Erasmus MC heeft ervoor gezorgd dat je Mijn BSL eenvoudig en snel kunt raadplegen. Je kunt je links eenvoudig registreren. Met deze gegevens kun je thuis, of waar ook ter wereld toegang krijgen tot Mijn BSL.
Om ook buiten de locaties van Erasmus MC, thuis bijvoorbeeld, van Mijn BSL gebruik te kunnen maken, moet je jezelf eenmalig registreren. Dit kan alleen vanaf een computer op een van de locaties van Erasmus MC.
Eenmaal geregistreerd kun je thuis of waar ook ter wereld onbeperkt toegang krijgen tot Mijn BSL.
Login
Als u al geregistreerd bent, hoeft u alleen maar in te loggen om onbeperkt toegang te krijgen tot Mijn BSL.
Scientific interest in compassion has increased in recent years. However, research highlights issues related to the measurement of compassion in children and young people (CYP). This systematic review aimed to describe and critically appraise measures of self- and other-compassion developed and/or validated with CYP.
Method
Medline, PsycInfo and Web of Science were searched from inception to 31 st July 2024 to identify English language, peer-reviewed psychometric papers on the development or validation of compassion measures for use with CYP. The measures were critically reviewed and rated for quality.
Results
Nineteen articles were included that outlined psychometric properties of eight compassion measures: Compassionate Engagement and Actions Scales—Adolescence (CEAS-A); Compassion Scale – Adolescence (CS-A); Compassion Scale—Child (CS-C); Self-Compassion Scale (SCS); Self-Compassion Scale – Child (SCS-C); Self-Compassion Scale—Youth (SCS-Y); Self-Compassionate Reactions Scale – Child (SCRS-C); Shortened Self-Compassion Scale—Adolescence (S-SCS-A). Quality ratings ranged from 2 to 9 out of 14. The CEAS-A received the highest rating (9), followed by the SCS-Y (8), SCS (7) and CS-C (7). No measure captured all five compassion elements and evidence for test–retest reliability and floor and ceiling effects was largely absent. Most papers assessed self-compassion and used (translated, adapted versions of) the SCS and SCS-Y.
Conclusions
This review suggests there are no psychometrically robust measures that comprehensively assess self- and other- compassion in CYP, potentially impeding research. The psychometric quality of existing compassion measures for CYP should be improved, and/or a new measure should be developed, and include the measurement of compassion for others.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
There has been increased scientific interest in compassion in recent years, with recognition of its importance in both education (Compassion in Education Foundation, 2016) and healthcare (Department of Health, 2013); there is growing evidence to suggest that compassion-focused training and intervention for teachers and/or pupils improves student wellbeing, cognitive functioning, resilience and peer relationships (e.g. Kappelmayer et al., 2022; Seekis et al., 2023), teacher stress and wellbeing (e.g. Matos et al., 2022). In healthcare, studies have shown that (training in) compassionate care reduces stress healthcare professionals’ stress and improves patient outcomes and experience (e.g. Sinclair et al., 2021).
Compassion is considered an evolutionary emotion, which is reproductively beneficial and has evolved as a crucial element of the caregiving system to nurture and protect offspring (Goetz et al., 2010). It is generally considered to be equivalent for others (other- compassion) and self (Feldman & Kuyken, 2011; Gilbert, 2014). Whilst understanding of this relationship is more limited in children and young people (CYP), findings suggest these orientations of compassion are overlapping constructs in CYP (Cunha et al., 2023). Several definitions exist in the literature, which reflect that compassion can be directed inward toward the self, or outward towards others (Roeser et al., 2018). Gilbert’s (2014) definition, “the sensitivity to suffering in self and others, with a commitment to try to alleviate and prevent it” (p. 19), encapsulates both. While Gilbert (2014) adopts an evolutionary, biopsychosocial perspective on compassion, Neff’s (2003a) conceptualisation draws on Buddhist philosophy and focuses on self-compassion; her conceptualisation of self-compassion involves three main components: self-kindness (being kind and understanding toward oneself rather than self-critical in instances of pain or failure), common humanity (perceiving one’s experiences as part of the larger human experience rather than as isolating), and mindfulness (holding painful thoughts and feelings in balanced awareness rather than over-identifying with them. These differences capture the wide scope of compassion research, which encapsulates self-compassion and interpersonal compassion and explores their associations with professional and family caregiving (e.g. Liao et al., 2022; Singh et al., 2018), social relationships (e.g. Lathren et al., 2021), and psychological wellbeing, (e.g. Zessin et al., 2015), including in the context of interventions that aim to improve compassion (e.g. Kirby et al., 2017).
Current understanding of the development of compassion is informed by research investigating the development of empathy-related responding and prosocial behaviour (Eisenberg, 2017; Roeser et al., 2018). The beginnings of compassion are thought to exist from the first year of life, shown through infants’ mirroring others’ emotions, e.g. crying when others cry (Gilbert, 2015), followed by children starting to develop rudimentary empathy, expressing concern and prosocial behaviour towards others, e.g. comforting their caregivers, in the second year of life (Eisenberg, 2000), associated with (further) self-other differentiation. In early childhood (ages 3–5 years), improving language and cognitive skills support the development of the capacity to understand others’ emotions and intentions, with children starting to show greater empathy and moral reasoning. Prosocial behaviour, particularly in familiar social environments, becomes more frequent and complex. Perspective-taking skills become more firmly established around six years of age and continue developing into adolescence. These skills are important for the development of compassion as they allow children to understand the emotions of others (Selman, 1980). Peer relationships and social awareness deepen during this time. Prosocial behavior becomes more deliberate and guided by moral reasoning and social norms (Eisenberg, 2000).
During adolescence, perspective taking skills approach adult levels. Compassion begins to mature during adolescence, as a more complex understanding of the self and others develops and as the ability to view experiences of the self and others from a third-person or societal perspective strengthens (Bengtsson et al., 2016; Selman, 1980). In line with the biopsychosocial perspective on compassion, individual differences in prosocial development are associated with temperamental factors like sociability, emotionality and shyness (including via their impact on emotion regulation) (Roeser et al., 2018). Family influences on prosocial development include attachment security, parenting styles and practices and parental affection, while broader societal influences include schools, communities, faith-based institutions (Roeser et al., 2018). Cultural influences on (expression of) compassion have also been noted (Koopmann-Holm et al., 2021).
Adolescence is a turbulent time as young people explore their identity and face societal challenges (e.g. education, future career), in the context of significant neurological restructuring. It is associated with increased vulnerability to mental health difficulties (Bluth & Clepper-Faith, 2023). Neff (2003a) proposed that self-compassion can decrease in adolescence due to increased self-reflection in relation to ideals and standards that may fuel self-criticism and feelings of isolation, which is supported by some studies (e.g. Bengtsson et al., 2016). Research also suggests gender differences, as adolescent females reported greater compassion for others and lower self-compassion than adolescent males (Bengtsson et al., 2016). Some research has shown age interacts with gender as older adolescent females report lower self-compassion than younger adolescent females and adolescent males (e.g. Bluth et al., 2017).
Whilst most compassion research has been conducted with adults (18 + years of age), research with CYP (defined as younger than 18 years of age) is growing and reflects findings with adults, suggesting self-compassion is important for the psychological and social wellbeing of CYP. For example, self-compassion in CYP has been associated with enhanced wellbeing, happiness, mindfulness, resilience, social connectedness, and mastery-oriented goals, in addition to reduced anxiety, depression, and performance-oriented goals (Marsh et al., 2018; Neff & McGehee, 2010; Neff et al., 2021). Self-compassion has also been shown to negatively correlate with peer victimization (Hatchel et al., 2019), and to moderate the relationship between peer victimization and self-harm (Jiang et al., 2016). Whilst research on other-compassion is more limited for CYP than adults, a recent study with adolescents showed positive associations with emotion regulation, wellbeing, and mindfulness (Heidary et al., 2022).
A key challenge in compassion research is lack of consensus in its definition and measurement (Gilbert & Van Gordon, 2023; Neff, 2023). The definition by Gilbert, noted above, has been operationalised in the Compassionate Engagement and Action Scales (CEAS) (Gilbert et al., 2017), which assess compassionate engagement and action in relation to self-compassion, compassion for others, and the ability to receive compassion from others, respectively. Neff’s (Neff, 2003b) Self-Compassion Scale (SCS), on the other hand, the most widely used measure of self-compassion (Neff, 2023), operationalises her definition and measures self-compassion across three overlapping bipolar domains (self-kindness vs self-judgment, common humanity vs isolation, mindfulness vs over-identification) that, together, capture a bipolar continuum from uncompassionate to compassionate self-responding when distressed (Neff, 2023). These differences in definition and measurement pose a significant challenge for compassion research (Fig. 1).
In response to this challenge, Strauss et al. (2016) consolidated eight existing, major, published conceptualizations and definitions in the literature (spanning Buddhist and Western psychological perspectives), including those by Neff and Gilbert cited above, into one multi-faceted definition of compassion as a cognitive, affective, and behavioral process consisting of five elements: (1) recognizing suffering; (2) understanding the universality of human suffering; (3) feeling empathy for the person suffering and connecting with the distress (emotional resonance); (4) tolerating uncomfortable feelings aroused in response to the suffering person (e.g., distress, anger, fear); (5) motivation to act/acting to alleviate suffering. While each element was captured across the eight published definitions, no existing definition explicitly captured all five elements. Gu et al. (2017) empirically examined the underlying conceptual structure of compassion by conducting a series of exploratory (Study 2) and confirmatory (Study 3) factor analyses with student samples, using items from existing self-report compassion measures reviewed for face validity by 15 mindfulness experts by training or experience, who also generated additional items to ensure a sufficient item pool for factor analysis (54 from four existing measures, 26 new items) that captured the five elements of the definition by Strauss et al. (2016) (Study 1). Findings supported the five-factor hierarchical structure of compassion, consistent with the definition by Strauss et al. (2016). This definition of compassion is therefore both grounded in theory and empirically supported and will be used throughout this review, to assess the content validity of measures of compassion developed and/or validated with CYP. Strauss et al. (2016) systematically reviewed nine self-report compassion measures for adults, finding that no measure comprehensively measured compassion with acceptable levels of reliability and validity. This highlighted a significant barrier to compassion research with adults and suggests similar issues may exist for measures for CYP.
Compassion research with CYP (i.e. younger than 18 years of age) is more limited than research with adults (18 years old or older), and there is particularly little research on other-compassion in CYP. This may be due to the limited validated, psychometrically robust self-report compassion measures for CYP (Neff et al., 2021) as research has mostly been conducted with older adolescents using measures developed for adults (Self-Compassion Scale [SCS]; Neff, 2003b; SCS-Short Form; Raes et al., 2011) (Neff et al., 2021). These self-report measures may use language that is too complex for CYP and/or may not capture the nuances of how CYP experience and express self-compassion and compassion for others and are therefore not necessarily suitable for assessing compassion in CYP. Measures that do not fully capture the nature of compassion, as defined by Strauss et al. (2016), and that have other psychometric weaknesses are likely to hinder research findings, thereby limiting advancements in research and practice with CYP (e.g. limiting the capacity to accurately evaluate the effectiveness of compassion-based interventions for CYP).
