Psychometric Validation of the Brazilian Portuguese Version of the Derriford Appearance Scale-24 (DAS-24) for People Living with HIV/AIDS

The changes in appearance of people living with HIV/AIDS (PLHA) interferes with how people around them react to their body, how social interactions take place, and how each person perceives and accepts their body. The definition of itself can be severely challenged when the body changes as a result of illness and the person does not look healthier anymore. People living with HIV/AIDS (PLHA) are an especially vulnerable group when it comes to “distress” and the psychosocial impact of appearance, yet the assessment of body image changes in these people was subjective in Brazil. The aim of this paper was to assess the psychometric properties of the Brazilian version of Derriford Appearance Scale 24 (DAS-24) for a sample of Brazilians living with HIV/AIDS. A sample of 400 patients were recruited from an HIV/AIDS ambulatory, aged between 18 and 78 years, of both sexes. The psychometric properties of DAS-24 were investigated while using confirmatory factor analysis (CFA), with unweighted least square estimation and listwise deletion for missing data. The adjustment of three structural models previously established for DAS-24 (single-factor, two-factor, and three-factor) was investigated. Evidences of construct validity—convergent and discriminant—and internal consistency—Cronbach’s alpha and construct reliability—were also generated for the measure model. The results showed that the one-factor model had the best adjustment, after eliminating items 8, 17, and 20, and accepting the covariance of errors between items 4 and 10; 9 and 23; 11 and 14; and, 14 and 22. Additionally, validity and reliability evidence were satisfactory for the model. The Brazilian Portuguese version of DAS-24 seems to be a psychometrically sound scale for measuring body image distress for people living with HIV/AIDS (PLHA).

systematic assessment of interventions that are made in the treatment system is important for assessing their effectiveness in the patient's overall health.

