Trust but verify: An analysis of redundant publications from two major psychiatry journals in India.

Background: No analysis of redundant or duplicate publications, deemed unethical and unscientific, has been undertaken in psychiatric literature. Aim: To analyze the proportion and patterns of redundant publications associated with index articles published in two major Indian psychiatry journals.
Methods: Index articles were original papers published in the Indian Journal of Psychiatry and the Indian Journal of Psychological Medicine between 2015 and 2017. Using a systematic search strategy that combined author names and article keywords, we combed the literature to identify and characterize redundant publications related to these index articles. Redundant publications were classified into one of the following categories using a priori definitions: dual, suspected dual, salami slicing, meat extender, and extended sample publication.
Results: From 324 index articles screened, a total of 27 articles (8.4%) were identified to have 32 associated redundant publications of the following types: dual (n = 3), suspected dual (n = 2), salami slicing (n = 22), meat extender (n = 3), and extended sample publication (n = 2). A majority of the redundant articles (n = 23, 71.9%) failed to clearly cross-reference the prior publication(s). We also identified nine non-redundant but related publications with no proper cross-referencing in five of them.
Conclusion: Redundant publications are a common practice in the psychiatry journals screened. Salami slicing is the most common form of redundancy, with no proper cross-referencing in most cases. Concerted efforts are needed to detect and deal with this concerning practice that undermines both science and ethics. Copyright:

INTRODUCTION

As far back as 1969, in an influential editorial in the New England Journal of Medicine (NEJM), the then editor Joseph Ingelfinger made a famous proclamation now known as the Ingelfinger rule: A manuscript cannot be published in the NEJM if its substance has been submitted or published elsewhere.[1] Since then, many other journals have followed this policy though there were some dissenting voices too, who argued for modifying the Ingelfinger rule considering the evolving publications environment.[2]

Academic publications are now a necessity for career advancement, obtaining tenured faculty positions, and securing research grants.[3] Indeed, the annual output of peer-reviewed research publications is a significant measure of scholarly productivity.[4] Of the many undesirable offshoots of this development, the practice of redundant publications is an unethical, yet widespread, issue. Whereas plagiarism and data fabrication are regarded as major crimes in academia, redundancy in publications is considered a misdemeanor and a less serious issue within the scientific community.[5] Nevertheless, the phenomenon has been widely condemned as it can lead to possible misinterpretation of the research findings.[6789] However, its true extent and impact across specialties remains unclear.

Redundant or duplicate publishing refers to a practice wherein two or more articles with substantially overlapping information, that is, more than two-third overlap in content, is published by the same set of author(s), typically without full cross-references to the prior publication(s).[10] However, there are gray areas here and what constitutes as substantial overlap is somewhat arbitrary. Nonetheless, the practice of redundant publications is considered unethical because it “junks” the scientific literature and wastes time, effort, and resources of editors, peer reviewers, and readers. Furthermore, if there are undetected redundant publications from the same data set, investigators conducting a meta-analysis may unintentionally include all of these papers. This may result in data from the same patient being entered more than once in the meta-analysis which compromises the validity of conclusions drawn.[11] Thus, such infractions compromise both science and ethics.

Several forms of redundant publications exist, such as duplicate publications,[12] salami publications,[13] and meat extender publications,[14] to name a few. A series of publications, mostly from Europe and America, have shown that the incidence of redundant publications ranges between 2%–20%, across specialties, such as pediatrics,[15] plastic surgery,[16] general surgery,[17] and otolaryngology.[18] However, little is known about the nature and extent of this practice in the field of psychiatry. Hence, the objectives of the present study were firstly, to determine the incidence of redundant publications in psychiatric literature and, secondly, to characterize the nature of these redundant publications.

METHODS

Study setting and data collection

We selected two major Indian psychiatry journals, the Indian Journal of Psychiatry (IJP) and the Indian Journal of Psychological Medicine (IJPM); both of these are Scopus-indexed and included in PubMed Central and provide immediate open access to their content. All original articles published in these two journals between January 2015 and December 2017 were retrieved from their respective websites and identified as index articles (reference articles). We excluded editorials, reviews, letters to the editor, case report/series, viewpoints/commentaries, abstracts of conference proceedings, and book reviews from the analysis because they neither involve substantive data sets nor research methods, which are used to classify redundant publications. Next, “potential redundant” articles were identified initially through a PubMed (for MEDLINE) search by combining the first or second or last author’s surname and a keyword from the title of the index article based on consensus between authors; this method has been used previously for similar research.[1617] The search was not restricted by time limits. For the searches which yielded more than 100 titles and abstracts, an additional keyword from the index article was added to narrow the search results. If the PubMed search did not yield any result, a Google Scholar search was carried out using the same search strategy for PubMed. We defined an article as a “potential redundant” if it matched the keywords and author surnames of the index article.

