Trends in Academic "Ghost Publications" in Plastic Surgery Residency Applications: A 3-Year Study.

Plastic surgery is an attractive specialty to medical students. Residency training programs have the luxury of selecting their trainees from the "cream of the crop" from United States medical schools. Because of the steep competition for PGY-1 integrated program positions, the temptation exists for applicants to falsify parts of their applications, particularly those parts that are difficult to verify.
METHODS: A retrospective analysis of the Integrated Plastic Surgery applications from the years (2010-2013) was done. Two reviewers manually and independently handsearched each of the articles in the databases (Medline, Scopus, Clinical trials, Google scholar) additionally, a specialized medical librarian corroborated. A ghost article was defined as the inability to find the listed applicant in the authorship list of the claimed article/abstract/chapter or the inability to find the submitted article. Misrepresentation was defined as a change in authorship order. Data were summarized and analyzed, generalized estimating equations model was used. SAS software, v9.4.
RESULTS: All 392 applicants were included, 159 (2010-2011), 120 (2011-2012), and 119 (2012-2013). The number of manually reviewed records was 2,124. "Ghost" authorship was found in 234 articles out of 2,124 (11.02%). The overall rate of "Ghost" authorship in applicants to our program was found to be 34.4%, 135 applicants and misrepresentation in 5 cases (1.28%).
CONCLUSIONS: Ghost publications are present in Plastic Surgery applications, its trend is similar through the years, "protective" factors are: first authorship and published peer reviewed abstract/article.

INTRODUCTION

Plastic surgery is an attractive specialty to medical students. Residency training programs have the luxury of selecting their trainees from the “cream of the crop” from United States medical schools. Using data from the National Resident Matching Program, successful applicants to Plastic surgery tend to have above average USMLE steps 1 and 2 scores.[1,2] Many applicants are members of AOA and have significant research experience. Extracurricular activities and service to the community are often impressive.

The information gathered through the Electronic Residency Application System (ERAS) is provided by each applicant. Each individual must certify that, “My ERAS application is complete and accurate to the best of my knowledge. I understand that any false or missing information may disqualify me from consideration for a position, or if employed, may constitute cause for termination from the program, and may result in expulsion.” Because of the steep competition for PGY-1 integrated program positions, we hypothesized that temptation exists for applicants to falsify parts of their applications, particularly those parts that are difficult to verify. To assess the frequency of this phenomenon, our study focuses on not only the frequency in a single group of applications, but also its trend over time and factors associated with it by evaluating the application cycles that each began in 2010, 2011, and 2012.

METHODS

After approval of the Baylor Scott and White Institutional Review Board (IRB approval number: 150035), a retrospective analysis of the Texas A&M, Baylor Scott and White Integrated Plastic Surgery applications from the 2010 to 2011, 2011–2012, and 2012–2013 application cycles. Data were collected and stored in an encrypted-access, electronic spreadsheet Microsoft Excel 2010 (Redmond, Washington) which included AAMC number, year of application, number, number and type of research articles and first or second authorship. Two reviewers (N.A.R.-U., N.D.W.) manually and independently hand-searched for each of the articles in the following databases: Medline, Scopus, Clinical trials, Google Scholar. We utilized searches of all the authors, the applicant and various key words from the reported publication title to attempt to identify any title changes and small changes in author order or publication venue. If the article could not be identified by the primary reviewers a secondary review by a specialized Medical Librarian was performed. No limitation in terms of language was used. Each reviewer noted whether a publication could be found and the author order. For a reported publication to be marked as “Ghost” or “Misrepresentation” both primary reviewers had to agree.

A ghost article was defined either as one in which the listed author was not named in the authorship list of the claimed articles/abstract/chapter/other publication or if the submitted piece could not be found at the time of the review with an interval of at least 5 years since the application was submitted.

Misrepresentation was defined as claiming to be first author when actually named later in the authorship list.

The statistical analyses will be reported using summary tables. Statistical tests were 2-sided with an alpha 0.05 significance level and 2-sided 95% confidence interval. Categorical variables were summarized by counts and by percentages of subjects in corresponding categories. To take into account repeated measures from the same applicant (eg, >1 ghost article), data were analyzed with the use of generalized estimating equations, a multivariate analogue of linear regression for multiple publications. Each applicant that reapplied in subsequent years was removed from subsequent year’s analysis which prevented them from being double counted. This contributed to the gradual decline in application number seen in the final results. Analyses were carried out with the use of SAS software, version 9.4 (The SAS Institute, Cary, N.C.).

