How can we help PhD students cope with mental health issues and drop-out ? An extensive analysis of discourse


 There is a growing concern regarding the increase of the mental health problems among PhD students worldwide. This problem is worrying, and it remains a major issue for research teams and labs. However, the particularity of this environment has not been explored in consistent enough ways to provide a clear way forward for universities and health services to answer this problem. Therefore, we carried out a large online survey and collected 480 testimonies on health issues and work conditions from doctoral students and young doctors. Our aim was to identify levers and key factors for intervention. A lexicometric analysis of the discourse was carried out, using the Iramuteq package for R. Results highlight a need to intervene at different levels, in terms of instruction, prevention, care and follow-up to reduce mental health problems among PhD students. It also highlights the perceived importance of the university and feeling of belonging to a community as possible factors to reduce these problems and reduce drop out probability. More than anything, the study reveals that there are no differences in the discourse of doctoral students according to their field, which reveals the generality of these processes.


Theoretical Approach
Theoretical approach There is a growing concern regarding the increase of the mental health problems among PhD students, as it is an issue observed worldwide but despite calls to action, effective ways to prevent this issue remain to be designed. For example, a recent study of Evans, Bira, Gastelum, Weiss and Vanderford (2018) shows that this trend is found in the 26 countries of the study. It should be added that different comparisons reveal that PhD students are more depressed than other students or quali ed workers 2,3 .
Indeed, repercussions of mental health problems are likely to include lower research productivity, diminished workforce talent and overall 'lost economic and social potential' if doctoral students fail to complete their research 2,4 . This problem is costly, humanly and economically and PhD drop-out remains a major problem for research teams and labs 2,4 . In France, the drop-out rate varies between 5% (in scienti c and technical elds) and 45% 5,6 . The same trend is observed worldwide (i.e. a dropout rate from 30% to 50%) 7 . Also, it should be noted that PhD students are estimated to carry out half of the research conducted by universities 8 .
To answer this issue, recent research suggests that their high levels of stress could be caused in part or exacerbated by aspects of the doctoral education environment ( eld of research, type of supervision, grant… for a complete review of these aspects, see Waight & Giordano, 2018). However, the particularity of this environment has not been explored in consistent enough ways to provide a clear way forward for universities and health services to respond to this issue. Most of the available research on this topic present issues with the study design, sometimes involving very small sample sizes, convenience sampling among academic communities, lack of diversity or non-validated and non-standardized instruments for measurement or follow-up evaluation. Also, most studies conducted in North America do not distinguish PhD students from other students and when they do, they sometimes extrapolate results from much smaller samples within the larger study and it can be difficult to extract speci c factors related to health during PhD studies. Finally, if certain studies work on qualitative data, they mostly focus on speci c biographical or academic factors, and aim to understand coping mechanisms among speci c groups (LGBT or foreign PhD students). Due to this limitation, their results cannot be transferred beyond these populations.

Research question and objective
In a few words, working conditions during PhD studies affect both the PhD students (their health and well-being, their social environment) and their institutions 2,3 . Indeed, PhD students are also part of research teams and contribute to the research process. In these conditions, a deterioration of working conditions or health has an impact on the quality of the work, the career prospects of the PhD students, and also on the functioning of the research teams, and on a larger scale it constitutes a loss for research and innovation 2,4 . However, most of the studies rely on closed quantitative methodologies or extremely limited qualitative samples. Therefore, we carried out a large qualitative survey and collected testimonies from doctoral students and young doctors. The objective of such an approach was to precisely identify the organization of health issues associated with a PhD and to contextualize them to identify levers and key factors for intervention.

