Compassion and Self-Compassion: Counterfactors of Burnout in Medical Students and Physicians

Abstract

Objective

To examine the relationship between self-compassion, compassion for others and Burnout in medical students and physicians.

Methods

A cross-sectional study was conducted. Medicine students and general physicians from two Colombian cities participated (n = 359). The Compassion Scales and the Maslach Inventory were administered. An Exploratory Structural Equation approach was used for validating new measures and testing for relationships between latent variables.

Results

Most participants were students (85.9%), mean age was 22 years (SD = 7), 55.2% were female, 62,6% dedicated more than 48 weekly hours to study or practice, while physicians had worked a mean of 10.34 years (SD = 8.67). Self-compassion and Compassion for others action subscales were validated, but engagement subscales of were not. Participant´s compassion actions for others and self-compassion actions are negatively related to depersonalization and emotional exhaustion, respectively. Additionally, compassion dimensions were positively associated with professional accomplishment.

Conclusion

Our findings indicate that compassion and self-compassion actions inversely relate to different components of Burnout and could constitute protective factors against the stress of healthcare. Compassion and self-compassion training programs for medical students and physicians might be an alternative to avoid Burnout, diminishing physicians’ depersonalization and emotional exhaustion while enhancing their professional accomplishment.

Keywords

Self-compassion compassion for others burnout medicine students physicians

Introduction

Burnout (BO) is described as an aggregation of emotional exhaustion, depersonalization, and low personal accomplishment (Rotenstein et al., 2018). In the case of physicians, BO seems to be caused by chronic stress related to medical practice (West et al., 2018). It affects both physicians and patients. Hence, it is of primary concern. The literature signals the negative consequences of BO on physicians' productivity, wellbeing, health and the care they provide to patients, representing a current public health issue (West et al., 2018; Yates, 2020). Yet, some scholars still call for a consensus definition and warn against drawing final conclusions concerning its prevalence and relationships with other variables (Rotenstein et al., 2018; Yates, 2020).

Given the impact of BO on wellbeing and health outcomes both for physicians and patients, solutions against it call the attention of practitioners and researchers (West et al., 2018). A growing body of evidence recognizes the potential role of compassion and self-compassion and related practices to counteract BO effects (Hashem & Zeinoun, 2020; Kalra et al., 2018; Yates, 2020). In fact, compassion has been found to reduce moral distress, burnout and occupational stress in healthcare providers, and also increases work engagement, job satisfaction, and retention (Sinclair et al., 2021). On the other way, excessive stress, burnout, and depersonalization have been considered inhibitors of compassion in healthcare providers (Crawford et al., 2014). Therefore, while some authors suggest that compassion affects burnout, other scholars indicate that the former is affected by the latter. These contrasting perspectives call for empirical research investigating the relationship between compassion and burnout.

Physician compassion has called the attention of scholars for several decades now. They posit that there are specific relationships between BO and compassion, a cornerstone of the Buddhist medical literature (Carmel & Glick, 1996; Kalra et al., 2018). Whereas exposure to stress may lead to Compassion Fatigue (CF), self-compassion should help combat such a state. This is due in part to the fact that self-compassion implies being kind towards oneself when encountering suffering and failure. It also has been associated with positive affect and healthy behaviors (Sirois et al., 2015).

Before determining if and how compassion reduces physicians’ BO, there is a need for a more precise conceptualization of compassion and empirical evidence about the existence of the link between compassion and physicians’ BO. Our study aims at furthering the conceptualization of compassion and disentangling its relationship with BO. The antecedents of BO seem to be both individual and institutional (Montgomery, 2014). Concerning the individual-level correlates, several studies are focused on the relationship between BO and compassion. Most of those studies inquire about the relationship between CF, Compassion Satisfaction (CS), and BO (Fernando & Consedine, 2014). Now, the concept of CF is blurry and sometimes confounded with BO (Fernando & Consedine, 2014), while CS is limited to represent an emotional state (fulfillment).

Recently, Gilbert proposed a new approach to study compassion empirically (Gilbert, 2017, 2019; Gilbert et al., 2017). From the author´s perspective, compassion is defined as a “sensitivity to suffering in self and others with a commitment to try to alleviate and prevent it” (Gilbert, 2019). This new conceptual perspective is broader than CS and can refresh BO studies mainly focused on its relationship, still unclear, with CF.

We believe that there is an opportunity to contribute to this emerging and broader conceptualization of compassion and, in that way, help understanding the relationships between physicians’ BO and compassion. To our knowledge, there are no studies about the relationship between compassion (from this perspective) and BO. There is a lack of studies validating this new conceptualization of compassion in samples of physicians and, virtually, no research has validated this construct in Latin American Countries. Furthermore, although some studies inquire about physicians’ compassion and burnout, such studies use first-generation statistical analyses and are carried out in developed countries (Gribben, Kase, et al., 2019). These analytical approaches prevent a simultaneous validation of the hypotheses and the properties of the corresponding scales. However, construct validation is fundamental when describing, predicting, and explaining psychological phenomena (Furr, 2011).

Compassion

As a concept, compassion is rooted in ancient philosophical and spiritual traditions and has been considered a main trait of a great physician by the Buddhist medical literature (Kalra et al., 2018). Substantial research on the neurophysiological, psychological, and social dimensions of compassion and compassion training has been conducted in the last 3 decades, highlighting its value to counteract suffering and anti-social behavior (Gilbert, 2020; Gilbert et al., 2017). There are different theoretical approaches to compassion and its application in psychotherapy, some of which have identified a particular core cluster of psychological processes and attributes. For instance, Gilbert has proposed a biopsychosocial process approach to compassion, by which it is considered an evolved strategy necessary for survival and reproduction, and a basic, personally experienced motivation. This evolutionary approach seeks the origins of compassion in the evolution of caring motives and behavior and understands compassion as a multi-faceted process entailing different competencies (Gilbert, 2020).

