To stay or go? A mixed methods study of psychiatry trainees’ intentions to leave training

Abstract

This mixed methods research study aimed to test a tailored version of the job demands-resources (JD-R) model to uncover what factors contribute to psychiatry trainees’ intentions to leave their training and how. A Web-based survey measured psychiatry trainees’ work conditions, well-being, occupational commitment, and intentions to leave training. The results were analyzed using structural equation modeling featuring validated constructs. Narrative interviews were analyzed using thematic analysis following the tailored JD-R model. Of 159 current London trainees who completed the questionnaire, 22.1% were thinking a lot about leaving training. Trainees with higher job demands, fewer resources, and less ability to detach from their work experienced higher burnout levels. More engaged and less burned-out trainees were more committed to their occupation and less inclined to leave training. The interviews identified that trainees’ decision to leave was not linear and took time to make. Trainees found their work environment challenging and reported reduced well-being and rethinking their career paths. The JD-R model is a useful tool to understand how medical trainees’ job demands and resources need to be balanced to maintain their well-being and, in turn, how this affects their commitment to the occupation and training.

Keywords:

• Burnout
• intentions to leave
• job demands and resources
• professional commitment
• psychiatry

Introduction

A healthy workforce in psychiatry is more important than ever, considering that mental health problems affect a large proportion of the population (McManus et al., 2016) and are expected to peak in the future (Pfefferbaum & North, 2020). Unfortunately, staff shortages in psychiatry worldwide are common (e.g., Royal College of Psychiatrists, 2019), and retention issues among trainees may be a culprit: only two thirds (65.8%) of psychiatry trainees plan to stay in psychiatry (Barras & Harris, 2012). Researching psychiatry trainees’ intentions to leave training and understanding what determines these intentions is therefore a crucial step toward increasing retention and building a sustainable workforce.

We propose the well-researched job demands-resources (JD-R) model (Demerouti et al., 2001) to support such research. This model divides psychosocial work conditions into job demands and resources and proposes that these conditions affect employees’ well-being and outcomes such as staff turnover (Bakker & Demerouti, 2007). The application of this model in health care has shown that job demands cause staff to feel burned out, less engaged (J. N. Scanlan & Still, 2019; Wen et al., 2016), and more inclined to leave their jobs (van der Heijden et al., 2018). For psychiatry training specifically, research highlighted several job demands (e.g., rota gaps) that reduced their satisfaction with training, and particularly morale and well-being (Barras & Harris, 2012; Ryland et al., 2020). Lai and Plakiotis (2018) found that resources (e.g., supportive supervision) are equally important to retain psychiatrists. Based on the JD-R model and the aforementioned research, our main hypothesis is that trainees working in more demanding environments with fewer resources will be less engaged and experience higher burnout levels and, in turn, will more likely consider leaving training (Hypothesis 1).

An advantage of the JD-R model is that it is an open heuristic model, meaning that the resources, demands, and outcomes represented in the model can be tailored to the study context (Schaufeli & Taris, 2014). We used the model’s flexibility by selecting those job resources and demands that were most relevant to psychiatry and by adding two additional concepts: occupational commitment and psychological detachment (see Figure 1). Both concepts have previously been used in the JD-R model (Collie et al., 2018; Dicke et al., 2018), albeit in different study contexts and not in combination.

Figure 1. Presentation of the tailored JD-R model that will be tested quantitatively (Aim 1) and qualitatively (Aim 2). Note. Descriptions of constructs are presented in Table 1.

Trainees’ occupational commitment refers to their affective reaction to their training (Lee et al., 2000). This commitment seems to be extremely important during medical training as junior doctors are forming their identity as specialists in a certain area. This affects their career decisions: if a psychiatry trainee does not feel committed to becoming a consultant, they might consider changing their career path. There is limited evidence available that occupational commitment links to outcomes such as turnover intentions (Arslan Yürümezoğlu et al., 2019), but studies have shown that reduced burnout and greater engagement increase occupational commitment (Dicke et al., 2018; Lee et al., 2000). Therefore, we hypothesize that occupational commitment will act as a mediator between trainees’ well-being and their intentions to leave training (Hypothesis 2).

