Associations of emotional burden and coping strategies with sick leave among healthcare professionals: A longitudinal observational study

Objectives: To investigate 1) whether care-related regrets (regret intensity, number of recent regrets) are associated with sick leave, independently of personality traits, perceived safety climate, and physical activity; and 2) whether these associations were mediated or moderated by coping strategies. Methods : Using a longitudinal international observational study (ICARUS), data were collected by the means of a weekly web survey. Descriptive and generalized estimation equations were performed. Results : A total of 276 newly practicing healthcare professionals (nurses, physicians, others) from 11 countries were included in this study. The average proportion of weeks with at least one day of sick leave was 3.2%. Nurses’ sick leave increased with number of care-related regrets (Relative Risk [RR]=1.52; 95% Confidence Interval [CI]=[1.18; 1.95], p=.001), while physicians’ sick leave increased with intensity of care-related regret (RR=1.21; 95%CI=[1.00; 1.21], p=.049). Coping was associated with lower risk of sick leave for nurses (RR problem-focused strategies = 0.53; 95%CI=[0.37; 0.74], p=.001, and RRphysical [...] CHEVAL, Boris, et al. Associations of emotional burden and coping strategies with sick leave among healthcare professionals: A longitudinal observational study. International Journal of Nursing Studies, 2021, vol. 115, p. 103869 DOI : 10.1016/j.ijnurstu.2021.103869

• Healthcare professionals are at high risk of experiencing various psychological and physical health problems, which in turn increase the likelihood of sick leave. • Previous studies suggest that emotional burden, notably carerelated regrets, might be associated with increased absenteeism, though these studies relied on cross-sectional designs and did not adjust for other critical factors such as safety climate or personality traits.

Introduction
The psychological and physical health problems likely to be encountered by healthcare professionals are well established and https://doi.org/10.1016/j.ijnurstu.2021.103869 0020-7489/© 2021 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ) extensive. Being a healthcare professional increases the risk of experiencing, back pain ( Ibrahim et al., 2020 ), sleep problems , burnout ( Dyrbye and Shanafelt, 2016 ), and depression ( Mata et al., 2015 ). These negative health outcomes have an impact on work efficiency ( Melnyk et al., 2018 ), job satisfaction ( Cheval et al., 2019 ), turnover ( Cheval et al., 2019, Addor et al., 2017, absenteeism and long spells of sick leave ( Dewa et al., 2014 ). In turn, these consequences have a detrimental impact on the organization and quality of patient care ( Letvak et al., 2012 ), and represents a considerable economic burden ( Letvak et al., 2012 ), thereby jeopardizing the sustainability of the healthcare system ( Dyrbye and Shanafelt, 2011 ). A better understanding of the factors supporting retention and reducing absenteeism and sick leave is thus warranted. The current study focus on self-reported absenteeism (sick leave) because it is an important problem complicating unit management and quality of care, and is known to predict future job quitting ( Daouk-Öyry et al., 2014 ).
Multiple factors are likely to contribute to sick leave, such as healthcare professionals' work environment (high workload, night shifts, time pressure) or psychosocial factors (work conflicts, lack of perceived support and lack of an encouraging work culture) ( Roelen et al., 2018 ). In addition, there have been an increasing growth of studies that focus on the emotional burden of healthcare work ( Oh and Gastmans, 2015 ).Emotional burden involves various emotional states such as perception of inappropriate care provided to patients ( Piers et al., 2011 ), moral distress ( Lamiani et al., 2017 ), loss of control during the process of patient care ( Shapiro et al., 2011 ), involvement in medical errors ( Sirriyeh et al., 2010 ), as well as care-related regret. ( Courvoisier et al., 2013 ).
Regret is a normal and frequently experienced emotion ( Frijda, 1994 ). Regret can be defined as the emotion that individuals feel when they believe that the outcome would have been better if they had acted or decided differently ( Zeelenberg and Pieters, 2007 ). Pertinent to sick leave, previous studies revealed that care-related regrets have a negative influence on self-rated health ( Cullati et al., 2017 ), and sleep Schmidt et al., 2015 ) as well as job satisfaction and turnover intention ( Cheval et al., 2019 ;von Arx et al., 2021 ). However, these associations between regret and outcomes are less strong when accounting for the coping strategies used by healthcare professionals ( Cullati et al., 2017 ;Carver and Connor-Smith, 2010 ). While adaptive coping strategies, such as problem solving and acceptance, can be beneficial by helping managing stressful situations, maladaptive strategies, such as rumination, can have deleterious consequences ( Courvoisier et al., 2014 ). Hence, coping strategies can partly explain whether and how a regretted event will influence health and work outcomes.
To the best of our knowledge, only one study has investigated the associations between care-related regrets (i.e., regret intensity, number of recent regrets, and coping strategies) and self-reported absenteeism among healthcare professionals ( Cullati et al., 2017 ). In this study, care-related regret intensity was associated with more frequent sick leave in nurses, whereas adaptive emotionfocused strategies were associated with less frequent sick leave in physicians ( Cullati et al., 2017 ). However, the study relied on cross-sectional data, and could not investigate the dynamics of the associations between regretted experiences over time and sick leave. In addition, the associations were not adjusted for critical factors such as perceived safety climate ( Quillivan et al., 2016 ), personality traits ( Goldberg, 1993 ), or physical activity. For instance, a healthcare professional with a lack of emotional stability may experience higher number or more intense regrets compared a professional with greater emotional stability ( Allen et al., 2014 ). Likewise, a lack of perceived safety climate likely reduces the possibility to report potential problems ( Miller et al., 2019 ), the willingness to speak up about safety ( Alingh et al., 2019 ), or to discuss ones own mistakes in a context of benevolence. This, in turn, decreases the quality of care ( Bower et al., 2003 ;Hann et al., 2007 ), increases the risk of errors ( Haynes et al., 2011 ), and reduces the healthcare professionals' ability to effectively cope with them ( Quillivan et al., 2016 ;Vifladt et al., 2016 ). Finally, physical activity has been shown to have a positive effect on sick leave ( Proper et al., 2006 ;Rongen et al., 2013 ) and to protect both physical and mental health ( Warburton et al., 2006 ;Rebar et al., 2015 ). It is likely that physical activity could help people cope with healthcare related stress situations and mitigate their detrimental effects on health. Thus, in summary, evidence of an independent association between regrets and sick leave are still lacking.
The objective of the study was therefore to investigate the interplay of sick leave with intensity and number of care-related regrets, accounting for critical confounding factors such as personality traits, and perceived safety climate. A second objective was to examine the mediating or moderating effect of coping strategies. Based on previous literature that has shown associations between care-related regret and various outcomes including absenteeism, we hypothesized that regret should be associated with an increased risk of sick leave. Finally, we hypothesized that coping strategy should mediate and/or moderate the associations between regret and sick leave.

