Self-Compassion, Wellbeing, and Distress in Adolescents and Young Adults with Chronic Medical Conditions: the Mediating Role of Emotion Regulation Difficulties.

The sample consisted of 107 (21 male, 81 female, 5 transgender or gender-diverse) adolescents and young adults with a mean age of 21.27 ( SD  = 2.59). To meet inclusion criteria, participants needed to live in Australia, be aged 16–25, and self-identify as having a chronic physical medical condition. The most commonly reported CMCs were asthma ( n  = 33; 30.8%), chronic pain ( n  = 29; 27.1%), chronic fatigue syndrome/myalgic encephalomyelitis ( n  = 27; 25.2%), and allergies ( n  = 26; 24.3%), with 45.8% of participants reporting two or more CMCs. A complete list of CMCs reported by participants and the prevalence of these conditions in the sample is reported in Table 1 . Table 2 displays participant characteristics including location, duration of illness, and the number of chronic medical conditions and mental health conditions reported per participant. Table 1 Frequency of chronic medical conditions Diagnosis n % Asthma 33 30.8 Chronic pain 29 27.1 Chronic fatigue syndrome 27 25.2 Allergies 26 24.3 Chronic skin conditions 9 8.4 Arthritis 7 6.5 Inflammatory bowel disease 7 6.5 Cancer 6 5.6 Endometriosis 6 5.6 Alopecia 5 4.7 Burns 5 4.7 Ehlers-Danlos syndrome 4 3.7 Epilepsy 4 3.7 Postural orthostatic tachycardia syndrome 4 3.7 Type 1 diabetes 4 3.7 Cystic fibrosis 3 2.8 Cerebral palsy 2 1.9 Coeliac disease 2 1.9 Fibromyalgia 2 1.9 Joint hypermobility 2 1.9 Multiple sclerosis 2 1.9 Nemaline myopathy 2 1.9 Axial spondylitis 1 0.9 Bronchiectasis 1 0.9 Cystic fibrosis-related diabetes 1 0.9 Chiari malformation 1 0.9 Congenital heart disease 1 0.9 Connective tissue disorder 1 0.9 Functional neurological disorder 1 0.9 Gastroesophageal reflux disease 1 0.9 Hip dysplasia 1 0.9 Hyper IgE syndrome 1 0.9 Klippel-Feil syndrome 1 0.9 Migraine 1 0.9 Polycystic ovarian syndrome 1 0.9 Sickle cell disease 1 0.9 Supraventricular tachycardia 1 0.9 Type 2 diabetes 1 0.9 Participants could select more than one condition Table 2 Medical, mental health, and geographic characteristics of participants Characteristic n % Number of chronic medical conditions a 1 58 54.2 2 20 18.7 3 13 12.1 4 12 11.2 5 3 2.8 7 1 0.9 Number of mental health conditions b 0 52 48.6 1 17 15.9 2 15 14.0 3 18 16.8 4 2 1.9 5 2 1.9 6 1 0.9 Time since diagnosis c Less than 1 month 1 0.9 1–3 months 2 1.9 4–6 months 3 2.8 7–11 months 4 3.7 1–2 years 12 11.2 3–4 years 26 24.3 More than 5 years 58 54.2 Not specified 1 0.9 Location Western Australia 31 29.0 Victoria 28 26.2 New South Wales 22 20.6 Queensland 6 5.6 South Australia 5 4.7 Australian Capital Territory 5 4.7 Tasmania 3 2.8 Not specified 7 6.5 N  = 107 a Number of chronic medical conditions reported per participant b Number of mental health conditions reported per participant c For illness of longest duration reported by each participant Frequency of chronic medical conditions Participants could select more than one condition Medical, mental health, and geographic characteristics of participants N  = 107 a Number of chronic medical conditions reported per participant b Number of mental health conditions reported per participant c For illness of longest duration reported by each participant Data were collected using an online survey hosted by Qualtrics (Qualtrics, Provo, UT). Participants were recruited via organizations representing young people with CMCs. Promotional materials containing the survey link were distributed through emailing lists, online newsletters, and social media posts. Participants followed the link directly to the survey. They were required to check a box indicating that they understood the information sheet and consented to participate in order to proceed. The survey took approximately 20–30 min to complete. Participation could be terminated at any time