Compassion has long been esteemed as an essential virtue in the major contemplative traditions (Dahlsgaard et al., 2005). Compassion is a fundamental human quality that generates altruism and generosity, as it motivates people to help alleviate suffering so resulting in a greater connection with others (Lama, 2002). It has been suggested that this sense of connection that compassion generates is a protective factor of mental health, due to its positive relationship with well-being and its negative relationship with depressive symptoms and stress (Gilbert et al., 2017). Consequently, compassion has become an area of special interest in many different fields, such as psychology, nursing, medicine, and education in recent years (Seppälä et al., 2017).

Compassion has been defined in many ways. Generally, it is understood as the awareness of another’s suffering and the desire to alleviate it (Chochinov, 2007; Goetz et al., 2010; Klimecki & Singer, 2013; Sprecher & Fehr, 2005). A review of the literature showed five principal dimensions of compassion: the recognition of another’s suffering, the understanding of another’s suffering, the feeling of empathy and concern for the person suffering, the tolerance of the distress caused by seeing the person suffer, and the motivation to reduce the suffering of the other (Gu et al., 2020; Jazaieri et al., 2013; Strauss et al., 2016).

Researchers’ efforts to understand compassion in clinical contexts have yielded significant insights, such as higher job satisfaction, mental health protection, and patient well-being (Galiana & Sansó, 2019; Klimecki & Singer, 2013; Matos et al., 2022; Seppälä et al., 2017); however, challenges remain from an educational perspective, where there is a lack of knowledge about how students practice compassion. Being compassionate to others is considered essential in all settings, yet is especially true in the healthcare context, and even more so within nursing (Bickford et al., 2019; Sinclair et al., 2018). The nursing profession is defined by its close contact with people’s suffering, stemming from health problems and negative emotional experiences. Understanding the suffering of others and the attempt to alleviate it is essential within this profession, where it is necessary to cater to patient needs and to enhance patient care. Ultimately, it can be argued that compassion is needed to improve the overall quality of healthcare delivery (Kelley et al., 2014). This is reflected in the six Cs of caring, where compassion is identified as one of its key pillars—compassion, competence, confidence, conscience, commitment, and comportment (Roach, 2002)—thus emphasizing its importance in nurse education.

Although compassion is essential for nurses and, therefore, should be addressed during their training, research studying compassion among nursing students, along with its determinants and consequences, is scarce. For example, some studies have focused on the role of optimism, hope, or resilience (Jarden et al., 2021), but have been descriptive in nature, without delving into their role as sources of compassionate healing. Although the literature on nursing students has expanded in recent years, it continues to be extremely poor in terms of addressing compassionate care. Increasing our understanding of compassion as a resource is key, especially in light of recent studies that indicate higher levels of burnout and work-induced stress in the nursing profession, a factor which has been shown to increase compassion fatigue and erode the capacity for compassion (Galiana & Sansó, 2019; Jinpa, 2015; Trzeciak & Mazzarelli, 2020). The little available evidence also suggested that nursing students’ caring behavior decreased as they progressed through their studies (Curtis, 2014).

With consideration of the highly stressful challenges faced by the nursing profession today and the documented erosion of the capacity to deliver care with compassion as a result of it, fostering compassion is essential. Especially among students, who are to be the future of the profession. It is of note then that there remains limited evidence of compassion in the vision, mission, and value statements of many nursing institutions. In a review carried out by Younas and Maddigan (2019), no evidence of the importance of compassion in nursing schools’ guiding statements was found. A closer look at our context, which included nursing education curricula from the University of Valencia and the University of the Balearic Islands, leads us to believe that this is also the case for the Spanish curricula. The lack of attention regarding compassion and compassionate resources by nursing schools is concerning, as compassionate care is needed in every nurse-patient interaction, as seen by repeated calls to action to foster compassion in students. Nursing education is formative in the socialization of nurses’ attitudes and values, particularly that of compassion. Nursing education is therefore an opportunity to improve the current situation, where compassion is not regarded as the powerful resource it could be.

One of the ways to improve compassion among nursing students is to introduce mindfulness-based interventions as part of their educational program. Mindfulness has been repeatedly shown to increase feelings of compassion among those who practice it, so could benefit students who are about to enter potentially stressful situations, not only with their own compassionate self-care but also for their patients (Conversano et al., 2020; Dariotis et al., 2023). Mindfulness has been defined as “paying attention in a particular way: on purpose, in the present moment, and nonjudgmentally” (Kabat-Zinn, 1994, p. 4). Mindfulness can also occur during interpersonal interactions and has been termed “interpersonal mindfulness” (Pratscher et al., 2019). Being present with others, here and now, is a prerequisite for recognizing pain and distress and therefore being capable of compassion.

