*3.2. Highly-Frequent MeSH*/*Subheading Combination Terms and Their Cluster Pattern*

From the articles included, 26 and 39 high-frequency major MeSH/subheadings combination terms were extracted for the first and second decade, respectively, based on the method for the g-index of the word frequency mentioned above, with a cumulative frequency percentage of 32.95% and 37.39%, respectively (Tables S2 and S3 in Supplementary Materials). Furthermore, these terms were subject to a co-word biclustering analysis to reveal the research hotspots for climate change and infectious diseases in the past two decades.

The high-frequency terms were classified into four clusters for the first decade and five for the second decade using the biclustering analysis, as presented in Figures 3 and 4. These two figures also show the mountain and matrix visualization of these terms. In the mountain visualization, each 3D peak labeled by the cluster number contains a cluster of terms, of which the location on the plane, volume, height, and color are used to portray information about a cluster. The distance between two peaks on the plane reflects the relative similarity of two clusters. There is positive correlation between the peak's height and the cluster's internal similarity. The volume of a peak is positively correlated with the number of terms classified into a cluster. In addition, the peak's color represents the internal standard deviation of a cluster's terms. Blue represents a high internal standard deviation of the objects inside, whereas red represents a low internal standard deviation. In the matrix visualization, the high-frequency terms are listed on the right side. The number before each term represents its serial number (See Tables S2 and S3). The top and left hierarchical trees display the relationships among the included articles and those among the high-frequency terms, by which the themes of all of the clusters have been able to identify and summarize, and insights into the representative articles of each cluster could be attained as well. The hotspots of climate change and infectious diseases revealed by the cluster analysis of high-frequency terms during the two periods are presented in Table 2.

**Figure 3.** Mountain and matrix visualization of biclustering of highly-frequent major medical subject headings (MeSH)/subheading combination terms and articles on climate change and infectious diseases from 1999 to 2008.



13

**Figure 4.** Mountain and matrix visualization of biclustering of highly-frequent major MeSH/subheading combination terms and articles on climate change and infectious diseases from 2009 to 2018.