Finally, the full texts of these potentially redundant articles were retrieved and their study objectives, methodology, results (tables), and conclusions were scrutinized and compared with the index articles to identify redundant publications. To define an article as redundant, we used the criteria by Kumar et al.[19] which reads as follows: “the number or sample sizes are similar; the methodology is identical or nearly so; the results are similar; at least 1 author is common to both reports; and little or no new information is made available.”

Four categories of redundant publications were defined a priori: dual, suspected dual, salami-slicing, and meat extender publications; the definitions used for each of them were based on criteria laid down by prior researchers[1720] as well as recommendations given by the Committee on Publication Ethics[21] [Table 1]. Every redundant publication identified were slotted into one of these four categories. A fifth “miscellaneous” category was also created to accommodate redundant articles which did not fit any of the a priori categories. This process of scrutiny and classification as a redundant publication was done primarily by one author with prior experience in reviewing and publishing (NV); any doubts during this stage were discussed with senior authors (VM and SKP) before arriving at a final decision about the redundant nature of the publication.

Table 1

Definitions used for different types of redundant publication [as per Schein and Paladugu (2001), Rosenthal et al (2003), and Committee on Publication Ethics (2005)]

Type of redundant publication	Definition used
Dual publication	An article with identical or nearly identical data that reported the same results and conclusions as to the index article
Suspected dual publication	Minor additions (<50%) to the data set that yielded
	similar results and the same conclusions
Salami slicing	If a portion of the manuscript, including the methodology, was exactly similar to the index article and they addressed related, or closely related, research questions.
	If two papers examined different, unrelated hypothesis, they were not classified as salami publication even though they may have been drawn from the same dataset.
Meat extender	Minor additions (<50%) to the data set that yielded different results and conclusions

Data extraction

The following data were abstracted for every index original article: journal name, country of origin (India or abroad), type of study (observational/interventional), type of interventional study (Randomized controlled trial-RCT/non-randomized), level of evidence (defined below), type of first author (psychiatrist/non-psychiatrist), discipline of origin of the paper (psychiatry/psychology/social work/other medical or surgical specialties/social sciences/others), total number of authors, associated with redundant publication (yes/no), number of redundant publications, type of redundant publication (dual/suspected dual/salami/meat extender publication), if salami whether adequate cross-referencing was made to prior related articles (yes/no), associated with non-redundant but related publication (yes/no), if non-redundant whether adequate cross-referencing was made to prior related articles (yes/no), whether index article is redundant (yes/no; defined by presence of a prior related publication), year of publication of redundant article(s), time interval between publication of the index and redundant article (in months), name and CiteScore of the journal (from Scopus) in which the redundant article(s) was/were published and link to redundant publication(s). The levels of evidence were graded as per Sackett[22]: Level I was large randomized trials with clear-cut results; level II was small randomized trials with uncertain results; level III was non-randomized, contemporaneous controls; level IV was non-randomized, historical controls; and level V was no controls, case series only.

We defined a non-redundant, but related, publication as those that examined a different, unrelated hypothesis even though they may be drawn from the same study or dataset. If the hypothesis was not explicitly mentioned, we examined the study objectives to assess overlap. In all such instances, we additionally noted whether subsequent papers clearly and adequately cited previous papers from the same dataset. This aspect was also recorded for all types of redundant publications. A cross-reference was defined as unclear and inadequate if it was mentioned along with other studies in introduction/discussion or briefly cited in the method section (e.g.: “as previously described”).[2023]

Abstracted data were entered in a standardized data collection proforma by a single investigator (NV). Data were cross-checked by a senior investigator (VM) who reviewed the accuracy of variables entered. Discrepancies and disagreements (about classification as a redundant publication and subsequent subtyping of such publications) were resolved through mutual discussion with all investigators till consensus.

Statistical analysis

We used descriptive statistics (frequencies and percentages) to depict data. Characteristics of index articles and those that led to redundant publications were examined separately.

Ethical aspects

The Institutional Review Board approved the conduct of the present study (vide letter no JIP/IEC/2021/167).