Contemporary notions of compassion describe it as a process triggered by the perception of suffering and having cognitive, affective, and behavioral dimensions towards suffering alleviation (Dutton et al., 2014; Jazaieri et al., 2013). Compassion may be experimented for others, towards one’s self, or from other people (Gilbert et al., 2017). Compassion and self-compassion are considered positive emotions that contribute to psychological flexibility and resilience to support the person´s health and wellbeing as well as that of others. They have been shown to enhance self-care (Mills et al., 2018) and confidence in healthcare staff, and diminish stress (Kemper et al., 2019) and burnout (Crawford et al., 2014; Sinclair et al., 2021).

Burnout

BO is defined as an occupational phenomenon, resulting from exposure to continuous and unmanaged work-related stress (WHO, 2019). BO is also a multidimensional construct composed of three different dimensions: Depersonalization (D), Emotional Exhaustion (EE), and Professional Accomplishment (PA). While there is controversy over the exact figures of BO in physicians (Rotenstein et al., 2018), a general agreement exists regarding its high prevalence in that population.

Most studies inquiring about the relationship between compassion and BO have considered the concept of CF. This form of secondary stress and its link with BO have been studied mainly in developed countries. Gribben et al. (2019; Gribben, MacLean et al., 2019) studied CF, CS, and BO in pediatric critical care and emergency physicians in the United States. Defined as an emotional fulfillment coming from caring for patients (Stamm, 2002), CS is deemed a remedy against CF, and evidence shows that it is negatively related to BO (Gribben et al., 2019, 2019).

The concept of CF has some issues to consider. Fernando and Consedine (2014) have called to the attention that it seems to contradict the construct of compassion itself as it implies that being compassionate is tiring. On the other hand, Gilbert (2009, 2019) and Gilbert & Choden (2013) proposed a broader definition of compassion, as outlined before. This approach to compassion in physicians is relatively new (Gribben et al., 2019, 2019). And, as in the case of studies on compassion in physicians in general, unexplored in developing countries. It also implies two different psychological processes: engagement and motivated action (Gilbert et al., 2017)

Another problem associated with the concept of CF regards its differentiation from BO. The relationship between these two concepts is not clear, and they are often confounded (Fernando & Consedine, 2014). Although the correlations between CF and BO are high, they do not reach levels to suspect discriminant validity problems (Gribben et al., 2019, 2019); BO is considered by some scholars as an aspect of CF, and the corresponding measures of the latter include the former as a subscale (Bride et al., 2007).

In turn, CS is related to both CF and BO. In fact, the Professional Quality of Life (ProQOL) instrument measures the three concepts and treat them as subscales (Cavanagh et al., 2019). Variables associated with CS are positive and include age, experience, and work environments (Cavanagh et al., 2019). Some studies conducted in the United States indicate that CS has negative correlations with BO (Gribben et al., 2019, 2019; Weintraub et al., 2016). These results suggest that compassion as a general construct would be negatively related to some dimensions of BO, namely D and EE, while its relationship with PA would be positive.

Gilbert et al. (2017) seem more in line with the general definition of compassion than CF. Different from CS, which is conceived as a single dimension of emotional fulfillment, the conceptualization proposed by these authors implies two different general dimensions, namely, self-compassion and compassion for others. The former refers to having a supportive and compassionate rapport with oneself. Compassion for others is deemed to be “the most basic focus for compassion” and implies, among other traits, a motivation to be helpful and to be able to do or try doing something to prevent and relieve suffering (Gilbert et al., 2017). When operationalizing compassion for others, the authors divided the dimension into two subscales: Compassion for others Engagement (COE) and Compassion for Others Action (COA). Similarly, Self-compassion is divided into two other subscales: Self-compassion Engagement (SCE) and Self-Compassion Action (SCA). A third General dimension, compassion from others, was also posited by Gilbert (Gilbert et al., 2017) and validated with samples of US, UK, and Portugal students. We were interested in studying compassion from the lenses of the professional delivering the medical services, so in this study, we omitted the Compassion from others’ dimensions. In addition, self-compassion and compassion for others have the same focus (for self, for others) as CS and CF, to which prior evidence suggest BO is related. Conversely, compassion from others has a different compassion focus than the compassion dimensions to which physician BO is associated with in the literature (Bride et al., 2007).

In their validation of the Compassion Scales with students participating in psychology programs in the US and other students from non-specified programs in the UK and Portugal, Gilbert et al. (2017) found that self-compassion was negatively with anxiety, depression and stress, and positively related to wellbeing. The results of the same study suggest that compassion for others is positively related to wellbeing. Based on these findings, and given that BO is a term used to describe job stress in health practice environments (Rotenstein et al., 2018), we decided to test the relationship between self-compassion, compassion for others and BO. We posit that compassion is negatively related to D and EE and positively related to PA. Specifically, our hypotheses are:

H1a

There is a negative relationship between physicians’ self-compassion engagement and depersonalization.

H1b

There is a negative relationship between physicians’ self-compassion action and depersonalization.

H2a

There is a negative relationship between physicians’ self-compassion engagement and emotional exhaustion.

