General health and stressors.

General health was assessed by two items of the Perceived Health Questionnaire (PHQ-2) [26]. Participants were asked to indicate how frequently they had experienced depressed moods and anhedonia over the past four weeks on a scale from 1 (not at all) to 4 (nearly every day). Additionally, they were presented with seven items of the Generalized Anxiety Disorder scale (GAD-7) [27] capturing the severity of various anxiety signs like nervousness, restlessness, and easy irritation on a scale from 1 (not at all) to 4 (nearly every day). Both scales were used in this combination in a previous study in German higher education [28]. Furthermore, we included two binary questions on whether the participants are currently in psychotherapy and if they have ever been diagnosed with a mental disorder.

The condensed version of the Perceived Stress Scale (PSS) [29] was used to get the degree of stressful situations in life in the last twelve months or since the start of the Ph.D. [30]. The response scale ranged from 0 (never) to 4 (very often), the following being a sample item: “… how often have you felt that you were unable to control the important things in your life?” To check the internal consistency of the four items, we calculated Cronbach’s alpha which was .79.

Job satisfaction and life satisfaction.

Three items on a scale from 1 (strongly disagree) to 5 (strongly agree) were used to measure job satisfaction [31], where a higher mean score indicated higher job satisfaction. A sample item is: “I am satisfied with my job.” Cronbach’s alpha was .86. Additionally, we added one item concerning general life satisfaction [adapted from 32] with the same response categories to get a more holistic insight.

Job insecurity.

To assess the fear of losing the job itself, quantitative job insecurity was measured with three items (e.g., “I am worried about having to leave my job before I would like to.”) [33] on a scale from 1 (strongly disagree) to 5 (strongly agree). We calculated a mean score with higher scores indicating higher job insecurity. Cronbach’s alpha was .80.

Institutional and systemic stressors.

For institutional stressors, we focused mainly on the role of supervision and included eight questions, four were framed using positive wording and four with negative wording, each with a scale from 1 (not at all) to 5 (all of the time). We summarized these questions in two constructs (positive support/negative support) which had Cronbach’s alphas of .85 and .76, respectively. As for systemic stressors, we included two questions on long-term contracts and on future perspectives, again using a scale from 1 (strongly disagree) to 5 (strongly agree).

COVID-19.

To cover the potential impacts of the COVID-19 pandemic and the implemented regulations, we included two questions to evaluate whether the pandemic affected the students’ general situation. On the one hand, participants were asked to pick the statement that best describes the effects of the pandemic in general (“yes, it improved my general situation”, “yes, it worsened my general situation”, “yes, but it neither worsened nor improved my general situation”, “no”), and on the other hand, they were asked to evaluate whether the particular answers provided in this survey had been affected by the pandemic from 1 (very likely) to 5 (very unlikely).

Causes of stress and potential solutions.

We included three open-ended questions in the questionnaire to get a deeper understanding of the perceived causes of stress, potential ways to improve mental health, and ways to improve the overall situation of Ph.D. students. The questions were: (1) “What is/are the cause(s) of your stress?” (2) “What would need to change to improve your mental health status?” (3) “What could be done to improve your situation?” Participants could mention as many points as they wanted (without any word limit). To analyze these questions, we built categories by following the model of inductive category development [34]. Two raters screened the first and last 20 responses in the data set and created categories for reoccurring topics (for a list containing all categories see S5–S7 Tables). In the next steps, two new raters rated all open answers with the developed categories and added additional categories if needed. Applicable categories were rated with 1 (“category was mentioned”) or 0 (“category was not mentioned”). For example, the following response to question (1) “[My] supervisor is on maternity leave with open end, i.e. I have no one to talk to about my topic and have almost nothing so far […] I feel like I’m not good enough at this, not sure I will be able to succeed–everyone else has other projects and publications except me–no topic-related network” was rated with 1 in the following four categories: supervision (quality & quantity), social integration & interactions (private & professional), self-perception (internal factors), and perceived lack of relevant competences & experience–(sense) of progress and success. The full list of categories and inter-rater reliability as measured by Krippendorf’s Alpha is reported in Table 3 [35].

Causes of stress.

