SHORT REPORT
Uncovering Publisher-level Self-citation Variations: A Study in General Internal Medicine
2026 年 8 巻 4 号 p. 135-138
詳細
2026 年 8 巻 4 号 p. 135-138
Biased citation practices by authors, reviewers, and editors raise ethical concerns1),2). Journal-level self-citation (JSC) to increase the impact factor (IF) has been reported1),3), leading to citation manipulation and distorted academic evaluation1). The IF is influenced by JSC; therefore, Journal Citation Reports4) also includes the metric “IF excluding JSC” to discourage intentional journal self-citation practices3).
Recently, the practice of “publisher-level self-citation” (PSC), where journals published by the same publisher cite each other to circumvent direct self-citation, has attracted increasing attention3),5),6). As citation practices vary across fields, examining PSC within a single research category is essential. However, its actual extent remains unclear. Additionally, the surge in coronavirus disease 2019-related publications in recent years has affected citation metrics in medical journals3). Therefore, this study aimed to examine the variability in PSC rates at the publisher level within the field of general and internal medical research.
We performed a descriptive cross-sectional study using data from general internal medicine journals listed in the 2023 Journal Citation Reports4). A total of 333 journals classified under “Medicine, General & Internal” were identified from this source. Journals with a 2023 IF reported as “<0.1” or “N/A” were excluded from the analysis.
Between December 2024 and January 2025, we extracted the citation counts and publisher information for each journal from Journal Citation Reports. The number of PSCs for each journal was defined as the number of citations received from other articles by journals within the same publisher group (including the citing journal). Citation data were aggregated at the publisher level (Supplementary Table 1), and PSC rates were calculated by dividing the total number of PSCs per publisher by the total number of citations within each publisher group. In addition, for journals classified under “Medicine, General & Internal” within each publisher, we calculated the median IF (interquartile range) at the publisher level. When a publisher had only a single journal in this category, the IF of that journal was reported. The mean number of citations and PSC rates per publisher, along with their standard deviations, were calculated. Funnel plots were generated to identify outlier publishers with unusually high PSC rates; the control limits were set at 95% and 99.8%, with the 99.8% limits used to distinguish statistically significant deviations from the expected variability7). Furthermore, to compare the distribution of JSC with that of PSC, we generated an additional funnel plot in which JSC rates were plotted for each journal. Further details are provided in the Supplementary Methods.
The study protocol, including the analysis plan, was preregistered and made publicly available on the Open Science Framework on February 26, 2025 (https://osf.io/j3ut6). All procedures, including data preprocessing and analysis, were performed using R version 4.3.2 (The R Foundation for Statistical Computing, Vienna, Austria).
Of the 333 target journals identified, six were excluded due to an IF of <0.1, while four were excluded as their IF was listed as “N/A.” Thus, 323 journals with 178 publishers were included in the analysis (Fig. 1). Of the 178 publishers analyzed, the mean number of citations per publisher was 3,012.9 (standard deviation: 11,847.7), whereas the mean PSC rate was 0.13 (standard deviation: 0.17). Fifteen publishers owned more than 100 journals across all categories, with PSC rates among these 15 publishers ranging from 0.01–0.27 (Supplementary Table 1). Notably, journals from the top 20 highest-PSC publishers mostly had IFs below 1.2, which was the median of the IFs in this category (Supplementary Table 1).
IF, impact factor.
As illustrated in Fig. 2, most publishers exhibited PSC rates below 0.20, whereas only a small number demonstrated rates exceeding 0.50, indicating a skewed distribution. Fifteen publishers (8.43%) exceeded the 99.8% upper control limit and were classified as upper outliers. Of these, one publisher had more than 10,000 citations, whereas the remaining 14 had fewer.
The plot shows the variability in publisher-level self-citation rates (vertical axis) relative to total citation volume (horizontal axis, log-transformed) by each publisher.
Since the plots were clustered at lower citation counts, the horizontal axis was log-transformed to improve visualization.
The dashed navy lines represent the 95% control limits based on an overdispersed binomial model, while the solid pink lines represent the 99.8% control limits.
Journals from the 15 PSC outlier publishers (n = 23) are highlighted in yellow in Supplementary Fig. 1. Among these, 14 journals exceeded the 99.8% upper control limit in the JSC funnel plot, and 9 journals (from 3 of the 15 outlier publishers) fell within the control limits.
We performed a descriptive study of publisher-level self-citation, using data from 323 journals across 178 publishers in the “Medicine, General & Internal” category from the 2023 Journal Citation Reports. We identified unusually high PSC rates among publishers with both low and high citation counts. Moreover, we detected 9 journals from 3 outlier publishers that could not be identified by JSC alone but were revealed through PSC analysis. To the best of our knowledge, this is the first study to assess the PSC within a single medical research category.
In this study, the range of PSC rates among the publishers of more than 100 journals was 0.01–0.27. A previous study reported that PSC rates among publishers owning more than 100 journals across all categories ranged approximately from 0.1 to 0.45). Hence, PSC rates in the “Medicine, General & Internal” category demonstrated a similar or slightly lower trend compared to those observed across all categories.
Among publishers with low citation counts, many oversee only one or a few journals. In such cases, high levels of self-citation at the journal level were observed to inflate the overall PSC rate at the publisher level, leading to outlier classifications. This indicated that certain journals exhibited excessive self-citation, either deliberately or due to limited citation sources.
Conversely, among publishers with a high volume of citations, a high PSC rate was observed even when their individual journals were not detected as outliers in the JSC funnel plot. This finding suggests that cross-citation between journals within the same publisher group may be systematically occurring, rather than just self-citation within individual journals. Such practices may artificially enhance bibliometric indicators, including the impact factor, undermining the fairness and integrity of academic publishing. These forms of citation manipulation not only favor certain journals or publishers but also risk distorting researcher behavior, including journal selection and academic evaluation5),6). Therefore, ongoing monitoring and increased transparency of citation practices are warranted.
This study has several limitations. First, journals not indexed in Journal Citation Reports were excluded, potentially leading to incomplete citation data. Second, PSC, as defined in this study, includes JSC that cannot be distinguished. This is particularly relevant for publishers with only a single journal, where the PSC and JSC are identical. However, by additionally generating a funnel plot of JSC and comparing it with PSC, we were able to partly examine this issue within our analysis. Further investigation is required to determine whether a high PSC reflects a coercive citation strategy.
This study identified considerable variation in PSC rates among publishers within the field of general and internal medicine, with some publishers exhibiting unusually high values. Our analysis suggests the importance of monitoring citation practices at both journal and publisher levels, though it is essential to recognize that self-citation is not inherently inappropriate and that citation metrics should be considered alongside multiple indicators to comprehensively assess academic contribution.
The authors declare no conflicts of interest in relation to the work presented in the manuscript.
The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
We thank Clarivate for their support in providing the Unification Publisher list and updating the inactive hyperlink in the Journal Citation Reports release notes. We would like to thank Editage (www.editage.jp) for English language editing.
Supplemental materials for this article are available online.
Data are available in Supplementary Table 1.
YT and SF conceptualized the research. TT and KS retrieved the datasets, performed the data analysis, and wrote the original manuscript; TT also wrote the R codes. YT, TA, and SF critically revised the manuscript for important intellectual content. All authors contributed to the study’s methodology and approved the final version.