Furthermore, we hypothesize that trainees’ psychological detachment from work will mediate the effect of work demands on well-being (Hypothesis 3). Researchers argue that detachment from work can positively affect well-being because it helps to restore lost or threatened resources during a working day (Sonnentag & Fritz, 2007; Steed et al., 2021). In particular, the inability to mentally switch off has a considerable effect on employees’ well-being (Bennett et al., 2018) and is a core mechanism by which job stressors transform into poor well-being (Sonnentag & Fritz, 2015). Considering the emotional effect of work on psychiatry trainees (Qayyum et al., 2021), detachment from work might be challenging and therefore is an important aspect to investigate.

Quantitative testing of the tailored JD-R model (Figure 1) is the focus of this article. It aims to explore what factors contribute to psychiatry trainees’ intentions to leave training and through what mechanisms (Aim 1). For concurrent validity, we also performed a qualitative investigation to determine whether we could recognize the factors and relationships analyzed in the JD-R model in trainees’ narrative stories about their training journeys (Aim 2). If so, we wanted to investigate qualitatively how these factors were experienced and explore how trainees make complex decisions to leave training (Aim 3). The addition of this qualitative aspect of the study provides a novel, in-depth understanding of the applicability of the theoretical framework and the complexity of decisions to leave that would not be possible to capture through quantitative methods alone.

Methods

Sample and recruitment

Psychiatry training in the UK consists of 3 years of core training and at least 3 years of specialty training. We invited current core and specialty psychiatry trainees in London (via emails from the London deanery and social media), as well as any trainees who had been in the London training program but had paused or left (via social media and snowball sampling), to complete a web-based survey and/or take part in an interview between July and November 2019. Participants who completed the survey were also invited to participate in an interview. Those wishing to participate in interviews completed a sociodemographic questionnaire, which was used by the researchers to recruit a sample with varied characteristics. The aim was to recruit trainees with a wide variety of views with different demographic characteristics and from different stages of training (core trainees, specialty trainees, and trainees who had left or were planning to leave training). This was done as part of a wider project (Medisauskaite et al., 2020).

Ethics approval

The project received ethics permission from the University College London (UCL) Research Ethics Committee (Reference No. 10121/001). All survey and interview participants provided written consent to participate in the study. At the data analysis stage, survey participants’ consent forms were separated from the survey answers. The project also received Data Protection approval (Reference No. Z6364106/2019/06/28) and was performed in compliance with the General Data Protection Regulation 2018. The data were stored securely on the UCL S drive, which is only accessible via password by UCL research staff directly involved in this study.

Survey

The survey consisted of questions mapping participants’ sociodemographic characteristics and validated questionnaires scored using Likert scales (see Table 1), which were chosen to investigate the JD-R model, as presented in Figure 1. The most relevant job demands and resources to the research context (psychiatry training) were selected through in-depth discussions with Royal College of Psychiatrists members with different levels of involvement in psychiatry training and College work. These members represented various stakeholders in psychiatry training (i.e., specialty registrar/psychiatric trainees’ committee member, consultant/head of School of Psychiatry, senior member of the College). The unstructured discussion was led by the researchers (A.M., A.R., L.K.) asking about the aspects of psychiatry training that might influence trainees’ satisfaction with their training and career decisions. Amendments to the questionnaire were then made based on the suggestions from a few psychiatry trainees who took part in a pilot study to provide us with their feedback on the most appropriate work environment to measure and on the presentation of the questionnaire items.

Table 1. Description of measures.

