TY - JOUR
AU1 - Brunson, Jason Cory
AU2 - Laubenbacher, Reinhard C
AB - Abstract Objective To survey network analyses of datasets collected in the course of routine operations in health care settings and identify driving questions, methods, needs, and potential for future research. Materials and Methods A search strategy was designed to find studies that applied network analysis to routinely collected health care datasets and was adapted to 3 bibliographic databases. The results were grouped according to a thematic analysis of their settings, objectives, data, and methods. Each group received a methodological synthesis. Results The search found 189 distinct studies reported before August 2016. We manually partitioned the sample into 4 groups, which investigated institutional exchange, physician collaboration, clinical co-occurrence, and workplace interaction networks. Several robust and ongoing research programs were discerned within (and sometimes across) the groups. Little interaction was observed between these programs, despite conceptual and methodological similarities. Discussion We use the literature sample to inform a discussion of good practice at this methodological interface, including the concordance of motivations, study design, data, and tools and the validation and standardization of techniques. We then highlight instances of positive feedback between methodological development and knowledge domains and assess the overall cohesion of the sample. network analysis, graph theory, secondary use, administrative data, EHR BACKGROUND AND SIGNIFICANCE As electronic information systems have increased in capacity, efficiency, and accessibility, digitized clinical recordkeeping has made routinely collected health care data (RCHD) of unprecedented depth, scope, and variability available to researchers. The computer science revolution behind this progress has also facilitated the development and implementation of rapid, accurate algorithms to perform calculations on mathematical graphs and sample from distributions. This has led to an explosion in applications of network analysis (NA), which have proved successful in many domains at extracting conceptual insights and predictive power from large and messy datasets. A recent systematic review of data acquisition and analysis in systems medicine identified RCHD such as electronic medical records and public databases as the most common sources of data and NA as the most common modeling paradigm.1 In this article, we review the work being done at the interface of these tools. OBJECTIVE We set out to assess the motivations, substance, contributions, and needs of network analyses of routinely collected health care datasets (NARCHD): What problems have motivated this work, and what research programs have emerged? What has this research contributed to knowledge and methodology, and what further advances can be made? The paper provides a comprehensive survey of this literature and an evaluation of ongoing projects. Our focus on methodologies complements several results-focused reviews that cover much of the same territory. Researchers seeking to build upon this knowledge base stand to benefit from a critical assessment of the methods in common use. By restricting our assessment to studies using data collected for nonresearch purposes, we showcase research designs that do not require collection of new data, which can be resource-intensive and require specialized expertise. MATERIALS AND METHODS Search strategy We operationalized our inclusion criteria as follows: The study must have made use of health care–related data that were collected as part of routine operations and were not collected or demarcated as part of an intervention. These data must have been modeled as mathematical graphs that were themselves objects of study investigated using tools from network analysis. The search protocol is summarized in Figure 1 and detailed in the supplementary material. We omitted from consideration studies whose use of NA consisted in building artificial neural networks and Bayesian networks, which we felt were distinct subfields deserving separate treatment. Figure 1. View largeDownload slide Search protocol and results. We based our keyword search on a “seed set” of publications that motivated our review, then adapted it to each database. Records were excluded based in most cases on title and abstract, and on full texts where necessary, though 8 desired full texts could not be found. Reference and citation chaining were performed twice in sequence (R, C, R, C). Figure 1. View largeDownload slide Search protocol and results. We based our keyword search on a “seed set” of publications that motivated our review, then adapted it to each database. Records were excluded based in most cases on title and abstract, and on full texts where necessary, though 8 desired full texts could not be found. Reference and citation chaining were performed twice in sequence (R, C, R, C). Analysis process While reading the final sample studies, we grouped them by commonality of study design, with an emphasis on the setting in which data were collected and the techniques used to analyze them, and to a lesser extent the motivating questions. When a study employed multiple datasets or analytical frameworks, we focused on the stage we identified as NARCHD. We then identified and described coherent research programs within each group. In addition to our qualitative analysis, we performed a scientometric summary, reported in the supplementary material. Both informed our discussion of the sample. Table 1 Table 1. Common network analysis terminology used in this manuscript Concept  Definition  Network  A system of nodes (members or parts) and links (relationships or connections) between them, usually modeled as a mathematical graph; links can be directed, weighted, or of different types, in which case the network is multilayer  Motifs  Small subgraphs consisting of a set of nodes and links among them, such as dyads (2 nodes) and triads (3 nodes)  Neighborhood  The set of nodes (alters) within a fixed number of hops along links, usually 1, from an index node (ego)  Meso-structure  Network structure that is not detectable locally (within neighborhoods) but does not require global information to detect, such as community structure (discerned from community detection, a family of node clustering methods for graphs) and distance effects  Distance effects  Phenomena that depend on hops along links through a network, such as paths (sequences of incident nodes and links), cycles (paths that end where they begin), betweenness centrality (the proportion of shortest paths on which a node lies), and closeness centrality (the reciprocal harmonic average distance from a node to other nodes)  ERGM  Exponential random graph (also p*) model, a family of recursive logistic regression models developed to measure the effects of specific generative processes, such as assortativity (homophily), transitivity, and triad closure on the global structure of social networks  Concept  Definition  Network  A system of nodes (members or parts) and links (relationships or connections) between them, usually modeled as a mathematical graph; links can be directed, weighted, or of different types, in which case the network is multilayer  Motifs  Small subgraphs consisting of a set of nodes and links among them, such as dyads (2 nodes) and triads (3 nodes)  Neighborhood  The set of nodes (alters) within a fixed number of hops along links, usually 1, from an index node (ego)  Meso-structure  Network structure that is not detectable locally (within neighborhoods) but does not require global information to detect, such as community structure (discerned from community detection, a family of node clustering methods for graphs) and distance effects  Distance effects  Phenomena that depend on hops along links through a network, such as paths (sequences of incident nodes and links), cycles (paths that end where they begin), betweenness centrality (the proportion of shortest paths on which a node lies), and closeness centrality (the reciprocal harmonic average distance from a node to other nodes)  ERGM  Exponential random graph (also p*) model, a family of recursive logistic regression models developed to measure the effects of specific generative processes, such as assortativity (homophily), transitivity, and triad closure on the global structure of social networks  defines network terminology used throughout the sample and in our discussion. RESULTS Search Our final sample comprised 138 journal articles, 52 conference presentations (papers and extended abstracts), 9 book sections, and 1 electronic preprint (see Figure 2 ). These 200 publications reported results from 189 distinct studies. In the sections below, we give an overview of each group and the major research programs within it, emphasizing projects that were primarily NARCHD. Figure 2. View largeDownload slide Number of publications each year in our sample, using the earliest date known to have been available. Figure 2. View largeDownload slide Number of publications each year in our sample, using the earliest date known to have been available. Institutional exchange networks Fifty studies analyzed interorganizational health care systems as institutional exchange networks, the most common instantiation of which were patient-transfer and patient-sharing networks. Most of these studies took a network flow approach,2 focusing on the movement of information and resources between providers, though several emphasized that established transfer patterns are an important part of health care infrastructure.3 The studies sorted roughly into 2 investigative frameworks: interorganizational network analysis,4 which views institutions as competitors and coordinators in a health care market and patient-sharing as a proxy for niche overlap or information exchange, and network epidemiology, which views patients as vectors for health care–associated infections.5,6 This is reflected in a methodological divide, with static models employed on the economical side and dynamic simulation on the epidemiological side. Major programs One research program within this group focused on the role of interhospital patient mobility in improving outcomes and increasing efficiency. Some studies sought to explain patient transfers in terms of resource and performance differentials, geography, and other factors.7–15 Others examined the structural positions of transfer partners7,11,12,16,17 and investigated transfer patterns that suggested inefficiencies.8,18 Another program tested theory-driven hypotheses of competitive interdependence, most notably similarities in performance between highly collaborative hospitals and in performance rankings between structurally equivalent ones.16,19–25 Some additional studies measured the impact of market-based incentives on administrative decisions.26,27 The largest program in the group examined how outbreaks spread through networks of patient transfer, referral, sharing, or interpersonal contact. Early observations that these networks had properties theorized to facilitate outbreaks7,28 were followed by several simulation29–36 and observational37–41 studies of the structural determinants of infection spread and the viability of surveillance42–45 and resource allocation46,47 strategies. Physician collaboration networks Another group of 33 studies modeled