The CEASY-SE was translated into Swedish with permission from the original authors of the CEAS (Gilbert et al., 2017). A back-translation method was used to ensure semantic and content equivalence in cross-cultural research (Hambleton, 2004). Henje et al. (2020) provide a detailed description of the translation procedure.

Participants were recruited from four child and adolescent psychiatric clinics, one primary care youth clinic, and one general primary care health clinic in four towns in northern Sweden. After approval from the operational managers, information flyers were placed in the clinics' waiting rooms. Additionally, patients with self-reported anxiety and depression, who were receiving care at the clinics, received study invitations via SMS or mail. Research assistants contacted both responders and non-responders by telephone to provide comprehensive verbal and written information about the study. This included details about the study's purpose, procedures, potential risks, benefits, and the option for voluntary participation. Informed consent was obtained from all participants, with additional consent from caregivers for those under 18 years of age. The first 65 participants completed paper-based assessments between October 2018 and May 2020. Between May 2020 and August 2021, 290 participants were invited to complete online assessments. Due to the impact of the COVID-19 pandemic on data collection, as well as delays in data processing, analysis, and manuscript preparation, the completion of final analyses and manuscript preparation occurred later than initially planned.

Participants were asked if they received mental health care and to report any current diagnoses. For those recruited from psychiatric clinics, diagnoses were extracted from medical records by a research assistant after data collection. Diagnoses were classified according to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) (American Psychiatric Association, 2013) or the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10) (WHO, 2019), and grouped into clusters.

Confidentiality was ensured through secure data storage and anonymization. Paper-based assessments were securely stored at Umeå University. Online data were collected using Research Electronic Data Capture (REDCap) (Harris et al., 2009) hosted at Umeå University. Upon completion, participants were reimbursed with a gift card worth 200 SEK (approximately 20 EUR).

A total of 582 participants were invited to participate, of whom 358 (61 %) agreed. This sample size is considered adequate for psychometric evaluation in cross-sectional designs (Evers et al., 2013), and the relatively high response rate (61%) enhances the generalizability of the findings. Attrition (n = 224) was primarily due to ineligibility, mostly related to age restrictions (n = 204; 90%). The eligibility criteria were as follows: (1) aged 16–22 years, (2) currently receiving care at one of the recruiting sites, (3) self-reported or parent-reported symptoms of depression and/or anxiety (all comorbidities were allowed), (4) for individuals with a recent history of suicide attempt or psychiatric inpatient-care (a minimum of 3 months had to have passed since the suicidal event or hospitalization discharge), and (5) fluent in written Swedish and able to complete the self-assessment.

The age range of 16 to 22 years was selected to capture a developmentally sensitive period marked by heightened emotional vulnerability. This range also aligns with the structure of Swedish youth healthcare services and with our randomized controlled trial (RCT) intervention study on compassion-focused therapy for young people (Vestin et al., 2025).

For comparative purposes, a normative sample of Swedish young people (N = 316; 213 girls, 67%) aged 15–20 years (M = 17.07, SD = 1.36) was collected by our research group (Henje et al., 2020). This sample was used to contextualize the clinical findings and assess differences in compassion-related constructs between clinical and non-clinical youth populations.

All analyses were conducted using SPSS, version 29.0 (Corp, 2023), and R, version 4.4.1 (Core, 2021).

The final sample included 355 participants, of whom 287 were female (81%). Participants' ages ranged from 16 to 22 years (M = 18.12, SD = 2.08). Parental SES indicated that none were unemployed, students, or on sick leave. Detailed demographic information is presented in Table 1.

Descriptive statistics for CEASY-SE items and subscales were computed (see Supplementary material), revealing that corrected item-total correlations (rit-c) were ≥ 0.30 across total and sex subscales, except for the item “I can stand my own different types of feelings” in the Self-Compassion Engagement subscale in the male sample (rit-c = 0.25). Despite this, the item was retained to preserve the conceptual breadth of the scale.

