The effect of time and day of the week on burnout-related experiences: an experience sampling study

ABSTRACT

Burnout has traditionally been characterized as a relatively stable construct, leaving the question of whether and how burnout-related experiences fluctuate within and between days unaddressed. In the current study, we assess the effect of time of day (expressed as external time, internal time, or time awake) and day of the week on momentary experiences of the two core components of burnout, i.e., exhaustion and disengagement. We employed a 7-day experience sampling method in the field among 65 working employees, with seven momentary assessments per day. Results indicated that a large proportion of variance in burnout-related experiences occurred between moments (46%-68%), with only minor variance occurring between days within participants (2%-6%). Notably, experiences related to the disengagement component showed no clear pattern over the day, while exhaustion remained relatively stable throughout the morning and then increased moderately towards the end of the day. We conclude that burnout-related experiences typically fluctuate between moments, supporting the view of burnout as a dynamic rather than a purely static state. Furthermore, much of the variance in momentary burnout-related experiences remains to be explained in absence of a structural temporal pattern.

KEYWORDS:

• Burnout
• disengagement
• experience sampling
• exhaustion
• momentary states

Introduction

Burnout is often described as a prolonged response to stressors at work (Maslach et al., 2001) characterized by high levels of both exhaustion and disengagement, deemed the energetic and motivational components of burnout, respectively. Exhaustion refers to feelings of depletion and strain, whereas disengagement is defined as a withdrawal reaction whereby employees distance themselves mentally from their work (Demerouti et al., 2001). Traditionally, burnout has been characterized as a relatively stable construct with research questions focused on how “stable” work characteristics (e.g., job demands and resources) or individual traits are related to or predict the development of burnout (Xanthopoulou & Meier, 2014). Yet, there is accumulated evidence that employee well-being and burnout have both a stable component that is slow to change and a dynamic component undergoing reversible changes more rapidly (within persons) as a response to changing circumstances or events (McCormick et al., 2020; Ohly et al., 2010; Sonnentag, 2015; Xanthopoulou & Meier, 2014).

Within-person research on the dynamic component of burnout has largely assessed covariation between two or more variables (i.e., proximal predictors) at particular moments in time (i.e., used the differential approach), and thus little is known about if and how daily and momentary experiences related to burnout fluctuate systematically within a person over time (i.e., the temporal perspective) (Fisher & To, 2012; Hülsheger, 2016). The temporal approach focuses on the type of dynamics that occur in a construct of interest and can thus shed light on the typical ebb and flow of such experiences (Fisher & To, 2012; Hülsheger, 2016). Such an approach can provide insights into how momentary burnout-related experiences develop or unfold over the course of a day and week in healthy participants and allows investigation of what factors may cause changes in typical fluctuations. Therefore, we adopt the temporal approach to investigate time profiles within and between days in specific experiences related to both the exhaustion and disengagement components of burnout: exhaustion (exhaustion-related), boredom and (a lack of) feeling positively challenged (disengagement-related). Specifically, we aim to 1) quantify the proportion of variance in these momentary experiences between persons, between days, and between moments, 2) assess whether these experiences fluctuate in structural patterns as a function of day of the workweek (Monday through Friday), 3) investigate and compare whether within-day fluctuations in these experiences occur in a linear or curvilinear (parabolic and third-degree polynomial) pattern over time and 4) explore how and to what extent within-day time profiles are impacted by trait burnout levels. Although previous research adopting the temporal approach has demonstrated systematic fluctuations both within- (Egloff et al., 1995; Golder & Macy, 2011; Hülsheger, 2016) and between-days (Beal & Ghandour, 2011; Egloff et al., 1995; Hülsheger et al., 2014, 2022; Larsen & Kasimatis, 1990; Luta et al., 2019; Pindek et al., 2021; Rook & Zijlstra, 2006; Weigelt et al., 2021) in a number of phenomena relevant to burnout experiences, temporal patterns differ by phenomena and thus it remains unclear if and which patterns can be replicated in specific experiences related to both burnout dimensions. Moreover, in contrast to previous studies that have examined time profiles in energy-related states (Hülsheger, 2016; Sonnenschein et al., 2007), we also include items assessing specific experiences related to disengagement. This is important for understanding the dynamic component of burnout, as research shows that exhaustion and disengagement are strongly related at the between person (Demerouti et al., 2001; Maslach et al., 2001) and daily level (Basinska & Gruszczynska, 2020), yet, it remains unclear whether experiences related to exhaustion and disengagement fluctuate in similar patterns within days.

Furthermore, insights from chronobiology have shown that time can also be conceptualized as internal time and time since awakening (Roenneberg et al., 2003), and highlighted the importance of accounting for variations in behaviour as a function of general and daily sleep-wake patterns (Facer-Childs & Brandstaetter, 2015; Schmidt et al., 2007). Internal time reflects the phase of entrainment to the light/dark cycle (Åkerstedt et al., 2009; Roenneberg et al., 2019) that manifests as differences in preference for sleep-wake timing (i.e., chronotype), ranging from “late” to “early” types (Roenneberg et al., 2003; Schmidt et al., 2007). Time since awakening is closely related to the homoeostatic sleep process (Åkerstedt et al., 2009; Borbely, 1982) (i.e., the drive to sleep) which increases with time spent awake, and decreases during the following sleep episode (Borbely, 1982; Fuller et al., 2006). Constructs related to momentary burnout experiences have been found to vary with internal time and to decrease with time spent awake (Åkerstedt et al., 2009). As adjusting for these intra- and interindividual differences may have substantial implications for time profiles of momentary experiences related to burnout, we include, and are the first to compare, temporal dynamics in burnout-related experiences as functions of internal time and time since awakening in addition to local clock time.

This study thus makes several contributions. By examining within-person fluctuations in burnout-related experiences amongst sub- and non-clinical populations, we can uncover “typical” time profiles in exhaustion, boredom, and feeling positively challenged while working in (seemingly) healthy participants. In order for future studies to ensure that a construct of interest caused changes in such momentary experiences, information regarding the frequency and magnitude of changes in such experiences, and the presence of repeating patterns or trends is needed so that effects can be accurately attributed to a causal factor and are not masked by time-of-day or type-of-day effects (Gabriel et al., 2018; McCormick et al., 2020). An understanding of temporal dynamics is also needed to be able to formulate theories explaining causal relationships across time so that increases or decreases are described as deviations from natural fluctuations (Gabriel et al., 2018). Moreover, previous studies demonstrated that there may be substantial differences in time profiles of energy-related states in clinical burnout patients and healthy controls, suggesting the development of burnout may impact structural patterns (Sonnenschein et al., 2007). By investigating trait burnout as a cross-level moderator, we can explore if and how daily time trajectories (i.e., dynamic components) are dependent on trait levels (i.e., stable components) of burnout in persons still at work: do persons with high vs. low levels of trait burnout demonstrate different patterns throughout the day?

Burnout experiences across days

Burnout-related experiences are likely to change within-persons from one day to another, either in response to a situation or event, or due to reoccurring cycles resulting from the socially constructed environment (Beal & Ghandour, 2011; Dalal et al., 2014; Xanthopoulou & Meier, 2014), such as the repetitive structure of the workweek (five days of work followed by two days of rest). In fact, recent diary studies on burnout demonstrated that the majority of variance in both disengagement and exhaustion occurred within persons (approximately 55–72%), indicating that these experiences fluctuate more between days within persons than between participants (Basinska & Gruszczynska, 2020; Gruszczynska et al., 2021; Mäkikangas et al., 2016). Yet, Basinska and Gruszczynska (2020) demonstrated that neither exhaustion nor disengagement increased or decreased systematically throughout the 10-day study period (two consecutive five-day workweeks) in a linear pattern. While an earlier study by Sonnenschein et al. (2006) showed that exhaustion decreased linearly across 14 days in both clinical and non-clinical participants, the effect size was small (R² = 0.2%), implying that there was a negligible decrease in exhaustion throughout the study period. Nevertheless, these studies assessed linear structural patterns as a function of study day and not day of the week, and thus it remains unclear if and how burnout-related experiences demonstrate entrainment to days of the week.