institutional and regional physician communities as physician collaboration networks. These studies frequently took a network architecture approach,2 emphasizing the determinants and effects of patient-sharing ties and motifs among doctors. This reflects the wider physician collaboration literature, which seeks to identify patterns associated with higher-quality care and better outcomes48–50 and the dissemination of best practices.49,51 Major programs Several studies in this group addressed the accurate measurement and prediction of patient-sharing. One compared the concordance of physician recollection and billing records52 and others used demography, geography, affiliation, and attitudes at the physician or dyad level53–58 and sociostructural tendencies55,56,59 to account for shares and referrals. Another subset linked the motifs and meso-structure of patient-sharing networks to differences in physician practice,59–63 patient outcomes,56, 64–70 and health disparities.71,72 Clinical co-occurrence networks A distinctive group of 59 studies concerned relational models of clinical events mined from patient- or encounter-level health records (clinical co-occurrence networks) most often used to generate etiological hypotheses. Half of these studies investigated graph models of disease co-occurrence,73,74 often called comorbidity networks. Other studies investigated more general networks of diagnoses, lab tests, procedures, prescriptions, ingredients, and clinical terms mined from patient encounter notes.75 Data mining and other exploratory methods dominated, with many techniques borrowed from other domains or developed ad hoc. Analyses were mostly local; only a handful invoked community structure or distance effects. Major programs Three coherent programs emerged from the studies on disease graphs. One used disease co-occurrence to identify candidate disease dependencies for lab research.76–82 Another assessed the ability of known genetic,78,79,83–86 proteomic,82,85 environmental, lifestyle,87 or a combination of factors87–89 to account for co-occurrences. A third aggregated patient timelines into temporal graphs, directed graphs in which links encode time differentials, that were used to describe patient trajectories and predict future diagnoses.90–96 Several other studies employed clustering algorithms, in some cases to recapitulate formal disease ontologies,84,96,97 but in others to develop risk-predictive measures,98 improve standard classifications,99 and stratify patients.97,100,101 Workplace interaction networks The remaining 47 studies fit into a highly modular group investigating workplace interaction networks, including network models of shared record access, interprovider communication, interpersonal contacts, and patient handoffs. Most of these studies were institution-specific and drew from health information system access or process logs. Dominant research themes used signature NA methods, eg, process mining to model workflow.102 Some of these modules were integrated, by citation or common authorship, into 1 of the groups discussed previously. Most others lay within a larger literature on the social network analysis (SNA) of health care wards and teams.103 Major programs A unique program appearing in this group concerned anomaly detection. These studies used data from user access logs and focused on inappropriate access to patient records104–107 and fraud detection.108 Another subset used similar datasets to model infection spread through contact networks. By incorporating patient-staff contact,109,110 surveillance cultures,109 and spatial proximity,111,112 these studies extended the scope of the epidemiological studies discussed above. The largest program emerged from the process management literature. These studies used patient handoff data to characterize the interactions of care teams113–117 and ward staffs,118–122 and to identify prototypical clinical pathways.107,120,123–128 DISCUSSION The studies included in our sample took a wide range of approaches, summarized in Table 2 Table 2. Summary of conceptual and methodological frameworks of journal articles in our sample Conceptual framework  Methodological framework  Studies  NA  NA  Anderson and Talsma (2011),118 Hripcsak et al. (2011),129 Lee et al. (2011),14 Lee et al. (2011)130  NA  ERGM  Moen et al. (2016)59  NA  Hypothesis generation  Hanauer et al. (2009),77 Hanauer and Ramakrishnan (2013)93  NA  Mixed methods  Barnett et al. (2012)57  NA  Regression  Boyer et al. (2005),13 Landon et al. (2012)54  NA  Rule mining  Malin et al. (2011)131  NA  Text mining  Finlayson et al. (2014)132  Care coordination  NA  Pham et al. (2009),133 Gray et al. (2010),121 Siden and Urbanoski (2011),113 Uddin and Hossain (2011),114 Mandl et al. (2014),134 Uddin and Hossain (2014),135 Merrill et al. (2015),127 Soulakis et al. (2015),116 Uddin et al. (2015)117  Care coordination  Mixed methods  Chen et al. (2014)122  Care coordination  Regression  Barnett et al. (2012),66 Pollack et al. (2013),115 Spear (2014),136 Casalino et al. (2015),69 Ong et al. (2016),62 Uddin (2016)137  Clinical ontology  NA  Aprile et al. (2008),138 Botsis et al. (2015)139  Clinical ontology  Complexity reduction  Lyalina et al. (2013),140 Jing and Cimino (2014)141  Clinical ontology  Feature extraction  Chen et al. (2015)142  Collaboration and competition  Agent-based modeling  Mascia and Di Vincenzo (2013)24  Collaboration and competition  ERGM  Lomi and Pallotti (2012),16 Pallotti et al. (2013)25  Collaboration and competition  Regression  Pallotti and Lomi (2011),19 Mascia et al. (2012),23 Mascia et al. (2015),17 Pallotti et al. (2015),22 Lee et al. (2016),27 Tranmer et al. (2016)21  Collaboration and competition  Stochastic modeling  Stadtfeld et al. (2016)26  Collaborative practice  NA  Manuel et al. (2011),143 Uddin et al. (2012),67 Landon et al. (2013),144 Lubloy et al. (2016)65  Collaborative practice  ERGM  Uddin et al. (2013),145 Paul et al. (2014)55  Collaborative practice  Feature extraction  Zhang et al. (2015)128  Collaborative practice  Mixed methods  Barnett et al. (2011)52  Collaborative practice  Regression  Pollack et al. (2014)70  Collaborative practice  Survival analysis  Lomi et al. (2014),12 Hussain et al. (2015)64  Comorbidity  NA  Kim et al. (2016),146 Liu et al. (2016)147  Comorbidity  Complexity reduction  Schafer et al. (2014)148  Comorbidity  Feature extraction  Sideris et al. (2016)98  Comorbidity  Natural language processing  Roque et al. (2011)97  Comorbidity  Software development  Moni and Lio (2015)87  Decision support  Natural language processing  Nikfarjam et al. (2013)149  Decision support  Process mining  Rossille et al. (2008)123  Decision support  Rule mining  Zhou et al. (2010)150  Decision support  Software development  Heer and Perer (2014),151 Li et al. (2015),152 Warner et al. (2015)153  Disease progression  NA  Chmiel et al. (2014),154 Jensen et al. (2014)90  Disease progression  Hypothesis generation  Hidalgo et al. (2009),92 Kannan et al. (2016)95  Disease progression  Probabilistic modeling  Chen et al. (2009)91  Disease progression  Survival analysis  Xu et al. (2015)89  Epidemiology  NA  Liljeros et al. (2007),28 Donker et al. (2010),29 Huang et al. (2010),38 Walker et al. (2012),39 Ohst et al. (2014),34 Geraci et al. (2016),155 Takahashi et al. (2016)63  Epidemiology  Agent-based modeling  Lee et al. (2011),30 Donker et al. (2012),31 Lee et al. (2012),42 Curtis et al. (2013),111 Bartsch et al. (2014),33 Donker et al. (2014),32 van Bunnik et al. (2015)45  Epidemiology  Regression  Geva et al. (2011),109 Ke et al. (2012),37 Simmering et al. (2015),41 Gibbons et al. (2016)40  Epidemiology  Stochastic modeling  Ueno and Masuda (2008),110 Karkada et al. (2011),46 Lesosky et al. (2011),35 Cusumano-Towner et al. (2013),112 Ciccolini et al. (2014),43 van den Dool et al. (2016)36  Health surveillance  NA  Ball and Botsis (2011),156 Patel and Kaelber (2014),157 Scott et al. (2014),158 Franchini et al. (2015)159  Health surveillance  Text mining  Roitmann et al. (2014)101  Inappropriate access  Anomaly detection  Chen et al. (2012),105 Chen et al. (2012),104 Zhang et al. (2013),107 Menon et al. (2014)106  Molecular biology of disease  NA  Park et al. (2009),83 Davis and Chawla (2011),84 Park et al. (2012),85 Paik et al. (2014)82  Molecular biology of disease  Hypothesis generation  Bagley et al. (2016)79  Molecular biology of disease  Probabilistic modeling  Rzhetsky et al. (2007),76 Blair et al. (2013)86  Molecular biology of disease  Rule mining  Chen and Xu (2014)160  Molecular biology of disease  Software development  Liu et al. (2014)88  Organizational effectiveness  NA  Minerba et al. (2008),161 Iwashyna et al. (2009),7 Iwashyna et al. (2009),11 Puggioni et al. (2011),9 Abbasi et al. (2012),162 Tighe et al. (2014)119  Organizational effectiveness  Mixed methods  Veinot et al. (2012)15  Organizational effectiveness  Process mining  Baumgart et al. (2009),120 Rebuge and Ferreira (2012)125  Organizational effectiveness  Regression  Iwashyna et al. (2010),8 Butala et al. (2015)163  Population health  NA  Feldman et al. (2016),164 Glicksberg et al. (2016)94  Population health  Regression  Hollingsworth et al. (2015)71  Population health  Rule mining  Holmes et al. (2011)165  Social capital and social influence  NA  Kwan et al. (2015)166  Social capital and social influence  ERGM  Fattore and Salvatore (2010)167  Social capital and social influence  Regression  Fattore et al. (2009),168 Pollack et al. (2012),60 Hackl et al. (2015),58 Pollack et al. (2015),61 Geissler et al. (2016)72  Conceptual framework  Methodological framework  Studies  NA  NA  Anderson and Talsma (2011),118 Hripcsak et al. (2011),129 Lee et al. (2011),14 Lee et al. (2011)130  NA  ERGM  Moen et al. (2016)59  NA  Hypothesis generation  Hanauer et al. (2009),77 Hanauer and Ramakrishnan (2013)93  NA  Mixed methods  Barnett et al. (2012)57  NA  Regression  Boyer et al. (2005),13 Landon et al. (2012)54  NA  Rule mining  Malin et al. (2011)131  NA  Text mining  Finlayson et al. (2014)132  Care coordination  NA  Pham et al. (2009),133 Gray et al. (2010),121 Siden and Urbanoski (2011),113 Uddin and Hossain (2011),114 Mandl et al. (2014),134 Uddin and Hossain (2014),135 Merrill et al. (2015),127 Soulakis et al. (2015),116 Uddin et al. (2015)117  Care coordination  Mixed methods  Chen et al. (2014)122  Care coordination  Regression  Barnett et al. (2012),66 Pollack et al. (2013),115 Spear (2014),136 Casalino et al. (2015),69 Ong et al. (2016),62 Uddin (2016)137  Clinical ontology  NA  Aprile et al. (2008),138 Botsis et al. (2015)139  Clinical ontology  Complexity reduction  Lyalina et al. (2013),140 Jing and Cimino (2014)141  Clinical ontology  Feature extraction  Chen et al. (2015)142  Collaboration and competition  Agent-based modeling  Mascia and Di Vincenzo (2013)24  Collaboration and competition  ERGM  Lomi and Pallotti (2012),16 Pallotti et al. (2013)25  Collaboration and competition  Regression  Pallotti and Lomi (2011),19 Mascia et al. (2012),23 Mascia et al. (2015),17 Pallotti et al. (2015),22 Lee et al. (2016),27 Tranmer et al. (2016)21  Collaboration and competition  Stochastic modeling  Stadtfeld et al. (2016)26  Collaborative practice  NA  Manuel et al. (2011),143 Uddin et al. (2012),67 Landon et al. (2013),144 Lubloy et al. (2016)65  Collaborative practice  ERGM  Uddin et al. (2013),145 Paul et al. (2014)55  Collaborative practice  