Exploring the world through the lens of science inspires curiosity and innovation, but it also presents a distinct set of challenges - challenges that demand persistence, critical thinking, adaptability, and resilience in the face of setbacks and rejection. These challenges aren’t merely anecdotal; they are reflected in the data. In fact, 60% of students starting in science, technology, engineering, and mathematic disciplines (STEM¹) and pre-med will change their major, at a rate 2 times higher than other disciplines (Frank, 2012; Drew, 2011). Grades for the same or comparable individuals often vary between STEM and non-STEM courses, with STEM courses typically showing lower averages (Coe, 2008).

But science is also challenging for other reasons. After surviving the undergraduate STEM attrition effect, many of aspiring scientists face mental health crises in the field caused by a myriad of academic stressors. These include imposter fears (vs. imposter syndrome, as the term syndrome seems to denote something is wrong with the individuals or that it is abnormal), isolation, constant looming deadlines, navigating complex relationships with advisers, intense competition, lack of work-life balance, uncertain job prospects, burnout, and bullying and harassment. For example, a survey of 4,300 academic scientists worldwide reported that at least half struggled with self-reported depression and anxiety, and 67% reported witnessing bullying or harassment, with 43% directly experiencing either bullying or harassment (Abbott, 2020). This effect is most pronounced among graduate students and trainees who are at 6 times greater risk for depression and anxiety than the general population (Evans et al., 2018). In Nature’s survey of doctoral students and their experiences (Woolston, 2019), 36% respondents reported seeking help for depression and anxiety that stem from their training. Almost half (45%) said their satisfaction with their PhD trajectory decreased as they progressed in their training. Similarly, Nagy et al. (2019) found that nearly half of the surveyed biomedical students met the criteria for at least one clinical diagnosis—a prevalence significantly higher than that of the general U.S. population and their same-age peers. The study also identified high levels of burnout, which were strongly associated with thoughts of dropping out and concerns about employment prospects.

With COVID-19, these negative effects were exacerbated by increased social isolation, truncated lab time, and financial stressors. In fact, in a survey of medical scientists including medical and graduate students, 23% of respondents considered leaving academia post COVID-19, due to lack of work-life balance (Matulevicius et al., 2021). This effect is even more pronounced among biomedical scientists (Chan et al., 2020). A survey conducted in the Netherlands during the height of the pandemic from March to May 2020 revealed that 47% of PhD trainees were at risk for psychiatric disorder, and approximately 40% experienced severe burnout symptoms (Mattijssen et al., 2020). As Chan et al. (2020) described it “…imagine the mental resilience needed to maintain focus on solving that equally important mystery in oncology, cardiology, neuroscience, or any other field that has been put on temporary hold due to the pandemic.”

It is no surprise that these stressors frequently contribute to burnout among biomedical students (Hish et al., 2019; Plieger et al., 2015), exacerbating the already high attrition rate in PhD programs, and the issue is more pronounced with women and underrepresented minority (URM) trainees, who leave science and academia at a disproportionate rate (Grogan, 2019; Maher et al., 2020). Without practical and effective interventions to address the stresses experienced by biomedical trainees, which exacerbate mental health crises, the field will suffer a brain drain and lose talented future scientists and potential for innovation. Now, more than ever, practical and effective interventions are vitally needed to address the core causes of biomedical trainee stress and to increase resilience among trainee populations.

While many of the findings, both for scientists in general and biomedical trainees in particular, raise alarms and call for immediate intervention to help those in the scientific workforce pipeline, academia’s responses have been largely muted. Inside Higher Ed declared a mental health crisis in graduate education and stated, “it is only with strong and validated interventions that academia will be able to provide help for those who are traveling through the bioscience workforce pipeline” (Flahtery, 2018). However, available interventions are often lacking, especially for biomedical science trainees. In fact, the results from the Graduate Student Depression and Anxiety Survey led by Evans and colleagues to recommend National Institutes of Health’s (NIH) own train-the-trainer model, where faculty, administrative, and support staff are trained by mental health professionals to recognize and respond to the trainee’s needs, be adapted “to help today’s PhDs compete in the ‘vast and ever-changing job market’” (Evans et al., 2018).

