*2.3. Data Analysis*

We used STATA version 15.0 (STATACorp., College Station, TX, USA) to analyze the final data in step 4 using the following information on the articles: authors' affiliations, titles, journals' names, keywords, total number of citations, and abstracts.

Publications' general characteristics, namely year of publication, the number of papers per year of each country, accumulative citations from published year to 2018, yearly mean citation rate, total usage in the last six months and the last five years, and mean use rate in the last six months and the last five years, were described. We used STATA to calculate the number of papers by country in abstracts.

A network graph showing the co-occurrence of terms in title and abstracts was established by VOSviewer (version 1.6.11, Center for Science and Technology, Leiden University, Leiden, The Netherlands). A co-occurrence network can be described as the collective interconnection of terms and is generated by connecting pairs of terms based on a set of criteria that defines co-occurrence. In this study, when term A and term B both appear in title and/or abstract of a particular publication, they are said to "co-occur". Additionally, another paper may contain terms B and C, and so on. By linking the co-occurrence of the identified terms, we created a co-occurrence network of terms.

Latent Dirichlet Allocation (LDA) was applied to classify papers into corresponding topics. LDA is a common topic modeling algorithm for text mining to determine the relationships of text documents. In LDA, each term is regarded as a random vector with the probability of drawing the words/texts associated with that term, in other words, that vector. The probability of a paper is calculated based on the probability of the terms, and papers with similar probability will be classified into one group [20–25]. Thus, by applying LDA, we would be able to recognize the interdisciplinary structure of research development as well as obtain an in-depth view of hidden themes of research on interventions in asthma [26]. We reviewed and read through the most cited papers of each group and assign the labels for each topic manually. In addition, to analyze the relationship of the research area, we utilized coincidence analysis using the STATA command 'precoin' and presented the results in the form of a dendrogram. Dendrograms graphically illustrate the information concerning which observations are grouped together at various levels of similarity.

Analytical techniques for each data types are summarized below (Table 1):

