Chapter 1 Introduction and scope of work
1.4. Scope of work
1.4.2. Overview of chapters
Table 1-2 provides an overview of the chapters included in this thesis. Chapter 1 gives a general introduction and overview of the methods, while Chapters 2 to 5 represent the phases of the study as shown in Figure 1-2. Each of these chapters contains a short introduction and rationale, methods and results section, as well as the relevant references. Chapter 6 provides an overall summary of findings from Chapters 2 to 5 and an integrated discussion on relevant aspects. The concluding chapter proposes priorities for further research and activities to promote research integrity in LMICs. The individual chapters therefore do not represent complete publishable articles, do not stand on their own and should be read in context of the other chapters.
Table 1-2 Overview of chapters included in thesis
Chapter Overview
Chapter 1 Short introduction on research integrity and the rationale for conducting this research
Chapter 2
(Phase 1) A comprehensive synthesis of exsiting literature related to occurrence of research misconduct amongst health researchers from LMICs, using pre-specified objectives and inclusion criteria.
Chapter 3 (Phase 2)
Mixed-method study of LMIC health researchers’ perceptions on research reporting practices. Online survey amongst Cochrane authors living in LMICs, using scenarios to elicit responses on perceptions and occurrence of research reporting practices. Follow-up, in-depth interviews with willing respondents.
Chapter 4 (Phase 3)
Cross-sectional study of actual practices in African biomedical journals.
Measuring the presence and extent of plagiarism, assessment of journal policies and author guidelines on authorship, plagiarism and conflicts of interest, and author adherence to these.
Chapter 5
(Phase 4) Development, implementation and evaluation of a workshop to introduce research integrity and publication ethics, offered in Malawi and Nigeira.
Chapter 6 Summary and integrated discussion on the findings of Chapters 2-5.
Chapter 7 (Phase 5)
Concluding chapter. Implications for future activities to promote research integrity and implications for future research on research integrity in LMICs.
Ethical considerations
13 1.5. Ethical considerations
Ethical clearance was obtained from the Stellenbosch University Health Research Ethics Committee prior to commencement of the study (N14/12/158). Appendix 1.2 contains the approved protocol for the study. We obtained ethics exemption for Phase 3 (X17/08/010). As research misconduct is a sensitive topic, the candidate and her supervisors were aware of several potential ethical dilemmas. We were prepared to deal with such dilemmas in the following way:
Should individual cases of misconduct come to light during Phase 2 and 4, participants would be counselled and advised to use appropriate channels to report such misconduct. In case we found severe plagiarism in research articles during Phase 3, we would alert the editor of the journal to our findings. Institutions where workshops were held (Phase 4) were provided with general, anonymised feedback on researchers’ awareness of research integrity and any concerns about research misconduct.
1.6. References
1. Steneck, N.H. and Bulger, R.E. The history, purpose, and future of instruction in the responsible conduct of research. Acad Med, 2007. 82(9): p. 829-34.
2. Anderson, M.S., Global research integrity in relation to the United States' research-integrity infrastructure. Account Res, 2014. 21(1): p. 1-8.
3. Singapore statement, 2010. www.singaporestatement.org
4. Kombe, F., Anunobi, E.N., Tshifugula, N.P., Wassenaar, D., Njadingwe, D., Mwalukore, S., Chinyama, J., Randrianasolo, B., Akindeh, P., Dlamini, P.S., Ramiandrisoa, F.N., and Ranaivo, N. Promoting Research Integrity in Africa: An African Voice of Concern on Research
Misconduct and the Way Forward. Dev World Bioeth, 2013.
5. Horn, L. Promoting Responsible Research Conduct: A South African Perspective. Journal of Academic Ethics, 2016. 15(1): p. 59-72.
6. Vasconcelos, S.M., Sorenson, M.M., Watanabe, E.H., Foguel, D., and Palacios, M. Brazilian Science and Research Integrity: Where are We? What Next? An Acad Bras Cienc, 2015.
87(2): p. 1259-69.
7. Ana, J., Koehlmoos, T., Smith, R., and Yan, L.L. Research misconduct in low- and middle-income countries. PLoS Med, 2013. 10(3): p. e1001315.
8. The Expert Panel on Research Integrity, Canadian Council of Academics. Honesty, accountability and trust: Fostering research integrity in Canada. 2010: Ottawa, Canada.