Construct	Content	Origin	Items	Likert scale	Cronbach’s alpha	Mean (SD)
Sociodemographic characteristics	Personal (e.g., gender, age); training context (e.g., work pattern)	Self-developed	9	—	—	—
Intentions to leave psychiatry	Thinking about leaving occupation and/or looking for another job	Intentions to leave (Cohen, 1998)	3	5-point (strongly disagree–strongly agree)	.9	1.86 (0.95)
Affective occupational commitment	Positive emotional attachment to the profession	Three-component Model of Commitment Subscale (Meyer et al., 1993)	6	7-point (strongly disagree–strongly agree)	.8	5.81 (0.89)
Work engagement^a	Vigor (mental resilience), absorption (happily engrossed in one’s work), dedication (proud of one’s work)	Utrecht Work Engagement scale (Schaufeli & Bakker, 2003)	9	7-point (never–always/every day)	.9	3.54 (0.86)
Burnout^b	Exhaustion (loss of energy), mental distancing (cynicism), emotional impairment (feeling overwhelmed), cognitive impairment (attention/concentration deficits)	Burnout Assessment Tool (Schaufeli et al., 2019)	23	5-point (never–always)	.9	2.14 (0.45)
Job demands^b	Emotional demands (e.g., high emotional labor, dealing with people whose problems touch one emotionally)	Demand-Induced Strain Compensation Questionnaire Short Version Subscale (Jonge et al., 2009)	6	5-point (never or very rarely–very often or always)	.8	3.38 (0.73)
	Job effort (time pressure, disruptions)	Effort–Reward Imbalance Questionnaire Short Version Subscale (Siegrist et al., 2014)	3	4-point (strongly disagree–strongly agree)	.8	2.67 (0.60)
	Aggression from patients (e.g., violence, suicide)	Selected items from the Perception of Prevalence of Aggression questionnaire (Oud & Adm, 2000)	5	5-point (never–frequently)	.7	2.25 (0.69)
Job resources^b	Support from the supervisor (e.g., advice, feedback)	Social Support Questionnaire (O’Driscoll et al., 2004)	4	5-point (never–all the time)	.9	3.81 (0.99)
	Autonomy (e.g., choosing how work is carried out)	Autonomy and control (Haynes et al., 1999)	6	5-point (not at all–a great deal)	.9	3.21 (0.84)
	Task significance (e.g., work influences the lives of others)	The Work Design Questionnaire Subscale (Morgeson & Humphrey, 2006)	4	5-point (strongly disagree–strongly agree)	.8	3.80 (0.62)
Psychological detachment from work	Mental disengagement from work, not thinking about work after work hours	Recovery Experience Measure Subscale (Sonnentag & Fritz, 2007)	4	5-point (do not agree at all–fully agree)	.8	2.94 (0.81)

Notes. Satisfactory internal consistency (Cronbach’s alphas >.7).

^a Subconstructs were combined into one overarching score because subconstructs were part of one measurement scoring the same overall construct (approach supported by authors who developed these questionnaires).

^b Subconstructs were assessed as individual subconstructs because each subconstruct was measured using a different tool.

Interviews

A narrative style of interviewing was chosen to allow trainees to tell their stories of training with little interruption. The interviews started with a question asking participants to think back to when they decided to specialize in psychiatry and just talk through their experiences from then to the time of the interview. The interviewer occasionally asked broad, open-ended questions and provided prompts to obtain more detail or guide the conversation, particularly encouraging seminal and impactful moments in trainees’ experiences (including the decision to leave training) to be discussed in depth (Connelly & Clandinin, 1990). Interviews were conducted using a semistructured interview guide either in person or by phone by A.R., K.A., and L.K.

Data analysis

Quantitative

Descriptions of missing values, construction of scales, and the correlation matrix are presented in Supplementary Material 1. Data were normally distributed (skewness and kurtosis between 2 and −2). Linear multiple regressions were used to explore relationships between constructs and informed the subsequently specified structural equation model (SEM). SEM was chosen to enable the testing of all chosen factors and relationships in one comprehensive statistical model. The model included two latent variables: job demands (indicated by the subconstructs of emotional demands, job effort, and aggression from patients) and job resources (indicated by support from the supervisor, autonomy, and task significance). The purpose of the model was to estimate regression coefficients between the constructs included in the model. The model tested whether/how job demands link to participants’ burnout (directly and through psychological detachment) and whether/how job resources link to burnout and engagement and, in turn, whether/how burnout and well-being link to intentions to leave directly and through affective occupational commitment. The Sobel test was used to explore the mediating relationship in more detail.

SPSS v26 (IBM Corp., 2019) was used for all analyses except for the Sobel test and SEM, which were conducted using the R Project for Statistical Computing v3.6.2 (R Core Team, 2021) (package: Lavaan). The SEM showed an acceptable to good fit for the data (Comparative Fit Index = 0.934; Tucker-Lewis Index = 0.905; standardized root mean square residual = 0.069; Root Mean Square Error of Approximation = 0.068; Brown, 2015).

