International Journal of Environmental Research and Public Health Article How the Lagged and Accumulated Effects of Stress, Coping, and Tasks Affect Mood and Fatigue during Nurses’ Shifts Fermín Martínez-Zaragoza 1 , Jordi Fernández-Castro 2, * , Gemma Benavides-Gil 1 and Rosa García-Sierra 3,4 1 Department of Behavioural Sciences and Health, University Miguel Hernández, 03202 Elch, Spain; f.martinez@umh.es (F.M.-Z.); gbenavides@umh.es (G.B.-G.) 2 Departament de Psicologia Bàsica, Evolutiva i de l’Educació, Universitat Autònoma de Barcelona, 08193 Barcelona, Spain 3 Research Support Unit Metropolitana Nord, University Institute Foundation for Research in Primary Health Care Jordi Gol i Gurina (IDIAPJGol), 08303 Mataró, Spain; yedra4@hotmail.com 4 Department d’Infirmeria, Universitat Autònoma de Barcelona, Campus de Bellaterra, 08193 Barcelona, Spain * Correspondence: jordi.fernandez@uab.es; Tel.: +34-93581-1469 or +34-93581-1377  Received: 1 September 2020; Accepted: 28 September 2020; Published: 5 October 2020  Abstract: Nurses experience significant stress and emotional exhaustion, leading to burnout and fatigue. This study assessed how the nurses’ mood and fatigue evolves during their shifts, and the temporal factors that influence these phenomena. Performing a two-level design with repeated measures with moments nested into a person level, a random sample of 96 nurses was recruited. The ecological momentary assessment of demand, control, effort, reward, coping, and nursing tasks were measured in order to predict mood and fatigue, studying their current, lagged, and accumulated effects. The results show that: (1) Mood appeared to be explained by effort, by the negative lagged effect of reward, and by the accumulated effort, each following a quadratic trend, and it was influenced by previously executing a direct care task. By contrast, fatigue was explained by the current and lagged effect of effort, by the lagged effect of reward, and by the accumulated effort, again following quadratic trends. (2) Mood was also explained by problem-focused and emotion-focused coping strategies, indicative of negative mood, and by support-seeking and refusal coping strategies. (3) Fatigue was also associated with direct care and the prior effect of documentation and communication tasks. We can conclude that mood and fatigue do not depend on a single factor, such as workload, but rather on the evolution and distribution of the nursing tasks, as well as on the stress during a shift and how it is handled. The evening and night shifts seem to provoke more fatigue than the other work shifts when approaching the last third of the shift. These data show the need to plan the tasks within a shift to avoid unfinished or delayed care during the shift, and to minimize accumulated negative effects. Keywords: stress; nurses; coping; mood; fatigue; burnout; ecological momentary assessment; lagged effects; accumulated effects 1. Introduction Nurses are exposed to significant stress at work, which can lead to burnout, provoking a deterioration in nurses’ health, a decrease in the quality of their care, and an increase in their intention to quit [1–5]. Two widely studied models of work stress have been tested in the context of nursing—Karasek’s Demand-Control model [6] and Siegrist’s Effort-Reward Imbalance (ERI) model [7]. Thus, nurses suffering high demands or an excessive workload and with little control or autonomy when carrying out their tasks tend to present greater health problems [8–10]. In addition, the lack of balance Int. J. Environ. Res. Public Health 2020, 17, 7277; doi:10.3390/ijerph17197277 www.mdpi.com/journal/ijerph Int. J. Environ. Res. Public Health 2020, 17, 7277 2 of 15 between the effort made at work and the rewards obtained (i.e., recognition, salary, and opportunities for promotion) would also generate a greater risk of burnout in the medium and long-term, and greater vulnerability to different pathologies [11–14]. By combining both of the aforementioned models in tasks and situations that involve high levels of demand/effort with low levels of control/reward, negative mood increases significantly. However, if the level of control/reward was high, then the mood remained positive, even if the degree of demand/effort was high [15]. Fatigue and negative mood are considered to be an effect of work stress. Some of the consequences of fatigue are lapses in attention, an inability to stay focused, increased reaction time, reduced processing capacity, confusion, compromised problem solving, memory lapses, impaired communication, irritability, indifference, and lack of empathy [16,17]. Moreover, fatigue has been related to excess effort and lack of reward [18]. In the nursing setting, some interesting results link fatigue and mood to task demand, control, effort and reward. For example, high demands and lack of control or autonomy have an absorbing effect on fatigue [19]; nursing tasks are less rewarding and mood is worse when exhaustion or fatigue are detected [20]; and mood decreases with tasks that involve high levels of demand/effort and low levels of control/reward, whereas mood remains positive when the level of control/reward is high, even when the degree of demand/effort is also high [15]. It is also important to take into account how the fatigue and mood levels in nurses change throughout the work shift and the consequences