Study design and participants
Our analyses used data from the Impact of CAre-related Regret Upon Sleep (ICARUS) cohort study . In short, ICARUS is an international cohort study of newly practicing healthcare professionals working in acute care hospitals and clinics including 52 repeated measurements over 1 year (i.e., weekly assessment). The main goal is to examine the real-time associations between care-related regret and multiple health-related and psychological variables. A random sample of hospitals and clinics were selected from French, English, German, or Danish-speaking countries (e.g., Australia, Austria, Botswana, Canada, Denmark, France, Haiti, Ireland, Kenya, United Kingdom, United States of America). Participants fulfilling all the following criteria were eligible for the study: 1) newly practicing healthcare professionals, 2) speak French, English, German, or Danish, and 3) completed at least one web survey. Participants who had not provided care to patients in the last 6 months were excluded. As incentive, a small donation to a charity (Theodora Foundation) of 0.5 Switzerland franc (SFr) for each completed survey was made.

Primary outcome
Sick leave was measured by asking for each of the seven days of the week if it was a "day work", "night work", "day off" or "sick leave". To fully present the occurrence of sick leave, it is reported in three different ways in the study. First, on a weekly basis, we used a binary variable indicating whether there were at least one day of sick leave on a given week (week with sick leave). This variable was used as outcome of the inferential models, since its frequency of assessment corresponds to the frequency of regret assessment. Two other ways of reporting sick leave were done for descriptive purposes. Specifically, we reported the proportion of participants with at least one sick leave during the study duration ( at least 1 sick leave ). This variable was used in our Table 1 . Finally, we used the proportion of sick leave days out of total shifts worked in a week to account for part-time work ( % sick leave shifts ). This variable was reported to illustrate our data in Table 1 and Fig. 1 . Note. The proportion of night shifts worked , day shifts worked, and sick leave were out of total shift worked.

