Slower information processing speed is associated with persistent burnout symptoms but not depression symptoms in nursing workers

Abstract

Introduction:

Burnout and depression both occur with chronic work-related stress, and cognitive deficits have been found when symptom severity results in work disability. Less is known about cognitive deficits associated with milder symptoms among active workers, and few studies have examined whether cognitive deficits predict persistent burnout and depression symptoms. The goal of this study was to examine the association of information processing speed and executive function performance to burnout and depression symptoms at baseline and 12-month follow-up in a sample of actively working individuals (N = 372).

Method:

The design was prospective with laboratory cognitive data at baseline, and burnout and depressive symptoms assessed at baseline and monthly follow-ups. Information processing speed and executive functions were assessed in a task-switching paradigm, including single-task reaction time (RT), switching costs, and mixing costs. Burnout was assessed with the Exhaustion subscale of the Oldenburg Burnout Inventory and depression with the Patient Health Questionnaire-9.

Results:

Slower RT was modestly associated with higher levels of burnout symptoms both cross-sectionally and prospectively, but switching costs and mixing costs were not associated with burnout symptoms. None of the cognitive measures were associated with depression symptoms cross-sectionally or prospectively.

Conclusions:

Despite statistically significant findings of slowed RT in acute exhaustion-related burnout, the proportion of the variance accounted for in the models was small and did not predict clinically significant levels of distress. The absence of statistically significant findings for depression symptoms suggests the cognitive profile associated with the exhaustion dimension of burnout may be distinct from that of depression, which reflects a more heterogeneous symptomatology. Our data suggest that the clinical impact of burnout symptoms on actively working individuals is marginal; nonetheless, it is important to screen and intervene on burnout and depression symptoms in the workplace because they can lead to other forms of work impairment.

Stress can disrupt psychosocial function and cognition in many contexts, including the workplace (Birnbaum et al., 2010). Two common manifestations of workplace stress are burnout and depression. Burnout has been characterized as a condition of emotional exhaustion in reaction to the cumulative effects of stress, particularly workplace stress (Demerouti et al., 2001). Depression, in contrast, is characterized by the key symptoms of sadness and anhedonia, but also includes additional emotional, and somatic and cognitive symptoms that are heterogeneous across individuals (American Psychiatric Association, 2013). Although both burnout and depression are associated with chronic stress, burnout is theorized to be context-dependent to occupational stress, while depression is theorized to be context-free (Bakker et al., 2000). Common aspects of burnout and depression are that both can interfere with work ability in a number of ways (Hatch et al., 2018; Jain et al., 2013), and can be chronic when untreated, which makes persistent burnout and depression symptoms a threat to both work and well-being.

A brief context and history of burnout

Burnout may be a less familiar construct to neuropsychologists than depression; thus, we present a brief review. Burnout is a clinically relevant condition because it is associated with multiple adverse health outcomes, including heart disease (Toker et al., 2012) and diabetes (Melamed et al., 2006), and is a driver of work disability (Ahola et al., 2009). Burnout is recognized in the ICD-11 as a mental health condition (World Health Organization, 2018), and is compensable for disability in several European countries (Lastovkova et al., 2018). The original academic definition of burnout is often credited to Freudenberger, who reflected on the sense of emotional exhaustion experienced by those working in free clinics (Freudenberger, 1975). Research by Maslach and colleagues was early and influential in formalizing the assessment and theoretical framework of burnout (Maslach et al., 2001). This group defined burnout as composed of symptoms related to exhaustion, depersonalization, and reduced sense of personal accomplishment. As with Freudenberger’s definition, burnout symptoms were viewed as the result of job-related stressors that most often occurred in human service professions like nursing, medicine, and social work. Other researchers have posed alternative models of burnout involving both unique and overlapping dimensions (Demerouti et al., 2001; Shirom & Ezrachi, 2003), with a systematic review including multiple models of burnout identified the common construct across these models to be emotional exhaustion (Seidler et al., 2014). Although burnout measures are correlated with measures of depression and anxiety, leading some researchers to question whether it is a distinct construct (Bianchi et al., 2015), most research has validated its distinctiveness (Koutsimani et al., 2019). Prevailing models of burnout are in accordance that exhaustion-related burnout is largely caused by chronic exposure to psychosocial stressors related to the cognitive, emotional, and physical demands of work (Demerouti et al., 2001; Maslach et al., 2001; Shirom, 2003).

