We used The Preferred Reporting Items for Systematic review and Meta-Analysis Protocols (PRISMA-P) 2015 statement [27, 28] and the Joanna Briggs Reviewers manual for systematic reviews of prevalence and incidence [29, 30] to develop and report this protocol. This manuscript is registered in the Open Science Framework (OSF) [31] and not in PROSPERO [32], because our research questions do not meet the inclusion criteria for the latter register.
Eligibility criteria
Table 2 presents eligibility criteria for domain, study designs, participants, survey instruments, outcomes, time point, setting, language, publication status, and publication dates [27, 28].
Table 2
Inclusion and exclusion criteria
Item | Inclusion criteria | Exclusion criteria |
Domain | • Health sciences as defined in Table 1. | |
Study designs | • Studies including at least one survey according to its definition in Table 1. | |
Participants | • Any author on the author list of a scientific publication, e.g., first, last, corresponding author etc., that was invited to participate in a survey on at least one of our authorship items. | |
Survey instruments | • Surveys based on questionnaires for self-completion. • Surveys administered by email, internet platforms, by post, or by hand. • We will only consider closed surveys, i.e., surveys open to a specific sample of participants selected by the investigators. • Surveys with or without incentives to complete it. | • Focus groups discussions and one-to one interviews. |
Outcomes | • One or more of the outcomes on authorship issues listed in our objectives. • Both self-and non-self-reported outcomes on authorship issues. | • Outcomes that were not reported as prevalence statistics or were not given in a format that such statistics could be calculated. |
Time point | • Any time point for measuring outcomes will be eligible, i.e., we will not set exclusion criteria whether an article on which the surveyee was questioned was published 1, 2, 3 etc. years previously. | |
Setting | • Any | |
Language | • Any | |
Publication status | • Peer-and non-peer-reviewed manuscripts. | |
Publication dates | • Articles published from bibliography inception onwards. | |
Information sources and search strategy
PRISMA-S was consulted to report the literature search [33], as well as a previous systematic review on the meaning, ethics and the practices of authorship published by Marušić et al. [25], which used “authorship” as the only word for their search strategy. We will perform a systematic search in the following electronic databases from inception onwards: PubMed, Lens.org, and Dimensions.ai, with no language or date filters, but applying health sciences filters for Lens.org and Dimensions.ai for full search strategies. Additionally, all included papers will be checked for additional refences mentioned in their introduction or discussion sections. The full search strategies are presented in additional file 2. These strategies were developed by one of the authors (MM) with the aim to capture all surveys on authorship, as honorary authorship might have been only one question in those survey, or a secondary outcome that was not mentioned in the abstract. Additionally, these searches captured all studies we identified in the pilot, expect 4 that did not have indexed abstracts and were short reports. These 4 however were referenced in studies we identified, so would have been captured by checking the introduction or discussions or papers, which, as mentioned above, we will do for all included studies.
Data management and selection process
Two authors (RMR and NDG) will be calibrated a priori through pilot tests. These investigators will screen titles and abstracts independently to select potentially relevant papers. Identified records will be imported in Rayyan [34] and duplicate records will be removed. Rayyan will be subsequently used for the initial screening of titles and abstracts to identify potentially relevant papers. In the initial search phase the selection of studies will be overinclusive [35]. Full texts of potentially eligible manuscripts will be retrieved and assessed for eligibility. We will implement Cochrane’s strategies to identify multiple reports from the same study [35]. We will further assess whether relevant studies were retracted, fraudulent, or whether errata or comments were given [35].
After these assessments we will make final decisions on study eligibility. Reference lists and citing articles of the selected eligible papers will also be crosschecked for additional relevant studies. A list of excluded studies with rationales for exclusion will be given. This list will focus on all studies that initially appear eligible, but after further inspection do not meet the eligibility criteria [35]. Our search strategy will not include a filter for manuscripts published in the health sciences. All non-health science manuscripts that assess authorship issues will also be presented in the list of excluded studies with the reason for exclusion. The study selection procedures will be reported in a PRISMA flow chart [36, 37].