of this. A decrease in the effectiveness of clinical decision-making as the working day progresses and over multiple work shifts has been observed [21,22]. The effort associated with long and continuous work shifts provokes high levels of fatigue, which negatively influence performance and patient safety [23–25]. Variations in mood have also been observed between the beginning and end of each shift [26,27]. Significantly, attempts have been made to predict the levels of fatigue from the values of physical energy expended, as well as the demand, control, and reward accumulated in the previous 90 min, through ecological momentary assessment throughout a nurse’s work shift. Results show that fatigue does not necessarily depend on the physical energy expended or on the perceived demands but, rather, on the perceived control over the work and the rewards associated with it [28]. However, nurses must perform different tasks in their work and not all of them have the same implications or are perceived in the same way. In this regard, physically demanding patient care tasks, logistic/management/organization tasks, and multitasking demands are those that appear to contribute to fatigue most frequently [29,30]. Likewise, not all coping strategies work in the same way in all situations and/or for all people, such that the risk of physical and psychological pathologies increases with the use of maladaptive coping strategies, including burnout [31–36]. Whilst problem-focused coping behaviours are not associated with fatigue in nurses, avoidance coping such as drinking alcohol, avoiding situations, or blocking emotions appear to predict fatigue [37]. Moreover, the use of avoidance coping and the perception of excessive demands on nurses are thought to be important predictors of mood disturbances [38]. Indeed, an important challenge in stress research is to understand how stress, coping, and health change in individuals over time [39], with insufficient progress in this field to date. Therefore, it is not yet established to what extent the mood and fatigue that nurses experience during their work shifts are determined by the delayed and accumulated effect of the type of task performed; the perception of demand, control, effort, and reward of the task; and the type of coping used when performing them. Analyzing the influences between these variables in real time by ecological momentary assessment, is necessary to be able help improve the performance of nursing staff and their physical and emotional well-being, this being the main objective of the present study. 2. Aim/Research Questions The initial aim of this study was to examine the relationship between stress, mood, and fatigue in nurses. A recent study concluded that fatigue does not depend on the demands or on the energy expended but, rather, on the perceived control and its associated reward [28]. Thus, it remains unclear Int. J. Environ. Res. Public Health 2020, 17, 7277 3 of 15 if previous (lagged effect) or accumulated stress (demand/control, effort/reward) influences the mood and fatigue of nurses. Thus, we hypothesized that: 1. Accumulated effort will worsen mood and will increase fatigue, while reward will produce the opposite effect, as predicted by Siegrist’s model. The second aim was to study the effect of coping on mood and fatigue. It has been shown that higher mean scores in problem-solving, avoidance, and seeking social support were associated with higher levels of compassion fatigue [40]. Thus, the question remains as to whether the coping strategy previously used or accumulated over time affects the mood and fatigue of nurses. Hence, we hypothesized that: 2. The use of problem-focused and seeking-support strategies will have a delayed and accumulated effect on mood and fatigue, producing an improvement in both; and 3. The use of emotion-focused and refusal strategies will have a delayed and accumulated effect on mood and fatigue, worsening them. The third aim of this study was to examine the effect of the task on mood and fatigue. It has been shown that physically demanding patient care tasks most frequently produced physical and mental fatigue in nurses [29]. Thus, the question arises as to whether the previous task or those accumulated over time affect the mood and fatigue of nurses. Accordingly, we hypothesized that: 4. Accumulated direct care tasks improve mood and increase fatigue; 5. Administrative and organizational tasks will have a delayed and negative effect on mood, presumably due to tedium. 3. Materials and Methods 3.1. Design Within-person and between-person relationships were identified by multilevel statistical analyses. A two-level design with repeated measures was used [41], in which level 1 was established from the moments at which the outcome variables were measured. In level 2, the moments were nested at the person level. 4. Participants, Settings, and Procedure A random sample of 113 nurses was recruited from the following wards at two University hospitals, selecting 80% of the nurses on each ward: internal medicine, surgery, traumatology, oncology, cardiology, neurology, nephrology, pneumology, rheumatology, digestive, gynaecology, geriatrics, palliative care, paediatrics, and psychiatry. Critical care services and emergency services were excluded from the cohort due to the distinctive tasks involved. The nurse–patient ratio ranged from 10 patients in the day shift to 30 patients in the night shift. Of the nurses recruited, 17 finally refused to participate in the study, and thus the final sample was 96 nurses with a response rate of 85% (96 of possible 113). Data were collected between January and December of 2015, excluding holiday periods. 