Burnout, depression and cognition

Persistent burnout and depression symptoms both are known to develop in the context of chronic psychosocial stress, but less is known about differences that may contribute at the level of the individual. One important individual difference may be cognitive function. With respect to burnout, several studies have found a relationship between clinically significant burnout symptoms and cognitive deficits. For instance, a systematic review of burnout and cognition found the largest and most consistent deficits in the cognitive domains of executive functions, attention, and memory (Deligkaris et al., 2014). Other studies, however, have found deficits in the domain of information processing speed (Oosterholt et al., 2014; Osterberg et al., 2009), which some have argued to be broadly influential to performance in other cognitive domains, including the domains of executive function, attention, and memory (Krail & Salthouse, 1994; Mella et al., 2015; Zaremba et al., 2019). The majority of studies on burnout and cognition have been cross-sectional and compared individuals with clinically significant burnout to healthy comparison participants, in which the burnout group was on leave from work due to the severity of their symptoms. In many of these studies, individuals in the burnout group perform worse than the non-clinical comparison groups (Oosterholt et al., 2012; Osterberg et al., 2009; Rydmark et al., 2006). The limitation of these studies is that the focus on impaired workers may be less generalizable to the emotional experiences and cognitive performances of actively working individuals.

Fewer studies have examined the relationship between burnout and cognition among individuals who are actively working. One such study compared a group with clinical burnout (n =33; 28 of whom were not working or on reduced time) to a working group without clinical symptoms but stratified into high (n = 29) and low (n = 30) levels of burnout symptoms (Oosterholt et al., 2014). The authors found differences between the clinical burnout group and the non-burnout group in generalized slowing of reaction time (RT), but not among executive functions. A few studies have found associations between burnout and cognition within a sample entirely composed of actively working individuals. One study of nursing home employees (N = 81), stratified participants by high vs. low exhaustion-related burnout scores, and found worse performance among the high burnout group on higher demand conditions of N-back and Stroop tasks (Diestel et al., 2013). Another study of working individuals across several occupational categories (N = 76) examined a task-switching paradigm, and found that higher exhaustion-related burnout scores were associated with slower RT on task switching trials, with a trend toward significance in non-switching RT trials as well (Gajewski et al., 2017). Studies of among working individuals generally find smaller and less consistent associations with cognitive deficits than among individuals whose burnout results in clinical impairment and work absence; however, studies that focus more specifically on the exhaustion dimension of burnout appear more likely to detect cognitive differences than studies examining burnout symptoms more broadly.

As with burnout, research on cognitive deficits in clinically severe depression has shown that individuals with Major Depressive Disorder (MDD) as a group tend to have worse performance relative to nondepressed controls on a number of neuropsychological measures (Lee et al., 2012; McDermott & Ebmeier, 2009; Snyder, 2013). Similar to findings with clinical burnout, the largest and most consistent associations are found in the domains of executive functions and information processing speed (Snyder, 2013), and studies find that cognitive differences between depressed and non-depressed groups are smaller and more equivocal when comparing those with milder depression symptoms to those with higher symptom severity (Hammar, 2003; McDermott & Ebmeier, 2009). Unlike research on burnout and cognition, there are few studies directly examining the effects of depression symptoms on cognition in an occupational setting. This is likely due to the fact that MDD is associated with impairments in domains of social function beyond work (Plieger et al., 2016), whereas burnout is defined as primarily occurring in the context of work. This explanation, however, does not fully explain the lack of studies examining depression symptoms and cognitive performance among actively working individuals. This is an area that needs additional objective assessment in light of several studies showing that working individuals with depressive symptoms report high rates of subjective cognitive complaints, and that these complaints are associated with worse functional performance in the workplace (Hammer-Helmich et al., 2018; Wang & Gorenstein, 2014).

There is some evidence to suggest that cognitive performance may represent a marker for the severity or persistence of emotional distress, such as burnout or depression symptoms. For instance, if the need to conserve emotional resources is such that cognitive performance worsens with increased severity of burnout or depression symptoms, this may reflect a compensatory response that predicts the severity or chronicity of burnout and depressive symptoms over time. In support of this, there is evidence that deficits in information processing speed and executive functions in the context of acute depression may be an indicator of the persistence of depressive symptoms (Abo Aoun et al., 2019; Groves et al., 2018). There is limited research examining the association between cognitive deficits and persistent burnout symptoms; however, there is evidence that deficits in information processing speed at baseline are associated with higher odds of general psychological distress over a 7-year follow up (Gale et al., 2016). This study was based on the UK Health and Lifestyle Survey (N = 3088) and the measure of psychological distress was the General Health Questionnaire; thus, while these findings suggest slowed information processing speed may predict higher general distress over time, it is not specific to depression or burnout.