A recent systematic review noted that compassionate interventions increase levels of mindfulness and reduce interpersonal conflict (Conversano et al., 2020). In turn, compassion is also strengthened through mindfulness exercises (Huppert, 2018). Both mindfulness and compassion have many benefits, including reducing negative emotions (Don et al., 2022; Han & Kim, 2023), increasing positive emotions (Tran et al., 2022), and helping to build a more enriching family environment (Dariotis et al., 2023), as well as reducing caregiver burden (Juberg et al., 2023).

Several mindfulness-based interventions have been proven to improve well-being among nurses and nursing students. For example, the review carried out by van der Riet et al. (2018) showed that mindfulness meditation had a positive impact on stress, anxiety, depression, burnout, sense of well-being, and empathy. However, there has been less empirical research on the relationship between mindfulness and compassion among nurses and nursing students. Erkin and Aykar (2021) found an increase in nursing students’ self-compassion after a yoga course. More recently, Baminiwatta et al. (2023) found a positive correlation between trait mindfulness and compassion in nurses, with the latter partially mediating the effects of trait mindfulness on helping and avoidance. Other studies, however, have failed to find an effect of meditative-based interventions on nursing students’ compassion levels (Joseph & Raque, 2023). Despite this last study, which had several limitations (brief intervention, post-intervention measures only), it has been widely demonstrated that compassion has numerous benefits (Matos et al., 2022). This is why in recent years there has been an increasing interest in the development of scales that can provide a reliable and valid measurement of this construct.

There are various scales that measure the construct of compassion for others: the Compassionate Love Scale (CLS; Sprecher & Fehr, 2005), a 21-item self-reporting scale that measures compassionate love for others, both strangers and close contacts, across three areas (kindness and affection, acceptance and understanding, and help and sacrifice); the Relational Compassion Scale (RCS; Hacker, 2008), a 16-item scale that measures compassion for self and others; the Santa Clara Brief Compassion Scale (SCBCS; Hwang et al., 2008), a 5-item self-reporting scale that measures compassionate love felt toward strangers; the Compassion Scale (CS; Pommier et al., 2020), a 16-item scale made up of four subscales that measures: compassion for others, common humanity, mindfulness, least indifference, and most kindness; the Compassion Scale (CS-M; Martins et al., 2013), a 10-item self-reporting scale that measures five domains of compassion: generosity, hospitality, sensitivity, tolerance, and objectivity; and finally, the Compassionate Engagement and Action Scale (CEAS; Gilbert et al., 2017), a 10-item self-reporting scale that assesses engagement, compassion for others and compassion for self.

Scales assessing compassion for others within the healthcare context, as reported by healthcare providers, have also been recently developed and published. The Compassion Competence Scale (CCS; Lee & Seomun, 2016) consists of 18 items measuring three dimensions of compassion for nursing professionals (communication, insight, and sensitivity). The Bolton Compassion Strengths Indicators (BCSIs; Durkin et al., 2020) consists of 48 items that measure the strength of compassion in nursing students, including self-care, empathy, commitment, and competence.

Yet despite the development of compassion measures, there remains a lack of consensus on the definition of compassion and its distinguishing characteristics, which impedes the development of reliable measures. Strauss et al. (2016), in a review of definitions, demonstrated that compassion for others consisted of five elements: (1) recognizing suffering, (2) understanding suffering, (3) caring and empathizing with the suffering person, (4) tolerating the distress generated by another’s suffering, and (5) being motivated to alleviate the other’s suffering. The scales described above fail to capture these five dimensions, or fully measure compassion (Gu et al., 2020; Strauss et al., 2016). Furthermore, none of them assessed compassion in a methodologically rigorous manner, and few employed robust psychometric measures (Strauss et al., 2016).

To address these shortcomings, Gu et al. (2020) developed the Sussex-Oxford Compassion for Others Scale (SOCS–O), a 20-item self-administered scale that measures compassion, including the five dimensions discussed above. Items score in a 5-point Likert-type scale ranging from 1 (not at all true) to 5 (always true). Unlike previous self-report scales, this measure demonstrates robust psychometric properties including reliability, interpretability, convergent and discriminant validity, and internal consistency (Gu et al., 2020), factors which allow for its use in practice and research. In further studies, this five-factor with correlated factors structure has been replicated (i.e., de Krijger et al., 2022; Sarling et al., 2024). A specific example of this was Sarling et al. (2024) who also tested a unidimensional structure and a three-factor with correlated factors structure, in which items of Caring and empathizing with the suffering person; Tolerating the distress generated by another’s suffering; and Being motivated to alleviate the other’s suffering were explained by a Universal latent variable, which both resulted in a poorer fit.