by exiting the survey. A $10 voucher for Big W or JB Hi Fi was offered to each participant upon completion. Data were collected between January and September 2020. Demographic questions included age, gender, location, chronic medical diagnosis, mental health diagnosis, and time since diagnosis. The Self-Compassion Scale – Short Form (SCS-SF; Raes et al., 2011 ) is a 12-item abbreviated form of Neff’s ( 2003b ) 26-item Self-Compassion Scale (SCS). Participants responded to items such as, “When I’m feeling down I tend to obsess and fixate on everything that’s wrong” on a Likert scale ranging from 1 ( almost never ) to 5 ( almost always ). Some items were negatively worded and required reverse coding. The SCS-SF contains six subscales, with self-kindness, common humanity, and mindfulness reflecting compassionate self-responding and self-judgment, isolation, and over-identification reflecting uncompassionate self-responding. Responses from all items were summed and averaged to give a total mean self-compassion score (Raes et al., 2011 ). Possible scores range from 1 to 5, with a high score indicating that an individual is highly self-compassionate. Internal consistency in this sample was high, α = 0.91. The Difficulties in Emotion Regulation Scale–Short Form (DERS-SF; Kaufman et al., 2016 ) is an 18-item abbreviated form of Gratz and Roemer’s ( 2004 ) Difficulties in Emotion Regulation Scale. Participants responded to items such as, “I have difficulty making sense out of my feelings” on a Likert scale ranging from 1 ( almost never ) to 5 ( almost always ). The DERS-SF has six subscales labelled strategies, non-acceptance, impulse, goals, awareness, and clarity. The use of a total score reflecting global emotion regulation difficulties has been validated in multiple studies (e.g., Hallion et al., 2018 ). Three items required reverse coding. Responses from all items were summed to give a total score. Possible scores range from 18 to 90, with high scores indicating substantial emotion regulation difficulties. Internal consistency in this sample was high, α = 0.92. The World Health Organization Wellbeing Index (WHO-5; Bech, 1996 ) is a five-item unidimensional indicator of general wellbeing. Participants considered their feelings over the last 2 weeks and responded to items such as, “I have felt active and vigorous” on a Likert scale ranging from 5 ( all of the time ) to 0 ( at no time ). Scores were multiplied by four to give a total score out of 100. Possible scores range from zero, indicating the poorest wellbeing imaginable, to 100, indicating the best wellbeing imaginable. Internal consistency in this sample was high, α = 0.87. The Kessler Psychological Distress Scale (K-10; Kessler et al., 2002 ) is a 10-item measure of non-specific psychological distress, with items covering several domains including depression, anxiety, fatigue, and physical symptoms of arousal (Andrews & Slade, 2001 ). Participants were asked, “In the past 30 days how often…” and responded to items such as “did you feel nervous?” on a five-point Likert scale ranging from 5 ( all of the time ) to 1 ( none of the time ). Possible scores range from 10, indicating no distress, to 50, indicating severe distress. Internal consistency in this sample was high, α = 0.90. Symptom severity was measured using a single item, “On a scale of 1–10, how severe would you say your current physical symptoms are?” Participants responded on a 10-point Likert scale ranging from 1 ( not severe at all ) to 10 ( extremely severe ). Lu et al. ( 2020 ) found that participants’ scores on a single item measuring headache severity were highly correlated with their total score on a six-item measure, r  = 0.08, providing support for the validity of single-item measures. The use of a general, non-symptom-specific question was necessitated by the inclusion of many different diagnoses in the present study.

