Factors associated with scientific misconduct and questionable research practices in health professions education

Researchers engaging in scientific misconduct and questionable research practices (QRPs) is a noted problem across many fields, including health professions education (HPE). To mitigate these practices, fields outside HPE have implemented strategies based on researcher characteristics and practice factors. This study investigated associations between HPE researcher characteristics (e. g., gender, years of practice); publication pressure; and frequency of self-reported scientific misconduct and QRPs. Findings suggest that publication pressure, as well as several researcher characteristics, explain significant variance in investigator misconduct and QRPs in HPE. These findings provide an evidence base from which HPE might tailor strategies to address irresponsible research practices.

In health professions education (HPE), a recent international survey of researchers reported that 90% of almost 600 respondents admitted to engaging in scientific misconduct and questionable research practices (QRPs). These practices ranged from serious infractions, such as data fabrication and plagiarism, to less severe violations, like excessive self-citations, inappropriate data storage, and so-called ‘salami slicing’ [1]. Similarly, a study of authorship in HPE unearthed a number of questionable practices, such as granting honorary authorship and denying authorship to individuals who deserved it [2]. These recent studies highlight a potential problem in HPE, a problem that can harm the research enterprise by wasting limited resources, damaging the scientific record, disadvantaging unwitting researchers, and setting a poor example for trainees [3].

To inform approaches that might mitigate research misconduct and QRPs, researchers in biomedicine [4, 5], psychology [6], and business [7] have attempted to identify relationships between personal characteristics, practice factors, and researchers’ engagement in irresponsible research behaviours. To date, these studies have considered variables such as gender, career stage, geographical location, primary research methodology used, and pressure to publish [6, 8]. For example, based on a self-report survey of biomedical researchers, Tijdink and colleagues found that publication pressure and early career stage were associated with an increased likelihood of engaging in QRPs [5]. In another study that examined authors of retracted biomedical journal articles, investigators found that junior researchers were more likely to have had a paper retracted; however, this study did not identify publication pressure as a predictor of paper retraction [8]. These findings and others have led for calls for increased transparency in research and the utilization of open science practices in order to mitigate misconduct and QRPs [9].

Although we now have a sense of the frequency of scientific misconduct and QRPs in HPE [1], we do not yet know which personal characteristics or practice factors may explain a researcher’s engagement in these irresponsible behaviours. Additionally, as a relatively new field ‘without a roadmap to navigate these challenges’ [10], HPE is populated by multidisciplinary researchers who draw upon and combine varied backgrounds (e. g., biomedicine, psychology, education). Therefore, we propose that related findings from other fields may not generalize to the HPE context. This dearth of knowledge hampers our efforts to address misconduct and QRPs in HPE. Thus, the present study aims to address this gap by examining the associations between HPE researcher characteristics (e. g., gender, years of practice, academic background, geographic location, and primary methodology used in research); a potentially important practice factor (publication pressure); and the frequency of self-reported scientific misconduct and QRPs.

In 2017, we conducted an anonymous, cross-sectional survey of HPE researchers as a component of a larger program of research on responsible research practices. The survey’s purpose was twofold: (1) to measure the frequency of self-reported misconduct and QRPs in HPE and (2) to determine which researcher characteristics and practice factors, if any, might explain the frequency of these irresponsible behaviours. In 2018, we chose to address each of these purposes separately, as we felt that each warranted their own exploration and analysis, and each told a different yet equally valued story [11]. Therefore, we first published an article detailing the creation and execution of the survey, and we reported on the frequency of research misconduct and QRPs, thus confirming that these practices are present in HPE [1]. In the current article, we now focus our attention to the relationships between researcher characteristics, practice factors, and irresponsible research practices, including both deliberate scientific misconduct and QRPs, to help unearth important characteristics and practices that might be targeted for intervention. Ethical approval for this study was granted by the Ethical Review Board Committee of the Netherlands Association for Medical Education (Dossier #937).