Regarding reliability estimates, the ones reported by Sarling et al. (2024) ranged from 0.70 for Tolerating the distress generated by another’s suffering to 0.87 for Recognizing suffering. Similar values were found by de Krijger et al. (2022), with values ranging from 0.68 for Recognizing suffering to 0.88 for Being motivated to alleviate the other’s suffering.

Evidence of measurement invariance for gender has been gathered, with varying results. Whereas Kim and Seo (2021) found partial evidence of measurement invariance, Sarling et al. (2024) found evidence of metric invariance, but pointed out some concerns with scalar invariance. De Krijger et al. (2022), in turn, found evidence of scalar invariance across gender in a sample of crisis line volunteers and in a sample of soldiers. However, they did not test gender invariance for nursing students.

The SOCS–O has been validated in different countries and languages, namely Swedish (Sarling et al., 2024), Dutch (de Krijger et al., 2022), Korean (Kim & Seo, 2021), Persian (Nikgoo et al., 2022), Slovak (Halamová & Kanovský, 2021), and Italian (Lucarini et al., 2022). However, few studies have examined the psychometric properties of the SOCS–O scale in the Spanish language, the fourth most widely used language in the world (after English, Mandarin and Hindi), spoken by more than 500 million people worldwide.

It is important that future nurses have the competency for compassion in order to provide high-quality, compassionate patient care. The first step towards this is to have a robust and valid instrument that assesses compassion in nursing students. Therefore, the main objective of the present study is to examine, for the first time, the psychometric properties of the SOCS–O in a large sample of Spanish undergraduate nursing students. Our specific aims were (1) to provide evidence of nursing students’ levels of compassion; (2) to study the internal factor structure of the Spanish version of the SOCS–O; (3) to provide evidence of the reliability of the scale; (4) to test its measurement invariance across gender; and (5) to evaluate criterion validity of the SOCS–O by means of structural equation modelling in which students’ levels of mindfulness are related to their levels of compassion, while simultaneously controlling for gender and age.

Method

Participants

Raosoft® software was used to calculate the minimum sample size needed in order to adequately represent the nursing students’ population from the two participating universities. The population of the first year of nursing students was calculated to be n = 892, with a confidence interval of 95% (alpha = 5%) with an error limit of 5%, and p = q = 0.50; the number of elements of the sample to obtain was n = 269.

For this specific study on the psychometric properties of the SOCS–O, we also calculated the sample size required for a confirmatory factor analysis. The calculation was based on the original structure, given the number of observed (20) and latent variables in the model (5), the anticipated effect size (0.50), and the desired probability and statistical power levels (0.05 and 0.90, respectively). For this purpose, we used Daniel Soper’s Free Statistics Calculator version 4.0. The recommended minimum sample size was n = 100.

The final sample was composed of 683 nursing students. 83.5% (n = 570) were women. The mean age was 22.74 years old (SD = 7.65). 54.8% (n = 374) were University of Valencia students, and 45.2% (n = 309) were students of the University of the Balearic Islands. The majority of the students were not working at the time of the survey (70.4%, n = 481). 13.5% (n = 92) were working as healthcare professionals. For more details, see Table 1.

Table 1 Sample description
Full size table

Procedure

The study had a cross-sectional design. The first-year nursing students at the University of Valencia and the University of the Balearic Islands (Spain) were encouraged to participate. They were identified via the universities’ list of registered students. In order to be included, participants had to be nursing students in the first year of their degree. The students completed the questionnaire online in approximately 20 min. Data were gathered in May 2022 and May 2023.

Measures

The present study included, together with sample description characteristics, two main measurement instruments: the Spanish version of the Sussex-Oxford Compassion for Others Scale (SOCS–O; Gu et al., 2020) and the Freiburg Mindfulness Inventory (FMI; Walach et al., 2006) (Spanish version; Pérez-Verduzco & Laca-Arocena, 2017).

For the translation of the SOCS–O, we used the backward and forward translation process. First, the scale was translated into Spanish by a professional native speaker; it was then translated back into English by another native professional speaker. The final version was revised by three experts in psychometrics and mindfulness. In the final version, no differences were found. The resulting Spanish version of the scale can be consulted in Table 2. Reliability estimates in this sample for the original structure of the scale were as follows: α = 0.87 and ω = 0.87 for Recognizing suffering; α = 0.88 and ω = 0.88 for Understanding the universality of suffering; α = 0.82 and ω = 0.82 for Feeling for the person suffering; α = 0.74 and ω = 0.88 for Tolerating uncomfortable feelings; and α = 0.88 and ω = 0.79 for Acting to alleviate suffering.