Feature extraction  Zhang et al. (2015)128  Collaborative practice  Mixed methods  Barnett et al. (2011)52  Collaborative practice  Regression  Pollack et al. (2014)70  Collaborative practice  Survival analysis  Lomi et al. (2014),12 Hussain et al. (2015)64  Comorbidity  NA  Kim et al. (2016),146 Liu et al. (2016)147  Comorbidity  Complexity reduction  Schafer et al. (2014)148  Comorbidity  Feature extraction  Sideris et al. (2016)98  Comorbidity  Natural language processing  Roque et al. (2011)97  Comorbidity  Software development  Moni and Lio (2015)87  Decision support  Natural language processing  Nikfarjam et al. (2013)149  Decision support  Process mining  Rossille et al. (2008)123  Decision support  Rule mining  Zhou et al. (2010)150  Decision support  Software development  Heer and Perer (2014),151 Li et al. (2015),152 Warner et al. (2015)153  Disease progression  NA  Chmiel et al. (2014),154 Jensen et al. (2014)90  Disease progression  Hypothesis generation  Hidalgo et al. (2009),92 Kannan et al. (2016)95  Disease progression  Probabilistic modeling  Chen et al. (2009)91  Disease progression  Survival analysis  Xu et al. (2015)89  Epidemiology  NA  Liljeros et al. (2007),28 Donker et al. (2010),29 Huang et al. (2010),38 Walker et al. (2012),39 Ohst et al. (2014),34 Geraci et al. (2016),155 Takahashi et al. (2016)63  Epidemiology  Agent-based modeling  Lee et al. (2011),30 Donker et al. (2012),31 Lee et al. (2012),42 Curtis et al. (2013),111 Bartsch et al. (2014),33 Donker et al. (2014),32 van Bunnik et al. (2015)45  Epidemiology  Regression  Geva et al. (2011),109 Ke et al. (2012),37 Simmering et al. (2015),41 Gibbons et al. (2016)40  Epidemiology  Stochastic modeling  Ueno and Masuda (2008),110 Karkada et al. (2011),46 Lesosky et al. (2011),35 Cusumano-Towner et al. (2013),112 Ciccolini et al. (2014),43 van den Dool et al. (2016)36  Health surveillance  NA  Ball and Botsis (2011),156 Patel and Kaelber (2014),157 Scott et al. (2014),158 Franchini et al. (2015)159  Health surveillance  Text mining  Roitmann et al. (2014)101  Inappropriate access  Anomaly detection  Chen et al. (2012),105 Chen et al. (2012),104 Zhang et al. (2013),107 Menon et al. (2014)106  Molecular biology of disease  NA  Park et al. (2009),83 Davis and Chawla (2011),84 Park et al. (2012),85 Paik et al. (2014)82  Molecular biology of disease  Hypothesis generation  Bagley et al. (2016)79  Molecular biology of disease  Probabilistic modeling  Rzhetsky et al. (2007),76 Blair et al. (2013)86  Molecular biology of disease  Rule mining  Chen and Xu (2014)160  Molecular biology of disease  Software development  Liu et al. (2014)88  Organizational effectiveness  NA  Minerba et al. (2008),161 Iwashyna et al. (2009),7 Iwashyna et al. (2009),11 Puggioni et al. (2011),9 Abbasi et al. (2012),162 Tighe et al. (2014)119  Organizational effectiveness  Mixed methods  Veinot et al. (2012)15  Organizational effectiveness  Process mining  Baumgart et al. (2009),120 Rebuge and Ferreira (2012)125  Organizational effectiveness  Regression  Iwashyna et al. (2010),8 Butala et al. (2015)163  Population health  NA  Feldman et al. (2016),164 Glicksberg et al. (2016)94  Population health  Regression  Hollingsworth et al. (2015)71  Population health  Rule mining  Holmes et al. (2011)165  Social capital and social influence  NA  Kwan et al. (2015)166  Social capital and social influence  ERGM  Fattore and Salvatore (2010)167  Social capital and social influence  Regression  Fattore et al. (2009),168 Pollack et al. (2012),60 Hackl et al. (2015),58 Pollack et al. (2015),61 Geissler et al. (2016)72 , which reflected differences in their conceptual and material needs. In this section we document some strengths and weaknesses of these studies and suggest some standards of good practice. We focus on 2 fronts: the choice of framework and the model construction and validation. We then discuss several achievements and needs, focusing on the mutual benefits to knowledge and methodology and on the overall cohesion of the sample. Framework assignments were subjective, and rare assignments were combined for ease of reference. Every study incorporated (social) network analysis conceptually and methodologically; when we identified no more specific framework, we listed “NA.” Choice of framework Every analysis decision requires justification, starting with the choice of framework. The network conceptual model is most illustratively called into question in the clinical co-occurrence setting: There are widely recognized problems with collapsing co-occurrence data to unipartite network models, but few studies of the “diseaseome” addressed them. The most popular use of these models was visualization; indeed, several studies analyzed co-occurrence data geometrically, for instance by using hierarchical clustering rather than community detection, but visualized them as graphs.97,101,169 Force-directed layout algorithms exploit the binary nature of graphs and can place unlinked dyads at any distance and orientation from each other; for count data such as these, visualizations like correspondence analysis biplots that minimize distortion are arguably more appropriate.170 Additionally, a defining element of NA is the conceptualization of transmission channels as links and of functional dependencies as motifs, which predicts distance effects between nonadjacent nodes and dependencies between node attributes and neighborhood characteristics. These concepts do not follow as naturally from disease co-occurrence as from, say, patient-sharing. Some studies demonstrated the utility of visualizations for electronic dashboards, though only at the neighborhood level.80,116,153 While a few studies made valuable use of distance effects81,90,91 or motif mining,84 most were primarily dyadic. Thus, despite their conceptual popularity, disease networks themselves have received little analytic attention. Once a framework is adopted, consequential considerations remain, among them concordance between the theoretical constructs of interest and the measures used to detect them. In the (S)NA setting, some constructs will be structural, and the corresponding measures should be theoretically grounded. For example, early epidemiological studies identified features of contact and transfer networks known from existing work to have implications for infection spread.11,28 Some studies employed network measures without specific motivation, eg, degree and betweenness centrality as possible social determinants of methadone treatment continuation.166 In this case, a discernible effect of degree was given a reasonable interpretation, but an indiscernible effect of betweenness was not commented upon; had it been theoretically motivated, an account of this result would clearly be required. Occasionally, an SNA framing seemed incidental to the analysis being conducted. For example, the number of visits a doctor makes to a patient and the share of the visit costs coming from hospital claims might be expected to predict total visit costs by indicating severity of illness or resource use; when termed “connectedness” and “tie strength,” they instead suggested a causal relationship with care team coordination.114 In this case, SNA theory seemed ill-suited to the hypotheses being tested. Model construction and validation Network models were constructed, with few exceptions, by linking nodes according to 3 data patterns: co-occurrence on records (physicians’ patients, patient diagnoses), sequential occurrence on chronological records (patient admissions and discharges, staff HIS access), and source/target designation on transmission records (patient transfers, staff handoffs). Higher-order multipartite structures were often reduced to simple graph models for conceptual or computational reasons. For example, early interorganizational analyses relied on network measures of collaboration and competition, for which chronological patient-hospital admission/discharge data were flattened to unipartite graphs.16,19,20,23,25 More recent contributions expanded their scope to multilevel models of organizations and activities26 and of organizations and their departments,21 in both cases grounding the motifs of interest in theory and building statistical inference models around them. Studies not strongly grounded in theory usually focused on machine learning–based prediction, using a wider range of motifs in higher-order incidence structures and developing efficient algorithms to mine for them. This is evident from several studies that mined these structures for recurring patterns indicative of standard protocols128,171 or for deviations suggestive of security breaches104,105 or fraudulent practices.108,172 Studies collapsed higher-order incidence structures into more manageable graphs using various techniques, none of which have become domain standards. In the research program on physician patient-sharing, for example, different studies adopted different statistics and thresholds on 2 physicians’ shared patients for link determination, including raw counts,70 mutual percentages,55 caseload-relative thresholds,54 and average numbers of visits by shared patients.59 Graph properties are sensitive to these decisions, as illustrated by 2 inaugural studies that sought to quantify the number of alters with whom an ego physician must coordinate care. Based on shared Medicare patients for whom either ego or alter was deemed primary, a representative sample of physicians in the United States were found to have typically hundreds of peers.133 In contrast, based on ego and alter seeing common patients at ≥1% each of their total visits, physicians within professional groups in Ontario had on average 2.2 peers.143 A related concern is the dependence of graph models on data sources, each having its own scope, bias, and format. Comorbidity graphs were constructed using patient-level data from several institution-specific electronic health records (EHRs)77,79,80,84,147 and from both public and private billing claims,83,86,146,148,154,173 and using incident-level data from adverse event reports.78 “Co-occurrence” itself ranged in meaning from a simple correlation92 or odds ratio146 to cosine similarity97,101 to a statistical signal from a probabilistic model,76 and the aggregated parts were not always dyads.78,148 These models, of the same theoretical construct, were never directly compared. A few validation studies did address such concerns. An early study of physician collaboration networks estimated the accuracy with which the number of Medicare patients 2 physicians share predicts their self-reporting of a professional relationship, using different raw-number thresholds,52 and a follow-up study characterized physicians’ stated reasons for referring patients to their data-mined colleagues.57 A similar study examined the decision process whereby clinical teams at sending hospitals maintained primary transfer relationships with recipient hospitals.15 A later study of health care worker interactions compared the relative likelihood that patients’ records would be accessed by staff in different departments with those staff members’ expectations.122 By drawing from nationwide patient admission datasets from 2 countries, 1 study of infection surveillance was able to report on the robustness of their optimal strategy to the difference between the health care systems.43 With respect to disease graphs, 2 studies compared the