The Present Study
Among the instruments that evaluate "distress" and body discomfort, the Derriford Appearance Scale (DAS) stands out [27]. DAS was originally designed by researchers in the United Kingdom to be an objective measure of the spectrum of psychological stress and body image dysfunction in the characteristic aspects of disfigurement, deformities, and aesthetic problems [28]. Its original formulation consists of 59 items (DAS-59), organized into five factors-awareness of general appearance; awareness of social appearance; sexual and bodily awareness of appearance; negative self-concept; and, awareness of easy appearance. DAS-59 has an introductory section that allows the subject to identify and describe which aspect of his appearance he is most sensitive or inhibited and this is referred to as his "reference" to the responses of the scale items. Fifty-seven items of the scale evaluate psychological distress and psychological dysfunctions, and two items evaluate physical distress and physical dysfunctions [28].
Because it was thought to be used in research and clinical practice, its psychometric studies extended to subjects of both sexes, of all age groups from 16 years, healthy or not. Attention should be paid to the specific audience to whom it is addressed: to people who have had their body parts disfigured and wait for reconstruction surgery or who are already under treatment [29].
Five years after the creation of the DAS-59, a short version with 24 items was proposed DAS-24 [27]. This short version was elaborated for clinical use and routine follow-up of people with these disorders [27]. The DAS-24 items were removed from the DAS-59, and their choice was given by both what they could observe-concerns and behaviors regarding appearance-and by their psychometric properties-factor load, correlation with the total scale score-and clinical utility. The initial psychometric study of DAS-24 used a sample of 518 participants, of both sexes, between 18 and 70 years. The appearance issues in the sample were due to the trunk, sex organs, upper limbs, lower limbs, face, head, or neck. Multivariate psychometric analysis was not performed-exploratory or confirmatory factor analysis. However, in terms of reliability, in this study the DAS-24 showed good temporal stability in the re-test with an interval of six months (r = 0.82). In terms of evidence of validity, satisfactory evidence of concurrent validity was generated with DAS-59 (r = 0.88), discriminant validity with the measurement of positive affect (r = −0.24). Finally, evidence of convergent validity was also generated with measures of anxiety (r = 0.50), depression (r = 0.45), social avoidance (r = 0.53), fear of negative evaluation (r = 0.50), negative affect (r = 0.50), and shame (r = 0.66) [27].
A second psychometric study, with multivariate methods of analysis, was conducted later in United Kingdom [30]. The total of 1265 participants-614 in the community and 651 in hospital treatment-of both sexes aged between 18 and 91 years participated in this new study. Exploratory factor analysis was performed for a subsample of 500 participants, while using axis factoring extraction and oblique rotation. The exploratory factor analysis generated a bi-factorial structure, with factor 1, with 18 items, eingenvalue of 9.73-9.98, and factor 2 with six items, eingenvalue 1.85-1.91, the others being below 1, both in clinical and non-clinical samples, respectively [30]. Confirmatory factor analysis was then performed with (1) the rest of the sample; (2) only the clinical sample; (3) only the non-clinical sample; and, (4) a pool of 500 participants that were randomly selected from the entire sample. In the four cases, the bi-factorial model was confirmed. The mean of the Root Mean Square Error of Approximation (RMSEA) value was 0.06; Goodness of fit index (GFI) = 0.86; Adjusted Goodness of Fit Index (AGFI) = 0.83; Normed Fit index (NFI) = 0.84; Non-Normed of Fit Index (NNFI) = 0.86; and, Comparative Fit Index (CFI) = 0.88. Cronbach's alpha values ranged from 0.93 to 0.79 between the factors and samples [30].
DAS-24 has already been validated in Thailand [31] and Portugal [32]. In Thailand [31], the psychometric study used a sample of 208 participants, with visible changes in appearance, of both sexes, aged 19−80 years. The preliminary analysis of the data resulted in the elimination of items 14, 17, and 22 because of their asymmetry, and items 1 and 8 for their low correlation with the total score of the scale. Exploratory factor analysis was then performed, using main component analysis with oblique and varimax rotations as extraction. The factor solution of the DAS-24 Taiwanese version, now with 19 items, has three factors-social distress, social avoidance, and negative affect-generated by the main components analysis with oblique rotation. Cronbach's alpha value was only calculated for the instrument as a whole-unlike the recommendations [33]-and a value of 0.91 was found.
The validation study of DAS-24 in Portugal [32] was conducted with a sample of 508 participants with no specific clinical condition, between 17 and 89 years old, of both sexes. Exploratory factor analysis was performed with main components analysis with varimax rotation, forced for one factor in order to verify the correlation pattern between the items. Items 8, 15, 20, and 24 presented weak (0.50) and moderate (>0.50 to 0.80) correlations. The confirmatory factor analysis, performed with the same sample used in the previous analysis-unlike the recommendations [33], presented an initial adjustment that was tolerable-χ2/df = 4.7; RMSEA = 0.09; GFI = 0.80; CFI = 0.79; NFI = 0.77 for the one-factorial model. The adjustment of the model was due to the acceptance of several communalities, which enabled the preservation of all items. However, it should be noted that some items had very low factorial load, such as item 15 (λ = 0.15), item 8 (λ = 0.21), item 20 (λ = 0.21), and item 24 (λ = 0.33). In addition, the acceptance of error covariance foresees theoretical reason that is linked to the mathematical advantages to fit the model. What can be observed in the authors' decisions was leniency in this criterion-for example, the acceptance of the "feeling rejected" item 14 and item 18 "worrying about not wearing the favorite clothes. With these modifications, satisfactory adjustment was achieved, χ2/df = 2.5; RMSEA = 0.05; GFI = 0.91; CFI = 0.93; NFI = 0.91.
Our literature review indicated that there was a cross culturally adapted version of DAS-24 into Brazilian Portuguese, but no psychometric assessment was carried out [34]. The psychometric validation of this scale for PLHA will make it possible to know the nature, extent, and determinants of individual differences in disease impact and diagnosis. It may also allow different health professionals to monitor the impact of their interventions on the affective dimension of body image in PLHA, adjusting and adapting them in order to make them a resource for the restructuring of body identity.
Recognizing the importance of having an outcome measure to allow for tracking changes of body image during treatment and for having the methodological gap of a valid instrument and appropriate to obtain such a measure, the aim of this research is to assess the psychometric properties of the Brazilian Portuguese version of Derriford Appearance Scale 24 (DAS-24) with a reference sample of PLHA.

