Background

The most common way to assess the impact of an article is based on the number of citations [1]. The mean number of citations for all articles published in a journal in the preceding 2 years is called the journal’s impact factor [1]. However, the number of citations and the journal’s impact factor do not precisely reflect whether the message of the article is reaching a wider audience [2]. Currently, social media is being used to disseminate the contents of scientific articles [3, 4]. However, until recently, the impact of scientific articles on social media was not quantified. To measure this type of impact, a new score (called Altmetric) was created [3, 4].

Altmetric measures the impact of each article through the attention attracted online [3]. Moreover, the Altmetric score reveals the instantaneous attention attracted online for articles in news outlets, comments on blogs, number of tweets, and mentions on social media. There are two types of Altmetric scores. The Altmetric-mentioned score includes data sources involving social media (e.g., Facebook, Twitter), newspapers, encyclopedias (e.g., Wikipedia), online platforms (e.g., Faculty1000 and publication peer reviews), videos on YouTube, question-and-answer sites (e.g., Q&A stack overflow), and policy documents in PDF form available over the internet. The Altmetric reader score includes data sources involving reference managers available online (e.g., Mendeley, CiteULike, and Connotea). The Altmetric score can be graphically represented by a “donut.” The different colors of the Altmetric donut represent the number of mentions on each specific online media source. For example, mentions on Twitter are represented in blue (Fig. 1).

Fig. 1
figure 1

Description of the Altmetric donut

Full size image

Research about Altmetric has been increasing and becoming more popular in recent years [5]. However, most articles about Altmetric published to date are only introductory tutorials or editorials [1, 3, 4, 6, 7]. Patthi et al. [2] published a systematic review in the field of dentistry that aimed to analyze the correlations between journal citations and Altmetric scores. The review concluded that journal citations and Altmetric scores are positively correlated (with Pearson’s r ranging from 0.30 to 0.61).

Recent articles from several research fields [8, 9] showed that the number of article citations and Altmetric score are positively correlated. Finch et al. [10] showed that the number of tweets (i.e., an Altmetric component) could predict citations within the first 3 days of article publication. Araujo et al. [11] found that number of citations and journal’s impact factor were positively associated with Altmetric [11]. These authors also found that the number of years since publication and having a descriptive title (i.e., a title describing the aim of the study but not revealing the main conclusions) were negatively associated with Altmetric [11]. Therefore, it is assumed that the publishing journal and publishing article variables, such as citation counts, journal impact factor, access counts (considered the sum of HTML views and PDF downloads), papers published as open access, time since publication, and press releases generated by the publishing journal, are likely to be associated with Altmetric [11]. This systematic review aims to summarize all available evidence on the associations between the publishing journal and publishing article variables and Altmetric scores.

Methods

Research question

What publishing journal and publishing article variables are associated with Altmetric scores?

Search strategy for identification of studies

Systematic searches were conducted on MEDLINE, EMBASE, CINAHL, CENTRAL, and Cochrane Library, as per the Cochrane Handbook [12], including publications from the inception of these databases until March 31, 2021, without language restrictions. As the topic is novel, we used only two search terms (Altmetric OR Altmetrics) in all databases to ensure a more sensitive search strategy.

Inclusion and exclusion criteria

We included any original research studies that measured any type of association between the publishing journal and/or the publishing article with Altmetric scores, such as citation counts (i.e., number of citations), journal impact factor, access counts (considered the sum of HTML views and PDF downloads), papers published as open access, time since publication, and press releases generated by the publishing journal. Studies that did not have at least one of these variables were excluded. Letters to the editor, editorials, and conference abstracts were also excluded. Moreover, we excluded articles that included a subset of highly cited papers or with extremely high Altmetric scores.

Data collection

Two review authors (AA and AV) independently screened all studies for eligibility and data extraction. All discrepancies identified during the stages and throughout the review were resolved via discussion or through arbitration provided by another investigator (DN). The study selection process included (1) screening the titles and abstracts and (2) screening of full-text articles.

Data extraction

Two review authors independently extracted the following data: (1) authors, (2) year of publication, (3) research field, (4) sample size of studies analyzed, (5) study design of the included studies, (6) study aims, (7) study results, and (8) study conclusions. Variables about the publishing journal included (9) journal impact factor, (10) access counts (considered the sum of HTML views and PDF downloads), (11) papers published as open access, and (12) press releases generated by the publishing journal. Variables about the publishing articles included (13) citation counts (i.e., number of citations), and (14) time since publication. We also collected data related to (15) the Altmetric mentioned score and (16) the Altmetric reader score. We contacted authors by email to request additional information that was not reported in the original manuscripts.

Ethics and registration

No ethical approval was required for this study. As this review has no health-related outcomes, no registration was needed [13].

Data analysis and quality of studies

Due to a large data heterogeneity, meta-analysis was not possible. For this reason, our results are reported descriptively. The quality criteria of included articles were analyzed using an adapted version of the Appraisal Tool for Cross-sectional Studies (AXIS) [14]. This tool was developed to systematically assess the quality of cross-sectional studies by assessing 20 items. Each item is rated as “yes,” “no,” or “don’t know/no comment” [14]. The AXIS was the tool that best covered the included studies. We adapted the AXIS by excluding items 7 (Were measures undertaken to address and categorize non-responders?), 9 (Were the risk factor and outcome variables measured correctly using instruments/measurements that had been trialed, piloted or published previously?), 13 (Does the response rate raise concerns about non-response bias?), 14 (If appropriate, was information about non-responders described?), and 20 (Was ethical approval or consent of participants attained?), as these items are unrelated to the aims of our review.