The question “What is/are the cause(s) of your stress?” was answered by n = 446 participants. To cover the breadth of the responses, we built 18 categories. The most frequently mentioned categories were Workload & Time Pressure (mean rating frequency = 211), Self-Perception (M = 132.5), Job-Insecurity (M = 93), Social Integration & Interactions (M = 91), and Supervision Quality & Quantity (M = 88.5). The category Workload & Time Pressure includes all responses referring to the amount of work and/or deadlines. The category Self-Perception includes responses that indicate a perceived lack of competences or other personal doubts, concerns, and worries (e.g., “Since I started my Ph.D. I have almost constantly felt stupid”, “feeling like not belonging in academia, lack of self-confidence, feeling of making too little progress”). The category Job Insecurity reflects responses regarding contract length and general uncertainty about future employment (e.g., “scholarship is to be ended”, “Not knowing how things will work out after the PhD”, “Hopelessness of scientific career because there are too few full-time positions”). The category Social Integration & Interactions covers responses regarding the integration and sense of belonging in the work environment (e.g., “not valued by colleagues”, “being socially isolated at work”) as well as social issues in the private life (e.g., “Mostly my personal life, or often the lack thereof”, “problems with parents”). The category Supervision Quality & Quantity was used to capture all supervision-related responses including comments about the lack of support, feedback, frequency of meetings, or supervisors’ interest in the topics (e.g., “no clear communication with supervisor”, “lack of support from supervisor, even gossiping about me behind my back”).

Potential ways to improve the mental health status.

When asked “What would need to change to improve your mental health status?”, the Ph.D. students’ responses (n = 307) included various topics, some addressing compensation and income-related aspects, others highlighting supportive supervision. Overall, the responses lead to twelve different categories. Most answers referred to Supportive Supervision (M = 98.5), followed by Job Security/Contract (M = 59). Sample quotes with respect to supervision are e.g., “more feedback from supervisor or even more interest in my topic” or “more regular support by supervisor”. The category Job Security/Contract contains comments with respect to contract length and aspects for future employment (e.g., “no more worries about not being able to get my contract renewed”). The category Manageable Workload (M = 56.5) includes all responses around work-life balance (e.g., “having also activities beside work”, “clear work hours”). The fourth category was Compensation & Financial Security (M = 35) and included all income- and compensation-related aspects of the job (e.g., “Be paid 100% would be a start”, “Get paid for all the time at work”). The category Less Additional Tasks (M = 27.5) was used to specifically cover responses mentioning the number of additional tasks within the job (“Less work in teaching/work unrelated to PhD”).

Ways to improve the personal situation.

In addition to the previous question, which focused on general ways to improve the mental health status, we asked the Ph.D. students the following question: “What could be done to improve your situation?” Based on the themes and topics mentioned in the responses (n = 281) we built eleven categories. The categories mentioned the most were Job-Security & Compensation (M = 85.5), followed by Supportive Supervision (M = 68), Services and Support System (M = 39.5), Decrease Pressure to Perform (M = 39.5), and Manageable Workload (M = 36). The category Job-Security & Compensation includes responses like “chances of getting a long-term job in academia, not just the three-year programs” or “Fair payment (half of students get 50% others 65% even at the same institute)”. For the category Supportive Supervision “Regular meetings with people who are supportive & have an expertise in my research topic” can serve as a sample quote. The category Services and Support System was built to cover the responses named a solution outside the working group and team, such as “it would be helpful to see a university-based psychologist outside of the regular working hours” or “more courses (or better communications about them) about stress management”. The next category was labeled Decrease Pressure to Perform and included all responses that highlighted a high level of perceived pressure, such as “the performance pressure (every talk at a seminar is a job talk) is a big problem” or “Instead of pressuring academics to publish as much as possible, there should be more focus on the quality instead of the quantity of their articles/publication”. The last category, Manageable Workload, contained answers with respect to the amount of work (e.g., “Normal working hours, having really free-time without having the feeling that I should be working, it should be normal to take all vacation days”).

Summary of the qualitative answers.

With respect to the open answers, it can be summarized that the factors named as causes for stress and the possible solutions cover a wide range of topics. However, there are reoccurring topics across all three questions, such as supervision, workload, and job security. The role of supervision is a reemerging motif in the qualitative content analysis. While the quality and quantity of supervision were seen as a cause of stress, supportive supervision has a positive impact on the mental health status as well as the whole situation of the Ph.D. students. Furthermore, job insecurity was mentioned as an important stressor, while stable contracts and appropriate compensation for the work and fewer extra tasks were also added for improvement. Workload and time pressure were the most often stated causes of stress, followed by self-doubts and worries about not having enough competencies for the job. A manageable workload, fewer additional tasks, and a lower pressure to perform were indicated by the participants as valuable improvements.