Here, we propose that to alleviate the current mental health crisis within graduate education, an effective intervention needs to address the unique challenges of academia, target the sources of depression and anxiety, and increase trainees’ ability to cope with stressors and adversity. One promising approach is to provide tailored training that increases trainees’ resilience. For example, a systematic review of resilience training and interventions (Joyce et al., 2018) revealed a large body of evidence highlighting the benefits of resilience training for mental health and well-being by mitigating the impact of stress and adversity. In a recent study of women who are thriving in undergraduate STEM majors, resilience was identified as a common and integral trait that allowed them to succeed (Thoman et al., 2020). In exploring why trainees withdraw early from their biomedical PhD programs, Maher et al. (2020) found that self-efficacy components related to resiliency played a role (see Diekman et al., 2017 for a broader analysis of reasons why women exit STEM majors).

Resilience is a form of mental and psychological strength that enables a person to adapt and adjust to difficult or stressful situations. According to the American Psychological Association, resilience is “the process of adapting well in the face of adversity, trauma, tragedy, threats, or significant sources of stress—such as family and relationship problems, serious health problems or workplace and financial stressors. It means ‘bouncing back’ from difficult experiences” (American Psychological Association, 2020). Because of their ability to adapt, resilient individuals tend to better regulate their behaviors, have a more optimistic outlook with greater life satisfaction, maintain positive self-views. They are also less likely to be depressed and anxious (Cohn et al., 2009; Fredrickson et al., 2008; Mak et al., 2011). Resilient individuals are also less likely to engage in presenteeism—that is, they remain fully functioning in the workplace rather than working while distracted (Thogersen-Ntoumani et al., 2017)—and they also avoid other self-defeating work behaviors (Seligman et al., 1986; Van Katwyk et al., 2000). Importantly, resilience is malleable and can be learned and nurtured (Kim-Cohen, 2007).

While resilience training is gradually gaining traction in academic settings, it has already been shown to be effective in workplace and military settings (e.g., U.S. Army’s Ready and Resilient Campaign, Reivich et al., 2011). Studies have found that resilience training can increase positive affect and a sense of well-being while reducing symptoms of depression and anxiety (Forbes and Fikretoglu, 2018; Robertson et al., 2015). Additionally, it has been associated with improved work performance and organizational commitment (Youssef and Luthans, 2007). An in-depth systematic review of resilience training interventions in the workplace demonstrated increased personal resilience, mental health, and subjective well-being. Participants also experienced other tangible benefits including increased self-efficacy, optimism, and performance (Robertson et al., 2015). It has been estimated that the cost-savings for resilience intervention training is $1,846 per person over 8-weeks due to the reduction in stress tied to depression or trait anxiety, as well as increased presenteeism of workers (Johnson et al., 2015).

Given the constant stressors and frequent rejection inherent in scientific academic settings, along with the high prevalence of depression and anxiety among graduate trainees, there is a pressing need for resilience training. Such training needs to focus on increasing individuals’ ability to adapt to adversity, effectively “bounce back,” and reduce stress, depression, and anxiety, and other self-defeating work behaviors, with a particular emphasis on the unique challenges faced by biomedical trainees. The objective of the current paper is to investigate whether a novel resilience program tailored for biomedical trainees and scientists can enhance their resilience, and hence, their persistence in science.

All analyses were conducted using SPSS. Prior to addressing study aims, we assessed whether there were observed differences across trainee populations (i.e., NIH intramural trainees and extramural trainees) and BRS round (i.e., BRS1 and BRS2). There were no differences observed with regards to consistent participation, primary and secondary outcomes, and program satisfaction/perceived changes. Thus, data were collapsed across trainee populations and BRS round.

To address aims 1 and 2, we matched pre-and-post surveys and conducted a paired sample t-test to assess pre-program changes in resilience (primary outcome) and secondary outcomes at the post-program assessment. Furthermore, an independent sample t-test was used to explore the impact of consistent and inconsistent attendance on primary and secondary outcomes at the post-program assessment. All analyses include effect size (Cohen’s d) to highlight the magnitude of an effect, offering insights into its practical significance and real-world relevance. We followed the convention of using 0.2 as small, 0.5 as medium, and 0.8 as large effect sizes.

To address aim 3, we conducted a bootstrapped mediational analysis (Preacher and Hayes, 2004) to test indirect effect of consistent and inconsistent attendance on the secondary outcomes mediated by resilience.