5. Measures 5.1. Level 1 Measures (moment) Ecological Momentary Assessment Stone and Shiffman [42,43] have described ecological momentary assessment, in which data of interest are recorded frequently or intensively, in real time, and in a critical environment. A Samsung Galaxy Mini Smartphone with an Android operating system was specially developed for this study and used to obtain ecological momentary measurements (all the measures in this study). Int. Int.J.J.Environ. Environ.Res. Res.Public PublicHealth 2020,17, Health2020, 17,7277 x 44 of of15 18 min. This meant that in terms of hygiene, if the task they were involved in was direct care, it could The nurse carried the device during five consecutive work shifts. Data entry was prompted by vibration be completed and the nurses could wash their hands before touching the screen. However, if the or a buzzing alarm and, if busy, the nurse could postpone the response for 10 min. This meant that in question remained unanswered for 20 min then this moment was registered as missing data. Answers terms of hygiene, if the task they were involved in was direct care, it could be completed and the nurses were presented on analogue scales and the participants were given “tips” as to how to select their could wash their hands before touching the screen. However, if the question remained unanswered for responses to the questions. The software was designed with a menu to help answer any queries by 20 min then this moment was registered as missing data. Answers were presented on analogue scales just touching the screen. The measurements taken at each evaluation point are listed below. and the participants were given “tips” as to how to select their responses to the questions. The software Demand, effort, control, reward. Four questions were designed to appraise the different aspects was designed with a menu to help answer any queries by just touching the screen. The measurements of work stress: demand, control (labelled as autonomy and skills development), effort, and reward. taken at each evaluation point are listed below. Each question labelled the term to be evaluated and was followed by a simple question as to how far Demand, effort, control, reward. Four questions were designed to appraise the different aspects each concept could be applied to the characteristics of the task being performed at the time. The four of work stress: demand, control (labelled as autonomy and skills development), effort, and reward. questions were answered using a visual analogue scale from 0 to 10 in order to evaluate the intensity Each question labelled the term to be evaluated and was followed by a simple question as to how far of the response, one of the most usual response formats for single-item questions [44]. each concept could be applied to the characteristics of the task being performed at the time. The four Momentary coping. A 10-item coping questionnaire was designed ad-hoc to assess nursing questions were answered using a visual analogue scale from 0 to 10 in order to evaluate the intensity coping in an ecological momentary assessment context (MoCoping) [45] and based on the COPE of the response, one of the most usual response formats for single-item questions [44]. Inventory [46,47]. The questionnaire followed a structure of one item for each strategy, grouped into Momentary coping. A 10-item coping questionnaire was designed ad-hoc to assess nursing four types: a problem-focused approach (including one item of active coping and one of planning), coping in an ecological momentary assessment context (MoCoping) [45] and based on the COPE an emotion-focused approach (including one item of acceptance, one of reinterpretation, and one of Inventory [46,47]. The questionnaire followed a structure of one item for each strategy, grouped into distraction), support-seeking (including one item of emotional support and one of instrumental four types: a problem-focused approach (including one item of active coping and one of planning), support), and a refusal approach (including one item of denial, one of venting, and one of self-blame). an emotion-focused approach (including one item of acceptance, one of reinterpretation, and one The variables were coded as binary—i.e., the coping strategy was used or not. of distraction), support-seeking (including one item of emotional support and one of instrumental Nursing task. The task the nurse was involved in was coded based on an adaptation [20] of the support), and a refusal approach (including one item of denial, one of venting, and one of self-blame). WOMBAT classification (Work Observation Method by Activity Timing: [48,49,50]), and classified The variables were coded as binary—i.e., the coping strategy was used or not. as: direct care, indirect care, medication, documentation, communication, and social/resting tasks. Nursing task. The task the nurse was involved in was coded based on an adaptation [20] of WOMBAT establishes 10 categories: direct care (communication with patient or family, applying the WOMBAT classification (Work Observation