Generalization.

While we aimed for our study to reflect the current situation for Ph.D. students as best as possible, there are points that are limiting the generalization of the results or are beyond the scope of this survey. First, we collected the data between October and December 2021, a time at which the ordinance on protection against risks of infection with the SARS-CoV-2 virus (“Coronavirus-Schutzverordnung”) [48] was still in place in Germany and influenced private and working life. About one-third (33.5%) of our study population stated that it is very likely or likely that the pandemic affected their answers. Nonetheless, a pandemic is a situation that can reoccur and is only one more reason to proactively set up a resilient Ph.D. graduation system. Another research group [49] investigated how mental health care should change as a consequence of the COVID-19 pandemic and concluded that the pandemic could even be seen as a chance to improve mental health services [49]. Nevertheless, we would like to point out that generalizing from a mental health study conducted during a pandemic may be difficult.

Overall, around 23% of all Ph.D. students at the University of Tübingen [50] participated in our study, which is slightly below the response rate in other similar studies [e.g., 16]. Considering that university students are very frequently invited to various questionnaires and studies, and given that our survey lasted approximately 20 minutes, it can be argued that the participants were motivated to invest time into their responses. However, our study population remains small compared to the total number of Ph.D. students in Germany. Moreover, we want to emphasize the likely sample bias in our data. We recruited participants mainly via mailing lists and our project therefore probably has especially appealed to people who are already interested in health or aware of mental health issues. However, given our relatively large coverage of almost a quarter of all Ph.D. students at the University of Tübingen, even a selective sample can give us insights into overall tendencies. The transferability of our results to other German universities or even universities in other countries is also not guaranteed as the academic systems can largely differ. Additionally, the results of this study are influenced by the overall living conditions the Ph.D. students experience. As Tübingen is a small town in the southwest of Germany, a comparison to larger cities or other countries might not be viable as the conditions probably differ largely.

Finally, even within one university, the generalization of our results is further limited by the uneven distribution of the participants across faculties. Most participants (61.8%) were from the Science Faculty, which is also the largest department (in terms of the highest total number of students) at the University of Tübingen. This skewness limits the faculty-wise comparisons, and we would expect to find interesting insights into the different graduate programs by conducting detailed comparisons. These differences could not only arise from different academic traditions but also from the highly varying expectations on the scope of a Ph.D. thesis. It follows that more detailed and systematic monitoring and data collection in national and international surveys are needed.

Methodology.

In a cross-sectional study, we investigate the current situation of Ph.D. students. While this is a valid and important instrument to access the current state, it cannot give us information about the dynamic changes in the transition phase between undergraduate studies and the Ph.D. as well as across the Ph.D. [51]. To track these changes or make comparisons over time, a longitudinal study design or propensity score matching procedures [52] could give further insights. It is therefore desirable to establish regular surveys and monitoring systems either on a university level or in a national survey to provide information on the impact of undertaken actions and implemented changes. We used a mixed quantitative and qualitative research approach. While this provides information on distinct levels, there are some pitfalls. For example, the open answer categories were defined post-hoc. While this gives the possibility for the participants to express their thoughts freely, it makes a systematic analysis more difficult, and the analysis might be biased by the evaluators. Overall, it is important to summarize and statistically analyze our study results on an overall level, but it must not be forgotten that every person and Ph.D. project is individual.

Research.

The overall scarce data, paired with worrisome flashlights on the mental health situation of Ph.D. students in different countries, highlights the need for more systematic monitoring of mental health in academia. For this purpose, standardized as well as domain-specific scales for Ph.D. students need to be established and longitudinal data needs to be collected. This would enable researchers to measure the effect of larger environmental changes (such as the COVID-19 pandemic or economic developments) and to measure the impact of interventions targeted to improve the situation. At the same time, we propose including qualitative measurements to assess unknown variables and the unique situation each Ph.D. student faces. These could also inform the development of additional quantitative measurable constructs to reflect the dynamic situation in academia. Such monitoring systems can either be implemented at the university level to give detailed insights into the situation at a specific university or on a national level to get an overall impression of Ph.D. students’ health issues. Optimally, a survey should be promoted from an independent self-governing institution dedicated to advancing science and research. While the demands for a better mental health situation for Ph.D. students are obvious, systematical and political changes need to be addressed in the research community and in academia.