RESULTS

All 392 applications from the years 2010, 2011, and 2013 were included, a total of 159 applicants from 2010 to 2011, 120 from 2011 to 2012, and 119 from 2012 to 2013. The number of manually reviewed reported publications was 2,124. The greatest number were distributed in peer reviewed journal articles/abstracts 1,187 (55.89%), followed by peer reviewed online publication 805 (37.90%), peer reviewed journal articles/abstracts other than published 70 (3.30%), and lastly, peer reviewed book chapters 62 (2.92%) (Table 1).

Table 1.

Distribution of Publication Articles by Type

Publication Type
	Totals	Percent
Peer reviewed journal articles/abstracts	1,187	55.9
Peer reviewed book chapter	62	2.9
Peer reviewed journal articles/abstracts (other than published)	70	3.3
Peer reviewed online publication	805	37.9

Most of the 2,124 articles were peer-reviewed journal articles/abstracts or online publication >90% (combined), a small minority were book chapters 2.9% and articles other than published 3.3%.

“Ghost” status was found for 234 reported publications out of 2,124, which constitutes 11.02% of the articles pool. “Misrepresentation” was found in 21 articles (0.99%).

The overall rate of “ghost” articles at our institution during those 3 years was found to be (34.44%) or in 135 applications out of 392. There were only misrepresentations in 5 applications (1.28%). The trend of “ghost” authorship through the applications years of 2010, 2011, and 2012 did not change over time (P =0.27) (Table 2).

Table 2.

Trend of “Ghost” Publication by Years 2010–2013

Trend of “Ghost” Publication by Years 2010–2013*
Authorship	Year
	2010–2011	2011–2012	2012–2013	Total
Applicants without “Ghost” (%)	98 (25)	85 (21.7)	74 (18.9)	257 (65.6)
Applicants with “Ghost” (%)	61 (15.6)	35 (8.9)	39 (9.9)	135 (34.4)
Total applicants (%) “Ghost”	159 (38.4)	120 (29.2)	119 (32.8)	392

*The trend of “Ghost publications” did not change significantly through the years (p 0.27).

Different publication types were strongly correlated to “ghost” status, with peer reviewed online publications the highest at 43.7%, peer reviewed journal articles/abstracts (other than published) 32.1%, peer reviewed book chapters 22.2%, and peer reviewed journal articles/abstracts 14% (P =0.0001) (Table 3).

Table 3.

Distribution of “Ghost Authorship” Publications by Type

Distribution of Applicants Authorship by Type*
Authorship	Publication Type
	Peer Reviewed Journal* Articles/Abstracts	Peer Reviewed Book Chapter	Peer Reviewed Journal Articles/Abstracts (Other than Published)	Peer ReviewedOnline Publication	Total
Verified author (%)	79 (20.1)	21 (5.4)	19 (4.9)	138 (35.2)	257 (65.6)
Ghost author (%)	13 (3.3)	6 (1.5)	9 (2.3)	107 (27.3)	135 (34.4)
Total (%)	92 (23.5)	27 (6.9)	28 (7.1)	245 (62.5)	392

The distribution of “Ghost” article is strongly associated by different publication type (p < 0.0001*).

Most applications that contained “Ghost” publications had only 1 [61.19%] or 2 articles [22.39%] where this occurred. On the other hand, we found a subset of 5 [1.28%] applicants who had 6 or more “Ghost” publications.

A multivariate linear regression was used to evaluate for each author type, type of publication and year of publication. A total of 2,124 publications were included in the final analysis. This analysis yielded a significant association for first author status and publication type, with P-values of 0.0005 and <0.0001, respectively (Table 4).

Table 4.

“Ghost” Publication Characteristics from Applicant Pool Using a GEE Model

Publication Type	Publication Type	Estimate	Standard Error	z Value	Pr > z	Exponentiated
Peer reviewed book chapter*	Peer reviewed journal articles/abstracts	1.3251	0.5257	2.52	0.011	3.76
Peer reviewed journal articles/abstracts (other than published)*	Peer reviewed journal articles/abstracts	2.0972	0.4421	4.74	<0.0001	8.14
First author status†	Yes/no	0.820	0.2355	3.48	<0.0005	2.27

*The baseline reference for article type was peer reviewed article abstract.

