Impact of on-call shifts on working memory and the role of burnout, sleep, and mental well-being in trainee physicians

ABSTRACT

Background

Optimal cognitive functions, including working memory (WM), are crucial to enable trainee physicians to perform and excel in their clinical practice. Several risk factors, including on-call shifts, poor mental health, burnout, and sleep problems, can impair clinical practice in trainee physicians, potentially through cognitive impairment; however, these associations have not been fully explored.

Objective

This study investigated the effect of on-call shifts on WM among trainee physicians and its association with burnout, depression, anxiety, affect, and sleep.

Materials and methods

This cross-sectional study involved 83 trainee physicians (45% male). We measured demographic and training-related factors including on-call shifts and working hours. We also assessed depressive symptoms (PHQ-9), both state and trait anxiety (STAI total score), burnout (OLBI total score), positive and negative affect scores (PANAS), and sleep disturbances (PSQI total score). WM was evaluated using spatial working memory (SWM) strategy scores that reflected performance and total error counts.

Results

Trainee physicians with more on-calls per month had significantly worse depressive symptoms, burnout scores, and sleep, as well as more negative affect. While controlling for covariates, being on-call more times per month was significantly associated with worse WM. Worse depressive symptoms and burnout scores were also significantly associated with impaired WM.

Conclusion

Working more on-call shifts is associated with compromised WM. Trainee physicians who experienced more depressive symptoms and burnout had worse WM.

KEYWORDS:

• Physicians
• cognitive functions
• working memory
• spatial working memory
• burnout
• depression
• sleep

1. Background

Trainee physicians must balance education and training activities with clinical workload. Long working hours and on-call demands are obligatory in medical training [1]. Long working hours have been previously linked with negative health outcomes [2] and associated with self-reported exhaustion, stress, and burnout [3]. Studies also suggest that disruptive work schedules, such as night shifts, are associated with cognitive impairment [4] and sleep disturbance [5], which may adversely affect trainees’ well-being and careers. This strain is compounded by mental health challenges, including high rates of depression, in medical trainees [6].

Cognitive functions, such as attention and working memory (WM), are core skills enabling physician performance and are directly linked to clinical reasoning and decision-making in clinical settings [7]. Previous studies have reported that individual differences in cognitive abilities among trainee physicians may contribute to challenges in medical training, clinical practice, and patient care, specifically in those reassessed for medical practice competency [8].

WM is the process of not only retaining but also manipulating short-term information [9]. It is the ability to remember, process, and effectively use relevant information in the current context to guide reasoning and decision-making [10]. Hence, WM is a core part of the clinical reasoning enabling a physician to reason in different clinical scenarios [11]. WM operates by directing attentional resources to relevant information [12] but has a limited capacity [13]. The workload and exhaustion experienced in clinical settings have a profound neurobiological impact on several cognitive functions, including WM [14]. cognitive load, defined as ‘the amount of information a person holds and processes within working memory’ [15], can be negatively impacted by physicians being overloaded or burned out by clinical duties. Hence, WM can be impaired or overloaded during clinical duties, potentially affecting the trainees’ ability to learn and use information, hindering their clinical training and practice [16].

Burnout is common among medical professionals, especially those at the postgraduate level [16]. In a recent survey performed by the General Medical Council in the UK of more than 45,000 trainees, more than one-third of the responding sample reported they felt burned out from work [17]. Clinical burnout is associated with poor cognitive function, including WM [16,18]. Hence, WM performance is expected to worsen with excessive and tiring work hours, especially when practitioners are on-call, and may impair workplace performance [19].

Anxiety and depression impair WM and result in lower academic grades among healthy students [20]. Depression is not only prominent among healthcare workers but also associated with worse cognitive abilities including WM [21]. Additionally, sleep disturbance has been found to significantly affect mood, processing speed, and WM of non-medical trainee shift workers [22]. The literature on cognitive impact of workload and on-call on medical trainees is sparse [23]. A comprehensive understanding of the relationship between these factors and their association with WM may enable better design of interventions to improve the well-being and medical performance of trainee physicians. In this study, we evaluated the effect of self-reported workload (on-calls and working hours) on WM while accounting for the role of burnout, mental well-being (depression and anxiety), affect, and sleep in trainee physicians.

