Li Jiahui,Zhao Peiyao,Yuan Xiaoliang,2＊

(1.School of Business Administration,Shenyang Pharmaceutical University,Shenyang 110016,China;2.Research Institute of Drug Regulatory Science,Shenyang Pharmaceutical University,Shenyang 110016,China)

Abstract Objective To study the research status,research hotspots and development trends in the field of real-world data(RWD) through social network analysis and knowledge graph analysis.Methods RWD of the past 10 years were retrieved,and literature metrological analysis was made by using UCINET and CiteSpace from CNKI.Results and Conclusion The frequency and centrality of related keywords such as real-world study,hospital information system (HIS),drug combination,data mining and TCM are high.The clusters labeled as clinical medication and RWD contain more keywords.In recent 4 years,there are more articles involving the keywords of data specification,data authenticity,data security and information security.Among them,compound Kushen injection,HIS database and RWD are the top three keywords.It is a long-term research hotspot for Chinese and western medicine to use HIS to study clinical medication,clinical characteristics,diseases and injections.Besides,the research of RWD database has changed from construction to standardized collection and governance,which can make RWD effective.Data authenticity,data security and information security will become the new hotspots in the research of RWD.

Keywords:social network analysis;knowledge graph;real-world data;data specification;technical specification

Real world study (RWS) originated from the supplement and development of randomized controlled trials (RCTs),which was organized by the British Medical Research Council in 1948.But the strict research conditions made it difficult to generalize conclusions.It was not until 1993 that Kaplan et al proposed the concept of the real world for the first time in a study involving 591 patients with hypertension.RWS was widely concerned after the Global Registry of Acute Coronary Events was launched in the United States in 1999[1].Then,the concepts of real-world data (RWD) and real-world evidence (RWE) were derived.RWD and RWE were formally introduced into China in 2010[2].RWD refers to a variety of data collected daily related to patients’health status and/or diagnosis and treatment and health care services[3].The application of RWE relies on high-quality RWD[4].Therefore,the research of RWD is vital to the development of RWS and the domestic medical field.The research of RWD has also received attention from the Chinese government.In 2015,the State Council issued the “Outline of Action to Promote the Development of Big Data”,which greatly promoted the development and application of big data[1].In January 2020,National Medical Products Administration (NMPA) issued “Guidance for Drug R&D and Evaluation Supported by Using Real-World Evidence (Trial)”[5].In August 2020,the Center for Drug Evaluation issued “Technical Guidance for Real-World Study to Support Children’s Drug R&D and Evaluation (Trial)”[6].In April 2021,the Center for Drug Evaluation issued “Guidance of Real-World Data for Generating Real-World Evidence (Trial)”[3].The frequent release of guidelines shows that China has paid more attention to RWD,which results in the fast development of RWD.Through social network analysis and the visual knowledge graph,the hotspots and trends of the RWD research in the past ten years in China are comprehensively presented,objectively reflecting the overall situation of the research of RWD.

1 Data sources and methods

1.1 Data sources

The data was retrieved from China National Knowledge Infrastructure (CNKI) for the subject or article named RWD,and the retrieval time was from the establishment of the database to May 31,2021.The selected journals were academic journals,and the source categories were Chinese Social Science Citation Index (CSSCI),core journals and EI.A total of 277 articles were obtained through retrieval.Then these articles were strictly screened to eliminate papers that did not belong to RWD or non-research papers that included solicits,notices,announcements and so on.138 articles,which were published from 2011 to May 31,2021,were selected as the research objects.

1.2 Methods

Social network analysis (SNA) and knowledge graph analysis were combined to make a visual analysis of research hotspots and development trends of RWD.SNA focused on the analysis from the perspective of social relations among individuals and the network structure formed by social relations.SNA could excavate the association patterns hidden in the complex social environment.SNA included centrality analysis and cohesive subgroup analysis[7].Knowledge graph (KG) was a series of different graphs that showed the process of knowledge development and the structural relationship.KG described the knowledge resources and their carriers with visualization technology,which was a method to mine,analyze,construct,draw and display knowledge and their interconnections.The knowledge graph drawing tools were UCINET and CiteSpace[8,9].

1.3 Normative approach

In the literature obtained from CNKI,keywords with the same meaning were uniformly standardized.For example,hospital information management system,and hospital information system/HIS were uniformly replaced with hospital information system(HIS).Electronic medical record was replaced with EMR and so on.