As there is a lack of consensus in the literature surrounding the factor structure of the SCS-SF (e.g., Neff et al., 2018 ) and DERS-SF (e.g., Moreira et al., 2020 ), factor analyses were conducted to determine the appropriateness of using a total score. Principal axis factoring with Promax rotation supported a single-factor structure for the SCS-SF. A single-factor model using principal axis factoring with Promax rotation produced high factor loadings for all items of the DERS-SF except 1, 4, and 6, which form the “awareness” subscale. The literature commonly reports low reliability and poor factor loadings for the awareness subscale in both adolescent (e.g., Neumann et al., 2010 ) and young adult (e.g., Bardeen et al., 2012 ; Tull et al., 2007 ) populations, to the extent that several papers recommended that the subscale is excluded from the total score (e.g., Hallion et al., 2018 ; Moreira et al., 2020 ). Consistent with previous research, items 1, 4, and 6 in the present study had low initial and extracted communalities, and internal reliability of the total scale was higher without these items. These items were removed, and a second analysis using principal axis factoring with Promax rotation supported a single-factor structure. As such, subsequent analyses were conducted using a total score comprised of the 15 retained items. Pearson’s correlations between each measure are reported in Table 3 , alongside the mean, standard deviation, and Cronbach’s alpha. Table 3 Bivariate correlations, means, standard deviations, and Cronbach’s alpha for symptom severity and measures in mediation models 1 2 3 4 M (SD) α Symptom severity  − .10 .20*  − .23* .24* 5.66 (2.02) 1. SCS-SF a - 2.71 (0.80) .91 2. DERS-SF b  − .70** - 40.54 (12.04) .92 3. WHO-5 c .46**  − .47** - 39.66 (20.75) .87 4. K10 d  − .61** .62**  − .74** - 27.04 (8.18) .90 N  = 107 a Self-Compassion Scale–Short Form (Raes et al., 2011 ) b Difficulties in Emotion Regulation Scale–Short Form (Kaufman et al., 2016 ) c World Health Organization Wellbeing Index (Bech, 1996 ) d Kessler Psychological Distress Scale (Kessler et al., 2002 ) * Correlation is significant at the .05 level (2-tailed) ** Correlation is significant at the .01 level (2-tailed) Bivariate correlations, means, standard deviations, and Cronbach’s alpha for symptom severity and measures in mediation models N  = 107 a Self-Compassion Scale–Short Form (Raes et al., 2011 ) b Difficulties in Emotion Regulation Scale–Short Form (Kaufman et al., 2016 ) c World Health Organization Wellbeing Index (Bech, 1996 ) d Kessler Psychological Distress Scale (Kessler et al., 2002 ) * Correlation is significant at the .05 level (2-tailed) ** Correlation is significant at the .01 level (2-tailed) Two simple mediation models were analyzed using the PROCESS macro (Hayes, 2013 ) for SPSS (IBM Corp., 2019 ). PROCESS conducts path analysis using ordinary least squares regression (Hayes, 2013 ). Model 4 was selected with a 95% confidence interval and 10, 000 bias-corrected bootstrap samples. Gender and symptom severity were included as control variables in both models. Unstandardized ( B ) regression coefficients, 95% CI, and standard error estimates for model one are presented in Fig.  1 . Self-compassion and emotion regulation difficulties in combination accounted for a statistically significant proportion of unique variance in distress, R 2  = 0.46, F (4, 101) = 21.83, p  < 0.001. This is a large effect according to Cohen’s ( 1988 ) conventions ( f 2  = 0.85). Hypothesis one was supported, with self-compassion predicting a significant proportion of unique variance in emotion regulation difficulties, a  =  − 10.06, LLCI/ULCI \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\ne$$\end{document} ≠ 0, p  < 0.001. Emotion regulation difficulties also predicted a significant proportion of unique variance in distress, b  = 0.24, LLCI/ULCI \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\ne$$\end{document} ≠ 0, p  < 0.05. Hypothesis two was supported, with evidence of a significant direct negative association between self-compassion and distress, c’  =  − 3.58, LLCI/ULCI \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\ne$$\end{document} ≠ 0, p  < 0.01. Hypothesis three was also supported, with the indirect effect of self-compassion via emotion regulation difficulties predicting a significant proportion of unique variance in distress, ab  =  − 2.38, LLCI/ULCI \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\ne$$\end{document} ≠ 0, p  < 0.05. The total effect of the model was also significant, c  =  − 5.95, LLCI/ULCI \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\ne$$\end{document} ≠ 0, p  < 0.001. Effect size was determined by calculating the ratio of indirect effect to total effect (Hayes, 2013 ). The indirect pathway accounted for 40% of the total effect of self-compassion on distress. Fig. 1 Statistical diagram of mediation model one: distress (K10) as outcome