Data collection process and data items
For the development of our data extraction forms we consulted the reporting checklists of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement for reporting cross-sectional studies [38], the Checklist for Reporting Results of Internet E-Surveys (CHERRIES) for reporting the survey [39, 40], and the Checklist for polls by Bethlehem [41]. To ensure consistency in the data collection process we will conduct calibration exercises for both data extractors a priori. These investigators (RMR and NDG) will independently extract the pertinent data from each eligible study. The pilot-tested data collection forms are listed in Additional file 3A. A description of each data item is given in these forms. Disagreement between investigators on study inclusion and issues regarding data extraction will be resolved through discussions. Persistent disagreements will be resolved through arbitration by a third author (GTR) or through contacting the authors of the pertinent publications [28]. We will document when and why this was deemed necessary.
Outcomes and prioritization
The prevalence of researchers perceiving other co-author(s) as honorary author(s) on a publication and the prevalence of researchers having been approached by others to include honorary author(s) on a publication will be the primary outcomes. The prevalence of researchers admitting being an honorary author on a publication, the prevalence of researchers admitting adding an honorary author(s) on a publication, and the prevalence of researchers admitting having approached others to include honorary author(s) on a publication will be the secondary outcomes. The definitions of these outcomes and the pertinent numerators and denominators are given in Table 3. The various response rates measured are also reported in these tables.
Table 3
Definition of response rates and primary and secondary outcomes
Outcome | Definition |
Number of emails with questionnaires on HA issues sent (N1) | The total number of emails with questionnaires on HA issues sent. |
Number of emails with questionnaires on HA issues not bounced (N2) | The total number of emails with questionnaires on HA issues sent that had surveyees with valid email addresses. |
Number of questionnaires for which the surveyee was available (N3) | The total number of emails with questionnaires sent to assess HA issues with surveyees with valid email addresses and for which the surveyee was available. Unavailability can be the result of, e.g., automated responses such as ‘out of office’, ‘study leave’, ‘on strike’, ‘vacation leave’, ‘maternity leave’ etc. |
Number of partly or completely answered questionnaires (N4) | The total number of questionnaires on HA issues received back in which the questions were answered (either partial or completely). |
Number of completely answered questionnaires (N5) | The total number of questionnaires on HA issues received back in which all questions were answered. |
Overall response rates in questionnaires on HA issues | N4 or N5/N1, N2, or N3 |
Number of questionnaires that answered the question on review item 1* (N6) | The total number of questionnaires received back in which the question on review item 1* was answered. |
Response rate on review item 1* | N6/N1, N2, N3, N4 or N5 |
Number of questionnaires in which the respondents reported review item 1* (N7) | The number of questionnaires in which the respondents reported review item 1*, i.e., perceiving other co-author(s) as honorary author(s) on a publication. |
Prevalence of review item 1* (Primary outcome) | N7/N6 |
Number of questionnaires that answered the question on review item 2** (N8) | The total number of questionnaires received back in which the question on review item 2** was answered. |
Response rate on review item 2** | N8/N1, N2, N3, N4, or N5. |
Number of questionnaires in which the respondents reported review item 2** (N9) | The number of questionnaires in which the respondents reported review item 2**, i.e., having been approached by others to include honorary author(s) on a publication. |
Prevalence of review item 2** (Primary outcome) | N9/N8 |
Number of questionnaires that answered the question on review item 3*** (N10) | The total number of questionnaires received back in which the question on review item 3*** was answered. |
Response rate on review item 3*** | N10/N1, N2, N3, N4 or N5 |
Number of questionnaires in which the respondents reported review item 3*** (11) | The number of questionnaires in which the respondents reported review item 3***, i.e., admitting being an honorary author on a publication. |
Prevalence of review item 3*** (Secondary outcome) | N11/N10 |
Number of questionnaires that answered the question on review item 4**** (N12) | The total number of questionnaires received back in which the question on review item 4**** was answered. |
Response rate on review item 4**** | N12/N1, N2, N3, N4, or N5. |
Number of questionnaires in which the respondents reported review item 4**** (N13) | The number of questionnaires in which the respondents reported review item 4****, i.e., admitting adding an honorary author(s) on a publication. |
Prevalence of review item 4**** (Secondary outcome) | N13/N12 |
Number of questionnaires that answered the question on review item 5***** (N14) | The total number of questionnaires received back in which the question on review item 5***** was answered. |
Response rate on review item 5***** | N14/N1, N2, N3, N4 or N5 |
Number of questionnaires in which the respondents reported review item 5***** (15) | The number of questionnaires in which the respondents reported review item 5*****, i.e., admitting having approached others to include honorary author(s) on a publication. |
