Analysis of scientific collaboration in Chinese psychiatry research

Background In recent decades, China has changed profoundly, becoming the country with the world’s second-largest economy. The proportion of the Chinese population suffering from mental disorder has grown in parallel with the rapid economic development, as social stresses have increased. The aim of this study is to shed light on the status of collaborations in the Chinese psychiatry field, of which there is currently limited research. Methods We sampled 16,224 publications (2003-2012) from 10 core psychiatry journals from Chinese National Knowledge Infrastructure (CNKI) and WanFang Database. We used various social network analysis (SNA) methods such as centrality analysis, and Core-Periphery analysis to study collaboration. We also used hierarchical clustering analysis in this study. Results From 2003-2012, there were increasing collaborations at the level of authors, institutions and regions in the Chinese psychiatry field. Geographically, these collaborations were distributed unevenly. The 100 most prolific authors and institutions and 32 regions were used to construct the collaboration map, from which we detected the core author, institution and region. Collaborative behavior was affected by economic development. Conclusion We should encourage collaborative behavior in the Chinese psychiatry field, as this facilitates knowledge distribution, resource sharing and information acquisition. Collaboration has also helped the field narrow its current research focus, providing further evidence to inform policymakers to fund research in order to tackle the increase in mental disorder facing modern China.


Background
China is now the world's second largest economy country, having seen profound changes in recent decades. The proportion of the Chinese population suffering from mental disorder has grown in parallel with the rapid economic development, as social stresses have increased. According to recent data from the World Health Organization (WHO), the burden of mental disorder is highest, surpassing that of cardiovascular disease, respiratory system disease and malignancy [1]. As it drew the attention of Chinese researchers studying psychiatry, the prevention and control of mental disorder became a huge challenge for them to overcome. Since this field permeates several arenas of biomedicine, no single individual is trained to perform all specialty tasks. Thus, scientific collaboration, which improves communication, facilitates sharing of expertise, and provides opportunities for the emergence of new scientific ideas, is indispensable for the growth of the field of psychiatry in modern China. There has been increased collaboration within and between different scientific fields over the last decade. Co-authorship is a frequently used and reliable measure of research collaboration [2]. In 2001, American researcher Newman used Social network analysis (SNA) to study on the structure of scientific collaboration networks in fields such as biomedicine, physics and computer science [3][4][5]. German scholar Kretschmer applied the method of science, information metrology and psychology in scientific collaboration skillfully and acquired a series of achievements by using SNA [6,7]. SNA has been used to study collaboration in bibliographic co-authorship networks [8,9]. A social network is defined as a set of social entities, such as people, organizations, and countries, with some pattern of relationship between them [10]. These networks are usually modeled by graphs, where nodes represent the social entities and lines represent the ties established between them. The underlying structure of such networks is the object of study of SNA.
There has, however, been a lack of publications on scientific collaboration within the Chinese psychiatry field. Therefore, this study was designed to measure the activities of this field at the micro level (authors), meso level (institutions) and macro level (regions).

Methods
We collected our data from 10 major journals in China National Knowledge Infrastructure (CNKI) and WanFang Database (2003Database ( to 2012. Together, CNKI and WanFang Database cover the majority of journals from China. Each bibliographic record includes information such as the title, author names, abstract and key words. The 10 major journals include: (1) (Table 1). We believe that these 10 publications are sufficient to determine the structure of collaboration in the field of Chinese psychiatry research.
Centrality, which reflects status and rights of activities in their social network, is one of the most important metrics in network analysis. There are three common measures of centrality: degree centrality, betweenness centrality and closeness centrality [11]. In collaborative networks, if there is a direct connection between an actor and others, the actor is in the central position with greater rights. Degree centrality is equal to the number of nodes that connect with a central node. If an author/ institution/region has the highest degree centrality, it is considered a central author/institution/region in the collaboration network [12]. In collaborative networks, if an actor is between two points, that actor is in the important posiition. Betweenness centrality is the number of shortest paths that pass through a given node [13]. In our study, having the highest betweenness centrality would indicate that the author/institution/region possesses and controls a great deal of resources for research. Finally, in collaborative networks, if all paths between an actor and others respectively are the shortest, this actor is in the core position. Closeness centrality of a given node is equal to the reciprocal of the total distance from this node to all other nodes. Thus, the closer a node is to all other nodes, the higher its closeness centrality. Having the lowest closeness centrality indicates that the author/institution/region is at the core of the entire network.
In Core-Periphery structure analysis, the network is divided into two arears-core area and periphery area. The nodes which are in the core area are in the important position.
Hierarchical clustering, which creates a hierarchy of clusters that can be represented by a dendrogram, has been used to extract subgroups from the co-authorship network in many studies. In this tree structure, the root of the tree consists of a single cluster containing all authors, while the leaves correspond to specific individuals.
SNA was used to analyze the collaboration structure of Chinese psychiatry research. We then used the UCI-NET program and Netdraw to produce a visual representation of the network. In addition, the software for frequency analysis was SATI (Statistical Analysis toolkit For Informetrics) published by Zhe Jiang University in China (Http://sati.liuqiyuan.com/#sati).
Firstly, we imported all the data into SATI and produced the co-occurrence matrix. Then, the information of the matrix was imported into UCINET to analysis of centrality, hierarchical clustering and core-periphery structure. Finally, we imported them to Netdraw to the analysis of visualization and formed the mapping knowledge domain.
In order to elucidate the main co-authorship structure of the network, we selected the 100 most prolific authors, the top 100 institutions and 32 regions from 2003 to 2012. This threshold resulted in the top 100 prolific authors who must have published 32 co-authorship papers to be included (two authors who have not collaborated with others were deleted, so the co-authorship map, a visualization of authors' collaboration network, includes 98 authors). The top 100 institutions appeared more than 24 times (25 institutions which have not collaborated with others were deleted, so the collaborative map, a visualization of institutions' collaboration network, includes 75 institutions) and there were 32 regions (there were no collaborations from Tibet, so it was excluded, leaving us with 31 regions).