Four hundred and sixty-three respondents from the curated sample completed at least a portion of the survey, resulting in a response rate of 28.2% (potential respondents n = 1,641). Based on the wave analysis, we identified a nonresponse bias statistic of 0.36. Since we utilized a six-point, frequency-response scale, this bias statistic represents a 6% difference between the scores from respondents in the last wave compared with those in the first. For practical purposes, this small difference is unlikely to have a meaningful effect on the results [15].

Our social media recruitment netted an additional 127 responses. Using a multivariate analysis of variance, we compared the curated sample with those in the social media sample and found statistically significant differences F(5, 524) = 6.67, p < 0.01. Post-hoc analyses indicated that those in the curated sample were slightly older (M = 47.4 years) and more experienced as HPE researchers (M = 11.0 years) than those in the social media sample (M = 40.7 years; M = 7.5 years). However, the groups were similar in the mean number of articles published and percentage of work dedicated to HPE research. Thus, because we aimed to understand the relations between HPE researcher characteristics, practice factors, and scientific misconduct and QRPs among a diverse, international sample, we combined the responses from the two groups and analyzed all participant responses together.

Descriptive statistics for the variables measured on a ratio scale are displayed in Tab. 1 and those on a nominal scale reported in Tab. 2. Results from the reliability analysis of the nine publication pressure items indicated that the item scores had a reasonably high internal consistency reliability coefficient (Cronbach’s alpha = 0.83)[16].

The results of the three-step hierarchical multiple linear regression are presented in Tab. 4. When the weighted misconduct score was used as the outcome variable, the normal probability plot of standardized residual indicated unsatisfactory model fit. Thus, we transformed the variable by taking its square root. The normal probability plot of standardized residual after transformation indicated a much improved model fit. Moreover, the variance inflation factor for all of the explanatory variables ranged from 1.06 to 3.52, suggesting very little multicollinearity [17].

All the regression coefficient estimates in Tab. 4 were calculated using the square root of weighted misconduct score as the outcome variable. In block 1, researcher age had a significant negative association with the outcome variable (b = -0.01, β = -0.28, t = -3.86, p <0.01); this suggests that older researchers tended to have lower misconduct scores. In total, the three variables entered in block 1 explained 5% of the variance in the outcome. In block 2, number of publications (b = 0.001, β = 0.13, t = 2.37, p < 0.05) had a significant positive association with the outcome when both age and years involved in health professions or medical education were controlled for in the preceding block. Compared with researchers in the region of North America, researchers in Asia tended to have higher misconduct scores (b = 0.28, β = 0.16, t = 3.53, p < 0.01). In addition, compared with those who defined their work role as clinician, those who defined their work role as researcher tended to have higher misconduct scores (b = 0.12, β = 0.13, t = 2.15, p < 0.05). Taken together, the six variables entered in block 2 explained an additional 6% of the variance in the outcome. Finally, in block 3, the publication pressure composite score was added and emerged as the strongest individual predictor of misconduct (b = 0.20, β = 0.34, t = 7.82, p < 0.01), controlling for all other explanatory variables in the model. As a group, all of the variables explained 21% of the variance in the misconduct score. Moreover, above and beyond the other explanatory variables already entered into the regression model, publication pressure explained an additional 10% of the variance in the misconduct score.

The present study examined the associations between HPE researcher characteristics, publication pressure, and the frequency of self-reported scientific misconduct and QRPs. Our findings suggest that the variables of age, number of publications, geographical location, work role, and publication pressure explain considerable variance in the frequency of researchers’ self-reported irresponsible research behaviours. Taken together, these findings provide the field with an evidence base from which to begin tailoring strategies to address scientific misconduct and QRPs in HPE.

Researchers have long described the pressure to publish as a contributor to the ‘dark side of science’ [18]. In biomedicine, researchers experiencing publication pressure are more likely to admit engaging in misconduct and QRPs [4, 8, 19, 20]. The findings reported here align with and extend this previous work. We found that both perceived publication pressure and the number of publications were associated with misconduct and questionable research conduct. To mitigate the pressure to publish, researchers have suggested modifying promotion and tenure structures, promoting and rewarding research transparency, and training senior researchers to act as responsible role models by demonstrating positive (and ethical) research practices [8, 9, 21]. Integrating such approaches in HPE may be beneficial; however, in many cases, HPE researchers are not directly assigned to a department of HPE, but rather are often members of clinical departments (e. g., Departments of Medicine, etc.). Therefore, for role modelling to have a positive impact on HPE researchers, such practices would likely need to be integrated at the institutional level.