9. All European Academcis (ALLEA). European Code of Conduct for Research Integrity Revised Edition. 2017: Berlin.
http://www.allea.org/wp-content/uploads/2017/03/ALLEA-European-Code-of-Conduct-for-Research-Integrity-2017-1.pdf
10. Steneck, N. ORI Introduction to the responsible conduct of research. 2007.
11. A consensus statement on research misconduct in the UK. BMJ, 2012. 344: p. e1111.
Chapter 1 - Introduction and scope of work
14
12. Hermeren, G., The Swedish Research Council's expert group on ethics. Good Research Practice. 2011: Stockholm, Sweden.
13. Fanelli, D. Redifine misconduct as distorted reporting. Nature, 2013. 494: p. 149.
14. Martinson, B., Anderson, M., and de Vries, R. Scientists behaving badly. Nature, 2005. 435:
p. 737-738.
15. Bouter, L.M., Tijdink, J., Axelsen, N., Martinson, B.C., and ter Riet, G. Ranking major and minor research misbehaviors: results from a survey among participants of four World Conferences on Research Integrity. Research Integrity and Peer Review, 2016. 1(1).
16. De Vries, R., Anderson, M.S., and Martinson, B.C. Normal Misbehaviour: Scientists Talk about the Ethics of Research. J Empir Res Hum Res Ethics, 2006. 1(1): p. 43-50.
17. Ferris, L.E. and Fletcher, R.H. Conflict of interest in peer-reviewed medical journals: The World Associaton of Medical Editors (WAME) position on a challenging problem. 2009.
18. World Association of Medical Editors (WAME). Conflict of Interest in Peer-Reviewed Medical Journals. 2009 [cited 2017 6 October]; Available from:
http://www.wame.org/about/conflict-of-interest-in-peer-reviewed-medical.
19. PLoS Medicine Editors. Making sense of non-financial competing interests. PLoS Med, 2008.
5(9): p. e199.
20. Wager, E. Defining and responding to plagiarism. Learned Publishing, 2014. 27(1): p. 33-42.
21. Bazdaric, K., Bilic-Zulle, L., Brumini, G., and Petrovecki, M. Prevalence of plagiarism in recent submissions to the Croatian Medical Journal. Sci Eng Ethics, 2012. 18(2): p. 223-39.
22. Cameron, C., Zhao, H., and McHugh, M.K. Perspective: publication ethics and the emerging scientific workforce: understanding "plagiarism" in a global context. Acad Med, 2012.
87(1): p. 51-4.
23. Wheeler, G. Plagiarism in the Japanese universities: Truly a cultural matter? Journal of Second Language Writing, 2009. 18(1): p. 17-29.
24. Li, Y. Text-based plagiarism in scientific writing: what Chinese supervisors think about copying and how to reduce it in students' writing. Sci Eng Ethics, 2013. 19(2): p. 569-83.
25. Satyanarayana, K. Plagiarism: a scourge afflicting the Indian science. Indian J Med Res, 2010. 131: p. 373-6.
26. Roig, M. Avoiding plagiarism, self-plagiarism, and other questionable writing practices: A guide to ethical writing. 2006.
27. World Association of Medical Editors (WAME). Recommendations on Publication Ethics Policies for Medical Journals. 2017 [cited 2017 8 August]; Available from:
http://www.wame.org/about/recommendations-on-publication-ethics-policie.
28. Bird, S. Self-plagiarism, recycling fraud, and the intent to mislead. J Med Toxicol, 2008. 4(2):
p. 69-70.
29. International Committee of Medical Journal Editors (ICMJE). Recommendations for the conduct, reporting, editing and publication of scholarly work in medical journals. 2016.
30. Marusic, A., Bosnjak, L., and Jeroncic, A. A systematic review of research on the meaning, ethics and practices of authorship across scholarly disciplines. PLoS One, 2011. 6(9): p.
e23477.
31. Albert, T. and Wager, E. How to handle authorship disputes: a guide for new researchers.
2003.
32. Fang, F., Steen, R., and A, C. Misconduct accounts for the majority of retracted scientific publications. PNAS, 2012. 109(42): p. 17028-33.
33. Fanelli, D. How many scientists fabricate and falsify research? A systematic review and meta-analysis of survey data. PLoS One, 2009. 4(5): p. e5738.