H2b

There is a negative relationship between physicians’ self-compassion action and emotional exhaustion.

H3a

There is a negative relationship between physicians’ compassion for others engagement and Emotional Exhaustion.

H3b

There is a negative relationship between physicians’ compassion for others action and Emotional Exhaustion.

H4a

There is a negative relationship between physicians’ compassion for others engagement and depersonalization.

H4b

There is a negative relationship between physicians’ compassion for others action and depersonalization.

H5a

There is a positive relationship between physicians’ self-compassion engagement and professional accomplishment.

H5b

There is a positive relationship between physicians’ self-compassion action and professional accomplishment.

H6a

There is a positive relationship between physicians’ compassion for others engagement and professional accomplishment.

H6b

There is a positive relationship between physicians’ compassion for others action and professional accomplishment.

Method

A study with a cross-sectional design was conducted. While having limitations, these designs are suitable when investigating a new variable in an old field, and there is uncertainty about the pattern of relationships and timeframe between the study variables (Spector, 2019), as in our case. Besides some studies inquiring about CF and BO, the knowledge about the relationship between physicians’ compassion and BO is still scarce. Furthermore, we employed a new operationalization of compassion, and the relationships of these new dimensions of compassion with BO are, to date, unexplored.

Participants and Procedures

Medicine students and general physicians from two Colombian cities were invited to participate. Inclusion criteria for medical students: being enrolled in the second or fifth year of the medical faculty of Universidad Pontificia Bolivariana (UPB, Medellín) or Universidad de Córdoba (Montería), voluntarily willing to participate, filling the informed signed consent. The recollection time was defined based on the students’ direct contact with patients (second year having none or fewer contact; fifth year having frequent contact). Inclusion criteria for general physicians: actively working at the University Clinic of UPB as a general practitioner, voluntarily willing to participate, filling the informed signed consent. All students and practicing physicians were approached during academic or clinical activities and were invited to participate. As a procedural remedy to common method variance, we protected the respondent’s anonymity and tried to reduce evaluation apprehension (Podsakoff et al., 2003). The respondents anonymously and voluntarily filled a paper and pencil version of the instruments, after signing the informed consent. Once the participants finished their responses, they handed the questionnaire in a sealed envelope. We also asked the respondents to be as honest as possible in their responses. Ethical approval was obtained from UPB. From a total of 359 participants, most were students (85.9%), and 14.2% were general physicians. Their mean age was 22 years (SD = 7), and 55.2% were female. On average, physicians had a professional experience of 10.34 years (SD = 8.67). Regarding hours per week dedicated to study/practice, 37.5% reported less than 48 hours, 37.2% between 48 and 60, and 25.4% more than 60 hours.

Measures

The Compassion Scales developed by Gilbert et al. (2017) were used. They assess three general factors (Self-compassion, Compassion for other, Compassion from others), each divided into two subscales (engagement and action). In this study, we used the self-compassion and compassion for others’ scales. We revised an adapted version to Spanish of the Self-compassion Scale and back translated all the items the Compassion Scale with permission from the authors. Before data collection, the authors of the original version reviewed all the back-translated sub-scales of compassion. The measure was validated with samples from the US, Britain, and Portugal, presenting a slightly different factorial structure than the theoretical factors (Gilbert et al., 2017). Gilbert and colleagues’ scales were recently used in a study conducted in Ecuador, but the study did not test the scales’ validity. The authors reported satisfactory reliability for both the global measure and subscales (Davalos-Batallas et al., 2020). Since to our knowledge no validation of the measure has been performed in Colombia, we followed the theoretical factor structure presented in 2017 (Gilbert et al., 2017). The participants tapped their responses to the 26 items using a ten-points Likert scale (1 = Never; 10 = Always).

To assess BO, we selected an adaptation of the MBI-HSS to the Colombian context. In their study with health professionals from one institution, Córdoba et al. (2011) reported good reliability for the EE dimension (alpha = .83). The other two, D and PA, showed low reliabilities (alphas of .57 and .52). Despite this, we decided to use the instrument given its wide use in healthcare (Rotenstein et al., 2018) and the current debate regarding Cronbach´s alpha (Dunn et al., 2014). More recently, the MBI-HSS was used in Colombia without validating it (Domínguez et al., 2019), which calls for validation studies in medicine students/professionals necessary. The MBI-HSS consists of 22 items using a seven-point Likert scale (0 = never, 6 = every day).

Analysis

Mplus statistical package was used to analyze the data (v.8) (Muthén & Muthén, 2017).

To test the relationship between physicians’ compassion, BO, and PA, we selected the Exploratory Structural Equation (ESEM) approach. With ESEM, we specified bidirectional relationships between BO and PA based on prior literature, our hypotheses, and our data. Specifying causal links between these variables would have implied ignoring the cross-sectional nature of our data.

Conversely to traditional exploratory factor analysis, ESEM allows validating new measures and testing relationships between latent variables adjusting for measurement error. Unlike Confirmatory Factor Analyses (CFAs), ESEM does not imply constraining items to load on a single factor, which may cause problems when evaluating discriminant validity (Marsh et al., 2014).

In view of some violations of the normality assumption, we selected the Maximum Likelihood with Robust standard errors (MLR) estimator. We used the CF-Varimax rotation option. We adopted the criteria suggested by Hu and Bentler (1999) for the CFI, RMSEA and SRMR coefficients to assess the goodness of fit of several ESEM models. Despite the high sensitivity to sample size, we also report the chi square coefficient (χ²). A non-significant chi square suggests good model fit. We adopted the criteria indicated by Hair et al. (2016, 2011) to evaluate the factor loadings and cross-loadings with samples sizes higher than 350 (i.e., >.30). Finally, to control for common method bias, we specified a model with a factor accounting for common method variance (Podsakoff et al., 2003).

