Abstract
BACKGROUND:
Coping with regret has a substantial impact on wellbeing and mental health, but has rarely been investigated in an occupational setting.
OBJECTIVE:
To translate the Regret Coping Scale for Health-Care Professionals (RCS-HCP) and explore internal consistency, construct-, criterion- and predictive validity.
METHODS:
The instrument was translated using forward- back method. The qualities were evaluated with a sample of 2758 social educators using exploratory and confirmatory factor analysis as well as Cronbach’s alpha, Pearson correlation, and multivariable regression.
RESULTS:
The translated instrument showed a trend similar to the original instrument. A 10-item version resulted from the research being reported. The reduced RCS-HCP showed improved fit (Full model, 15 items); CFI = 0.91, TLI = 0.89, RMSEA = 0.66, PClose = 0.000 and BIC = 1392 vs. (Reduced instrument, 10 items); CFI = 0.97, TLI = 0.96, RMSEA = 0.05, PClose = 0.499 and BIC 307. This instrument had acceptable internal consistency for short scales (Cronbach’s alpha = 0.65, 0.69 and 0.84 respectively). The subscales correlated as expected with measures of health and occupational factors, coefficient ranging from 0.182 to 0.399. Also, the RCS-HCP predicted stress three month later ΔF[3,2747] = 15.1, p < 0.001, but with very small effect ΔR2 = 0.01, p≤0.001.
CONCLUSIONS:
The 10-item Danish version of the RCS-HCP is a valid instrument for measuring coping with regret in health related work.
Introduction
Regret is a psychological reaction to a decision or action that has unwanted or negative effects [1, 2]. The experience of regret is comprised of negative thoughts and emotions of self-blame and a wish to have acted differently [3]. Regret is often a negative experience but also a factor of motivation for correcting a mistake and doing better in the future [4]. Experiencing regret is a normal and everyday event, typically regulated either by seeking to change our decisions, by seeking alternative solutions or directly regulating our emotions [5, 6]. However, when regret is poorly regulated, with constant rumination and self-blame, regret has been associated with reduced quality of life and stress [7] as well as depression [8] and anxiety [9]. Furthermore, since rumination and self-blame are distinct but closely related phenomena [10], poor regret regulation might also be associated with the development of posttraumatic stress, since rumination is a known risk factor for PTSD [11].
The way we cope with regret at work is not solely dependent on individual factors, but also on the organizational context that influences our behaviors and the evaluation of the decisions made. In an occupational setting, how you cope with a problem is influenced and can be formed by organizational values and norms, as well as social and managerial pressure [12–15]. Also, general demands such as workload, role ambiguity, and control can be pivotal to decisions made in a work setting [16–19]. Self-blame, as part of regret, in an organizational context, may interfere with professional identity and can cause self-devaluation of one’s professional skills. It can be argued that professional self-devaluation influences the worker’s appraisal of coping resources. In this perspective, coping with work-related regret is an important factor in understanding development of mental health problems at work.
Existing research on regret primarily focuses on the nature of regret and decision making [3, 21], and to a lesser extent on how we handle regret (regret regulation) [3]. Research on regret regulation in an occupational setting is almost non existant, and to our knowledge, only the studies regarding the development of the Regret Coping Scale for Health-Care Professionals (RCS-HCP) [22, 23] specifically target regret in an occupational context. The RCS-HCP is an instrument that assesses coping strategies among health-care workers, when experiencing regret for actions or decisions made in relation to patient care. The RCS-HCP was developed as a consequence of research showing that regret among health care professionals had an impact on both patient care and wellbeing of health care professionals [23]. The RCS-HCP showed great promise in the original study, were it was validated on a sample of nurses and physicians at a general hospital in the French speaking region of Switzerland. The RCS-HCP has also been validated in a German sample of healthcare workers in a hospital setting, in the German speaking region of Switzerland [24]. However instrument properties across other patient-related occupations and nationalities, are still unexplored.
The applicability of RCS-HCP to a broader spectrum of patient related work, such as psychiatric institutions, and care homes, as well as residences and homecare for impaired adults and elders, could be valuable since these occupations are known to have increased rates of stress, depression, anxiety and PTSD [25, 26]. Coping with regret in patient care might be a new way to understand and prevent these negative health outcomes.