RESULTS

Characteristics of index articles

A total of 324 articles were identified as original index articles, published between 2015 and 2017, in the two journals. Of these, 144 articles (44.4%) were published in the IJP and the remaining (n = 180, 55.6%) were published in the IJPM. A majority of the index articles were from India (n = 276, 85.2%), belonged to the observational study type (n = 304, 93.8%), and consisted of level V evidence studies (n = 277, 87.5%).

Most of the index articles that were associated with redundant publications (n=27) originated from India (n = 22, 81.5%) and were observational studies (n = 24, 88.9%). More than two-thirds were cases series or studies without a control group (n = 19, 70.4%). The discipline of origin was commonly psychiatry (n = 20, 74.1%) followed by psychology (n = 4, 14.8%). Of the 27 index articles, 23 indexes (85.2%) had only one redundant publication, 3 indexes (14.7%) were associated with two each, and 1 index (2.9%) with three redundant publications. Sixteen indexes were published in the IJP while 11 were from the IJPM. Table 2 depicts characteristics of index articles that were associated with or not associated with redundant publications.

Table 2

Characteristics of index articles

Study characteristics	Total index articles (n=324, %)	Associated with redundant publications (n=27, %)	Not associated with redundant publications (n=297, %)
Articles published in journal
IJPM	180 (55.6)	11 (40.7)	169 (56.9)
IJP	144 (44.4)	16 (59.3)	128 (43.1)
Country of origin
India	276 (85.2)	22 (81.5)	254 (85.5)
Abroad	48 (14.8)	5 (18.5)	43 (14.5)
Type of study
Observational	304 (93.8)	24 (88.9)	280 (94.3)
Interventional	20 (6.2)	3 (18.5)	17 (5.7)
Type of interventional study (n=20)
RCT	17 (85)	3 (11.1)	14 (82.4)
Non-RCT	3 (15)	-	3 (17.6)
Type of evidence
Level I	13 (4)	3 (11.1)	10 (3.4)
Level II	5 (1.5)	1 (3.7)	4 (1.3)
Level III	18 (18)	4 (14.8)	14 (4.7)
Level IV	4 (1.2)	-	4 (1.3)
Level V	277 (87.5)	19 (70.4)	258 (86.9)
Others (qualitative studies)	7 (2.2)	-	7 (2.4)
Number of authors*	4 (2)	4 (2)	4 (3)
Type of author
Psychiatrist	221 (68.2)	21 (77.8)	200 (67.3)
Non-psychiatrist	103 (31.8)	6 (22.2)	97 (32.7)
Discipline of origin
Psychiatry	218 (67.3)	20 (74.1)	198 (66.7)
Psychology	27 (8.3)	4 (14.8)	23 (7.7)
Social work	4 (1.2)	-	4 (1.3)
Others	75 (23.2)	3 (11.1)	72 (24.3)

Values expressed as frequency (percentage) or *median (interquartile range); IJPM, Indian Journal of Psychological Medicine; IJP, Indian Journal of Psychiatry; RCT, Randomized controlled trial

Frequency and patterns of redundant publications

Initial search for redundant articles using the methodology described above yielded 79 articles (24.4% of index articles) which covered similar topics as the index articles: These were considered as potential redundant articles. The full texts of all of these articles were retrieved and compared against the index article, following which 32 articles (corresponded to 27 out of 324 index articles, i.e., 8.3%) were found to have some type of redundancy [Figure 1]. Among the 32 redundant articles, 6 were published in IJP and 5 in IJPM itself. Thus, 34.4% of redundant articles identified were published in the two index journals itself.

Figure 1

Flow diagram for search and retrieval of redundant publications

We retrieved nine publications drawn from the same sample and conducted during the same study period, but because they addressed non-overlapping objectives and had different results and conclusions, these were not classified as redundant publications. Of these, four were published subsequent to the publication of the index article, three in the same year as the index article, and two appeared in literature prior to the index article. Cross-referencing to the prior paper(s) were seen in four out of these nine publications.

Of the 32 redundant publications, a majority were salami publications (n = 22, 68.8%), three (9.4%) were classified as dual and meat extender publications each, and two each (6.2%) as suspected dual and a miscellaneous category. The latter did not fit any of the a priori definitions that we used to classify different types of redundant publication. The two miscellaneous publications added a sample size more than 50% of the original index article but arrived at similar results and conclusions as the index article. We named this redundant category as “extended sample publication”.

The distribution of types of redundant publications in the two Indian journals were as follows: IJPM (N = 6, all were salami publications); IJP (N = 5, three were salami publications, one each was dual and suspected dual publication).