Prevalence of review item 5***** (Secondary outcome) | N15/N14 |
* Review item 1: Researchers perceiving other co-author(s) as honorary author(s) on a publication |
** Review item 2: Researchers having been approached by others to include honorary author(s) on a publication |
*** Review item 3: Researchers admitting being an honorary author on a publication |
**** Review item 4: Researchers admitting adding an honorary author(s) on a publication |
*****Review item 5: Researchers admitting having approached others to include honorary author(s) on a publication |
Assessment of methodological quality
The methodological quality of each survey will be assessed with a survey checklist of 14 items that was tailored to our research questions on HA issues. To develop this quality checklist we conducted pilot tests and consulted existing appraisal tools [29, 30, 39–46]. To rate the overall confidence in the results of the survey we adopted the rating scheme reported for the AMSTAR 2 critical appraisal tool [47], which is based on an assessment of critical-and non-critical items. As in the AMSTAR 2 tool we labeled 7 of the 14 items of our quality checklist as ‘critical’, because we believe that these items can critically affect the validity of a survey. In congruence with AMSTAR 2 we will assign 4 ratings: ‘High’, ‘Moderate’, ‘Low’, and ‘Critically low’ rating of the overall confidence in the results of the survey. A detailed description of this survey checklist and guidance on the rating scheme are given in additional file 3B. We will list the critical appraisal scores for each included survey and will calculate the prevalence of Yes scores (All Yes scores/Number of surveys) for each critical appraisal question (Additional file 3C) [48]. Two reviewers (RMR and NDG) will independently implement the methods reported for the 14 item quality checklist. This tool will be used for each of the 5 outcomes of this review. A third reviewer (GTR) will be consulted in the case of persistent disagreements. We will report when and why these consultations were necessary. Four surveys will be used a priori to calibrate the operators.
Data synthesis
Criteria for a quantitative synthesis
Outcomes in this review are prevalence statistics (proportions), which can be quantitatively synthesized. We may preclude meta-analyses for the following scenarios: (1) only 1 or no included studies (2) very different definitions of outcomes (3) incomplete reporting of proportions (4) biased evidence such as ‘Low’, and ‘Critically low’ ratings of the overall confidence in the results of the survey (5) explained and unexplained heterogeneity [49]. We will consider a I2 larger than 50% as an approximate rule of thumb for not conducting meta-analyses. When applying this rule we will consider that the value of I2 depends on the direction and magnitude of the outcomes and the strengths of the evidence for the identified heterogeneity [50]. Prior to precluding meta-analyses we will assess if solutions are possible for dealing with one or more of these limiting criteria [49].
Summary measures for a quantitative synthesis
The prevalence proportions and their exact (method = Wilson) 95% confidence limits across studies will be visually displayed in a forest plot. If the criteria for calculating a pooled estimate are met, we will pool the proportions and report them with their 95% confidence intervals. Double arcsine transformation will be performed prior to any statistical pooling. Summary estimates will be displayed after back-transformation [51]. These calculations will be performed using the metaprop command in Stata 16 [52]. We plan to implement a random-effects model, because between-study variance is expected.
Unit of analysis issues
To address unit-of-analysis issues we will assess in each study whether surveyees underwent more than one survey, e.g., surveys conducted at multiple time points on the same individuals.
Dealing with missing data
To address missing data we will contact the corresponding author and the author that was acknowledged as involved in the statistical analysis of the pertinent research studies. Such authors will be contacted by email. A first reminder will be sent one week after the first one and a second after two weeks. We will then wait for two weeks and accept the data to be missing and proceed.
Assessment of heterogeneity
We will assess the presence and the extent of heterogeneity. In the forest plots we will assess the overlap of the confidence intervals for the results of the individual surveys. We will conduct Tau2 (Estimate of between study variance) and Chi2 tests to measure statistical heterogeneity and I2 to quantify inconsistency [50].
Investigation of heterogeneity
We will assess survey-related and methodological diversity [50]. We will conduct subgroup analyses and meta-regression to investigate heterogeneity. The following explanatory variables for these investigations will be considered for these analyses:
Survey-related diversity
-
the type of authors that were surveyed, i.e., first authors versus corresponding or any other author in between (Table 1)
-
career levels of the surveyee, e.g. PhD students, post doctorates, department chairpersons etc.
-
the field of research on which the surveyee was interviewed, e.g., radiology, urology, dentistry etc.
-
the country of the first institution listed in the manuscript
-
the journal impact factor, e.g, above or below 5.00
-
the year of conducting the survey
-
the method of survey delivery, e.g., administered by email, internet platforms, by post etc.