2 Results

2.1 Co-word network analysis based on keywords

The literature data in Refworks format was imported through CiteSpace software.269 keywords in literature were analyzed by a co-word network.The analysis results were shown in Fig.1 and Table 1.The shade of the line color on the map represented different years in Fig.1.The later the keywords appeared,the lighter the line color.The bigger the node,the more frequently it appeared.RWS,HIS,drug combination,RWD,association rules,data mining and electronic medical data were the hot topics of RWD research in the last 10 years.

Fig.1 Co-word network graph of RWD based on keywords from 2011 to 2021

Table 1 High-frequency keywords of RWD research (Threshold:frequency ≥ 4)

(to be continued)

Continued Table 1

As can be seen from Table 1,there were 11 keywords related to medicines before 2016.These 11 keywords were concentrated in the field of traditional Chinese medicine (TCM) research,and 8 of them were related to the dosage form of TCM injections.In 2016,the first western medicine named bicyclol tablets appeared.Afterwards,there were 29 keywords related to medicines,including 13 TCM and 11 western medicines.Among them,there were 8 TCM injections and 4 western medicine injections.Therefore,drug research was mainly about TCM,especially TCM injections,and the research of western medicines was increasing gradually.Compound Kushen injection appeared in 2017 for the first time,with a relatively high frequency of 7.It was a hot research drug in recent years.In the study from 2011–2021,there were 62 keywords involving diseases in the data,including cancer,circulatory system disease,central nervous system disease,respiratory system disease,immune system disease,digestive system disease,endocrine system disease,urinary system disease and reproductive system disease.Cancer as the keyword was widely studied,there were 13 types of cancer.“Comorbidities” and “cerebral infarction”were hotspots in the research of diseases because of their high frequencies.There were 21 keywords about TCM,including TCM syndrome,throttle,syndrome differentiation and treatment,and solar terms,etc.The frequency of TCM syndrome was 9.TCM syndrome was a hot topic in the field of TCM.

2.2 Centrality analysis

There are 3 indexes commonly used in the centrality analysis of SNA:degree centrality,closeness centrality and betweenness centrality.

Degree centrality refers to the number of nodes that are directly related to a node.The higher the value of degree centrality,the wider range of this node[10,11].Ucinet was used to analyze keywords degree centrality of RWD research.The results were shown in Table 2.In addition to RWS and RWD,HIS,drug combination,association rule,data mining and other keywords had high value of degree centrality.RWD research in recent 10 years paid more attention to RWD data mining and processing technology.

Closeness centrality refers to the sum of distances from one node to all other nodes.In 1950,Bavelas defined closeness centrality as the reciprocal of distance.The smaller the value of closeness centrality,the shorter distance to the center of network[10,11].It can be seen from Table 2 that apart from RWS and RWD,HIS,TCM syndrome,clinical characteristics,safety,type 2 diabetes mellitus were less close to the center,which indicated that keywords in clinical medicine were in a more central position in the network.

Betweenness centrality refers to the number of shortest paths passing through a node.The higher the betweenness centrality value is,the stronger control ability of this node has[10,11].As shown in Table 2,in addition to RWS and RWD,HIS,data mining,adverse drug reactions,TCM and others had high degree of centrality.These keywords had been studied more frequently in different articles in RWD research network.They played the role of connecting other keywords to form a research network.

Table 2 The centrality analysis results of keywords in RWD research (Top 10)

Table 3 Top 14 keywords with the strongest citation bursts

Based on the results of 3 indexes of centrality analysis,the following conclusions were obtained:(1) HIS had more advantages in the 3 centralities analysis.It was an important research hotspot of RWD.There were 95 of the 138 articles related to HIS,and 80 of the 95 studies were the analysis of clinical characteristics and prescription rules of drugs or diseases based on HIS.In these studies,the rational use of HIS to make it an important source of RWD occupied a mainstream position;(2)Keywords related to TCM research generally had good closeness centrality and betweenness centrality.RWD was also widely studied in the field of TCM;(3) The centrality of keywords related to drug combination and data mining was second only to that of HIS.They often appeared together with HIS or keywords of TCM in literature and were highly popular in research.