Tests of normality indicated that several CEASY-SE scales were significantly skewed (p < 0.05), particularly among females. The only subscale that met normality assumptions across both sexes was the Self-Compassion Engagement subscale (Shapiro-Wilk p = 0.526 for females, p = 0.246 for males; Kolmogorov-Smirnov p = 0.200 for both).

Descriptive statistics and group comparisons for CEASY-SE total and subscale scores by sex are presented in Table 2. Females scored significantly higher than males on the Compassion for Others total scale (d = 0.50, p < 0.001) as well as on the Engagement (d = 0.42, p = 0.002) and Action subscales (d = 0.50, p < 0.001). Conversely, males scored significantly higher than females on the Self-Compassion total scale (d = −0.41, p = 0.002), the Engagement (d = −0.33, p = 0.014), and Action subscales (d = −0.44, p = 0.001). No significant sex differences were found for the Compassion from Others total scale (d = −0.02, p = 0.878).

To further examine the effects of both sex and age, a two-way ANOVA was conducted (see Table 3). Significant main effects of were found for the Compassion for Others total scale: F_{(1, 351)} = 13.74, p < 0.001, partial η² = 0.04, and the Self-Compassion total scale, F_{(1, 351)} = 7.16, p = 0.008, partial η² = 0.02 (see Table 3). Additionally, significant interaction effects between sex and age were observed for both Compassion for Other total scale, F_{(1, 351)} = 12.35, p < 0.001, partial η² = 0.03, and the Self-Compassion total scale, F_{(1, 351)} = 6.39, p = 0.012, partial η² = 0.02. Self-Compassion increased with age among females but decreased among males. A similar interaction pattern was observed for Compassion from Others, whereas Compassion for Others remained relatively stable across age groups. Age was analyzed as a categorical factor with two groups (16–17 and 18–22 years) to reflect development stages.

Table 4 presents descriptive statistics for total scales of Self-Compassion, Compassion for Others, and Compassion from Others across psychiatric diagnoses, based on combined clinician-assigned and self-reported diagnoses. As expected, participants with depression (M = 42.44, SD = 14.01) and eating disorders (M = 40.36, SD = 14.69) reported the lowest levels of Self-Compassion. In contrast, those with ADHD (M = 48.42, SD = 17.38) and autism (M = 49.00, SD = 17.04) showed relatively higher Self-Compassion scores. Compassion for Others was generally high across all groups, with the highest score in the depression group (M = 79.26, SD = 9.99). In contrast, Compassion from Others was lowest among participants with autism (M = 47.41, SD = 17.65). These findings suggest that compassion profiles vary across different diagnostic categories.

No significant correlations were found between SES and any of the three CEASY-SE scales (all p > 0.05). A Chi-square test showed no significant differences in SES between sexes, χ²(5, N = 318) = 1.686, p = 0.891.

CFA supported the proposed two-factor structure for all three CEASY-SE scales:

Model 1: Compassion for Others

The model demonstrated good fit (χ²(26) = 83.89, χ²/df = 3.22, CFI = 1.00, TLI = 0.99, RMSEA = 0.08 (90% CI [0.06, 1.00]), SRMR = 0.05), confirming the model's suitability. Standardized factor loadings exceeded 0.40 for all items (Figure 1, Model 1).

Model 2: Compassion from Others

The initial two-factor model showed poor fit (χ²(26) = 111.18, χ²/df = 4.28, CFI = 0.99, TLI = 0.99, RMSEA = 0.10 (90% CI [0.08, 0.12]), SRMR = 0.05). After allowing residual correlations between (1) item 1 and item 5 from the Engagement subscale and (2) item 1 and item 2 from the Action subscale, model fit improved (χ²(24) = 62.54, χ²/df = 2.61, CFI = 1.00, TLI = 1.00, RMSEA = 0.07 (90% CI [0.05, 0.09]), SRMR = 0.04). Factor loadings remained above 0.40 (see Figure 1, Model 2).