To address aim 4, we constructed descriptive statistics for program satisfaction assessments and reported the percent rating the program “good” (=3) to “excellent” (=5). Independent t-tests were conducted to assess differences in program satisfaction and self-perceived changes following program completion between those who attended the program workshops consistently and those who were inconsistent attenders.

Given the attrition of females and underrepresented trainees in science, we further explored how trainee’s gender and race may influence the effects of BRS as an ancillary analysis.

The results show that the BRS can be an intervention program with a moderate to large impact on the trainees. In addition to being highly rated, the program appears to enhance resilience, self-efficacy, and self-awareness. For those who completed both pre-and-post program measures, BRS improved their self-perceived ability to shift and persist, self-efficacy, and self-awareness while decreasing self-reported anxiety, depression, perceived stress, and presenteeism. After the intervention, participants reported that they were better managing stress and conflict, that they found what they learned valuable, and that they had become more resilient.

As expected, we saw greater effects among those who attend more sessions indicating that the program is effective and is especially beneficial for those who consistently attended (i.e., “higher dose”). In the post-program survey only sample, consistent attenders showed increased resilience, ability to persist, self-efficacy, proactive and reflective self-development components of self-awareness, and saw decreases in perceived stress and marginal decrease in depression. We found that the program’s impact of resilience fully mediated its effects on depression, ability to persist, self-efficacy, and self-awareness, and partially mediated perceived stress. This hints at the underlying mechanism that the increase in resilience is driving the corresponding changes on the secondary outcomes. We speculate that perceived stress may only be partially mediated by resilience, because BRS had another important component—a sense of community and that the trainees are not alone. In fact, in our comment section, the most frequently mentioned comments were that the trainees were glad that they are not alone or that their experiences are not unique. Part of the mechanism driving the reduction in perceived stress may not only be from the increase in resilience but also from a sense of relief or comfort that the experience is shared by others (see Neff, 2003). That sense of not being alone may also reduce feelings of shame which may further improve one’s ability to ask for help and work on building effective coping strategies. Furthermore, of those who attended more than half also found the program more valuable and were more likely to recommend to their friend or colleague; however, it should be noted that those who attended less than half still rated the series highly, found it valuable, and likely to recommend it to a friend or a colleague.

Although BRS seems to be beneficial for all trainees taken together, it seems to have been especially beneficial for trainees from racially and ethnically diverse groups. Compared to white trainees, racially and ethnically diverse trainees self-reported and perceived much greater changes in their perception of resiliency, gained important skills that helped them in their work and home, learned to manage stress and conflict better, and reported that they became a better scientist as a result, despite not showing greater changes in resilience measures. We speculate that, similar to the not being alone effect noted above, that racially and ethnically diverse trainees may find comfort in knowing that their experiences and challenges are shared and recognized by others. BRS explicitly address the challenges that marginalized trainees face in science and research community—especially the role of bias, microaggression, stereotype threat, and how it can lead to attributional ambiguity (Crocker et al., 1991). These findings indicate that BRS could be an important and effective tool in retaining diverse trainees in biomedical science.

It appears that BRS can fill a critical void that currently exists in the biomedical training community. It is empowering trainees to effectively deal with the stressors of academics and research by giving them a sense of agency and the best strategies and tools to cope with those stressors, as well as setbacks and other adversities. It is a step toward reducing the mental health crisis among trainees in biomedical sciences, and when scaled up, could provide a large benefit, and help prevent attrition of trainees in science.

The BRS series is a program designed to help trainees struggling with stressors and raise their resilience during a time when the alarm bells started sounding regarding the graduate mental health crisis in science and when the pandemic was exacerbating that effect. The program was an attempt at meeting the crucial needs of the trainees, and helping them alleviate and address common stressors, providing them with coping skills, and ultimately increasing resilience for those pursuing biomedical science. And it appears that the program can be successful at meeting those goals and is addressing the needs of the trainees and improving their well-being.

However, some may question if the beneficial changes in this intervention program were primarily driven by the changes in the pandemic-stressors and passage of time. The BRS1 started before the COVID-19 vaccine was available, at the height of the pandemic when trainees were grappling with uncertainty and isolation and concluded during a potentially more hopeful period of as vaccines started becoming widely available. While this alternative explanation cannot be ruled out, if this were the case, BRS2 should not show the same results. BRS2 started as the vaccines were becoming widely available and pandemic-related pressures were easing. Yet we did not see any differences between BRS1 and BRS2. This consistency between two distinct cohorts suggests that the observed improvements are unlikely to be solely attributable to the effects of the pandemic (i.e., effects of history or regression to the mean) and that it is more likely that the intervention is targeting the needs of the trainees.