Explanatory variables
Care-related regrets variables included the number of recent regrets, regret intensity, and care-related coping strategies. The number of recent regrets was assessed by using the following single item: "During the last week, how many patient care situations have there been in which you experienced regret?", with an open (numerical) answer. The regret intensity was measured by using the following single item: "What would you say is the average level of intensity of your feelings of regret in the situations that happened last week?". Here response was given on a visual analog scale (VAS), with response options ranging from 0 (null) to 10 (very high). If participants reported no regret over the last 7 days, the intensity question was not asked and an intensity of 0 was imputed. Care-related coping strategies were assessed with the Care-related Regret Coping Scale for Health-care Professionals (RCS-HCP). The scale is validated in French, German and Danish ( Courvoisier et al., 2014 ;Pihl-Thingvad et al., 2018 ). The scale measures the frequency of use of different coping strategies in relation to regretted experiences in healthcare work. The RCS-HCP contains 15 items divided in 3 subscales: problem-focused strategies (e.g., "I try to find concrete solutions to the situation"), emotion-focused adaptive strategies (e.g., "I try to see the positive side of things"), and emotion-focused maladaptive strategies (e.g., "I turn these situations over in my mind all the time"). Respondents answered each item on a four-point Likert scale ranging from 1 (never or almost never) to 4 (always or almost always). In our sample Cronbach's alpha were 0.84, 0.88, and 0.88, for problem-focused adaptive, emotional focused adaptive, and emotion focused maladaptive strategy respectively.
In line with previous studies , we defined three types of healthcare professionals according to the coping strategies they typically relied on. "Adaptive copers" (i.e., low use of emotion-focused maladaptive strategies), "Maladaptive copers" (i.e., high use of emotion-focused maladaptive strategies), and "Mixed copers" (high use of maladaptive and of adaptive/problem focused coping strategies). Specifically, we used the healthcare professional mean value of each type of coping strategy over time to characterize the strategies each individual typically relied on during the whole study duration. "Adaptive copers" were characterized by a low use of emotion-focused maladaptive coping strategies ( < 1.8 on average on the four-point Likert scale) and by the frequent use of emotion-focused adaptive and/or problemfocused coping strategies ( ≥ 2.2 on average of the two scales). "Mixed copers" were characterized by a high use of emotionfocused maladaptive coping strategies and, concomitantly, by a high use of emotion-focused adaptive and/or problem-focused coping strategies. "Maladaptive copers" were characterized by a high use of emotion-focused maladaptive coping strategies and, concomitantly, by a low use of both emotion-focused adaptive and problem-focused coping strategies. These cut-offs were based on the median scores of each subscale within the current sample consistent with previous studies .
Physical activity focused on moderate-to-vigorous physical activity and was assessed by using two items: "During the last week, how many days did you do moderate (vigorous) physical activity for at least 10 min", with a scale ranging from "no day" to "7 days". The number of days spent in either moderate or vigorous physical activity were used. Associations between care-related regrets and sick leave, according to coper types for nurses and physicians. Illustration of the associations of emotional burden number (left panels) and intensity (right panels) with sick leave, according to coper types, for nurses (top panels) and physicians (bottom panels), with weeks observed indicated below each bar. For readability, bars at 0 on the Y axis are drawn at -1 to make their color visible. The width of the bars represents the frequency of each coping type.
Perceived safety climate was assessed using 3 scales from the Safety Attitudes Questionnaire (SAQ) ( Sexton et al., 2006 ): teamwork climate (e.g., "In this area it is difficult to speak up if I perceive a problem with patient care"), safety climate (e.g., "Medical errors are handled appropriately in this clinical area"), and working conditions (e.g., "All the necessary information for diagnostic and therapeutic decisions is routinely available to me"). Respondents answered each item on a five-point Likert scale ranging from 1 (disagree strongly) to 5 (agree strongly). A possibility to answer "not applicable" was also provided for each item.

Covariates
The following variables were assessed: gender, profession (physicians, nurses, other healthcare professions, e.g., occupational therapists, midwife), night shifts, number of days off, number of perceived medical errors (whether the event eliciting the most important regret in the week was related to an error or not), and consequences for the patient of the regretted situation.