2. Methods

2.1. Design and procedure

2.1.1. Ethics statement

Ethical approval for the study was provided by the institutional review board at King Abdulaziz University Hospital (Reference No. 391–22). All participants signed an electronic consent form prior to participation in this study. Participants were given the right to ask questions and withdraw at any point in the study. All participant information was anonymized.

2.1.2. Participants

Participants were recruited from the local physician trainee population at King Abdulaziz University Hospital, a national trainee center rotating with other hospitals. Recruitment was mainly through electronic fliers distributed among trainee physicians that briefly explained the study nature and requirements. Participants were invited from different departments and training years. Those individuals who showed interest in the study were then screened, and the study procedure was explained to them. If participants agreed, they consented to participate in the study. Inclusion criteria were any physician wishing to be enrolled in the study and speaking fluent English. We excluded trainees with a documented neurological or psychiatric illness or sleep disorder.

2.1.3. Study design and participant information

This cross-sectional study was conducted between August and December 2022. The participants first completed a set of demographic questionnaires to determine their age; sex; marital status; number of children, if any; year of training; and specialty. Participants were also asked about any associated medical conditions or medication use, smoking, and consumption of caffeinated or energy drinks. Thereafter, participants were surveyed about their workload, including working hours per day, regular working hours or shift hours, day or night shifts or both, and number of on-calls per month.

2.2. Measures

2.2.1. Study questionnaires

Participants were given the Patient Health Questionnaire (PHQ)-9 [24]. PHQ-9 is a 9-item questionnaire that assesses the severity of depression based on nine criteria from the Diagnostic and Statistical Manual-IV (DSM-IV) on a Likert scale ranging from 0 (not at all) to 3 (nearly every day); higher scores indicate more severe depression [24]. The State-Trait Anxiety Inventory (STAI) [25] has 20 items, with separate items for trait and state anxiety. All items are scored on a 4-point scale where 1 is ‘almost never’ and 4 is ‘almost always’ [25]; higher scores on the scale indicate greater anxiety. The Oldenburg Burnout Inventory (OLBI) [26] is a 16-item self-report tool used to assess work-related burnout; a higher scale indicates more severe burnout. Items are split between two subscales assessing disengagement and exhaustion [26], and questions are scored on a 4-point Likert scale where 1 is ‘strongly agree’ and 4 is ‘strongly disagree.’ The Positive and Negative Affect Schedule – Short Form (PANAS-SF) [27] comprises 20 items, with 10 items assessing positive affect (e.g. excited) and 10 items assessing negative affect (e.g. guilty). Each item is measured on a scale from 1 to 5, where 1 means ‘very slightly or not at all’ and 5 means ‘extremely’ [27]. Each 10 items are scored separately, with higher scores per item corresponding to a higher emotion in that category [27]. Sleep quality was assessed through the Pittsburgh Sleep Quality Index (PSQI) [28], which is a 10-item scale with subcomponents, all of which are scored from scored from 0 (no difficulty) to 3 (severe difficulty) [28]. The cutoff points of these scales are presented in the results section below.

2.2.2. Working memory assessment

WM was assessed using the CANTAB cognitive research software (Cambridge Cognition, Cambridge, UK). CANTAB is a widely used computer-based neuropsychological assessment for identifying cognitive functions [29]. The software runs several cognitive function assessment tests, all of which are computerized and presented on a touch screen, and has been used and validated in neurocognitive research for over two decades. The spatial working memory (SWM) CANTAB battery was selected to assess WM in this study and has been validated in the literature [30]. The test examines individuals’ ability to remember spatial information (tokens hidden in boxes presented on the screen). Participants are required to develop strategies to search for these hidden tokens inside the boxes, and the test automatically scores the number of errors they make in finding them; thus, if they open an empty box or one that has already been opened, an error is recorded, and the more errors, the worse the SWM performance. The test also measures participants’ search strategies in finding these tokens, and the lower the strategy score, the better their WM is. The test difficulty increases with the number of tokens. The test also changes the color and sequence of boxes in each trial to increase the difficulty level. The maximum duration of the test is 6 minutes.