However, it is important to acknowledge the limitations of COVID-related confounds in this study. Although we included measures of perceived stress, we did not collect COVID-specific stress data, which limits our ability to disentangle the effects of pandemic-related stressors from the intervention’s impact. Additionally, the lack of a control group prevents us from definitively attributing the observed increases in resilience to the intervention alone. Regression to the mean, in particular, is a plausible explanation given that the program began at a time when baseline resilience may have been unusually low due to pandemic-related challenges. However, the consistent results observed in two separate cohorts of BRS starting at different times during the pandemic with varying stress levels, significantly reduce the likelihood that regression to the mean alone accounts for the findings. Such replication across the two distinct groups suggests that the observed improvements are not random fluctuations or statistical artifact but rather that the intervention is contributing meaningfully to the observed improvements.

These limitations underscore the importance of interpreting the results with caution. While the consistency of findings across BRS1 and BRS2 strengthens the case for the intervention’s effectiveness, future studies should incorporate a control group and include COVID-specific stress measures or other relevant contextual factors to better isolate the program’s impact and to rule out confounds, especially regression to the mean. Such methodological improvements will help clarify the unique contribution of the intervention to trainees’ resilience.

Another alternative explanation for the result could be that the participants who were consistent attenders were inherently different at baseline. In order to rule out this explanation, we conducted pre-program measure differences between consistent vs. inconsistent attenders on all primary and secondary measures. We did not find any significant pre-measure differences (all ps > 0.2). Therefore, it appears that both consistent and inconsistent attenders were equivalent on our primary and secondary measures at the baseline (at the beginning of the program) and the program attendance or dosage effect is seemingly driving the effect. On the other hand, we cannot say for sure that those who attended all sessions and those who missed one or more were identical in all possible ways. Given the severity of the pandemic, including widespread isolation, salient racial injustice issues, and concerns for various other mental health issues of the trainees at that time, a decision was made to make the program open for all trainees who wanted to participate. Hence, a critical decision was made to forgo a control comparison condition (i.e., waitlist control) and, therefore, drop the quasi-experimental design to allow for broad and open participation. A more ethical decision was made at the expense of establishing the causal impact of the intervention program, and we are unable to rule out some confounding variables which underscores the need for cautious interpretation. Thus, acknowledging this weakness, future studies might compare the efficacy of this intervention with that of other interventions or introduce a waitlist control.

While we utilized a well-known and validated survey instruments for this study, we had to make compromises in the interest of survey length and administration time. We deliberately selected only a subset of questions that were face valid on the instruments and calculated means accordingly. This approach may have impacted the psychometric properties of the measures, as they were originally designed to be evaluated as a whole. While the subset of questions retained their essential reliability and validity, the overall robustness of the measures may have been affected. This limitation underscores the need for caution when interpreting the results, especially in cases where a more comprehensive evaluation of psychometric properties is required (i.e., clinical anxiety vs. non-physical component anxiety measured in this study). However, these limitations were necessary to balance practicality and participation in our study, and we are confident that they do not detract from the overall effectiveness of the program.

Although the program is largely successful, there was significant attrition. In both rounds of BRS, we started with approximately 600 trainees, and by the end of the series, we had approximately 250 trainees in attendance. We speculate that there are two reasons for this. First is the time commitment and pressure from others to work while in the lab during the workday. As the attrition occurred, small group facilitators informally reached out to a subset of trainees, querying why they stopped attending. Many trainees mentioned that the time commitment interfered with lab work commitments, and they chose to prioritize lab work. Some said their PI/supervisor did not support them attending a webinar during the day when they should be focused on lab work. It is essential to acknowledge that the impact of this time commitment and the pressure to prioritize lab work may introduce a selection bias in our study. It is possible that our sample may not fully represent trainees who are heavily or overcommitted to lab work, those who prioritize it over self-care/improvement, or those who lack support for their well-being and resilience from their PIs/supervisors. However, we contend that this underscores the potential benefits of mandating resilience training as a vital component of the curriculum. This step is crucial as we strive to foster a cultural shift in the field of science. Many trainees were hesitant to commit 3–6 h per month to improving their resilience because of fears that participation would interfere with their lab work. We need to change the culture of long hours and complete dedication to lab work to also emphasizing the importance of self-care and self-improvement. After all, one must be well to do well.