Results

We first tested an ESEM with all compassion and BO sub-scales and items. According to two, out of the four goodness of fit coefficients, that model did not obtain satisfactory results (c² = 1366.05, df = 813, p = .00; CFI = .89). A closer examination of the compassion sub-scales indicated that the two engagement dimensions were not supported, as a reduced number of factor loadings reached the expected value. For instance, in the self-compassion engagement subscale, only item 4 (I am emotionally moved by my distressed feelings or situations) had a factor loading >.30. Prior studies have shown that the self-compassion engagement subscale has variations concerning the initial theoretical dimensionality of compassion. For example, Gilbert et al. (Gilbert et al., 2017) respecified the factor structure of that dimension. In fact, these authors found that the item 4 of the self-compassion engagement dimension loaded in a sub-factor named sensitivity to suffering (Gilbert et al., 2017). Since the lower order dimension of sensitivity to suffering of the self-compassion engagement factor was only reflected by items 2 and 4, and given that the factor loading of item 2 did not reach the cut-off value, we had no evidence to support that conceptualization of self-compassion engagement. Furthermore, most of the items of both compassion engagement scales had low factor loadings. Hence, we left both subscales out of the subsequent analyses. Following author recommendation (Gilbert et al., 2017), we did not include the reversed items in our analyses.

Concerning the MBI-HSS, the factor loadings of some items (items 4, 5, 6, and 20) also resulted below the critical value. In a previous validation of the MBI-HSS, the authors deleted some items due to the same cause (Córdoba et al., 2011). Residual correlations led to the deletion of other MBI-HSS items (e.g., items 13 and 14 had a residual correlation of .90). The deletion of these items did not imply a change in the dimensionality of BO as operationalized by the MBI-HSS.

After the deletion of the MBI-HSS items and the compassion engagement subscales we specified a five-factor model M1, namely, two compassionate action factors (self-compassion action and compassion to others action; respectively SCA and COA) and three BO dimensions (D, PA, and EE; D, PA, and EE, respectively). Scholars suggest avoiding the use of global measures of BO instead of using the subscales (Rotenstein et al., 2018). However, and aiming to have an alternative model to help validate the M1 model, we compared it with a model including a general compassionate action dimension and a global Maslach factor, henceforth M2.

The M1 obtained satisfactory fit in the four goodness of fit coefficients (Table 1), except for the χ². Although non-significant, the p value associated to the χ² did not reach the .05 level. Therefore, we inspected the residual correlations to assess the local fit of the model. Positive residual correlations higher than .30 suggest response dependency (Heffernan et al., 2020). The highest residual correlation between items 1 and 2 of the SCA scale was .27. In the absence of evidence of local misfit, and given the poor fit showed by the M2 model (see Table 1), we accepted M1.

Table 1.

Goodness of Fit ESEM Models.

Model	χ²	df	p value	CFI	RMSEA	SRMR
M1	119.49	100	.08	.99	.02	.02
M1covariates	144.46	115	.03	.99	.03	.02
M2	782.40	151	.00	.70	.11	.09

After validating the M1 model, we calculated the alphas for the resulting subscales. The alphas of the compassion subscales and the EE dimension of BO resulted satisfactory (SCA alpha = .86; COA alpha = .86; EE alpha = .80). The alpha of the PA subscale was acceptable (alpha = .64). In line with the study of Cordoba et al. (2011), the D factor obtained a low internal consistency coefficient (alpha = .48). According to the meta-analysis of the alpha for the MBI conducted by Wheeler et al. (2011) the alphas of the PA and D dimensions are below the recommended levels for high-stakes decision making (e.g.BO diagnosis). The authors of this meta-analysis argue that one of the reasons for the low reliability of the PA and Depersonalization subscales is the use of non-English versions of the BO inventory. Referring to studies in Spain, Gil-Monte (2005) indicates that alphas for the D are sometimes low, falling in the range between .42 and .64. As mentioned above, there is a debate among researchers regarding the use of the Cronbach´s alpha coefficient (Dunn et al., 2014). In view of these prior research findings, and the continued use of the MBI-HSS in healthcare research (Rotenstein et al., 2018), we decided to continue with our analyses.

Subsequently, we proceeded to include demographic and work-related variables into de model. Prior research has found that variables such as age, sex and workload are related to physician BO (Azam et al., 2017). We also found that prior studies report gender differences in some compassion dimensions (Gilbert et al., 2017), a relationship between years as medical service provider and compassion satisfaction (Gribben et al., 2019) and associations between self-compassion and age (Hwang et al., 2016; Neff & Vonk, 2009; Potter et al., 2014). In ESEM, as a default, the covariates should be specified as relating to all the focal constructs. The model with covariates (M1 covariates) also obtained satisfactory goodness of fit coefficients (Table 1). Conversely, the model with a common method factor did not converge. To remedy the no convergence, we increased the number of iterations, but the χ² could not be computed. Looking for the causes for no convergence and no identification of the model with the additional method factor, we inspected the correlations among the items and did not find excessively higher correlations (<.78). We concluded that the complexity added to the model versus the size of our sample generated the problems when validating the model with the additional method factor. Since we also used procedural remedies for common method bias, we continued inspecting the M1 model.