Workplace stress and the resilience of cognitive performance

One theoretical model often associated with burnout and depression symptoms in the context of work is Conservation of Resources Theory (COR, Hobfoll, 1998), COR is a prominent theory on human behavioral response to stress, and one of the core principals is that individuals strive to preserve what they find psychologically valuable (Hobfoll, 2010). From this perspective, it can be argued that individuals lacking sufficient resources in one dimension, such as emotional energy due to burnout or depression symptoms, may experience disproportionate decline when taxed in a different dimension, such as cognition. A contrasting perspective can be found in the work of Kahneman (1973), who argued in his seminal treatise on attention that it is difficult to significantly impair work-related cognitive performance, due to the human ability to maintain goal-directed attention when required (Kahneman, 1973). If the COR perspective is correct, we would expect cognitive performance to be vulnerable to levels of burnout and depression symptoms. If Kahenman’s perspective is correct, we would expect goal-directed cognition to be resilient to burnout and depression symptoms.

The goal of the current study was to examine both cross-sectional and prospective associations between cognitive performance and symptoms of burnout and depression. Examining both burnout and depression symptoms in the same sample is important because there is debate about the conceptual distinctiveness of burnout relative to depression (Bianchi, Schonfeld, & Laurent, 2015). A few studies have explicitly compared burnout and depression symptoms in the same sample with respect to cognitive performance, and thus similarities or differences between the two constructs in relation to information processing speed or executive functions may inform this debate over their distinctiveness. We chose the task-switching paradigm for this study because it is a widely researched cognitive paradigm (Monsell, 2003; Rogers & Monsell, 1995), including previous studies of burnout (Gajewski et al., 2017; Oosterholt et al., 2014) and depression (Gajewski et al., 2017; Meiran et al., 2011). In addition, the tasks can be decomposed to differentiate information processing speed from the executive function components of switching costs and mixing costs. Switching costs are presumed to measure cognitive rigidity due to difficulty abandoning a currently irrelevant task causing proactive interference, whereas mixing costs are presumed to reflect the ability to maintain multiple tasks in working memory (Rogers & Monsell, 1995). From the perspective of ecological validity, task-switching is an important ability in highly active or complex work environments, in which multiple discrete activities require ongoing monitoring and response.

The central questions of this study were: 1) whether cognitive differences in information processing speed and executive functions are differentially associated with burnout and depression symptoms at baseline, and 2) whether baseline differences in these cognitive functions predict persistent burnout and depression symptoms over 12-month follow-up. We hypothesized that a) slower single–task reaction time (RT, an indicator of information processing speed) and b) greater switching and mixing costs would be associated with higher levels of burnout symptoms (Hypothesis 1 or H1) and depression symptoms (H2). We also hypothesized that slower single-task RT and greater switching and mixing costs would be associated with persistently higher levels of burnout symptoms (H3) and depression symptoms (H4) over 12-months. We did not have specific hypotheses about the relative strength of associations of burnout and depression symptoms to cognitive performance.

Materials and methods

Participants

We tested our hypotheses in a sample of nursing workers, which is an occupation with higher-than-average rates of burnout, depression, and job turnover due to stressful working conditions (Aiken et al., 2013; Letvak et al., 2012). Participants were employees from three hospitals within single health system. We used a convenience sampling approach based on recruitment via email addresses obtained from hospital administration. Responding individuals were sent an eligibility survey. Eligibility criteria were: 1) actively working in a nursing field (including certified nursing assistants, and similar positions), 2) minimum two years’ experience in a nursing field, 3) minimum 25 years old, and 4) no neurological conditions potentially confounding to cognitive performance (seizures, severe brain trauma, and stroke). Of 862 screening survey responses, eligible participants (92%) were contacted to schedule in-person study visits. Visits were scheduled on a rolling “first-come-first-served” basis until we reached our target sample of 400 completed baseline visits (N = 402). For the current study we included only participants who were registered nurses (n = 372) in order to maintain similarity of educational/training experience and stresses related the job; non-nurses recruited for the study such as nursing assistants and certified medical assistants have different training, job demands, level of responsibility, and potentially different contexts contributing to burnout. Mean age for the nurse sub-sample was 41.7 years (SD = 11.4, range: 25–67), and mean education was 15.7 years (SD = 1.2, range: 14–20). The sample was 83% Caucasian, 9% African American, and the remaining 8% reported other racial or ethnic identities. Monthly completion rates for self-reported burnout and depression symptoms ranged from 82–100% (mean: 90%) and 79%−95% (mean: 86%), respectively.

Although hospital administration provided email addresses for recruitment, they were provided no information about individual enrollment or responses. This study was approved by the Institutional Review Board of the participating health system.

Design

This study used a cross-sectional cognitive and psychosocial assessment paired with prospective monthly assessments of burnout and depression symptoms over 12 months. Information on demographics, burnout, and depression were collected at baseline via computerized self-report questionnaires. Task-switching performance was obtained at baseline via a computerized task. Follow-up assessments on burnout and depressive symptoms were administered via 12 monthly emails linked to a secure online questionnaire (Harris et al., 2009).