Meta-analyses of compassion-based intervention (CBI) randomized controlled trials (RCTs) in adults indicated that compassion is a skill that potentially can be cultivated (Ferrari et al., 2019; Kirby et al., 2017), although such findings are limited by the psychometric weaknesses of compassion measures already noted. Moreover, these meta-analyses highlighted the lack of RCTs assessing CBIs in CYP. RCTs evaluating CBIs with CYP are therefore needed, and this should include examination of potential mechanisms of action, including compassion. This further highlights the need for valid and reliable compassion measures suitable for CYP, to advance our understanding of the effectiveness and mechanisms of change of CBIs.
To the authors’ knowledge, no systematic review to date has examined the psychometric properties of self-report compassion measures developed and/or validated with CYP. This systematic review, therefore, aimed to describe and critically appraise the psychometric properties of self-report self- and other-compassion measures developed and/or validated with CYP, to facilitate understanding of their suitability for use in research and clinical practice and to identify areas for future research.
Method
Eligibility Criteria
To be selected, papers had to meet the following inclusion criteria: (i) peer-reviewed article, (ii) published in English, (iii) psychometric paper outlining the development and/or validation of a self-report measure, (iv) description of the measure is related to compassion or self-compassion, (v) majority of participants under 18. The following exclusion criteria were applied: (i) measure did not specifically assess global compassion (e.g., body compassion), (ii) exclusively adult sample, (iii) unclear if the majority of participants were under 18, (iv) non-questionnaire measure of compassion, or subscale used. Therefore, to be selected, (one of) the aim(s) of the paper had to develop and/or validate a compassion measure for CYP. In line with best practice guidance by Boateng et al. (2018), papers that validated existing self-report measures with CYP were expected to include evaluation of factor structure, (aspects of) reliability and validity. Papers that included a compassion measure, e.g. to assess the association between compassion in CYP and other variables, but did not aim to evaluate its measurement properties and/or solely reported internal consistency, were excluded.
Search Strategy
Electronic databases Medline, Psychinfo and Web of Science were searched from inception to 31 st July 2024 to identify relevant papers. Abstracts were searched using the following search terms: (compassion* OR self-compassion*) AND (measure* OR scale* OR questionnaire) AND (“young people” OR youth OR child* OR adolesce* OR teen*) AND (valid* OR reliab* or psychometric* OR “factor analys*”).
All records from the electronic search were imported to RefWorks (n.d.). After removing duplicates, titles and abstracts of remaining records were screened against the eligibility criteria. If eligibility was not clear from the title/abstract, the full-text paper was sought for retrieval and, if retrievable, read to determine eligibility. The reference lists of selected papers were hand-searched when a paper referred to additional scales and potentially relevant papers were retrieved and screened as set out above. The search and paper identification conducted by the first author were independently replicated by the second author for 80% of records, using the provided search string across the three electronic databases set out above. Records were screened against the eligibility criteria, discussed beforehand to ensure shared understanding. There was 100% agreement on selected papers across the electronic and hand search. Therefore, no reconciliation of differences was required.
Assessment of Quality
The quality of measures was reviewed and rated based on adapted quality criteria for health status measures by Terwee et al. (2007) and Barker et al. (2002) recommendations for psychological measures (Strauss et al., 2016). In addition to Terwee et al. (2007), the factor structure (i.e., structural validity) was also evaluated. The following measurement properties were evaluated:
(1)
Content validity: The extent to which compassion was comprehensively sampled by questionnaire items. This was evaluated against the five-element definition of compassion put forward by Strauss et al. (2016) (i.e. recognising suffering (R), understanding the universality of suffering (U), Feeling for the person suffering (F), tolerating uncomfortable feelings (T), acting/being motivated to act to alleviate suffering (A)), rather than the definition by the measure’s authors. The requirement for both experts by experience and training to be consulted on the items (see Table 1) reflects that expert consultation can enhance the robustness of content validity and ensure that items adequately capture all facets of compassion (Boateng et al., 2018).
(2)
Factor structure (structural validity): Whether the factor structure proposed by the authors of the original measure, reflecting how items are combined into (sub-scale and/or total), was examined and supported.
(3)
Internal consistency: The extent to which items in a (sub)scale are inter-correlated, thereby measuring the same construct.
(4)
Test–retest reliability: The extent to which participants' performance is repeatable, i.e., how consistent scores are across time.
(5)
Construct validity: The extent to which scores on the compassion measure relate to other measures consistent with theoretically derived hypotheses about the constructs being measured.
(6)
Floor and ceiling effects: The number of respondents obtaining the highest or lowest possible scores. The presence of floor or ceiling effects indicates limited content validity (Terwee et al., 2007).
(7)
Interpretability: The extent to which qualitative meaning can be attached to quantitative scores or how differences in scores on the measure can be interpreted.
Terwee et al. (2007) also provide criteria for criterion validity and responsiveness.
Table 1
Quality criteria for rating measures
Property
Definition
Quality Criteria
1. Content validity
The extent to which the domain of interest is comprehensively sampled by questionnaire items. Strauss et al.’s (2016) definition of compassion was considered, rather than the author of the measure’s definition. Components of compassion are referred to as follows:
Recognising (R) = recognising suffering
Universality (U) = understanding the universality of suffering Feeling (F) = feeling for the person suffering
2: Five elements of compassion captured by items AND items generated in consultation with both experts and young people;
1: Five elements of compassion captured by items OR items generated in consultation with both experts and young people;
0: Five elements of compassion not captured AND items not generated in consultation with both experts and young people
2. Factor structure
Whether the factor structure for the compassion measure is examined and supported
These criteria were additional to those proposed by Terwee et al. (2007) and based on the criterion proposed by Strauss et al. (2016). Exploratory structural equation modelling (ESEM) and Principal Component Analysis (PCA) were also added to these criteria to aid rating. The bifactor-ESEM framework has been developed to provide more precise psychometric examination of measures (Morin et al., 2016) and is a combination of Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA) (Morin et al., 2013). PCA has a greater emphasis on data reduction than interpretation and EFA and CFA are therefore recommended to examine the factor structure (Alavi et al., 2020)
2: EFA followed by CFA and conducted in independent samples OR CFA
conducted if factor structure was theoretically proposed previously OR ESEM AND factor analyses support the proposed factor structure;
1: EFA conducted without CFA and EFA supports the factor structure;
0: Factor analysis not conducted OR conducted but does not support the proposed factor structure OR PCA conducted
3. Internal consistency
The extent to which items in a (sub)scale are inter-correlated, thereby measuring the same construct
2: Factor analyses conducted on an adequate sample size (7 * number of items AND N > 100) AND Cronbach’s alpha for each identified factor between 0.70 and 0.95;
1: Factor analyses not conducted on adequate sample size OR Cronbach’s alpha for each identified factor <.70 OR > 0.95;
0: No information on internal consistency
George and Mallery’s (2003) rules of thumb for internal consistency are used when describing internal consistency in text:
The extent to which participants'performance is repeatable, i.e., how consistent scores are across time
These criteria were based on Terwee et al. (2007). However, as these criteria did not include Pearson’s r, Barker et al.’s (2002) guidance was also drawn on
2: ICC OR weighted kappa ≥ 0.70;
1: ICC OR weighted kappa < 0.70 OR r ≥ 0.70;
0:r < 0.70 or no information on test–retest reliability
5. Construct validity
The extent to which scores on a measure relate to other measures in a way that is consistent with theoretically derived hypotheses about the constructs being measured
These criteria were based on Terwee et al. (2007). However, as Terwee et al. (2007) do not take into account the strength of these correlations, Barker et al’s (2002) guidance was drawn on in relation to this
2: Specific hypotheses given AND at least 75% of results in line with these hypotheses AND a minimum of two correlations of r ≥ 0.50 to demonstrate convergent validity;
1: Specific hypotheses given AND less than 75% of results in line with these hypotheses AND/OR less than two correlations of r ≥ 0.50 to demonstrate convergent validity OR other clear methodological weakness;
0: No specific hypotheses given OR no information on construct validity
Cohen’s (1988) benchmarks for effect sizes are used when describing the strength of correlations in text:
r = 0.1–0.3 were considered small, r = 0.3–0.5 were considered medium, r = over 0.5 were considered large
6. Floor/ceiling effects
The number of respondents obtaining the highest or lowest possible scores
These criteria were based on Terwee et al. (2007) and adapted to be more specific regarding whether floor/ceiling effects were specifically examined to aid rating
2: Floor/ceiling effects specifically referenced AND ≤ 15% of the respondents achieved the highest or lowest possible scores;
1: Floor/ceiling effects specifically referenced AND > 15% of the respondents achieved the highest or lowest possible scores OR floor/ceiling effects not specifically referenced but data indicates ≤ 15% of the respondents achieved the highest or lowest possible scores;
0: No information on floor/ceiling effects
7. Interpretability
The extent to which qualitative meaning can be attached to quantitative scores or how differences in scores on the measure can be interpreted
These criteria were based on Terwee et al. (2007) with some adaptations in line with Strauss et al. (2016). Terwee et al. (2007) require minimal important change to be defined. However, as this was not considered relevant to the measures in this review, consideration was instead given to how scale scores might be interpreted
2: Mean and standard deviation scores provided for at least four relevant subgroups AND indication of how scale scores might be interpreted;
1: Mean and standard deviation scores provided for less than four relevant subgroups OR no indication of how scale scores might be interpreted provided;
0: No information on interpretability
However, criterion validity cannot be assessed in the absence of a gold-standard compassion measure. Furthermore, data was typically not available for responsiveness. Terwee et al. (2007) suggests assigning positive (+), indeterminate (?), or negative (−) ratings, or a rating of 0 where no information regarding the relevant criteria is provided. However, to make the findings of this review more directly comparable to the review of adult measures of compassion (Strauss et al., 2016), measures were given a rating of 2 if there was evidence for the criterion being fully met, 1 if the criterion was partially met, and 0 if the criterion was not met or if no relevant data was provided, Quality criteria are outlined in Table 1.
Quality ratings were then collated to give each included measure an overall rating. The total possible rating was 14. Where multiple papers evaluated the same measure, the following scoring criteria were applied when calculating the overall quality rating: a score of 2 if at least one study provided evidence that met the 2 criterion and no studies met a 1 or a 0 criterion that indicated evidence against the criterion (i.e. any 1 or 0 ratings would be for lack of evidence, not evidence against the criterion); a score of 1 if at least one study provided evidence that met the 1 criterion and no studies met a 0 criterion that indicated significant evidence against the criterion (i.e. not simply indicating lack of evidence), or where the “2” criterion above was not met and the criterion for “0” below was also not met; a score of 0 if any study met a 0 criterion indicating significant evidence against or all studies met a 0 criterion indicating lack of evidence.