sets of comorbid pairs identified using the same link determination rule on multiple datasets,79,86 and 1 matched observed comorbidities with co-occurrences in the medical literature.165 Several additional studies discussed biases in their data and tested the sensitivity of their results to different sources and thresholds, but these efforts were far narrower than the breadth of methods used, and none discussed concomitant differences in the resulting network structure. Knowledge gains and methods development Several studies used (S)NA to advance knowledge in ways that other methods would not be expected to, eg, invoking distance effects and motifs. Here is a partial list: An investigation of hospital transfers identified “cascades” of inefficiencies, measured as temporal paths through a transfer network in which a transfer by one sending hospital to a nonprimary recipient apparently results in a primary sending partner of the recipient redirecting a transfer of its own, and so on.18 One approach to accounting for population comorbidities using genetic commonalities was to perform a triad census on the multilevel network of population comorbidities and genetic overlaps.84 Another was to select diseases of high betweenness with respect to 2 index diseases (obesity and colorectal cancer) and identify their shared genetic associations.81 One application for co-occurring clinical text fields was to distinguish “phenotypic signatures” of mental illnesses, which pose difficulties for criteria-based diagnosis.140 A study of operating room staff interactions following a layout redesign observed a shift from sequential to parallel performance of perioperative tasks, suggesting new vulnerabilities to staffing changes.120 A study in a neonatal intensive care unit linked longer handoff cycles to lower reported patient satisfaction, suggesting a quantitative measure of care continuity.121 Despite its inherent limitations, RCHD was likewise often well-suited to a study’s needs. One instance was the rigorous testing of hypotheses about interorganizational collaboration and competition derived from economic theory, which was made possible in the health care sector by the exceptionally thorough and reliable documentation on both information and resource exchange (patient transfers) and quality of performance (discharge rates or lengths of stay) contained in administrative datasets.19,23 Another was the use of empirical data on institutional infection rates or individual cultures by several inter-29,31,34,37,40,45 and intraorganizational109,155 epidemiological modeling studies, respectively, to calibrate or validate their network models. Third, patient-sharing networks provided an exceptionally detailed setting in which the social diffusion of practice could be measured.59,61,62 These studies focused on relationships between structural and attributive variables, which often require considerable expense and effort to generate but are common elements of RCHD. Conversely, though most studies used NA tools “out of the box,” several domain-specific research questions led to advances in methodology, attesting to the reciprocal value to network science of health care applications. The need to adapt individual-level epidemiological models to populated institutions led to fractional models of immunization,47 and the simulation framework (and sometimes the software) developed for these studies was also used to test surveillance42–45 and resource allocation46,47 strategies. Predictive modeling of population comorbidity spurred the development of tools for exploration, link prediction, and stratification in multilayer disease networks combined from EHR and molecular data.84,87,88 Efforts to model disease progression and clinical workflow inspired several original aggregations of ordered pairs and longer chronological disease sequences into temporal graphs,90,91,128,174 a procedure still young in network science.175 In the domain of interorganizational networks, the availability of exceptionally thorough and reliable documentation on patient admissions and transfers made health care a testing ground for novel graph-theoretical statistical inference methods, including new specifications of ERGMs,16,25 extensions to stochastic actor-oriented models,26 and adaptations of multiple membership–multiple classification models to multilevel networks.21 Cohesion While research at this interface shows no signs of slowing (Figure 2), a scientometric analysis (see supplementary material) showed that the literature is highly fragmented, roughly along the groups identified above. The research programs identified above often had “home journals” where several research teams published, for instance Medical Care for physician collaboration and Infection Control and Hospital Epidemiology for patient transfer epidemiology, but we found very little overlap in venues between programs, and no central hub for NARCHD. Several early publications that anticipated the approaches that later gained traction were not cited by any others in our sample,9,10,13,113,123,143,161,176,177 nor did they have any authors in common with them (until recently79). Seldom, too, did active teams form new collaborations.86,174,178 A stepwise ERGM of the graph of citations failed to explain the extent of fragmentation based on publication dates, common authorship, and simple generative processes.179,180 This fragmentation is not so surprising from a young field, but is also in part an artifact of our review criteria. The cohesive research programs described above, as evidenced by several domain-focused reviews,5,6,48–51,74,75,102,103,181,182 are merely the contours along which much larger research programs intersect our sample. The vast majority of SNA in health care settings relies on original survey data,103 and the scale of the health care environment amenable to network modeling far exceeds that of RCHD.183 Our sample may therefore underestimate the cohesion of this research community. We did observe several instances of corroborations and useful comparisons with previous work. Several regression and ERGM analyses (and some visual inspections) observed that patients tended to be transferred to target hospitals with more care-relevant resources7–10 and better patient outcomes,8,11–13 and the best-equipped and -performing target hospitals tended to be more central.7,11,17 Transfer partnerships were reliably associated with competitive interdependence, even after controlling for geography, attributive similarity, and generative processes.16,23,24 Several interorganizational epidemiological studies corroborated associations between a hospital’s in-degree and infection rate29–31,34,40,41 and between its overall connectedness and overall pathogen prevalence30,33,37 (with 1 exception35). Some applications saw interestingly inconsistent results: One study identified cohesive communities from the commonalities in conditions diagnosed by an institution’s providers, which reflected departmental structure in some ways but not others.161 Another study mined access logs to produce a communication network dominated by multispecialty teams, more consistent with cohesive floor staff and service teams,129 and a third observed from a similar dataset that collaboration was more prevalent within departments than between them.162 How best to identify and characterize cohesive units from institution HIS data appears to be an open question. Yet researchers in different domains faced many similar challenges, and opportunities may have been missed to build upon each other’s progress and to recombine techniques developed independently. A case in point is the aggregation of temporal sequences of clinical events into a directed graph. From a clinical informatics perspective, Patnaik et al.174 explored a top-down approach starting with a partial order on a collection of diagnosis and procedure codes, before Hanauer and Ramakrishnan93 (of the same team) adopted a bottom-up approach of statistically identifying individual temporal pairs. From a data security perspective, H. Zhang et al.107 constructed a temporal graph from access logs for the purpose of characterizing routine behavior and identifying deviations from it, by aggregating all patient record–specific access sequences, analogous to Patnaik et al. At a more granular level, Nikfarjam et al.149 developed a text-mining technique to encode clinical notes into encounter-level temporal graphs. From a prognosis perspective, Liu et al.126 constructed similar patient-level temporal graphs from timelines of coded clinical events, which they clustered using feature extraction methods from linear algebra, and Jensen et al.90 took the approach, an extension of Zhang et al.’s, of first clustering patient timelines and then aggregating these in each cluster into a cohort-level temporal graph. Meanwhile, more sophisticated tools to produce similar summary structures had been developed and implemented by business researchers for process management and applied to EHR process logs by a different community of researchers.120,124,184–186 Y. Zhang et al.128 proposed a clustering-based alternative to these standard tools, much in the spirit of Jensen et al., and Finney et al.187 adopted an existing theoretical and methodological framework within which to contribute an efficient implementation. None of these studies cited those that came before (excluding the process mining studies). In summary, the disconnectedness and inconsistency of this literature suggests the possible benefit to be gained from detailing how the motivating problems, study designs, data sources, and network tools reviewed here “hang together.” FUNDING This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sector. COMPETING INTERESTS The authors have no competing interests to declare. CONTRIBUTORS JCB and RL conceived and designed the review. JCB designed and conducted the search. JCB and RL designed, and JCB performed, the synthesis. JCB designed and performed the scientometric analysis. JCB and RL contributed to the discussion, wrote and approved the final manuscript, and take responsibility for the integrity of the work. SUPPLEMENTARY MATERIAL Supplementary material is available at Journal of the American Medical Informatics Association online. Acknowledgments We thank Jenny Miglus for assistance in constructing and refining literature searches and 2 anonymous reviewers for suggestions that greatly improved the manuscript. APPENDIX: CASE STUDIES Institutional exchange networks In an effort to structurally characterize hospitals that suffered the highest rates of methicillin-resistant Staphylococcus aureus (MRSA), Ohst et al. (2014)34 drew upon a comprehensive regional dataset of care episodes to construct a contact network of care units based on patient transfers and to assign each unit a MRSA prevalence rate based on the observed frequency of positive tests. As candidate measures of each unit’s structural position, the authors used 5 notions of centrality: in-degree and daily patient turnover (based on immediate contacts), weighted and unweighted betweenness (based on shortest paths), and PageRank (based on random walks). The contact network evolved over time, so units’ centralities were calculated as of the time of each positive test. Averaging their results over the different strains of MRSA, the authors observed 2 paradoxical trends: more central units were