Although phenomena related to burnout (e.g., job satisfaction and job stressors, fatigue, sleep quality and detachment, work engagement, positive and negative affect, and daily hedonic level) have demonstrated entrainment to day of the week, patterns vary substantially between the different constructs with some showing linear (Hülsheger et al., 2014; Pindek et al., 2021) and curvilinear patterns (Beal & Ghandour, 2011; Larsen & Kasimatis, 1990; Luta et al., 2019; Rook & Zijlstra, 2006), some revealing a dichotomous distinction between workdays and weekend days (Egloff et al., 1995), while others show no clear pattern as a function of day of the (work)week at the group level (Egloff et al., 1995). Notably, negative affect (and a lack of positive emotions) (Maslach & Leiter, 1997) and fatigue are closely related the burnout experience and thus provide valuable insights into temporal dynamics of momentary burnout-related experiences. Fatigue has been demonstrated to peak at the beginning of the week, then decrease abruptly on Friday (Weigelt et al., 2021) while positive and negative affect show opposite fluctuations throughout the workweek, with positive affect increasing, and negative affect decreasing linearly from Monday through Friday (Hülsheger et al., 2022; Larsen & Kasimatis, 1990).

Despite the relevance of affect and fatigue for daily and momentary burnout-related experiences, several conceptual differences warrant investigation into whether patterns detected in fatigue and especially positive and negative affect can be replicated in specific burnout-related experiences. Whereas neither fatigue nor affect are necessarily work-related, burnout-related experiences are per definition experienced as a consequence of work (Xanthopoulou & Meier, 2014). Moreover, affect does not include a motivational component, which is a key feature of disengagement. Thus, traditional mood items do not assess the work-related and motivational states that should be the core of momentary experiences related to burnout.

Given these conceptual differences, the different patterns across (week)days found in related phenomena and because of the absence of research on disengagement and exhaustion specifically, it is still unclear if and how burnout-related experiences are entrained to the workweek. Thus, several feasible structural patterns are considered in the current study. In line with the Effort-Recovery Theory, employees may have replenished motivational and energetic resources after the weekend, which is a substantial period of recovery (Fritz & Sonnentag, 2005). In this case, employees are expected to have experienced the lowest intensity of burnout-related experiences on Monday, when resources were still available. As effort is expended throughout the workweek and employees draw on resources to meet work demands (Meijman & Mulder, 1998), the intensity of burnout-related experiences should have increased in a linear or parabolic function, reaching its highest peak on Friday. Alternatively, considering that the sample consisted of healthy employees, participants may have been able to recover adequately between workdays, thus preventing an accumulation of exhaustion- and disengagement-related experiences with day of the week. This would result in rather stable levels of burnout-related experiences between days of the week as resources are fully restored to baseline after work. Another possibility, often used to explain weekly patterns in mood, is that employees experience the “Monday blues” due to the anticipation of stressors that will be encountered throughout the week, thus rationing resources to be able to meet demands throughout the workweek (Hülsheger et al., 2022; Pindek et al., 2021). As anticipation of work tasks decreases with the approaching weekend, the intensity of burnout-related experiences should decrease accordingly, with peak intensities occurring on Monday and troughs on Friday. This is in line with results from Hülsheger et al. (2022) demonstrating that anticipation of work and negative affect decreased, and positive affect increased, in linear patterns throughout the workweek. Lastly, more abrupt changes may occur at transitions from the workweek to the weekend and vice versa due to the anticipation of leisure time (Weigelt et al., 2021). This would suggest that burnout-related experiences are high on Monday, remain relatively stable from Monday to Thursday, and decrease abruptly on Friday.

Expected fluctuations in burnout-related experiences within days

Burnout-related experiences are also unlikely to be static within days, probably fluctuating from moment to moment depending on reoccurring and situational factors that vary with time (Xanthopoulou & Meier, 2014). Indeed, as one of the only studies to explicitly explore the exhaustion component of burnout within days, Sonnenschein et al. (2007) found that 42% of the variance in momentary exhaustion occurred between moments. Previous research has demonstrated that energy-related constructs (e.g., fatigue, sleepiness, vitality, activation, and alertness) seem to fluctuate in U-shaped patterns, with a peak in energy around late morning or noon, and a decrease towards the evening, both in field (Egloff et al., 1995; Hülsheger, 2016; Smolders et al., 2013; Sonnenschein et al., 2007) and in laboratory studies (Daurat et al., 1993; Rüger et al., 2006). In line with this, Sonnenschein et al. (2007) demonstrated a significant (although modest) relation between local clock time and exhaustion: exhaustion decreased slightly at the beginning of the day, remained relatively stable throughout the day, then increased again towards the end of the evening. Notably, the U-shaped pattern their one-item exhaustion measure had a considerably lower amplitude (i.e., was more flattened) than fatigue, suggesting exhaustion is a more extreme and stable form of fatigue (Sonnenschein et al., 2007). Although time-dependent fluctuations in disengagement have not been explored explicitly within days, motivation and mood-related constructs have been demonstrated to fluctuate in various structural patterns within days. For instance, valence and tension fluctuated in u-shaped diurnal patterns (Zhang et al., 2018), pleasantness increased linearly over the course of the day (Egloff et al., 1995) and Twitter messages of users around the globe indicating positive and negative affect decreased and increased throughout the day respectively (Golder & Macy, 2011).

As constructs related to burnout did not provide a clear answer regarding what patterns to expect for exhaustion- and disengagement-related momentary experiences specifically we aimed to assess and compare the predictive strengths of three feasible patterns as a function of time: linear, parabolic, and third-degree polynomial. The intensity of momentary disengagement-related experiences (i.e., the lack of motivation to engage in work-related activities) should be a function of available energetic resources and momentary task demands (Meijman & Mulder, 1998) and fluctuate accordingly, thus disengagement- and exhaustion-related experiences are expected to fluctuate in similar patterns.

According to both the Effort-Recovery theory (Meijman & Mulder, 1998) and the two-process model of sleep regulation (Borbely, 1982), either a linear or parabolic increase can be expected in the time profiles of exhaustion and disengagement-related experiences. The Effort-Recovery theory postulates that employees start their workday in an optimal state, with mental and energetic resources at their disposal (Meijman & Mulder, 1998). As employees mobilize available resources to meet task demands, psycho-physiological changes are induced, resulting in the experience of temporary load reactions such as exhaustion or boredom. Research shows that recovery processes (such as sleep or detachment) primarily occur following work episodes (Geurts & Sonnentag, 2006). This suggests that employees have little opportunity for recovery from load reactions during their workday, and are likely required to invest compensatory effort following the initial experience of load reactions, thereby increasing the intensity of subsequent load reactions (Meijman & Mulder, 1998). Investing effort to meet task demands should thus trigger a loss spiral, with momentary experiences related to burnout intensifying or accumulating in a linear (additive) or parabolic (exponential) pattern as more compensatory effort is needed to maintain work performance.

Moreover, according to the two-process model of sleep regulation, sleep propensity at any given moment depends on circadian and homoeostatic processes (Borbely, 1982; Schmidt et al., 2007). Given the similarity between the construct of sleepiness and especially energy-related experiences related to burnout, it seems likely that similar processes may regulate momentary exhaustion. Since sleep is one of the primary recovery mechanisms for restoring psycho-physiological systems overnight (Åkerstedt et al., 2009; Sonnentag, 2018), load reactions should dissipate during this time and increase linearly throughout the waking day in accordance with the homoeostatic process. The circadian process is generally entrained to the 24 hr light/dark cycle, synchronizing internal rhythms to environmental time using indicators of external time (such as light/dark cues) (Roenneberg et al., 2003, 2019). This process, which manifests as oscillatory patterns, is regulated by the internal clock to optimize functioning (e.g., not having a surplus of energy at bedtime or falling asleep during lunch) (Appleman et al., 2013; Wright et al., 2013). In fact, the circadian system regulates many aspects of human physiology and behaviour, among which mood, alertness, vitality, sleepiness, fatigue, and performance (Åkerstedt et al., 2009; Baron & Reid, 2014; Chellappa et al., 2011; Fisk et al., 2018; Vetter, 2020), that share conceptual similarities with burnout components. Given that we only assess burnout-related experiences throughout the waking day, and thus can only capture part of the 24-hour cycle, the circadian process would predict that exhaustion- and disengagement-related experience fluctuate in a parabolic pattern.

A third-degree polynomial pattern would demonstrate that a second change in the relationship occurs, either increasing or decreasing again after an initial change. As mentioned previously, abrupt changes occur in well-being during the transition from Thursday to Friday, due to the anticipation of leisure activities (Weigelt et al., 2021). Similarly, the anticipation of the end of the workday and leisure activities during off-work time may decrease burnout-related experiences abruptly towards the end of the workday. Thus, experiences related to exhaustion and disengagement may first decrease throughout the morning, then increase towards noon as resources deplete, and finally decrease again in anticipation of the end of the workday.