Measures

Task-switching

Task-switching is defined as the ability to switch between task sets, and is seen as a central aspect of executive control (Norman & Shallice, 1986). In a typical task-switching paradigm, differences in RT are compared across two types of tasks: those requiring a single type of response (single-task blocks), and those requiring two or more types of responses (mixed-task blocks). Two key parameters in task-switching paradigms are switching costs and mixing costs (Gajewski et al., 2010). Switching costs are defined as the difference in RT between a switch trial (e.g., task 1 to task 2) and a repeat trial (e.g., task1 to task 1, or task 2 to task 2) occurring within the mixed block. Mixing costs are defined as the difference in RT between trials in the single-task block and between repeat trials in the mixed block. Generally, response to a task takes longer on a trial that requires switching to a new task than on a trial that repeats the previous task.

The task-switching paradigm used in the current study was adapted from previous work (Gajewski et al., 2017; Gajewski et al., 2010; Schapkin et al., 2014). The task is composed of two single-task blocks and one mixed-task block. The stimuli in each block were the digits 1–9, excluding the number 5. Stimuli were presented in white on a black computer screen with a cue indicating the relevant set. In one of the single-task blocks, the cue “NUM” (representing “Number”) prompted for a response indicating whether the number presented was greater or less than 5. In the other single-task block, the cue “PAR” (representing “Parity”) prompted for a response indicating whether the number presented was odd or even. The cue was presented in the single-task blocks only. In the memory-based mixed block, a cue of “XXX” was presented instead of the informative cue. There were 65 single-task trials each for the NUM and PAR blocks. The single task blocks are important for assessing RT, and to become familiar with the two tasks individually before they are administered together in the mixed block. In the memory-based mixed block, the two tasks were presented in mixed order and the participants were instructed to switch the task after every two trials in the following order: NUM, NUM, PAR, PAR, NUM, NUM, PAR, PAR (and so forth to end of trial). A stimulus prompt of “XXX” was presented instead of the actual cue, which forced participants to mentally track the actual trial sequence. When two consecutive errors were made, or no response given within the 2500 ms interval, cues were presented for the next two trials to help participants re-establish the task sequence. An example of a single-task sequence and a mixed-task sequence is presented in Figure 1. The mixed block consisted of 130 trials. The frequency of task-switching in the mixed block was 50% of trials. The participants were provided with on-screen instructions and the research technician administering the task verbally queried participants for understanding of the task before beginning. Variables of interest were single-task RT (mean of NUM and PAR blocks), switching costs (switch trial RT – repeat trial RT in mixed block), and mixing costs (repeat trial RT in mixed block – single-task RT). For the analysis of single-task RT, we removed trials with responses faster than 100 ms or slower than 2500 ms.

Figure 1.

Schematic example of single task and mixed-task trial

Note. Single-task sequence appears at top of figure. NUM cue prompts participants to indicate whether the number appearing is greater or less than 5. Mixed-task sequence appears at bottom of figure. XXX represents either the NUM or PAR cue, the sequence of which participant has to maintain in memory. ms = milliseconds.

Burnout symptoms

Burnout symptoms were assessed with the Exhaustion subscale of the Oldenburg Burnout Inventory (OLBI, Demerouti et al., 2001). The OLBI Exhaustion scale assesses self-reported resources across emotional, physical, and cognitive domains, which provides a broader survey of the exhaustion construct than some other measures. The eight items are scored on a four-point Likert scale, 3 in a positive direction and 5 in a negative direction in order to reduce response bias. Items worded in a negative direction were reverse-scored before items were averaged, yielding scale scores ranging from 1 to 4, with 4 indicating the highest level of severity. Chronbach’s alpha was reported by the authors as 0.82.

Depression symptoms

Depression symptoms were assessed with the Patient Health Questionnaire-9 (PHQ-9, Kroenke et al., 2001). This self-report measure assesses the frequency of nine depression symptoms over the previous two weeks (diminished interest in pleasurable activities, feeling down, sleep problems, diminished energy, appetite problems or overeating, feelings of failure, concentration problems, slowed speech or actions, and suicidal thoughts). Individual items assess frequency of symptoms on a 0–3 scale, with a total score range from 0–27. Chronbach’s alpha was reported by the authors as 0.89. Several participants had missing individual items on the PHQ-9; rather than imputing scores with missing items, we chose not to include individuals with an incomplete PHQ-9.

Analysis

Statistical analyses were conducted in SAS (SAS Institute Inc., 2012). We produced descriptive statistics including mean and standard deviation (SD) to examine the characteristics of study variables. We produced Pearson correlation coefficients to examine bivariate associations between variables.