Table 2 Items’ content and descriptive statistics for the Spanish version of the SOCS–O
Full size table

The FMI is a 14-item scale, assessing a general factor of mindfulness. Responses were made on a Likert-type scale, and ranged from 1 (almost never) to 4 (almost always). Reliability estimates in this sample were α = 0.84 and ω = 0.88.

Data Analyses

Firstly, descriptive statistics for the items and the total scores of the Spanish version of the SOCS–O were calculated. These included mean, standard deviation, and minimum and maximum scores.

Secondly, for the study of the internal structure, we used a sequence of models, including three confirmatory factor analyses that were hypothesized, estimated, and tested. A first five-factor with correlated factors model was hypothesized, in which five factors—Recognizing suffering, Understanding the universality of suffering, Feeling for the person suffering, Tolerating uncomfortable feelings, and Acting to alleviate suffering—explained the 20 items of the scale. The structure followed both the original work (Gu et al., 2020) as well as later translations (de Krijger et al., 2022; Sarling et al., 2024). Based on the high intercorrelations observed among certain subscales, two simpler structures were tested. The first one hypothesized a general factor of compassion for others that explained the 20 items of the scale. The last one hypothesized three-correlated factors of compassion for others: Recognizing suffering, Understanding the universality of suffering, and Universal or Behavioral compassion. The first two factors were based on the original structure of the scale, Recognizing suffering and Understanding the universality of suffering, whereas the third factor explained Items 3, 4, 5, 8, 9, 10, 13, 14, 15, 18, 19, and 20, corresponding to the original factors Feeling for the person suffering, Tolerating uncomfortable feelings, and Acting to alleviate suffering. This latest factor was renamed Behavioral compassion, following the classification used by Halamova and Kanovsky (2021).

To assess the model’s overall fit, several criteria were used: the chi-square statistic, the Comparative Fit Index (CFI), the Standardized Root Mean Square Residual (SRMR), and the Root Mean Square Error of Approximation (RMSEA). CFI and TLI values above 0.90 (better over 0.95) and SRMR and RMSEA values below 0.08 (better under 0.06) were indicative of a good fit (Hu & Bentler, 1999; Perry et al., 2015). Particular relationships within the model were also examined, including factor loadings and correlations between factors. In addition, following the recommendations of Kline (2023), modification indices were checked to ensure the absence of cross-loadings and unmodeled relationships, as well as a visual inspection of the residuals.

Model comparison was done using delta indices. Specifically, differences between models’ CFI were calculated. CFI differences of 0.05 or less (Little, 1997) or 0.01 or less (Cheung & Rensvold, 2002) can be interpreted as negligible. Each model was compared to the simpler one: the three-factor with correlated factors model was compared to the one-factor model in terms of ∆CFI, and the five-factor with correlated factors model was compared to the three-factor with correlated factors model.

Thirdly, the best fitting structure for the SOCS–O was tested for measurement invariance by using multi-group models that compared invariance between gender (women, n = 570, and men, n = 108). As a previous step, and once the best-fitting model had been retained, the structure was tested separately in samples of women and men. Because the model fitted both sets of data adequately, the invariance routine was developed. Invariance was assessed using a sequential strategy testing the invariance at different levels. First, a configural or unconstrained model was evaluated. This model imposed no equality constraints on parameters and provided a baseline model for comparing the more restrictive models (Byrne, 2012). If this model holds, the metric or weak model is tested. This model is nested into the configural model and examines the extent to which the magnitude of item factor loadings is the same across groups (Brown, 2006). Finally, if metric invariance holds, the scalar or strong model is evaluated. This model tests for the evidence that thresholds (intercepts) for the items are invariant across groups (Brown, 2006). If scalar invariance is held, then latent means can be compared across groups.

To compare the nested models in the invariance routine, we used χ2 differences (∆χ2). However, as this statistical comparison presented the well-known problem of being too sensitive to trivial differences (Cheung & Rensvold, 2002), we also calculated delta indices. CFI differences (ΔCFI) and changes in RMSEA and SRMS were considered. Regarding the interpretation of ∆CFI, differences lower than 0.01 or 0.05 are usually used as cut-off criteria for equivalence across groups (Cheung & Rensvold, 2002; Little, 1997). For adequate metric invariance, ∆RMSEA and ∆SRMR should be ≥ 0.01 and ≥ 0.025, respectively (Chen, 2007). While evaluating scalar invariance changes ≥ 0.01 in RMSEA and ≥ 0.005 in SRMR are considered indicators of invariance (Chen, 2007).