Individual differences in within-day patterns

While a diurnal pattern in momentary experiences related to burnout may be normal amongst healthy participants, it is nevertheless likely that substantial differences in these patterns exist between persons. Patterns in fatigue have been demonstrated to vary depending on sleep quality the previous night (Hülsheger, 2016) and differ between clinically burned out and healthy control participants (Sonnenschein et al., 2007). Sonnenschein et al. (2007) demonstrated that clinical burnout patients show flattened u-shapes in their pattern of fatigue throughout the day, with a more stable state of high fatigue throughout the entire day. Similarly, de Longis et al. (2022) examined the inertia of negative emotions and found that chronic exhaustion moderated the strength of the inertia insofar that those experiencing high levels of chronic exhaustion demonstrated a stronger relationship between previous negative emotions and current negative emotions. De Longis and colleagues (2022) suggested that this individual stability in negative emotions may signal resource depletion. Similarly, a lack of fluctuations in momentary burnout-related experiences may also signal a pathological process, i.e., the slowing of dynamics or fluctuations (de Longis et al., 2022) in burnout-related experiences. Therefore, we additionally explored whether trait burnout moderates the linear relationship between time and burnout-related experiences to obtain insights regarding the dependency of within-day patterns on trait exhaustion or disengagement.

We expect trait burnout to relate to time profiles in one of two ways. According to the Effort-recovery theory, those experiencing higher levels of trait burnout should begin the day with suboptimal levels of energetic and motivational resources and hence start their workday already exhausted and/or unmotivated. This state requires employees to immediately invest compensatory effort, resulting in increased load reactions or momentary burnout-related experiences (Meijman & Mulder, 1998). This could potentially result in ceiling effects (or stable levels of experiences related to exhaustion and disengagement throughout the day) for those experiencing high levels of burnout (in-line with results from Sonnenschein et al. (2007) and de Longis et al. (2022)). Alternatively, considering that the current study investigates non-clinical populations and thus that ceiling effects are quite unlikely to occur, higher levels of trait burnout could result in a faster depletion of energetic and motivational resources within the day, manifesting as a stronger relationship between time and momentary experiences related to burnout.

Methods

Participants

A sample of employees was recruited via the networks of Master students in a relevant course at the University and via advertisements on various media and newsletters. The inclusion criteria were as follows: participants were required to be over the age of 30 (to ensure sufficient work experience and exclude students with part-time jobs), employed at least four days a week, have access to a phone or tablet, and speak Dutch.

72 participants took part in the study. However, 65 (28 female, 37 male) participants completed the momentary questionnaires during moments when they were working at least three times on one day, and were thus included in the final sample. Participants lived and worked in the Netherlands at the time of the study. The participants’ mean age was 48.9 years old (SD = 9.2, range: 30–65), and they worked on average 41.2 hours per week (SD = 9.34, range: 16–70). Differences in working hours before and during COVID-19 were minor, with a mean of 41.7 (SD = 10.05, range: 16–70, n = 45) hours per week before and 40.1 (SD = 7.61, range: 30–55, n = 20) hours per week during the nationwide restrictions. Participants had varying levels of education: 23 (35%) participants obtained a university degree or higher, 25 (38%) completed an applied sciences education, and 17 (26%) obtained some form of high school degree or lower. Their most frequent employment sectors were research (16.9%), business services (16.9%), healthcare (12.3%), industry (9.2%), and government (9.2%). The mean chronotype of the sample was 3.62 (SD = 0.91, range: 0.92–6.18), indicating that the participants’ midsleep (corrected for sleep dept) occurred at 3:37 AM on average. General levels of disengagement and exhaustion ranged between 1 and 3.38 and between 1.12 and 3.12 respectively (mean disengagement = 2.04, SD = 0.46; mean exhaustion = 2.09, SD = 0.41).

Procedure

Ethical approval was obtained via the Ethical Review Board of the university (Code: 1021). Data collection occurred in multiple measurement waves from October 2019 through July 2020. Due to the COVID-19 pandemic and associated measures (e.g., social restrictions and the advice to work from home) some data collection waves were postponed from March to July 2020.

The study lasted seven consecutive days and consisted of: (1) an intake questionnaire, (2) a daily morning diary, and (3) seven brief momentary questionnaires each day. Before participation, employees followed instruction sessions and were asked to read and sign the informed consent. During this session, a mobile application called MetricWire was installed on participants’ phones, enabling participants to receive notifications and complete the daily and momentary questionnaires on their mobile devices.

The intake questionnaire was completed once in the week preceding the study. Participants accessed the questionnaire via an anonymous link sent to their personal email addresses. Every morning after waking, participants completed the diary on their phones. The diary was triggered with a silent notification at 6:00 and remained available until 12:00. Throughout the waking day (8:00–22:00 on weekdays, 9:00–23:00 on weekends), participants received seven notifications at semi-random intervals (>90 minutes apart) to complete the momentary questionnaires. These questionnaires remained available for 30 minutes after the notification, with reminders after 10 and 20 minutes. After the sampling week, participants were debriefed and thanked for their participation.

This study was part of a larger project on the relationship between light, sleep, and burnout-related experiences; thus, more measurements were taken (for a complete overview, please see our preregistration on the Open Science Framework: mask authors).

Measurements

All questionnaires, diaries, and ESM measures were administered in Dutch.

Intake questionnaire

The intake questionnaire consisted of items assessing trait-level variables that included demographics (age, gender, education level, hours of work per week based on contract, actual hours of work per week, job type), and general levels of disengagement and exhaustion assessed via the Oldenburg Burnout Inventory (OLBI; Demerouti et al., 2003). The response scale of OLBI ranges from 1- “completely disagree” through 4- “completely agree”. The reliability of both subscales of the OLBI was sufficient to good (Exhaustion: Cronbach’s alpha = 0.78; Disengagement: Cronbach’s alpha = 0.79).

Chronotype was assessed via the Munich Chronotype Questionnaire (MCTQ; Roenneberg et al., 2003) and calculated using the formula by Roenneberg et al. (2003).

Diary questionnaires

Participants were asked to complete the core Consensus Sleep Diary (CSD) by Carney et al. (2012) every morning after waking, pertaining to the previous night’s sleep episode. From the CSD, the single item “At what time did you wake up?” was used to calculate the variable time since awakening.

Momentary questionnaires

ESM items were adapted from trait-level questionnaires to apply to a momentary state at single moments in time.

Momentary experiences related to burnout

Valid and reliable questionnaires assessing momentary burnout-related experiences were not yet available. Thus, momentary items were selected based on factor loadings obtained via a previous diary study (unpublished) that translated the OLBI (Demerouti et al., 2003) to a daily assessment level. This resulted in two-item scales for both exhaustion and disengagement, each consisting of one negatively- and one positively- framed item, with the response scales ranging from 1-“completely disagree” to 7-“completely agree”. Exhaustion was measured with: (1) “I have enough energy for my current activity” and (2) “Right now, I feel emotionally drained” (r = .15–.68, mean = .46¹). Disengagement was assessed via (1) “Right now, I feel bored by my activity” and (2) “At this moment, I find my activity to be a positive challenge”.

Correlations between the two items for the disengagement scale were insufficient (r = −0.37–0.39, mean = .101), indicating that the two items do not measure the same construct. Thus, we examined the two disengagement items in separate analyses, henceforth referred to as momentary (1) boredom and (2) positive challenge.

Measures of time

Local clock time

To assess local clock time, time stamps of the completion of the ESM questionnaires were used.

Internal time

To calculate the proxy of internal time, we used each individual’s chronotype, which reflects the habitual preferred midsleep of each participant. Chronotype was subsequently subtracted from local clock time to obtain time since preferred midsleep, i.e., time since the middle of the subjective night for each participant.

Time since awakening

To obtain the measure of time that operationalizes time as the amount of time that has passed since waking, we subtracted the daily wake time of each participant from local clock time. This results in the amount of time that has elapsed since the participant woke up that morning.

Data analysis

Data processing and analyses were performed using RStudio (2020) using the dplyr, lme4, lmerTest, stats, emmeans and jtools packages. Figures were created using the package ggplot2Before data analysis, the distributions of variables were visually examined, and Skewness, Kurtosis, and the Shapiro-Wilk test W-statistic were calculated. Analyses were conducted with and without removing outliers (defined as ± 3 standard deviations from the mean; please see supplementary material for the results of the statistical analyses with outliers excluded). Results did not differ substantially with the exclusion of outliers.