Method by Activity Timing: [48–50]), and classified dressings, bathing, etc...), indirect care (planning care, reviewing results, etc...), medication tasks as: direct care, indirect care, medication, documentation, communication, and social/resting tasks. (medication preparation, administration, documentation, etc...), documentation (on paper or WOMBAT establishes 10 categories: direct care (communication with patient or family, applying electronic), professional communication (with other healthcare professionals), ward-related activities dressings, bathing, etc...), indirect care (planning care, reviewing results, etc...), medication tasks (coordinating beds, coordinating staffing, etc...), in transit (time between patients and between tasks), (medication preparation, administration, documentation, etc...), documentation (on paper or electronic), supervision (of students, nurses newly introduced to the service, etc...), social (breaks, meal times, professional communication (with other healthcare professionals), ward-related activities (coordinating etc...), and other. The variables were coded as binary—i.e., they are undertaking the task or not. beds, coordinating staffing, etc...), in transit (time between patients and between tasks), supervision (of Mood. Mood was measured on a single-item, five-point visual analogue scale from a happy face students, nurses newly introduced to the service, etc...), social (breaks, meal times, etc...), and other. to a sad face, where higher values reflect a poor mood (see Figure 1). The variables were coded as binary—i.e., they are undertaking the task or not. Fatigue. Fatigue was measured by a single-item, five-point visual analogue scale from a full Mood. Mood was measured on a single-item, five-point visual analogue scale from a happy face battery to an empty battery, where higher values mean more intense fatigue (see Figure 1). to a sad face, where higher values reflect a poor mood (see Figure 1). Figure 1. Visual analogue scales for mood and fatigue. Figure 1. Visual analogue scales for mood and fatigue. Int. J. Environ. Res. Public Health 2020, 17, 7277 5 of 15 Fatigue. Fatigue was measured by a single-item, five-point visual analogue scale from a full battery to an empty battery, where higher values mean more intense fatigue (see Figure 1). The order of each record at the person level (level 1, moment) was automatically recorded by the device, registering a mean of five records per shift (every half an hour). 5.2. Level 2 Measures (Person) Questionnaires Ad-hoc questionnaire. We recorded the work shift, gender, age, marital status, number of children, years of experience, and professional status of the subjects. The ad-hoc questionnaire was administered just before the smartphone was programmed in a room prepared for it. 5.3. Data Analyses No missing data imputation was carried out and descriptive statistics were obtained using the R Statistical Package [51]. 5.3.1. Multilevel Modelling (MLM) Multilevel analysis allows to control for the variance associated with random factors without data aggregation. As fixed effects, we entered the moment of time, demand-control, and effort-reward, and four different types of coping strategies for momentary use, nursing tasks, mood, and fatigue into the model. As random effects, we tested random intercepts and slopes for the effect of the same variables, allowing these to be varied randomly across groups. p-values were obtained by likelihood ratio tests of the full model with the effect in question compared against the model without the effect in question. Z-values were obtained to test the significance for fixed effects (estimates and standard error data in the tables). Trends in the longitudinal data were studied by testing the linearity of the model for all the outcomes. 5.3.2. Assumptions Visual inspection of q-q plots using the car R package [52] did not reveal any obvious deviations from homoscedasticity or normality in the dependent variables. The momentary coping strategies were categorized and coded as dummy variables (logit models), with mood and fatigue as quantitative variables. 5.3.3. Initial Models, Model Fit, and Fit Criteria Model comparison was achieved following the guidelines set out by Bliese and Ployhart [53] and Bliese [54], beginning the process by examining the nature of the outcome. To test the significance of the person effects, we carried out a likelihood ratio test comparing the null multilevel model (unconditional model) with a null single-level model, thereby testing the null hypothesis that there were no group differences. Subsequently, the intraclass correlation coefficient (ICC) was estimated in order to calculate the between/within variation ratio. The model’s fit was assessed using chi-squared tests on the log-likelihood values to compare the different models, and using the Akaike’s information criterion (AIC: [55]), a relative goodness of fit index. According to the change in these fitness indices, the last model with significant changes was chosen for each analysis. The data were analysed using the R Statistical Package [51]. The lme4 R package [56] was used to analyse the quantitative outcomes variables and to obtain a multilevel longitudinal growth curve model of the relationship between stress (demand/control, effort/reward), coping strategies (momentary), tasks, and mood and fatigue. Linearity