To our knowledge, this is the first study to find a relationship between publication pressure and irresponsible research practices in HPE. Not only has publication pressure been connected with greater misconduct and QRPs, it has also been tied to researchers’ reluctance to share their work openly and to partner with other researchers and has been linked to higher levels of researcher burnout [22, 23]. Overall, these findings suggest that the topic of publication pressure may be ripe for further exploration in HPE.

Our findings indicate that older HPE researchers report misconduct and QRPs less frequently than young researchers. These results are compelling, especially considering the fact that older researchers have been involved in HPE research for much longer (r = 0.81) and thus have had more opportunities to act unethically. And yet, these results suggest they do not. Taken together, these findings provide some evidence for the idea that junior scientists may be particularly vulnerable to the lure of misconduct and QRPs in HPE or, alternatively, that junior researchers may simply be unfamiliar with responsible research practices [8]. In the biomedical and social sciences, this finding has led funders, such as the National Institutes of Health, to mandate that all grantees, including junior and senior researchers, complete responsible conduct of research training [24‐26]. While such an approach is a positive step, it is worth nothing that most HPE research is unfunded, making it possible that our junior researchers may not be offered or exposed to such training. Furthermore, researchers have wrestled with the scope of this training, raising concerns that it often falls short of addressing the increasingly diverse nature of research [24, 27, 28]. As noted by Keune et al., when studying common HPE topics, such as examining residents and trainees, this training’s lack of diversity can mean many researchers are operating without guidance on how to navigate common ethical challenges within HPE [10].

Our results suggest that HPE researchers in Asia reported higher frequencies of irresponsible behaviours than researchers in North America. It is important to note, however, that this does not mean Asian researchers are more unethical than those from other regions, since the observed differences could have resulted from any number of factors, including differences in survey item interpretation, researchers’ awareness of their own behaviours, and respondents’ truthfulness. Nonetheless, in the wake of concerns that research misconduct in Asia is on the rise, multiple research teams have undertaken Asia-specific programs of research [29] to further examine the problem. For example, a team of Chinese scientists recently conducted a survey and found that 40% of research from China may be tainted by researcher misconduct [30]. In that particular study, the respondents attributed misconduct to lack of training and institutional oversight, and to extremely high pressures to publish. In studying geographical influence more broadly, Fanelli suggests that geographical variation in research misconduct may be related to the presence or absence of policies and procedures [8].

Respondents identifying primarily as researchers reported higher frequencies of QRPs and misconduct than those identifying primarily as clinicians or ’other’ (e. g., teachers or administrators). Although speculative, this finding could be related to the idea that the stakes tend to be much higher for investigators whose research outputs and published papers represent the primary measures of success. On the other hand, a clinician who conducts education research ‘on the side’, may be less incentivized to cut corners. Regardless of the explanation, we would advocate for future research to more closely explore this relationship. Related areas of exploration could include the impact of widespread and long-standing ethics training in clinical education [31, 32] in contrast to approaches taken (or not taken) by graduate programs in HPE. Future studies might also consider differences in the incentive structures and other systematic pressures for researchers as compared with clinicians.

The results of this study suggest that a researcher’s age, number of publications, geographical location, and work role are associated with the frequency of self-reported misconduct and QRPs in HPE. We also found a positive association between perceived publication pressure and irresponsible research behaviours. Both scientific misconduct and QRPs have the potential to seriously damage the quality of our scientific work, as well as the public’s belief in its credibility. Ultimately, we hope these results can help academic leaders, journal editors, and others involved in educational research develop targeted measures to minimize scientific misconduct and QRPs in HPE.