Qualitative

Interview audio recordings from the wider study (n = 28) were transcribed verbatim (median length = 59 min). Because we were interested in those trainees who explicitly discussed their intentions to leave training, for our analysis we identified participants who had made plans to leave training or intended to leave as soon as possible, were out of training with no plans to return, or had decided to leave and subsequently returned (n = 9).

Transcripts were coded deductively (NVivo 12) (QSR International Pty Ltd., 2018) via thematic analysis (Crabtree & Miller, 1992), using the JD-R model as a coding framework. The two objectives of this study were as follows:

1. Explore whether the factors presented in the JD-R model held when examining nuanced, qualitative understandings of trainees’ training experiences and how these were experienced by the trainees.

2. Because the quantitative JD-R model suggests a linear trajectory toward leaving, with a set order, examine how this would play out in trainees’ narratives and whether this trajectory could be identified.

The first three transcripts (33%) were coded concurrently by K.A., A.M., and M.S. The team reviewed the themes and critically discussed which factors/relationships from the JD-R model were recognized. Differences in coding were discussed and agreement was reached. The remaining transcripts were then coded independently using the same framework, with researchers meeting to discuss any differences and reach a consensus.

Although thematic analysis attempts to avoid removing coded items from their context, inevitably the act of grouping text into themes can mean that these become disconnected from their overarching narrative. Each transcript was thus then reread with the codes included to holistically understand how the JD-R constructs were experienced by each individual throughout their training, in which order, and in what connection to one another.

In the Results section, two participant stories as told to the interviewer are presented, with the JD-R constructs indicated. We chose this method to present the data because a narrative presentation will give the reader an appreciation of the granularity of each participant’s experience over time and elucidate how the JD-R model constructs interacted across the trainees’ changing training placements and situations. These two participant stories were selected for presentation as a useful illustration, because they focus on the same key themes but illustrate these interacting in distinct ways for each individual. Following this, a summary of findings is presented, providing an overview of the extent to which the two objectives described above were met, through an analysis of all nine transcripts.

Results

Demographic characteristics

A total of 159 trainees took part in the survey: 62.9% were female, 64.8% were White, and 79.9% were UK graduates (Table 2). A total of 22.1% of trainees agreed or strongly agreed that they were thinking a lot about leaving the profession. The sociodemographic characteristics of the trainees who took part in interviews are also presented in Table 2.

Table 2. Sociodemographic characteristics of the study sample.

Variables	Quantitative, % (N)	Qualitative, % (N)
Gender (female)	62.9 (100)^a	88.9 (8)
Ethnicity
White	64.8 (103)^b	66.6 (6)^c
Asian/Asian British	22 (35)	22.2 (2)
Black/African/Caribbean/Black British	3.1 (5)	0 (0)
Arab/Arab British	2.5 (4)	0 (0)
Mixed/Multiple ethic groups	5 (8)	0 (0)
Dependents (no)	69.8 (111)^d	66.7 (6)
Region of primary medical qualification (UK)	79.9 (127)^e	66.7 (6)^d
Training level^f
Core	56.6 (90)	66.6 (6)^g
Specialty	41.5 (66)	33.3 (3)
Training region: London
South West	10.7 (17)	n/a
South East	36.5 (58)	n/a
North East	15.7 (25)	n/a
North West	12.6 (20)	n/a
North Central	24.5 (39)	n/a
Work pattern (full-time)^e	82.4 (131)	n/a
Trainees’ intentions to leave training (agree/strongly agree)
Thinking a lot about leaving	22.1 (35)	n/a
Thinking to leave as soon as possible	4.4 (7)	n/a
Actively looking for a job outside psychiatry	8.2 (13)	n/a
Burnout level^h
Low	11.9 (19)	n/a
Average	63.5 (101)	n/a
High	23.9 (38)	n/a
Very high burnout levels	0.6 (1)	n/a

Notes. ^a2.5% (4) missing/other.

^b 1.3% (2) missing.