Two authors independently rated the measures against these criteria, following initial discussion of the (papers on) these criteria to ensure shared understanding; any discrepancies were resolved collectively. Cohen’s kappa, calculated across the 19 selected papers and 7 properties (i.e. 133 ratings in total, ratings 0–2), was 0.94 (SE = 0.03, 95% CI [0.89, 0.99]), indicating almost perfect agreement (Landis & Koch, 1977). Where applicable, ratings across papers (i.e. for the SCS, SCS-Y, CS-C) were summated independently by the first and second author and subsequently discussed to reach consensus.
For papers involving translation, translation processes were evaluated broadly in line with WHO (Ustun et al., 2010) and Sousa and Rojjanasrirat (2010) best practice guidance, which include having a bilingual and monolingual (group of) translator(s), overseen by (a) bilingual subject-field expert(s) closely familiar with the measure and associated construct: (1) forward translation by a qualified translator(s) familiar with the construct and measure, supported by the bilingual experts. The aim is conceptual rather than word-for-word translation; (2) review of the translation by bilingual experts, followed by review of the translation by the monolingual group (speaking the target language). Discussion and potential redraft of the translation by the bilingual group; (3) Back-translation by independent professional translator(s) not involved in forward-translation who was blind to the original version. The bilingual group compares back-translation to the original for conceptual equivalence, resolving any differences, and; (4) pre-testing and cognitive interviewing—at least 10 respondents from the target population were invited to review the wording of the questionnaire.
Results
Selected Studies
After removing duplicates (n = 110), titles and abstracts of 149 remaining records were screened against the eligibility criteria. Subsequently, the full text of 26 papers was read to determine eligibility, resulting in 18 selected papers. The hand search resulted in one additional paper Fig. 1. The 19 studies selected for the review examined eight compassion measures. Five measured self-compassion, three (also) measured other-compassion, one (also) measured receiving compassion from others, and one (also) measured other-compassion for other living things. Three papers developed a compassion measure for CYP (Self-Compassion-Scale-Youth (SCS-Y), Compassion Scale – Child (CS-C), Self-Compassionate Reactions Scale – Child (SCRS-C)), six papers examined (adapted (parts of)) the adult SCS (n = 5) and CEAS (n = 1) for use with CYP, nine papers translated (and adapted) the SCS (n = 2), SCS-Y (n = 6) and CS-S (n = 1), and one paper adapted an existing Portuguese version of an adult measure of compassion (Compassion Scale – Adolescence; Sousa et al., 2022). See Table S1 in the Supplementary Information for an overview.
See Table S2 for an overview of participant demographic information. Measure quality ratings are shown in Table 2. Tables 3, 4 and 5 provide an overview of the psychometric properties outlined in each paper. Each paper and measure are described in greater detail below. See Table S3 for quality ratings for the individual papers. Table S4 provides a brief overview of translation processes employed in nine papers, information also summarised under evaluation of ‘content validity’ of a measure, where applicable.
ns non-significant, NR Not reported, CEAS-A Compassionate Engagement and Actions Scales – Adolescence, CS-A Compassion Scale – Adolescence, CS-C Compassion Scale – Child, SCS Self-Compassion Scale, SCS-C Self-Compassion Scale – Child, SCS-Y Self-Compassion Scale – Youth, SCRS-C Self-Compassionate Reactions Scale – Child, S-SCS-A Shortened Self-Compassion Scale – Adolescence, R recognising suffering, U understanding the universality of suffering, F feeling for the person suffering, T tolerating uncomfortable feelings, A acting or being motivated to act to alleviate suffering (elements of compassion captured from the definition by Strauss et al., 2016)
Table 4
Overview of psychometric properties (test–retest reliability, floor/ceiling effects and interpretability)
Measure
Authors
Test–retest reliability: r or ICC (time between testing)
Floor and ceiling effects
Interpretability:
- subgroups tested for differences
-mean and standard deviation provided for number of subgroups
Grade: Trend for self-compassion to decrease in higher grades but not significant (Study 1 and 2). No
difference (Study 4)
Gender: No difference (Study 1). Males had higher levels of self-compassion (Study 2 and 4) but not significant (study 2). (Study 4: males: M = 3.18, SD = 0.54; females
M = 2.99, SD = 0.68)
Gender interaction: No interaction between gender and age, or gender and grade (Study 1, 2 and
4). However, trend for males to score slightly higher than females, and for females to slightly decrease in self- compassion with age (study 2)
Gender: Female participants in community sample had higher levels of compassion (M = 4.28, SD = 0.47) when compared to male participants (M = 3.93, SD = 0.56). Female participants scored higher on all subscales, except indifference, where male participants had higher levels
Kindness: males: M = 3.74, SD = 0.82; females: M = 4.23, SD = 0.69
Common humanity: males: M = 4.18, SD = 0.71; females: M = 4.40, SD = 0.63
Mindfulness: males: M = 3.94, SD = 0.63; females: M = 4.27, SD = 0.55
Indifference: males: M = 2.17, SD = 0.76; females: M = 1.76, SD = 0.69
Community vs forensic sample: Male participants in the forensic sample had lower levels of compassion towards others (M = 3.77, SD = 0.67) than male participants in the community sample (M = 3.93, SD = 0.56). They also had lower levels of mindfulness (M = 3.67, SD = 0.82 vs community sample male participants: M = 3.94, SD = 0.63)
Number of diagnoses: Indifference subscale associated with
Gender: No differences for total sample. Female participants (M = 27.97, SD = 6.84) had lower levels of self- compassion than male participants (M = 31.54, SD = 6.11), and for self-kindness subscale (males: M = 10.29,
SD = 2.80; females: M = 8.89, SD = 2.76) in older adolescent sample (15–17 years)
ns non-significant, NR Not reported, CEAS-A Compassionate Engagement and Actions Scales – Adolescence, CS-A Compassion Scale – Adolescence, CS-C Compassion Scale – Child, SCS Self-Compassion Scale, SCS-C Self-Compassion Scale – Child, SCS-Y Self-Compassion Scale – Youth, SCRS-C Self-Compassionate Reactions Scale – Child, S-SCS-A Shortened Self-Compassion Scale – Adolescence, R recognising suffering, U understanding the universality of suffering, F feeling for the person suffering, T tolerating uncomfortable feelings, A acting or being motivated to act to alleviate suffering (elements of compassion captured from the definition by Strauss et al., 2016)
Table 5
Overview of psychometric properties (construct validity )
Measure
Authors
Construct validity:
correlation (Pearson’s r) of compassion measure with related constructs
CEAS-A self-compassion (SC) total FSCSRS: r = 0.42 (self-reassurance), r = − 0.17 (self-criticism); SCS: r = 0.60 (positive), r = − 0.20 (negative); SLSS: r = 0.36
CEAS-A compassion for others (CFO) total
FSCSRS: r = 0.40 (self-reassurance), r = 0.05 ns (self-criticism); SCS: r =.39 (positive), r = 0.05 ns (negative); SLSS: r = 0.24
CEAS-A compassion from others (CfO) total
FSCSRS: r = 0.41 (self-reassurance), r = − 0.11 (self-criticism); SCS: r = 0.59 (positive), r = − 0.07 (negative); SLSS: r = 0.36
STAIC: r = − 0.38; CDI: r = − 0.53; UCL-A: r = 0.50 (active tackling), r = 0.07 ns (palliative reacting), r = − 0.07 ns (avoidance), r = 0.42 (social support seeking), r = − 0.47 (passive reacting), r = − 0.29 (expression of emotion), r = 0.41(reassuring thoughts)
SCS self-judgement
STAIC: r = 0.61; CDI: r = 0.58; UCL-A
A: r = − 0.25 (active tackling), r = 0.18 (palliative reacting), r = 0.19 (avoidance), r = − 0.25 (social support seeking), r = 0.56 (passive reacting), r = 0.25 (expression of emotion), r = − 0.11 (reassuring thoughts)
SCS common humanity
STAIC: r = − 0.19; CDI: r = − 0.32; UCL-A: r = 0.37 (active tackling), r = 0.13 ns (palliative reacting), r = 0.00 ns (avoidance), r = 0.22 (social support seeking), r = − 0.33 (passive reacting), r = − 0.23 (expression of emotion), r = 0.47 (reassuring thoughts)
SCS isolation
STAIC: r = 0.58; CDI: r = 0.53; UCL- A: r = − 0.25 (active tackling), r =—0.02 ns (palliative reacting), r = 0.22 (avoidance), r = − 0.28 (social support seeking), r = 0.53 (passive reacting), r = 0.17 ns (expression of emotion), r = − 0.09 ns (reassuring thoughts)
SCS mindfulness
STAIC: r = − 0.25; CDI: r = − 0.38; UCL-A: r = 0.48 (active tackling), r = 0.22 (palliative reacting), r = − 0.09 ns (avoidance), r = 0.30 (social support seeking), r = − 0.36 (passive reacting), r = − 0.21 (expression of emotion), r = 0.35 (reassuring thoughts)
SCS overidentification
STAIC: r = 0.58; CDI: r = 0.49; UCL-A: r = − 0.25 (active tackling), r =—0.06 ns (palliative reacting), r = 0.09 ns (avoidance), r = − 0.08 ns (social support seeking), r = 0.56 (passive reacting), r = 0.26 (expression of emotion), r =
SWLS: r = 0.30; BRS: r = 0.42; SPANE: r = − 0.35 (negative), r = 0.42 (positive); STAIC: r = − 0.19 (state), r = − 0.44 (trait); SHS: r = 0.36; DASS: r = − 0.27 (depression), r = − 0.23 (stress)
SCS-Y self-kindness
SWLS: r = 0.35; BRS: r = 0.27; SPANE: r = − 0.29 (negative), r = 0.38 (positive); STAIC: r = − 0.05 ns (state), r = − 0.25 (trait); SHS: r = 0.32; DASS: r = − 0.11 ns (depression), r =—0.08 ns (stress)
SCS-Y self-judgment
SWLS: r = − 0.16; BRS: r = − 0.23 SPANE: r = 0.22 (negative), r = − 0.21 (positive); STAIC: r = 0.18 (state), r = 0.30 (trait); SHS: r = − 0.15; DASS: r = 0.19 (depression), r = 0.21 (stress)
SCS-Y common humanity
SWLS: r = 0.01 ns; BRS: r = − 0.10 ns; SPANE: r = 0.04 ns (negative), r =.12 ns (positive); STAIC: r = 0.21 (state), r = 0.09 ns (trait); SHS: r = 0.06 ns; DASS: r = 0.12 ns (depression), r = 0.15 (stress)
SCS-Y isolation
SWLS: r = − 0.14 ns; BRS: r = − 0.31; SPANE: r = 0.25 (negative), r = -
0.11 ns (positive); STAIC: r = 0.25 (state), r = 0.36 (trait); SHS: r = − 0.17; DASS: r = 0.30 (depression), r = 0.25 (stress)
SCS-Y mindfulness
SWLS: r = 0.27; BRS: r = 0.24; SPANE: r = − 0.19 (negative), r = 0.39 (positive); STAIC: r = − 0.12 ns (state), r = − 0.24 (trait); SHS: r = 0.31; DASS: r = − 0.13 ns (depression), r =—0.13 ns (stress)