Data from moments when participants indicated they were not working and on weekends were excluded from the main analyses. Moreover, observations for a specific day were removed when participants completed less than three momentary questionnaires that day while working, as three is the minimum number of observations needed to detect a time-dependent pattern. The distribution of residuals of each model was assessed for normality using the Shapiro-Wilk test (W of .97 as the cut-off value for normality). Natural log transformations were necessary for models using momentary exhaustion, momentary boredom and daily exhaustion as outcome variables, as the residuals were not normally distributed (all W < .97). Using the log-transformed variable as the outcome variable in these models resulted in normally distributed residuals for the models investigating momentary boredom and daily exhaustion. Although a natural log transformation for momentary exhaustion did not fully succeed in normalizing the distribution of the residuals of the statistical models (W = .962–.964), we nevertheless proceeded with the analyses using these transformed scores given the close proximity of the resulting W-statistic to the cut-off value.

Multilevel analysis was used to investigate the amount of variance in burnout-related experiences to be explained within days, across days, and between persons. Separate null (unconditional) models were run for exhaustion, boredom, and positive challenge. The nesting was as follows: momentary measures were nested in days, which were nested in participant ID. Intra-class correlations were computed to determine the variance existing at the different levels (momentary, daily, and person level).

To analyse day-to-day variations in burnout-related experiences, we performed 2-level multilevel models by aggregating the momentary values for exhaustion, boredom, and positive challenge to the day level. These average burnout values subsequently became the dependent variable in the multilevel models, with Day (Monday-Friday) included as an ordinal categorical variable (i.e., day of the week was represented as categories or levels). This resulted in the inclusion of dummy variables for each day of the week in the model. Subsequently, to assess whether day of the week had an effect on the variance of each burnout-related experience, the main effect of the Day variable was assessed with an ANOVA (taking into account the hierarchical structure of the data). The corresponding F-statistic and p-value test the null hypothesis that the categorical variable for day did not significantly impact the burnout-related experiences assessed in the study. Pairwise comparisons were subsequently conducted among all days of the week to enable detection of significant differences between any days of the week and thus the presence of any kind of pattern (e.g., abrupt increases or decreases). Post hoc contrasts were corrected using the Tukey method.

Three-level multilevel models were also run to test for structural time-of-day-dependent variations in momentary burnout-related experiences. Local clock time, local clock time squared (for parabolic function), and local clock time to the third power (for 3rd-degree polynomial function) were added one after another as covariate(s) to the model. Subsequently, the linear predictor of time was tested for random slopes within participants. Adding additional random slopes for the non-linear functions (quadratic and 3^rd-degree term) of time resulted in convergence issues due to the increased complexity of the model. For this reason, we were unable to test slope variances for the non-linear predictors of time across participants.The same strategy was employed in the models with internal time and time since awakening as time metrics. Models were compared using Chi-square tests, and the best time-related predictors were identified by comparing the marginal R-squared values of the models. For multilevel models, it is important to use R-squared measures that align with, or are sensitive to, the kind of parameter added to the more complex model for meaningful interpretation of effect size (Rights & Sterba, 2021). In the current study, we compare models on marginal R-squared values, which represent the explained variance of the combined fixed predictors of the model out of the total outcome variance.

When the linear predictor of time demonstrated random slopes within participants, we subsequently tested whether these interindividual differences in linear time slopes were moderated by trait burnout. For these models, the main effects for the linear function of time and trait burnout were included in the model as well as the interaction between time and trait burnout. Moderators were defined in accordance with the item dimension (i.e., trait exhaustion was included as the moderator when the outcome was momentary exhaustion) and centred at the grand mean.²

Results

Variations in burnout-related experiences

The final sample completed on average 76.97% (SD = 16.37, range = 31.43% − 100%) of all ESM questionnaires throughout the workweek (an overview of the descriptive statistics is available in Table 1). However, only 36.88% (SD = 15.71, range = 8.57% − 71.43%) of these ESM questionnaires were completed during working moments on days with at least three observations, and thus included in the final analysis. After excluding (a total of 68) days when participants did not complete a minimum of three momentary questionnaires while working, we were left with 213 days during which participants completed an average of 3.94 (SD = .92, range = 3–7) momentary questionnaires per day during moments when they were working.

Table 1. Descriptive statistics of the main parameters of interest.

Variable	M (SD)	Range	Skewness	Kurtosis	W (p-value)
Momentary exhaustion	1.90 (1.02)	1–7	1.40	2.17	.83 (<.01)
Momentary exhaustion (log)	.52 (.48)	0–1.95	.47	−.80	.88 (<.01)
Momentary positive challenge	4.31 (1.63)	1–7	−.32	−.62	.94 (<.01)
Momentary boredom	2.36 (1.43)	1–7	1.03	.38	.84 (<.01)
Momentary boredom (log)	.69 (.59)	0–1.95	0.19	−1.19	.87 (<.01)

The percentage of variance in momentary exhaustion, positive challenge, and boredom at the three levels of assessment is presented in Table 2 (for the complete null models, please see Appendix 1). For all three experiences, a large proportion of the variance occurred between moments, and the least occurred between days. Hence, all burnout-related experiences fluctuated substantially within days and between persons but remained relatively stable between measurement days.

Table 2. Percentage of variance in the dependent variable occurring between persons, days, and moments.

	Between persons	Between days	Between moments
Log momentary exhaustion	50% (<.01)	4% (.012)	46%
Momentary positive challenge	34% (<.01)	6% (<.01)	60%
Log momentary boredom	30% (<.01)	2% (.319)	68%

Note: p-value for the random intercepts within and between days in brackets. The values for the momentary assessment level (i.e., between moments) are expressed as the percentage of the residual variance in the dependent variable.

We examined the within- and between-person correlations among the study variables (see Table 3 for a correlation matrix). Average levels of momentary exhaustion and boredom showed moderate to strong correlations with both the exhaustion (momentary exhaustion: r = .57 p < .001; momentary boredom: r = .29 p = .020) and disengagement (momentary exhaustion: r = .37, p = .002; momentary boredom: r = .36 p = .003) subscales of the OLBI. Moreover, how positively challenged participants felt on average showed no significant relationships with either of the OLBI subscales (Exhaustion: r = .06 p = .637; Disengagement: r = −.16, p = .206), suggesting that feeling positively challenged on a momentary basis was, on average, unrelated to trait disengagement. Within-person correlations revealed that momentary and daily exhaustion was unrelated to momentary and daily experiences of feeling positively challenged (momentary: r = .01 p =.693, daily: r = .06 p = .497), and weakly correlated with boredom (r = .15 p < .001), but only at the momentary level (and not at the day level; r = .05 p = .564). Moreover, boredom demonstrated moderate average correlations with feeling positively challenged within-persons, both at the daily (r = −.36 p < .001) and momentary level (r = −.38 p < .001).

Table 3. Within- and between-person correlations between parameters of interest.

	1.	2.	3.	4.	5.
1. Trait exhaustion	-
2. Trait disengagement	.52***	-
3. Momentary exhaustion	.57*** (.60***)	.37** (.40***)	-	.01 (.06)	.15*** (.05)
4. Momentary positive challenge	.06 (.06)	−.16 (−.16)	.29* (.32*)	-	−.38*** (−.36***)
5. Momentary boredom	.29* (.37**)	.36** (.42***)	.40*** (.55***)	−.23 (−.21)	-

Note: * = p<.05, ** = p<.01, *** = p<.001, correlations below the diagonal are at the between-person level (n = 65); Correlations in brackets reflect correlations with the daily aggregates of the momentary variables; n = 213 for within-person between-day (i.e., daily level) correlations; n= 839 for within-person within-day (i.e., momentary level) correlations.

Between-day fluctuations in experiences related to burnout

Table 4 presents the results from the ANOVA, run on the output from the multilevel models investigating between-day fluctuations based on day of the week (Monday – Friday). As can be seen from the table, the analysis revealed no main effect for exhaustion (F = .88, p = .476), boredom (F = 1.47, p = .214) or feeling positively challenged (F = 2.14, p = .079), indicating that day of the week did not significantly impact daily burnout-related experiences. This suggests that these specific experiences did not fluctuate systematically as a function of week day in the current sample (see Figure 1 for a figure of momentary experiences by day of the week). Inspection of the pairwise comparisons using Tukey corrections for multiple comparisons also revealed that the burnout-related experiences assessed in the current study did not differ significantly from one another on any of the days of the week (for all: p > .05) (see Appendix 2 for a table of the post hoc contrasts).