Methods
Participants and procedure 480 testimonies were collected as part of a wider research on health issues and work conditions of French PhD students (mean age = 28.13 years, 68.3% women). These testimonies were collected from a large panel of PhD students and represent all research elds which exist in France. Participants were asked to answer a wide question about their PhD, "In this part, we invite you to review your experience as a doctoral student (for example: your daily life, your motivations, your di culties, the image of the doctorate, etc.)". In addition to the sociodemographic variables (age, gender), an exhaustive set of academic variables were requested and introduced in our analysis: academic trajectory, elds of study, duration of PhD studies, self-assessment of the stage of completion of the thesis, type of supervision, the main source of funding. All details about the sample are presented in Table 1.
Participants were recruited by email using academic networks, specialized websites, and social networks. Several French universities and research groups also relayed our questionnaire. Only French PhD students and young doctors (less than 1 year since the graduation) were retained for the study, but all research elds were investigated. Explicit consent was requested at the beginning and at the end of the questionnaire. Note. N = 480. Mean age in the sample is 28.13 y.o. (SD = 3.06). a : co-tutelle is a particular type of cosupervision with a supervisor in a French institution and a supervisor abroad. b : a subjective assessment of the stage of completion was asked to complete the enrollment year and introduced in the analysis.

Data Analysis
In this study, we opted for a software approach to analyze the qualitative data. Indeed, this approach allowed us to analyze and compare many and large texts to identify the main themes in these texts and how the different themes are related to each other. Furthermore, the software offered a signi cant contribution in terms of data visualization. In a few words, a lexicometric analysis of the interviews was carried out, using the Iramuteq package for R (R Interface for Multidimensional Analyses of Texts and Questionnaires) 9,10 . Iramuteq software splits the corpus into Elementary Context Units (ECUs), consisting of segments from 10 to 15 words each. Firstly, we performed a Descendant Hierarchical Classi cation (DHC). The DHC searches for the most frequently associated words within these ECUs and regroups the most similar ECUs within classes -also called thematic pro les 11 . At each step, the largest class is divided in two in order to form a set of classes that are as exhaustive and exclusive as possible. Within each class, the ECUs as well as the independent variables are tested on the basis of a χ² test to measure the strength of their association with the class. Then, the χ² and its signi cance value make it possible to identify the words that structure the thematic classes and interpret them (Figure 1).
In a few words, the DHC makes it possible to obtain coherent thematic classes from the discourse of participants to interpret them, as well as the signi cant and most representative categories among the independent variables inserted in the analysis (sociodemographic and academic variables) associated with these classes.

Results
The DHC reveals that the discourse is organized according to six thematic classes (Figure 1) which indicates a strong heterogeneity in the discourse of the participants. In a few words, the analysis reveals a clear split between Node 1 ("Health") that regroups health problems (Theme 1 : "Physical, mental health and impostor syndrome", 10.6% of all ECUs ; Theme 2 : "Stress and malaise at work", 18.7%) and personal problems related to work (Theme 3 : "Work/life interface", 21.1%) and Node 2 ("Work") that regroups thematic classes related to work (Theme 3 : "Work organization and additional responsibilities", 12.4%), professional identity (Theme 4 : "Young researcher identity", 17.4%) and future perspective (Theme 5 : "Perspective, future and precariousness", 19.9%).