Model 3: Self-Compassion

The initial two-factor model showed suboptimal fit (χ²(26) = 103.02, χ²/df = 3.96, CFI = 0.99, TLI = 0.99, RMSEA = 0.09 (90% CI [0.07, 0.11]), SRMR = 0.05). After introducing residual correlations between (1) item 3 and item 5 from the Engagement subscale and (2) item 1 and item 2 from the Action subscale, model fit improved (χ²(24) = 62.81, χ²/df = 2.62, CFI = 1.00, TLI = 1.00, RMSEA = 0.07 (90% CI [0.05, 0.09]), SRMR = 0.04). All factor loadings were above 0.40 (see Figure 1, Model 3).

The internal consistency of the CEASY-SE was generally good to excellent across all scales concerning the total sample, with Cronbach's alpha values ranging from 0.75 to 0.92 and McDonald's omega values showing similar strength. The only exception was the Self-Compassion Engagement subscale among males, which showed lower reliability (α = 0.62, ω = 0.60), suggesting potential variability in how males interpret or respond to these items. All other subscales demonstrated acceptable to excellent reliability, supporting the scale's robustness across most domains and subgroups. Table 5 summarizes the internal consistency estimates for each subscale, by sex, and for the total sample.

Spearman's correlation analyses revealed strong internal relationships between Engagement and Action subscales within each compassion scale. For example, the correlation between Engagement and Action in the Compassion for Others scale was r = 0.65 (p < 0.01), indicating these components are closely related. In contrast, correlations between different scales (e.g., Self-Compassion and Compassion for Others) were weak or absent (r = 0.00–0.13), suggesting that these constructs are relatively independent in this sample. That is, being self-compassionate does not necessarily imply being more compassionate toward others, and vice versa. Table 6 presents the intercorrelations between the subscales.

CFA supported the proposed two-factor structure for each of the three CEASY-SE scales. While the Compassion for Others scales showed good model fit without modifications, the Compassion from Others and Self-Compassion scales required minor adjustments (residual correlations) to improve fit. These findings align with previous validations of the CEAS and CEASY-SE both in young and adult populations (Cunha et al., 2023; Gilbert et al., 2017; Henje et al., 2020). The need for model refinement highlights the complexity of measuring compassion, particularly in clinical youth populations, and suggests that further refinement may enhance model fit in future studies.

Reliability, in terms of internal consistency, was good to excellent across most scales and subscales except the Self-Compassion Engagement scale among males. This mirrors findings from earlier studies and suggests that males may interpret or respond to self-compassion differently (Cunha et al., 2023; Gilbert et al., 2017; Henje et al., 2020). This discrepancy should be taken into account when interpreting the results for this subgroup and may warrant refinement of the scale.

The strong correlations within the Compassion for Others, Compassion from Others, and Self-Compassion scales support the internal consistency of the scales. Specifically, the strong correlations between the Engagement and Action subscales within each category suggest that these aspects of compassion are closely related. In contrast, the weak correlation between Self-Compassion and Compassion for Others subscale indicates that these are relatively independent constructs. These support the theoretical distinction between the three flows of compassion and underscore the importance of assessing them separately.

As expected, CEASY-SE scores were positively associated with self-compassion and wellbeing and negatively associated with anxiety and depression. These findings support the scale's convergent validity and reinforce the protective role of compassion in mental health. Interestingly, the Compassion for Others scale showed small positive correlations with anxiety and depression, which may reflect emotional burden or empathic distress in highly compassionate individuals, a pattern also observed in previous CEAS research (Gilbert et al., 2017; Brophy et al., 2024). This finding suggests that outward compassion, although generally adaptive, may be associated with emotional strain in vulnerable populations.

Divergent validity was supported by weak or non-significant correlations with unrelated constructs, such as physical activity and functional impairment, indicating that the CEASY-SE measures distinguish between psychological domains. This strengthens the argument that the scale captures specific compassion-related constructs rather than general distress or functioning. This result aligns with earlier studies (Brophy et al., 2024; Gilbert et al., 2017). Similarly, the initial validation study of the CEASY-SE found low correlations with unrelated constructs, supporting the divergent validity of the scales (Henje et al., 2020).