The specific results of the M1 with covariates model are displayed in Figure 1. The results indicate that COA and the SCA sub-scales are positively related. Furthermore, the M1 model shows that COA dimension has a negative relationship with D, while SCA is negatively associated with EE. Moreover, COA and SCA resulted positively related to positive aspects measured by the MBI-HSS, namely, PA. In addition, EE resulted negatively related with PA. These latter results support the theoretical relationships between the MBI-HSS dimensions.

Figure 1.

M1 model with covariates. Note. N = 359; MLR estimator; CF-Varimax rotation; standardized paths; ** p < .01 and *p < .05. Variances of the factors fixed to 1 as default in the ESEM approach, hence double headed arrows are equal to the correlations between the subscales.

In summary, after simultaneously examining some of the psychometric properties of the compassion scale and MBI-HSS, and validating the relationship between the constructs, our findings partially support the hypothesis of a negative relationship between compassion and BO in physicians. Specifically, the hypotheses including engagement sub-scales (i.e. H1a, H2a H3a, H4a, H5a, and H6a) were not validated given that the corresponding subscales did not show good psychometric properties. Regarding the relationships between COA and SCA with the BO dimensions, we find support for a negative relationship between physicians’ COA and D (H4b) and between physicians’ SCA and EE (H2b) and a positive relationship between physicians’ SCA and PA (H5b), and COA and PA (H6b). Conversely, our results do not support a negative relationship between physicians’ COA and EE (H3b).

According to Gilbert et al. (Gilbert et al., 2017), the sum of all the items of the self-compassion subscale gives an aggregate score for that dimension. However, from our analysis only the SCA subscale was retained. Thus, we calculated the SCA score by summing the retained items. Following this procedure, we calculated also the COA score and found that the means of both subscales (Msca = 27.81, SD = 5.27; Mcoa = 28.62, SD = 5.34) were similar to those presented in Gilbert et al. (Gilbert et al., 2017) original validation. We calculated the composite scores of the three MBI-HSS factors in the same way and using only the retained items. We report the means of the three score, but following Rotenstein et al. (2018), we warn the readers against taking these values as a reference (Mee = 15.59, SD = 5.34; Md = 3.94, SD = 3.72; Mpa = 23.02, SD = 5.12).

Finally, as shown in Figure 1, age resulted positively related to SCA and PA, and negatively associated with EE. Furthermore, we found that hours per week dedicated to study/practice was negatively correlated with EE. We did not find a significant relationship between sex, years in professional activity, and compassion or BO.

Discussion

Our study aims at contributing to the literature on compassion and BO, particularly from the viewpoint of medical students and professionals. Previous studies about this relationship are mainly focused on the link between CF and BO. We formulate our hypotheses based on a relatively new conceptualization of compassion (Gilbert, 2009, 2017; Gilbert & Choden, 2013). In this study, we simultaneously validate a new compassion scale (Gilbert et al., 2017) and test the relationships between compassion for others, self-compassion, EE, D, and PA.

According to our results, the factor structure of both the compassion and BO measures changed concerning the theoretical factor structure. This is not surprising since in the original validation using a US sample, the factor structure of the self-compassion engagement diverged from the expected one (Gilbert et al., 2017). Both action subscales of the Self-compassion and Compassion for others’ Scales were validated and these factors are positively correlated. Conversely, we could not validate the engagement subscales of neither the scales. Although both action and engagement subscales showed good reliability in the original validation, the actions subscales obtained higher reliability coefficients. In fact, in their factor analyses, Gilbert et al. (2017) found that the items of self-compassion engagement loaded on two factors. The authors also found that one of those factors was represented by only two items. From the psychometric point of view, this finding can be problematic since a two-item factor is usually considered weak and unstable (Costello & Osborne, 2019). Therefore, we believe that our results indicate the need to further examine the factor structure of the compassion engagement sub-scales.

Regarding our hypotheses, we found that COA and EE, a dimension of BO, were not related. This finding was, to some extent, expected since a previous study did not find any relationship between Compassion for others and stress (Gilbert et al., 2017). Now, our results indicate that the same dimension of compassion (COA) is significantly related with another factor of BO, namely, D. In this sense, we suggest studying the relationships between compassion for others and stress, specifically BO, at the subscale level (actions and engagement separately). Our findings also indicate that the same reasoning should be applied when studying the relationship between self-compassion and stress or BO. For instance, Gilbert et al. (2017) found that self-compassion as a general dimension was negatively and significantly correlated with stress, while we found that SCA was negative and significantly associated to EE but not to D.

Our findings also suggest that age is associated with some SCA, PA and EE, while workload is related to EE. The literature suggests that in the general population self-compassion are positively related with emotional distress (Neff & Vonk, 2009; Potter et al., 2014). The results of our study sum up to this evidence by showing that this relationship also exist in the medical community, and that at the same time age is also positively related to PA, and negatively related to EE. This evidence can be useful for other researchers interested in inquiring about more complex relationships between BO, self-compassion and age in the medical community. For instance, Hwang and colleagues (2016) studied the interaction between age and self-compassion. However, they collected data from the general population and, since their data was cross-sectional, they suggest reading these results with care.