Missing data imputation
Data on covariates were imputed using multiple imputation with the joint modelling approach ( Schafer and Yucel, 2002 ), using the R mitml package and the jomo R packages ( Schafer and Zhao, 2014 ;Grund et al., 2016 ). The jomo package allows to perform imputation of continuous and categorical variables at the first and second level of a given multilevel structure, while the mitml provides a user interface to the former. To ensure valid inference of the statistical analysis ( Black et al., 2011 ), the inherent multilevel structure of the longitudinal data was taken into account in the imputation model. All variables with missing data were imputed with random intercepts, considering the individuals as clusters and profession, number of sick leaves, number of night shifts and number of days off as the independent variables of the imputation model. With these variables included in the imputation model, missing data was assumed to be missing at random. Variables being constant at the individual level, such as personality, were imputed at the second level of the multilevel imputation with a dependence on the profession.

Statistical analyses
We first performed raw and adjusted generalized estimation equations (GEE) with a Poisson log-link, the individual as a cluster, and an autoregressive correlation structure. These models were used to examine the associations of care-related regrets (i.e., number of regrets, regret intensity, and care-related coping strategies), perceived safety climate, personality traits, and physical activity with sick leave in a given week (having at least one day of sick leave on a given week). When testing binary outcomes, the Poisson regression with robust standard error is recommended to accurately estimate the relative risk ( Zou, 2004 ). The advantage of using this model is that relative risks are easier to interpret as a multiplicative coefficient of the baseline risk, compared to odds ratio. The autoregressive correlation structure allows taking into account the longitudinal nature of the data, considering that observations closer together in time are more correlated than observations further apart. Variables measuring care-related regrets, physical activity as well as covariates measured on a weekly basis were decomposed into an intra-individual and an inter-individual component, to identify the contribution of both inter-individual differences and intra-individual dynamics. Specifically, individuals' mean value of a particular variable estimates inter-individual differences. For instance, an individual answering the survey 3 weeks with values in number of regrets of 3, 6, and 0 would receive a mean number of regrets of 3. In contrast, individual's deviation from this mean value at each time point (i.e., week) estimates intra-individual changes (yielding intra-individual changes of 0, 3, and -3 for the same individual as above). For example, healthcare professionals with, on average, lower number of regrets (e.g., 1 by week) may have lower risk of sick leave, compared to their counterparts who experienced, on average, higher number of regrets (e.g., 5 by week) (i.e., inter-individual effects). Likewise, for a given healthcare professional, having a lower level of number of regrets than usual (e.g., 1 regret in a given week relative to 3 regrets on average) may be associated with a lower risk of sick leave (i.e., intra-individual effects).
To examine the role of coping strategies (i.e., maladaptive, mixed, adaptive), we performed three adjusted models. The adjusted models include only the significant predictors of the raw models, with the exception of profession which was adjusted for in all models. This strategy was used for two reasons. First, the number of events (in this case the number of weeks with sick leave) was relatively low, and we could not include all covariates because of issues in the estimation. Second, other covariates could be correlated to a certain extent, potentially leading to multicollinearity issues. In case of non-significance, we not only relied on the arbitrary cut-off of p < .05 which do not reflect the variable importance ( Hayat et al., 2019 ). We also carefully checked the effect size to ensure that the strengths of non-significant associations were close to the null effects. Note that we did not include both number and intensity of regret also to avoid collinearity problems. The first model (M1) included either number of regrets or regret intensity, as well as potential confounders and profession. Then, the selected predictors were included. In a second adjusted model, coping style was added as independent variable to assess potential mediating processes (M2). Finally, in a third adjusted model, the three coping styles and an interaction term between the coping style and regrets was added to M1 to assess potential moderating processes of the coping style (M3). All analyses were first performed without stratifying on professions. Then, adjusted models were stratified on professions (i.e., nurses, physicians, and other). Finally, to examine if respondent's context (i.e., the country in which they work) influence the results observed, we performed a sensitivity analysis where country was included in the models. We restricted the analytical sample (N = 220) by including only countries with at least ten participants (i.e., Austria, Denmark, France, Germany, and Switzerland).

Ethical approval and informed consent
This study was approved by all the relevant local Ethics Committees, and all participants signed informed consent forms. See  for more details on the ICARUS protocol.