Statistical diagram of mediation model one: distress (K10) as outcome Unstandardized ( B ) regression coefficients, 95% CI, and standard error estimates for model two are presented in Fig.  2 . Self-compassion and emotion regulation difficulties in combination accounted for a statistically significant proportion of unique variance in wellbeing, R 2  = 0.31, F (4, 101) = 11.47, p  < 0.001. This is a large effect according to Cohen’s ( 1988 ) conventions ( f 2  = 0.45). Hypothesis four was supported, with self-compassion predicting a significant proportion of unique variance in emotion regulation difficulties, a  =  − 10.06, LLCI/ULCI \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\ne$$\end{document} ≠ 0, p  < 0.001. Hypothesis five was supported, as the direct path between self-compassion and wellbeing was significant, c’  = 6.04, LLCI/ULCI \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\ne$$\end{document} ≠ 0, p  < 0.05. Emotion regulation difficulties also predicted a significant proportion of unique variance in wellbeing, b  =  − 0.41, LLCI/ULCI \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\ne$$\end{document} ≠ 0, p  < 0.05. Hypothesis six was not supported as the unstandardized coefficient for the indirect effect was non-significant, ab  = 4.13, LLCI/ULCI \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$=$$\end{document} = 0, p  > 05. Fig. 2 Statistical diagram of mediation model two: wellbeing (WHO-5) as outcome Statistical diagram of mediation model two: wellbeing (WHO-5) as outcome

The aim of the current study was to examine the relationship of self-compassion to distress and wellbeing. A further aim was to extend findings of an emotion regulation model of self-compassion to a population of young people with CMCs. The participants’ average distress rating is considered “high” against population norms (Australian Bureau of Statistics, 2012 ), as well as in comparison to samples of young people with CMCs, such as adolescents and young adults with cancer, who reported a mean K10 score of 19.2 (McCarthy et al., 2016 ). Participants also reported wellbeing considerably lower than the mean WHO-5 score of 63.38 observed in a sample of adolescents with type 1 diabetes (de Wit et al., 2007 ). The high distress and low wellbeing reported by this sample highlights the importance of identifying correlates of distress and wellbeing for adolescents and young adults with CMCs. Self-compassion and emotion regulation difficulties together accounted for nearly half of the unique variance in distress. This is consistent with findings from a hierarchical multiple regression that self-compassion and emotion regulation, alongside age, gender, insulin use, and symptom severity, predicted 38.5% of the variance in diabetes-related distress in adults with type 2 diabetes (Kane et al., 2018 ). There was a significant direct negative association between self-compassion and emotion regulation difficulties. This was expected in light of previous research linking self-compassion and emotion regulation (e.g., Bluth et al., 2016a , b ; Raes, 2010 ). There was a significant direct positive association between emotion regulation difficulties and distress. This is consistent with previous literature suggesting that adolescents (Garnefski et al., 2009 ; Kraaij & Garnefski, 2012 ) and adults (Trindade et al., 2017 ) with CMCs who experience high levels of emotion dysregulation experience greater distress. As predicted, there was a significant direct negative association between self-compassion and distress. This is consistent with findings from meta-analyses using general community samples (Macbeth & Gumley, 2012 ) and adolescent samples (Marsh et al., 2018 ) which suggest that individuals high in self-compassion are likely to experience less distress than individuals with an uncompassionate style of self-responding. Findings from the current study further contribute to an understanding of the mechanism through which self-compassion is related to distress. As has been observed in previous samples of psychologists (Finlay-Jones et al., 2015 ), adults with depression (Diedrich et al., 2017 ; Krieger et al., 2013 ), and young adults (Barlow et al., 2017 ; Johnson & O’Brien, 2013 ; Raes, 2010 ), a significant negative indirect association was found between self-compassion and distress, operating via emotion regulation difficulties. This suggests that emotion regulation difficulties mediate the relationship between self-compassion and distress. Of note, this pathway accounted for 40% of the total effect of self-compassion on distress. This large effect is comparable to that observed in a sample of adults with depression, where the indirect pathway between self-compassion and depression symptoms via emotion regulation accounted for 46.63% of the total effect (Diedrich et al., 2017 ). These findings further strengthen support for an emotion regulation model of self-compassion (e.g., Finlay-Jones, 