Discussion
The results reveal the predominance of negative elements in the representation of the PhD studies among PhD students and young doctors. Starting with the three themes regrouped under Node 1 (health), we can distinguish several aspects of mental health issues among which Theme 1 ("Physical, mental health and impostor syndrome") and Theme 2 ("Stress and malaise at work") seem to oppose one another. While Theme 2 refers to a certain level of stress, it also refers to tasks and expected di culties, considered as intrinsically part of PhDs and research: competitiveness, high levels of complexity, pressure to perform...
On these questions, research teams, labs and institutions have little room to maneuver. Most of this stress is related to the amount of work and di culties to maintain a high level of competence to carry out these tasks over the long term but it should be noted that it concerns mainly the 1 st year of PhD. A lack of skill, work organization, experience or self-e cacy could explain the predominance of this topic at the beginning of PhDs.
On the other hand, the rst theme depicts a psychopathological picture. It regroups terms associated with the impostor syndrome which relates to a mental state of self-doubt. Despite evident abilities and academic success, those suffering from this condition are unable to feel competent and accomplished as they live in a persistent fear of being discovered as intellectual frauds 12,13 . In addition, a certain number of psychological and behavioral symptoms linked to anxiety and depression are also regrouped in this theme (physical pain, sleeping problems, sick leaves…). In this regard, it should be noted that this early detection of anxious or depressive symptoms can be especially di cult if we consider the stress associated with research (as previously identi ed in Theme 2). Care should therefore take different forms.
When possible, it seems necessary to provide more information to doctoral students and their supervisors on anxiety and depressive symptoms, as well as information on the impostor syndrome which seems to be associated with these symptoms. Though, it is known that students are unwilling to seek or receive help from health services because they perceive emotional distress as a potential threat for a successful career progression if psychiatric issues are revealed, especially among students who belong to minorities [14][15][16][17][18][19] . It is accompanied by poor knowledge of these services and the type of support they can provide 17,20 . However, it seems essential to work on the representations and the lack of knowledge associated with these services to promote care and limit the long-term repercussions on physical and mental health, and at a broader level on general functioning and the ability to maintain activities, work and avoid dropout.
The same could be said for inability to maintain a satisfying work-life balance 2,21 . This challenge is often di cult to achieve in research and it cannot be considered as a part of PhD experience solely. However, acting at this level can limit personal and professional exhaustion and reduce the guilt and negative feelings associated with the inability to maintain this balance. In most cases, supervision focuses on methodological aspects or the research eld and leaves aside work organization and support. Universities can intervene on this topic and train doctoral students in e cient work methods and project management to reduce this inability. Such training should not only have a positive impact on the shortterm but also it should help PhD students to maintain their motivation and ability to work despite highly demanding tasks.
With regard to the factors linked to the work (Node 2), one of the key factors related to the doctorate experience seems to be the social environment (Theme 4) within which the doctoral student evolves, as well as the professional identity and the feeling of belonging he develops. The intellectual stimulation, the institution and the level of responsibility associated with his work are positive aspects which encourage the pursuit of the doctorate. In a way, these aspects reward the high level of demand and investment required. In fact, it appears more than ever necessary to take these aspects into consideration to promote the reduction of mental health problems among PhD students. In a nutshell, cognitive behavior therapy already demonstrated that dysfunctional thinking (such as the impostor syndrome) can alter mood, behavior and has a negative effect on health and general functioning 22 . It can also produce an underestimation of one's ability or a belief of incompetence. By emphasizing positive aspects of PhD studies (skills developed, feeling of belonging, intellectual stimulation…) it would be possible to reduce these dysfunctional beliefs and produce positive effects at these different levels.
Finally, the lack of security and con dence regarding post-doctorate and professional career (Theme 5) is another key element that needs to be addressed 21,23 . The integration of young doctors into society is more than ever necessary, whether in the public or private elds. Concrete actions to develop and support the post-thesis projects could bene t at several levels: reduce anxiety, increase well-being (and therefore reduce drop-out probability), establish new bridges between research and society. Unsurprisingly, this theme is associated with the end of the thesis. Especially if we consider the large number of ancillary tasks that seems intrinsically associated with PhD studies, among which we found tasks related to events, communication, teaching, etc. and leads PhD students to develop skills in a large number of elds outside of their eld of expertise. An approach of their professional career in terms of skills rather than expertise seems to be an interesting way to develop self-e cacy, con dence and to reduce anxiety.
Although this is a convenience sample the extensive amount of qualitative data should prompt both academia and policy makers to consider intervention strategies. This study highlights a need to intervene at different levels, in terms of instruction, prevention, care and follow-up to reduce mental health problems among PhD students. It also highlights the perceived importance of the university and feeling of belonging to a community as possible factors to reduce these problems and reduce drop out probability.