Figure 1. Daily burnout-related experiences as a function of day of the week.

Note: Violin plots showing the mean (blue dot), boxplot (white box), and distributions (shaded grey area) of the predicted experiences related to burnout (y-axis) per day of the week (x-axis). The boxplot denotes the median (black middle line), interquartile range (white box), ±1.5*interquartile range (black whiskers), and outliers (black dots).

Table 4. Fluctuations in burnout-related experiences as a function of day of the week.

	Exhaustion (log)			Positive Challenge			Boredom
	F	df	p	F	df	p	F	df	p
Day of the week	.88	149.9	.476	2.14	151.7	.079	1.47	154.7	.214
Level 2 Variance (Person)	.12			.99			.54
Level 1 Variance (Within-day)	.04			.50			.38
AIC	115.03			603.62			532.19
BIC	138.56			627.15			555.72
LogLik (df)	−50.52 (7)			−294.81 (7)			−259.10 (7)
R^2 (Conditional)	.74			.67			.59
R^2 (Marginal)	<.01			.02			.01
W	.98			.98			.99

Note: all models used 213 observations, AIC = Akaike Information Criterion, BIC = the Bayesian Information Criterion, LogLik= log-likelihood, R² (Conditional) = explained variance of the complete model, R² (Marginal) = explained variance of the predictors, W = Shapiro-Wilk test statistic.

Time-dependent fluctuations in burnout-related experiences

Table 5 shows the results from the multilevel models with momentary exhaustion as the dependent variable regressed on the different time metrics according to linear and curvilinear functions. The linear and parabolic functions of all three time-related metrics were significant (all p < .05), suggesting that exhaustion increases exponentially throughout the day (see Figure 2). In contrast, the third-degree polynomial function was not significantly related to momentary exhaustion for all time metrics, and including it did not improve respective model fits (all p > .05), implying that, on average, exhaustion did not decrease after the increase indicated by the parabolic function. The chi-square test comparing the parabolic function of all three time metrics revealed no statistically significant differences between the models (all p > .05), suggesting that the predictive strength of all three time metrics was equal. However, the explained variance of all three time-related metrics was low, indicating that time only accounted for approximately 3% of the variance in momentary exhaustion. The time profiles suggest that momentary exhaustion increased slightly at the beginning of the day, followed by a moderate increase in the evening.

Figure 2. Momentary exhaustion as a function of time metrics.

Note: Log momentary exhaustion as a parabolic function of local clock time, internal time (hours elapsed since preferred midsleep), and time since awakening. Note that the regression line does not account for the multilevel structure in the data.

Table 5. Time dependent variations in (log) momentary exhaustion.

	Local Clock Time			Internal Time			Time Since Awakening
Time	.023*** (<.01)	−.057* (.03)	−.111 (.17)	.022*** (<.01)	−.037 (.02)	.002 (.08)	.022*** (<.01)	−.015 (.01)	−.036 (.04)
Time^2		.003** (<.01)	.007 (.01)		.003**(<.01)	−.001 (.01)		.003** (<.01)	.006 (.01)
Time^3			<-.001 (<.01)			<.001 (<.01)			<-.001 (<.01)
Intercept	.20** (.07)	.727*** (.19)	.968 (.75)	.289*** (.06)	.566*** (.11)	.445 (.26)	.359*** (.05)	.468*** (.06)	.506*** (.09)
Level 3 Variance (Person)	.12	.12	.12	.12	.12	.12	.12	.12	.12
Level 2 Variance (Day)	.01	.01	.01	.01	.01	.01	.01	.01	.01
Level 1 Variance (Within-day)	.10	.10	.10	.10	.10	.10	.10	.10	.10
AIC	691.69	697.44	713.93	692.94	697.22	713.84	692.89	699.01	715.71
BIC	715.35	725.83	747.06	716.60	725.61	746.96	716.55	727.40	748.84
LogLik (df)	−340.84 (5)	−342.72 (6)	−349.97 (7)	−341.47 (5)	−342.61 (6)	−349.92 (7)	−341.44 (5)	−343.50 (6)	−350.86 (7)
R^2 (Conditional)	.571	.576	.576	.576	.582	.582	.576	.581	.581
R^2 (Marginal)	.021	.027	.027	.022	.028	.028	.022	.027	.027
W	.963	.962	.962	.963	.962	.962	.964	.962	.962

Note: * = p<.05, ** = p<.01, *** = p<.001, all models used 839 observations, standard error noted in parentheses, AIC = Akaike Information Criterion, BIC = the Bayesian Information Criterion, LogLik= log-likelihood, R² (Conditional) = explained variance of the complete model, R² (Marginal) = explained variance of the predictors, W = Shapiro-Wilk test statistic.

Including random slopes for the linear component significantly improved model fits (Local clock time: X² = 42.64, p < .001; Internal time: X² = 42.37, p < .001; Time since awakening: X² = 42.54, p < .001) indicating that the relationship between time of day and momentary exhaustion differs between persons (Local clock time: slope variance = 0.002, SD = 0.04; Internal time: slope variance = 0.002, SD = 0.04; Time since awakening: slope variance = 0.002, SD = 0.04).

In contrast to exhaustion, none of the tested time metrics or patterns were substantially related to either disengagement item (see Tables 6 and 7 for parameter estimates, Appendix 3 for t-values and df), suggesting that there was, at the group level, no systematic time dependency in momentary positive challenge or boredom according to a linear, parabolic or third-degree function. Although boredom was significantly related to the linear function of local clock time and internal time, the explained variance was less than 0.5%, and is thus negligible. Including random slopes for the linear function of each time metric did not improve model fits for either disengagement-related item (Positive challenge: Local clock time: X² = 2.02, p = .365; Internal time: X² = 1.79, p = .408; Time since awakening: X² = 1.78, p = .411; Boredom: Local clock time: X² = 3.41, p = .182; Internal time: X² = 3.03, p = .220; Time since awakening = X² = 1.93, p = .380), indicating that the relationship between disengagement-related experiences and time did not differ significantly between persons.

Table 6. Time dependent variations in momentary positive challenge.

	Local Clock Time			Internal Time			Time Since Awakening
Time	−.021 (.02)	.034 (.11)	.471 (.67)	−.022 (.02)	.030 (.08)	.451 (.32)	−.020 (.02)	.037 (.06)	.179 (.16)
Time^2		−.002 (<.01)	−.033 (.05)		−.003 (<.01)	−.044 (.03)		−.004 (<.01)	−.025 (.02)
Time^3			.001 (<.01)			.001 (<.01)			.001 (<.01)
Intercept	4.632***(.24)	4.273*** (.78)	2.325 (3.06)	4.569*** (.20)	4.326*** (.41)	3.031** (1.05)	4.484***(.17)	4.320*** (.23)	4.058*** (.35)
Level 3 Variance (Person)	.943	.943	.940	.939	.937	.934	.937	.938	.935
Level 2 Variance (Day)	.152	.151	.152	.152	.151	.154	.154	.152	.154
Level 1 Variance (Within-day)	1.645	1.647	1.648	1.645	1.647	1.644	1.645	1.646	1.645
AIC	2996.59	3007.51	3020.87	2996.36	3007.22	3019.50	2996.79	3007.04	3020.32
BIC	3020.25	3035.91	3054.00	3020.02	3035.62	3052.63	3020.45	3035.43	3053.45
LogLik (df)	−1493.30 (5)	−1497.76 (6)	−1503.44 (7)	−1493.18 (5)	−1497.61 (6)	−1502.75 (7)	−1493.40 (5)	−1497.52 (6)	−1503.16 (7)
R^2 (Conditional)	.401	.400	.400	.400	.399	.400	.400	.400	.400
R^2 (Marginal)	.002	.002	.002	.002	.002	.003	.001	.002	.003
W	.997	.997	.997	.997	.997	.997	.997	.997	.997

Note: * = p<.05, ** = p<.01, *** = p<.001, all models used 839 observations, standard error noted in parentheses, AIC = Akaike Information Criterion, BIC = the Bayesian Information Criterion, LogLik= log-likelihood, R² (Conditional) = explained variance of the complete model, R² (Marginal) = explained variance of the predictors, W = Shapiro-Wilk test statistic.

Table 7. Time dependent variations in (log) momentary boredom.