The CEASY-SE effectively distinguished between clinical and non-clinical youth. While clinical participants reported higher on the Compassion for Others scale, they scored significantly lower on the Self-Compassion and Compassion from Others scales. These findings suggest that although young people in psychiatric care may be attuned to others' suffering, they struggle more with receiving and directing compassion inward. This pattern is consistent with clinical observations and empirical findings showing that self-criticism and shame are common among youth with internalizing disorders (Mcintyre et al., 2018). It underscores the importance of targeting self-compassion and receptivity to support in therapeutic interventions for young people in clinical psychiatric care.

Additionally, scores on the Self-Compassion scale varied across different psychiatric diagnoses. As expected, the lowest levels were observed among participants with depression and eating disorders, which are conditions commonly associated with high self-criticism and shame. In contrast, participants with autism and ADHD reported relatively higher self-compassion. These differences may reflect diagnostic-specific emotional processing styles, such as reduced self-monitoring in ADHD or unique cognitive-emotional processing in autism. However, these interpretations remain speculative and require further research (Cai et al., 2023). Scores on the Compassion for Others scale were generally high across all groups, with the highest scores in the depression group, which possibly reflects heightened sensitivity to others' suffering. Meanwhile, scores on the Compassion from Others scale were lowest among participants with autism, potentially indicating difficulties in perceiving or receiving social support. However, this shows the potential value of tailoring compassion-based interventions to specific diagnostic groups, particularly those with internalizing disorders.

These findings are consistent with Gilbert's theoretical model, which suggests that the three flows of compassion, self-compassion, compassion for others, and compassion from others, are only moderately correlated and may function independently (Gilbert et al., 2017; Gilbert, 2017). For instance, an individual may be highly self-compassionate but not necessarily receive much compassion from others. This aligns with our findings, which showed that while the Self-Compassion scale was strongly correlated with mental health outcomes, Compassion for Others and Compassion from Others scales had more varied relationships. These distinctions highlight the importance of evaluating each compassion flow independently in both research and clinical practice.

Self-Compassion scores significantly predict the likelihood of depressive symptoms. The logistic regression model explained 26% of the variance in depression symptoms, correctly classifying 80.3 % of cases. This suggests that self-compassion may serve as a clinically meaningful indicator of vulnerability in youth. Previous studies have consistently shown that higher levels of self-compassion are associated with lower levels of negative emotional states (Gilbert, 2020; Marsh et al., 2018). These findings confirm the clinical relevance of self-compassion and support its use as a target for screening or intervention in youth mental health care. It also supports the use of CEASY-SE in clinical settings to assess and enhance self-compassion as a protective factor against mental health issues.

Our findings revealed sex and age differences in compassion, as well as developmental patterns. Females reported higher levels on the Compassion for Others scale, while males scored higher on the Self-Compassion scale. Given the predominance of female participants, the observed sex differences should be interpreted with caution, as the smaller number of male participants may have influenced variability and statistical power. However, these results are consistent with previous research in normative samples (Henje et al., 2020; Cunha et al., 2023), suggesting that females may be more attuned to the needs of others, potentially at the expense of their own self-compassion. This aligns with the broader literature, which indicates that females often exhibit higher levels of empathy and caregiving behaviors (Pang et al., 2023).

We observed a sex-by-age interaction: self-compassion increased with age among females but decreased among males. This pattern diverges from previous findings by Bluth et al. (2017), who reported that self-compassion tends to decline with age among adolescent girls, while remaining relatively stable among boys. One possible explanation for this discrepancy is that our sample included older adolescents and young adults (16–22 years), whereas Bluth's study (Bluth et al., 2017) focused on a younger adolescent (14–17 years). Differences between clinical and community samples may also influence the development of compassion across different age groups.