In sum, our results indicate that mobilizing behaviors towards compassion and self-compassion inversely relates to different components of burnout and could constitute a protective factor against the stress related to healthcare activities. Other authors have found similar results (Hashem & Zeinoun, 2020; Kemper et al., 2020; Olson et al., 2015; Sinclair et al., 2021). For instance, Kemper et al. (2020) found that self-compassion and confidence in providing compassionate care were associated with a lower risk of burnout; while Hashem and Zeinoun (2020) reported an association between self-compassion and EE, which is particularly consistent with our results

Education programs for physicians should include tools allowing professionals to cope with stress. For example, mindfulness training has shown moderate effects on stress (Yates, 2020) while training in compassionate responses increases negative affect and increases positive affect, denoting a potential strategy to prevent burnout (Klimecki et al., 2014). Particularly, considering the findings of our study, compassion training programs for medical students and physicians might constitute an alternative to avoid BO in this population. Such programs should seek to boost self-compassion and compassion for others as they seem to be proper mechanisms against physicians’ depersonalization and emotional exhaustion while enhancing their professional accomplishment.

Limitations and Future Research

Although we justified using the cross-sectional design in our study, future studies could adopt different designs to isolate the direction of the relationship between compassion and BO. For example, experimental designs using random samples and control groups to test the effect of compassion training on medical students and physicians in challenging clinical practices. Such studies would also overcome another limitation of our research: the use of convenience sampling, which hinders the external validity of our study.

Future studies inquiring about the relationship between BO and compassion should employ additional remedies to common method biases. For instance, although using different information sources is not always feasible given the nature of the variables (Podsakoff et al., 2003), some of our variables could have been measured using external raters. Even though self-compassion and EE can be better inquired using self-reported measures, in further studies information about facets of compassion for others could be gathered from colleagues and/or patients.

Most of our study participants were students. Reaching physicians was a difficult task given their time constraints. In this sense, future studies should go deeper on examining response rates and missing values of studies involving BO to check its relationship with missingness. Such studies may reveal an essential aspect of attrition in physicians’ compassion and BO studies. We did not compare the psychometric properties of the measures, prevalence, and relationships between the variables under study across groups of students and physicians because we had highly unbalanced sub-sample sizes. Future studies with more balanced samples and bigger sample sizes should verify if the measures we used in this study are invariant across physicians and students. Multigroup analyses would be helpful to test invariance across this groups and provide a more concrete validation of our hypotheses in these two populations.

Finally, we want to add a cautionary note regarding the operationalization of BO. The internal consistency of some scales of non-English versions of the MBI-HSS regularly falls below the recommended rules of thumb. As in the case of previous research results involving the use of the non-English version of the MBI-HSS, our results may be affected by low internal consistency of the depersonalization sub-scale. We recommend that future studies willing to measure BO using the MBI-HSS should review the content validity of the scale to check if improvements in language and cultural adaptations can be conducted.

Footnotes

Acknowledgments

We thank the participating institutions in Medellín and Montería for their contributions and L Garavito, LF Álvarez, C Duque, and J Acevedo for their invaluable help in data gathering.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Universidad Pontificia Bolivariana seccional Monteria 245M-07/19-G003.

ORCID iD

Alicia Krikorian

Author Biographies

Juan P. Román-Calderón PhD, Area of Organization, Direction and Strategy, Universidad Eafit, Medellin, Colombia.

Alicia Krikorian PhD, Pain and Palliative Care Group, School of Health Sciences, Universidad Pontificia Bolivariana, Medellin, Colombia.

Erika Ruiz, Faculty of Psychology, Universidad Pontificia Bolivariana, Monteria, Colombia.

Ana M. Romero, Faculty of Psychology, Universidad Pontificia Bolivariana, Monteria, Colombia.

Mariantonia Lemos, Psychology Department, Universidad Eafit, Medellin, Colombia.

References

Azam

Khan

Alam

M. T.

(2017). Causes and adverse impact of physician burnout: A systematic review. Journal of the College of Physicians and Surgeons--Pakistan: JCPSP, 27(8), 495–501.

Bride

B. E.

Radey

Figley

C. R.

(2007). Measuring compassion fatigue. Clinical Social Work Journal, 35, 155–163. https://doi.org/10.1007/s10615-007-0091-7.

Carmel

Glick

S. M.

(1996). Compassionate-empathic physicians: Personality traits and social-organizational factors that enhance or inhibit this behavior pattern. Social Science & Medicine, 43(8), 1253–1261. https://doi.org/10.1016/0277-9536(95)00445-9

Cavanagh

Cockett

Heinrich

Doig

Fiest

Guichon

J. R.

Page

Mitchell

Doig

C. J.

(2019). Compassion fatigue in healthcare providers: A systematic review and meta-analysis. Nursing Ethics, 27(3), 639–665. https://doi.org/10.1177/0969733019889400

Córdoba

Tamayo

J. A.

González

M. A.

Martínez

M. I.

Rosales

Barbato

S. H.

(2011). Adaptation and validation of the maslach burnout inventory-human services survey in Cali, Colombia. Colombia Medica. https://doi.org/10.25100/cm.v42i3.874.

Costello

Osborne

(2019). Best practices in exploratory factor analysis: Four recommendations for getting the most from your analysis. Practical Assessment, Research, and Evaluation, 10(1). https://doi.org/10.7275/jyj1-4868

Crawford

Brown

Kvangarsnes

Gilbert

(2014). The design of compassionate care. Journal of Clinical Nursing, 23(23–24), 3589–3599. https://doi.org/10.1111/jocn.12632

Davalos-Batallas

Vargas-Martínez

A.-M.

Bonilla-Sierra

Leon-Larios

Lomas-Campos

M.-M.

Vaca-Gallegos

S.-L.

de Diego-Cordero

(2020). Compassionate engagement and action in the education for health care professions: A cross-sectional study at an Ecuadorian university. International Journal of Environmental Research and Public Health, 17(15). https://doi.org/10.3390/ijerph17155425

Domínguez

L. C.

Sanabria

Á.