2.2.3. Statistical analysis

Associations between variables were tested using Pearson correlation coefficients, with significance indicated at α = 0.05. Comparison of continuous measures between two groups used independent samples t-tests. General linear modeling was used to test for associations between mental well-being, burnout, sleep, cognition, and the number of monthly calls while controlling for covariates. All models controlled for the effects of age, sex, number of hours worked, and experience level (intern or resident). One model was created per outcome, and false discovery rate-adjusted p-values were reported to control for multiple comparisons. Multicollinearity was checked using the variance inflation factor (VIF) and was deemed negligible (all VIF < 1.50). All models were confirmed to have normally distributed residuals both by using the Kolmogorov – Shapiro test for normality (α = 0.05) and visually inspecting QQ plots. All statistical analyses were completed using R version 4.1.1.

3. Results

3.1. Descriptive data

The sample comprised of 83 physicians with a mean age of 26.3 years (SD = 3.3, 45% male). Table 1 outlines the sample characteristics. Those working both day and night shifts during the current rotation (n = 50, 60%) worked an average of 8.8 hours daily (SD = 1.9), and those who only worked day shifts had a similar average (mean = 8.5, SD = 1.4, t(80) = 0.87, p = 0.385).

Table 1. Sample demographics and characteristics.

Sample size	83
Age – Mean (SD)	26.29 (3.33)
Sex (M) – Num (%)	37 (44.6)
Marital status (Married) – Num (%)	16 (19.2%)
Current rotation – Num (%)
Emergency medicine	19 (23%)
Medicine	22 (27%)
Surgery	22 (27%)
Obstetrics and gynecology	10 (12%)
Pediatric	9 (11%)
Year – Num (%)
Intern	38 (45.8)
First	12 (14.5)
Second	14 (16.9)
Third	7 (8.4)
Fourth	12 (14.5)
Hours daily – Mean (SD)	8.77 (1.80)
Calls monthly – Mean (SD)	3.94 (3.61)

The PHQ-9 scores suggested moderate depressive symptoms in 20.5% of participants and moderately severe or severe symptoms in an additional 15.6%. The PSQI scores indicated sleep disturbances (scores of 5 or greater) in 85.5% of participants. Overall, individuals with at least moderate depression were on-call more times per month (mean = 5.3, SD = 4.1) than those who were not moderately depressed (mean = 3.2, SD = 3.1). Depressive symptoms (PHQ-9), state and trait anxiety, sleep disturbances (PSQI scores), and burnout (OLBI disengagement, exhaustion, and full-scale) scores are further outlined and broken down by clinically significant cutoffs, where available, in Table 2.

Table 2. Descriptive data of the study variables.

Sample size	83
PHQ-9 – Mean (SD)	8.41 (5.16)
PHQ-9 categories – No. (%)
Minimal (1–4)	22 (26.5)
Mild (5–9)	31 (37.3)
Moderate (10–14)	17 (20.5)
Moderately severe (15–19)	10 (12.0)
Severe (20–27)	3 (3.6)
STAI state – Mean (SD)	44.05 (12.20)
STAI state categories – No. (%)
None/Low (20–37)	26 (31.3)
Moderate (38–44)	18 (21.7)
High (45–80)	39 (47.0)
STAI Trait – Mean (SD)	43.73 (11.12)
STAI trait categories – No. (%)
None/Low (20–37)	26 (31.3)
Moderate (38–44)	14 (16.9)
High (45–80)	43 (51.8)
PSQI – Mean (SD)	8.31 (3.60)
PSQI categories – No. (%)
No sleep disturbances (<5)	12 (14.5)
Sleep disturbances (5+)	71 (85.5)
OLBI full-scale categories – No. (%)
Low (<1.62)	3 (3.6)
Medium (1.63–2.68)	52 (62.7)
High (>2.68)	28 (33.7)

Participants had SWM strategy scores of 13.3 (SD = 4.45) and made, on average, 9.8 (SD = 9.1) total errors. Those who were at least moderately depressed showed worse (higher) strategy scores (mean = 14.3, SD = 4.0) and made more errors (mean = 11.2, SD = 9.8) than those who were not (strategy mean = 12.8, SD = 4.6; error mean = 9.1, SD = 8.6), although this difference was not statistically significant.

3.2. Intercorrelations for on-call shifts

Pearson correlation coefficients showed that being on-call more times per month was significantly correlated with a worse WM strategy (r = 0.290) and more WM errors (r = 0.318). This was also correlated with worse depressive symptoms (r = 0.317), disengagement (r = 0.244), burnout (r = 0.242), and worse sleep (r = 0.266). Critically, burnout was also significantly correlated with worse WM strategy (r = 0.234), worse sleep (r = 0.527), and depressive symptoms (r = 0.574).