Results
We retrieved 16,224 papers spanning 2003-2012 from the ten psychiatry journals. Among those, 13,669 were co-authored. The percentage of co-authored papers rose from 78 to 80 % over the last 10 years, suggesting an increase in scientific collaboration in the Chinese psychiatry field.

Analysis on collaboration at the micro (authors) level
Of all the publications from the Chinese psychiatry field in 2003 to 2012, more than 80 % were published by two or more collaborators and the output of achievements in scientific research by way of collaboration was consistent with the total output. This suggested that most published research was collaborative. In this study, the map was composed of 5 independent sub-networks. The line value and the distance between two vertices represent collaborative strength, while thickness of the line represents the number of co-authorship papers. In this authors' collaboration network, the researcher working in Shanghai Mental center has the highest degree centrality of 33, indicating that with 33 collaborators, he was the most key author. In a collaboration network, betweeness centrality reflects the role that the author plays within the network. Without the author with the highest betweeness centrality, the collaborative network would be disrupted. Since the researcher working in Nanjing Medical University was the author with the highest betweeness centrality, he played an instrumental role in maintaining the cohesion of the network and thus had the power to influence collaborative relationships. In a collaborative network, the closer two authors are to each other, the more easily information is communicated and the more likely the two are to collaborate. The researcher working in Shanghai Mental health center had the lowest closeness centrality, indicating that he contributed a great deal of research and was in a central position within the entire network.
Using hierarchical clustering analysis, we divided the 98 authors into 5 sub-networks. The largest sub-network included 90 nodes and 1132 lines. In this sub-network, the average path length was 2.495 and the average clustering coefficient was 3.288, indicating a clustering effect. A researcher, who is in this sub-network, worked in the Shanghai Mental Health Center and had the highest centrality degree mainly due to his research focusing on neuroimaging of patients with mental disorders. The researcher, who also worked in the Shanghai Mental Health Center, had the second highest centrality degree mainly due to his research focusing on the neuropathology of mental disorders. The researcher, with the third highest centrality degree, mainly researched treatments for schizophrenia. The second sub-network included two researchers, who both worked in the Tian shui Mental Hospital of Gansu and researched pharmacological treatments of depression. The third sub-network included the researcher from the Fourth Hospital of Linyi in Shandong and the researcher from the Linyi Mental Health Center of Shandong, and their research focused on the classification of depression. The fourth sub-network included the researcher from Anhui Medical University and the researcher from The Fourth People's Hospital of Hefei in Anhui, and they researched the neurophysiology of mental Fig. 1 The structure map of the institutional collaboration network on Chinese psychiatry research (The geographic name and the name of university adopted the official abbreviation, see Table 2 ) disorders. Finally, the fifth sub-network included two researchers from Xuanwu Hospital in Beijing, where they researched neurology.

Analysis on collaboration at the meso (institutions) level
Papers with multi-institutional collaborators accounted for more than half of all papers. Universities, research institutions and hospitals were the main institutions conducting research in this field. As collaborations among institutions increased, the total output in terms of scientific publications also increased from 2003 to 2012. In this study, there were 7,672 papers from SCI which demonstrated inter-institution collaboration from 2003 to 2012. This number increased from 418 in 2003 to 587 in 2012. These papers included 8,382 institutions that appeared a total of 32,410 times. The largest collaboration in our sample involved 20 institutions. We formed a map visualizing the structure of institutions' collaboration network in the field of psychiatry during 2003 to 2012 (Fig. 1), the abbreviations of the geographic name and the name of university were listed in Table 2. The size of the node represents centrality in collaborative network. Shandong Mental Health Center and Mental Health Center of the Second Xiangya Hospital in Central South University had the highest degree centrality and betweenness centrality, and the lowest closeness centrality (Table 3). This indicated that these institutions were highly collaborative. The distance and thickness of the line between two nodes represents collaborative strength and the number of collaborative papers, respectively. We found that Shandong Mental Health Center, Mental Health Center of the second Xiangya Hospital in Central South University and Mental Health Center of Peking University were in the center of the collaborative network and thus played an influential role in the development of psychiatry. In contrast, The eighth People's Hospital of Zhenzhou and The Fourth People's Hospital of Liaocheng in Shandong were on the edge of the collaborative network. Compared to that of those institutions in the center, the scientific research strength of institutions on the edge were slightly weaker. Further, collaborations among the institutions in the center reflect 'the center of the obvious effect' in the process of coauthorship while the institutions on the edge collaborated more loosely. Analyzing Core-Periphery was a quantitative way to study networks. We applied this method to the collaboration network and found that the 'correlative value of collaborative network' was 0.698 (UCINET 6.0). A welldelineated Core-Periphery structure appeared (Fig. 2), indicating that the collaborative network was regional, whereby research institutions in close geographical proximity were more likely to collaborate, the abbreviations of the geographic name and the name of university were listed in Table 2. Shandong Mental Health Center and Shandong University, with the greatest number of collaborations, had a close collaborative research relationship.

