Meta-Analysis of Correlations between Altmetric Attention Score and Citations in Health Sciences

Introduction. In recent years, several controversial reports of the correlation between altmetric score and citations have been published (range: -0.2 to 0.8). We conducted a meta-analysis to provide an in-depth statistical analysis of the correlation between altmetric score and number of citations in the field of health sciences. Methods. Three online databases (Web of Science, Scopus, and PubMed) were systematically searched, without language restrictions, from the earliest publication date available through February 29, 2020, using the keywords "altmetric," "citation," and "correlation." Grey literature was also searched via WorldCat, Open Grey, and Google Scholar (first 100 hits only). All studies in the field of health sciences that reported on this correlation were included. Effect sizes were calculated using Fisher's z transformation of correlations. Subgroup analyses based on citation source and sampling methods were performed. Results. From 27 included articles, 8 articles comprise several independent studies. The total sample size was 9,943 articles comprised of 35 studies. The overall pooled effect size was 0.19 (95% confidence interval 0.13 to 0.26). Bivariate partial prediction of interaction between effect size, citation source, and sampling method showed a greater effect size with Web of Science compared with Scopus and Dimensions. Egger's regression showed a marginally nonsignificant publication bias (p=0.055), and trim-and-fill analysis estimated one missing study in this meta-analysis. Conclusion. In health sciences, currently altmetric score has a positive but weak correlation with number of citations (pooled correlation=0.19, 95% C.I 0.12 to 0.25). We emphasize on future examinations to assess changes of correlation pattern between altmetric score and citations over time.