Ramírez

Vargas

Pacheco

JiménezGermán . (2019). Desgaste profesional en residentes colombianos de cirugía: Resultados de un estudio nacional. Revista Colombiana de Cirugía, 32(2), 121–127. https://doi.org/10.30944/20117582.16

10.

Dunn

T. J.

Baguley

Brunsden

(2014). From alpha to omega: A practical solution to the pervasive problem of internal consistency estimation. British Journal of Psychology (London, England : 1953), 105(3), 399–412. https://doi.org/10.1111/bjop.12046

11.

Dutton

J. E.

Workman

K. M.

Hardin

A. E.

(2014). Compassion at work. Annual Review of Organizational Psychology and Organizational Behavior, 1(1), 277–304. https://doi.org/10.1146/annurev-orgpsych-031413-091221

12.

Fernando

A. T.

Consedine

N. S.

(2014). Beyond compassion fatigue: The transactional model of physician compassion. Journal of Pain and Symptom Management, 48(2), 289–298. https://doi.org/10.1016/j.jpainsymman.2013.09.014

13.

Furr

R. M.

(2011). Scale construction and psychometrics for social and personality. Sage.

14.

Gilbert

(2009). The compassionate mind:A new approach to life’s challenges. New Harbinger Publications.

15.

Gilbert

(2017). Compassion: Concepts, research and applications. Routledge. https://derby.openrepository.com/handle/10545/622328

16.

Gilbert

(2019). Explorations into the nature and function of compassion. Current Opinion in Psychology, 28, 108–114. https://doi.org/10.1016/j.copsyc.2018.12.002.

17.

Gilbert

(2020). Compassion: From its evolution to a psychotherapy. Frontiers in Psychology, 11, 586161. https://doi.org/10.3389/fpsyg.2020.586161.

18.

Gilbert

Catarino

Duarte

Matos

Kolts

Stubbs

Ceresatto

Duarte

Pinto-Gouveia

Basran

(2017). The development of compassionate engagement and action scales for self and others. Journal of Compassionate Health Care, 4(1), 4. https://doi.org/10.1186/s40639-017-0033-3

19.

Gilbert

Choden (2013). Mindful compassion. Hachette UK.

20.

Gil-Monte

P. R.

(2005). Factorial validity of the maslach burnout inventory (MBI-HSS) among Spanish professionals. Revista De Saude Publica, 39(1), 1–8. https://doi.org/10.1590/s0034-89102005000100001

21.

Gribben

J. L.

Kase

S. M.

Waldman

E. D.

Weintraub

A. S.

(2019a). A cross-sectional analysis of compassion fatigue, burnout, and compassion satisfaction in pediatric critical care physicians in the United States. Pediatric Critical Care Medicine: A Journal of the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies, 20(3), 213–222. https://doi.org/10.1097/PCC.0000000000001803

22.

Gribben

J. L.

MacLean

S. A.

Pour

Waldman

E. D.

Weintraub

A. S.

(2019b). A cross-sectional analysis of compassion fatigue, burnout, and compassion satisfaction in pediatric emergency medicine physicians in the United States. Academic Emergency Medicine, 26(7), 732–743. https://doi.org/10.1111/acem.13670

23.

Hair

Hult

G. T. M.

Ringle

C. M.

Sarstedt

(2016). A primer on partial least squares structural equation modeling (2nd ed.). SAGE Publications, Inc.

24.

Hair

J. F.

Ringle

C. M.

Sarstedt

(2011). PLS-SEM: Indeed a silver bullet. Journal of Marketing Theory and Practice, 19(2), 139–152. https://doi.org/10.2753/MTP1069-6679190202

25.

Hashem

Zeinoun

(2020). Self-compassion explains less burnout among healthcare professionals. Mindfulness, 11, 2542–2551. https://doi.org/10.1007/s12671-020-01469-5.

26.

Heffernan

Weinstein

B. E.

Ferguson

M. A.

(2020). Application of rasch analysis to the evaluation of the measurement properties of the hearing handicap inventory for the elderly. Ear and Hearing, 41(5), 1125–1134. https://doi.org/10.1097/AUD.0000000000000832

27.

L. T.

Bentler

P. M.

(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6(1), 1–55. https://doi.org/10.1080/10705519909540118

28.

Hwang

Kim

Yang

J.-W.

Yang

(2016). The moderating effects of age on the relationships of self-compassion, self-esteem, and mental health. Japanese Psychological Research, 58(2), 194–205. https://doi.org/10.1111/jpr.12109

29.

Jazaieri

Jinpa

G. T.

McGonigal

Rosenberg

E. L.

Finkelstein

Simon-Thomas

Cullen

Doty

J. R.

Gross

J. J.

Goldin

P. R.

(2013). Enhancing compassion: A randomized controlled trial of a compassion cultivation training program. Journal of Happiness Studies, 14(4), 1113–1126. https://doi.org/10.1007/s10902-012-9373-z

30.

Kalra

Priya

Grewal

Aye

T. T.

Waraich

B. K.

SweLatt

Khun

Phanvarine

Sutta

Kaush

Manilka

R. S.

Kalra

(2018). Lessons for the health-care practitioner from buddhism. Indian Journal of Endocrinology and Metabolism, 22(6), 812–817. https://doi.org/10.4103/ijem.IJEM_286_17

31.