We wanted to introduce the RCS-HCP in the “everyday violence” study, a prospective cohort study on a sample of social educators working with adults. The aim of the current study was to translate the original version of the RCS-HCP into Danish and to evaluate the Danish instrument in relation to internal consistency, construct-, criterion- and predictive- validity.
Methods and materials
Instrument description
The RCS-HCP is a 15-item scale developed by Courvoisier and colleagues. It assesses coping strategies among health-care workers, when regretting actions or decisions made in relation to patients and their relatives [23]. The RCS-HCP is formulated in French, and describes the feeling of regret as the emotion experienced when a decision made in regards to patients turns out to be either inappropriate, ineffective or futile [23]. The RCS-HCP covers three constructs of coping with regret; Problem-focused coping (e.g., talking to supervisors or colleagues and finding specific solutions), Maladaptive emotional coping (e.g., rumination and self-blame) and Distancing emotional coping (e.g. re-appraising the situation by focusing on the wider context or trying to create emotional distance to the situation). Each construct is measured as the mean of a 5-item scale, where each item pertains to a given coping strategy, and is answered using a 4- point Likert scale where 1 = “never/almost never” and 4 = “always/almost always”.
In the original study, each scale showed good internal consistency with Cronbach’s Alpha 0.89 for each of the three scales separately, and good test retest reliability with coefficients of 0.78, 0.82, 0.79 respectively. Also, the instrument showed good construct validity both in regards to factor loadings in the PCA analysis and in association with criterion variables such as self-rated health, depression, sleep problems, and job- and life quality. Here both problem-focused and distancing coping were associated with positive health and life quality outcomes and maladaptive emotional coping was associated with negative health and life quality outcomes [23].
Translation
The RCS-HCP was translated using forward and back translation. The forward and back translation were carried out by two bilingual translators both native Danish but educated in French at master degree level with more than 25 years of experience in translating and teaching French. One translator made the forward translation and the other made the back translation. The translators worked independent of each other, and were instructed to make conceptual translations aimed at a broad audience rather than word by word translation. The French to Danish translation was revised and minor adjustments were made to make the questions suitable for a Danish survey. The back translation was then made and presented to the original author for acceptance. The Danish translation was tested on a small group of clinical assessment psychologists. They were interviewed in regards to their understanding of the questions, and if they thought the question could be worded differently to obtain a better understanding. The interview results were analyzed by both the first and second authors and minor revisions were made, constituting the final Danish version of the RCS-HCP.
Participants
For assessment of reliability and validity we used the baseline sample of “the everyday violence project”. The sample consisted of social educators (N = 2758) working with adults with special needs, e.g., physically and mentally disabled, psychiatric patients living outside a hospital setting, or adults who are socially marginalized (e.g. alcoholics or substance abusers and homeless people). Social educator training consists of a 3-year undergraduate degree. The function of the social educator is to train and strengthen clients’ social skills and daily functioning either in institutionalized settings, in private homes or through proactive contact seeking out their clients where they live, e.g. in the streets and in shelters. Aside from social training, occupational functions include: assessment and planning of social and healthcare initiatives, documentation and, for some, tasks such as assisting in bathing, dressing and eating.
Data collection
Participants were invited through their union which represents approximately 80% of the Danish social educators working with adults. Surveys were distributed in May and June 2016, through email, with a total of three reminders. Of the 12,070 potential participants invited, 3, 212 filled out the baseline questionnaire. The baseline sample was compared to both non-respondents, normative data [27] and a comparable group of working people from the 2014 Danish worker cohort study [28]. The baseline sample did not differ in regards to gender, work area, level of burnout, self- rated performance ability or self-rated health (non published results – Authors may be contacted for detailed information). However, the sample was significantly older than the non-respondents (sample mean age = 47.9 SD = 9.9, non-respondent mean age = 46.5 SD = 10.7). This was expected due to a general trend in Danish survey response rates [29].
Of the total sample, we excluded all individuals who were either not currently employed (n = 20), were employed in a leadership position (n = 297) or did not have Danish as native language (n = 141). The final sample consisted of 2758 respondents.