Study Design and Sample Size Calculation
This is a methodological study with a non-probabilistic sample [35] having, as target population, Brazilian people living with HIV/AIDS. The sample size was calculated according to the recommendations of Hair et al. (2009), which recommend that the sample size should be five participants per assessed parameters. When considering that the largest model of DAS-24 tested has 51 parameters (24 observable variables; 24 errors; three first-order latent variables), the minimum sample size for this research should be 255 participants.
The participants that were included in the study met the following criteria: being over 18 years of age, being a HIV seroreagent individuals, and being under regular treatment for HIV/Aids. Comorbidities were not screened. Patients returning from treatment after a period longer than six months of drop-out were excluded.

Instruments
Brazilian Portuguese version of Derriford Appearance Scale 24 (DAS-24) for people living with HIV/AIDS in Brazil [34]. DAS-24 is originally from the United Kingdom and it was created in order to evaluate the distress and difficulties of those who have changes in appearance [27]. Items 1,4,8,10,11,14,17,20, and 22 are arranged on a four-point Likert scale (1 = not at all, 4 = extremely; 1 = never, 4 = always) as are items 2, 3,5,6,7,9,12,13,15,16,18,19,21,23, and 24 on the which only adds the "not applicable" option (with zero score). There are two additional items, which are not included in the structural model or in the sum of the score, but that provide information regarding the intensity of pain and the frequency of physical limitations that the respondent may have due to the change in their appearance. The higher the mean score, the greater the anguish and concern for appearance. [27]. The Brazilian Portuguese version of DAS-24 was made according to the guidelines that were recommended by Beaton, Bombardier, Guillemin, and Ferraz (2002) [36], having PLHA individuals on their sampling.
Demographic Questionnaire: specially designed for this research and only consists of the identification of biological sex, gender, age, marital status, and ethnicity.

Procedures and Ethical Aspects
This study was approved by the Research Ethics Committee of the ABC School of Medicine and it is registered under number: 092257/2014. Patients from the HIV/AIDS ambulatory of São Bernardo do Campo were invited to participate in the research in an individualized approach, by the professionals of the nursing team with which they were familiar and felt confidence. Only those who agreed to participate voluntarily were given the informed consent form (ICF). The questionnaires were answered by the participants, in the outpatient clinic, in a private space. The participants individually completed the instruments, taking between 15-25 min. to complete.

Investigated Models
In this research, model fit was investigated in three structural models, previously established for the DAS-24. The first model is the one-factor solution established the study of the development of the instrument [27] and ratified in the Portuguese version of the instrument (Mendes et al., 2016). In this model, low factor loadings for items 8, 15, 20, and 24 were noted.

Statistical Analysis
The PRELIS TM 2 module of the LISREL ® system was used to prepare the data for the Confirmatory Factor Analysis (CFA), adopting the listwise deletion criterion for missing data [37]. After generating the PRELIS TM 2 file, the SIMPLIS model, which is a coding of the LISREL ® system, was used. The CFA was then generated, enabling the estimation of the parameters of the construct measurement model. Given that the data distribution was not in line with the normal multivariate distribution, the Unweighted Least Square extraction method was used, because it was not sensitive to this violation [38].
In the analysis of the measurement model, the construct validity (discriminant and convergent) and internal consistency were considered. In order to establish the internal consistency, Cronbach's alpha (α) and construct reliability (CR) were calculated and values that were above 0.70 were considered to be acceptable [33,39]. The CR is given by the formula: (square of the sum of the factor loadings)/(square of the sum of the factor loadings) + (sum of the errors of the observable indicators).
In order to establish convergent validity, factor loadings and the Average Variance Extracted (AVE) were considered. The AVE is given by the formula: (sum of the square of the factor loads)/(sum of the square of the factor loads) + (sum of the errors of the observable indicators). Values that are above 0.50 were accepted (Hair et al., 2009) and factor loadings (λ) above 0.50 are preferable and above 0.40 are acceptable (Hair et al., 2009).