Cross-sectional analysis.

To test the associations of task-switching variables to burnout symptoms and depression symptoms, cognitive tasks were modelled as independent variables in regression models, along with control variables of age and education. Burnout (H1: OLBI) and depressive symptoms (H2: PHQ-9) were modelled as dependent variables. We tested three models for our hypothesis related to baseline burnout symptoms, one each for single-task RT, switching costs, and mixing costs. We similarly tested three models for our hypothesis related to baseline depression symptoms, one each for single-task RT, switching costs, and mixing costs. To assess multicollinearity, we determined the variance inflation factor (VIF) for each predictor. All predictors had a VIF less than 10, indicating no multicollinearity.

Prospective analysis

As with the cross-sectional analysis, cognitive tasks were modelled as independent variables in regression models, along with control variables of age and education. Persistent burnout (H3: OLBI) and persistent depressive symptoms (H4: PHQ-9) were modelled as dependent variables. Persistent symptoms were defined on the mean value of OLBI Exhaustion (H3) and PHQ-9 scores (H4) over the 12 follow-up months, and excluding the baseline assessment scores. We tested three models for our hypothesis related to persistent burnout symptoms, one each for single-task RT, switching costs, and mixing costs. We similarly tested three models for our hypothesis related to persistent depression symptoms, one each for single-task RT, switching costs, and mixing costs. As with the cross-sectional models, all predictors had a VIF less than 10, indicating no multicollinearity.

Results

Task characteristics

Table 1 provides means and standard deviations for RT for each of the task-switching components. Additionally, there was no significant difference in RT between the two types of single task conditions (PAR vs. NUM). Table 1 also provides means and standard deviation of the OBLI (Exhaustion scale) and PHQ-9.

Table 1.

Descriptive characteristics of sample and study variables

Task-switching	Mean (SD)
RT single trial: Combined mean	597.2 (105.1)
RT Switching costs	94.4 (108.7)
RT Mixing costs	189.6 (140.7)
Burnout
OLBIb (baseline)	2.62 (0.47)
OLBI (persistentc)	2.49 (0.42)
Depression
PHQ-9d (baseline)	5.49 (4.86)
PHQ-9 (persistentc)	4.99 (3.77)

^aRT = reaction time (milliseconds).

^bOLBI: Oldenburg Burnout Inventory, Exhaustion subscale.

^cPersistent: mean of 12 follow-up months per each participant.

^dPHQ-9 = Patient Health Questionnaire-9.

Simple bivariate correlations revealed that burnout and depression measures were correlated with each other, both at baseline, r(352)=.58, p<.001, and persistently over 12 months, r(369)=.62, p<.001. Burnout symptoms were correlated between baseline and 12-month mean, r(369)=.82, p<.001. Depression symptoms were correlated between baseline and 12-month mean, r(351)=.78, p<.001. In addition, higher age was associated with lower burnout symptoms at baseline, r(369)= −10, p=.046, and with lower persistent symptoms, r(368)=−.12, p=.03, but was not significantly associated with depression symptoms. Education level was not significantly associated with burnout or depression symptoms.

Cross-sectional findings

Baseline burnout symptoms were regressed on each of the three task-switching components, controlling for age and education level (H1). Higher burnout symptoms were associated with slower single-task RT, β=0.13, t=2.23, p=.03, but were not significantly associated with switching costs or mixing costs. Baseline depression symptoms were regressed on each of the three task-switching components, controlling for age and education level (H2). Single-task RT, switching costs and mixing costs were not significantly associated with baseline depression symptoms. These models are presented in Table 2.

Table 2.

Baseline burnout and depression on cognitive tasks

	Burnout
	Single trial RT		Switching costs		Mixing costs
	b	b	b	b	B	b
Task performance	0.0005* (0.0002)	0.13* (0.06)	0.0002 (0.0002	0.05 (0.05)	−0.00008 (0.0002)	−0.03 (0.05)
Age	−0.007** (0.002)	−0.17** (0.06)	−0.005* (0.002)	−0.13* (0.05)	−0.005* (0.002)	−0.13* (0.05)
Education
Bachelors	−0.03 (0.06)	−0.06 (0.12)	−0.03 (0.06)	−0.07 (0.12)	−0.03 (0.06)	−0.07 (0.12)
Masters or Doctorate	0.18* (0.08)	0.38* (0.18)	0.19* (0.08)	0.40* (0.18)	0.18* (0.08)	0.39* (0.18)
R2	.05		.03		.03
	Depression
Task performance	0.004 (0.003)	0.09 (0.06)	0.004 (0.002)	0.10 (0.06)	0.0009 (0.002)	0.03 (0.06)
Age	−0.04 (0.02)	−0.10 (0.06)	−0.03 (0.02)	−0.07 (0.05)	−0.03 (0.02)	−0.08 (0.06)
Education
Bachelors	−0.76 (0.61)	−0.16 (0.13)	−0.69 (0.61)	−0.14 (0.13)	−0.76 (0.61)	−0.16 (0.13)
Masters or Doctorate	0.83 (0.92)	0.17 (0.19)	0.97 (0.92)	0.20 (0.19)	0.95 (0.93)	0.20 (0.19)
R2	.02		.02		.02

Note.