Descriptive results
We followed a total of 276 healthcare professionals (87.4% females; mean age = 30.4 ±7.6 years; 29.0% physicians, 52.9% nurses, 18.1% other healthcare professions, e.g., physiotherapists, social assistants) for 23 weeks on average. Physicians had a significantly lower conscientiousness than the other professions, while nurses had a significantly higher level of extraversion. On average, participants experienced over 1 regret per week and the average regret intensity when experiencing a regret was around 4. Nurses used all types of coping strategies less frequently. This was especially true for adaptive strategies (nurses: 2.3, physicians: 2.8, other healthcare professions: 2.6, p < 0.001). When characterizing participants as using mostly adaptive, mostly maladaptive or mixed strategies, the "coping type" differed significantly between professions. Only 45% of physicians used mostly adaptive strategies, compared to more than 50% of nurses and other healthcare professionals. Of note, 32.4% of nurses used mostly maladaptive coping strategies, compared to less than 20% of physicians and other healthcare professionals ( Table 1 ).
Reporting the three estimates of sick leave, the average proportion of weeks with at least one day of sick leave (week with sick leave ), over the whole follow-up, was 3.7%, with 3.8% in physicians, 3.7% in nurses, and 3.4% in the other professions. On average, the proportion of sick leave days out of total shifts worked in a week (% sick leave shifts ) was 1.4%. Overall, 77 (27.9%) participants had at least one sick leave during the study duration ( at least 1 sick leave ), 22 (27.8%) physicians, 43 (29.5%) in nurses, and 12 (23.5%) in the other professions. Fig. 1 illustrates the association of regret number (left panels) and intensity (right panels) with sick leave, according to coper types, for nurses (top panels) and physicians (bottom panels). For readability, bars at 0 on the Y axis are drawn at -1 to make their color visible. The width of the bars represents the frequency of each coping type. Maladaptive coping types were more frequent when reporting more regrets or more intense regrets. The height of the bars represents the frequency of sick leave days out of total shifts worked ( % sick leave shifts ). The Fig. illustrates how nurses' sick leave increased especially with number of regrets among maladaptive copers and physicians' sick leave increased especially with intensity of regrets among maladaptive copers.
In contrast, conscientiousness was negatively associated with sick leave (RR = 0.76; 95%CI = [0.59;0.97], p = 0.026). Also, the mean level and intra-individual variation in number of days off was Table 2 Generalized estimation equations testing the associations of mean level and intra-individual variation in number of regrets with sick leave and the mediating role of coping strategies in healthcare professionals. In other words, one extra point on conscientiousness on the 1-5 scale was associated with 32% (i.e., 1/0.76) reduction in the risk of sick leave. A higher number of days off on average (inter-individual differences) or in a given week (intra-individual changes) were associated with a 59% and a 61% reduction of the risk of sick leave, respectively. And a higher level of physical activity on average or in a given week were associated with a 16% and a 25% reduction of the risk of sick leave, respectively. No significant associations emerged with safety climate, nor with the participants-reported characteristic of the event (considering the regret-eliciting event as a medical error, consequences to the patients, feeling responsible for the event), nor with any of the other covariates. Coping style did not moderate the association between sick leave and regret intensity ( p s > 0.562).

Adjusted models M1 and M2 stratifying on professions
Due to the small sample size, stratified analyses were done only for nurses and physicians, but not for other healthcare professionals.

Regret intensity.
The pattern of results for regret intensity was similar than for number of regrets, except that the associations were stronger for physicians instead of nurses. In M1, sick leave was associated with regret intensity among physicians (RR = 1.24; 95%CI = [1.08; 1.42], p = .002), but not among nurses (RR = 1.10; 95%CI = [0.97;1.26], p = .136). Table 3 Generalized estimation equations testing the associations of mean level and intra-individual variation in regret intensity with sick leave and the mediating role of coping strategies in healthcare professionals. Note. The adjusted models were run separately for regret intensity and number of regrets. The adjusted model includes only the significant predictors of the raw models, with the exception of profession which was adjusted for. Of note, in case of non-significance, authors carefully checked the effect size to ensure that the strength of non-significant associations were close to the null effects. Gender, number of night shifts, medical errors, consequence for the patients, other personality dimensions, and perceived safety climate were tested in the raw models, but were not associated with sick leave. The generalized estimation equations used a Poisson log-link and an autoregressive correlation structure.