trends were also analysed by studying the quadratic and cubic trends, and interactions between time and predictors were examined. The p-values of the lme4 outputs were obtained with the lmerTest package [57], the ICC was calculated using the sjstats package [58], and graphic data processing was performed with the ggplot2 [59] R package. Int. J. Environ. Res. Public Health 2020, 17, 7277 6 of 15 5.3.4. Mood and Fatigue from Lagged and Current Values of the Predictors To predict the lagged effects on mood and fatigue, the lag-1 measures of the predictors were calculated (i.e., the measures taken in the moment immediately prior to the outcome measure). The current and lagged effects of the predictors were included in a two-level model with the observations over time nested into participants. These analyses were divided into three parts depending on the type of predictors used: stress, coping, or task. Each analysis was performed separately with each dependent variable, such that six analyses were performed to study the lagged effects. The predictors included were: stress (first analysis) (intercept; time; and current and lag-1 measures of demand, control, effort, and reward), coping (second analysis) (intercept; time; and current and lag-1 measures of problem-focused, emotion-focused, support-seeking, and avoidance), task (third analysis) (intercept; time; and current and lag-1 measures of direct care, indirect care, medication, documentation, social, and communication). Intercept and time were allowed to vary randomly, while the rest of the variables were considered as fixed effects. 5.3.5. Mood and Fatigue from Accumulated Values of the Predictors The accumulated values were calculated by summing the previous values of the variable at each time point within each shift. The analyses were again divided into three parts, depending on the type of predictors used: stress, coping, and task. Each analysis was performed separately with each dependent variable, such that six analyses were used to study the accumulated effects. The effects of the predictors were included in a two-level model with the time of observation nested into the participants. The predictors included were: stress (first analysis) (intercept; time; and accumulated demand, control, effort, and reward), coping (second analysis) (intercept; time; and accumulated problem-focused, emotion-focused, support-seeking, and refusal coping), task (third analysis) (intercept; time; and accumulated direct-care, indirect care, medication, documentation, social, and communication). Intercept and time were allowed to vary randomly, while the remaining variables were considered as fixed effects. 5.4. Ethics All the nurses participating in the study provided their informed consent, and they were aware that they could leave the study at any moment. Approval was granted by the Ethics and Clinical Research Committee (CEIC) at the University Hospital of Elche (Spain) (CEIC-HGUE 23/01/2013) and at the Hospital of Terrassa (Spain) (CEIC-CST 18/06/2013). 6. Results 6.1. Sample Description In the cohort of nurses studied, 89.90% were female, with a mean age of 40.22 years (SD = 8.50). In terms of the shift patterns, 45.45% worked rotating shifts and the remainder worked fixed shifts—either mornings, evenings, or nights. Moreover, 5.20% of the nurses had an additional job. The distribution between the two hospitals was 47.42%/52.58%, and the mean number of years the nurses had been working at their current job was 9.86 (SD = 7.99), with a mean of 17.40 years of experience as a nurse (SD = 8.36). Tenured staff made up 77.31% of the sample and 52.53% had a specialty in nursing beyond their university degree. In addition, 50.51% were in a relationship; 16.16% were single; and the remaining were separated, divorced, or widowed. In terms of the coping strategies used, problem-focused coping was used in 46.23% of the moments measured, emotion-focused in 44.89%, support-seeking in 8.13%, and avoidance in 2.70%. Int. J. Environ. Res. Public Health 2020, 17, 7277 7 of 15 6.2. Data Analyses Initially, the effects of the work shift on the dependent variables were described prior to testing the hypotheses, and the data were analysed to see if there were any interaction effects between work shift and the time into the shift when fatigue can be predicted (see Johnston et al. [28]). Mood was explained by theInt.effect of time J. Environ. (E =Health Res. Public p <17, 0.09,2020, 0.01), x and its interaction with work shift was only marginal7 (E = 0.07, of 18 p = 0.07 for the evening work shift, and E = 0.07, p = 0.09 for the night work shift), displaying a linear explained by the effect of time (E = 0.09, p < 0.01), and its interaction with work shift was only trend with an ICC = 0.37. Fatigue was explained by the effect of time (E = 2.36 × 10−1 , p < 0.000) and its marginal (E = 0.07, p = 0.07 for the evening−2 work shift, and E = 0.07, p = 0.09 for the night work shift), −1 interaction withathe displaying work linear shift trend (E =an9.83 with ICC×=10 p < 0.01was 0.37., Fatigue for the evening explained bywork timeE(E shift,ofand the effect = =1.30 2.36×x 10 , p < 0.01 for the night work shift), again with a linear trend and an ICC = 0.42. 10-1, p < 0.000) and its interaction with the work shift (E = 9.83 x 10-2, p < 0.01 for the evening work The shift,evolution and E = 1.30ofx mood10-1, p