Further, we used longitudinal data from the cohort study for one of our analyses. Here all respondents included from the baseline sample, who had also completed the monthly short survey at three months, were included (N = 1881, 68%). Drop out analysis was not made on this subsample.
Statistical analysis
All analyses were made using SPSS 24 and Amos 16. Missing data amounted to 0.97% distributed on 46% of the respondents, with no identifiable pattern. We therefore used Expectation maximization as the method for replacement of missing values [30].
Since the Danish RCS-HCP was introduced in another working culture, we chose to use both exploratory and confirmatory analysis, to evaluate the factor structure of the 15 items.
We randomly divided the sample in two groups (N = 1379). In one group we conducted Principal Component Analysis similar to the original validation study. Extraction was based on eigenvalues >1 and scree plot evaluation. We used oblique rotation (Promax), for possible factor correlations, with suppression of factor loadings <0.3. Different models were assessed based on the total variance explained, and factor coefficients, where items with small or ambiguous loadings (loading on more than one factor) were eliminated one by one re-running the analysis each time. This method of exploration have been used in other studies [31–33]. Based on the explorative factor analysis we ended up with an adjusted 3 factor model. This model was tested against the original model using confirmatory factor analysis, with Maximum Likelihood estimation. The confirmatory analysis was made in the second group. The models were compared in relation to comparative fit indices, the Tucker-Lewis fit index (TFI) and the Comparative fit index (CFI). The models were also compared with absolute fit index, the Root mean square error of approximation (RMSEA). Both the x2/df. and the goodness of fit index, were omitted because of sensitivity to large sample sizes [34]. The models were further compared directly, using the Bayesian information criteria (BIC), with a lower BIC indicating a better model fit.
To evaluate scale consistency, we used Cronbach’s alpha statistics on each of the three scales. We further assessed criterion validity with Pearson’s correlation statistics measuring correlation with “self-rated health”, “work burnout”, “mental health problems” and “workplace social capital”. Finally, we assessed predictive validity for perceived stress after three months. Here we used multivariable regression modelling using each of the RCS-HCP dimensions as predictive variables and “stress” after three month as outcome variable. A two-step hieratical regression model was used. At step one the following known predictors were entered: gender, age, years of experience, workload, social capital- linking and- bonding and mental health problems at baseline. At step two the RCS-HCP subscales were entered. Forced entries were used at both step 1 and step 2.
Measure of “mental health problems” and “work burnout” did not meet the criteria of normal distribution and bootstrapping with 5000 samples were used in both correlation and regression analysis, to secure robust measures.
Measures
All measures were part of the larger “every day violence survey” consisting of a total 180 items.
The RCS-HCP consisted of the Danish translation of the original French version described in Section 2.1. For the exploratory and confirmatory factor analysis the full 15 item instrument was used. In the following analysis the reduced 10 item version, defined through the factor analysis was used.
Self-rated health was included for direct comparison with the original validation study, using the first item of the Danish version of the SF36 single item “How would you describe your health in general” answered on a 5- point Likert scale where 1 = “bad” and 5 = “excellent”, as a continuous variable.
Mental health problems were assessed with the Global severity index (GSI) of the SCL-10, a validated short form of the SCL-90, and considered a valid tool for measuring common mental health disorders [35]. This scale consists of 10 items stating different symptoms answered on a 5- point Likert scale where 1 = “not at all” 5 = “extremely much”. The GSI score is the mean of all items, and treated as a continuous variable.
Work Burnout was measured with the Copenhagen Burnout Inventory (CBI) that has been validated on different Danish samples [27]. The inventory consists of three constructs e.g., personal-, work- and client-burnout. We chose to use work burnout, since this measure is related directly to the perception of wellbeing and functioning at work. The scale consists of 7 items related to the symptoms of burnout and answered on a 5-point Likert scale where 1 = “never/almost never” and 5 = “always”. Burnout was measured as the mean of all 7 items, and was treated as a continuous variable.
Workplace social capital was included since the RCS-HCP dimension of “problem focused coping” includes involvement of both colleagues and managers. Further workplace social capital is associated with well-being at work [36] and thus comparable to the “satisfaction with work” measure used in the original study. We used the validated workplace social capital survey from the Danish National research institute of work environment [36]. We chose to use scales for social capital within coworker groups (bonding social capital) and between employees and daily management (linking social capital) since these adheres directly to the RCS-HCP questions. Both scales are measured as the mean of the 4 items all answered on a 5 point Likert scale and normalized on a 0–100 scale according to the test manual [36].