Sensitivity analysis
Overall, results of the sensitivity analyses were consistent with those of the main analysis. However, some differences should be noted. First, the main effect of the number of regrets on sick leave was no longer significant. Yet, coping style moderated this association, with a positive association between sick leave and number of regrets significantly higher in maladaptive copers compared to adaptive copers ( p = .028). Furthermore, conscientiousness was no longer associated with sick leave ( ps > .285) and the effect of regret intensity did not remain significant in the adjusted models ( ps > .246). Finally, regarding coping strategies, the negative association between sick leave and problem-focused strategies was no longer significant in either the raw or adjusted models ( ps > .101). On the contrary, emotion-focused maladaptive strategies became marginally associated with sick leave in the adjusted models.

Main findings
This study examined the associations between care-related regrets and sick leave in a multicenter, international, prospective cohort study of novice healthcare professionals. It also investigated whether coping strategies mediated or moderated these associations. Overall, the prevalence of sick leave was relatively low compared to other studies, as could be expected in this cohort of novice healthcare professionals. Interestingly, sick leave prevalence was similar for nurses, physicians, and other healthcare professionals. Yet, this result can be explained by the fact that we assessed sick leave in newly practicing healthcare professionals that had just started working with patients, with on average, a follow-up duration of 24 weeks. Significant differences in absenteeism between professions are more likely to be observable in more experienced healthcare professionals ( Rugless and Taylor, 2011 ).
Number (for nurses) and intensity (for physicians) of regrets were associated with an increased number of sick leaves. In addition, conscientiousness, as well as individuals' differences and intra-individual changes in physical activity, were associated with a decreased number of sick leaves. Finally, the association between number of regrets and sick leave were slightly mediated by coping styles, with problem-focused strategies marginally associated with a decreased number of sick leaves. It is important to note that the models stratified on professions (nurses and physicians) showed that the association between number of regrets and sick leave, as well as the mediating role of problem-focused strategies, were only observed among nurses. At least two differences between nurses and physicians can explain these differences in the results observed between professions. First, the clinical activity differs between these two professions, with nurses spending more time in direct contact with patients. Because of this proximity, nurses may be more emotionally implicated in the care of patients compared to physicians. Second, a qualitative study revealed that professional socialization also differed between nurses and physicians. Unlike physicians, nurses massively used social resources (i.e., cope with their close colleagues) to deal with regrets ( Courvoisier et al., 2011 ), and considered the support of their colleagues as essential. In contrast, physicians relied more on acceptance ( Goldberg et al., 2002 ), a cognitive coping strategy which is rarely observed in the general population ( Zeidner and Endler, 1995 ).

Comparison with other studies
Our finding that care-related regret is associated with higher levels of sick leave is in line with a previous cross-sectional study ( Cullati et al., 2017 ). However, the previous study found an association of sick leave with regret intensity, but not with the number of regrets among nurses. Moreover, consistent with our results, this research found that adaptive coping strategies were associated with less frequent sick leave among nurses. However, unlike our study, it did not find evidence of associations between coping strategies and sick leave among physicians. This discrepancy could be explained by the fact that our study involved a sample of novice healthcare professionals, whereas the previous study involved experienced physicians and nurses, often with 10 years or more of experience. This latter population is thus more exposed to the risk of experiencing difficult events related to patient care. Moreover, this population is more likely to include individuals who deal more effectively with healthcare-related difficult events and situations, and in particular who have learned to use problem-focused strategies, such as talking to colleagues and supervisor to avoid the situation re-occurring. In contrast, novice healthcare professionals may still show a gradient of knowing how to navigate the hospital institution to deal with regret-inducing situations.
Adjustment for coping strategy only partly explained the association between number of regrets and sick leave among nurses. These findings are in line with previous studies that have shown a mediating role of coping strategies between care-related regrets and health outcomes, including for instance psychological wellbeing ( Sirriyeh et al., 2010 ), and self-rated health ( Cullati et al., 2017 ). However, the mediation was partial, suggesting that regret experience is still very relevant for sick leave. It is also possible that novice healthcare professionals rely on coping strategies not measured by the coping scale used in the current study. Of note, in the sensitivity analyses, we observed a moderating effect of coping strategies on the association between number of regrets and sick leave. Here, the positive association between sick leave and number of regrets was more pronounced in maladaptive copers compared to adaptive copers. These findings suggest that coping strategies can not only partly explain the association between regrets and sick leave (i.e., mediation), but also explain the boundary conditions in which the effect of regrets will be particularly pronounced vs. attenuated (i.e., moderation).
Interestingly, the lower level of conscientiousness in physicians, a personality dimension playing a protective role in the emergence of sick leave, contrasts with the usually lower prevalence of sick leave observed in this profession. As a consequence, these findings may suggest that protective personality traits could not be enough to effectively temper the aversive effect of emotional burden after several years of work. This effect must be interpreted with caution because it did not remain significant in the sensitivity analysis. Likewise, the absence of association between the perceived safety climate and sick leave may also reflect the study timeframe. That is, the working conditions related to safety may require some time to have a significant impact (either positive or negative) on the healthcare professionals, which may not be reached in this longitudinal study lasting approximately 6 months. Future research should examine longer-term influence of coping strategies, personality dimensions, and working environment on the emergence of sick leave. Finally, the protective influence of physical activity on the number of sick leaves are in line with previous literature demonstrating the numerous health benefits associated with a more physically active lifestyle ( Godlee, 2019 ). Promoting physical activity among healthcare professional may represent a particularly effective way to reduce the prevalence of sick leave among healthcare professionals.