^c 11.1% (1) missing.

^d 2.5% (4) missing.

^e 0.6% (1) missing.

^f 1.9% (3) missing.

^g These trainees had completed their core training but not progressed to specialty.

^h Cutoff values for burnout were provided by Schaufeli et al. (2019).

Quantitative results

The results from the SEM are presented in Figure 2 and Table 3. Trainees’ intentions to leave were stronger when they were less committed to psychiatry (β = −0.46, p < .001) and were less engaged in their work (β = −0.18, p = .017). There was no significant direct effect of burnout on intentions to leave. Trainees experiencing stronger work engagement (β = 0.47, p < .001) and lower burnout (β = −0.25, p < .001) levels were more committed to their occupation. Furthermore, we tested the mediating effect of occupational commitment on the relationship between burnout and trainees’ intentions to leave as well as on the relationship between work engagement and trainees’ intentions to leave (Supplementary Material 2). For the relationship between burnout and trainees’ intentions to leave there was borderline full mediation (p = .046) by occupational commitment (Z = 5; p < .001). This means that the effect of burnout on trainees’ intentions to leave manifests solely through trainees’ occupational commitment. In other words, trainees who experience more signs of burnout are less committed to their profession and therefore more likely to intend to leave their training. The relationship between work engagement and trainees’ intentions to leave was partially mediated by occupational commitment (Z = −5.2; p < .001). Trainees who are engaged at work and committed to their work are less likely to have intentions to leave training.

Figure 2. Visual representation of the SEM model testing the tailored JD-R theoretical model and the relationships between the constructs. Note. The dotted constructs indicate the latent variables job demands and resource.

Table 3. Regression coefficients and test statistics for the relationships between the constructs used in the structural equation modeling.

Outcome	Predictor	Unstandardized coefficient (b)	Standardized coefficient (β)^a	Z value	p
Intentions to leave	Affective commitment	−0.49	−0.46	−5.9	<.001
	Engagement	−0.30	−0.18	−2.4	.017
	Burnout	0.22	0.10	1.5	.135
Affective commitment	Engagement	0.48	0.47	7.0	<.001
	Burnout	−0.50	−0.25	−3.8	<.001
Engagement	Job resources	1.52	0.62	3.4	.001
Burnout	Job resources	−0.42	−0.33	−2.7	.007
	Job demands	0.34	0.42	3.8	<.001
	Psychological detachment	−0.10	−0.17	−2.4	.017
Psychological detachment	Job demands	−0.61	−0.41	−4.2	<.001

Note. ^aCompletely standardized solution in which both the observed and latent variables are standardized. Bold numbers indicate significant effects.

Trainees were more engaged at work when they had better job resources (β = 0.62, p = .001). Trainees who had fewer job resources (β = −0.33, p = .007), had more job demands (β = 0.42, p < .001), and were less able to psychologically detach from work in their free time (β = −0.17, p = .017) experienced higher levels of burnout. Trainees experiencing high job demands were less able to psychologically detach from work in their free time (β = −0.41, p < .001). We tested the mediating effect of psychological detachment from work for the relationship between each individual job demand and burnout (Supplementary Material 2). Both linear multiple regression and the Sobel test confirmed that there was partial mediation of psychological detachment on the relationship between trainees’ job demands and burnout (emotional demands: Z = 2.9; p = .004; job effort: Z = 2.4; p = .015; aggression: Z = 2.8; p = .006). This means that these three work demands have less of a negative effect on trainees’ burnout levels if trainees are able to psychologically detach from work in their free time. Because psychological detachment is only a partial mediator, there is also a direct effect of trainees’ job demands on burnout; that is, high job demands still have a negative effect on burnout.

Qualitative results

Participant stories

To holistically illustrate the two stories and maintain their integrity (Connelly & Clandinin, 1990), a section of the interview is presented alongside the summary to provide a description in the participant’s own words (Bamberg & Demuth, 2016). Findings have been presented as stories in other medical education literature to good effect (G. M. Scanlan et al., 2018). Participant names are pseudonyms.