To measure stress at three month we used the single stress item from the Occupational Stress Questionnaire that has been validated in several different working populations [37]. The item consists of a short narrative on the symptoms of stress and asks to the degrees of symptoms within the latest days. The item is answered on a 5 point Likert scale where 1 = “not at all” and 5 = “very much”, and is treated as a continuous variable.
Based on existing literature we chose age, gender and years of experience [25], as well as workload and mental health problems at baseline [38], as predictors in the regression model.
Age was measured as a continuous variable in whole years. Also, years of experience was measured as a continuous variable of total years working as a social educator. Workload was measured with the scale of “quantitative demands” from the Copenhagen psychosocial questionnaire (COPSOOQ) [39]. The COPSOQ is validated across several cultures and working populations [40–43]. The scale consists of 4 items answered on a 5 point Likert scale, calculated as the mean of the four items, normalized to 0–100 scale according to the test manual [39].
Ethics
According to Danish law, survey based studies do not require approval by the Scientific Ethics committee. All respondents gave their informed consent as part of their participation in the electronic questionnaire. The project procedure for treatment of sensitive data, was approved by the Danish Data Protection Agency; Journal no.15/96549.
Results
Translation
The original author approved the back translation but felt that on item 6 “I feel worthless and incompetent”, had too strong a valence, which is why we changed it to “I feel worthless”. Also, item 15 “I try to put the situation in perspective” was commented on as a little different in respect to meaning. This item was adjusted to a Danish wording closer to “I try to see the importance of the situation in a larger perspective”. When piloted on the panel of psychologists item 13 “I expose the situation to colleagues to improve our practices” was commented on, in regards to the word “practices”, and this wording was adjusted. All other items were deemed to be easy to understand and capturing the intended meanings. The face validity of the Danish instrument was acceptable.
Population socio-demographics
The baseline sample consisted primarily of women (80.5%) with a mean age of 47 years and a mean of 16 years of experience. Mean levels of burnout (1.38), mental health problems (0.49) and self-rated health (2.46), was as expected as these people were currently employed. The rated experience of workload (mean: 50.2), and social capital (mean: 62 and 67 respectively) indicates that work environments were about average and not extremely stress full. Ratings on the three dimensions of the RCS-HCP indicates that maladaptive emotional coping is generally lower than both problem solving and distancing emotional coping with a mean of 0.7 vs. 1.6 and 1.8 (Authors may be contacted for detailed information).
Exploratory analysis
Principal component analysis on the full 15 items had good sampling adequacy KMO = 0.83 with none of the single items below 0.73. Three factors were extracted based on eigenvalues >1 and they explained 52% of the variance combined. The scree plot also indicated 3 factors. As can be seen in Table 1, the items loaded on the same dimensions as in the original study, but 4 items had rather low factor loadings <0.6 and some had loadings on more than 2 factors indicating a poor fit of the full model. Omitting items with low factor coefficients, step by step, left a 10-item model. The new model still presented three factors with eigenvalues >1, explaining 64.9% of the total variance combined. Sampling adequacy was still good, KMO = 0.79 with none of the single items rated below 0.7.
Factor coefficients for each item on the three factors: problemsolving coping (Prob.), maladptive emotional coping (Emo.), and distancing emotional coping (Distan.), for both the full and the 10-item instrument. (The English translation is from the original article and not a translation from the Danish items)
Factor coefficients for each item on the three factors: problemsolving coping (Prob.), maladptive emotional coping (Emo.), and distancing emotional coping (Distan.), for both the full and the 10-item instrument. (The English translation is from the original article and not a translation from the Danish items)
Table 1 show that the pattern matrix was unambiguous and clearly stating three separate factors with acceptable factor loadings >0.7. All the items loaded on the same dimensions as in the original study, and the 10-item model seemed to have a better fit to the data.
Extracting at eigenvalues >0.7, and looking at the scree plot could suggest a possible 4 factor model which was explored but then rejected due to small loading coefficients and only one item loading on factor 4, but also loading on factor 3. Also, a 2-factor model was tested due to double loadings of several factors in the full model. The 2-factor model was also rejected due to low variance explained 43%, 2 items with no factor loadings and 5 ambiguous items loading heavily on more than one factor.