Characteristics of redundant publications

Of the redundant publications, 13 (40.6%) were published in the same year as the index articles. Of these, eight appeared before the index article and five were published after the index articles. Sixteen articles (50%) were published in the years before the index article within a time frame of nine months to five years and three (9.4%) appeared in the succeeding years within a time frame of ten months to four years.

A majority of the index articles (n = 17, 53.1%) did not cite the prior paper(s) or referred to it vaguely (n = 1, 3.1%); only six index articles (18.8%) clearly cross-referenced prior publications from the same dataset. Of the eight redundant publications that were published after the index articles, a clear cross-referencing was seen only in three of them, while one provided a vague cross-reference, and four did not cite the prior paper(s). Overall, 23 of the 32 redundant publications did not make a clear cross-referencing to prior papers from the same dataset. Sixteen redundant papers (50%) were published in journals with a CiteScore higher than that of the index article [Table 3].

Table 3

Characteristics of redundant publications (n=32)

Variable	Dual (n=3, %)	Suspected dual (n=2, %)	Salami slicing (n=22, %)	Meat extender (n=3)	Extended sample publication (n=2)	Total (n=32)
Cross-reference
Clear	1 (33.3)	0	7 (31.8)	0	1 (50)	9 (28.1)
Vague	0	0	1 (4.5)	0	1 (50)	2 (6.3)
Not referenced	2 (66.7)	2 (100)	14 (63.7)	3 (100)	0	21 (65.6)
Published in
year prior to index article	1 (33.3)	1 (50)	10 (45.5)	3 (100)	1 (50)	16 (50)
same year as index article	2 (66.7)	1 (50)	10 (45.5)	0	0	13 (40.6)
year after index article	0	0	2 (9)	0	1 (50)	3 (9.4)
Published in Journals with
CiteScore
>than index article	1 (33.3)	1 (50)	11 (50)	2 (66.7)	1 (50)	16 (50.0)
<than index article	0	0	2 (9.1)	0	0	2 (6.2)
cannot determine *	1 (33.3)	1 (50)	4 (18.2)	1 (33.3)	0	7 (21.9)
same as index article	(33.3)	0	5 (22.7)	0	1 (50)	7 (21.9)

Values expressed as frequency (percentage); *redundant article published in a journal not listed in Scopus

DISCUSSION

The present paper sought to analyze the frequency and patterns of redundant publications related to index articles published in two major Indian psychiatry journals. We found that 8.3% of the index articles were associated with some form of redundant publications in literature. In a majority of them, the work originated from India and the redundant papers too were published in the same index journals. Salami publication was the most common form of redundancy noted in more than two-thirds of the redundant publications. Further, four out of five redundant publications did not fully cross-reference the prior publication(s) emerging from the data set.

Despite universal acknowledgment about the unethical nature of redundant publications, it remains a persistent and poorly understood transdisciplinary phenomenon. In one of the earliest evaluations of redundant publications, Schein and Paladugu, in 2001, screened more than 600 articles in three prominent general surgery journals and found a redundancy rate of 14%.[17] Subsequently, authors from ophthalmology,[24] plastic surgery,[16] orthopedics,[25] radiology,[26] otorhinolaryngology,[1820] and pediatrics[15] have all noted instances of redundant publication. Of these, redundancy rates were low (less than 5%) in the fields of orthopedics,[25] ophthalmology,[24] radiology,[26] and pediatrics[15] while higher rates, comparable to what we found, were noted in journals belonging to plastic surgery[16] and otorhinolaryngology.[1820] One analysis of otorhinolaryngology publications found a high redundancy rate of 20.9%.[18]

The most common form of redundancy observed in the present, as well as prior analyses,[151820] was salami publications. This was notwithstanding the use of a more restrictive definition of salami publication in our analysis, where papers that addressed unrelated hypotheses were not classified as salami though they were drawn from the same study dataset, unlike earlier research.[151718] A salami publication can be roughly understood as a publication of multiple articles drawn from a single study or single dataset.[6] Articles of this type typically split the larger study data into smaller units that are just sufficient to yield reasonable results and conclusions; in other words its “minimum publishable units”.[27]

A salami publication may not always be characterized in terms of black and white, but has several shades of gray and is often difficult to deduce based purely on information available in published papers. Sometimes, there are legitimate salami publications; for instance, in case of large, multicentric studies where several independent hypotheses are examined, it is nearly impossible to present all of the findings in a single paper. In such cases, the authors may legitimately decide to analyze and present different hypotheses in separate papers; nine such instances were noted in the present analysis. This is not considered unethical, per se. As an example, the large multiphase Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) trial yielded several legitimate papers that tested many independent hypotheses.[282930]