SCS-Y overidentification
SWLS: r = − 0.04 ns; BRS: r = − 0.36; SPANE: r = 0.19 (negative), r = -
0.14 ns (positive); STAIC: r = 0.21 (state), r = 0.32 (trait); SHS: r = -
0.12 ns; DASS: r = 0.24 (depression), r = 0.22 (stress)
Reported to be positively related to resilience and well-being, and negatively related to depression but examined by network analysis, and r values not given
Item discrimination values for all SCS-Y items ranged
between 1.70 and 2.74, suggesting a very high discrimination
ability for the items. All subscales reported an average variance extracted (AVE) higher than 0.50, corresponding to an acceptable discriminant validity
(general), r = 0.39 (school); RI: r = 0.45; SWLS-C: r = 0.40; PANAS-C: r = 0.42 (positive), r = − 0.09 ns (negative); SPQ: r = − 0.13 (anxiety), r = − 0.22 (depression); IRP: r = 0.42 (empathetic concern), r = 0.54 (perspective taking); SGQ: r = 0.60
SCS-C negative
MAAS-C: r = − 0.41; SDQ: r = − 0.10 ns (general), r = − 0.05 ns (school); RI: r = − 0.39; SWLS-C: r = − 0.25; PANAS- C: r = − 0.16 (positive), r = 0.46 (negative); SPQ: r = 0.51 (anxiety), r = 0.43(depression); IRP: r = 0.21
(empathetic concern), r = 0.08 ns (perspective taking); SGQ: r = 0.09 ns
r = 0.47 (SAS), r = 0.28 (MAAS-C), r = 0.60 (RSES), r = 0.03 (DDDT)
SCRS-C compassionate reaction r = 0.11 ns (SAS), r = 0.06 ns (MAAS- C), r = 0.28 (RSES), r = 0.06 ns (DDDT)
SCRS-C uncompassionate reaction r = − 0.56 (SAS), r = − 0.34 (MAAS-C), r = − 0.58 (RSES), r = 0.01 ns (DDDT)
ns non-significant, NR Not reported, CEAS-A Compassionate Engagement and Actions Scales – Adolescence, CS-A Compassion Scale – Adolescence, CS-C Compassion Scale – Childm SCS Self-Compassion Scale, SCS-C Self-Compassion Scale – Child, SCS-Y Self-Compassion Scale – Youth, SCRS-C Self-Compassionate Reactions Scale – Child, S-SCS-A Shortened Self-Compassion Scale – Adolescence, BFI- 10 Brief 10‑Item Big Five Inventory (Rammstedt & John, 2007), BRS Brief Resilience Scale (Smith et al., 2008), BYI Beck Youth Inventories (BYI) (Beck et al., 2001), CAMM Child and Adolescent Mindfulness Measure (Greco et al., 2011), CDEQ-SC Children’s Depressive Experiences Questionnaire – self-criticism subscale (Abela & Taxel, 2001), CDI Children’s Depression Inventory (Kovacs, 1981), CD-RISC- 10 The 10-item Connor-Davidson Resilience Scale (Campbell-Sills & Stein, 2007) CES-DC The Center for Epidemiological Studies Depression Scale for Children (Faulstich et al., 1986), DASS Depression Anxiety Stress Scales: (Stalikas & Flora, 2012), DASS- 21 Depression, Anxiety and Stress Scales (Lovibond & Lovibond, 1995, Portuguese version: Pais-Ribeiro et al., 2004), DDDT Dirty Dozen Dark Triad (Jonason & Webster, 2010), DERS‑SF Difficulties in Emotion Regulation Scale – Short Form (Kaufman et al., 2016), EMWSS Early Memories of Warmth and Safeness Scale (Richter et al., 2009, Portuguese version for adolescents: Cunha et al., 2014), ERQ-CA Emotion Regulation Questionnaire-Children and Adolescents (Gullone & Taffe, 2012), FSCRS The Forms of Self-Criticizing/Attacking & Self-Reassuring Scale (Castilho & Pinto-Gouveia, 2011), FSCSR The Forms of Self-Criticizing and Self-Reassuring Scale (Gilbert et al., 2004, Portuguese version: Silva & Salvador, 2010), GAD- 7 Generalized Anxiety Disorder Scale (Spitzer et al., 2006), ICQ− 15 15-item Interpersonal Competence Questionnaire (Coroiu et al., 2015), ILK Inventory for measuring quality of life in children (Mattejat & Remschmidt, 2006) IRP Interpersonal Reactivity Index (Davis, 1983, Oberle et al., 2010), MAAS-C Mindful Attention Awareness Scale – Children (Lawlor et al., 2014), OASB-A Other as Shamer Scale Brief – Adolescent version (Cunha et al., 2015), PALS Patterns of Adaptive Learning Scales (Midgley et al., 1998), PANAS-C Positive and Negative Affect Schedule for Children (Laurent et al., 1999), PHQ- 2 Patient Health Questionnaire (Kroenke et al., 2003), PHQ- 9 Patient Health Questionnaire- 9 (PHQ- 9) Kroenke et al., 2001), PROMIS Patient Reported Outcome Measurements Information System (PROMIS) (Cella et al., 2007), RBDI Revised Beck Depression Inventory (Finnish modification, Kaltiala-Heino et al., 1999), RCADS-R Revised Child Anxiety and Depression Scale (RCADS) RCADS long scale (Chorpita et al., 2005) RI Resiliency Inventory – optimism subscale (Song, 2003), RSES 10-item Rosenberg Self-Esteem Scale (Chinese version) (Rosenberg, 1965), RSPWB Ryff Scale Psychological Wellbeing (Khanjani et al., 2014), SAS Self-Acceptance Scale (Cong & Gao, 1999), SCARED Screen for Child Anxiety Related Emotional Disorders (Birmaher et al., 1997), SCS Self-Compassion Scale (Cunha et al., 2016), SDQ Self-Description Questionnaire (Marsh, 1988), SDQ Strength and Difficulties Questionnaire (Goodman, 2001) SEQ Self-Esteem Questionnaire (Rosenberg, 1965), SEQ-C Self-Efficacy Questionnaire for Children (Muris, 2001), SGQ Social Goals Questionnaire – prosocial goals subscale (Wentzel, 1993), SHS Subjective Happiness Scale (Lyubomirsky & Lepper, 1999), SLSS Students’ Life Satisfaction Scale (Marques et al., 2007), SPANE Scale of Positive and Negative Experience (Diener et al., 2010), SPPC Self-Perception Profile for Children (Harter, 1985), SPQ Seattle Personality Questionnaire for Young School-Aged Children – anxious and depressive symptoms subscales (Kusche et al., 1988), STAI State-Trait Anxiety Inventory, STAIC State-Trait Anxiety Inventory for Children (Spielberger, 1973), SWLS Life Satisfaction Scale (Stalikas & Lakioti, 2012), SWLS-C Satisfaction with Life Scale – Children (Gadermann, 2009), UCL-A The Utrecht Coping List for Adolescents (Bijstra et al., 1994), WHO- 5 5-item World Health Organization Well-being Index (Blom et al., 2012), YLOT Youth Life Orientation Test (Ey et al., 2005), YSR-SF Youth Self Report – short form (Cruz et al., 2014)
Compassionate Engagement and Action Scales–Adolescence (CEAS-A)
The CEAS is three self-report scales, assessing self-compassion (SC), compassion for others (CFO), and the ability to receive compassion from others (CfO), respectively. Each has two sections: (1) compassionate engagement (8 items); and (2) compassion action (5 items). The CEAS (and CEAS-A) use a 10-point response scale (ranging from never to always).
Population
The main sample was 674 school students in Portugal, aged 12–19 (M = 14.88, 61% female). Construct validity was assessed with a subsample of 336 participants (M = 15.48, 73% female). Test–retest reliability was assessed with 76 participants (M = 15.87, 89% female).
Content Validity
Cunha et al. (2023) adapted the adult version of the (Portuguese version of the) CEAS (Gilbert et al., 2017) to use with young people. Items were discussed with a group of adolescents (n = 18) and revised to make them more understandable and easier for adolescent use. However, experts were not consulted. Items related to 4/5 elements of compassion (not universality).
Factor Structure and Reliability
CFA supported the factor structure of the CEAS-A that was found for the adult version, including two distinct processes: 1) engagement with suffering, 2) an action component to alleviate or prevent suffering. The three-order factor model was a very good fit. The higher-order factor compassion for self, included two second-order factors: the engagement and actions subscales. The engagement subscale comprised the sensitivity to suffering and engagement with suffering dimensions. However, the authors noted that the model fit resulted from error term correlations between two pairs of items of the actions subscale, which they suggested might be related to similar phrasing in the Portuguese version of Items 4 and 5. For the CFO and CfO scales, a higher-order factor encompassed two first-order factors: the engagement and actions subscales. The model presented a good fit to the data, after removal of item 4 on the engagement subscale, for CFO. Cunha et al. (2023) found good–excellent internal consistency for the total scales (α = 0.84 to 0.94), acceptable-excellent internal consistency for subscales (α = 0.70 to 0.92) and excellent test–retest reliability for the three total scales (ICC = 0.97 to 0.98).
Construct Validity and Interpretability
Congruent with hypotheses, the CEAS-A scales were positively correlated with one another, and had positive associations with self-reassurance, life satisfaction, and the self-compassion scale positive factor, with small-large effects. In relation to discriminant validity, Cunha et al. (2023) reported that correlations between the CEAS-A scales and negative variables were weaker than positive variables. SC and CfO scales had negative associations with the self-compassion negative factor and self-criticism, with small effects. Ten correlations were r ≥ 0.50.
Means and standard deviations were reported for the total sample, males, and females for total and subscale scores. Subgroup analyses were conducted. Females scores higher than males on the CFO scale, and lower than males on the SC scale. Floor and ceiling effects were not examined.
Summary
The CEAS-A was rated 9/14 for quality. Whilst respondents were consulted, experts were not, and the’ understanding the universality of suffering’ element of compassion was not captured by items. Experts could be engaged to develop and refine items related to ‘universality of suffering’. CFA supported the proposed factor structure, and internal consistency and test–retest reliability were good. The error term correlations identified in the ‘actions’ subscale items should have been addressed; these items should have been revised to improve model fit without compromising conceptual integrity. There was evidence for construct validity and interpretability was facilitated. Floor and ceiling effects were not assessed.
Self-Compassion Scale (SCS)
The SCS was developed by Neff (2003b) to assess self-compassion in adults and is the most widely used measure of self-compassion (Neff et al., 2019). It consists of 26 items on a 5-point response scale (ranging from almost never to almost always) and has six subscales (self-kindness, mindfulness, common humanity, self-judgment, isolation, over- identification). Five papers examined the psychometric properties of the SCS (Cunha et al., 2016; Gruber et al., 2023; Kumlander et al., 2018; Muris et al., 2018; Stolow et al., 2016).