Finally, criterion-related validity evidence was studied by predicting the SOCS–O scores through students’ levels of mindfulness, while controlling for gender and age. For this purpose, a structural equation model was hypothesized, estimated, and tested, in which gender, age, and a latent factor of mindfulness, as measured with the Freiburg Mindfulness Inventory, predicted the three-correlated factors measured by the SOCS–O. Correlations between mindfulness and gender and age and between the dimensions of compassion for others were also estimated.

Normality was tested for the variables under study. Non-normal distributions were found for all the items. Therefore, all the models were estimated using maximum likelihood with robust standard errors. This is the recommended option for non-normal outcome variables, such as the one under study (Kline, 2023).

For the statistical analyses, SPSS version 28 (IBM Corp, 2021) and Mplus version 8.4 (Muthén & Muthén, 2017) were used.

Results

Spanish Version of the Sussex-Oxford Compassion for Others Scale

The SOCS–O was translated into Spanish. The final version was reviewed by three experts in psychometrics and mindfulness, who agreed on the adequacy of the item content in the Spanish version of the SOCS–O. The resulting Spanish version of the scale can be consulted in Table 2.

Descriptive Statistics

Descriptive statistics were calculated for the dimensions and the items of the SOCS–O. Regarding the dimensions of compassion for others, the lowest mean was found for Recognizing suffering (M = 15.75, SD = 2.72), whereas the highest was for Understanding the universality of suffering (M = 17.80, SD = 17.80). Item means ranged from 3.57 (Item 19, “When someone else is upset, I can be there for them without feeling overwhelmed by their distress”) to 4.50 (Item 2, “I understand that everyone experiences suffering at some point in their lives”). More details can be consulted in Table 2.

Construct Validity and Reliability

For the study of the psychometric properties of the SOCS–O, three confirmatory factor analyses were specified, estimated, and assessed, using the structures detailed in the “Method” section. These models were as follows: a one-factor model, a three-factor with correlated factors model, and a five-factor with correlated factors model. In Table 3, model fit indices for the three CFAs are shown.

Table 3 Results of the confirmatory factor analyses and measurement invariance routine
Full size table

When the models’ fit was studied, one-factor model was discarded, as model fit was not acceptable. The three-factor with correlated factors model and five-factor with correlated factors model showed excellent fit, with no CFI differences. Taking this evidence into account, together with the fact that very high correlations were observed in the five-factor with correlated factors model between the dimensions of Feeling for the person suffering, Tolerating uncomfortable feelings, and Acting to alleviate suffering, the three-factor with correlated factors model was retained as the best structure to represent the Spanish version of the SOCS–O internal structure.

When the particular relationships within the retained model were examined, evidence of adequate factor loadings were found. Indeed, factor loadings ranged from 0.75 (Item 1 and Item 11) to 0.84 (Item 6) for Recognizing suffering; from 0.76 (Item 2) to 0.83 (Item 12) for Understanding the universality of suffering; and from 0.34 (Item 19) to 0.81 (Items 13, 5, 10, and 15) for Behavioral compassion. All of these results were statistically significant (p < 0.01). Correlations among latent variables were positive, high and statistically significant, as expected (Fig. 1). No cross-loadings or unmodeled relationships emerged in the modification indices. The pattern of the residuals was also inspected, with no evidence of misspecification in the model.

Fig. 1
figure 1

Analytical fit of three-factor with correlated factors model for the Spanish version of the SOCS–O. Notes: All factor loadings and correlations among factors were statistically significant (p < 0.01). For the sake of clarity, standard errors are not shown

Full size image

To study the reliability of the SOCS–O scale, Cronbach’s alpha and McDonald’s omega were calculated, with values of 0.87 for Recognizing suffering; 0.88 for Understanding the universality of suffering; and 0.93 for Behavioral Compassion; thus, estimations indicated adequate reliability for all the dimensions.

Measurement Invariance Across Gender

Multi-group analyses were conducted to examine the measurement invariance between women and men for the three-factor with correlated factors model. Table 3 shows the results of these multi-group analyses. The configural model demonstrated adequate fit, and so was used as the basis for testing more constrained models. Accordingly, the metric model was tested, in which factor loadings were constrained to be equal across gender. This model did not present statistically significant differences from the configural model (∆χ2 = 26.53, ∆df = 20, p = 0.15) and showed no deterioration of model fit (∆CFI = 0.00; ∆RMSEA = 0.00). Next, intercepts were constrained to be equal across gender. Again, no statistically significant chi-square differences were found when compared to the metric model (∆χ2 = 20.05, ∆df = 17, p = 0.27), nor was any worsening of the model’s fit observed (∆CFI =  − 0.01; ∆RMSEA = 0.00). Thus, it can be seen that the Spanish version of the SOCS–O demonstrated measurement invariance across gender.