	Local Clock Time			Internal Time			Time Since Awakening
Time	.013* (.01)	−.021 (.04)	−.362 (.25)	.011* (.01)	−.024 (.03)	−.167 (.12)	.010 (<.01)	−.017 (.02)	−.081 (.06)
Time^2		.001 (<.01)	.026 (.02)		.002 (<.01)	.016 (.01)		.002 (<.01)	.011 (.01)
Time^3			−.001 (<.01)			<-.001 (<.01)			<-.001 (<.01)
Intercept	.518*** (.09)	.739* (.29)	2.258* (1.14)	.576*** (.07)	.741*** (.15)	1.182** (.39)	.622*** (.06)	.700*** (.08)	.817*** (.13)
Level 3 Variance (Person)	.106	.105	.105	.107	.107	.106	.107	.106	.106
Level 2 Variance (Day)	.009	.009	.009	.008	.009	.009	.009	.009	.009
Level 1 Variance (Within-day)	.231	.231	.231	.232	.231	.231	.232	.231	.231
AIC	1315.81	1328.31	1342.18	1316.78	1328.58	1343.13	1317.83	1329.36	1344.20
BIC	1339.47	1356.71	1375.30	1340.44	1356.97	1376.26	1341.49	1357.75	1377.32
LogLike (df)	−652.91 (5)	−658.16 (6)	−664.09 (7)	−653.39 (5)	−658.29 (6)	−664.56 (7)	−653.91 (5)	−658.68 (6)	−665.10 (7)
R^2 (Conditional)	.333	.333	.334	.336	.337	.336	.335	.336	.335
R^2 (Marginal)	.004	.005	.006	.004	.005	.006	.003	.005	.005
W	.988	.988	.989	.988	.988	.989	.988	.988	.989

Note: * = p<.05, ** = p<.01, *** = p<.001, all models used 839 observations, standard error noted in parentheses, AIC = Akaike Information Criterion, BIC = the Bayesian Information Criterion, LogLik= log-likelihood, R² (Conditional) = explained variance of the complete model, R² (Marginal) = explained variance of the predictors, W = Shapiro-Wilk test statistic.

Trait exhaustion did not moderate the relationship between time and momentary exhaustion for any of the metrics of time assessed in the current study (all p > .05) (see Table 8). Note that the number of observations was limited to the person level in these analyses, meaning there was insufficient power to detect small effect sizes.

Table 8. Cross-level moderation by trait burnout.

	Local clock time	Internal time	Time since awakening
Time	.02 (.01)***	.02 (.01)***	.02 (.01)***
Trait Exhaustion	.19 (.21)	.27 (.17)	.33 (.13)*
Interaction	.02 (.02)	.02 (.02)	.02 (.02)
Intercept	.17 (.09)	.27 (.07)***	.35 (.05)***
Random effects
Level 3 Variance (Person)	.330	.214	.136
Time	.002	.002	.002
Level 2 Variance (Day)	.013	.013	.014
Level 1 Variance (Within-day)	.084	.084	.084
AIC	644.44	645.25	646.58
BIC	687.03	687.84	689.17
LogLik (df)	−313.22 (9)	−313.63 (9)	−314.29 (9)
R^2 (Conditional)	.66	.66	.67
R^2 (Marginal)	.20	.19	.20
W	.97	.97	.97

Note: * = p<.05, ** = p<.01, *** = p<.001, all models used 839 observations, standard error noted in parentheses, AIC = Akaike Information Criterion, BIC = the Bayesian Information Criterion, LogLik= log-likelihood, R² (Conditional) = explained variance of the complete model, R² (Marginal) = explained variance of the predictors, W = Shapiro-Wilk test statistic.

Discussion

This study aimed to investigate time profiles of exhaustion, boredom, and feeling positively challenged within and between days. Moreover, we aimed to explore if trait levels of burnout influenced the temporal pattern in these momentary experiences. Results demonstrated that a large proportion of the variance in exhaustion, boredom, and feeling positively challenged occurred between moments, with an approximately equal amount of variance occurring between persons. These results align with a study conducted by Sonnenschein et al. (2007), which found that a substantial part of the variance in their one-item exhaustion measure occurred within persons. The large proportion of variance between moments demonstrates that these specific momentary burnout-related experiences fluctuate substantially within days and thus supports a shift away from the conceptualization of burnout purely as a stable trait. Nevertheless, results showed that the overall mean levels of burnout-related experiences varied substantially between persons – showing that some participants felt more positively challenged, bored, and exhausted than others.

Moreover, the current study showed that exhaustion, boredom, and feeling positively challenged remained relatively stable between workdays. A negligible amount of variance occurred between days for all momentary experiences assessed in this study, none of which varied substantially depending on day of the week. This lack of entrainment to days of the week is in line with a previous study by Mäkikangas et al., (2014), demonstrating that for most participants exhaustion did not increase or decrease across the five-day work week (Mäkikangas et al., 2014). Nevertheless, these findings contradict research on closely-related constructs such as fatigue and mood that have shown entrainment to days of the week, fluctuating in accordance with the anticipation of leisure time and work(load), respectively. Fatigue has been demonstrated to increase abruptly on Monday (when it is at it’s peak) remain relatively stable, and then decrease abruptly on Friday due to the anticipation of leisure time (Weigelt et al., 2021). Moreover, positive and negative affect have been shown to increase and decrease, respectively, in-line with the hypothesis that employees experience the “Monday blues” due to the anticipation of work tasks or stressors employees will be faced with throughout the week, which decreases with the approaching weekend (Hülsheger et al., 2022; Larsen & Kasimatis, 1990). Taken together with the result that a substantial amount of variance occurred within days and between persons, our results on weekly trajectories suggests that while exhaustion, boredom, and feelings of being positively challenged vary from moment to moment, and person to person, averaged daily levels of exhaustion, boredom, and feeling positively challenged are relatively stable within-persons. This stability between days may be explained by the fact that while fluctuations within-days are normal and to be expected, our sample of healthy participants was able to recover adequately between work days, thus starting the next workday with psychophysiological systems restored to baseline. Indeed, according to the Effort-Recovery theory, changes in mean levels between days should only occur if recovery from effort expended throughout the workday is inadequate (Meijman & Mulder, 1998). It is also possible that burnout-related experiences only appeared stable due to large differences between persons. Some participants may have experienced an increase in the intensity of burnout-related experiences throughout the workweek, while others experienced a decrease, which resulted in an average null effect. Unfortunately, the sample size of this study did not provide sufficient power to test for such interindividual differences across days using random slope models.

The diurnal pattern of momentary experiences related to burnout

Our study uncovered that, following a minimal increase during the beginning of the day, exhaustion began to increase more substantially towards the end of the day in a U-shaped pattern, similar to energy-related constructs such as sleepiness (Åkerstedt et al., 2013), fatigue (Hülsheger, 2016; Sonnenschein et al., 2007) and exhaustion as measured by Sonnenschein et al. (2007). This pattern suggests that exhaustion among non-clinical (i.e., seemingly healthy) populations intensifies over time. We should note that this time-dependent nature was modelled with equally modest predictive strengths for all three time metrics. This suggests that interindividual and daily variations in sleep-wake patterns do not substantially impact the structural temporal pattern of momentary experiences of exhaustion.

Although the linear relationship between time and exhaustion differed significantly between persons, as displayed in statistically significant random slope variance, trait exhaustion did not moderate the linear relationship between time and exhaustion. This shows that scoring high vs. low on trait exhaustion did not substantially impact the trajectory of momentary exhaustion throughout the day. However, given that these analyses were limited to the person level (N = 65), the current study did not have sufficient power to detect medium to small effect sizes, and conclusions cannot be drawn regarding whether trajectories are more subtly impacted by trait burnout. Future studies are needed to investigate the role of trait burnout in time-of-day dependent trajectories of exhaustion with larger sample sizes.

Exhaustion- and disengagement-related experiences did not fluctuate in similar structural patterns. In contrast, neither of the items assessing momentary experiences related to disengagement showed meaningful structural patterns within days as a function of any of the included time metrics. This is contrary to studies on various mood-related constructs that have been demonstrated to fluctuate in u-shaped (for valence and tension; Zhang et al., 2018), as well as decreasing (positive affect; Golder & Macy, 2011) and increasing (pleasantness, negative affect; Egloff et al., 1995; Golder & Macy, 2011) linear patterns throughout the day. Although most of the variance in feeling positively challenged and bored occurred between moments, this variance was almost entirely unexplained by time. Boredom and feeling positively challenged may vary in the absence of a clear temporal pattern because these experiences are largely task-related. This is in line with episodic engagement research demonstrating that engagement varied with activity (Reina-Tamayo et al., 2017). Moreover, the Effort-Recovery Theory states that the lack of willingness to invest effort (i.e., a lack of motivation or disengagement) is dependent on both the psycho-physiological state of the individual and the demands of the current task. Thus, among non- and sub-clinical populations, feeling bored and positively challenged may depend largely on the characteristics of the current activity.