In this study, age was analyzed using two predefined categories: adolescents (16–17 years) and young adults (18–22 years), reflecting key developmental stages. This approach enabled us to investigate whether compassion-related processes vary between late adolescence and early adulthood. The observed interaction may reflect distinct developmental trajectories, where older females benefit from greater emotional regulation and self-reflection, whereas males might be more influenced by social or cultural factors that inhibit self-compassion. Future studies could build on these findings by combining categorical and continuous age analyses to capture more nuanced developmental patterns.

These findings suggest that the ability to give and receive compassion develops differently across age groups in clinical populations. Developmental stages present unique challenges: younger people often focus on identification and social relationships, while older people develop greater emotional regulation and self-reflection through more life experiences and emotional resilience (Tavares et al., 2024). Tailoring interventions to address these sex and age differences could improve their effectiveness. For instance, programs for females might focus on building self-compassion, while those for males could emphasize the importance of receiving compassion from others. Future research should continue to explore these patterns in diverse clinical and non-clinical populations.

This highlights the importance of adopting a multidisciplinary perspective when studying compassion in youth. Developmental psychology, cultural psychology, and evolutionary theory each offer complementary insights into how compassion emerges and functions across contexts. Evolutionary models emphasize caregiving systems and social affiliation as foundational mechanisms (Gilbert, 2017), whereas cultural frameworks point to variations in how compassion is expressed, valued, and internalized across societies (Jiang et al., 2023; Strauss et al., 2016). Adolescence is a particularly dynamic period in which identity formation intersects with cultural norms, gender roles, and social expectations, all of which may shape the development and expression of compassion. Integrating these perspectives can enhance our understanding of compassion as both a universal and culturally embedded phenomenon and may inform the generalizability of findings beyond the Swedish clinical context.

CEASY-SE is a promising tool for assessing compassion in clinical youth populations. It can help clinicians identify compassion-related strengths and vulnerabilities, guide treatment planning, and monitor therapeutic progress. Its multidimensional structure enables a nuanced understanding of how compassion is experienced and expressed across various relational contexts. Given the strong link between self-compassion and mental health, compassion-focused interventions (e.g., CFT) may be particularly beneficial for this group. Future studies should also explore the scale's sensitivity to change in response to intervention to further establish its clinical utility.

Several limitations should be noted. First, the sample was drawn from clinics in northern Sweden and was predominantly female, which may limit generalizability. Second, the cross-sectional design precludes causal inferences. Third, the reliance on self-report measures introduces potential biases such as social desirability. Finally, although the sample encompassed a range of psychiatric diagnoses, subgroup sizes were small, which limited the ability to draw specific conclusions for each diagnosis.

Future research on CEASY-SE should aim to replicate these findings in a more diverse and balanced sample, including younger adolescents and underrepresented sexes. Studies should also focus on specific psychiatric diagnoses to better understand the tool's applicability. Longitudinal studies are necessary to investigate how compassion evolves over time and in response to treatment. Further refinement of the Self-Compassion Engagement subscale for males may also be warranted. Additionally, exploring the scale's sensitivity to change in intervention studies would strengthen its utility in clinical practice.

This study aimed to validate the psychometric properties CEASY-SE in a clinical psychiatric sample of young people. The findings demonstrate that CEASY-SE is a reliable and valid instrument for assessing compassion across three relational flows—toward others, from others, and toward oneself. Despite some limitations, the scale showed good to excellent internal consistency, a replicable factor structure, and substantial evidence of convergent, divergent, construct, and criterion-related validity.

In addition to confirming the scale's psychometric soundness, this study contributes novel insights into how compassion varies across psychiatric diagnoses and interacts with age and sex. Notably, self-compassion was found to decrease with age among males, while increasing among females, highlighting the need for sex-sensitive interventions. Furthermore, the substantial predictive value of self-compassion for depressive symptoms underscores its clinical relevance as a screening and intervention target.