When latent means were compared, men repeatedly exhibited lower levels in all the compassion dimensions: Recognizing suffering (mean difference =  − 0.40, standard error = 0.12, p < 0.01), Understanding the universality of suffering (mean difference =  − 0.33, standard error = 0.11, p < 0.01), and Behavioral compassion (mean difference =  − 0.57, standard error = 0.12, p < 0.01).

Criterion-Related Validity

Evidence for criterion-related validity was studied with structural equation modelling, in which gender, age, and a latent factor of mindfulness predicted the three dimensions of the Spanish version of the SOCS–O. The model showed an adequate overall fit: χ2(581) = 1345.79 (p < 0.01), CFI = 0.92, TLI = 0.92, RMSEA = 0.04 (90% CI = 0.04, 0.04), and SRMR = 0.06. The measurement part of the model pointed to adequate factor loadings for the factors of mindfulness and compassion for others. Mindfulness showed positive and statistically significant correlations with gender, which meant higher scores for men, who were coded as 1 (whereas women were coded as 0), and age, with higher levels of mindfulness for older students. The predictive part of the model showed statistically significant effects of gender on the three dimensions of compassion for others, as already pointed out in the latent means comparison; a negative effect of age in all the dimensions of compassion for others, except for the Understanding the universality of suffering; and a positive effect of mindfulness on the three dimensions of compassion for others. Details are displayed in Fig. 2.

Fig. 2
figure 2

Structural equation modeling predicting compassion for others dimensions with gender, age, and mindfulness. Notes: *p < 0.05; **p < 0.01. All factor loadings were statistically significant (p < 0.01). For the sake of clarity, they are not shown

Full size image

Discussion

The aim of this study was to explore the SOCS–O scale using a large sample of Spanish undergraduate nursing students. We examined the psychometric properties of this scale with a particular focus on the five elements of compassion, initially identified by Strauss et al. (2016) and further developed by Gu et al. (2020).

Our specific aims were to measure nursing students’ levels of compassion; to study the internal factor structure through confirmatory factor analysis (CFA); to provide evidence of the reliability of the scale in the Spanish language; to investigate its measurement invariance across gender; and, finally, to evaluate criterion validity through the relationship between SOCS–O and gender, age, and mindfulness.

Descriptive statistics indicated high mean results (ranging from 3.57 to 4.50), in line with the findings by Lucarini et al. (2022), who found values ranging from 3.58 to 4.21. The dimension of compassion with the highest mean results was that of understanding the universality of suffering. Perhaps this is not so surprising, as nurses are exposed to others suffering on a daily basis as part of their work, so are accustomed to the presence and ubiquity of suffering (Kelley et al., 2014). The areas with the lowest means were Recognizing suffering and the management of one’s own emotions in the face of another’s distress. Similar conclusions have been drawn in other language validations. In Lucarini et al. (2022), for example, two themes were drawn from the data on compassion: “universality” (of feelings/suffering) and “acting” on these feelings. In Halamova and Kanovsky (2021), a distinction was drawn between rational compassion (i.e., to recognize and understand suffering) versus emotional or behavioral compassion (which concerned feelings, tolerance, and the desire to act). Similarly, Kim and Seo (2021) noted these trends in their Korean validation, where most variation was within the category of Tolerating uncomfortable feelings, and most agreement (least variations) was found in identifying and being concerned over the suffering of others. The fact that similar patterns were identified across various validations suggests the underlying psychometric robustness of the SOCS–O.

In order to support the underlying robustness ascertained with descriptive statistics, construct validity was examined using competitive confirmatory factor analyses. Evidence pointed to an excellent fit for the three and five-factor with correlated factors solutions. In order to maintain simplicity and taking into account the high correlations observed among the dimensions of Feeling for the person suffering, Tolerating uncomfortable feelings, and Acting to alleviate suffering, the three-factor model with correlated factors was retained as the best representation of the data. Therefore, the structure of compassion, as measured with the SOCS–O, was better represented in Spanish nursing students by the following three dimensions: Recognizing suffering, Understanding the universality of suffering, and Behavioral compassion. These results are in line with the ones presented by Halamova and Kanovsky (2021) for the SOCS–S in Slovakia. The authors defended a rational versus behavioral approach to compassion, the latter composed of items from the original dimensions of Feeling for the person suffering, Tolerating uncomfortable feelings, and Acting to alleviate suffering, as in the present work. Previous literature has already pointed to these different components of compassion (Galiana & Sansó, 2019), distinguishing between a more cognitive component, which involves awareness of others’ suffering (Jinpa, 2015) and refers to recognizing and understanding others’ suffering (Galiana & Sansó, 2019); and a behavioral component, which includes the actions we perform to address suffering, and are always preceded or accompanied by feelings towards and tolerance of the person suffering (Galiana & Sansó, 2019).