more likely to have at least 1 positive test, but those that did had lower prevalence on average. This led the authors to conclude that the net relationship between centrality and prevalence has limited practical value. Simulations confirmed the relationship and further showed that the measures based on immediate contacts outperformed those based on geodesics (betweenness) or walks (PageRank), and were themselves outperformed by the network-agnostic attribute of patient turnover. Physician collaboration networks Moen et al. (2016)59 provide a richly detailed case study of physician collaboration, as reconstructed on the basis of patient-sharing observed in Medicare claims. The authors tailored their analysis, including the selection of comparison regions, the sample of beneficiaries and physicians, and the choice of network measures and models, to the question of why regional health care networks differ in their adherence to evidence-based defibrillator guidelines. Building upon a literature with mixed emphasis on physician-, hospital-, and region-level network structure, they incorporated effects of individual and structural attributes at the physician and hospital levels. Their multistage design used a patient-level logistic regression framework with physician-level and hospital-level random effects, and incorporated as predictors of interest the hospital referral region and several hospital-, physician-, and patient-level covariates. Among these were estimated effects of homophily by physician specialty, controlling for the distribution of colleague counts (degree distribution), and any of several measures of the centrality of hospitals within the region and of physicians within hospitals. Among other findings, they concluded that physicians’ frequency of patient-sharing (node strength) and closeness centrality in their hospital together accounted for much of the variation between hospital referral regions, indicating that the social positions of physicians among their colleagues is important to their institution’s ability to articulate good practice. Clinical co-occurrence networks In an early contribution, Davis and Chawla (2011)84 demonstrated that data-driven structural analysis could generate useful targets for experimental investigation. From the medical histories of patients served by a regional health system, they constructed a multimorbidity graph in which a link between 2 diseases was weighted by mutual information. Analogously, they constructed a disease graph whose links indicated shared genetic factors, based on gene-disease associations compiled from the experimental literature. (The most frequent database used for this purpose is the Online Mendelian Inheritance in Man, www.omim.org.) Their analysis of the combined multilayer disease network yielded several results: First, population-level co-occurrence (“phenotypic links”) correlated strongly with known genetic overlap (“genetic links”), corroborating the genetic deterministic paradigm. Second, roughly the same subsets of nodes in both networks formed strongly linked communities, which recapitulated clinically meaningful disease groups. Finally, a probabilistic link prediction model, built on a generalization of the triad census, revealed that a combination of phenotypic and (known) genetic links achieved greater predictive accuracy of unknown genetic links than either source alone. Workplace interaction networks Zhang et al. (2013)107 addressed a security problem that distinguishes EHRs from many other information systems: The frequency with which users will need to violate any manually curated set of access rules, in order to maintain patient care in hectic and unpredictable circumstances, makes postliminary auditing impractical. They proposed aggregating patient-specific sequences of EMR access into temporal graph models, for which they derived statistics to measure the irregularity of a given access or access sequence with respect to the corpus. Whereas the temporal graphs would be constructed iteratively from continually generated access logs, this approach is learning-based but relies on an assumption that care processes are consistent over time. A test of the method on an institutional EHR revealed that rates of irregularity, and of the success of their approach, varied across hospital services, accentuating the need to carefully tailor anomaly detection methods. References 1 Gietzelt M, Löpprich M, Karmen C, Knaup P, Ganzinger M. Models and data sources used in systems medicine. A systematic literature review. Methods Inf Med.  2016; 55 2: 107– 13. Google Scholar CrossRef Search ADS PubMed  2 Borgatti SP, Lopez-Kidwell V. Network theory. In Scott J, Carrington PJ, eds. The SAGE Handbook of Social Network Analysis . London: SAGE Publications, Ltd; 2011. 3 Iwashyna TJ. The incomplete infrastructure for interhospital patient transfer. Crit Care Med.  2012; 40 8: 2470– 78. Google Scholar CrossRef Search ADS PubMed  4 Bergenholtz C, Waldstrøm C. Inter-organizational network studies—a literature review. Industry Innovation.  2011; 18 6: 539– 62. Google Scholar CrossRef Search ADS   5 Danon L, Ford AP, House T. Networks and the Epidemiology of Infectious Disease. Interdisciplinary Perspect Infect Dis.  2011; 2011: 28. 6 van Kleef E, Robotham JV, Jit M, Deeny SR, Edmunds WJ. Modelling the transmission of healthcare associated infections: a systematic review. BMC Infect Dis.  2013; 13 1: 1– 13. Google Scholar CrossRef Search ADS PubMed  7 Iwashyna TJ, Christie JD, Kahn JM, Asch DA. Uncharted paths: hospital networks in critical care. Chest.  2009; 135 3: 827– 33. Google Scholar CrossRef Search ADS PubMed  8 Iwashyna TJ, Kahn JM, Hayward RA, Nallamothu BK. Interhospital transfers among Medicare beneficiaries admitted for acute myocardial infarction at nonrevascularization hospitals. Circulation: Cardiovasc Qual Outcomes.  2010; 3 5: 468– 75. Google Scholar CrossRef Search ADS   9 Puggioni D, Cappellini G, Di Bisceglie M, Cappellini C, Minerba L. Use of the complex network theory to study inpatient flow between health facilities in Sardinia (Italy). Igiene e Sanita Pubblica.  2011; 67 3: 253– 79. Google Scholar PubMed  10 Chanut C, Boyer L, Robitail S, et al.   Applying social network analysis to the health system. Sante Publique.  2005; 17 3: 403– 15. Google Scholar CrossRef Search ADS PubMed  11 Iwashyna TJ, Christie JD, Moody J, Kahn JM, Asch DA. The structure of critical care transfer networks. Medical Care.  2009; 47 7: 787– 93. Google Scholar CrossRef Search ADS PubMed  12 Lomi A, Mascia D, Vu DQ, Pallotti F, Conaldi G, Iwashyna TJ. Quality of care and interhospital collaboration: a study of patient transfers in Italy. Medical Care.  2014; 52 5: 407– 14. Google Scholar CrossRef Search ADS PubMed  13 Boyer L, Chanut C, Horte C, Mabriez JC, Auquier P. Analysis of transfer from one hospital to another in patient with myocardial infarction in the Provence–Alpes–Cote-d’Azur. Annales de Cardiologie et d’Angeiologie.  2005; 54 5: 233– 40. Google Scholar CrossRef Search ADS PubMed  14 Lee BY, McGlone SM, Song Y, et al.   Social network analysis of patient sharing among hospitals in Orange County, California. Am J Public Health.  2011; 101 4: 707– 13. Google Scholar CrossRef Search ADS PubMed  15 Veinot TC, Bosk EA, Unnikrishnan KP, Iwashyna TJ. Revenue, relationships and routines: The social organization of acute myocardial infarction patient transfers in the United States. Soc Sci Med.  2012; 75 10: 1800– 10. Google Scholar CrossRef Search ADS PubMed  16 Lomi A, Pallotti F. Relational collaboration among spatial multipoint competitors. Soc Networks.  2012; 34 1: 101– 11. Google Scholar CrossRef Search ADS   17 Mascia D, Angeli F, Di Vincenzo F. Effect of hospital referral networks on patient readmissions. Soc Sci Med.  2015; 132: 113– 21. Google Scholar CrossRef Search ADS PubMed  18 Unnikrishnan KP, Patnaik D, Iwashyna TJ. Spatio-temporal Structure of US Critical Care Transfer Network. In AMIA Joint Summits on Translational Science . 2011; 2011: 74– 78. 19 Pallotti F, Lomi A. Network influence and organizational performance: The effects of tie strength and structural equivalence. Eur Manag J.  2011; 29 5: 389– 403. Google Scholar CrossRef Search ADS   20 Mascia D, Di Vincenzo F. Understanding hospital performance: The role of network ties and patterns of competition. Health Care Manag Rev.  2011; 36 4: 327– 37. Google Scholar CrossRef Search ADS   21 Tranmer M, Pallotti F, Lomi A. The embeddedness of organizational performance: Multiple Membership Multiple Classification Models for the analysis of multilevel networks. Soc Networks.  2016; 44: 269– 80. Google Scholar CrossRef Search ADS   22 Pallotti F, Tubaro P, Lomi A. How far do network effects spill over? Evidence from an empirical study of performance differentials in interorganizational networks. Eur Manag Rev.  2015; 12 3: 189– 208. Google Scholar CrossRef Search ADS   23 Mascia D, Di Vincenzo F, Cicchetti A. Dynamic analysis of interhospital collaboration and competition: Empirical evidence from an Italian regional health system. Health Policy.  2012; 105( 2–3): 273– 81. Google Scholar CrossRef Search ADS PubMed  24 Mascia D, Di Vincenzo F. Dynamics of hospital competition: social network analysis in the Italian National Health Service. Health Care Manag Rev.  2013; 38 3: 234– 47. Google Scholar CrossRef Search ADS   25 Pallotti F, Lomi A, Mascia D. From network ties to network structures: exponential Random Graph Models of interorganizational relations. Qual Quant.  2013; 47 3: 1665– 85. Google Scholar CrossRef Search ADS   26 Stadtfeld C, Mascia D, Pallotti F, Lomi A. Assimilation and differentiation: A multilevel perspective on organizational and network change. Soc Networks.  2016; 44: 363– 74. Google Scholar CrossRef Search ADS   27 Lee K-H, Lim S, Park J. Expelled uninsured patients in a less-competitive hospital market in Florida, USA. Int J Equity Health.  2016; 15 1: 1– 9. Google Scholar CrossRef Search ADS PubMed  28 Liljeros F, Giesecke J, Holme P. The Contact Network of Inpatients in a Regional Healthcare System. A Longitudinal Case Study. Mathematical Population Stud.  2007; 14 4: 269– 84. Google Scholar CrossRef Search ADS   29 Donker T, Wallinga J, Grundmann H. Patient referral patterns and the spread of hospital-acquired infections through national health care networks. PLoS Comput Biol.  2010; 6 3: e1000715. Google Scholar CrossRef Search ADS PubMed  30 Lee BY, McGlone SM, Wong KF, et al.   Modeling the spread of methicillin-resistant Staphylococcus aureus (MRSA) outbreaks throughout the hospitals in Orange County, California. Infect Control Hospital Epidemiol.  2011; 32 6: 562– 72. Google Scholar CrossRef Search ADS   31 Donker T, Wallinga J, Slack R, Grundmann H. Hospital networks and the dispersal of hospital-acquired pathogens by patient transfer. PLoS One.  2012; 7 4: e35002. Google Scholar CrossRef Search ADS PubMed  32 Donker T, Wallinga J, Grundmann H. Dispersal of antibiotic-resistant high-risk clones by hospital networks: changing the patient direction can make all the difference. J Hosp Infect.  