^*p<.05,

^**p<.01.

Reference group for education: Associate degree.

Prospective findings

Persistent burnout symptoms were regressed on each of the three task-switching components, controlling for age and education level (H3). Slower single-task RT predicted a higher mean level of burnout symptoms across the 12 follow-up assessments, β=0.20, t=3.35, p<.001, but the other cognitive variables were not associated with burnout symptoms. Persistent depression symptoms were regressed on each of the three task-switching components, controlling for age and education level (H4). Single-task RT, switching costs and mixing costs were not significantly associated with baseline depression symptoms. These models are presented in Table 3.

Table 3.

Follow-up burnout and depression symptoms on cognitive tasks

	Burnout
	Single trial RT		Switching costs		Mixing costs
	b	b	b	b	B	b
Cognitive task	0.0007** (0.0002)	0.20** (0.06)	0.0002 (0.0002)	0.05 (0.05)	0.00006 (0.0002)	0.02 (0.05)
Age	−0.007** (0.002)	−0.20** (0.06)	−0.005** (0.002)	−0.14** (0.05)	−0.005** (0.002)	−0.14** (0.05)
Education
Bachelors	−0.03 (0.05)	−0.06 (0.12)	−0.03 (0.05)	−0.07 (0.12)	−0.03 (0.05)	−0.08 (0.12)
Masters or Doctorate	0.16* (0.08)	0.37* (0.18)	0.17* (0.08)	0.39* (0.18)	0.17* (0.08)	0.39* (0.18)
R2	.06		.04		.03
	Depression
Cognitive task	0.002 (0.002)	0.05 (0.06)	0.002 (0.002)	0.05 (0.05)	0.0007 (0.001)	0.03 (0.05)
Age	−0.02 (0.02)	−0.06 (0.06)	−0.02 (0.02)	−0.05 (0.05)	−0.02 (0.02)	−0.05 (0.05)
Education
Bachelors	−0.48 (0.47)	−0.13 (0.12)	−0.48 (0.47)	−0.13 (0.12)	−0.49 (0.47)	−0.13 (0.12)
Masters or Doctorate	0.21 (0.69)	0.06 (0.18)	0.23 (0.69)	0.06 (0.18)	0.26 (0.69)	0.07 (0.18)
R2	.008		.008		.007

Note.

^*p<.05,

^**p<.01.

Reference group for education: Associate degree.

Analysis of dichotomized burnout and depression symptoms

As a follow-up to our a priori analyses using burnout and depression as continuous dependent variables, we examined burnout and depression symptoms dichotomized to better distinguish between clinical and non-clinical severity of burnout and depression symptoms. That is, do the cognitive variables predict “clinically meaningful” levels of burnout and depression symptoms? For depression symptoms, we dichotomized the sample into a higher severity group of PHQ-9 scores greater than 10 (n = 70), which is the cutpoint for moderate symptom severity (Kroenke & Spitzer, 2002), and a lower severity group with scores below 10 (n = 284). Because there is no comparable clinical definition of burnout severity on the OLBI, we defined high burnout severity as the top quintile of scores (n = 64), which reflects the same proportion of individuals who scored as moderate or higher depression symptom severity on the PHQ-9, and low severity as scores in the lower 4 quintiles (n = 308). We then ran dichotomized models comparable to our a priori models. In the cross-sectional models, none of the cognitive variables were significantly associated with either burnout symptoms or depression symptoms. In the prospective models, none of the cognitive variables were significantly associated with either burnout symptoms or depression symptoms. As with the a priori models, greater age was associated with lower burnout symptoms both cross-sectionally and prospectively, but was not associated with depression symptom severity.

Discussion

The current study explored how cognitive components of the task-switching paradigm are associated with burnout and depression symptoms, both cross-sectionally and prospectively over 12 months, in a sample of actively working nurses. Results indicate that in this healthy working sample, generalized cognitive slowing may be associated with higher and more persistent burnout symptoms, but not with depression symptoms. Specifically, we found that slower single-task RT was associated with higher baseline burnout symptoms, and also predicted higher mean severity of burnout symptoms over 12 months. While these associations were statistically significant, we note that the magnitude of these associations was small; R² was .02 for baseline burnout symptoms and .06 for persistent burnout symptoms. The switching and mixing components of the task were not associated with burnout symptoms, either cross-sectionally or prospectively. We found no significant associations depression symptoms and any of the task-switching variables. When we ran models dichotomized to approximate clinically meaningful levels of burnout and depression symptoms, none of cognitive variables were significant predictors, either cross-sectionally or prospectively. This suggests that single-task RT was sensitive to burnout severity across the full range of scores, but did not discriminate between groups based on a dichotomous cut-off score.