All factor loadings were greater than 0.50, except for 19 in the dimension of Behavioral compassion (“When someone else is upset, I can be there for them without feeling overwhelmed by their distress”). This may be due to the fact that this item not only measures compassion, but also emotional management skills, specifically, the management of managing one’s own distress. This was also observed in the original research results (Gu et al., 2020), with Item 19 being the only one with a factor loading < 0.50, so showing important similarities between the scale functioning in both studies.

Regarding the third aim of our study, internal consistency was established using the statistical measures of Cronbach’s alpha and omega. Results showed all reliability estimates to be excellent, indicating that participant responses across all questions were consistent. This then allows for validation of the research, as results within this study are not only internally consistent, but also comparable with those of the original research by Gu et al. (2020), as well as other validated models of this scale (Kim & Seo, 2021; Lucarini et al., 2022; Nikgoo et al., 2022).

The fourth aim was to study the measurement invariance of the scale across gender. Measurement invariance is a key element when making group comparisons when the groups can be understood as different populations (different genders, countries, races, cultures, professions, etc.), as it concerns whether scores have the same meaning under different conditions (Kline, 2023). For this purpose, a hierarchical set of increasingly restricted models was employed. The scale showed evidence of scalar invariance, which is an important prerequisite for group comparison (Putnick & Bornstein, 2016). This equivalence of compassion across gender, as assessed by the SOCS–O, has been previously demonstrated in other languages, such as Swedish (Sarling et al., 2024). However, the only study focused on nursing students had a very small sample of males so gender comparisons were not possible (de Krijger et al., 2022).

When latent means were compared, results consistently showed statistically significant differences in the levels of compassion for men and women, with women displaying higher values of Recognizing suffering, Understanding the universality of suffering, and Behavioral compassion. Although there were similar findings in the study by Gu et al. (2020), where females scored significantly higher on the compassion scale compared to male counterparts, other SOCS–O validations do not always show this. Kim and Seo (2021) found that there were no differences between Korean gender scores in relation to compassion for others. The Italian validation by Lucarini et al. (2022) did not use gender as a criterion for validity but did observe higher compassion scores among female participants. It is important to note however that this was a large random sample and not focused on nursing students or healthcare workers.

If we look to the wider literature however, women have traditionally shown higher levels of compassion. This can be seen anecdotally in popular culture (Yarnell et al., 2019), where women are frequently depicted as caring, emotional, and nurturing in contrast to men who stereotypically are described as feeling awkward talking about their emotions; as well as in more robust scientific studies, where women are often shown to better identify emotions (Connolly et al., 2019) and recognize suffering (López et al., 2018). It is interesting however to examine gender and compassion within the context of nursing, traditionally a caring field that now attracts both men and women. It could be argued that these results imply that men are less compassionate than women in nursing care. However, it is important to note culture and the socialization of genders, especially in the face of caring behaviors (Gilbert, 2014; Yarnell et al., 2019). Both genders may feel compassion but the way in which it is manifested may differ. For instance, in a study carried out by Mercadillo et al. (2011) when male and female participants were placed in an MRI scanner and exposed to compassion-evoking images, both genders demonstrated neural responses, but in different parts of the brain. The research suggested that certain moral elements are gender-relative, evolving from neural mechanisms of socially learned patterns. So, it is not that men and women experience more or less compassion, but that they may express it differently. Seppälä (2013) suggested that compassion has traditionally been portrayed as a soft feminized skill (perhaps reflected in the questions and results of compassion scales) and argued that men often show “aggressive compassion,” something our society does not frame as compassion but instead as protection, such as firefighters or police. We could potentially apply this perspective to the field of nursing, where compassion can be demonstrated (but perhaps is not always recognized) in different ways.

Finally, criterion-related validity was established by relating mindfulness to compassion, while controlling for the effects of gender and age. As found in the measurement invariance results, and previously discussed, gender predicted all the dimensions of compassion, with higher scores for women. In contrast, age was negatively related with the compassion dimensions of the SOCS–O, except for the Understanding of universal suffering. That is, younger nursing students showed greater levels of compassion. These results were unexpected, as compassion is considered to be a universal human quality, regardless of the age of the perceiver of suffering (Lama, 2002; Seppälä et al., 2017). In fact, in the few studies that examined age differences on compassion, results pointed to younger participants being less compassionate, especially with regard to self-compassion (Jarden et al., 2021; Murn & Steele, 2020; Tran et al., 2022). The only dimension that was not affected by age was the dimension of Understanding the universality of suffering, but it must be interpreted in its context: it may be the higher scores found on this dimension are due to low variability that has not allowed us to find an effect of age. Future research could explore the relationship between compassion and age and how it can affect the caring professions.