Story 1. Joanna (Participant 15) is a female trainee who at the time of the interview was in her first 6 months of specialty training. Upon entering psychiatry, her interest in psychiatry was grounded in the specialty’s holistic approach to illness, the understanding that psychiatry could meaningfully improve patients’ lives (task significance), and a feeling of “fit” with other “like-minded” psychiatrists (occupational commitment).

Conceptualized within the JD-R model, Joanna’s training included substantial job effort (e.g., demands beyond her competence and confidence levels, such as decisions on physical health matters) in an environment with few job resources (e.g., lacking support from supervisors and without wider support from other specialties). As per the model, high demands and lack of key resources led to burnout. However, upon her return to work, support from supervisors and occupational health helped her recover from burnout, by making the work seem manageable and rebuilding her confidence.

In specialty training, however, rota gaps and long waiting lists (job effort) and the strain of working with risk of self-harm and suicide patients (patient aggression) affected Joanna and again led to burnout. In her final placement, Joanna worked with a burned-out supervisor (lacked supervisor support). Joanna thus began to feel that the job was pointless (low task significance) and caused her anxiety and felt bitterness toward the specialty (burnout and reduced commitment), which finally led to an intention to leave (see Box 1).

Story 2. Aisha (Participant 21) is a female specialty trainee who described herself as proud to start training, very passionate about psychiatry, and usually loving the work (engagement).

In her first placement, the sudden death of a patient due to a physical condition that the unit was not equipped to manage caused her to feel angry at what she perceived as “unfair” care for patients. This led to low mood and nightmares (job effort and emotional demands resulting in burnout). Her desire to avoid appearing “not cut out to be a doctor” prevented her from seeking support from her supervisors; instead, she poured her energies into advocating for training and equipment, taking extra shifts and attempting to “block it out” with work (unsuccessful psychological detachment).

She felt that this caught up with her in specialty training, where she was placed in an extremely understaffed location and became the doctor of record for hundreds of patients (job effort). She began to feel disheartened and frustrated and constantly worried that the service did not serve patients (task identity). She sought support from her supervisors, but they dismissed her concerns with positive reassurance, which heighted her perception that she was not coping. She then experienced another incident of patient aggression, where a patient repeatedly followed her outside of work, which this led to high burnout levels (see Box 1).

Aisha thus handed in her notice and took a break but subsequently resumed training for financial reasons. She found the new post much improved (better work conditions), and she reported that her occupational commitment returned (see Box 1). This feeling was boosted by her acceptance that perhaps she did not have to base her identity so predominantly on being a psychiatrist and was actively pursuing other activities away from psychiatry that energized her (psychological detachment). These activities also actually highlighted what she enjoyed about psychiatry.

Thematic analysis

All of the JD-R model factors were identified from the deductive thematic analysis. Autonomy and task significance were mentioned as positives within a “good” work environment; however, it was the supervisory relationship that was pivotal. Supervisor support (or lack of) was crucial in shaping trainees’ decisions regarding their careers. Supervisor opinions of both the trainee (e.g., their competence and suitability for the career) as well as psychiatry (e.g., passion for, or disillusionment with, the specialty) commonly underpinned trainee perceptions of themselves and their career path.

She was full of energy, she was very kind and funny, and compassionate and … she led … she was very involved in what was going on, so she was very, like, on the floor. You felt you could ask her anything, she was very good at educating … and she felt very passionately about the patients and advocating for them and … and the ward rounds she ran were really … like, they were really enjoyable. (Participant 27, female, specialty trainee)

Nothing was explained—nothing. I had four days a week on the ward, half a day in the psychosis clinic, and half a day with this assessment clinic—I never once met the consultant that ran the assessment team in 6 months and never laid eyes on him. (Participant 28, male, core trainee)

The effect of job effort (e.g., workload, understaffing) on trainees’ career decisions was clearly evident, with trainees most commonly citing these as the largest and most constant job demands. There was noticeable effect of a combination of demands. Particularly when heavy job effort was combined with high emotional demands (e.g., empathizing with patients’ distress), trainees reported progressing to burnout if they were unable to take time away to recover or to alter the circumstances they perceived as demanding. Similarly, an act of aggression from a patient (even if more minor than previous assaults) could trigger a trainee under substantial effort/demands to decide to leave the program temporarily or permanently.