Our finding of statistically significant associations between slower single-task RT and higher burnout symptoms is consistent with some previous findings of generalized slowing in association with higher burnout symptoms, also without a finding corresponding association with measures of executive functions (Oosterholt et al., 2014; Osterberg et al., 2009). Unlike these studies, our sample was composed of actively working individuals rather than individuals with burnout-related work impairment, which suggests that slowed information processing is detectable in milder, non-clinical manifestations of burnout. Although information processing speed is often viewed as a lower order cognitive processes and executive functions as a higher order cognitive process (Diamond, 2013), studies in multiple populations suggest that information processing speed is a foundational component of higher order cognitive processes like executive functions. Prior studies have found mediation of executive function performance by information processing speed in a number of populations, including children (Mulder et al., 2011), adults with MDD (Zaremba et al., 2019), older adults (Salthouse, 1996), and individuals with cardiovascular disease (Liebel et al., 2017). It is possible that previous studies finding associations between burnout and executive functions were at least partly attributable to unmeasured underlying decrements in information processing speed that are often inherent in the task attributes and response output of many executive function measures. One strength and novel aspect of the current study is a task-switching paradigm that can better disambiguate information processing speed from executive function processes.

Contrary to our hypotheses, we did not find cognitive performance to predict persistent depression symptoms, which is where most of the research has been conducted on cognitive markers of persistent emotional distress. One explanation for this may be that the exhaustion-related construct assessed by the OLBI is more sensitive and/or specific to slowed information processing than the more heterogeneous depression construct assessed by the PHQ-9, at least when depression symptom severity is mild and largely sub-clinical. Some evidence suggests the exhaustion dimension of the burnout construct may be closer in character to the stress-related condition of chronic fatigue than to depression (Huibers et al., 2003; Leone et al., 2011), and research does find that chronic fatigue is associated with deficits in information processing speed (Cvejic et al., 2016; Deluca et al., 2004).

Our findings add novel cognitive evidence to the debate about the conceptual distinctiveness of burnout and depression (Bianchi et al., 2015; Schaufeli et al., 2001), and in support of the argument that cognitive differences exist between the two constructs. Few studies have examined both burnout and depression symptoms together with respect to cognition; but a previous study by our group with a different sample of working adults found significant slowing of RT during switch trials in association with burnout, but not in association with depressive symptoms (Gajewski et al., 2017). This study also found a distinct profile of brain activity during task-switch performance using event-related potentials for individuals with burnout symptoms compared to depressive symptoms. Our finding that single-task RT predicted mean burnout symptoms over 12-months suggests deficits in information processing speed under prolonged stress may be a trait of individuals with vulnerability to persisting exhaustion-related burnout. We recognize that the association between information processing speed and persistent burnout symptoms may be bidirectional, but this is something the current study was not designed to test. We also acknowledge that cognitive deficits in information processing speed and executive functions may emerge at higher levels of depression symptom severity, and that cognitive tasks measuring other dimensions of the executive function construct may potentially be more sensitive to depression symptoms than switching costs and mixing costs.

Our results seem to provide a nuanced perspective on whether cognitive performance is vulnerable to workplace stress. One the one hand, our findings regarding single-task RT would seem to support the COR perspective predicting a vulnerability of cognitive processes to the emotional perturbations of burnout. On the other hand, the small magnitude of the associations may support Kahneman’s argument for the resilience of goal-directed attention despite increased burnout symptoms. In this respect, the small magnitude of the associations is reassuring; it suggests that working individuals with burnout and depression symptoms maintain the resilience to exert sufficient cognitive effort when needed, particularly if removed from the proximate stressors of the workplace. We argue that the influence of work stress on cognition may be more closely tied to factors that lead to work disability than to measures of symptom severity per se. The current findings came from samples of actively working individuals, whereas the most robust cognitive findings tend to come from samples of individuals whose stress has required them to take leave from their work. There may be unique factors associated with disability that are also associated with cognitive vulnerability to stress.