The second possibility for attrition is motivation. Although many trainees were excited to start, they may have lacked the motivation and commitment to complete the entire series. Hence, it is possible that only the motivated trainees completed the program and benefited from it. Nevertheless, this explanation for attrition does not undermine the effectiveness of the BRS program. The social psychological literature on behavior and attitude change state that one needs to be willing (i.e., motivated) and able first before any actual changes can occur (e.g., theory of planned behavior; Ajzen, 1991). The effectiveness of BRS is likely driven by providing the tools and the skills the trainees need to implement and make the changes. Those who are willing and able are more likely to practice those skills and see corresponding changes in their resilience. Our data hint that this could be the case—those who attend more than half of the sessions (likely more motivated and committed) report more benefits than those who attend less than half. Furthermore, our pre-and-post sample group demonstrated the greatest positive changes in all secondary outcomes, and they were also the group who attended the majority the BRS sessions. In fact, while we saw changes in anxiety, work presenteeism, and the ability to shift during stressful times in our pre-and-post matched-sample, we did not find the corresponding changes between consistent and inconsistent attenders on post-program only sample. It is possible that these variables require consistent effort and motivation to improve. Hence, it is unlikely that any intervention program, no matter how effective, will have a large impact on those who are unmotivated and unwilling to change.

The results of the BRS align with existing research on strategies to prevent burnout among science trainees. For instance, Prendergast et al. (2024) demonstrated that a reflection-based intervention effectively reduced burnout in second-year medical students. Similarly, a survey by Hish et al. (2020) found that social engagement was rated as a highly effective method for alleviating burnout symptoms, while earlier work by Hish et al. (2019) highlighted the mediating role of mastery and advisor support in the relationship between stress and burnout. These findings raise an important question: Did BRS primarily reduce burnout? It is plausible that BRS contributed to mitigating burnout, as it incorporates strategies commonly found in burnout interventions, such as increasing self-reflection, mindfulness, social connection, and self-efficacy. However, while burnout and resilience are related concepts, they remain distinct constructs with unique definitions, characteristics, and implications. Burnout is characterized as a state of chronic, unmanaged work-related stress (see World Health Organization, 2019). In contrast, resilience refers to the capacity to adapt and thrive in the face of adversity. Importantly, resilience can act as a protective factor not only against burnout but also against other challenges, such as anxiety, depression, and stress and increases the ability to shift, persist, and self-awareness—all of which were assessed in our study. Since we did not directly measure burnout but rather examined some of its symptoms, it would be inappropriate to draw definitive conclusions about the program’s impact on burnout. Instead, our findings suggest that BRS may increase resilience, which in turn could help address multiple factors associated with burnout. This distinction underscores the broader applicability of resilience-based interventions beyond addressing burnout alone.

The BRS has shown its effectiveness as an intervention, and it could be a valuable tool for trainees as they navigate the unique challenges of academic scientific settings. Furthermore, the program is readily available to a wide range of trainees since it is free of charge and easy to participate in. Trainees have the option of joining live sessions twice a year with small discussion groups via Zoom, or they can watch recordings at any time at their convenience. The program’s affordability and accessibility are particularly advantageous to trainees who are constrained by financial and scheduling limitations. Given the program’s accessibility, there is little reason why trainees and extramural institutes should not explore the possibility of incorporating it in their training (see Supplementary material on how OITE can aid in BRS adaptation at various extramural institutions).

The BRS program has been shown to be effective in enhancing resilience among trainees, and especially beneficial for individuals from diverse backgrounds. By providing trainees with the tools to manage stress, cope with failure, and maintain a healthy work-life balance, the program has the potential to retain a talented biomedical workforce while nurturing a group of resilient future scientists. Considering the mental health crisis that many biomedical and science trainees face, the BRS program may be an important component of addressing these issues in the sciences more broadly. With its proven effectiveness, accessibility, and potential to improve trainee well-being, the BRS program offers a promising solution to some of the challenges facing the scientific community.