingUpperMICs to become more prominent players, andinvesting in moving away from a core-reliant co-authorship network structure towards a more distributednetwork structure.AbbreviationsGNI: gross national income; HPSR: health policy and systems research;LCC: largest connected component; LICs: low-income countries; LMICs: low-and middle-income countries; LowerMICs: lower-middle income countries;SDC: small disjoint chains; UpperMICs: upper middle-income countriesAcknowledgementsNot applicable.Fig. 7 A subset of the 1990–2015 co-authorship network (n = 1026 nodes) that includes the most prolific authors (with 15 papers and more), aswell as their first neighbours. The first neighbour of a node are those nodes that are directly connected to that original node by an edge. Thisfigure is information rich and shows more attributes per node, including the number of publications (body colour) the number of times anauthor’s work is cited (size) and the first-author’s economic region (border colour). Orange colour indicates low-income countries (LICs), green forlower middle-income countries (LowerMICs), pink for upper middle-income countries (UpperMICs) and blue for high-income countries (HICs)English and Pourbohloul Health Research Policy and Systems  (2017) 15:71 Page 9 of 10FundingThis work was supported by the WHO Alliance for Health Policy and SystemsResearch. The funders had no role or influence in the design or conduct ofthe study.Availability of data and materialsThe datasets generated and/or analysed during the current study areavailable publicly from PubMed, https://www.ncbi.nlm.nih.gov/pubmed/ afterentering the search criteria described in this paper.Authors’ contributionsKE and BP contributed to the conception, methodological design as well ascollection and analysis of data, interpretation of results and drafting of themanuscript. Both authors read and approved the final manuscript.Ethics approval and consent to participateNot applicable.Consent for publicationNot applicable.Competing interestsThe authors declare that they have no competing interests.Publisher’s NoteSpringer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.Author details1Complexity Science Lab, School of Population & Public Health, University ofBritish Columbia, Vancouver, British Columbia, Canada. 2Institute ofResources, Environment and Sustainability, University of British Columbia,Vancouver, British Columbia, Canada.Received: 1 February 2017 Accepted: 9 August 2017References1. World Health Organization. Report from the Ministerial Summit on HealthResearch: Identifying challenges, inform action, correct inequalities. Geneva:WHO; 2004. http://www.who.int/rpc/summit/documents/summit_report_final2.pdf, Accessed 21 Aug 2017.2. Bennett S. Briefing Note 1: What is Health Policy and Systems Research andwhy does it matter? 2007. World Health Organization, Alliance for HealthPolicy and Systems Research. http://www.who.int/alliance-hpsr/resources/Alliance%20HPSR%20-%20Briefing%20Note%201.pdf. Accessed 21 Aug 2017.3. World Health Organization. Changing Mindsets: Strategy on Health Policyand Systems Research. Geneva: WHO; 2012. http://apps.who.int/iris/bitstream/10665/77942/1/9789241504409_eng.pdf, Accessed 21 Aug 2017.4. Adam T, Ahmad S, Bigdeli M, Ghaffar A, Røttingen J. Trends in health policyand systems research over the past decade: still too little capacity in low-income countries. PLoS One. 2011;6(11):e27263.5. Yao Q, Chen K, Yao L, et al. Scientometric trends and knowledge maps ofglobal health systems research. Health Res Policy Syst. 2014;12:26.6. English K, Pourbohloul B. Increasing health policy and systems researchcapacity in low-and-middle income countries: results from a bibliometricanalysis. Health Res Policy Syst. 2017;15:64.7. Börner K, Dall'Asta L, Ke W, Vespignani A. Studying the emerging globalbrain: analyzing and visualizing the impact of co-authorship teams.Complexity. 2005;10:57–67. doi:10.1002/cplx.20078.8. Boyack KW, Klavans R, Börner K. Mapping the backbone of science.Scientometrics. 2005;64(3):351–74.9. Herr BW, Duhon RJ, Borner K, Hardy EF, Penumarthy S. 113 Years of PhysicalReview: Using Flow Maps to Show Temporal and Topical Citation Patterns.Information Visualisation 2008. 12th International Conference. IEEE; 2008. pp.421–26; doi:10.1109/IV.2008.97.10. National Center for Biotechnology Information, U.S. National Library of Medicine.Clinical Queries Filters. PubMed Help. 2015. http://www.ncbi.nlm.nih.gov/books/NBK3827/#pubmedhelp.Clinical_Queries_Filters. Accessed 21 Aug 2017.11. National Center for Biotechnology Information, U.S. National Library ofMedicine. Combining search terms with Boolean operators (AND, OR, NOT).PubMed Help. 2015. http://www.ncbi.nlm.nih.gov/books/NBK3827/#pubmedhelp.Combining_search_terms_with_B. Accessed 21 Aug 2017.12. National Center for Biotechnology Information, U.S. National Library ofMedicine. Species. PubMed Help. 2015. http://www.ncbi.nlm.nih.gov/books/NBK3827/#pubmedhelp.Species. Accessed 21 Aug 2017.13. National Center for Biotechnology Information, U.S. National Library of Medicine.Title/Abstract [TIAB] [Internet]. PubMed Help. 2015. http://www.ncbi.nlm.nih.gov/books/NBK3827/#pubmedhelp.TitleAbstract_TIAB. Accessed 21 Aug 2017.14. Cytoscape. http://www.cytoscape.org. Accessed 21 Aug 2017.15. The World Bank. Country and Lending Groups. http://data.worldbank.org/about/country-and-lending-groups. Accessed 21 Aug 2017.•  We accept pre-submission inquiries •  Our selector tool helps you to find the most relevant journal•  We provide round the clock customer support •  Convenient online submission•  Thorough peer review•  Inclusion in PubMed and all major indexing services •  Maximum visibility for your researchSubmit your manuscript atwww.biomedcentral.com/submitSubmit your next manuscript to BioMed Central and we will help you at every step:English and Pourbohloul Health Research Policy and Systems  (2017) 15:71 Page 10 of 10


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items