Depressive symptoms, anxiety (both state and trait), burnout, and worse sleep were all highly significantly correlated with one another (Table 3). Relative to women, men had significantly lower state [t(81) = −2.85, p  = 0.006] and trait [t(81) = −3.26, p  = 0.002] anxiety and less burnout [t(81) = −2.08, p  = 0.041].

Table 3. Intercorrelations of mental scores. All results are presented as r = correlation coefficient.

	PHQ total	STAI state total	STAI trait total	OLBI disengagement	OLBI exhaustion	OLBI full scale	PANAS Positive	PANAS Negative	PSQI total
STAI state total	0.688		0.809	0.541	0.570	0.621	−0.515	0.750	0.524
STAI trait total	0.658	0.809		0.545	0.600	0.639	−0.507	0.731	0.538
OLBI disengagement	0.408	0.541	0.545		0.604	0.899	−0.649	0.436	0.406
OLBI exhaustion	0.523	0.570	0.600	0.604		0.892	−0.597	0.465	0.539
OLBI full scale	0.520	0.621	0.639	0.899	0.892		−0.696	0.503	0.528
PANAS Positive	−0.315	−0.515	−0.507	−0.649	−0.597	−0.696		−0.389	−0.346
PANAS Negative	0.664	0.750	0.731	0.436	0.465	0.503	−0.389		0.646
PSQI total	0.575	0.524	0.538	0.406	0.539	0.528	−0.346	0.646

3.3. Impact of being on-call and mental well-being on cognition

When controlling for hours worked, experience level (intern or resident), age, sex. Being on-call more times in the last month were significantly associated with worse SWM strategy [t(76) = 3.23, p = 0.002] and more errors [t(76) = 3.15, p = 0.002]. We additionally investigated the impact of mental well-being on cognition (Table 4) and found that worse depressive symptoms and burnout were significantly associated with worse WM strategy.

Table 4. Associations between working memory strategy and errors and worse mental well-being, burnout, and sleep. A lower strategy is better.

	Working memory strategy		Working memory errors
	Coefficient (SE)	Significance	Coefficient (SE)	Significance
PHQ-9 Total score	0.29 (0.10)	t(76) = 2.90, p = 0.034*	0.29 (0.22)	t(76) = 1.36, p = 0.272
STAI State anxiety total score	0.07 (0.05)	t(76) = 1.61, p = 0.220	0.10 (0.10)	t(76) = 1.02, p = 0.343
STAI Trait anxiety total score	0.08 (0.05)	t(76) = 1.55, p = 0.220	0.14 (0.11)	t(76) = 1.31, p = 0.272
OLBI Burnout total score	2.89 (1.11)	t(76) = 2.61, p = 0.038*	4.03 (2.37)	t(76) = 1.70, p = 0.272
PANAS Positive affect score	−0.08 (0.06)	t(76) = −1.28, p = 0.285	−0.19 (0.13)	t(76) = −1.44, p = 0.272
PANAS Negative affect score	0.06 (0.06)	t(76) = 1.01, p = 0.314	0.17 (0.12)	t(76) = 1.41, p = 0.272
PSQI Total score	0.16 (0.14)	t(76) = 1.12, p = 0.311	0.28 (0.30)	t(76) = 0.95, p = 0.343

Significant associations are shown using asterisks (*, p  < 0.05).

Finally, given that WM was significantly associated with both being on-call more times per month and mental well-being, we tested the assoication between on-call and mental well-being. When controlling for all covariates, being on-call more often was significantly associated with worse depressive symptoms, burnout, affect, and sleep problems (Table 5).

Table 5. Associations between being on-call more often and worse mental well-being, burnout, and sleep, while controlling for hours worked, experience level, age, and sex.

	Coefficient (SE)	Significance
PHQ-9 Total score	0.45 (0.14)	t(77) = 3.17, p = 0.015*
STAI State anxiety total score	0.66 (0.33)	t(77) = 1.98, p = 0.052
STAI Trait anxiety total score	0.60 (0.30)	t(77) = 2.01, p = 0.052
OLBI Burnout total score	0.03 (0.01)	t(77) = 2.36, p = 0.036*
PANAS Positive affect score	−0.57 (0.24)	t(77) = −2.40, p = 0.036*
PANAS Negative affect score	0.62 (0.26)	t(77) = 2.37, p = 0.036*
PSQI Total score	0.23 (0.11)	t(77) = 2.17, p = 0.046*

Significant associations are shown using asterisks (*, p  < 0.05).

