Altmetric Analysis of the Online Attention Directed to Artificial Intelligence Literature in Ophthalmology

To the Editor:

Artificial intelligence (AI) has become an increasingly important topic in ophthalmology due to its potentially indispensable role in the rapid and efficient detection of ocular pathology.^1–3 Indeed, a bibliometric analysis of the most cited ophthalmology articles in Asia identified AI as an emerging area of investigation within the region.⁴ However, citation count captures scientific, not social, dissemination. Here, we provide a perspective distinct from traditional bibliometric analyses by evaluating the social impact of literature concerning AI in ophthalmology.

The Web of Science database was queried by “Topic” on October 9, 2022, using search terms and Boolean operators: Ophthalmol* OR Optometry AND “Artificial Intelligence” OR “Deep Learning” OR “Machine Learning.”⁵ Exclusion of duplicates yielded 2927 articles. Altmetric Attention Score (AAS), a measure of the online dissemination of an article,^6–8 was obtained using Altmetric Explorer, and articles were ranked. Each article was reviewed by 2 investigators (T.B. and P.A.P.), and only the top 100 discussing the application of AI primarily in ophthalmology were included in this analysis (Supplementary Digital Content, Table 1, https://links.lww.com/APJO/A207). Pertinent article, author, and journal characteristics of the top 100 articles were extracted. Pearson correlation coefficient, Mann-Whitney U test, Kruskal-Wallis test, and Fisher exact test were calculated where appropriate using GraphPad Prism 9 (San Diego, CA). P<0.05 was considered statistically significant.

The median AAS of selected articles was 36.5 (range: 20–1231), predominantly driven by mentions on Twitter (median: 33.5; range: 0–796). Robust positive correlations were observed between AAS and citation count (r=0.58; P<0.001) and citations per year (r=0.54; P<0.001). Ninety-nine percent of articles were published within the past 7 years, with the majority (54%) published between 2020 and 2022 (Fig. 1), highlighting the relative recency of social interest in the area. Eighty-three percent of articles were open-access, a significant difference from the 35.1% of open-access articles identified in our initial search (P<0.001).

FIGURE 1: Trends in the 100 most mentioned articles examining artificial intelligence and ophthalmology. The bar graph, measured against the left axis, indicates the number of most mentioned articles published in a particular year, and the line graph, measured against the right axis, indicates their median Altmetric Attention Score (AAS).

Ophthalmology-specific journals were the source of 54% of articles (Supplementary Digital Content Table 2, https://links.lww.com/APJO/A208), and those published in JAMA Ophthalmology (16%) and Ophthalmology (13%) collectively garnered higher median AAS relative to counterparts in other ophthalmology-specific journals (P=0.023). As depicted in, Supplementary Digital Content, Table 3, https://links.lww.com/APJO/A209, most articles examined the application of AI for pathologies principally relevant to medical and/or surgical retina (52%). Across articles, there were significant differences in median AAS by subspecialty (P=0.039).

Females comprised 21% of first authors and 16% of senior authors. While no significant differences in median AAS were observed by senior author gender (P=0.24), articles by male first authors trended toward having a higher median AAS (P=0.077). Senior authors were from 19 countries (Supplementary Digital Content, Table 4, https://links.lww.com/APJO/A210). Aggregated by region, they primarily originated from North America (43%), Europe (29%), and Asia (24%). There were no disparities in median AAS of articles by region (P=0.77). The authors with the most articles in the top 100 were D.S.W. Ting (14%) and T.Y. Wong (12%). Eighty-four percent and 58% of articles were the product of multi-institutional and multinational collaborations, respectively.

Despite limitations associated with AAS,^6–8 measuring the social impact of research offers complementary insights to traditional bibliometrics. The present investigation demonstrates that literature regarding AI in ophthalmology is widely discussed on online platforms. Of particular social interest is the utility of AI for the detection of retinal pathologies such as diabetic retinopathy and age-related macular degeneration. As evidenced here, authors from Asian countries have adopted a significant role in the production of socially impactful scientific literature concerning this topic. By increasing engagement with social media, investigators exploring AI in ophthalmology have the potential to disseminate research to a substantially broader audience.