However, in all such cases, authors are mandated to provide the editor with the necessary information about published results from the dataset so as to facilitate an accurate judgment about the legitimacy of the submitted work. This must be achieved by mentioning about prior publications in the covering letter as well as by a clear cross-referencing of prior papers in the introduction or methods section of the subsequent papers so that editors and reviewers can independently evaluate the justification for publishing the manuscript. In this regard, our findings that a substantial proportion of the redundant publications were published in either of the two index journals itself, without adequate cross-referencing of “slices” in most of them, calls for introspection about the inability of editors to weed out such papers. While the decision on whether the publication of sliced papers should be supported is best left to the discretion of editors and reviewers, if permitted, slicing should be overt, not covert, with full cross-referencing to prior papers being obligatory.

Sometimes, because of delays in review and publication, additional manuscripts are prepared and submitted before the acceptance or publication of earlier ones. In such cases, providing the registration number of trials in case of randomized controlled trials, giving the clearest possible cross-referencing to papers that are in press, and updating the status of citations in papers under preparation based on change in status of preceding ones must be considered by authors while there is an opportunity during the publishing process. The bottom-line is that authors must make all efforts to leave editors, reviewers, and readers in no doubt about the relationship between different papers from the same dataset.[31]

Clear and full cross-referencing also clarifies the author’s honest intentions; this is very important because, as has been pointed, in such an area with several shades of gray, more than the duplicity, it is the intention to deceive that is more incriminating.[32] In this context, our observation that less than one-third of redundant publications made a clear cross-reference to the prior paper reflects poorly on the concerned authors; it not only appears deceitful but also makes it hard to justify salami slicing of datasets when no reference is made to the other related publications. This observation is not an isolated phenomenon: two other studies, both from otolaryngology,[1820] have encountered a similarly high instance of failure to cross-reference the preceding publication(s).

Given these concerning findings across the spectrum of medical specialties, there is a need for concerted action by stakeholders to raise awareness about the unethical practice of redundant publications and minimize infractions. For starters, at an institutional level, institutions should conduct regular awareness programs for faculty and doctoral researchers on this issue. Universities should appoint research integrity officers and make them accessible to all departments so that they are best placed to detect and handle instances of research misconduct. Every institution should prepare an institutional policy document that expounds on various types of research misconduct, including redundant publications, and this should be posted on all relevant websites.

Better cooperation between journals and institutions in detecting, probing, and handling cases of research misconduct is essential; editors should promptly inform institutions about suspected misconduct; institutions should adequately investigate all such instances and respond to journal editors’ request for more information; and finally, journals should have clear policies for dealing with redundant publications and be ready to issue retraction notices, according to guidelines laid down by professional bodies such as Committee on Publication Ethics (COPE).[33] A nuanced approach is warranted in identifying and dealing with infractions because of the several gray areas and element of subjectivity involved in determining what constitutes research misconduct. A discussion and consensus approach involving senior academics and other members of the editorial board would help in this regard. Sometimes, it may be necessary to discuss with other editors on the COPE platform to resolve these ethical concerns.

Our work had certain limitations. The analysis was conducted on index articles published in two Indian psychiatry journals. These journals were selected because they have the highest CiteScores among Indian journals and are an attractive destination for authors to place their articles. However, this means that their rejection rates may also be high, as shown in a recent analysis of submissions in one of them.[34] Thus, extrapolating our findings to other journals with lower rejection rates should be done with caution. Further, we only searched two databases, MEDLINE and Google Scholar, and it is possible that some more redundant articles may have been published in journals not indexed in these databases or in the gray literature. Our search was more exhaustive than similar articles published earlier which employed search time limits of two years on either side of publication of the index article.[1526] The search methodology using the first or second or last author names as well as selected keywords from the index articles to locate potentially redundant publications was more comprehensive than prior similar publications.[18]

To conclude, nearly one-tenth of index articles published in two major Indian psychiatry journals were associated with redundant publications, of which most were salami publications with inadequate cross-referencing to prior papers. These substantial figures warrant further examination of the issue in other journals from the discipline to better understand its scope, extent, and impact. It also calls for coordinated efforts from journals, institutions, and other stakeholders to increase surveillance and put in place adequate deterrents to curb this practice that undermines both science and ethics.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.