4. Discussion

This study investigated the impact of on-call shifts on WM in trainee physicians and examined associations with burnout, mental health (depression and anxiety), affect, and sleep. Trainee physicians with more on-calls per month had worse WM. More on-calls were also associated with worse depression, and burnout scores more negative and less positive affect, and sleep disturbances. Individuals with more depressive symptoms and burnout had worse WM impairment compared to others.

The study results are consistent with previous research indicating that those working long hours – and night shifts, specifically – have more severely impaired WM [4,31]. Nevertheless, our study further investigated the roles of mental well-being, affect, and sleep disturbances in this relationship. Accounting for these factors may explain some of the contradictory results in the literature [18]. Our findings indicate that the association between worse WM and more on-calls per month may be related to individual differences in depression, which highlights the vulnerability of those affected to impaired WM. Nonetheless, this has not been previously investigated in trainee physicians. A longitudinal assessment would help to elucidate any causal or mediating role of depressive symptoms in the relationship between more on-calls per month and worsened WM.

Working night shifts is known to disturb the circadian mechanism, previously labeled shift-work disorder, a clinically defined condition resulting from working mainly at night or starting or working an alternating-shift schedule, a situation experienced by most health workers [32]. However, few studies have explored the impact of such disorders on cognitive functions. In clinical practice, managing training demands and handling patients requires high cognitive ability, including information processing, decision-making, and WM, and it is crucial to investigate these factors to ensure optimum patient care [33].

Previous work has suggested that WM characterizes a neurocognitive defect in depression [34]. Depression negatively impacts WM, apparently by impairing the ability to effectively react to environmental cues [35]. Specifically, depression may interfere with key regions for WM, such as the prefrontal and cingulate cortices [36]. The prefrontal cortex is crucial for WM and is impaired in depression, which further explains the association between depression and WM impairment found in the present study [37]. Nevertheless, while the present study found that being on-call predicted depressive symptoms in the studied population, being on-call additionally was associated with a complex set of consequences, whether cognitive, affective, or both. The implication is that an intervention targeting only depression may be insufficient to address the negative cognitive consequences of being on-call. Instead, separate cognitive interventions, for example WM training, may be needed.

The prevalence of burnout in our cohort was 95%, and a third of our sample reported high levels of burnout, replicating previous findings locally [38,39] and internationally [40–42]. Many efforts have been made to identify predictors of burnout. While some evidence suggests it is mostly linked to external work factors, such as work demands, work environment, and lack of supervisory or mentoring support [43], other studies have found that internal factors, such as self-control and personality, determine the impact of burnout on trainee physicians [3]. After adjusting for covariates, including sex and type of training program, our study found that being on-call more times per month was a significant predictor of burnout; that is, those with more on-calls were more likely to report burnout. Importantly, trainee burnout was also associated with reports of sleep disturbance, depressive symptoms, and WM impairment. Overall, these findings of the disruptive effects of burnout on physical, mental, and cognitive well-being not only expand on previous findings [43] but also highlight the need to develop interventional and strategic frameworks to manage burnout among trainee physicians, a suggestion also made in previous studies [44].

5. Limitations of the study

The main limitation of this study is the cross-sectional nature of the study design, which limits understanding of how WM changes over time in response to factors such as on-call time. Hence, future longitudinal studies incorporating a follow-up period are essential. Questionnaire recall bias is common in this type of study design, which may also have limited the study findings. Furthermore, about half of our sample were intern trainees (45.8%) with fewer on-calls; thus, the present findings may not generalize to non-trainee physicians with more on-calls per month, which should be a target for future investigations.

6. Conclusion

Trainee physicians with more on-call periods had worse WM and increased depressive symptoms and burnout. Moreover, we found that burnout is associated with worse WM and impaired sleep and emotional well-being. More efforts are needed to address these factors in workplace schedules and provide a more balanced working environment for trainee physicians to ensure better clinical care.

Declaration of financial/other relationships

The author has no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties. Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgments

The author would like to thank Maha H. Almuhaiyawi and Maha H. Alreemi for their assistance throughout in data collection.

Data availability statement

Data sharing is not applicable to this article as no new data were created or analyzed in this study.

Additional information

Funding

This paper was not funded.