References
15 34. Kaiser, M. The integrity of science - lost in translation? Best Pract Res Clin Gastroenterol,
2014. 28(2): p. 339-47.
35. Committee on Assessing Integrity in Research Environments, National Research Council, Institute of Medicine. Integrity in Scientific Research: Creating an Environment That
Promotes Responsible Conduct. 2002, Washington, D.C.: THE NATIONAL ACADEMIES PRESS.
36. Rossouw, T.M., van Zyl, C., and Pope, A. Responsible conduct of research: Global trends, local opportunities. South African Journal of Science, 2014. 110(1-2): p. 30-35.
37. Resnik, D.B., Rasmussen, L.M., and Kissling, G.E. An international study of research misconduct policies. Account Res, 2015. 22(5): p. 249-66.
38. Bunniss, S. and Kelly, D.R. Research paradigms in medical education research. Med Educ, 2010. 44(4): p. 358-66.
39. Denzin, N.K. and Lincoln, Y.S. The SAGE handbook of qualitative research. 2011: Sage.
40. World Bank. World Bank list of economies. 2016; Available from:
http://Siteresources.worldbank.org/DATASTATISTICS/Resources/CLASS.XLS
17
Chapter 2
Taking stock of existing research
A systematic review on the prevalence and factors associated with research misconduct
Summary
During the first phase of the PhD, we conducted a systematic review that aimed to summarise existing literature from low-and middle-income countries (LMICs) on health research reporting practices. We included sectional studies of health researchers from LMICs and
cross-sectional studies of biomedical research articles from LMIC authors or published in LMIC journals that assessed the prevalence of research misconduct related to research reporting practices; as well as factors influencing research misconduct. We searched a wide range of databases up until 16 February 2017, contacted experts in the field and checked reference lists of included studies.
One author screened titles and abstracts to exclude obviously irrelevant studies, and two authors independently screened all potentially relevant citations and full-texts. We extracted data using a pre-specified and piloted data extraction form and assessed risk of bias in duplicate using an adapted version of the tool by Hoy and colleagues. Due to heterogeneity between studies, we were not able to pool prevalence estimates and results were summarised narratively.
Chapter 2 - Taking stock of existing research
18
We included 32 studies reported in 33 publications, comprising cross-sectional studies of health researchers (n=22) and cross-sectional studies of biomedical research articles (n=10).
Cross-sectional studies of health researchers, comprising faculty members, health care practitioners and authors of research articles, were conducted in Latin America (n=3), sub-Saharan Africa (n=3), the Middle East (n=6) and South Asia (n=12) and addressed authorship practices, plagiarism, conflicts of interest and research misconduct in general. Overall risk of bias was judged to be high for 12 and moderate for 10 studies.
Cross-sectional studies of biomedical research articles represented research from Latin America (n=4), sub-Saharan Africa (n=2), Middle East (n=2), South Asia (n=1) and East Asia (n=4) and addressed authorship practices, plagiarism, conflicts of interest, redundant publication and research misconduct in general. Overall risk of bias was judged to be high for one, moderate for seven and low for two.
Studies mostly reported on the prevalence of research misconduct. Cross-sectional studies of health researchers generally reported on the proportion of participants that admitted to having engaged in misconduct or the proportion of participants admitting to knowing of others who have engaged in misconduct. Cross-sectional studies of biomedical research articles reported on the proportion of articles with evidence of misconduct. The reported prevalence of guest authorship ranged from 6% to 66% (n=10), ghost authorship from 6% to 43% (n=4), plagiarism from 5% to 89% (n=8), non-disclosure of conflicts of interest from 45% to 98% (n=7), non-disclosure of funding sources from 28% to 58% (n=4), redundant publication from 5% to 33% (n=6), and data fabrication or falsification from 10% to 91% (n=3). Few cross-sectional studies of health researchers reported on factors influencing research misconduct and reported that factors such as pressure to publish, academic expectations, lack of knowledge and inadequate punishment for research misconduct may influence behaviour.
19 Appendices
Appendix 2.1: Protocol of systematic review Appendix 2.2: Search strategies
Appendix 2.3: Data extraction form Appendix 2.4: Table of excluded studies
Appendix 2.5: Characteristics of included studies Appendix 2.6: Summary of risk of bias across studies Appendix 2.7: Prevalence of research misconduct
Appendix 2.8: Summary of results of secondary outcomes
Linked presentations
Rohwer A, Young T, Wager E, Garner P. Research integrity in low- and middle-income countries:
systematic review of prevalence of poor authorship practice, plagiarism and other misconduct.