The mechanism behind the parabolic increase in exhaustion over time may be due to a complex interaction of several processes. From an organizational psychology perspective, this increase is in line with the Effort-Recovery Theory, which would attribute it to an accumulation of load effects (Meijman & Mulder, 1998). Investing compensatory effort to offset the exhaustion arising from expending effort to complete work tasks could trigger an energetic loss spiral within days. Thus, to reduce the within-day accumulation of exhaustion and prevent the potential development of chronic burnout, features of the work environment (e.g., workload or recovery possibilities) should be addressed. Additionally, from a chronobiological perspective, circadian and homoeostatic processes may regulate feelings of exhaustion throughout the day. In this view, exhaustion increases regardless of what occurs within the day and the amount of effort expended but rather due to increasing sleep pressure and/or decreasing circadian drive. The almost linear increase throughout the beginning of the day in the current study suggests a homoeostatic component (i.e., an increase with time spent awake) as seen with sleepiness (Borbely, 1982). On the other hand, the more substantial increase towards the end of the day may suggest an additional influence of circadian rhythms in exhaustion, whereby the experience of momentary exhaustion is regulated depending on the individual’s circadian phase (i.e., the timing of an individual’s circadian clock). This is in line with results from studies on related constructs, such as fatigue, sleepiness, alertness, and mood, that have demonstrated circadian rhythmicity in constant routine laboratory studies (Daurat et al., 1993; Rüger et al., 2006). According to this view, preventing chronic exhaustion would require interventions that support circadian alignment, such as improving light exposure and sleep-wake patterns. This also aligns with indications from cross-sectional studies demonstrating that circadian misalignment, or the mismatch between internal and external time, is related to higher levels of trait burnout (Cheng & Hang, 2018; Mokros et al., 2018; Waleriańczyk et al., 2019).

The explanations offered by the organizational and chronobiological perspectives are not mutually exclusive. In fact, it is likely the combination of both effort-driven (organizational perspective), as well as circadian and homoeostatic processes (chronobiological perspective), that determine the daily trajectory of exhaustion. In line with the suggestion of the Effort-Recovery Theory that as the day goes on, resources that were initially available are spent (Meijman & Mulder, 1998), the amount of effort invested throughout the workday and time spent awake should increase in unison. However, there may be situations in which effort-driven and chronobiological processes are not aligned and thus interact to determine momentary exhaustion levels. Even when the drive to sleep is low (e.g., at noon), expending a large amount of effort on a difficult task may amplify levels of momentary exhaustion, thereby perhaps also influencing the trajectory of exhaustion for the remainder of the day because more compensatory effort will be required to accommodate the experienced resource loss. Thus, both time spent awake and circadian drive, as well as the effort invested on tasks or activities until that moment, are likely to interact to determine momentary states of exhaustion.

State and trait burnout: two separate constructs

Several results suggest that the state experiences related to burnout probed in this study and trait burnout may represent different constructs. Due to a lack of validated and reliable momentary burnout questionnaires, the items used to assess momentary experiences related to burnout via ESM were selected based on factor loadings of items used in a previous diary study and translated to apply to a momentary assessment level. The low correlations between items of the resulting exhaustion and disengagement scales imply that factor loadings may change substantially when transitioning from a daily to momentary assessment level, suggesting a lack of cross-level measurement invariance, i.e., that the latent construct of burnout differs depending on the time scale of the assessment. This is in line with results from a diary study examining the factorial invariance of a daily translation of the OLBI that found large differences in factor loadings within- and between- participants (Gruszczynska et al., 2021). Especially the disengagement items were unrelated on a momentary basis. This suggests that items assessing disengagement on a trait level, cannot be generalized to items measuring it on the momentary level. While the two items measuring disengagement (boredom and feeling positively challenged) seem to assess the same theoretical construct when participants reflect on work in general (trait level), being more bored in one moment was only moderately associated with feeling less positively challenged at that same moment within-persons. Perhaps feelings of boredom and being positively challenged are caused by different proximal predictors (e.g., different features of the work environment). Moreover, although the current study found moderate relationships between average momentary (state) and general (trait) assessments of burnout, the explained variance was only 13%-32% (for boredom and exhaustion respectively), indicating that a considerable amount of variance in trait levels of disengagement and exhaustion is unexplained by these particular momentary experiences. Similar effect sizes were reported by Sonnenschein et al. (2007) between the MBI exhaustion scale and average momentary exhaustion in healthy participants and those on partial sick leave. Thus, trait and state-level questionnaires and experiences may provide information on related but separate constructs. Future studies are needed to clarify the theoretical constructs of state disengagement and exhaustion and to specify which momentary experiences are central to these definitions.

Theoretical and practical implications

As argued in the introduction, using the temporal approach to examine momentary experiences related to the exhaustion and disengagement component of burnout provides a first step in understanding the type of temporal dynamics that can be expected in such experiences. Thus, we contribute to understanding the “typical” ebb and flow of momentary exhaustion, boredom, and feeling positively challenged. Dalal et al. (2014) defined three types of within-person variability: linear or non-linear “growth” curves, cycles, and discontinuous variability, i.e., sudden changes in direction and/or magnitude. While this study is rather descriptive in establishing only the first two types of within-person variability within days for a limited set of burnout-related experiences, we aimed to stimulate research into state burnout-experiences by demonstrating that meaningful fluctuations (and differences in fluctuations) exist in these experiences within-persons. Specifically, the results of this study indicate that a large amount of variance can be expected in exhaustion, boredom, and feeling positively challenged on a momentary basis – perhaps suggesting that a lack of variance in these experiences (i.e., stability) is atypical and may even signal the development of burnout. Moreover, in healthy participants, exhaustion seems to fluctuate in a parabolic pattern throughout the day (albeit accounting for only a modest amount of variance in momentary exhaustion), implying that the absence or flattening of such a pattern may be indicative of energetic resource depletion. This is in line with results from de Longis et al. (2022) and Sonnenschein et al. (2007), who demonstrated that participants experiencing high levels of chronic exhaustion and clinical burnout exhibited less variability in negative affect and fatigue, respectively. Based on our findings, we suggest that future studies should investigate environmental and social determinants that cause changes in time profiles in order to develop temporally precise theories at the within-person level.

The results of this study are also of practical relevance. The structural fluctuations, demonstrated in momentary exhaustion, albeit modest, are associated with the intensity of momentary burnout-related experiences. Thus, they should be considered in future studies as relationships between exhaustion and predictors of interest may partially depend on the timing of such predictors. Similarly, when monitoring momentary experiences or designing interventions aimed at reducing or preventing exhaustion, boredom, and a lack of feeling positively challenged, researchers and practitioners should keep in mind that the time of administering an assessment or intervention may have consequences for the results. Moreover, as these constructs vary substantially from moment to moment, also according to as yet unexplained dynamics, assessing these experiences repeatedly throughout the day is essential for researchers and organizations interested in obtaining accurate daily indicators of employee burnout. Lastly, detecting temporal patterns also has implications for job design, suggesting that the availability of energetic resources partially depends on the time of day. Perhaps more demanding tasks should thus be completed before noon when more energetic resources are typically available.

Limitations

Our approach to exploring time-dependent moderations in burnout-related experiences requires further consideration. Given the low correlations between the two items of the momentary disengagement scale, the selected items did not reliably capture the construct of disengagement as defined at the trait level, implying that current results based on the selected items may not be generalizable to the construct of momentary disengagement. The validation of questionnaires assessing the momentary experience of burnout is thus required to support research into within-day processes. Particularly, more research is needed to assess time dependency in other momentary experiences related to disengagement and the minor time dependency in momentary exhaustion. Additionally, it would be worthwhile to investigate how well the findings from this study extend to other populations, such as clinical burnout patients, extreme chronotypes, and persons with sleep disturbances. In fact, the relatively low levels of momentary exhaustion and boredom in our sample may have had implications for detecting temporal patterns and predictive strengths insofar that a sample experiencing higher levels of momentary exhaustion or boredom may have rendered different results. Exhaustion seems to be a more extreme form of fatigue that is likely less common amongst healthy participants, potentially resulting in floor effects (Sonnenschein et al., 2007), which was also reflected in the smaller effect sizes compared to growth models of fatigue (Hülsheger, 2016; Sonnenschein et al., 2007). It is also possible that participants only completed the ESM questionnaires when their schedule and mental capacity allowed for this, thus missing moments during which the intensity of these experiences was greater. This is in line with the relatively low number of ESM questionnaires completed during working moments.

