The methodological process of the adaptation and validation of CYRM-28 for Brazilian version

The adaptation and validation process of CYRM-28 for Brazil was carried out according to the recommendations proposed by ^{Borsa, Damásio, and Bandeira (2012)} which are in line with the contemporary theories of validity (^{American Psychological Association/APA, 2020}). This involves two processes: 1) translation and adaptation of the original version of the CYRM-28 for the Brazilian youth population and 2) evaluation of the psychometric properties of the adapted instrument. Ethical approval was obtained by the Research Ethics Committee of the University mentioned (National Report No. 1,684,065 - CAAE: 57860916.10000.5208).

Study 1

Study 2

Results

As mentioned above, CFA using the original CYRM-28 three-factor model was performed, but it resulted in an inadequate fit (χ² = 1750, df = 347, p < 0.001, CFI = 0.785, TLI = 0.766, RMSEA = 0.069). This suggested that the standard three-factor solution did not work well in the current sample. Therefore, this was followed by an AFE, including site-specific questions, using a random selection from half the sample (n = 416). Before EFA, the assumptions of the appropriateness of the data for the analysis were confirmed (Bartlett test; χ² = 2814, df = 253, p < 0.001; Kaiser-Meyer-Olkin test = 0.87).

When requesting parsimonious factorial solutions with eigen values >1, a three-factor model was returned. In this model, several items loaded on factors with values less than 0.32 (items 7, 9, 22, 27 and 28 of the CYRM-28). They were excluded as their values were too low to be considered good indicators of the factor. The analysis was then re-run to review the factors without the impact of these items. Item 1 of the CYRM-28 and site-specific item S7 were also removed, as they loaded on more than one factor. Then, items 4, 10, 15, 16, 19 and 25 of the CYRM were excluded, in addition to item site-specific S5, as they also loaded lower than the minimum acceptable level. The analysis was re-run without these items and no further adjustments were required (Table 1).

Factor 1 brought together all of the items from the original CYRM "relationship with primary caregivers" subscale, except item 7 ("If I'm hungry, there's enough to eat"), which had been removed. Factor 2 brought together four items of the "individual" dimension, in addition to site-specific questions S6, S8 and S9, which deal with self-perception of overcoming capacity problems, hope for a better life and self-confidence, respectively. These additional items also seem to be related to personal qualities involved in resilience, so the name of the factor was kept. The third factor included items 3 and 23 of the original contextual subscale (importance of education and service to the community, respectively), items 2, 14 and 18 of the individual subscale (cooperation with others and support from friends) and items S1, S2 and S4 on specific issues (school as a resource for a better life, the desire for formal work and the existence of prejudice in the neighborhood). The "contextual" name of this factor was also maintained, being a mixture of the individual and contextual dimensions of the CYRM-28.

A CFA was then performed on the emergent model using the other half of the sample. Initially, the fit statistics were not promising (χ² = 426, df = 182, p < 0.001; CFI = 0.90; TLI = 0.88; RMSEA = 0.06). However, when reviewing the model parameters according to Tim ^{Brown's (2015)} and Rex ^{Kline's (2016)}, it was found that the standardized estimate of item S4 was not significant (p > 0.05) and the value of the standardized estimate of item S10 was low (<0.40); therefore, these items were removed from the model. Some items also had several standard residual values above .10 and were also removed: initially item 18 followed by item 14. Finally, covariance between the following pairs of items s was allowed: 2 - 23 and 2 - S1. These modifications resulted in good fit model (χ² = 321, df = 147, p < 0.001; CFI = 0.92; TLI = 0.91; RMSEA = 0.05, CI 90% = 0.05 -06) (Table 2). To distinguish this version from other CYRM iterations, we call this version CYRM-19-Br.

After confirming the model, the descriptive and reliability statistics of CYRM-19-Br are presented, together with those of CD-RISC-10, for comparison (Table 3). All Cronbach's alpha coefficients were above .7, except for the contextual subscale, which was slightly lower than desired (α = 0.67). The correlations between items were appropriate for all subscales and comparable to those detected for the CD-RISC-10.