Global Evidence Summit 2017, Cape Town, South Africa, 12-16 September 2017 (poster presentation)
Rohwer A, Young T, Wager E, Garner P. Mapping the literature on health research reporting practices from LMICs. 5th World Conference on Research Integrity 2017, Amsterdam, The Netherlands, 28-31 May 2017 (poster presentation)
Chapter 2 - Taking stock of existing research
20
2.1. Introduction and rationale
Although researchers have studied research misconduct in high-income countries, there is little research on research misconduct in LMICs (1, 2). Ana and colleagues (2013) were the first to provide an overview of existing studies on research misconduct in LMICs, in form of a published essay (2). They found few publications from a search in MEDLINE, which related mostly to high-profile case reports. To supplement their search, they conducted a survey amongst research centres from various LMICs and found that there were limited discussions around the topic and that regulatory bodies were lacking.
Existing systematic reviews on research misconduct
Through a snowball search, we found four systematic reviews that reported on the prevalence of research misconduct. Three of these mostly included studies from high-income countries (3-5), while one only included studies from Brazil (6).
In 2009, Fanelli published a systematic review that assessed the prevalence of research
misconduct by synthesising data from surveys. Only studies examining data falsification and data fabrication were included and no other questionable research reporting practices were assessed.
Fanelli included 21 surveys in the review, of which 18 contributed data to the meta-analysis. He reported a pooled estimate of 1.97% (95%CI 0.96 to 4.45) for self-reported fabrication or
falsification and 14.12% (95%CI 9.91 to 19.72) for participants knowing about others who did this.
No studies from LMICs were included in the review (3).
In a subsequent review, Fanelli teamed up with a co-author and assessed the prevalence of
plagiarism but did not include any studies from LMICs (4). They included 17 surveys and reported a pooled estimate of 1.7% (95%CI 1.2 to 2.4) for participants admitting to having plagiarised
themselves and 30% (95% CI 17 to 46) for participants knowing about others who had done so.
Although this study was published in October 2014, the date of the last search was December 2011. As pointed out by the authors in the discussion, they therefore omitted recently published surveys on research misconduct from LMICs.
Marusic and colleagues (2011) examined authorship practices across scholarly disciplines and included 118 studies, of which 54% were related to health sciences (5). They reported a pooled estimate of 29% (95%CI 24 to 35) of researchers reporting misuse of authorship in self or others.
Objectives
21 The pooled estimate for studies conducted outside of the USA or UK, namely France, India,
Bangladesh and South Africa was 55% (95%CI 45 to 64).
Padua and colleagues (2015) conducted a systematic review of publications on scientific research integrity in Brazil (6). They included 19 articles that addressed plagiarism, conflicts of interest, authorship conflicts, rules and guidelines and general perceptions of research integrity. However, only three of the included articles represented original research and authors of the review did not report data on the prevalence of research misconduct.
To our knowledge, the study by Ana and colleagues (2013) is the only attempt to map the
literature on research integrity across LMICs. However, they limited their search to MEDLINE and did not aim to conduct a systematic review. There is thus currently no comprehensive, rigorous, up-to-date summary of the available literature on research integrity and misconduct in LMICs.
Taking stock of existing studies by conducting a thorough up-to-date search of various databases is important to inform further phases of this study.
2.2. Objectives
To summarise from existing studies:
• The prevalence of research misconduct in reporting research amongst health researchers in LMICs
• The factors influencing good and poor practices in LMICs
2.3. Methods
We conducted a systematic review according to pre-specified objectives and methods. The protocol is available in Appendix 2.1.
2.3.1. Criteria for considering inclusion of studies 2.3.1.1. Types of studies
We considered published and unpublished cross-sectional studies containing data on the prevalence of and the factors associated with research misconduct in a specific population. We included surveys of researchers as well as surveys of biomedical journal articles.
Chapter 2 - Taking stock of existing research
22
2.3.1.2. Types of participants
We considered studies conducted amongst health researchers in LMICs (as defined by the World Bank) for inclusion. Health researchers comprised faculty members or health care practitioners involved in research and authors of biomedical research articles. Journal articles reporting on health research with authors based in LMICs, or published in journals from LMICs were also included. Studies with participants from a variety of disciplines (i.e. not restricted to health researchers) were only included if results were stratified according to disciplines and data for health researchers could be extracted. Similarly, studies that were conducted across regions were only included if results were stratified according to regions and data for LMICs could be extracted.