Results

Search results

The initial search yielded 1109 potentially eligible studies. After screening by title and abstract and removing duplicates, we considered 42 potentially eligible studies for inclusion and retrieved full-text articles. Nineteen published studies [11, 15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32] met the inclusion criteria and were included in this review. The study flow diagram of the eligibility assessment is presented in Fig. 2.

Fig. 2
figure 2

Study flow diagram of the eligibility assessment

Full size image

Quality of studies

We did not consider the total score based on the instructions of the AXIS [14]. However, we observed that the studies included in general did not have good methodological quality (Table 1). We observed that most studies did not select a representative/random sample of a population (item 5), as most studies sampled the articles from main journals in their fields.

Characteristics of included studies

The 19 eligible studies were published between 2014 and 2021 and summarized a total of 573,842 articles. The study designs of the included articles were mixed research designs [15,16,17,18,19,20,21,22,23,24,25,26,27,28, 30,31,32] and randomized controlled trials [11, 29]. The research fields of these articles included biomedicine [20], burn care [31], ecology and conservation [30], emergency medicine [19], engineering and technology, gastroenterology and hepatology [26], general medicine [18], joint arthroplasty [29], medical education [15], medical and natural sciences [25], multidisciplinary [22], oncology [24], physiotherapy [11, 16], plastic surgery [17], psychiatry [23], radiology [32], rheumatology [21], social sciences and humanities, solid organ transplantation [27], and spine [28]. The main objective of the included studies was to assess the association between Altmetric scores and variables such as citation counts (i.e., number of citations), journal impact factor, access counts (considered the sum of HTML views and PDF downloads), papers published as open access, time since publication, and/or press releases generated by the publishing journal. A summary of the methods, data analysis, results, and conclusions is presented in Table 2.

Statistical analysis and associations of included studies

Different types of analyses were conducted in the included studies: correlation analysis [15, 17,18,19, 22, 23, 26,27,28,29, 31, 32], regression analysis [11, 16, 21], boosted regression trees analysis [30], principal component analysis, and factor analysis [20, 25]. The main results of the included studies demonstrated that the variables citation counts (i.e., number of citations), journal impact factor, access counts (considered the sum of HTML views and PDF downloads), papers published as open access, time since publication, and press releases generated by the publishing journal were associated with Altmetric scores. The magnitude of these associations ranged from weak to strong (Table 3).

Table 1 Quality assessment of the included studies by the AXIS tool
Full size table
Table 2 Summary of the objectives and methods according to the variables of interest in the review and author’s conclusions
Full size table

Summary of the association between variables of the publishing

Table 3 Summary of the association between variables of the publishing journal and the publishing articles with Altmetric scores
Full size table

Discussion

We aimed to summarize all available evidence on the associations between the publishing journal and publishing article variables and Altmetric scores. We found that citation counts (i.e., number of citations), journal impact factor, access counts (considered the sum of HTML views and PDF downloads), papers published as open access, time since publication, and press releases generated by the publishing journal were associated with Altmetric scores. The magnitude of these associations ranged from weak to strong. In addition, we observed that citation counts and journal impact factor were associated with Altmetric scores in all included studies [11, 15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32].

There is a previous systematic review about the correlation between citation counts and Altmetric in medical research [2]. Moreover, there are articles that have measured associations between citation counts and Altmetric scores [15, 32]. In accordance with the systematic review [2] and these articles [15, 32], we found a positive correlation between citation counts and Altmetric scores. Similarly, our overview indicated positive associations (ranging from weak to moderate) between citation counts and Altmetric scores [11, 15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32]. These results are similar to those related to a journal’s impact factor [11]. This is not surprising, because a journal’s impact factor is based on citation counts of scientific articles [1]. We also found that most included studies, with the exception of those by Araujo et al. [11, 16] and Knight [27], did not analyze the Altmetric reader score. Thus, our findings are largely based on the Altmetric-mentioned score. We strongly recommend that further investigations on Altmetric reader score be conducted.

Regarding the quality of studies, the main limitation we observed was the lack of reporting the methods in detail. Items related to sampling, selection criteria, and statistical analysis in general were poorly described. On the other hand, the articles were clear in terms of data analysis and results. Finally, most authors presented the limitations of the study in their discussion and disclosed their potential conflicts of interest.

No studies identified specific characteristics of articles, for example, analysis of studies that published popular/hot topics (e.g., studies on coronavirus, miraculous diets, cancer prevention, early life on earth, religious evidence). Moreover, there is no analysis of studies comparing whether the direction of the results (i.e., positive versus negative conclusions) influences Altmetric scores. These characteristics are likely to increase the number of people who access and share these articles on social media [11]. We recommend that future studies identify if these characteristics are associated with Altmetric scores.

Finally, we propose 4 suggestions to improve the social impact and visibility of scientific articles: (1) select high impact factor journals for submission of articles; (2) use provocative titles (titles expressing the results of the trial) or interrogative titles; (3) use social media (Twitter, Facebook, etc.), websites, and blogs to disseminate principal findings; and (4) post the article with its digital object identifier (DOI) or the journal’s link to the article to be captured by Altmetric. These simple strategies are likely to improve the visibility of articles to a larger readership [5, 33]. The major strength of this study is the inclusion of articles from all fields of the research (n = 565,352 articles analyzed). On the other hand, a possible limitation of this study is the large heterogeneity of the included studies. Because of this, the data were analyzed only descriptively. Another potential limitation of our review is related to the selection of the databases we chose. We decided to cover the most comprehensive databases, such as MEDLINE, CINAHL, EMBASE, Cochrane Library, and CENTRAL, and we might have missed some eligible articles published in smaller databases or gray literature.

Conclusion

Citation counts, journal impact factor, access counts (considered the sum of HTML views and PDF downloads), papers published as open access, time since publication, and press releases generated by the publishing journal were associated with Altmetric scores.