2014; 86 1: 34– 41. Google Scholar CrossRef Search ADS PubMed  33 Bartsch SM, Huang SS, Wong KF, Avery TR, Lee BY. The spread and control of norovirus outbreaks among hospitals in a region: a simulation model. Open Forum Infect Dis.  2014; 1 2: ofu030. Google Scholar CrossRef Search ADS PubMed  34 Ohst J, Liljeros F, Stenhem M, Holme P. The network positions of methicillin resistant Staphylococcus aureus affected units in a regional healthcare system. EPJ Data Sci.  2014; 3 1: 1– 15. Google Scholar CrossRef Search ADS   35 Lesosky M, McGeer A, Simor A, Green K, Low DE, Raboud J. Effect of patterns of transferring patients among healthcare institutions on rates of nosocomial methicillin-resistant Staphylococcus aureus transmission: a Monte Carlo simulation. Infect Control Hosp Epidemiol.  2011; 32 2: 136– 47. Google Scholar CrossRef Search ADS PubMed  36 van den Dool C, Haenen A, Leenstra T, Wallinga J. The role of nursing homes in the spread of antimicrobial resistance over the healthcare network. Infect Control Hosp Epidemiol.  2016; 37 7: 761– 67. Google Scholar CrossRef Search ADS PubMed  37 Ke W, Huang SS, Hudson LO, et al.   Patient sharing and population genetic structure of methicillin-resistant Staphylococcus aureus. Proc Natl Acad Sci.  2012; 109 17: 6763– 68. Google Scholar CrossRef Search ADS   38 Huang SS, Avery TR, Song Y, et al.   Quantifying interhospital patient sharing as a mechanism for infectious disease spread. Infect Control Hosp Epidemiol.  2010; 31 11: 1160– 69. Google Scholar CrossRef Search ADS PubMed  39 Walker AS, Eyre DW, Wyllie DH, et al.   Characterisation of clostridium difficile hospital ward-based transmission using extensive epidemiological data and molecular typing. PLoS Med.  2012; 9 2: e1001172. Google Scholar CrossRef Search ADS PubMed  40 Gibbons CL, van Bunnik AD, Blatchford O, et al.   Not just a matter of size: a hospital-level risk factor analysis of MRSA bacteraemia in Scotland. BMC Infect Dis.  2016; 16 1: 1– 7. Google Scholar CrossRef Search ADS PubMed  41 Simmering JE, Polgreen LA, Campbell DR, Cavanaugh JE, Polgreen PM. Hospital transfer network structure as a risk factor for clostridium difficile infection. Infect Control Hosp Epidemiol.  2015; 36 9: 1031– 37. Google Scholar CrossRef Search ADS PubMed  42 Lee BY, Bartsch SM, Wong KF. Simulation shows hospitals that cooperate on infection control obtain better results than hospitals acting alone. Health Affairs.  2012; 31 10: 2295– 303. Google Scholar CrossRef Search ADS   43 Ciccolini M, Donker T, Grundmann H, Bonten MJM, Woolhouse MEJ. Efficient surveillance for healthcare-associated infections spreading between hospitals. Proc Natl Acad Sci.  2014; 111 6: 2271– 76. Google Scholar CrossRef Search ADS   44 Fernandez-Gracia J, Onnela J-P, Barnett M, Eguíluz VM, Christakis NA. Spread of pathogens in the patient transfer network of US hospitals. arXiv:1504.08343v1 [physics.soc-ph]. 2015;13pp. Google Scholar CrossRef Search ADS   45 van Bunnik BAD, Ciccolini M, Gibbons CL, et al.   Efficient national surveillance for health-care-associated infections. BMC Public Health.  2015; 15 1: 1– 9. Google Scholar CrossRef Search ADS PubMed  46 Karkada UH, Adamic LA, Kahn JM, Iwashyna TJ. Limiting the spread of highly resistant hospital-acquired microorganisms via critical care transfers: a simulation study. Intensive Care Med.  2011; 37 10: 1633– 40. Google Scholar CrossRef Search ADS PubMed  47 Prakash BA, Adamic L, Iwashyna T, Tong Hanghang, Faloutsos C. Fractional immunization in networks. In SIAM International Conference on Data Mining. Austin, TX; 2013:659–67. 48 Cunningham FC, Ranmuthugala G, Plumb J, Georgiou A, Westbrook JI, Braithwaite J. Health professional networks as a vector for improving healthcare quality and safety: a systematic review. BMJ Qual Safety.  2012; 21 3: 239– 49. Google Scholar CrossRef Search ADS   49 Tasselli S. Social networks of professionals in health care organizations: a review. Med Care Res Rev.  2014; 71 6: 619– 60. Google Scholar CrossRef Search ADS PubMed  50 Bae S-H, Nikolaev A, Seo JY, Castner J. Health care provider social network analysis: A systematic review. Nursing Outlook.  2015; 63 5: 566– 84. Google Scholar CrossRef Search ADS PubMed  51 Lublóy Á. Factors affecting the uptake of new medicines: a systematic literature review. BMC Health Services Res.  2014; 14 1: 1– 25. Google Scholar CrossRef Search ADS   52 Barnett ML, Landon BE, O'Malley AJ, Keating NL, Christakis NA. Mapping physician networks with self-reported and administrative data. Health Services Res.  2011; 46 5: 1592– 609. Google Scholar CrossRef Search ADS   53 Bridewell W, Das AK. Social network analysis of physician interactions: the effect of institutional boundaries on breast cancer care. In AMIA Annual Symposium Proceedings . 2011; 152– 60. 54 Landon BE, Keating NL, Barnett ML, et al.   Variation in patient-sharing networks of physicians across the united states. JAMA.  2012; 308 3: 265– 73. Google Scholar CrossRef Search ADS PubMed  55 Paul S, Keating NL, Landon BE, O'Malley AJ. Results from using a new dyadic-dependence model to analyze sociocentric physician networks. Soc Sci Med.  2014; 117: 67– 75. Google Scholar CrossRef Search ADS PubMed  56 Uddin S, Hamra J, Hossain L. Mapping and modeling of physician collaboration network. Stat Med.  2013; 32 20: 3539– 51. Google Scholar CrossRef Search ADS PubMed  57 Barnett ML, Keating NL, Christakis NA, O'Malley AJ, Landon BE. Reasons for choice of referral physician among primary care and specialist physicians. J General Int Med.  2012; 27 5: 506– 12. Google Scholar CrossRef Search ADS   58 Hackl F, Hummer M, Pruckner GJ. Old boys’ network in general practitioners’ referral behavior? J Health Econ.  2015; 43: 56– 73. Google Scholar CrossRef Search ADS PubMed  59 Moen EL, Austin AM, Bynum JP, Skinner JS, O'Malley AJ. An analysis of patient-sharing physician networks and implantable cardioverter defibrillator therapy. Health Services Outcomes Res Methodol.  2016; 16 3: 132– 53. Google Scholar CrossRef Search ADS   60 Pollack CE, Weissman G, Bekelman J, Liao K, Armstrong K. Physician social networks and variation in prostate cancer treatment in three cities. Health Services Res.  2012; 47( 1 Pt 2): 380– 403. Google Scholar CrossRef Search ADS   61 Pollack CE, Soulos PR, Gross CP. Physician’s peer exposure and the adoption of a new cancer treatment modality. Cancer.  2015; 121 16: 2799– 807. Google Scholar CrossRef Search ADS PubMed  62 Ong MS, Olson KL, Cami A, et al.   Provider patient-sharing networks and multiple-provider prescribing of benzodiazepines. J General Int Med.  2016; 31 2: 164– 71. Google Scholar CrossRef Search ADS   63 Takahashi Y, Ishizaki T, Nakayama T, Kawachi I. Social network analysis of duplicative prescriptions: One-month analysis of medical facilities in Japan. Health Policy.  2016; 120 3: 334– 41. Google Scholar CrossRef Search ADS PubMed  64 Hussain T, Chang H-Y, Veenstra CM, Pollack CE. Collaboration between surgeons and medical oncologists and outcomes for patients with stage III colon cancer. J Oncol Pract.  2015; 11 3: e388– 97. Google Scholar CrossRef Search ADS PubMed  65 Lubloy A, Kereszturi JL, Benedek G. Formal professional relationships between general practitioners and specialists in shared care: possible associations with patient health and pharmacy costs. Appl Health Econ Health Policy.  2016; 14 2: 217– 27. Google Scholar CrossRef Search ADS PubMed  66 Barnett ML, Christakis NA, O'Malley J, Onnela J-P, Keating NL, Landon BE. Physician patient-sharing networks and the cost and intensity of care in US hospitals. Medical Care.  2012; 50 2: 152– 60. Google Scholar CrossRef Search ADS PubMed  67 Uddin S, Hossain L, Kelaher M. Effect of physician collaboration network on hospitalization cost and readmission rate. Eur J Public Health.  2012; 22 5: 629– 33. Google Scholar CrossRef Search ADS PubMed  68 DiazGranados D, Dow AW, Perry SJ, Palesis JA. Understanding patient care as a multiteam system. In Pushing the Boundaries: Multiteam Systems in Research and Practice . Emerald Group Publishing Limited; 2014: 95– 113. Google Scholar CrossRef Search ADS   69 Casalino LP, Pesko MF, Ryan AM, et al.   Physician networks and ambulatory care–sensitive admissions. Med Care.  2015; 53 6: 534– 41. Google Scholar CrossRef Search ADS PubMed  70 Pollack CE, Wang H, Bekelman JE, et al.   Physician social networks and variation in rates of complications after radical prostatectomy. Value Health.  2014; 17 5: 611– 18. Google Scholar CrossRef Search ADS PubMed  71 Hollingsworth JM, Funk RJ, Garrison SA, et al.   Differences between physician social networks for cardiac surgery serving communities with high versus low proportions of black residents. Med Care.  2015; 53 2: 160– 67. Google Scholar CrossRef Search ADS PubMed  72 Geissler KH, Lubin B, Marzilli Ericson KM. Access is not enough: characteristics of physicians who treat medicaid patients. Med Care.  2016; 54 4: 350– 58. Google Scholar CrossRef Search ADS PubMed  73 Emmert-Streib F, Tripathi S, de Matos Simoes R, Hawwa AF, Dehmer M. The human disease network. Syst Biomed.  2013; 1 1: 20– 28. Google Scholar CrossRef Search ADS   74 Capobianco E, Liò P. Comorbidity networks: beyond disease correlations. J Complex Networks . 2015; 3 3: 319– 32. Google Scholar CrossRef Search ADS   75 Jensen PB, Jensen LJ, Brunak S. Mining electronic health records: towards better research applications and clinical care. Nat Rev Genet.  2012; 13 6: 395– 405. Google Scholar CrossRef Search ADS PubMed  76 Rzhetsky A, Wajngurt D, Park N, Zheng T. Probing genetic overlap among complex human phenotypes. Proc Natl Acad Sci.  2007; 104 28: 11694– 99. Google Scholar CrossRef Search ADS   77 Hanauer DA, Rhodes DR, Chinnaiyan AM. Exploring clinical associations using ‘-omics’ based enrichment analyses. PLoS One.  2009; 4 4: e5203. Google Scholar CrossRef Search ADS PubMed  78 Chen Y, Xu R. Network analysis of human disease comorbidity patterns based on large-scale data mining. In Basu M, Pan Y, Wang J, eds. International Symposium on Bioinformatics Research and Applications . Zhangjiajie, China: Springer International Publishing; 2014: 243– 54. Google Scholar CrossRef Search ADS   79 Bagley SC, Sirota M, Chen R, Butte AJ, Altman RB. Constraints on biological mechanism from disease comorbidity using electronic medical records and database of genetic variants. PLoS Comput Biol.  2016; 12 4: e1004885. Google Scholar CrossRef Search ADS PubMed  80 Zamora M, Baradad M, Amado E, et al.   Characterizing chronic disease and polymedication prescription patterns from electronic health records. In IEEE International Conference on Data Science and Advanced Analytics. Paris: 2015; 1–9. 81 Chen Y, Li L, Xu R. Disease comorbidity network guides the detection of molecular evidence for the link between colorectal cancer and obesity. In AMIA Joint Summits on Translational Science . 2015; 2015: 201– 06. 82 Paik H, Heo H-S, Ban H-j, Cho SB. Unraveling human protein interaction networks underlying co-occurrences of diseases and pathological conditions. J Transl Med.  2014; 12 1: 1– 8. Google Scholar CrossRef Search ADS PubMed  83 Park J, Lee D-S, Christakis NA, Barabási A-L. The impact of cellular networks on disease comorbidity. Mol Syst Biol.  2009; 5: 262. Google Scholar CrossRef Search ADS PubMed  84 Davis DA, Chawla NV. Exploring and exploiting disease interactions from multi-relational gene and phenotype networks. PLoS One.  2011; 6 7: e22670. Google Scholar CrossRef Search ADS PubMed  85 Park S, Yang J-S, Kim J, et al.   Evolutionary history of human disease genes reveals phenotypic connections and comorbidity among genetic diseases. Scientific Reports.  2012; 2: 757. Google Scholar CrossRef Search ADS PubMed  86 Blair DR, Lyttle CS, Mortensen JM, et al.   