Statistical Analysis
The PRELIS TM 2 module of the LISREL ® system was used to prepare the data for the Confirmatory Factor Analysis (CFA), adopting the listwise deletion criterion for missing data [37]. After generating the PRELIS TM 2 file, the SIMPLIS model, which is a coding of the LISREL ® system, was used. The CFA was then generated, enabling the estimation of the parameters of the construct measurement model. Given that the data distribution was not in line with the normal multivariate distribution, the Unweighted Least Square extraction method was used, because it was not sensitive to this violation [38].
In the analysis of the measurement model, the construct validity (discriminant and convergent) and internal consistency were considered. In order to establish the internal consistency, Cronbach's alpha (α) and construct reliability (CR) were calculated and values that were above 0.70 were considered to be acceptable [33,39]. The CR is given by the formula: (square of the sum of the factor loadings)/(square of the sum of the factor loadings) + (sum of the errors of the observable indicators).
In order to establish convergent validity, factor loadings and the Average Variance Extracted (AVE) were considered. The AVE is given by the formula: (sum of the square of the factor loads)/(sum of the square of the factor loads) + (sum of the errors of the observable indicators). Values that are above 0.50 were accepted (Hair et al., 2009) and factor loadings (λ) above 0.50 are preferable and above 0.40 are acceptable (Hair et al., 2009). Discriminant validity was assessed by comparing the shared variance with the AVE of each latent variable. Evidence of discriminant validity is established when a AVE is ≥0.50 and greater than the shared variance between factors [40].
Finally, correlation analyses were performed between the DAS-24 scores and the participants' age, and differences in the DAS-24 score in relation to gender and gender role were analyzed, in order to generate some additional information. The part of the body that bothered them, the type of lipodystrophy, and the other body feature that bothered the participants other than the one first described. The discrete variables of this analysis (age and DAS-24 score) were adherent to the normal distribution; hence, Pearson's correlation test and Student's t-test were used. For all tests, a 95% confidence interval was considered. SPSS 21 and LISREL ® software were used for these analyzes.

Factorial Structure
The first estimation of model one showed adequate adjustment indices, except for RMSEA,    Table 1).
When considering all of the models after their fit, model two had the best structural fit, and model three the worse. However, for model two, the covariance of its factors was strong, having no statistical evidence that they are really distinct of each other regarding then (AVE < r 2 ). Thus, the decision is to adopt the model one as a structural model for the Brazilian Portuguese version of DAS-24.

Converged Validity and Internal Reliability of Model One
Regarding the convergent validity, when considering the AVE, it was below the acceptance limits (AVE = 0.46). However, all of the items were above the acceptable value, ranging from λ = 0.40-0.81. The construct reliability (CR = 0.94) and Cronbach's alpha (α = 0.94) values were above the reference value (α and CR ≥ 0.70). Discriminate validity was not evaluated, since Fornel and Larcker's criteria does not apply for this model

Extra Items
Through Pearson's correlation test, it was possible to verify that the two extra items ("The characteristic of my appearance that bothers me causes me physical pain/discomfort" and "The characteristic of my appearance that bothers me causes physical limitations to do what desire") have a high positive association with the scores of the Brazilian version of DAS-24, r = 0.61 (p < 0.05) and 0.68 (p < 0.05), respectively.