-
anonymity of the surveyee
-
definitions of HA
-
the magnitude of the response rates, e.g. above or below 25%
Methodological diversity
-
the method of sampling, e.g. randomly selected or not
-
sample size
-
the time point for measuring outcome (the recall period), e.g., before or after 1 year of publication of the manuscript on which authors were surveyed.
-
study quality, e.g., low, moderate, or high quality
-
response rate
We will visually display the individual effects for each planned prevalence outcome in stratified forest plots according to these explanatory variables. To avoid misinterpretation of the findings we will not give the combined effect estimate in these plots if the criteria for a meta-analysis are not met [53].
We will also build Generalized Linear Mixed Models (GLMMs) to assess what factors contribute to each of the 5 review outcomes on HA issues. For each study that reported prevalence data on these issues the following data will be extracted and tabulated in an electronic spreadsheet: number of respondents and the number of individuals reporting each outcome. Each individual respondent was listed as a row in the electronic spreadsheet. For the GLMMs the presence/absence of an HA issue will be the dependent variable and all the explanatory variables reported previously as the predictors. Study ID will be included as a random effect, in order to account for clustering at a study level. The other variables will be included as fixed effect. The linearity assumption of the continuous variables will be assessed as described elsewhere [54–56] (1) the variables will be binned in quartiles; (2) the multivariate model will be fitted replacing the continuous variables with the four-level binned variables; (3) log odds of the upper three quartiles (the lower quartile will be used as indicator) will be plotted versus the respective quartile midpoints; (4) the four plotted points will be connected with straight lines and the plot will be visually inspected for linearity. Specific categories of categorical predictors will depend on categories used by primary investigators in their surveys. Depending on the total number of individual response data available, different modeling approach will be employed to include/exclude independent variables and avoid overfitting of the model. We will apply the same methods for additional explanatory variables that will be identified during the review process. We will explain in the final review why these variables were added.
Qualitative synthesis
We will conduct a systematic narrative synthesis whether quantitative syntheses will be possible or not.
Tables will be developed to report the characteristics of included surveys as reported in Additional file 3A. We will consult these tables for the data synthesis and assess relationships and diversity within and between surveys.
Sensitivity analysis
We will undertake sensitivity analyses to investigate the impact of certain decisions on the outcomes of the systematic review. Such decisions could refer to the searching of studies, the choice of certain eligibility criteria, and the quality of the included studies. However, which specific issues to explore in sensitivity analyses will be decided during the review process [50]. We will report the findings of these analyses in a summary table [50].
Meta-biases
We will assess the presence of non-reporting biases. Such biases occur when results are missing from studies that should have been included in the syntheses of the review [57, 58]. Non-reporting biases can come in many forms such as publication, time-lag, language, citation, multiple publication, location, and selective (non-) reporting bias [57]. We will implement a variety of strategies to address the risk of non-reporting biases such as:
-
Use a broad spectrum search strategy and searching studies in a wide body of search engines.
-
Assess the availability of study protocols and registers and if available assess whether the planned outcomes in the protocols are the same as those reported in the included studies.
-
Contacting authors regarding issues such as multiple publications of research data, information on the availability of protocols, unpublished or ongoing studies.
-
After the implementation of these 3 strategies we will adopt the 6 step Cochrane framework for assessing risk of bias as a result of missing results in a synthesis, e.g. a meta-analysis [58]. For the sixth step of this protocol an overall judgment on the risk of bias as a result of missing results for each synthesis will be given.
Confidence in the cumulative evidence
We will present summary of findings tables that report the magnitude of the outcome, the certainty or quality of evidence for each primary and secondary outcome, and other key data (Additional file 3C). We will adopt the GRADE approach to assess the overall certainty of the body of evidence for each outcome that was sought in this review [59]. For this approach we will assess: (1) bias in the included surveys, i.e., the overall confidence in the results of each survey based on our 14 item quality checklist (2) heterogeneity or inconsistency of results (3) indirectness of evidence (4) imprecision of results and (5) meta-biases [59].
The GRADE approach assigns four levels of certainty: ‘High’, ‘Moderate’, ‘Low’, and ‘Very low certainty’ [59]. The rationale for assigning these ratings will be given.
Differences between the protocol and the review
Any differences between what is described in this protocol and the methods implemented in the final systematic review will be reported with rationale. We will also explain, if possible, the potential consequences of these changes for the direction, magnitude, and the validity of the results [60].