It’s put me off ever working there to be honest. I’m not the only one to be assaulted, there are quite a few, and I know the nurses got assaulted quite a lot. (Participant 17, female, core trainee)

The trajectory in the JD-R model showing high job demands and low job resources leading to burnout could be clearly traced in all of the trainees’ narratives. Likewise, if trainees were able to psychologically detach, there was evidence that this did help temper burnout, although for our participants a positive change in the balance between demands and resources was also necessary for them to not progress to disillusionment with training (low affective occupational commitment).

A cyclical pattern between the constructs in the JD-R model was seen multiple times within the same person’s narrative and for a number of trainees. This cycling was especially prominent between burnout, commitment (disillusionment), and turnover intentions. This indicates that the decision to leave training is not linear but rather includes several “cycles” as trainees move between placements and changing work environments. Those who decided to leave had experienced several such cycles before making the decision to leave. Notably, a drop in affective occupational commitment prior to choosing to leave was clearly evident in narratives.

Discussion

This study revealed that 22.1% of psychiatry trainees who responded to our survey were considering leaving the profession. A similar percentage has been reported in other specialties (18% of general practitioners; Owen et al., 2019) or overall (18% of UK doctors in 2021, which increased from 12% in 2019; General Medical Council, 2021). This mixed methods study revealed the importance of job environment and ability to detach in shaping trainees’ well-being and commitment to their occupation and, in turn, their decision to leave training. The fact that trainees in harmful work environments are more inclined to leave the profession might also have significant financial and patient care consequences (Misra-Hebert et al., 2004). Work environments therefore affect individual trainees but also the organization as a whole.

Corroborating findings from other studies (Lesener et al., 2019; Rattrie et al., 2020), our quantitative exploration (Aim 1) revealed that trainees working in more demanding environments with fewer resources were less engaged and experience higher burnout levels. In turn, they were more likely to consider leaving training (confirmed Hypothesis 1). Therefore, recognition of and resources to offset the heavy demands trainees encounter in training are vital. Furthermore, we aimed to explore the concurrent validity of the model and found that factors and relationships from the JD-R model were recognized in trainees’ narrative stories about their training experiences (Aim 2). Trainees particularly highlighted the role of supportive supervisors, and that support might take various forms. Practical support was valued, but supervisors’ own attitudes toward the profession underpinned trainees’ perceptions of themselves and their career path (Aim 3).

Doctors start forming their professional identity early on in their career (Wong & Trollope-Kumar, 2014), and being a doctor becomes a big part of who they are (Henderson et al., 2012). With this in mind and considering the time and emotional investment in becoming a doctor, it is not surprising that not (or no longer) being able to emotionally connect to the profession is an underlying mechanism through which trainees make the decision to leave training. Moreover, this study revealed that affective commitment served as a mechanism through which trainees’ reduced well-being led to intentions to leave training (confirmed Hypothesis 2). This mediating effect of commitment was particularly strong for burnout because it canceled the direct link to intentions to leave. Trainees’ stories also revealed that trainees in a challenging environment experiencing reduced well-being reported doubting whether they should continue with their training, questioning whether they are “cut out” to be psychiatrists (Aim 3).

The third hypothesis that psychological detachment from work protects trainees from burnout was also confirmed. In interviews, trainees shared that not being able to detach from work affected them in the short and/or longer term and that pursuing other activities had a positive effect on their well-being and views of their role as psychiatrists (Aim 3). A training program allowing more flexibility for detachment and recovery should be promoted to allow trainees the space and time essential to self-manage their well-being. Our study confirms other researchers’ conclusions that detachment from work can prevent the negative effect of job challenges (Sonnentag & Fritz, 2015) and showed that this conclusion is relevant in psychiatry training.

Finally, this study highlights how the process of deciding whether to leave training is an extremely complex one, influenced by positive and negative factors, and is a cyclical process that occurs over time. The qualitative data showed that participants did not move linearly from job challenges to intentions to leave; instead, their attitudes toward and experiences of their job characteristics, well-being, commitment, and intentions to leave fluctuated within and between placements and contexts: job demands wore them down and resources rebuilt them in an iterative cycle (Aim 3). This was particularly evident for the links between well-being, commitment, and intentions to leave. Again, a strong identity of being a doctor might explain the challenge in making this difficult decision.