Discussion
The aim of this paper was to assess the psychometric properties of the translated and culturally adapted version of Derriford Appearance Scale, with 24 items (DAS-24) [10,34], for a Brazilian sample of PLHA.
Of the three models tested, the one that concomitantly obtained the best structural adjustment and best fit of the measurement model was the one-factor model, after eliminating items 8, 17, and 20 and accepting the covariance of errors between items 4 and 10; 9 and 23; 11 and 14; and, 14 and 22. It is worth mentioning that the changes made do not imply the loss of the quality of the measure. The elimination of items in cross-cultural methodological studies is not uncommon, including attitudinal measures in health [41]. Specifically, for DAS-24, items 8 and 20 were also eliminated in the methodological study that was conducted in Portugal [32].
It is unlikely to have occurred a bias on the items during the cross-cultural adaptation process that may have caused this elimination, since the work was thoroughly conducted and accompanied by the original author of the instrument, as reported by Martins et al. (2019). It is also noteworthy that the cross-cultural adaptation study for Brazil also pointed out that item 8 was potentially fragile, due to its low correlation with the total score of the instrument [34], a fact that does not make its elimination strange. It is more plausible that cultural differences have led to this situation: specifically, it is possible that there are some aspects of PLHA body experiences that are specific to Brazil and that may have an impact on how they deal with changes in appearance, due to the chronicity of the disease and virus activity. From this perspective, the eliminated items could be considered as less relevant to the construct under study, or rather could be observable variables that are unexplained for the latent variable. Indeed, previous studies have suggested discrepancies between scales for assessing body image developed in the Western and translated into Brazilian Portuguese [42].
As for the covariances of errors that are accepted in this model, common causes include item redundancy (caused by similar content or social desire) and/or omission of an exogenous factor [43]. The acceptance of the covariance error should be theoretically supported rather than a purely statistical reason, for example, to improve model fit [38,44]. Having this clear criterion helps to avoid creating model caricatures and ignore the logic of confirmatory factor analysis. Social desirability is unlikely to be a factor for the covariance of error, as the conditions for data collection to reduce this bias (voluntary and anonymous participation) were guaranteed. It is also unlikely that an ignored latent variable will be absent in DAS-24, since the analyzed models have already been proposed and investigated by previous studies and the items of the scale already thoroughly investigated. It was the content analysis of the paired items that led us to consider that the existence of error covariance was due to a similar content between items.
Weak evidence of convergent validity of the measurement model was generated, but still relevant. Although the average variance extracted was within the citation limit, the factor loadings were adequate, which indicated that the items inserted in the single factor were coherent with each other. It was also observed that the additional analyzes made a point to discriminant validity of the measure, since they indicated differences between biological sex and gender identity, as predicted by the body image's literature [45,46]. The significant difference between the scores of participants who have some feature that bothers them and the one who does not was also relevant evidence, since it demonstrates that the measure can distinguish distress and beliefs caused by these changes in appearance, which becomes the theoretical essence of this measure [27,28].
The non-statistical significance of the impacts of different appearance changes leads us to resume the stigmatization that is generated by the signs of chronicity of the action of the HIV virus and disease [14]. What our evidence complains is that it does not matter what the change is (whether on the face, trunk, limbs, back, or other body location), but that it may be associated with HIV/AIDS. Finally, it is commented that the non-association of the score of the Brazilian Portuguese version of DAS-24 with age leads us to consider that distress and beliefs that are generated by appearance alteration are independent of age; thus, both young and old. Older people need to receive the same degree of support from the care service in order to promote the positive resignification of HIV/AIDS.
The present study validated DAS-24 for a specific clinical group of Brazilians: people living with HIV/AIDS. Despite the evidence psychometric evidence generated here, a number of limitations must be considered. The opportunistic method of participant recruitment means that the current results should only be considered for the reference population of the instrument. On the other hand, this study presents contributions to the improvement of DAS-24 by the possibility of further cross-cultural studies. Because appearance is considered to be an important part of the Brazilian body image [47], it would be interesting to have DAS-24 validated for other groups in which appearance changes, such as burns, vitiligo, psoriasis, with progressive hemifacial atrophy (Parry-Romberg syndrome), among others. As recruitment was done on an outpatient basis with limited data collection time, future studies could also examine the extent of relationships and associations of body image impact assessed by DAS-24 with other relevant constructs, such as body dysmorphia, acceptance of plastic surgery, social physical anxiety, quality of life, and physical attractiveness in the Brazilian context.

Conclusions
A satisfactory fit was achieved for the original model with the reference sample of PLHA, with weak evidence of construct validity and evidence of internal consistency for the Brazilian Portuguese version of DAS-24.