Factors Associated With Burnout and Stress in Trainee Physicians

Key Points Question What factors are associated with burnout/stress in trainee physicians? Findings This systematic review and meta-analysis of 48 studies included 36 266 trainee physicians. The odds ratios for the associations between workplace factors and burnout/stress were found to be higher compared with nonmodifiable and non–work-related factors such as age and grade. Meaning The findings of this study highlight the importance of improving training and work environments to possibly prevent burnout among trainee physicians and suggest that implementing multicomponent interventions to target major stressors uncovered in this study could be promising.


Culture and upbringing
no overall n/a n/a n/a n/a n/a n/a n/a

Obstetrics and gynaecology
2.12 (0.8, 5.64) 48.7%, p=0.14 n/a n/a n/a n/a n/a n/a Internal Medicine 1.2 (0.67, 2.14) 73.5%, p=0.002 1.20 (0.67, 2.14) 73.5%, p=0.002 n/a n/a n/a n/a Paediatrics 1.07 (0.7, 1.65) 16.5%, p= 0.30 n/a n/a n/a n/a n/a n/a Psychiatry 1.41 (1.1, 1.8) 22.8%, p=0.27 n/a n/a n/a n/a n/a n/a Surgery 1.46 (0.86, 2.49) 76.1%, p<0001 1.13 (0.77, 1.66) 53.7%, p=0.071 n/a n/a n/a n/a eFigure 1. forest plots of association between burnout and different factors burnout, and that trainee doctors are a particularly high-risk group. (2)(3)(4)(5). Burnout consists of three components: emotional exhaustion, reduced sense of personal achievement as well as depersonalisation (6), and one of its main contributors is prolonged occupational stress (7). High levels of burnout and stress have been found in trainee doctors working in the US and other countries such as Australia and Canada (8)(9)(10)(11)(12). This was also mirrored in a recent national survey of 51956 trainee doctors working in the United Kingdom (UK) has found that nearly a quarter of trainee doctors were experiencing burnout to a concerning degree. Another survey conducted by Health Education England found that 50% of trainees were experiencing symptoms suggestive of burnout and 80% of trainees felt their job caused excessive stress (13). Burnout has negative impact on the personal wellbeing, the family and professional relationships and the career prospects of trainee doctors and may jeopardise patient care. The wellbeing of trainee doctors is a key benchmark for the efficiency and sustainability of healthcare systems in this and the following decades (14,15). It is to our knowledge that no previous systematic reviews have been identified in the scientific literature and we did not identify any similar systematic reviews on the PROSPERO database. Hence, we propose to conduct a systematic review and meta-analysis that aims to identify occupational and non-occupational stressors that are associated with burnout/distress in trainee doctors.
We will use methods reported in line with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (16) as well as Meta-analysis Of Observational Studies in Epidemiology (MOOSE) (17). We will also present the results for this review in line with the PRISMA and MOOSE guidelines.

Search Strategy
We plan to search the following databases: • Cochrane Database of Systematic reviews We will search for eligible papers from inception until July 2018. The search strategy will include the following combinations of three key blocks of terms: • Stress/ burnout (determinant* or factor* or driver* or caus* or contributor* or stressor* or predictor* or predispos* or correlat* or associat* or risk*)

8 and 14 and 15 eTable 3. Search strategy
Database searches will be supplemented by hand searches of reference lists of included papers.

Eligibility Criteria
Studies were eligible for inclusion if they met the following criteria: • Population: Qualified doctors who are engaged in postgraduate training (i.e. trainee doctors). Studies that are based on a mix of trainee doctors and other doctors or health professionals were included if trainee doctors comprised of at least 70% of the sample.
• Stressors: Stressors can lead to stress which in turn can lead to distress and burnout. Stressors will include occupational and non-occupational contributors of stress such as work demands, specialty, work environment and demographics.
• Outcome: The main outcomes will be associations between stressors (occupational and non-occupational contributors of stress) and negative outcomes of stress such as burnout/distress. We will include both burnout measured with validated measures such as the Maslach Burnout Inventory as well as measures of distress because both are known negative outcomes of stress (15). A pooled analysis of burnout and distress will be undertaken. A separate analysis will be undertaken on burnout to examine burnout as an outcome of prolonged stress.
• Design: Quantitative research design such as observational studies including retrospective or prospective cohort and cross-sectional as well as case control studies.
• Context: Any healthcare setting including primary and secondary care.

Exclusion Criteria
• Studies not explicitly focusing on stress such as studies that explore the determinants of psychiatric conditions with specific diagnostic criterion e.g. depression and generalised anxiety disorder.
• Grey literature, conference abstracts and letters to the editor and studies not published in a peer-reviewed journal were excluded.

Study Selection
Searches will be exported onto Endnote and duplicate will be removed. Study selection will be completed in two stages. Firstly, the titles and abstracts of the identified studies will be screened and subsequently the full-texts of relevant studies were accessed and further screened against the eligibility criteria. Both MP and AZ will be involved in the screening. Disagreements will be resolved through discussions.

Data extraction
Data extraction will be done in an excel form and this will be initially piloted on 5 randomly selected studies. The following descriptive data will be extracted:

Methodological quality of the studies
We will use the Newcastle Ottawa scale to critically appraise the quality of the included studies (18) and will use the adapted version to undertake critical appraisal assessments of cross-sectional studies (19). This modified Newcastle Ottawa instrument shown below ( Fig. 1) provides scores from 0 to 10 with studies scoring ≥6 classified as high quality.

Data Analysis
The primary outcome of this review will be the association of stressors with burnout/distress in trainee doctors. We will calculate odds ratios (ORs) together with the 95% confidence intervals from each study using Comprehensive Meta-analysis (CMA) software (20). The pooled ORs and the forest plots will be computed using the metaan command in STATA (21). We will use ORs to pool the results because this was the most commonly reported estimated for effect in individual studies and because ORs are considered more appropriate for cross sectional studies compared with other estimates such as relative risk (22). In this study, OR>1 indicates that the stressor is associated with an increased risk of burnout or distress, whereas OR<1 indicates that the stressor is associated with a decrease risk. In accordance to recommendations (20), across studies reporting multiple measures of the same stressor category (e.g. different measures of job demands such as on call or long working hours), the median ORs will ve computed to ensure that each study contributed only one effect measure to each meta-analysis. The I 2 statistic will be used to assess heterogeneity among studies.
Two sensitivity analyses will be performed to examine whether the results are stable when only i) highly rated methodologically studies were retained in the analyses (a score of ≥6) and ii) studies using measures of burnout only. Including measures of only burnout would enable us to assess whether the results are stable when only prolonged stress (burnout) outcomes are included.
Potential for publication bias will be assessed on all pooled outcomes which included ≥9 studies (24). The possibility of publication bias will be examined by inspecting the symmetry of funnel plots and using Egger's test (25). Funnel plots were constructed using the metafunnel command and the Egger test was computed using the metabias command (26,27). All analyses will be performed in Stata version 14.