2017 ; Finlay-Jones et al., 2015 ), whereby self-compassion is related to lower distress by facilitating the use of adaptive emotion regulation strategies. The contribution of this study is to extend these findings to adolescents and young adults with CMCs. The current study also aimed to explore whether a meaningful relationship exists between self-compassion, emotion regulation, and wellbeing. Overall, the model accounted for nearly a third of the unique variance in wellbeing, suggesting that self-compassion and emotion regulation difficulties are relevant to the experience of positive psychological outcomes in young people with CMCs. As expected, there was a significant positive direct effect of self-compassion on wellbeing, suggesting that individuals who are self-compassionate are more likely to experience high wellbeing than individuals who are less self-compassionate. This was expected, given the positive correlations previously observed in a meta-analysis using general community samples (Zessin et al., 2015 ). A significant direct negative association was observed between emotion regulation difficulties and wellbeing, confirming our expectation that individuals with greater emotion regulation difficulties would experience lower wellbeing. However, no support was found for an indirect effect of self-compassion on wellbeing via emotion regulation difficulties. No previous research has investigated this relationship using wellbeing as an outcome. However, literature suggests that the bivariate correlations between self-compassion and wellbeing (e.g., Zessin et al., 2015 ), and emotion regulation difficulties and wellbeing (e.g., Hu et al., 2014 ), are smaller than their respective correlations with distress (e.g., Kraaij & Garnefski, 2012 ; Marsh et al., 2018 ). Given these findings, it is possible that an indirect effect exists but is smaller than the indirect effect observed in the distress model. The current study may have been underpowered to detect an effect of this size. Additional explanations may relate to the measure of wellbeing that was used. According to Diener ( 1984 ), subjective wellbeing consists of affective wellbeing, such as the experience of positive emotions, and cognitive wellbeing, such as experiencing a state of satisfaction with life. It is logical that emotion regulation difficulties are most closely related to affective wellbeing. With only two items, “I have felt cheerful and in good spirits” and “I have felt calm and relaxed,” that may relate to affective wellbeing, it is possible that the WHO-5 (Bech, 1996 ), was not sensitive enough to detect this relationship. Alternatively, emotion regulation difficulties may not act as a mediator between self-compassion and wellbeing. Emotion regulation may be more closely related to distress, due to its role in modulating the intensity and duration of negative affect (Thompson, 1994 ). The presence of decreased negative affect relative to positive affect may, over time, facilitate increased wellbeing (Diener, 1984 ). As such, the pathway between self-compassion and wellbeing may be more complex than what the simple mediation model used in the current study can explain. This possibility is consistent with the absence of an indirect effect, but presence of significant direct effects and significant unique variance explained by the model. Other variables within positive psychology such as hope (e.g., Yang et al., 2016 ) and self-efficacy (e.g., Sirois, 2015 ) may also be involved in the relationship between self-compassion and wellbeing in this population. These findings provide support for an emotion regulation model of self-compassion and its application among young people with CMCs. This is encouraging, as self-compassion and emotion regulation may meaningfully relate to the daily challenges experienced by this population. For example, Neff ( 2003a ) described how self-compassion may buffer against negative affect when receiving negative self-relevant information. Young people with CMCs may frequently encounter negative self-relevant information related to their health status. For example, a young person with type 1 diabetes required to test their blood glucose several times a day may frequently receive readings outside their target range. While the direction and causal nature of this relationship cannot be concluded from the current study, the hypothetical interaction of self-compassion, emotion regulation difficulties, and distress in a chronic illness population can be illustrated using the above situation as an example. In responding to this blood glucose reading with self-kindness, an individual may acknowledge that such experiences are stressful and reassure themselves that it is unrealistic to expect perfect readings every time. They may then find it easier to engage in positive emotion regulation strategies such as cognitive re-framing, reminding themselves that they have a plan to address such