A nondegenerate code of deleterious variants in mendelian loci contributes to complex disease risk. Cell.  2013; 155 1: 70– 80. Google Scholar CrossRef Search ADS PubMed  87 Moni MA, Lio P. How to build personalized multi-omics comorbidity profiles. Front Cell Develop Biol.  2015; 3 28: 00028. 88 Liu C-C, Tseng Y-T, Li W, et al.   DiseaseConnect: a comprehensive web server for mechanism-based disease-disease connections. Nucleic Acids Res.  2014; 42( W1): W137– 46. Google Scholar CrossRef Search ADS PubMed  89 Xu H, Moni MA, Lio P. Network regularised Cox regression and multiplex network models to predict disease comorbidities and survival of cancer. Comput Biol Chem.  2015; 59 ( Part B): 15– 31. Google Scholar CrossRef Search ADS PubMed  90 Jensen AB, Moseley PL, Oprea TI, et al.   Temporal disease trajectories condensed from population-wide registry data covering 6.2 million patients. Nat Commun.  2014; 5: 4022. 91 Chen LL, Blumm N, Christakis NA, Barabási A-L, Deisboeck TS. Cancer metastasis networks and the prediction of progression patterns. Brit J Cancer.  2009; 101 5: 749– 58. Google Scholar CrossRef Search ADS   92 Hidalgo CA, Blumm N, Barabási A-L, Christakis NA. A dynamic network approach for the study of human phenotypes. PLoS Comput Biol.  2009; 5 4: e1000353. Google Scholar CrossRef Search ADS PubMed  93 Hanauer DA, Ramakrishnan N. Modeling temporal relationships in large scale clinical associations. J Am Med Inform Assoc.  2013; 20 2: 332– 41. Google Scholar CrossRef Search ADS PubMed  94 Glicksberg BS, Li L, Badgeley MA, et al.   Comparative analyses of population-scale phenomic data in electronic medical records reveal race-specific disease networks. Bioinformatics.  2016; 32 12: i101– 10. Google Scholar CrossRef Search ADS PubMed  95 Kannan V, Swartz F, Kiani NA, et al.   Conditional disease development extracted from longitudinal health care cohort data using layered network construction. Scientific Rep.  2016; 6: 26170. Google Scholar CrossRef Search ADS   96 Steinhaeuser K, Chawla NV. A network-based approach to understanding and predicting diseases. In Social Computing and Behavioral Modeling . Boston: Springer; 2009: 1– 8. Google Scholar CrossRef Search ADS   97 Roque FS, Jensen PB, Schmock H, et al.   Using electronic patient records to discover disease correlations and stratify patient cohorts. PLoS Comput Biol.  2011; 7 8: e1002141. Google Scholar CrossRef Search ADS PubMed  98 Sideris C, Pourhomayoun M, Kalantarian H, Sarrafzadeh M. A flexible data-driven comorbidity feature extraction framework. Comput Biol Med.  2016; 73: 165– 72. Google Scholar CrossRef Search ADS PubMed  99 Fujita K, Akiyama M, Toyama N, Kamemori Y. Detecting effective classes of medical incident reports based on linguistic analysis for common reporting system in Japan. In Lehmann CU, Ammenwerth E, Nøhr C, eds. World Congress on Medical and Health Informatics . Copenhagen: IOS Press; 2013: 137– 41. 100 Bauer-Mehren A, LePendu P, Iyer SV, Harpaz R, Leeper NJ, Shah NH. Network analysis of unstructured EHR data for clinical research. In AMIA Joint Summits on Translational Science . 2013; 2013: 14– 8. 101 Roitmann E, Eriksson R, Brunak S. Patient stratification and identification of adverse event correlations in the space of 1190 drug related adverse events. Front Physiol.  2014; 5: 332. Google Scholar CrossRef Search ADS PubMed  102 Rojas E, Munoz-Gama J, Sepúlveda M, Capurro D. Process mining in healthcare: a literature review. J Biomed Inform.  2016; 61: 224– 36. Google Scholar CrossRef Search ADS PubMed  103 Chambers D, Wilson P, Thompson C, Harden M. Social network analysis in healthcare settings: a systematic scoping review. PLoS One.  2012; 7 8: e41911. Google Scholar CrossRef Search ADS PubMed  104 Chen Y, Nyemba S, Malin B. Detecting anomalous insiders in collaborative information systems. IEEE Transactions on Dependable and Secure Computing.  2012; 9 3: 332– 44. Google Scholar CrossRef Search ADS PubMed  105 Chen Y, Nyemba S, Zhang W, Malin B. Specializing network analysis to detect anomalous insider actions. Security Informatics.  2012; 1 1: 5. Google Scholar CrossRef Search ADS PubMed  106 Menon AK, Jiang X, Kim J, Vaidya J, Ohno-Machado L. Detecting inappropriate access to electronic health records using collaborative filtering. Machine Learning.  2014; 95 1: 87– 101. Google Scholar CrossRef Search ADS PubMed  107 Zhang H, Mehotra S, Liebovitz D, Gunter CA, Malin B. Mining deviations from patient care pathways via electronic medical record system audits. ACM Transactions on Management Information Systems.  2013; 4 4: 1– 20. Google Scholar CrossRef Search ADS   108 Liu J, Bier E, Wilson A. Graph analysis for detecting fraud, waste, and abuse in healthcare data. In Conference on Innovative Applications of Artificial Intelligence. Austin, TX: AAAI Press; 2015: 3912–19. 109 Geva A, Wright SB, Baldini LM, Smallcomb JA, Safran C, Gray JE. Spread of methicillin-resistant Staphylococcus aureus in a large tertiary NICU: Network analysis. Pediatrics.  2011; 128 5: e1173– 80. Google Scholar CrossRef Search ADS PubMed  110 Ueno T, Masuda N. Controlling nosocomial infection based on structure of hospital social networks. J Theoretical Biol.  2008; 254 3: 655– 66. Google Scholar CrossRef Search ADS   111 Curtis DE, Hlady CS, Kanade G, Pemmaraju SV, Polgreen PM, Segre AM. Healthcare worker contact networks and the prevention of hospital-acquired infections. PLoS One.  2013; 8 12: e79906. Google Scholar CrossRef Search ADS PubMed  112 Cusumano-Towner M, Li DY, Tuo S, Krishnan G, Maslove DM. A social network of hospital acquired infection built from electronic medical record data. J Am Med Inform Assoc.  2013; 20 3: 427– 34. Google Scholar CrossRef Search ADS PubMed  113 Siden H, Urbanoski K. Using network analysis to map the formal clinical reporting process in pediatric palliative care: a pilot study. BMC Health Services Res.  2011; 11: 343. Google Scholar CrossRef Search ADS   114 Uddin S, Hossain L. Social networks enabled coordination model for cost management of patient hospital admissions. J Healthcare Qual.  2011; 33 5: 37– 48. Google Scholar CrossRef Search ADS   115 Pollack CE, Weissman GE, Lemke KW, Hussey PS, Weiner JP. Patient sharing among physicians and costs of care: a network analytic approach to care coordination using claims data. J General Int Med.  2013; 28 3: 459– 65. Google Scholar CrossRef Search ADS   116 Soulakis ND, Carson MB, Lee YJ, Schneider DH, Skeehan CT, Scholtens DM. Visualizing collaborative electronic health record usage for hospitalized patients with heart failure. J Am Med Inform Assoc.  2015; 22 2: 299– 311. Google Scholar CrossRef Search ADS PubMed  117 Uddin S, Kelaher M, Srinivasan U. A framework for administrative claim data to explore healthcare coordination and collaboration. Australian Health Rev.  2015; 40 5: 500– 10. Google Scholar CrossRef Search ADS   118 Anderson C, Talsma A. Characterizing the structure of operating room staffing using social network analysis. Nursing Res.  2011; 60 6: 378– 85. Google Scholar CrossRef Search ADS   119 Tighe PJ, Patel SS, Gravenstein N, Davies L, Lucas SD, Bernard HR. The Operating Room: It’s a Small World (and Scale Free Network) After All. Connections.  2014; 34( 1 and 2): 29– 42. 120 Baumgart A, Denz C, Bender H-J, Schleppers A. How work context affects operating room processes: using data mining and computer simulation to analyze facility and process design. Qual Manag Health Care.  2009; 18 4: 305– 14. Google Scholar CrossRef Search ADS PubMed  121 Gray JE, Davis DA, Pursley DM, Smallcomb JE, Geva A, Chawla NV. Network analysis of team structure in the neonatal intensive care unit. Pediatrics.  2010; 125 6: e1460– 67. Google Scholar CrossRef Search ADS PubMed  122 Chen Y, Lorenzi N, Nyemba S, Schildcrout JS, Malin B. We work with them? Healthcare workers interpretation of organizational relations mined from electronic health records. Int J Med Inform.  2014; 83 7: 495– 506. Google Scholar CrossRef Search ADS PubMed  123 Rossille D, Cuggia M, Arnault A, Bouget J, Le Beux P. Managing an emergency department by analysing HIS medical data: a focus on elderly patient clinical pathways. Health Care Manag Sci.  2008; 11 2: 139– 46. Google Scholar CrossRef Search ADS PubMed  124 Bose RPJC, van der Aalst WMP. Analysis of patient treatment procedures. In Daniel F, Barkaoui K, Dustdar S, eds. Business Process Management Workshops. Clermont-Ferrand, France: Springer, Berlin Heidelberg; 2011: 165–66. 125 Rebuge A, Ferreira DR. Business process analysis in healthcare environments: a methodology based on process mining. Inform Syst.  2012; 37 2: 99– 116. Google Scholar CrossRef Search ADS   126 Liu C-R, Wang F, Hu J, Xiong H. Temporal phenotyping from longitudinal electronic health records: a graph based framework. In ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Sydney, NSW, Australia: ACM; 2015: 705–14. 127 Merrill JA, Sheehan BM, Carley KM, Stetson PD. Transition networks in a cohort of patients with congestive heart failure. A novel application of informatics methods to inform care coordination. Appl Clin Inform.  2015; 6 3: 548– 64. Google Scholar CrossRef Search ADS PubMed  128 Zhang Y, Padman R, Patel N. Paving the COWpath: Learning and visualizing clinical pathways from electronic health record data. J Biomed Inform.  2015; 58: 186– 97. Google Scholar CrossRef Search ADS PubMed  129 Hripcsak G, Vawdrey DK, Fred MR, Bostwick SB. Use of electronic clinical documentation: time spent and team interactions. J Am Med Inform Assoc.  2011; 18 2: 112– 17. Google Scholar CrossRef Search ADS PubMed  130 Lee BY, Song Y, Bartsch SM, et al.   Long-term care facilities: important participants of the acute care facility social network? PLoS One.  2011; 6 12: e29342. Google Scholar CrossRef Search ADS PubMed  131 Malin B, Nyemba S, Paulett J. Learning relational policies from electronic health record access logs. J Biomed Inform.  2011; 44 2: 333– 42. Google Scholar CrossRef Search ADS PubMed  132 Finlayson SG, LePendu P, Shah NH. Building the graph of medicine from millions of clinical narratives. Scientific Data.  2014; 1: 140032. Google Scholar CrossRef Search ADS PubMed  133 Pham HH, O'Malley AS, Bach PB, Saiontz-Martinez C, Schrag D. Primary care physicians’ links to other physicians through Medicare patients: the scope of care coordination. Ann Int Med.  2009; 150 4: 236– 42. Google Scholar CrossRef Search ADS   134 Mandl KD, Olson KL, Mines D, Liu C, Tian F. Provider collaboration: cohesion, constellations, and shared patients. J General Int Med.  2014; 29 11: 1499– 505. Google Scholar CrossRef Search ADS   135 Uddin S, Hossain L. Social networks in exploring healthcare coordination. Asia Pacific J Health Manag.  2014; 9 3: 53– 62. 136 Spear SE, Reducing readmissions to detoxification: an interorganizational network perspective. Drug Alcohol Dependence.  2014; 137: 76– 82. Google Scholar CrossRef Search ADS   137 Uddin S, Exploring the impact of different multi-level measures of physician communities in patient-centric care networks on healthcare outcomes: A multi-level regression approach. Scientific Rep.  2016; 6: 20222. Google Scholar CrossRef Search ADS   138 Aprile G, Ramoni M, Keefe D, Sonis S. Application of distance matrices to define associations between acute toxicities in colorectal cancer patients receiving chemotherapy. Cancer.  