Kemper

K. J.

McClafferty

Wilson

P. M.

Serwint

J. R.

Batra

Mahan

J. D.

Schubert

C. J.

Staples

B. B.

Schwartz

Pediatric Resident Burnout-Resilience Study Consortium . (2019). Do mindfulness and self-compassion predict burnout in pediatric residents? Academic Medicine: Journal of the Association of American Medical Colleges, 94(6), 876–884. https://doi.org/10.1097/ACM.0000000000002546

32.

Kemper

K. J.

Schwartz

Wilson

P. M.

Mahan

J. D.

Schubert

C. J.

Staples

B. B.

McClafferty

Serwint

J. R.

Batra

Pediatric Resident Burnout-Resilience Study Consortium . (2020). Burnout in pediatric residents: Three years of national survey data. Pediatrics, 145(1), Article e20191030. https://doi.org/10.1542/peds.2019-1030

33.

Klimecki

O. M.

Leiberg

Ricard

Singer

(2014). Differential pattern of functional brain plasticity after compassion and empathy training. Social Cognitive and Affective Neuroscience, 9(6), 873–879. https://doi.org/10.1093/scan/nst060

34.

Marsh

H. W.

Morin

A. J. S.

Parker

P. D.

Kaur

(2014). Exploratory structural equation modeling: An integration of the best features of exploratory and confirmatory factor analysis. Annual Review of Clinical Psychology, 10(1), 85–110. https://doi.org/10.1146/annurev-clinpsy-032813-153700

35.

Mills

Wand

Fraser

J. A.

(2018). Examining self-care, self-compassion and compassion for others: A cross-sectional survey of palliative care nurses and doctors. International Journal of Palliative Nursing, 24(1), 4–11. https://doi.org/10.12968/ijpn.2018.24.1.4

36.

Montgomery

(2014). The inevitability of physician burnout: Implications for interventions. Burnout Research, 1(1), 50–56. https://doi.org/10.1016/j.burn.2014.04.002

37.

Muthén

(2017). Mplus user’s guide (8th ed.). Author.

38.

Neff

K. D.

Vonk

(2009). Self-compassion versus global self-esteem: Two different ways of relating to oneself. Journal of Personality, 77(1), 23–50. https://doi.org/10.1111/j.1467-6494.2008.00537.x

39.

Olson

Kemper

K. J.

Mahan

J. D.

(2015). What factors promote resilience and protect against burnout in first-year pediatric and medicine-pediatric residents? Journal of Evidence-Based Complementary & Alternative Medicine, 20(3), 192–198. https://doi.org/10.1177/2156587214568894

40.

Podsakoff

P. M.

MacKenzie

S. B.

Lee

J.-Y.

Podsakoff

N. P.

(2003). Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of Applied Psychology, 88(5), 879–903. https://doi.org/10.1037/0021-9010.88.5.879

41.

Potter

R. F.

Yar

Francis

A. J.

Schuste

(2014). Self-compassion mediates the relationship between parentalcriticism and social anxiety | IJPSY. International Journal of Psychology and Psychological Therapy, 14(1), 33–43.

42.

Rotenstein

L. S.

Torre

Ramos

M. A.

Rosales

R. C.

Guille

Sen

Mata

D. A.

(2018). Prevalence of burnout among physicians: A systematic review. JAMA, 320(11), 1131–1150. https://doi.org/10.1001/jama.2018.12777

43.

Sinclair

Kondejewski

Jaggi

Roze des Ordons

A. L.

Kassam

Hayden

K. A.

Harris

Hack

T. F.

(2021). What works for whom in compassion training programs offered to practicing healthcare providers: A realist review. BMC Medical Education, 21(1), 455. https://doi.org/10.1186/s12909-021-02863-w

44.

Sirois

F. M.

Kitner

Hirsch

J. K.

(2015). Self-compassion, affect, and health-promoting behaviors. Health Psychology: Official Journal of the Division of Health Psychology, American Psychological Association, 34(6), 661–669. https://doi.org/10.1037/hea0000158

45.

Spector

P. E.

(2019). Do not cross me: Optimizing the use of cross-sectional designs. Journal of Business and Psychology, 34(2), 125–137. https://doi.org/10.1007/s10869-018-09613-8

46.

Stamm

B. H.

(2002). Measuring compassion satisfaction as well as fatigue: Developmental history of the compassion satisfaction and fatigue test. In Figley

C.R.

(Ed.), Treating compassion Fatigue (pp. 107–119). Routledge.

47.

Weintraub

A. S.

Geithner

E. M.

Stroustrup

Waldman

E. D.

(2016). Compassion fatigue, burnout and compassion satisfaction in neonatologists in the US. Journal of Perinatology, 36(11), 1021–1026. https://doi.org/10.1038/jp.2016.121

48.

West

C. P.

Dyrbye

L. N.

Shanafelt

T. D.

(2018). Physician burnout: Contributors, consequences and solutions. Journal of Internal Medicine, 283(6), 516–529. https://doi.org/10.1111/joim.12752

49.

Wheeler

D. L.

Vassar

Worley

J. A.

Barnes

L. L. B.

(2011). A reliability generalization meta-analysis of coefficient alpha for the maslach burnout inventory. Educational and Psychological Measurement, 71(1), 231–244. https://doi.org/10.1177/0013164410391579

50.

WHO . (2019). Burn-out an “occupational phenomenon”: International classification of diseases: WHO.

51.

Yates

S. W.

(2020). Physician stress and burnout. The American Journal of Medicine, 133(2), 160–164. https://doi.org/10.1016/j.amjmed.2019.08.034