occurrences. From the perspective of common humanity, they may recognize that they are not alone in struggling with their illness management and reach out to sources of social support such as friends and family members. Supporting both the self-compassion and emotion regulation skills of this population may be a beneficial focus for future intervention. Increases in self-compassion and wellbeing, and decreases in distress, have been observed following self-compassion training in healthy adolescent (Bluth et al., 2016a , b ) and adult chronic illness populations (Friis et al., 2016 ). A meta-analysis by Kılıç et al. ( 2020 ) found a medium to large effect size of self-compassion interventions across adult samples with CMCs. Given these findings, it is promising that the efficacy of an online self-compassion program for young people with CMCs is under investigation (Finlay-Jones et al., 2020 ). Several considerations should be made when interpreting the findings of this study. Firstly, data was collected during the COVID-19 pandemic. Research suggests that young people with CMCs experienced decreased wellbeing during this period, potentially related to high susceptibility to the virus and uncertainty surrounding treatment delivery (Košir et al., 2020 ). This may be an alternative explanation for the low wellbeing and high distress observed in this sample. An additional consideration is that medical diagnoses were self-reported by participants. To increase sampling rigor, future research may involve medical practitioners in the recruitment phase. The use of a non-specific measure of symptom severity, while facilitating the inclusion of young people with a range of CMCs, may have decreased its validity as a control variable. An individual’s perception of symptom severity may be influenced by the nature of their condition and whether this report reflected their physical symptoms as a whole or a specific primary complaint. Future research may utilize a measure with multiple items, more specific guidance for respondents, or involve medical practitioners in the assessment of symptom severity. Further, the use of several self-report measures in this study introduced the possibility of common methods bias (Mackenzie & Podsakoff, 2012 ). This bias describes the artificial inflation of correlations between constructs completed by the same respondent in a single survey, due to priming effects, social desirability, and personal response styles (Mackenzie & Podsakoff, 2012 ). Accordingly, it should be noted that the true correlations between the variables in this study may be lower than those observed using this data collection method. A limitation of this study is that reduced item fit of the awareness subscale of the DERS-SF necessitated the deletion of these items. This limits our ability to link the mindfulness component of self-compassion with increased emotional awareness, as proposed in our theoretical model. Further, we remain unable to provide support for the argument that emotional awareness is a mechanism through which self-compassion is related to distress. Reduced item fit of the awareness subscale may be due to methodological limitations associated with negatively worded items, as it is the only subscale that requires reverse coding (Moreira et al., 2020 ). Alternatively, it has been suggested that awareness is not assessing the same underlying construct as the other subscales (Bardeen et al., 2012 ). This is consistent with findings from Subic-Wrana et al. ( 2014 ) that awareness is a precursor to emotion regulation, rather than an emotion regulation strategy (Bardeen et al., 2012 ). Future research may use an alternative measure of emotional awareness such as the Levels of Emotional Awareness Subscale (Lane et al., 1990 ), alongside the DERS-SF, to explore the relationship between the mindfulness component of self-compassion, emotional awareness, and emotion regulation difficulties. As the research design is correlational, it is not possible to infer causation from these findings. Randomized control trials implementing self-compassion programs for young people with CMCs are therefore required. It is also recommended that future intervention studies include emotion regulation difficulties as an outcome measure to further evaluate its role as a mediator between self-compassion, wellbeing, and distress. In addition, future studies investigating emotion regulation may benefit from using a wellbeing measure such as the Positive and Negative Affect Schedule (Watson et al., 1988 ), which may be more closely aligned with the affective dimension of subjective wellbeing. Finally, replication with a larger sample could investigate whether an indirect effect between self-compassion and wellbeing was not detected due to insufficient power. Replication following the COVID-19 pandemic is also required to evaluate the extent to which the findings can be generalized to other samples of young people with CMCs.