Strengths/limitations and future research

Recent studies have applied the JD-R model to explore factors that link to burnout among medical students and junior doctors (Hariharan & Griffin, 2019). This mixed methods study goes one step further by investigating turnover intentions, tailoring the JD-R model to medical training by adding two additional factors: occupational commitment and psychological detachment. We addressed the complexity of turnover intentions in medical training by investigating a multitude of factors and mechanisms from two methodological perspectives. The fact that the widely tested and validated JD-R model was tested quantitatively and then found applicable to trainees’ stories enhanced the concurrent validity of the tailored model in the setting of medical training. We therefore advocate for the use of this model in other research looking at similar retention questions, even outside medical training, to explore its relevance and applicability.

One of the limitations was our inability to calculate a response rate for the survey because recruitment was carried out by a third party and social media and collected in Greater London only. We expect to find similar factors and mechanisms in studies with a larger sample size, especially because the constructs used have been previously validated. Future research should consider including a broader sample of trainees and increasing the sample size and explore intentions to leave longitudinally.

In addition to the suggested further explorations of trainees’ intentions to leave and well-being, future research should focus on appropriate interventions. Our study shows that the work environment is key to well-being and career intentions. Therefore, exploring which organizational-level interventions would effectively address these issues is crucial to improve the workplace and staff experiences.

Concerning the qualitative aspect, only a small number of interviewees had left or made plans to leave training, and only one of these participants was male. Although clear themes and commonalities were found, a larger group would have provided a wider range of experiences. It would be fruitful for future research to conduct and analyze narrative trainees’ exit interviews to appreciate the cumulative effect of trainees’ experiences as they move between very different placements.

Data accessibility statement

The data generated and analyzed during the current study are not available, because consent has not been granted by the participants.

Box 1. Interview excerpts.

STORY 1:

Joanna: And so you felt so impotent and unable to do anything for your patients. And yet again you were going to get blamed, and you felt so unable to help anyone and so dissolute with what psychiatry could offer that it was the most pointless speciality in the world because, you know, we only had the medication and it didn’t work, and people were saying, “Help me,” and you were like, “I can’t, I don’t know how”—you know, that’s how you felt. People were begging for help and you couldn’t—hadn’t got anything you could do to help them. And that is an awful position to be in.

Interviewer: Yes, it really is. So how did you cope with that post then and with that situation?

Joanna: It made me really anxious. It made me bitter and angry at psychiatry. […] There was no one big incident that suddenly meant I couldn’t carry on, it was just that inability to just keep doing this because it’s making me more and more anxious. But it was very clear—again, there was no decision—it was very clear that you couldn’t carry on doing this, you had to leave.

STORY 2:

Aisha describes how, at a point when her burnout levels were high, this incident affected her substantially.

Aisha: When that happened before [in core training] with the sexual assault, with the patient who burnt me with a lighter, with the patient who had hit me and broken my bleep […], not once had I ever thought about leaving, it had never occurred to me. Oh, I was a little bit battered and bruised, but the next day you just got up and got on with it, and I loved my work. And I think it was only when this incident happened […] I did not have the resilience anymore to just try to put it in perspective and get on with things.

During new post:

Aisha: So I’m probably, honestly speaking, about three quarters of the way back to feeling like I will get on with psychiatry and I will be a consultant and get on with, and enjoy it as much as I used to. But I am more—I’m not as afraid as I was before to say that actually that might not happen, cos I did see that as a massive failure if I did not become a consultant after all the [inaudible] and time and effort it’s taken to get through [medical school and training]. So I’m also proactively doing other things which are not related to psychiatry […], and in the most trickiest times that I’ve had I’ve always come back to [the] side of things [that I enjoy most] and thought, well, actually there is that […], and that is something that definitely keeps me going.

Acknowledgements

We thank the trainees who took part in this research. We also thank Dr. Kate Lovett and Dr. Vivienne Curtis for their valuable contributions in discussing psychiatry training processes and research findings.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the Royal College of Psychiatry and Health Education England (Award No. 179756).