Population
The sample in Cunha et al (2016) included 3165 CYP aged 12–19 from Portugal (M = 15.49, 53.8% female) and in Stolow et al. (2016) included 193 CYP aged 9–16 from the US (M = 13.00, 59% female, 58% Caucasian). The sample in Kumlander et al. (2018) included 1725 adolescents from Finland, with 75% aged 15–17 (M = 16.56, 50.3% female). Muris et al. (2018) recruited 130 CYP aged 15–19 from the Netherlands (M = 16.68, 66% female). Gruber et al. (2023) evaluated the SCS with a German community sample of 10–19-year-olds (M = 14.9, SD = 2.6, 59.2% female). No papers reported SES, and only Stolow et al. (2016) reported ethnicity.
Content Validity
Cunha et al. (2016) used the Portuguese version (Pinto-Gouveia, 2011) of Neff’s (2003b) SCS, adapted for adolescents (Cunha et al., 2013). A group of adolescents were consulted on the items but independent experts were not. Stolow et al. (2016) used a revised child-suitable version of the SCS, stating original meaning was maintained whilst ensuring items could be understood by younger children. However, no reference was provided, and it was not possible to assess the items used. Gruber et al. (2023) translated the child-revised version of the SCS by Stolow et al. (2016) to German, using an independent translator for backward translation and consulting independent mindfulness and compassion experts. The SCS was translated from English to Finish by Kumlander et al. (2018), by the research group and an independent professional translator. A further independent translator did the backward translation; no independent bilingual experts were consulted. No adaptations were reported by Muris (2017). None of the papers consulted both experts and CYP . Items related to 4/5 elements of compassion (not recognising). Whilst it was not possible to assess the items used in two papers, the aforementioned was assumed.
Factor Structure and Reliability
Cunha et al. (2016) conducted CFA, which supported the use of an overall self-compassion score and six subscale scores, congruent with the adult SCS (Neff, 2003b) and Neff’s (2003a) conceptualisation of compassion (also see Neff et al., 2019).
Stolow et al. (2016) conducted Principal Component Analysis (PCA), yielding two factors. The SCS-positive factor comprised of items from the positive subscales and the SCS-negative factor of items from the negative subscales. Neither EFA nor CFA were conducted. Kumlander et al. (2018) conducted CFA, which supported the six-factor model and two-factor model. However, high correlations among the three positive and three negative factors for the six-factor model showed strong multicollinearity, questioning the extent to which the six factors measured separate constructs. Therefore, use of two factors (self-compassion and self-coldness) was supported, and use of six subscales or total self-compassion score was not recommended. Findings were replicated in a second sample. Muris et al. (2018) did not conduct CFA but conducted joint PCA on all questionnaire data (including convergent validity measures). Two factors emerged. The positive subscales loaded on a factor also composed of adaptive coping styles, and the negative subscales loaded on a factor with scales measuring anxiety and depression and maladaptive coping styles. Whilst positive subscales were considered representative of self-compassion, they concluded negative subscales should be removed as they are indicators of vulnerability and inflate the relationship between self-compassion and symptoms of mental health difficulties. Use of the total score including negative subscales was cautioned. Gruber et al. (2023), however, tested a two- (positive and negative scales, respectively), three- (the three bipolar components of self-compassion) and six-factor solution using CFA and the six-factor solution was the best fit.
Cunha et al. (2016) found good internal consistency for the total SCS (α = 0.88) and acceptable internal consistency for subscales (α = 0.70 to 0.79). Kumlander et al. (2018) found good internal consistency for the two factors (ω = 0.87, 0.91). Stolow et al. (2016) found good–excellent internal consistency for the total positive and negative subscales (α = 0.87, 0.92) for the total sample, and for each school grade (α = 0.82 to 0.93). Muris et al. (2018) found internal consistency was good for the total scale (α = 0.89) and questionable-good for subscales (α = 0.61 to 0.84). Gruber et al. (2023) reported acceptable internal consistency for subscales (α = 0.72–0.78), except for the mindfulness subscale (α = 0.48), and good internal consistency for the total scale (α = 0.89). No paper reported test–retest reliability.
Construct Validity and Interpretability
As hypothesized by Cunha et al. (2016), the SCS was positively associated with early memories of warmth and safeness, and negatively associated with depression, anxiety, and stress. These were medium effects, and one correlation was r ≥ 0.50. Hypotheses were not provided for subscales. Positive subscales were positively associated with early memories of warmth and safeness, and negatively associated with depression, anxiety (except common humanity), and stress. The inverse direction was found for negative subscales. Effects for subscales were small-medium. Stolow et al. (2016) did not specify hypotheses for construct validity. The negative subscale was positively associated with depression and self-criticism, and negatively associated with self-esteem. The positive subscale was negatively associated with depression and positively associated with self-esteem, and not associated with self-criticism. Effects were small-large, and three correlations were r ≥ 0.50. As hypothesized, Kumlander et al. (2018) found the self-compassion factor was negatively associated with depression, and the self-coldness factor was positively associated with depression. Positive and negative subscales were negatively and positively associated with depression, respectively. Effects were medium-large and four correlations were r ≥ 0.50.
Results from Muris et al. (2018) were largely as hypothesized. Positive subscales were positively associated with adaptive coping and negatively associated with anxiety and depression. Negative subscales were positively associated with anxiety and depression and less helpful coping strategies. Effects were small-large, and ten correlations were r ≥ 0.50. Gruber et al. (2023) did not offer hypotheses, but findings showed medium-large associations between self-compassion (SCS) and mindfulness, quality of life, internalizing and externalizing symptoms, except for the ‘common humanity’ and ‘mindfulness’ subscales, which showed a small-medium association with the other variables. Twenty correlations were r ≥ 0.50.
Cunha et al. (2016) provided means and standard deviations for total and subscale scores for the total, male and female samples. Subgroup analyses showed males had higher levels of self-compassion, self-kindness, and mindfulness than females. Females had higher levels of self-judgement, isolation, and over-identification than males. Descriptive statistics and percentiles indicated no floor and ceiling effects. Stolow et al. (2016) provided means and standard deviations for subscale scores for the total sample and by school grade and gender. Females and older participants scored higher on the negative subscale than males and younger participants, respectively. There were no gender differences for the positive subscale. Gruber et al. (2023) provided means and standard deviations for total and subscale scores for the total, male and female samples. They found that boys had significantly higher total SCS scores than women and girls reported significantly higher SCS overidentification and isolation scores than men. No other gender differences were found. Interpretability was not facilitated by Kumlander et al. (2018) or Muris et al. (2018). Cunha et al. (2016) provided data which indicated floor and ceiling effects were not present. Gruber et al. (2023) reported that data was not normally distributed (highest mean score for self-judgment subscale, lowest for the overidentification subscales) but normality tests alone cannot confirm floor or ceiling effects.
Summary
The SCS was rated 7/14 for quality. No papers consulted both experts and recipients and the ‘recognizing’ element of compassion was not captured. Evidence for the factor structure was mixed, and two papers did not conduct EFA/CFA. Internal consistency was questionable-excellent, and test–retest was not conducted. There was some evidence for construct validity, subgroup comparisons were conducted, and floor and ceiling effects were not explicitly assessed.
Given the mixed evidence for Neff’s proposed factor structure, it should be examined further across different cultural adaptations to establish a more consistent factor model, taking into consideration the recommendation of Neff et al. (2019) to use bifactor ESEM rather than CFA to capture the multidimensionality of the scale. The internal consistency of subscales with questionable reliability (e.g., subscales with α < 0.70) should be improved and/or implications for recommending use of subscale scores should be considered.
Self-Compassion Scale–Youth (SCS-Y)
Neff et al. (2021) developed and validated a youth version of the SCS. The SCS-Y assesses self-compassion and consists of 17 self-report items with a 5-point response scale (ranging from almost never to almost always). The SCS-Y has six subscales (self-kindness, self-judgment, common humanity, isolation, and over-identification). Six papers validated the SCS-Y (Karakasidou et al., 2021; Deniz et al., 2022; Nazari et al., 2022; Huang et al., 2022; Cheung et al., 2023; Henje et al., 2024).
Population
Neff et al. (2021) developed the SCS-Y with 279 adolescents aged 11–15 (M = 12.17, 56.7% female), cross-validated the factor structure in a second sample of 402 adolescents aged 11–15 (M = 12.43, 48.8% female), examined test re-test reliability in a subset of 102 adolescents aged 11–14 (M = 12.52, 51% female) and examined construct validity with 212 adolescents aged 11–14 (M = 12.18, 42.5% female). All were school samples in the US. Participants were reported to be from socioeconomically diverse schools and the majority identified as White. Karakasidou et al. (2021) validated the SCS-Y with 193 Greek CYP aged 8–14 (M = 11.69, 51.3% male, 100% White), reported to be from socioeconomically diverse schools. Deniz et al. (2022) validated the SCS-Y with 450 Turkish adolescents aged 11–15 (M = 13.09, 61.8% female) and Nazari et al. (2022) with 532 Persian adolescents aged 12–15 (M = 13.57, 50.8% female). Huang et al. (2022) validated the SCS with a Chinese student sample aged 10–17 (M = 12.90, SD = 1.84, 52.5% female) while Cheung et al. (2023) used a community sample of 2881 Chinese (age: M = 16.58, SD = 1.32, 56.8% female, 98% Asian) and 1104 UK (age: M = 16.42, SD = 1.30, 76.1% female, 70% White) 12–18-year-olds. Henje et al. (2024) validated the SCS-Y with Swedish students aged 15–20 (M = 17.07, SD = 1.26, 67% female, 90% born in Sweden), with the majority < 18 (n = 205).
Content Validity
Neff et al. (2021) developed potential items based on the authors’ expertise in cognitive development. A small number of early adolescents and middle school teachers were also consulted. Karakasidou et al. (2021) and Deniz et al. (2022) translated the SCS-Y to Greek and Turkish, respectively, but did not consult with independent bilingual experts or respondents. Nazari et al. (2022) reviewed translated items (Persian) with an expert and students from the target sample. Huang et al. (2022) and Cheung et al. (2023) translated the SCS-Y to Chinese. Both papers used independent back-translators and (a) bilingual expert(s). Huang et al. (2022) consulted 15 CYP to establish readability. Henje et al. (2024) translated the SCS-Y to Swedish; the process of forward translation was not clear; a professional translator conducted the back-translation. It was not reported whether experts were consulted. Although four adolescent girls were consulted, this falls short of the recommended minimum of 10 (Sousa & Rojjanasrirat, 2010). Items related to 4/5 elements of compassion (not recognizing).