The reduced instrument had superior fit compared to the original instrument. The full model only met acceptable fit in CFI measure, CFI = 0.91, with TLI = 0.89 and RMSEA = 0.066, 90% CI (0.063–0.069) PCLOSE = 0.000. In comparison, the 10-item model showed approved and acceptable fit in all of the 3 fit measures; CFI = 0.97, TLI = 0.96 and RMSEA = 0.05, 90% CI (0.042, 0.058) PCLOSE = 0.499.
Further the reduced model improved the BIC from BIC = 1392 to BIC = 307.
scale reliability
Cronbach’s alpha for the three scales in the adjusted model were as follows; Distancing emotional coping = 0.65 (3 items), problem solving coping = 0.69 (3 items) and emotional maladaptive coping = 0.84 (4 items). Although both the problem solving and the distancing emotional scales had alpha coefficients less than the general accepted 0.7, it has been argued, that for scales with only few items alpha coefficients >0.6 are acceptable [44, 45].
Criterion validity
Overall the adjusted model showed good criterion validity. All three subscales correlated as expected and in the same pattern as in the original study, as shown in Table 2.
Pearson’s Correlation coefficients with Bootstrap 95% CI presented in brackets, based on 5000 samples
Pearson’s Correlation coefficients with Bootstrap 95% CI presented in brackets, based on 5000 samples
Note *p≤0.001.
Maladaptive emotional coping was negatively correlated with both linking and bonding social capital. At the same time, maladaptive emotional coping was positively correlated with negative health measures i.e. mental health problems, work burnout and self- reported health. For problem solving coping and distancing emotional coping the correlations were inverse, with positive correlations to both types of social capital, but negatively correlated with mental health problems, work burnout and self-rated health.
Table 3 shows the results from the two step hierarchical regression model for the 10 -item instrument. The Step one model significantly predicted stress after 3 month F (7, 2750) = 71.36, p < 0.001 with small to medium effect, R2 = 0.15. Entering the subscales of the RCS-HCP significantly improved the prediction model ΔF(3,2747) = 15.1 however with small effect ΔR2 = 0.01, p≤0.00. Only emotional maladaptive and distancing emotional coping acted as significant predictors in the full model.
A 2- step hierarchical regression model presenting regression coefficients (B), with 95% bootstrap CI, based on 5000 samples, standard errors, standardized coefficient(β) and level of significance
A 2- step hierarchical regression model presenting regression coefficients (B), with 95% bootstrap CI, based on 5000 samples, standard errors, standardized coefficient(β) and level of significance
Note: Step 1, R2 = 0.15, p≤0.001, Step 2, Δ R2 = 0.01, p≤0.001.
The goal of this study was to translate the RCS HCP into Danish and explore internal consistency and validity in relation to another working population. The original French version was translated adjusted and back translated with only minor comments on two items from the original author. The adjusted translation was approved by a panel of 5 assessment psychologists and was deemed to have acceptable face validity. In the exploratory principal component analysis, the Danish instrument showed the same trend as the original instrument. Three factors were extracted each with loadings from 5 items in the same pattern as seen in the original study. However, item 2,5,14 and 15 had factor loadings less than 0.6 and 14 and 15 also loaded heavily on another factor, indicating a less than optimal fit on our data. In comparison, only item 14 had a factor loading less than 0.6 in the original study, and this was also the only item with ambiguous loading.
Reducing the Danish instrument, by omitting items with small factor loadings one by one and rerunning the analysis each time, yielded a 10-item instrument with acceptable and unambiguous loadings above 0.7, indicating a better fit. Through confirmatory factor analysis it became evident, that only the 10-item instrument showed satisfactory fit in both the comparative and absolute fit indices, and also reduced the Bayesian information criteria substantially. The reduced instrument showed acceptable internal consistency for small scales in each of the three subscales, although Cronbach’s alphas were lower than in the subscales of the original study. Criterion validity was good and similar to the original study. Here the problem solving and distancing subscales both correlated negatively with measures of mental health problems, work burnout and self-rated health, whereas the maladaptive emotional coping scale correlated positively with these measures. In the original study, problem focused and distancing emotional coping correlated negatively with the self-rated health and with the CEDS-10 depression scale, and the maladaptive emotional subscale correlated positively with these measures.