Moreover, due to the one-week sampling period and the restricted sample size, there was insufficient power for between-day and person-level analyses to detect small effect sizes. Importantly, according to guidelines presented by Arend and Schäfer (2019)for two-level models, we could only detect large standardized effect sizes for the cross-level interaction.³ We thus cannot draw conclusions regarding whether trait burnout more subtly amplifies or weakens the within-day increase in exhaustion. Moreover, it would be worthwhile for future studies to investigate whether trait exhaustion moderates the non-linear functions for time within-days, as well as the relationship between daily burnout and day of the week which was not possible for this study due to the size of the dataset. Future studies with larger samples, longer sampling periods, and perhaps more assessments per day are needed for these investigations, and to obtain more conclusive results regarding small effect sizes.

We relied on self-report measures, with no objective measures of time of awakening and internal time. Future studies can aim to replicate these findings with more objective measures, such as actigraphy (to assess the time of awakening) and dim light melatonin onset (as a biomarker for internal time) to avoid inducing biases via introspection.

Conclusion

The current study contributes to burnout literature by demonstrating that, typically, exhaustion, boredom, and feelings of being positively challenged fluctuate substantially within days, with momentary exhaustion showing structural entrainment to time of day, regardless of the conceptualization of time (i.e., local clock time, internal time, and time since awakening). In contrast, boredom and feeling positively challenged did not demonstrate such structural time dependencies. Furthermore, given the relatively small effect sizes reported in the current study, much of the variance in these specific momentary experiences remains to be explained. More research on momentary burnout-related experiences is needed to understand the concept of state burnout experiences and their relation to the development of trait burnout.

Disclaimer

The views expressed in the submitted article are the authors own and not an official position of the institution they work for.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. Items were not centred at the person-mean. Momentary correlations are based on complete dataset.

2. Given the convergence and power issues when testing random slopes for the non-linear patterns, we did not explore moderations for the higher-order terms for time.

3. It should be noted that Arend & Schäfer’s (2019) guidelines were developed for two-level models, whereas this study employed three-level models, making this is a rough estimate of the effect sizes this study was able to detect.

Appendix 1.

Null models

Table

	Exhaustion (log)	Positive Challenge	Boredom (log)
Intercept	.50*** (.04)	4.36*** (.13)	.69*** (.04)
Level 3 Variance (Person)	.115	.938	.105
Level 2 Variance (Day)	.010	.154	.008
Level 1 Variance (Within-day)	.105	1.646	.233
AIC	716.97	2989.96	1310.28
BIC	735.90	3008.89	1329.21
LogLik (df)	−354.49 (4)	−1490.98 (4)	−651.14 (4)
R^2 (Conditional)	.54	.40	.33
R^2 (Marginal)	<.01	<.01	<.01
W	.95	.996	.99

Note: * = p < .05, ** = p < .01, *** = p < .001, all models used 839 observations, standard error noted in parentheses, AIC = Akaike Information Criterion, BIC = the Bayesian Information Criterion, LogLik= log-likelihood, R² (Conditional) = explained variance of the complete model, R² (Marginal) = explained variance of the predictors, W = Shapiro-Wilk test statistic.

Appendix 2.

Post hoc comparisons for daily analysis

Table

Post hoc comparisons	Mean difference	SE	df	t-value	p-value
Daily Exhaustion (log)
Monday – Tuesday	−0.07	0.04	150	−1.54	0.538
Monday- Wednesday	−0.04	0.05	150	−0.92	0.888
Monday – Thursday	−0.07	0.05	148	−1.43	0.612
Monday – Friday	−0.07	0.05	151	−1.46	0.592
Tuesday – Wednesday	0.02	0.05	151	0.50	0.987
Tuesday – Thursday	<-0.01	0.05	150	−0.04	1.000
Tuesday – Friday	−0.01	0.05	151	−0.16	1.000
Wednesday – Thursday	−0.02	0.05	149	−0.50	0.988
Wednesday – Friday	−0.03	0.05	150	−0.59	0.977
Thursday – Friday	−0.01	0.05	150	−0.11	1.000
Daily Positive Challenge
Monday – Tuesday	0.05	0.15	151	0.37	1.00
Monday – Wednesday	−0.34	0.16	152	−2.16	0.20
Monday – Thursday	−0.23	0.16	150	−1.43	0.61
Monday – Friday	−0.04	0.17	153	−0.25	1.00
Tuesday – Wednesday	−0.39	0.16	153	−2.51	0.09
Tuesday – Thursday	−0.28	0.16	151	−1.78	0.39
Tuesday – Friday	−0.10	0.17	153	−0.57	0.98
Wednesday – Thursday	0.11	0.17	150	0.64	0.97
Wednesday – Friday	0.30	0.18	152	1.67	0.46
Thursday – Friday	0.19	0.18	152	1.04	0.84
Boredom
Monday – Tuesday	−0.10	0.13	153	−0.79	0.93
Monday – Wednesday	0.18	0.14	154	1.29	0.70
Monday – Thursday	−0.03	0.14	152	−0.22	1.00
Monday – Friday	0.16	0.15	156	1.10	0.81
Tuesday – Wednesday	0.28	0.14	156	2.03	0.26
Tuesday – Thursday	0.07	0.14	154	0.49	0.99
Tuesday – Friday	0.27	0.15	156	1.79	0.38
Wednesday – Thursday	−0.21	0.15	153	−1.41	0.62
Wednesday – Friday	−0.01	0.16	155	−0.08	1.00
Thursday – Friday	0.20	0.16	154	1.23	0.73

Appendix 3.

t-values and df for time dependent variations in momentary burnout-related experiences

Table

Predictor	t-value	df	t-value	df	t-value	df	t-value	df
Exhaustion (log)
Local clock time
Intercept	3.06	274.14	3.75	756.05	1.29	693.87	1.95	56.18
Time	6.14	705.50	−2.04	709.85	−0.67	691.02	3.90	52.65
Time^2			2.87	715.62	0.58	693.78
Time^3					−0.33	697.59
Trait Exhaustion							0.88	54.96
Time × Trait Exhaustion							1.42	51.15
Internal Time
Intercept	5.03	163.35	5.38	662.60	1.72	724.64	3.82	58.16
Time	6.03	711.59	−1.94	698.51	0.03	693.19	3.80	52.67
Time^2			3.15	706.16	−0.13	696.49
Time^3					0.51	700.61
Trait Exhaustion							1.61	55.87
Time × Trait Exhaustion							1.39	51.20
Time since awakening
Intercept	7.02	105.24	7.33	234.29	5.53	571.01	6.42	60.84
Time	6.04	718.89	−1.11	710.88	−0.94	712.84	3.77	53.14
Time^2			2.84	724.15	1.07	717.63
Time^3					−0.59	725.85
Trait Exhaustion							2.56	58.50
Time × Trait Exhaustion							1.46	51.20
Positive Challenge
Local clock time
Intercept	19.29	447.38	5.46	756.01	0.76	711.06
Time	−1.39	723.68	0.30	728.22	0.70	710.13
Time^2			−0.48	734.13	−0.70	712.71
Time^3					0.66	716.37
Internal time
Intercept	23.22	259.89	10.53	755.41	2.88	729.18
Time	−1.47	734.59	0.38	717.35	1.39	712.04
Time^2			−0.67	724.86	−1.41	714.90
Time^3					1.34	718.72
Time since awakening
Intercept	27.15	143.69	19.00	395.63	11.56	730.99
Time	−1.32	739.24	0.66	728.58	1.15	730.45
Time^2			−1.05	742.65	−1.15	734.77
Time^3					0.98	742.54
Boredom (log)
Local clock time
Intercept	5.97	523.78	2.54	754.80	1.98	710.98
Time	2.24	727.65	−0.49	734.48	−1.44	711.13
Time^2			0.79	742.39	1.44	714.59
Time^3					−1.38	719.47
Internal time
Intercept	8.19	310.04	4.88	761.38	3.02	726.30
Time	2.02	739.60	−0.82	719.68	−1.38	713.36
Time^2			1.22	729.89	1.36	717.14
Time^3					−1.22	722.21
Time since awakening
Intercept	10.73	167.82	8.53	467.33	6.33	759.80
Time	1.74	746.33	−0.83	734.04	−1.39	737.19
Time^2			1.34	752.60	1.39	743.30
Time^3					−1.17	753.62