Strengths and weaknesses of the study
Among the strengths of the present study are the use of an intensive longitudinal design (i.e., 1-year weekly assessment), to better explore associations over time. Likewise, the inclusion of different healthcare professions (i.e., nurses, physicians, and other professions), the international recruitment in a random sample of medical and nursing schools in multiple countries, and the selection of newly practicing healthcare professionals reduces selection bias. In addition, the focus on regret, a normal and common emotion rather than abnormal and less frequent emotion, may enhance the usefulness of this study for all healthcare professionals and not only for people suffering from burnout. Finally, the adjustment of individuals' (i.e., personality) and environmental' (i.e., perceived safety climate) characteristics help reduce potential confounding. However, several limitations of this study should be con-sidered. First, sick leave was measured using self-reported retrospective data and may therefore by subject to recall bias or social desirability. However, recall bias was reduced thanks to the weekly data collection and the social desirability bias was minimized by the use of an internet-based survey, which allows participants to feel more comfortable reporting their sick leave and their experiences ( Sikorskii et al., 2009 ). Second, as participation is voluntary, a potential selection bias cannot be excluded. Nevertheless, participants who had at least one time of measurement were included in the analysis thanks to the statistical approach employed, which limited selection bias due to attrition. Third, because of a rather low sample size within each country, we were unable to provide a meaningful assessment of the differences between countries in the associations observed -although it should be noted that all participants were included in the analysis, regardless their number in a given country. Likewise, because the study did not recruit enough participants within each gender and professions strata, we were unable to assess whether gender differences between professional groups and expression of regret could be observed. Future studies with larger sample sizes for each country should be conducted in order to investigate any country or gender related differences. Finally, to guarantee the schools and participants anonymity, we did not capture any data on support programs. This feature limits the ability to evaluate how the content of the graduate programs can impact the associations between regrets and sick leave observed in this study. For example, we can expect that a support program emphasizing not only that regret is an unavoidable corollary of providing care, but also how to effectively cope with it, may be useful to prevent the future detrimental effects of regrets on healthcare professional own health.

Conclusion and implications
Accumulation of care-related regrets was associated with more sick leave among nurses, even after adjusting for individual and environmental factors. These associations were partially mediated by coping strategies. Because sick leave has serious impact on unit management, quality of patient care, and the healthcare system sustainability, helping healthcare professionals, especially nurses, to effectively deal with regretted situations is warranted. Integrating training modules in the curriculum aiming to prepare future healthcare professionals to effectively deal with difficult events and situations, especially by facilitating their use of problem-focused coping strategies, may lead to beneficial outcomes not only for healthcare professionals' own health and well-being, but also for the quality of patient care. Of note, physical activity, which has been shown to protect both physical and mental health on various context ( Warburton et al., 2006 ;Rebar et al., 2015 ), seems also effective in reducing the risk of sick leave among young healthcare professionals. Interventions that promote physical activity among healthcare professionals may be particularly suited to protect against psychological and physical health problems.