†When using GEE the first author status was using first author as baseline reference.

First author status and studies published in peer reviewed journals were “protective factors” against Ghost authorship.

GEE, generalized estimating equations.

DISCUSSION

Academic dishonesty in medicine is not isolated to the field of plastic surgery. This topic was originally studied in the field of gastroenterology applicants[3] and was subsequently expanded to include internal medicine, emergency medicine, radiology, orthopedic surgery, neurologic surgery, and general surgery.[4-9]

In this article, we present a 3-year cohort of applicants from 2010 to 2013 to a single academic program in Texas. Strengths of the study are its dependable methods and 2 independent screeners, well versed in the medical literature with no language, time or type of study restrictions. In addition, an interval of 5 years after the application period was chosen to allow for resubmission of a article/abstract/article in case it required revisions after it had been accepted. Considering the average period of 4–8 months from submission to acceptance in journals with an impact factor of 1–3, this would be plenty of time to account for subsequent changes and eventual publication.[10-12]

When considering the 2,124 screened publications, only 234 [11.02%] met the “ghost” authorship definition. However, when we look at the percentage of applicants through the 3-year period with at least one ghost publication, a surprising 34.4% value is obtained. This is similar to previously reported values of 18%–33% among surgical specialties.[8,13] Nevertheless, this is higher than earlier studies on plastic surgery applications which cited 2%–7%.[14,15] These studies were limited to a 1 year cohort, and <240 applicants. When more applicants were added and the length of the cohort was lengthened to 3 years, this percentage increases. Nevertheless, our percentage is still higher than the 28% found by Chung et al[16] among plastic surgery residents.

Since our cohort, there has been an increase in the number of claimed publications. There is now an average of 14.2 abstracts, presentations and publications per applicant in 2018.[1] With approximately 140 applicants per year, to review a projected 1,960 claimed publications per each year would be impractical. The aim of our study was to characterize the incidence and identify “risk factors” for “ghost” publications to help identify high-risk claims. We found 2 important factors. Not being listed as first author had 2.3 times the chance of being a ghost publication when compared with first authorship. Secondly, any other nontraditional research document (other than peer reviewed articles/abstracts, eg, book chapter or newspaper article) was 8 times more likely to be a ghost citation.

The risk factor of authorship makes intuitive sense. Authorship usually requires a level of involvement that is unlikely to result in not being included in the final publication even if there are bumps in the road and the author list change. One of our other proposed causes of such a high rate of ghost publication is that the role of the medical student during a research project may not be clearly established. During the development of a research plan with their fellow researchers and discussion of working on a paper, there may be a generalized misunderstanding of what constitutes authorship. For instance, a student might gather data for a retrospective study, or run one experiment, eg “western blot,” without participating in the article design, draft and final approval, assuming that 1 task alone constitutes authorship.

Nontraditional research documents such as book chapters and newspapers articles are the second significant risk factor. These are at high risk for not being indexed in google or Medline and have even less clear guidelines for authorship. Book chapters in particular may require a long time between submission and publication.

There are some limitations of our study. Ideally, we would have included all the applicants across the United States to obtain a more accurate estimation of the problem Consequences of academic dishonesty to someone’s career could be severe and could result in termination of a resident contract, a generalized bad reputation for the individual and the institution and loss of credibility.[17]

Plastic surgery as a specialty has been shown to attract strong applicants as evidenced by AOA membership, USMLE step 1 and 2 scores and research articles, but if we want a program to thrive we need men and women whose integrity is not compromised.[2] At least a partial, possible explanation to the problem could be found in the article by Momenis et al[18] about opinions on authorship, where 1/3 of plastic surgery residents were not aware of the rules which constitute true authorship, according to the International Committee of Medical Journal Editors guidelines. This should be discussed when helping a prospective applicant to build their application. Applicants should also be encouraged to make clear their level of involvement in studies that they claim on their application and place those where they were not authors instead in the section for research experiences.

As a result of our study, we have started to screen a random sample of research publications from the applicants to which we offer an interview. If there is any discrepancy between what is reported and what is discovered, we ask the applicant about the project and what the circumstances are that could explain the discrepancy.

CONCLUSIONS

Ghost publications are present in Plastic Surgery applications and the trend remains similar through the years. Characteristics associated with ghost publications are not being first author and not publishing a standard peer-reviewed article/abstract.