We found that problem focused and distancing emotional coping correlated positively and maladaptive emotional coping negatively with both measures of workplace social capital. Similar findings were reported in the original study where maladaptive emotional coping was found to correlate negatively with measures of work life quality, and the problem solving and emotional distancing coping, were positively correlated with this measure.
Finally, our study showed that the 10-item RCS-HCP instrument significantly improved prediction of stress three month later. However the effects size was very small and significance might be a result of large sample size. Further studies are needed in order to explore whether the RCS-HCP is in itself a predictor of stress or weather coping with regret interacts with the existing predictors of stress.
That the reduced model fitted the data better than the full model when translated into Danish could be due to differences in work culture between a hospital setting and social educator institutions. In general hospitals, there are procedures, regulations and accreditations in regards to reporting errors [46]. This is not the case in special educator institutions. Awareness of and a readiness to find solutions to mistakes can thus be expected to be more prominent at a general hospital and could influence responses to the omitted item 3 “I try to find concrete solutions to the problem”. Also, patients and clientele differs between the two work settings, which might affect possibilities of coping. For example “Discussing the problem with the patient and his/hers relatives” as suggested in the omitted item 2, might be less obvious and maybe impossible when working with mentally disabled or homeless people, compared to working with hospitalized patients.
Culturally different work environments could also be part of the explanation. In the Danish work-environment there is a strong tradition for democratic work processes and support is seen as an essential factor of workplace interaction [47]. Looking at the reduced problem solving scale, all items focuses on interactive and supportive problem solving, this might be a result of a general trend in handling problems in the Danish work environment. However, further research in cross cultural work environment is needed to test this explanation.
Limitations
There are several limitations to consider in the present study. In the translation process, we only used one translator in each step of the forward and back translation, as opposed to the recommended use of two translators [48]. The use of only one translator reduced the possibility of comparing different understandings in the translation, and could have influenced some of the nuances of the single items. However, when presented to the original author as well as to the group of assessment specialists, only small discrepancies were pointed out, on three items, and only one of these items posed a problem in the factor analysis. Also, all of the items loaded on the intended subscales, which indicates that the meaning of each item was preserved in the translation process.
Sampling might also have been affecting the study results. Using two completely separate samples would have been preferable to splitting the sample for exploratory and confirmatory analysis. Considering that the sample was representative for the population, that the sample split was randomized and that subsamples were large, we deemed that the sample split procedure was acceptable and not a cause for any systematic error in our results. Another problem with sampling is in regards to diversity of professions. Our exploration of the RCS- HCP in different types of health care was only performed on one other profession. It would have been preferable to have multiple work settings including a general hospital for better comparison and evaluation of the instrument qualities. Indeed further studies across work populations are recommended.
The overall framework of the surveys in which the RCS-HCP was presented might have an impact on our results. In the original study, the RCS-HCP were presented in relation to a regret intensity scale focusing on the most regretted situations in work life. In comparison, the RCS-HCP in the Danish version was presented in a survey focusing on violence, threats and work environment in general. These different placements of the survey items might influence the understanding and answering of the questions [49]. Although we tried to counteract this effect by using transitional texts, the different overall framework of the survey might be a concern in the present study. Further validation of the RCS-HCP should as a minimum include the regret intensity scale when presenting the RCS-HCP.
Conclusions
The RCS-HCP could be an important tool in discovering how coping with regret in health care influences the development of burnout and other work related mental health problems. In the original study the RCS-HCP showed good promise with strong measures of reliability, construct and criterion validity. These qualities are preserved in the Danish translation, although in a reduced 10-item version. The RCS-HCP is thus a measure, which can potentially enrich our understanding of the development of work related mental health problems, when working with people. However, since this study only explored the RCS-HCP with a sample of social educators, further studies across different types of health care settings, is recommended.
Conflict of interest
None to report.
Footnotes
Acknowledgments
This project was in part supported by a grant from the Swiss National Fund (100019_166010) and by The Danish Working Environment Research Fund (23-2015-03-20150018099).