As for the effect of mindfulness on the dimensions of compassion, this was repeatedly positive and statistically significant. This supports previous research that has repeatedly found that mindfulness, defined as the way of paying attention to the present moment without judgment (Bishop et al., 2004), could help nurses and nursing students understand and address patients’ suffering. Incorporating mindfulness into the nurses’ curriculum could improve students’ skills for compassionate care, which may also enhance the student experience (Walker & Mann, 2016).

The validation of the SOCS–O compassion scale in Spanish has shown not only a robust internal factor structure and criterion validity, but also serves as a reliable tool to measure compassion within the Spanish-speaking world. As discussed above, it is important to be able to measure compassion, as it is a previously unquantifiable element of nursing care that impacts positively on both the professional and the patient. Now this essential element of nursing can be accurately measured, akin to symptom management scales. All these more tangible factors of nursing have associated tools to measure their efficacy which allows for protocols and improvements in patients’ care delivery to be put in place. Now it will be possible to do this with compassion as well, something which has not always been possible due to conflicting terminology and lack of robust internal measures. With the aid of the validated SOCS–O, not only can compassion be more accurately measured in different languages, but there is also potential to further advance the delivery of patient-centered, compassionate care. It will now be possible to monitor how students (and in the long-run experienced nurses) acquire compassionate competencies throughout their training, so allowing for tailored educational activities regarding compassionate competencies to be designed and applied for future educational programs.

Limitations and Future Research

As regards the study's shortcomings, one potential limitation was that the validation of this scale was carried out at two universities. In order to gain a greater representation of potentially different manifestations of compassion, future studies could incorporate universities from more regions across the country. Additionally, although the samples of women and men were unbalanced, the total size of the men’s sample did not allow us to adjust group size. However, it is worthy to note that simulation studies testing factorial invariance with unbalanced samples have pointed out that the chi-square values are not much affected by measurement non-invariance (Yoon & Lai, 2018). Therefore, as we have found evidence of scalar invariance in terms of not statistically significant differences in the chi-square, it can be argued that it is indeed a gender-invariant scale. In any case, this limitation will be taken into account in subsequent studies, providing evidence in balanced samples.

Considering current results, it will now be possible to explore how nursing students manage compassion in a more robust and reliable way, using the SOCS–O. One area that would benefit from deeper research is the five dimensions of compassion in different cultural contexts. Another future direction, also noted by Gu et al. (2020), would be the development of a longitudinal study, to address any potential common method variance, often associated with self-reporting data (Podsakoff et al., 2003) as well as to further explore response bias and the tendency for participants to reply in accordance with social desirability expectations (Lucarini et al., 2022).

Finally, by validating the SOCS–O in Spanish, further cross-cultural exploration of compassion, as suggested by Gu et al. (2020) and Halamová and Kanovský (2021), will be possible in the Spanish-speaking world. The original sample consisted mainly of White (85.7%) females (87.9%) from southern England. Halamová and Kanovský (2021) also observed cultural differences in understanding the translated instrument within Slovakia, so highlighting the potential for further cross-cultural research with the validated Spanish tool.

The findings of this study demonstrate the robust psychometric properties of the Spanish version of the Sussex-Oxford Compassion for Others Scale. The descriptive results show that nursing students have high scores on all the dimensions of the SOCS–O. Validity and reliability results are excellent, and confirm the Spanish version of the measure is a valid and reliable tool to evaluate compassion for others in Spanish-speaking contexts. The SOCS–O offers a useful instrument to examine gender differences in compassion. The study showed women have consistently higher self-reported levels of compassion than men. Additionally, the positive relationship between mindfulness and compassion, as measured with the SOCS–O, was found on all the dimensions of the measure, suggesting the value of incorporating mindfulness in nurses’ curricula in order to support the delivery of compassionate care.

In conclusion, this study validates the SOCS–O scale in Spanish, and offers a reliable tool with which to measure compassion using a three-dimension subscale as well as exploring how student nurses manage compassion in the face of their placements. The main contribution of this study to the field of nursing is that it will facilitate the use of a robust, psychometric measurement of compassion among student nurses that could enable interventions to support and sustain compassion during nurse education.