The possible biological mechanisms for the current findings cannot be addressed directly within our study, but there is an evidence base that implicates broad adverse effects of stress-related inflammation on cognitive function. Burnout and depression are both endpoints of chronic psychosocial stress, and higher levels of these symptoms have been associated with higher levels of pro-inflammatory cytokines (Felger & Lotrich, 2013; Mommersteeg et al., 2006). A previous study by our group found that the cytokines IL-6 and IL-12 were associated with higher burnout symptoms in men both more strongly and consistently than with depression symptoms (Gajewski et al., 2017). Another study found that individuals with higher levels of fatigability had greater increase in the IL-6 cytokine in response to cognitive demand (Lin et al., 2014). Elevated cytokines and other markers of inflammation are associated with worse performances on speed of processing and executive functions (Heringa et al., 2014), and are associated with cognitive decline (Baune et al., 2008; Schram et al., 2007). The role of individual cytokines in various disease processes is complex, and may differ by characteristics such as gender and symptom phenotype (e.g., burnout vs. depression). Thus, it is biologically plausible that inflammation due to chronic psychosocial stress can contribute to slowed information processing and potential differences in the clinical manifestation of burnout and depression, however, further research is needed to elucidate these processes.

Practical Implications

Despite finding statistically significant associations between single-task RT and burnout symptoms both cross-sectionally and prospectively, the clinical relevance is more marginal. The magnitude of our findings was relatively modest and occurred in a laboratory setting removed the unique stressors of the workplace; as such, these measures may lack ecological validity in generalizing to workplace performance outcomes. An important next step is to examine how burnout and depression symptoms influence ecologically valid indicators of cognitive performance in workplace settings. Regardless our specific findings, it is known that both burnout and depression symptoms are associated with poor work performance, which could be a safety issue if untreated. For instance, one study of physicians found that those with high levels of burnout and depression symptoms were approximately 10 times more likely to characterize their work performance as deficient (Ruitenburg et al., 2012). In another study of physicians, higher levels of burnout were associated with higher levels of self-reported medical errors (Tawfik et al., 2018). Such individuals might be prioritized for interventions aimed at reducing burnout, which is important because: 1) chronic burnout has been found to predict depression (Armon et al., 2014), and 2) individuals who end up leaving the workforce due to stress-related exhaustion have shown persistent cognitive deficits (Jonsdottir et al., 2017). Given that research generally supports the conceptualization of burnout as a consequence of chronic job-related stressors, interventions may include both changing organizational sources of psychosocial job stress, as well as helping individuals better manage their individual responses to these stressors.

Limitations

One potential limitation to the current study is that we used a convenience sample of nursing workers in one health system, which may introduce potential for selection bias if healthy workers with low burnout and depression symptoms were more likely to participate. We note that our baseline burnout scores were comparable to those reported in a study of European nurses (Innstrand et al., 2011). We also note that approximately 20% of our sample had PHQ-9 depression scores of 10 or higher, which is often cited as a cutpoint for likelihood of MDD (Kroenke et al., 2001), and is a considerably higher proportion than in population-based samples such as NHANES (Shim et al., 2011). Both of these findings provide confidence that the sample was not unduly biased toward emotionally healthy workers, and that we captured variability in the range of burnout and depression symptoms. Our sample was demographically similar to the overall nursing workforce in the health system we studied, which suggests our convenience sample was broadly representative of the workforce population from which it was drawn.

Another potential limitation is that the study focused on a single occupation in a single organization. While this may limit generalization to other occupations, nursing is an emotionally demanding profession that is frequently studied across multiple countries with respect to burnout (Aiken et al., 2013; Poghosyan et al., 2010). As discussed previously, a prior cross-sectional study found an association of slower RT to burnout but not to depression symptoms in a broader sample of individuals with emotionally demanding jobs (Gajewski et al., 2017). We believe that the current results are likely generalizable to other emotionally demanding jobs, but this should be tested in future research.

A third potential limitation is that the current study only included the exhaustion domain of burnout. Exhaustion is widely considered to be the core dimension of burnout (Seidler et al., 2014), and it is the most common dimension of burnout studied with respect to cognition (Deligkaris et al., 2014); thus, our approach was consistent with the majority of current research. It is possible that the current results generalize less well to individuals who maintain low levels of exhaustion while scoring high on other burnout dimensions like cynicism or loss of personal efficacy (Maslach et al., 2001). Examining cognitive function related to multiple dimensions of burnout could be an informative direction of future research.

Conclusion

The main novel finding from this study was that slower information processing speed was modestly associated with higher baseline and persistent burnout symptoms over 12 months among actively working individuals, but was not associated with depression symptoms. Information processing speed may be more sensitive to the emotional exhaustion dimension of burnout and less sensitive to mild depressive symptoms, which are more heterogeneous in nature. Because burnout is a consequence psychosocial job stress, future research should examine whether specific types of job stressors are more associated with burnout symptoms and slowing of information processing speed.