To establish convergent validity, the association of scores in the subscales of CYRM-19-Br and CD-RISC-10 were examined. All were positively correlated (p < 0.01), as expected, given both are positively worded measures of resilience (Table 3). The correlations were of appropriate strength (neither too weak nor too strong; r = 0.27-0.67). The individual factor was more strongly correlated with CD-RISC-10 (r = 0.67, p < 0.01), which makes sense, given that CD-RISC-10 strongly focuses on individual qualities associated with resilience. The subscales of the CYRM-19-Br also correlated appropriately with each other, again, not too weakly indicating they were measuring distinct constructs and not so strongly that they suggested problematic overlap (r = .42-.43).

Considering the variables sex, age and income, it is possible to state that statistically significant differences are found for each subscale (p < 0.05), with the exception of sex in the individual dimension (t = 0.57, p > 0.05).

Discussion

The objective that CYRM-28 could be adapted to the Brazilian reality was achieved, resulting in a robust measure for a variety of children and youth in the country. The alterations to the measure led us to distinguish it by renaming it the CYRM-19-Br.

Overall, the CYRM-19-Br has demonstrated evidence of various kinds of validity, such as structural validity, although the factor structure does not correspond to the structure initially proposed, though this is expected, due to the cultural specificities of the studied sample (^{Jefferies et al., 2018}; ^{Van Rensburg et al., 2017}). Robustness is also encouraged based on the observed correlation pattern of the factors within the measure and to a comparable measure of resilience, and reliability based on Cronbach's alpha values (only one not excellent, but acceptable).

The fact that the model found in the present study is not identical to others found in adaptation studies of the CYRM-28 is not problematic, as the specific factorial structures are likely to reflect aspects of the cultural context (^{Langham et al., 2018}). For instance, ^{Govender, Cowden, Oppong Asante, George, and Reardon (2017)} sought to validate the CYRM-28 measure among a sample of South African adolescents (n = 1854) with an average age of 14.88 years. After comparing the adjustment indices and the standardized factorial item loads for each model, their confirmatory factor analysis arrived at a 24-item model composed of 3 factors. In the study by ^{Zand et al. (2017)}, the scores of the exploratory and confirmatory factor analysis resulted in 11 items distributed across three resilience factors, which presented satisfactory Cronbach's alphas.

In the CYRM-Br, Factor 1 brought together four items from the "individual" dimension of the original scale (03 from the personal skills cluster and one from social skills cluster), in addition to site-specific questions S6, S8 and S9, which deal with self-perception of overcoming capacity problems, hope for a better life and self-confidence, respectively. These additional items also seem to be related to personal qualities involved in resilience, so the name of the factor was kept. As it is an individual dimension, it can be said that this reflects the way young people consider their possibilities to deal with society in general. In this sense, the focus is less on "peers" and social skills and more on the possibility of looking for ways to negotiate and navigate with the world around; as can also be seen in studies by ^{Ungar (2019)}, ^{Ungar et al. (2020)} and ^{Ungar, Theron, Murphy, and Jefferies (2021)}.

Factor 2 contained all the items about "relationship with primary caregivers" from the original CYRM, and only item 7 ("If I am hungry, there is enough to eat") was removed. This raises the question of the relevance of food scarcity and or the difficulty of admitting this, which could lead to a "more stigmatized view" of individuals surrounded by scarcity, privation and "deficiency" (^{Brasil, 2019}; M. L. F. ^{Silva, 2019}; ^{Bezerra, 2015}). In Brazilian society, respect, protect and fulfil the rights of displaced children and youth starts acknowledging what was established at the Child and Adolescent Statute (ECA), for example. Unfortunately, it has been practiced a perverse policy (^{Takeuti & Bezerra, 2009}) in which youth are encouraged to consider social problems as their own. But, on the other hand, the misconception that youth's "empowerment" depends on individual efforts has been strongly criticized by researchers (^{Bezerra, 2021}; ^{Tommasi & Corrochano, 2020}).