Analysis on collaboration at the macro (regions) level
From 2003 to 2012, the most productive regions were Shanghai, Beijing and Shandong (Fig. 3), with 2,154 psychiatric research papers originated from Shanghai, making up 13.07 % of all such papers. Using the top 31 most productive regions in China to construct the map of scientific collaboration (Fig. 4), the network included 31 nodes and 226 lines. The average path length was 1.858 and the average clustering coefficient was 5.363, indicating a pronounced clustering effect. We applied Core-Periphery analysis and calculated the correlative value of the collaborative network to be 0.904 (UCINET 6.0). Once again, a well-delineated Core-Periphery structure appeared (Fig. 5).
We analyzed the effect of scientific collaboration on regions' scientific output by correlating nodes to their corresponding scientific research achievements. For each of the 31 regions, we examined the number of ties and number of papers produced (Table 4). We found that the quantity of international collaborations correlates with research output. This suggested that international scientific collaboration greatly influenced scientific output in this field, whereby countries that collaborated frequently had a greater research output.
A centrality analysis (Table 5) revealed that Beijing, with the highest degree centrality of 120, highest betweenness centrality of 177, and lowest closeness degree of 90, was the center of the scientific collaboration network of China's psychiatry field. Beijing's extensive research output made it a major producer of international publications.

Discussion
As China's economy has grown and social pressures have increased, the prevalence of mental disorder has also grown dramatically. Because of the diversity and complexity of diseases this encompasses, scientific collaboration is indispensable if progress is to be made in the treatment of mental disorder. Although several studies have shown that collaboration has increased at the level of authors, institutions and regions [14][15][16], few have reported this phenomenon within the psychiatry field. This study retrieved bibliographic data of Chinese psychiatry research from 2003 to 2012 from CNKI and WanFang Database. We constructed and analyzed the structure of scientific collaboration at the micro (authors), meso (institutions) and macro (regions) levels based on SNA and found that scientific collaboration was correlated with this field's development.
The authors who had the highest centrality were the central authors of the whole network, suggesting that they heavily influenced research in the Chinese Fig. 2 The core-periphery structure map of the institutional collaboration network on Chinese psychiatry research (The geographic name and the name of university adopted the official abbreviation, see Table 2) Fig. 3 Regions distribution of Chinese psychiatric papers (The geographic name adopted the official abbreviation, see Table 4) psychiatry field and were thus leader of the field. In the era of a knowledge-based economy, as the most important economic factors, intellectual resources become more obviously valuable. Thus, scientific collaboration played a large role in the emergence of a subject leader.
The institutions which had the highest centrality were the center of multi-institutional collaboration in the Chinese psychiatry field possessing and controlling a great deal of resources for research. From the analysis on Core-Periphery structure, collaborating academic institutions obviously demonstrated 'center effect' , while those collaborating with famous institutions demonstrated 'elite institutions assembling'. In other words, the phenomenon that institutional collaborations was within the same country showed geographical characteristics. Thus, the other research institutions need to collaborate with institutions which collaborated closely to strive for the more scientific research resource.  The structure map of collaboration network among regions on Chinese psychiatry research At a regional level, Beijing had the highest centrality and was thus in the most central position. Our analysis suggested that regional scientific collaboration was positively correlated with total output in terms of scientific research. From the analysis on Core-Periphery structure, developed cities such as Beijing and Shanghai collaborated closely. China's rapid economic development, which has encouraged collaborative behavior, has contributed to increases in research output. Higher-income regions prefer to collaborate with each other while lower-income regions prefer to collaborate with higherincome regions in order to increase the quality of their research.

Conclusion
This study described the collaborative behaviors of research in the Chinese psychiatry field at the micro (authors), meso (institutions) and macro (regions) levels. Based on the centralities of the author ranking, academic leaders will be selected more easily. Furthermore, studying these collaborations not only can help researchers to master the forefront of this field but also provide scientific evidences and suggestions for policymakers to guide and manage the Chinese psychiatry field in the future. Ningxia