Practice.

Our mixed methods research approach allows us not only to find out more about the issues of Ph.D. students but also to draw conclusions about what is needed to improve their situation. However, finding solutions to a recognized problem is not a straightforward task, and complex problems often require a step-by-step solution. Therefore, we assume that more practical implications, which could be indicated by an established monitoring system, will be necessary once the first steps have been taken.

In general, we can group interventions into at least four levels that can influence each other: the Ph.D. students themselves, the supervisors, the universities or research institutions, and the greater political context and academic culture. Building on the responses about potential improvements and additional services, we identified the following practical implications:

On an individual level, the main interventions could happen in capacity building (e.g., in time/project management, self-reflection or mental health awareness) but also by being more proactive about changing working modes (e.g., establishing collaborations or a peer counseling system) or by improving the social environment. This could additionally lead to a change in self-perception, for which direct interventions might be more difficult. At this point, we want to highlight that changes on the individual level aim to prevent the development of mental health problems and strengthen the resilience of Ph.D. students. They can at no point replace professional support once such problems have been manifested.

The level of supervision seems to be the most urgent and promising target for an improvement of Ph.D. students’ situation. As supervisors are usually defining a project and its goals, but also additional teaching or other tasks, they are responsible for setting the workload and time constraints. Not only the hard constraints of the working conditions but also the quality of supervision was often mentioned to be highly deficient. Possible interventions could target improving the skills in personnel management of supervisors. But also, clear supervision requirements and guidelines could be imposed by the university. Such agreements (including expectations on the thesis, supervision times and conciliation mechanisms) might be an option to enhance the agreements in a supervisor-student relationship. While these suggestions are not new, and some of them are theoretically established in some university departments, our study results suggest that they are often ignored or not properly implemented, and more binding agreements and control mechanisms need to be made. Establishing additional external supervision, where for example the personnel management is reflected, might also give new perspectives and enhance demanding situations. At this point, it has to be considered that there are strong dependencies between Ph.D. students and their supervisors since, in many cases, it is the supervisors who have a major impact on the outcome of a Ph.D. thesis, such as the final grade. It remains challenging how Ph.D. students can criticize the supervising situation without negatively impacting the personal relationship with their supervisors.

Further interventions on the level of universities and research institutions might include support in bureaucratic processes and providing more information on different contact points (e.g., for mental health services). It is obvious that the aforementioned interventions (such as capacity building courses for Ph.D. students and supervisors) are dependent on the support of the central facilities of the research institution. Furthermore, highlighting the high prevalence of mental health problems, for example, at mandatory introductory sessions for Ph.D. students, might help to raise awareness about this topic. This could help unexperienced young researchers to notice signs of anxiety and depression early on before these mental disorders manifest. Finally, public events on this topic could reduce the stigma associated with it, making it easier for affected Ph.D. students to seek help. Such events might also be used to remind the students that it is important to take care not only of their physical but also mental health, for instance, by strengthening social relationships and pursuing hobbies which are not work-related.

Lastly, there are also changes in the political setting and academic culture needed. This includes a fair payment system, reasonable control of contract lengths and extensions, and more perspectives for long-term positions in academia. Considering that the vast majority of Ph.D. students will end up in positions outside of academia, it could be beneficial to better prepare students for careers in alternative job markets, such as industry. Such interventions might directly influence the job insecurity and job dissatisfaction of Ph.D. students. In Germany, the current regulations for temporary academic employment are being evaluated [53], but even propositions from the conference of university rectors [54] seem not to be sufficient for fundamental changes. These changes would also need a shift in the academic culture [55], in which “publish or perish” is still a guiding theme leading to high pressure to perform. Working on a cultural shift is a task for all scientists. This will lead to a more sustainable work culture from which all stakeholders might benefit.

All in all, there is an interplay and dependence of all mentioned levels. Importantly, most problems mentioned in the survey can result from shortcomings on multiple levels, and therefore interventions on more than one level are needed for a satisfying solution. For example, changes to improve the mental health situation on an individual level can be dependent on the consent of the supervisor and can also be negatively impacted by already existing mental health issues. In addition to individual responsibility for health, it is important to systematically target prevention and change the system on the aforementioned levels so that Ph.D. students are better and more quickly supported when mental health problems arise.