2.3 Cluster analysis

CiteSpace was used to conduct clustering analysis on research data,and the data was divided into 11 cluster subgroups,as shown in Fig.2 and Fig.3.Each cluster was composed of multiple closely related keywords,and its Silhouette value (average profile value of the cluster) was greater than 0.7,which meant that the 11 clusters were reasonable and reliable.The smaller the clustering order number was,the more keywords it contained,and the more important the clustering field was[12].The number of keywords in cluster 0 and cluster 1 was 49 and 40 respectively.They were two important hot research directions.The distance between cluster 0 and cluster 1 was relatively far,which meant the research contents of the two cluster subgroups were less correlated and obviously independent.

Fig.2 Keywords clustering graph of RWD research

Fig.3 Keywords timeline diagram of RWD research

Cluster 0 was labeled as clinical medication,including:(1) Classification of patients such as children,elderly and adolescent patients;(2) Injection drugs and Yinhua Miyanling Pian and Suxiao Jiuxin Pills;(3) Diseases such as cardiovascular diseases,shock,cerebral infarction,cancer,kidney diseases and respiratory diseases;(4) Treatment dose,outcome and risk control;(5) HIS.Through studying literature and Fig.3,it can be seen that HIS used to analyze the clinical characteristics of drug treatment was a research hotspot from 2011 to 2021.Therefore,we can predict that the use of HIS to analyze clinical medication is the main research direction in the future.

Cluster 1 was labeled as RWD,including:(1) Data specification covering data authenticity,designing criteria,and regulation decision making,etc.;(2) Data collection covering research data infrastructure and RWD;(3) Data analysis covering statistical analysis;(4) Medical treatment covering tislelizumab and central venous catheter (CVC);(5)Diseases covering COVID-19 and advanced urothelial cancer.According to Fig.3 and literature,research on RWD started in 2013,and the content from 2013 to 2021 showed an increasing trend.Data specification was the key field of RWD research from 2016 to 2021.Data authenticity,data security and information security received more attention in 2021.Data specification to ensure data authenticity will be the main research direction in the future.

Cluster 2 had the same clustering label as cluster 0,but the research focus was different.The focus of cluster 2 on clinical medication was to analyze the clinical characteristics of TCM and Chinese patent medicine in the treatment of diseases by using TCM information sharing system.From 2011 to 2021,this clustering content showed an increasing trend.The application of RWD in the field of TCM may still be a hot research direction in the future.

2.4 Evolution direction analysis

Through the analysis of burst detection in CiteSpace,as shown in Table 3,from 2011 to 2021,the burst strength of HIS database,outcome research,TCM syndrome,drug combination,compound Kushen injection,Shenxiong glucose injection and RWD were relatively high.They were important in the RWD field.Compound Kushen injection had the highest burst strength,followed by HIS database.RWD emerged from 2019 to 2021,which received more attention in the past two years.At the same time,RWD research in the field of injections may become a hot spot in the future.

3 Conclusion

Based on SNA,KG and literature analysis,the early topics of RWD research included the following items such as information and data technology,clinical medicine,clinical/drug epidemiology,statistical analysis,artificial intelligence,health/pharmaceutical economics,policy and management,institutions[2].With the establishment and improvement of the database,problems such as lack of process quality control,incomplete data and inconsistent standards arose.Since 2016,structure,data governance,technical specification and standardization have become hot research topics.From 2019 to 2021,how to evaluate RWD for producing RWE to support drug regulatory decisions has become a key issue.Since 2019,the NMPA has successively issued guidance on RWD.So,research on RWD surged between 2019 and 2021,and new research directions were proposed in 2021,including data authenticity,data security and information security.The introduction of RWD promoted the development of technology and medicine.Shown from the timeline diagram in Fig.3,each stage had different technologies to develop and diseases to conquer,which stimulated the creativity of governments and companies.At the same time,technological progress and medical research and development also promoted the development of RWD.However,because the system of RWD was complex,the establishment of source data to research database lacked standardized technical guidance.Through the above analysis,conclusions are drawn:(1) It was found that the use of HIS in traditional Chinese and western medicines to explore clinical medication,clinical characteristics,diseases and injections has been a long-term research hotspot;(2) RWD database research should turn from construction to standardized collection and governance so as to make RWD real and effective;(3) Data authenticity,data security and information security will become the new hot research directions of RWD.