The third factor in the emergent measure was named "contextual", since it addresses school, work and spirituality, aspects which have been highlighted in the literature as important resilience resources (^{Brasil, 2019}; ^{Francisco & Coimbra, 2015}; ^{Ungar, 2011}, ²⁰¹⁹). Interestingly, of the five items most clearly related to culture, none. Youth point out the relevance of the S2 'signed employment contract'. Youth also links culture to financial stability. Citizenship in a culturally significant way is reduced to have a formal contract, reflecting the precariousness of the world of labor. Even so, it is still possible to recognize the importance of school, although a meaningful way of accessing a horizon of possibilities is not identified in it (M. L. F. ^{Silva, 2019}). The third factor brought together items 2 and 23 from the original contextual subscale (importance of education and service to the community, respectively), items 2, 14 and 18 from the individual subscale (cooperation with others and support from friends) and items S1, S2 and S4 (school as a resource for a better life, the desire for formal job and the existence of prejudice in the neighborhood). The cultural subscale of the CYRM-28 reveals the crisis in the current context of youth in Brazil, the difficulties to foresee pathways of the future, youth's feeling as been "lost" due to high mortality rates and non-existent access to leisure and jobs, crucial points of young people's life (^{Cerqueira & Bueno, 2020}). As for the items on spirituality, young people consider spirituality more focused on caring for/helping others, differentiating themselves from the traditional perspective of attending temples, etc. The distribution among religious denominations, as well as a perceptual of youth who call themselves without religion are in accordance with the literature in the area (Instituto Brasileiro de Geografia e Estatística, 2010; ^{Jahn & Dell'Aglio, 2017}).

There were differences between certain demographic subgroups completing the measure that merit further scrutiny. Such differences indicate the CYRM-28's focus on socio-ecological dimensions of resilience (^{Ungar, 2019}; ^{Ungar et al., 2020}; ^{Ungar et al., 2021}), which are likely to lead to variations in scores. ^{Langham et al. (2018)} worked with youth from the middle-east of Australia and discovered that men had a higher score in the item "to be treated fairly" while women differed in the items "proximity to parents (caregivers)", "pride of indigenous culture" and "friends, supporters". These data corroborate the findings of CYRM-19-Br, however, they also warn of its limitations regarding the investigation of specific themes, for example, including issues such as cis/transgenerity (F. C. ^{Silva, Souza, & Bezerra, 2019}) in studies of resilience.

Regarding thresholds for the as is often desirable with (as is often desirable with measures in order to know what is "good" or "good enough") the CYRM-19-Br follows formulation of predecessors, where what is "good" would depend on the contextual reality of each group studied. While this remains to be determined, score scan be contrasted within or between samples, i.e., high scores can be contrasted with low ones and potential reasons for this discrepancy must be formulated and investigated (^{Resilience Research Center, 2018}).

Finally, considering the design of research on resilience proposed by ^{Ungar (2019)} and ^{Ungar et al. (2020)}, the CYRM-19-Br meets the indicated requirements. In this sense, the measure focus on protective factors that can and should be linked to government programs.

In terms of limitations, the scale has not been tested in indigenous communities, quilombolas, riverside communities in the Amazon region, gypsies, nomadic groups such as circus residents, young people diagnosed with mental disorders, young refugees living in Brazil and not adapted for youth of special needs (e.g.: deef, blind). There was no sample from socioeconomic strata A and B used by the IBGE, which corresponds to the segment with the highest purchasing power in the country. Further discussions regarding the limits and possibilities of using resilience scales need to be considered. Such limitations also provide the opportunity for future studies to address each of the aspects mentioned.

Conclusion

The CYRM-19-Br scale fits criteria for validation and psychometric adaptation and is therefore indicated for the use of research on resilience in Brazil with youth, filling a gap in this field of studies.