2008; 112 2: 284– 92. Google Scholar CrossRef Search ADS PubMed  139 Botsis T, Scott J, Woo EJ, Ball R. Identifying similar cases in document networks using cross-reference structures. IEEE Journal of Biomedical and Health Informatics.  2015; 19 6: 1906– 17. Google Scholar CrossRef Search ADS PubMed  140 Lyalina S, Percha B, LePendu P, Iyer SV, Altman RB, Shah NH. Identifying phenotypic signatures of neuropsychiatric disorders from electronic medical records. J Am Med Inform Assoc.  2013; 20( e2): e297– 305. Google Scholar CrossRef Search ADS PubMed  141 Jing X, Cimino JJ. A complementary graphical method for reducing and analyzing large data sets. Case studies demonstrating thresholds setting and selection. Methods Inform Med.  2014; 53 3: 173– 85. Google Scholar CrossRef Search ADS   142 Chen Y, Ghosh J, Bejan CA, et al.   Building bridges across electronic health record systems through inferred phenotypic topics. J Biomed Inform.  2015; 55: 82– 93. Google Scholar CrossRef Search ADS PubMed  143 Manuel DG, Lam K, Maaten S, Klein-Geltink J. Using administrative data to measure the extent to which practitioners work together: “interconnected” care is common in a large cohort of family physicians. Open Med.  2011; 5 4: e177– 82. Google Scholar PubMed  144 Landon BE, Onnela J-P, Keating NL, et al.   Using administrative data to identify naturally occurring networks of physicians. Medical Care.  2013; 51 8: 715– 21. Google Scholar CrossRef Search ADS PubMed  145 Uddin S, Hossain L, Hamra J, Alam A. A study of physician collaborations through social network and exponential random graph. BMC Health Services Res.  2013; 13 1: 1– 14. Google Scholar CrossRef Search ADS   146 Kim JH, Son KY, Shin DW, et al.   Network analysis of human diseases using Korean nationwide claims data. J Biomed Inform.  2016; 61: 276– 82. Google Scholar CrossRef Search ADS PubMed  147 Liu J, Ma J, Wang J, et al.   Comorbidity analysis according to sex and age in hypertension patients in China. Int J Med Sci.  2016; 13 2: 99– 107. Google Scholar CrossRef Search ADS PubMed  148 Schafer I, Kaduszkiewicz H, Wagner H-O, Schön G, Scherer M, van den Bussche H. Reducing complexity: a visualisation of multimorbidity by combining disease clusters and triads. BMC Public Health.  2014; 14: 1285. Google Scholar CrossRef Search ADS PubMed  149 Nikfarjam A, Emadzadeh E, Gonzalez G. Towards generating a patient’s timeline: extracting temporal relationships from clinical notes. J Biomed Inform.  2013; 46( Suppl): S40– 47. Google Scholar CrossRef Search ADS PubMed  150 Zhou X-Z, Chen S, Liu B, et al.   Development of traditional Chinese medicine clinical data warehouse for medical knowledge discovery and decision support. Artificial Intell Med.  2010; 48( 2–3): 139– 52. Google Scholar CrossRef Search ADS   151 Heer J, Perer A. Orion: A system for modeling, transformation and visualization of multidimensional heterogeneous networks. Inform Visualization.  2014; 13 2: 111– 33. Google Scholar CrossRef Search ADS   152 Li Y-B, Zhou X-Z, Zhang R-S, et al.   Detection of herb-symptom associations from traditional Chinese medicine clinical data. Evid Based Complement Alternat Med.  2015; 2015: 270450. Google Scholar PubMed  153 Warner JL, Denny JC, Kreda DA, Alterovitz G. Seeing the forest through the trees: uncovering phenomic complexity through interactive network visualization. J Am Med Inform Assoc.  2015; 22 2: 324– 29. Google Scholar CrossRef Search ADS PubMed  154 Chmiel A, Klimek P, Thurner S. Spreading of diseases through comorbidity networks across life and gender. New J Physics.  2014; 16: 115013. 155 Geraci DM, Giuffrè M, Bonura C, et al.   A Snapshot on MRSA Epidemiology in a Neonatal Intensive Care Unit Network, Palermo, Italy. Front Microbiol.  2016; 7: 815. Google Scholar CrossRef Search ADS PubMed  156 Ball R, Botsis T. Can network analysis improve pattern recognition among adverse events following immunization reported to VAERS? Clin Pharmacol Therapeutics.  2011; 90 2: 271– 78. Google Scholar CrossRef Search ADS   157 Patel VN, Kaelber DC. Using aggregated, de-identified electronic health record data for multivariate pharmacosurveillance: a case study of azathioprine. J Biomed Inform.  2014; 52: 36– 42. Google Scholar CrossRef Search ADS PubMed  158 Scott J, Botsis T, Ball R. Simulating adverse event spontaneous reporting systems as preferential attachment networks: application to the Vaccine Adverse Event Reporting System. Appl Clin Inform.  2014; 5 1: 206– 18. Google Scholar CrossRef Search ADS PubMed  159 Franchini M, Pieroni S, Fortunato L, Molinaro S, Liebman M. Poly-pharmacy among the elderly: analyzing the co-morbidity of hypertension and diabetes. Curr Pharmaceutical Design.  2015; 21 6: 791– 805. Google Scholar CrossRef Search ADS   160 Chen Y, Xu R. Mining cancer-specific disease comorbidities from a large observational health database. Cancer Informatics.  2014 ( Supp 1: Computational Advances in Cancer Informatics (A)): 37– 44. 161 Minerba L, Chessa A, Coppola RC, Mula G, Cappellini G. A complex network analysis of a health organization. Igiene e Sanita Pubblica.  2008; 64 1: 9– 25. Google Scholar PubMed  162 Abbasi A, Uddin S, Hossain L. Socioeconomic analysis of patient-centric networks: Effects of patients and hospitals’ characteristics and network structure on hospitalization costs. Eur J Health Econ.  2012; 13 3: 267– 76. Google Scholar CrossRef Search ADS PubMed  163 Butala NM, King MD, Reitsma W, et al.   Association between organ procurement organization social network centrality and kidney discard and transplant outcomes. Transplantation.  2015; 99 12: 2617– 24. Google Scholar CrossRef Search ADS PubMed  164 Feldman K, Stiglic G, Dasgupta D, Kricheff M, Obradovic Z, Chawla NV. Insights into population health management through disease diagnoses networks. Scientific Rep.  2016; 6: 30465. Google Scholar CrossRef Search ADS   165 Holmes AB, Hawson A, Liu F, Friedman C, Khiabanian H, Rabadan R. Discovering disease associations by integrating electronic clinical data and medical literature. PLoS One.  2011; 6 6: e21132. Google Scholar CrossRef Search ADS PubMed  166 Kwan TH, Wong NS, Lee SS. Participation dynamics of a cohort of drug users in a low-threshold methadone treatment programme. Harm Reduction J.  2015; 12: 30. Google Scholar CrossRef Search ADS   167 Fattore G, Salvatore D. Network organizations of general practitioners: antecedents of formation and consequences of participation. BMC Health Services Res.  2010; 10: 118. Google Scholar CrossRef Search ADS   168 Fattore G, Frosini F, Salvatore D, Tozzi V. Social network analysis in primary care: the impact of interactions on prescribing behaviour. Health Policy.  2009; 92( 2–3): 141– 48. Google Scholar CrossRef Search ADS PubMed  169 Yesha R, Gangopadhyay A, Siegel E. A graph-based method for analyzing electronic medical records. In IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining. Paris: ACM; 2015: 1036–41. 170 Greenacre M, Correspondence Analysis in Practice , 2nd ed. CRC Interdisciplinary Statistics. Boca Raton: Chapman & Hall/CRC; 2007: 296. Google Scholar CrossRef Search ADS   171 Fabbri D, LeFevre K. Explanation-based auditing. In Proceedings of the VLDB Endowment . Istanbul; 2011; 5(1): 1– 12. Google Scholar CrossRef Search ADS   172 Chandola V, Sukumar SR, Schryver JC. Knowledge discovery from massive healthcare claims data. In ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Chicago: ACM; 2013:1312–20. 173 Lai Y-H. Network analysis of comorbidities: case study of HIV/AIDS in Taiwan. In Wang L, et al.  eds. Multidisciplinary Social Networks Research . Berlin Heidelberg:Springer; 2015: 174– 86. Google Scholar CrossRef Search ADS   174 Patnaik D, Butler P, Ramakrishnan N, Parida L, Keller BJ, Hanauer DA. Experiences with mining temporal event sequences from electronic medical records: initial successes and some challenges. In ACM SIGKDD Conference on Knowledge Discovery and Data Mining. San Diego, CA: ACM; 2011: 360–68. 175 Barrat A, Fernandez B, Lin KK, Young L-S. Modeling temporal networks using random itineraries. Phys Rev Lett.  2013; 110 15: 158702. Google Scholar CrossRef Search ADS PubMed  176 Butte AJ, Kohane IS. Unsupervised knowledge discovery in medical databases using relevance networks. In AMIA Annual Symposium . 1999; 1999: 711– 15. 177 Cuggia M, Rossille D, Arnault A, Bouget J, Le Beux P. Towards a decision support system for optimising clinical pathways of elderly patients in an emergency department. In Kuhn KA, Warren JR, Leong T-Y, eds. World Congress on Health (Medical) Informatics . Amsterdam: IOS Press; 2007: 840– 44. 178 Chen Y, Nyemba S, Zhang W, Malin B. Leveraging social networks to detect anomalous insider actions in collaborative environments. In IEEE International Conference on Intelligence and Security Informatics. Beijing; 2011;2011: 119–24. 179 Robins G, Snijders T, Wang P, Handcock M, Pattison P. Recent developments in exponential random graph (p*) models for social networks. Social Networks.  2007; 29 2: 192– 215. Google Scholar CrossRef Search ADS   180 Hunter DR, Handcock MS, Butts CT, Goodreau SM, Morris M. ERGM: A Package to Fit, Simulate and Diagnose Exponential-Family Models for Networks. J Stat Softw.  2008; 24 3: nihpa54860. Google Scholar CrossRef Search ADS PubMed  181 Piro RM. Network medicine: linking disorders. Human Genetics.  2012; 131 12: 1811– 20. Google Scholar CrossRef Search ADS PubMed  182 Vidal M, Cusick ME, Barabási AL. Interactome networks and human disease. Cell.  2011; 144 6: 986– 98. Google Scholar CrossRef Search ADS PubMed  183 Pescosolido BA. Of pride and prejudice: the role of sociology and social networks in integrating the health sciences. J Health Soc Behav.  2006; 47 3: 189– 208. Google Scholar CrossRef Search ADS PubMed  184 Chen ES, Cimino JJ. Automated discovery of patient-specific clinician information needs using clinical information system log files. In AMIA Annual Symposium Proceedings . Vancouver, BC: American Medical Informatics Association; 2003: 145. 185 Mans RS, Schonenberg MH, Song M, van der Aalst WMP, Bakker PJM. Application of Process Mining in Healthcare: A Case Study in a Dutch Hospital. In Fred A, Filipe J, Gamboa H, eds. International Joint Conference on Biomedical Engineering Systems and Technologies . Berlin Heidelberg: Funchal, Madeira, Portugal: Springer; 2009: 425– 38. Google Scholar CrossRef Search ADS   186 De Weerdt J, Caron F, Vanthienen J, Baesens B. Getting a grasp on clinical pathway data: an approach based on process mining. In Washio T, Luo J, eds. Pacific-Asia Conference on Knowledge Discovery and Data Mining , Kuala Lumpur: Springer Berlin Heidelberg; 2012: 22– 35. Google Scholar CrossRef Search ADS   187 Finney JM, Marquez LM. Generating temporal network paths from hospital data. In International Conference on Health Informatics. Rome: Springer; 2016: 263–68. © The Author 2017. Published by Oxford University Press on behalf of the American Medical Informatics Association. All rights reserved. For Permissions, please email: journals.permissions@oup.com
TI - Applications of network analysis to routinely collected health care data: a systematic review
JF - Journal of the American Medical Informatics Association
DO - 10.1093/jamia/ocx052
DA - 2017-09-13
UR - https://www.deepdyve.com/lp/oxford-university-press/applications-of-network-analysis-to-routinely-collected-health-care-0O9yZzYSHj
SP - 210
EP - 221
VL - 25
IS - 2
DP - DeepDyve
ER -