Factor Structure and Reliability
Neff et al. (2021) conducted bifactor exploratory structural equation modelling (ESEM) which supported the use of an overall self-compassion score and six subscale scores and was cross-validated in a second sample. Support for this factor structure was provided by Nazari et al. (2022) (ESEM) and Deniz et al. (2022) (CFA). Whilst providing support for this structure, Deniz et al. (2022) also found support for other factor structures and found invariance in all models. Karakasidou et al. (2021) did not support the factor structure reported by Neff et al. (2021), finding this model a poor fit, and instead supported a six-factor structure. Their sample was considerably smaller (n = 193) than for the other studies, which could be a contributing factor. Huang et al. (2022) tested nine models (CFA and ESEM) and found support for a global self-compassion factor and a six-factor (correlated) structure. Cheung et al. (2023) tested a single-, two- and six-factor model (CFA) and found that the six-factor model was the best fit. Henje et al. (2024) tested (two-)bi-factor models (six factors and one global factor and two global factors and six specific factors) using CFA and both proved a good fit.
Neff et al. (2021) found good internal consistency for the total scale (α = 0.82, 0.85), and acceptable internal consistency for five of six subscales (α = 0.70 to 0.80) in both studies. Internal consistency for overidentification (α = 0.66) and mindfulness (α = 0.67) were questionable for study 1 and 2, respectively. Deniz et al. (2022) reported acceptable internal consistency for the total scale (α = 0.79), while Karakasidou et al. (2021) found questionable internal consistency (α = 0.64). Neither reported internal consistency for subscales. Nazari et al. (2022) demonstrated good–excellent internal consistency for the total SCS-Y (α = 0.88) and subscales (α = 0.80 to 0.90). Huang et al. (2022) reported questionable internal consistency for self-judgment (α = 0.64) and common humanity (α = 0.69) and poor internal consistency for overidentification (α = 0.58), which they attributed to possible cultural differences. The remaining subscales and total scale had acceptable-good internal consistency. Cheung et al. (2023) reported acceptable to good internal consistency for the total and sub-scales (across the UK and Chinese samples). Henje et al. (2024) reported acceptable to good internal consistency for the total and sub-scales, apart from mindfulness (α = 0.69). Internal consistency was acceptable-good for girls. For boys, internal consistency was questionable for mindfulness (α = 0.62) and self-kindness (α = 0.68) but acceptable to good for the total and other subscales.
Neff et al. (2021) found good test–retest reliability for the total score (r = 0.83). In relation to subscales, two were adequate (self-kindness, self-judgment, r = 0.70, 0.71), three were questionable (common humanity, isolation, mindfulness, r = 0.63 to 0.65), and one was poor (overidentification, r = 0.51). Nazari et al. (2022) found adequate test–retest reliability for the total scale after 4 weeks (ICC = 0.60). Henje et al. (2024) reported adequate-good test–retest reliability for the total and sub-scales, apart from self-judgment (r = 0.68) and overidentification (r = 0.66), which were questionable. Test–retest reliability was not examined by Karakasidou et al. (2021), Deniz et al. (2022), Huang et al. (2022), and Cheung et al. (2023).
Construct Validity and Interpretability
Neff et al. (2021) found associations in line with hypotheses. Self-compassion was positively associated with happiness, life satisfaction, mindfulness, resilience, and mastery-approach goals, and negatively associated with depression, performance-avoidance goals, and performance-approach goals. Effects were small-large, and three correlations were r ≥ 0.50. At the subscale level, notable differences were a lack of significant association between the common humanity and mindfulness subscale and mindfulness and depression. Neff et al. (2021) attributes the lack of association between the mindfulness subscale and CAMM (mindfulness measure) to the subscale capturing balanced equanimity in relation to suffering rather than paying attention to present-moment experience more generally. Associations with performance avoidance and approach goals were nonsignificant for the positive subscales (self-kindness, common humanity, mindfulness).
Karakasidou et al. (2021) did not outline specific hypotheses, instead making vague predictions that positive associations with similar constructs would be indicative of convergent validity, and negative or no correlation with unrelated constructs would indicate discriminant validity. They found positive associations with life satisfaction, resilience, positive emotions, and happiness, and negative associations with negative emotions, anxiety, depression, and stress, with small-medium effects. The three negative factors (self- judgement, isolation, overidentification) had positive associations with negative emotions, stress, depression, and anxiety, and negative associations with positive emotions, happiness, life satisfaction, and resilience, with small-medium effects. Two positive factors (self- kindness and mindfulness) had positive associations with positive emotions, happiness, life satisfaction, and resilience, and negative associations with negative emotions, stress, and depression, with small-medium effects. Unexpectedly, the common humanity factor correlated positively with stress, anxiety, and negative SCS-Y factors. No correlations were r ≥ 0.50.
Deniz et al. (2022) reported that the SCS-Y was positively related to resilience and wellbeing, and negatively related to depression, congruent with hypotheses. However, this was examined by network analysis and no correlation values were given.
As expected, Nazari et al. (2022) found positive associations between positive factors of self-compassion with resilience and optimism, and negative associations with depression and neuroticism. Negative self-compassion factors also had positive associations with depression and neuroticism, and negative associations with resilience and optimism. Effects were small-medium, and no correlations were r ≥ 0.50. The SCS-Y total had negative associations with depression and neuroticism, and positive associations with resilience and optimism. However, this was examined with SEM analysis and r-values were not reported.
Huang et al. (2022), Cheung et al. (2023) and Henje et al. (2024) did not outline specific hypotheses. Huang et al. (2022) found large negative associations with anxiety and depression. and medium-large positive associations with interpersonal competence, wellbeing, and resilience. Cheung et al. (2023) reported high discriminant ability for items (graded response model) and acceptable discriminant validity for subscales (average variance extraction). The SCS-Y correlated highly with the SCS in both Chinese and UK samples. Henje et al. (2024) reported medium-large negative associations of the SCS-Y total with anxiety and depression, medium negative association with anger, and medium-large positive associations with wellbeing. Associations with functioning, and paediatric pain interference supported divergent validity. The correlation with the SCS was high (0.72). Subscale associations were largely in line with those for the total scale, except for the common humanity and mindfulness’ subscales, which showed small-medium associations with wellbeing, in- and externalising symptoms. The positive SCS-Y subscales showed weaker associations with the total SCS than the (reverse-coded) negative subscales.
Neff et al. (2021) reported means and standard deviations for the total sample and male and female participants for three grades, for three studies. Subgroup analyses were conducted for age, grade, gender, and gender and grade/age interactions explored. No significant differences were found for age or grade, but there was a trend for self-compassion to decrease in higher grades. Males had greater levels of self-compassion in two of three studies (one non-significant). No interaction was found between gender and age, or gender and grade, although one study showed a trend for males to score slightly higher than females, and for females to slightly decrease in self-compassion with age. Nazari et al. (2022) did not report means or standard deviations but also found males had higher levels of self- compassion than females. Karakasidou et al. (2021), Deniz et al. (2022) and Huang et al. (2022) did not aid interpretability. Cheung et al. (2023) compared Chinese and UK participants and tested gender differences (but did not provide M and SD). Boys scored higher on self-compassion than girls on all subscales. Chinese participants reported higher self-compassion subscale scores than UK participants. Henje et al. (2024) reported measurement invariance for age and ambiguous measurement invariance for gender. They reported means and standard deviations for the total, male and female (sub)samples and age groups (15–17 vs 18–20). Girls scored lower on total self-compassion than boys and higher than boys on self-judgment, isolation and over-identification. Age and socio-economic status were not significantly associated with self-compassion.
Normality data (Huang et al., 2022; Karakasidou et al., 2021; Nazari et al., 2022), descriptive statistics (Cheung et al., 2023; Henje et al., 2024) and percentile ranks (Karakasidou et al., 2021) suggested that floor and ceiling effects were not evident.
Summary
The SCS-Y was rated 8/14 for quality. Experts and respondents were mostly consulted on items, but the recognizing element of compassion was not captured. Various models were tested, reflecting the controversy around Neff’s originally proposed factor structure in the adult literature (e.g. Muris & Otgaar, 2020; Neff et al., 2019). However, all studies, except for the study by Karakasidou et al. (2021), supported Neff’s proposed factor structure. Karakasidou et al. (2021) employed a notably smaller sample, reported questionable internal consistency and some unexpected correlations. This suggests methodological weaknesses and/or possible cultural and/or language factors warranting further investigation. Some studies reported that a two-bifactor model was also a good fit. Internal consistency and test–retest reliability ranged from questionable to good. There was some evidence for construct validity, although some papers did not outline hypotheses, report r values, or find multiple large correlations. Floor and ceiling effects did not appear evident and interpretability was facilitated.
Further reliability testing is needed to ensure consistency across diverse cultural samples. While almost all studies confirmed Neff’s proposed factor structure, some also provided evidence for a two-bifactor model, suggesting that further investigation is warranted. Neff (2020) recommend that bifactor ESEM should be used, given the measure’s multidimensionality.
Compassion Scale–Child (CS-C)
Nas and Sak (2021) developed the CS-C, a 20-item self-report measure of compassion with a 5-point response scale (ranging from never to always) that has three subscales: self-compassion (9 items) and compassion for others (5 items) and towards other living things (six items) that consists of 20 self-report items. One paper (Heidary et al., 2020) validated the CS-C.
Population
Nas and Sak’s (2021) sample included 756 Turkish CYP aged 12–18 (M = not reported, 52.4% female) with varied SES. Heidary et al. (2022) validated the CS-C with 302 Iranian CYP aged 12–18 (M = 15.85, 85% female). Neither provided information on ethnicity.
Content Validity
Initial items selected by Nas and Sak (2021) were based on a review of key literature, including the measurement of compassion. Teachers and experts were consulted, and adjustments made following feedback. The measure was then piloted in schools with students who were representative of the target age group. Minor adjustments were made accordingly. Heidary et al. (2022) also consulted students and experts in the field and retranslated one item. Items related to 3/5 elements of compassion (universality, feeling, acting). Around a third of items relate to compassion towards other living things (e.g., environment and animals) and this was the only measure to include this orientation of compassion.
Factor Structure and Reliability
EFA yielded a three-factor structure (compassion toward other people, compassion toward oneself, compassion toward other living things). CFA indicated the three-factor structure was an acceptable fit. CFA by Heidary et al. (2022) supported the three-factor structure. Nas and Sak (2021) reported good internal consistency for the total scale (α = 0.89), and acceptable-good internal consistency for subscales (α = 0.75 to 0.86). Heidary et al. (2022) did not report internal consistency for the total scale, but subscales were acceptable-good (α = 0.70 to 0.87). Neither paper assessed test–retest reliability.
Construct Validity and Interpretability
Construct validity was not examined by Nas and Sak (2021). Congruent with hypotheses, Heidary et al. (2022) found medium-large positive associations with emotion regulation, mindfulness, and wellbeing. Two correlations were r ≥ 0.50. In relation to discriminant validity, they predicted the CS-C would correlate negatively with anxiety and found a negative non-linear association. Neither paper facilitated interpretability nor examined floor and ceiling effects.
