*2.2. Data Analysis*

The corrected data after the screening was imported to STATA for further analysis using the following information of the articles: authors' affiliations, the title of papers, the journals' name, authors' keywords, the number of citations, research areas, and abstracts.

Several basic characteristics of the data sets were included publication year, the number of papers /per year, total citations up to 2018, average citation rate per year, total number of downloads in the last six months/five years, and average number of downloads (mean use rate) the last six months/five years. Two network graphs showing the countries collaboration and co-occurrence terms in title and abstracts were established by VOSviewer (version 1.16.15, Centre for Science and Technology Studies (CWTS), Leiden University, Leiden, The Netherlands). Latent dirichlet allocation (LDA) was used for classifying papers into topics [41–45]. The titles and abstracts of most cited papers within each group were reviewed. After discussing with COPD specialists, the labels for each topic were named. In addition to the number and percentage of publications of each topic, these topics were ranked based on the total number of publications in the past five years to explore the research interests. Table 1 shows the methods and results for each kind of data.