2.3.1.3. Content
We included studies on research reporting practices, namely data fabrication, data falsification, plagiarism, conflict of interest, authorship practices, acknowledgement practices, redundant publication (duplicate publication, salami publication and text-recycling), delayed publication and accurate reporting of results.
We excluded studies that investigated research integrity when planning, conducting and reviewing research as well as studies focusing on academic integrity (i.e. student cheating).
2.3.1.4. Types of outcomes
We included studies that addressed the primary or secondary outcomes.
Primary outcome
Prevalence of any type of misconduct related to research reporting, reported as:
• The proportion of health researchers admitting to having engaged in poor practices
• The proportion of health researchers admitting to knowing about others who have engaged in poor practices
• The proportion of biomedical journal articles with evidence of poor practices
Secondary outcomes
Factors influencing research misconduct, reported in cross-sectional studies:
• Knowledge of good practices
• Attitude towards research misconduct
Methods
23
• Perceptions on research misconduct and on factors influencing research misconduct 2.3.2. Search methods for identifying studies
2.3.2.1. Electronic searches
We searched MEDLINE via PubMed, Scopus, CINHAL, ERIC, PsychInfo, Web of Science, LILACS and Africa-Wide (date of last search 16 February 2017). The search strategies contained a combination of text words and MeSH terms of the terms “research integrity” and “low-and middle-income countries”. The search strategies for the various databases are detailed in Appendix 2.2. We did not impose any restrictions based on language or publication status.
2.3.2.2. Searching other resources
We screened reference lists of included studies for potentially eligible studies. In addition, we contacted experts in the field to find out whether they were aware of any unpublished studies. We specifically contacted experts in China and India to enquire about relevant studies that might not have been identified in the literature search.
2.3.3. Study selection, data collection and analysis 2.3.3.1. Selection of studies
One author (AR) screened all titles and abstracts of search outputs to exclude all the clearly irrelevant studies. Two authors (AR and EW) independently screened titles and abstracts of potentially relevant studies. We retrieved full texts of these studies and two authors (AR and EW) independently screened full texts to determine eligibility. We included cross-sectional studies of health researchers and biomedical research articles from LMICs that addressed research reporting practices and reported on either prevalence or factors influencing research misconduct. We resolved discrepancies through discussion, contacted authors in case of missing information and listed reasons for excluding studies.
2.3.3.2. Data extraction and management
One author (AR) extracted data using a pre-specified and pre-piloted data extraction form (Appendix 2.3). We extracted descriptive data related to study design, objectives, participants, data collection methods and outcomes. Results related to prevalence of research misconduct and factors associated with research misconduct were also extracted. Where studies included
participants from various disciplines, only data related to health researchers were extracted.
Chapter 2 - Taking stock of existing research
24
Similarly, when studies included participants from LMICs and high-income countries, we only extracted data relevant to LMICs. We contacted authors of studies in case of missing data.
2.3.3.3. Assessment of risk of bias
For cross-sectional studies of health researchers, we adapted the tool by Hoy et al. (2012) (7) to make judgements about risk of bias of included cross-sectional studies. We assessed risk of
selection bias, non-response bias, measurement bias and bias related to the analysis by answering guiding questions for each domain (Table 2-1), according to the guidance provided by Hoy and colleagues (7). As the original tool was designed for population-based prevalence studies, we adapted the questions to be suitable for cross-sectional studies of research articles. This enabled comparison of risk of bias across all included studies. We answered each question with yes (indicating low risk of bias), no (indicating high risk of bias) or unclear (indicating unclear risk of bias). For each included study, we reported a summary risk of bias score for each domain (low risk, high risk or unclear risk) and an overall risk of bias score across domains (low risk, moderate risk, high risk). For overall risk of bias, we used the definitions as per Hoy et al. (2012) (6):
• Low risk of bias: Further research is very unlikely to change our confidence in the estimate
• Moderate risk of bias: Further research is likely to have an important impact on our confidence in the estimate and may change the estimate
• High risk of bias: Further research is very likely to have an important impact on our
• High risk of bias: Further research is very likely to have an important impact on our