Summary
The CS-C was rated 7/14. Experts and the target group were consulted on items but recognizing and tolerating elements of compassion were not captured. Despite initial consultation during item development, the target population should be included in future validation studies to refine items further. The three-factor structure was deemed acceptable, and internal consistency was good-acceptable. There was evidence for construct validity, but interpretability was not facilitated, and test–retest reliability and floor and ceiling effects were not assessed.
Compassion Scale–Adolescence (CS-A)
The CS-A assesses compassion for others and consists of 16 self-report items with a 5-point response scale (ranging from almost never to almost always). The CS-A has four subscales (kindness, mindfulness, common humanity, indifference) with four items each.
Population
Sousa et al. (2022) validated the Portuguese version (Sousa & Rojjanasrirat, 2010) of the adult CS (Pommier et al., 2020) with two adolescent samples in Portugal. The community sample (n = 658) were students from public schools, aged 14–18 (37.8% male). The forensic sample (n = 183) were aged 14–18 (M = 15.96, 100% male), recruited from foster care homes and juvenile detention facilities, and 80.3% had a primary diagnosis of conduct disorder. Participants were from socioeconomically diverse backgrounds. Ethnicity was not reported.
Content Validity
The original CS was a 24-item scale developed by Pommier (2010) for adults, then reduced to its current 16-items (Pommier et al., 2020). Three items were slightly reworded from this to be clearly understood by adolescents, without altering content. This was conducted by the research team with extensive experience with adolescence and measure validation procedures. Translation and backtranslation by a bilingual researcher assured similarity between the adult and adolescent versions. The target population were not consulted. As the final items were not presented in Sousa et al. (2022), the scale by Pommier et al. (2020) was reviewed, taking into account reworded items. Items related to 4/5 elements of compassion (not tolerating).
Factor Structure and Reliability
CFA supported the proposed structure of the CS in the original study (Pommier et al., 2020), of a hierarchical higher-order factor of compassion, and four subscales (kindness, common humanity, mindfulness, indifference).
Internal consistency for the total scale was good for both samples (α = 0.86, 0.88) and acceptable-good for subscales (α = 0.63 to 0.80). Test–retest reliability was not examined.
Construct Validity and Interpretability
As hypothesized, the CS-A had negative associations with external shame and self-criticism (hated self), and positive associations with self-compassion. However, there was not a significant association with a second self- criticism subscale (inadequate self). There was also a positive association between the CS-A subscales of common humanity, and mindfulness, with these same factors on the SCS-A. All effects were small, with no correlations of r ≥ 0.50.
Means and standard deviations for total and subscale scores were provided for community males, community females, and the total forensic sample. Measurement invariance across gender was established and subgroup analyses were conducted. Females in the community sample had higher levels of compassion when compared to males. Female scored higher on all subscales, except indifference, where males had higher levels. Males in the forensic sample had lower levels of compassion towards others than males in the community sample, and lower levels of mindfulness. The indifference subscale was associated with number of diagnoses in the forensic sample. Floor and ceiling effects were not examined.
Summary
The CS-A was rated 5/14. Experts, but not the target population, were consulted on items, and the tolerating element of compassion was not captured. CFA supported the hierarchical higher-order factor of compassion and four subscales, internal consistency was acceptable-good, and interpretability was facilitated. There was some evidence for construct validity, and test–retest reliability and floor and ceiling effects were not examined. Additional evidence is needed to support the hierarchical factor structure identified in the CFA. Alternative models could be explored and, as suggested by Neff et al. (2019), further bifactor analysis could be conducted to clarify the structure. Future studies should specify hypotheses for construct validity testing, to strengthen interpretations of associations with related constructs.
Self-Compassion Scale–Child (SCS-C)
Sutton et al. (2018) developed the SCS-C by modifying the SCS-Short Form (Raes et al., 2011), which has primarily been used with adults. The SCS-C consists of 12-items, rated on a 5-point response scale (ranging from never to always), including two items from each of the six SCS subscales.
Population
The sample included 382 CYP from Canada aged 8–12 (M = 11.3, 50% female, 71% English as first language).
Content Validity
Language of SCS-SF items was altered to be developmentally appropriate, but experts nor recipients were consulted. Items related to 4/5 elements of compassion (not recognizing).
Factor Structure and Reliability
CFA supported the use of a two-factor model as hypothesized, with positively and negatively worded self-compassion items forming two subscales (positive and negative). Internal consistency was good for the two subscales (α = 0.81, 0.83). Test–retest reliability was not assessed.
Construct Validity and Interpretability
Whilst associations were reported to be in expected directions and an aim was to see if the SCS-C related to other constructs in line with previous research, clearly predefined hypotheses were not outlined. The positive subscale was positively associated with self-concept, optimism, empathetic-related responding, prosocial goals, life satisfaction, and mindfulness, and negatively associated with depression and anxiety. The negative subscale was negatively associated with mindfulness, optimism, life satisfaction, and positive affect, and positively associated with negative affect, depression, anxiety, and empathic concern, but not associated with self-concept, perspective- taking, or prosocial goals. Effects were small-large, and three correlations were r ≥ 0.50. Interpretability was not facilitated, and floor and ceiling effects were not examined.
Summary
The SCS-C was rated 4/14 for quality. Experts and recipients were not consulted, and the recognizing element of compassion was not captured. CFA supported the two-factor structure and there was some evidence for construct validity, but predefined hypotheses were not outlined. Internal consistency was good. Few measurement properties were assessed, as test–retest reliability and floor and ceiling effects were not examined, and interpretability was not facilitated.
The S-SCS-A consists of 9 self-report items, on a 5-point response scale (ranging from never to always) and has three subscales (self-kindness, common humanity, mindfulness).
Population
The sample included 132 Dutch adolescents aged 12–17 (M = 14.8, 58% female), from socioeconomically diverse backgrounds (Muris et al., 2016).
Content Validity
S-SCS-A items were modified from the self-kindness, common humanity and mindfulness subscales of the SCS (Neff, 2003b), based on the authors’ critique that including the self-judgment, isolation and overidentification components of compassion inflates the association between compassion and anxiety and depression (also see Muris & Otgaar, 2020; Muris et al., 2024). The modification was guided by three CYP aged 12–15 with varied educational levels. They identified difficult items, and three psychologists modified them. Items related to 4/5 elements of compassion (not recognizing).
Factor Structure and Reliability
PCA yielded three factors (self-kindness, common humanity, mindfulness), with 8/9 items loading most substantially on their intended factor. Neither EFA nor CFA was conducted. Internal consistency was good for the total scale (α = 0.84), and acceptable for subscales (α = 0.74 to 0.79). Test–retest reliability was not examined.
Construct Validity and Interpretability
Vague predictions were made for a link between self-compassion and mental health problems. The S-SCS-A was positively associated with self-esteem and self-efficacy, and negatively associated with anxiety and depression, with small-medium effects. Subscales were positively associated with self-esteem (except common humanity) and self-efficacy, and negatively associated with anxiety (except self- kindness) and depression, with small-large effects. Three correlations were r ≥ 0.50.
Means and standard deviations for total and subscale scores were provided for the total sample, males, and females. No differences for gender or age were found. However, when older (15–17) and younger (13–14) adolescents were analyzed separately, females had lower self-compassion than males in the older sample. Floor and ceiling effects were not assessed.
Summary
The S-SCS-A was rated 4/14 for quality. Experts and the target group were consulted on items, but the recognizing element of compassion was not captured. Whilst PCA yielded three factors, EFA/CFA was not conducted. Internal consistency was good for the total scale and acceptable for subscales, but test–retest was not conducted. Although there was some evidence for construct validity, predefined hypotheses were not outlined. Interpretability was facilitated, but floor and ceiling effects were not assessed.
Self-Compassionate Reactions Scale–Child (SCRS-C)
The SCRS-C consists of four vignettes and 24-items (6 subscales: self-kindness, self-judgment common humanity, isolation, mindfulness, over-identification) informed by Neff’s conceptualisation of self-compassion. Vignettes are written in simple sentences, followed by six items describing possible reactions. Items are rated on a 6-point response scale (ranging from almost never to almost always).
Population
Zhou et al. (2019) piloted the SCRS-C with 161 children (M = 11.49, 36% male), and examined the factor structure with 637 children (M = 10.62, 51% male) and test–retest reliability with a subset of 77 children. All children were Chinese, aged 9–12, and attended school in China. Information about SES was not reported.
Content Validity
To inform vignettes, primary school teachers were interviewed, and children rated the frequency of stressful situations in the classroom. Children were interviewed about vignette situations to help inform items. Experts in compassion research evaluated the meaning of items (consistency with definition of self-compassion) and appropriateness of language for children. Items related to 4/5 elements of compassion (not recognizing).
Factor Structure and Reliability
CFA was conducted but EFA was not conducted beforehand. Whilst items were developed to be consistent with Neff’s (2003a) definition, Zhou et al. (2019) did not make explicit a proposed factor structure for this new scale. A two second-order factor model (compassionate reactions and uncompassionate reactions) with six first-order factors (self-kindness, common humanity, mindfulness, self-judgment, isolation, and over-identification) best fit the data. 0/2 was given for factor structure, as EFA was not conducted, and the factor structure of this newly developed scale was not proposed before CFA.
Internal consistency was good for the total scale (α = 0.81), acceptable-good for second-order subscales (α = 0.78 to 0.81) and unacceptable-acceptable for first-order subscales (α = 0.43 to 0.73). Only the self-kindness subscale reached an acceptable value. Test–retest reliability was poor for the total score (r = 0.46), and questionable for subscales (r = 0.58 to 0.61).
Construct Validity and Interpretability
Predefined hypotheses were not outlined. The SCRS-C total had positive associations with mindfulness, self-acceptance, and self-esteem. The compassionate reactions scale was positively associated with self-esteem but not mindfulness. The uncompassionate reactions scale was negatively associated with self- esteem, self-acceptance, and mindfulness. There were no associations with narcissism, which was used to assess discriminant validity. Overall, effects were small-large, and three correlations were r ≥ 0.50. Interpretability was not facilitated, and floor and ceiling effects were not examined.
Summary
The SCRS-C was rated 2/14 for quality. Experts and the target group were consulted on items, but the recognizing element of compassion was not captured. EFA was not conducted, nor was a factor structure proposed before CFA. CFA supported the use of two second-order factors, with six first-order factors. Test–retest reliability was poor for the total score and questionable for subscales, and internal consistency was good for the total scale, and unacceptable and acceptable for first and second-order subscales, respectively. There was some evidence for construct validity, but predefined hypotheses were not outlined. Interpretability was not facilitated, and floor and ceiling effects were not assessed.
Discussion
The systematic review aimed to describe and critically appraise the psychometric properties of self- and other-compassion measures that were developed and/or validated with children and/or young people (CYP, defined as being younger than 18 years of age). Congruent with a review of adult compassion measures for adults (Strauss et al., 2016), this review of eight measures highlights the lack of psychometrically robust compassion measures for CYP. It also demonstrates increased interest in compassion in CYP in recent years, with more than half of the papers published within the last three years.
Three measures (4 papers) assessed other-compassion and seven (18 papers) assessed self-compassion. The review found that the SCS and its adaptations (SCS-Y, SCS-C, S-SCS-A) also dominate the measurement of self-compassion in CYP. The SCS and other measures did not fully capture the consensual definition of compassion by Strauss et al. (2016) and many papers lacked consultation on items with both experts and recipients, contributing to poor content validity, with no measure scoring the maximum score of 2 for this criterion. Arguably, poor content validity is the most problematic of all criteria. A measure could be rated highly for all other psychometric properties, but if it fails to adequately capture the full breadth of the construct being evaluated, the value of the measure in research and practice is questionable. Therefore, psychometric evaluations of measures for CYP that would be rated highly (i.e. 2) for content validity are needed.
CFA was conducted for seven measures, but mixed evidence emerged for the factor structure of measures based on the SCS, apart from the SCS-Y, reflecting inconsistencies reported in the adult literature (Muris & Petrocchi, 2017; Muris & Otgaar, 2020), suggesting poor validity. Muris and colleagues (e.g. Muris et al., 2021, 2022) argue that including the (reduced) uncompassionate self-responding subscales inflates the relationship of the SCS with internalising symptoms, complicating the exploration of the protective role of compassion. It is also questionable whether lack of uncompassionate responding to the self is equivalent to being self-compassionate. Neff et al. (2019) refuted these points however, reporting empirical support for the use of a total score in adults (and six subscale scores when assessing mechanisms) rather than separate scores for uncompassionate and compassionate self-responding, based on their analysis of 20 (adult) study samples. In the current review, studies into the factor structure of the SCS-Y were largely aligned with this. However, some studies also found good evidence for the two-bifactor model, aligning with the concerns raised by Muris et al.and colleagues (2024). Therefore, as in research with adults, further research is needed to clarify the factor structure of the SCS as adapted for CYP.
Internal consistency was the strongest criterion. However, Cronbach’s alpha was unacceptable for some measures. Only five papers assessed test–retest reliability. All but two were rated zero, questioning the reliability over time of current measures.
Correlations with related constructs were reported for all measures. Three measures met the full criteria for construct validity. Whilst many correlations were weak and some findings inconsistent, correlations were generally significant, medium-large, in expected directions, and suggestive of the importance of self-compassion for wellbeing (Marsh et al., 2018). Other-compassion also had associations suggestive of benefits for wellbeing, but fewer papers examined this. This suggests that offering CBIs to CYP may offer the possibility of improving wellbeing and mental health outcomes. Few papers examined the relationship between other- and self-compassion. Findings were mixed, with small-large associations found (Cunha et al., 2023; Sousa et al., 2022). This suggests more limited understanding of this relationship in CYP and the need for further research.
No paper explicitly examined floor and ceiling effects. Nine papers did not aid interpretability of scores. At least one subgroup analysis was conducted for five measures, with gender the most common comparison. Subgroup analyses generally indicated that females had lower levels of self-compassion than males, but higher levels of compassion for others, supporting research with adults (Yarnell et al., 2015). There was also indication of a trend for self-compassion to decrease with age, particularly for females. Few measures had subgroup analyses conducted for multiple groups to aid interpretability.
The highest rated measure overall was the CEAS-A (9/14), followed by the SCS-Y (8), SCS, and CS-C (both 7/14). None of the measures, however, captured the five elements of compassion or explicitly examined floor and ceiling effects. CFA (and ESEM) was conducted for the CEAS, SCS-Y and CS-C, supporting factor structures, and construct validity was good. However, there was mixed evidence for the SCS factor structure, and inconsistent findings for construct validity. The CEAS-A was the only measure to fully meet criteria for test–retest reliability and the SCS and SCS-Y did not meet full criteria for internal consistency, due to inconsistent findings. However, the CEAS-A, in addition to the SCS, lacked consultation with experts and CYP. The SCS and SCS-Y also focus on self-compassion, meaning they are not suitable to assess compassion towards others.
Identified papers used samples from 11 different countries (Portugal, US, Finland, Netherlands, China, Canada, Turkey, Iran, Greece, Germany, Sweden). The majority of papers did not report socio-economic status or ethnicity of participants. However, of those that did, samples were generally diverse. Lack of consultation with CYP for items is particularly problematic for measures developed for use with adults (e.g., SCS), as it cannot be assumed the language used is developmentally appropriate and this contributes to low ratings for content validity for these measures. Whilst some measures were validated in different countries, with items translated to a different language, recipients/experts were not consulted on the translated versions in most included papers. This therefore poses a risk of items not being suitable for these populations.
One measure assessed compassion towards other living things, which does not fully align with the definition of Strauss et al. (2016) of compassion as being related to “human suffering” and feeling for the “person”. However, given the current climate emergency, this may be an important area for future research.
Some papers conducted PCA and did not conduct EFA or CFA. However, PCA has a greater emphasis on data reduction than interpretation. EFA is recommended when research aims to identify underlying factors, and CFA is recommended to test whether a set of observed variables represent underlying factors (Alavi et al., 2020). Neff et al. (2019) recommended bifactor ESEM to assess the factor structure of the SCS given its multidimensionality, however, as noted above, the inclusion of uncompassionate self-responding items has been questioned for both theoretical and empirical reasons (e.g. see Muris et al., 2024).
Whilst Pearson and Spearman correlation coefficients were reported in two papers for test–retest reliability, these do not take into account systematic differences. The intraclass correlation coefficient (ICC) is reported to be a more suitable measure of reliability as it reflects the degree of correlation and agreement between measurements, taking into account systematic differences (Koo & Li, 2016; Terwee et al., 2007). However, only three (Cunha et al., 2023; Henje et al., 2024; Nazari et al., 2022) papers reported the ICC.
Many papers did not outline predefined hypotheses for construct validity. Where hypotheses were made, these were generally about the direction of correlations (positive or negative), but not about the strength of correlations. However, hypotheses should be as specific as possible (Terwee et al., 2007) as without, there is a high risk of bias as it can be tempting to give alternative explanations for low correlations, instead of concluding the measure has poor validity.
Strengths Limitations and Future Research
A strength of this review is that a broad, stringent and replicable approach was taken to assessing psychometric quality, allowing for comparison of quality across measures and between measures used with CYP and measures used with adults. Whilst a range of definitions of compassion exist, the Strauss et al. (2016) definition was used in this study as it aligns with theories of compassion, was developed to reach consensus across multiple definitions and is supported by factor analysis of items from multiple measures of compassion (Gu et al., 2017). Ratings of content validity are dependent on the adopted definition of compassion. However, this did not affect the rank order of measure quality or alter the outcome that none of the measures obtained the maximum quality score.
This review highlighted the need for valid and reliable measures to assess both self-compassion, and compassion towards others, and for the psychometric properties to be examined across a diverse range of CYP, including CYP from different socio-economic, cultural/language and ethnic backgrounds. Existing measures (including following translation, where applicable) should be revised through thorough consultation with both experts and young recipients during item development/translation, as a critical step towards enhancing content validity, ensuring that items are culturally, linguistically and developmentally appropriate for CYP across diverse backgrounds. Questionnaire items should align with a comprehensive definition of compassion, such as the framework by Strauss et al. (2016), to capture the multidimensional nature of compassion adequately and therefore to ensure strong content validity (arguably, as noted above, the most critical criterion). Gu et al. (2020) developed the Sussex-Oxford Compassion Scales (SOCS) with adults to align with the definition of compassion by Strauss et al. (2016), and this measure has strong psychometric properties in adults (Gu et al., 2020). Muris et al. (2022) advocated for the use of measures such as the SOCS (Gu et al., 2020) with CYP to advance understanding of compassion as a protective construct, given limitations they highlight with regard to the SCS and its derivatives, the currently most widely used measure of compassion with CYP. However, the psychometric properties of the SOCS with CYP are as yet unexamined, with the exception of establishing internal consistency (Muris et al., 2022), and this is an important avenue for future research.
A more rigorous approach to confirmatory factor analysis (CFA) is needed to address multicollinearity, particularly in factor models with high intercorrelations between subscales, to ensure that constructs are distinct and not redundant. Exploratory factor analysis (EFA) could be reported followed by CFA (with data from independent samples contributing to each) when developing new measures or validating existing ones, to ensure that underlying factors are identified and assessed using both bottom-up and top-down approaches.
Future studies should detail hypotheses for construct validity, specifying the strength and direction of correlations, to mitigate bias and support more accurate assessments of construct validity. Test–retest reliability should be evaluated, using intraclass correlation coefficients (ICC), to account for systematic differences, to offer insight into the stability of compassion measures over time, particularly during adolescence. More comprehensive reporting of internal consistency of scales/subscales (Cronbach's alpha) is essential to examine reliability across various populations and settings.
Future studies should also conduct subgroup analyses across multiple demographic groups (e.g., gender, age, socio-economic status) and report the necessary details, such as means, standard deviations, and subgroup analyses, to enhance the interpretability and replicability of findings and provide insights into how compassion measures function across diverse populations of CYP. Consistent reporting of socio-economic status and ethnicity of participants is also recommended to better understand of how these factors might be associated with compassion.
This review underscores the importance of examining and reporting floor and ceiling effects, to understand the range and distribution of compassion scores, ensuring measures are sensitive across the entire range of possible scores, aiding the interpretation of results. Measures (including when translated) should be validated across different cultural contexts and languages, with thorough consultation with local experts and recipients to ensure cultural and linguistic relevance and appropriateness.
Whilst poor test–retest reliability could be seen to undermine the reliability of measures of compassion reviewed here, this may also suggest that compassion is less stable during development and in adulthood. Therefore, future studies should explore developmental aspects of compassion among CYP, including its stability over time and the relationship between self-compassion and other-compassion. Evaluating the psychometric properties of measures such as the SOCS, which measure both self-compassion and other-compassion with potentially stronger content validity than in the measures reviewed here, in CYP could aid this understanding. This exploration could inform interventions aimed at promoting compassion and well-being.
Declarations
Conflict of Interest
CS is co-lead for Sussex Mindfulness Centre and has received funding from the National Institute of Health Research and from the Economic and Social Research Council (ESRC) to evaluate and develop mindfulness and compassion-based interventions and measures. She is the lead author on the Strauss et al. (2016) review and a co-author on the Gu et al. (2017, 2020) articles referenced in the current paper. TL has received ESRC funding to develop and evaluate mindfulness-based interventions. JH declares that they have no conflicts of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
share
DELEN
Deel dit onderdeel of sectie (kopieer de link)
Optie A:
Klik op de rechtermuisknop op de link en selecteer de optie “linkadres kopiëren”
Met BSL Psychologie Totaal blijf je als professional steeds op de hoogte van de nieuwste ontwikkelingen binnen jouw vak. Met het online abonnement heb je toegang tot een groot aantal boeken, protocollen, vaktijdschriften en e-learnings op het gebied van psychologie en psychiatrie. Zo kun je op je gemak en wanneer het jou het beste uitkomt verdiepen in jouw vakgebied.