Cultural Adaptation and Psychometric properties of Brazilian and Peruvian versions of the Behavioral Activation for Depression Scale Short Form (BADS-SF)

Background: Behavioral Activation (BA) is an evidence-based treatment that aims to help the individual to stay active and reduce avoidance behaviors, as a means to reduce depressive symptoms. This study aims to describe the adaptation process and evaluate the psychometric properties of the Behavioral Activation for Depression Scale Short Form (BADS-SF) in its Brazilian and Peruvian version. Methods: Data were collected as part of a randomized trial with 880 participants in Brazil and 432 in Peru. The content validity was assessed using the Content Validity Index (CVI). Principal Component Analysis (PCA) method was applied to evaluate the factorial distribution. Sampling adequacy was assessed by Bartlett’s test of Sphericity and Kaiser-Meyer-Olkin measure. Cronbach’s alpha coecient was calculated to assess internal consistency. Results: CVI in Brazil was 0.92 and in Peru 0.87. The two-factor solution of the original scale is sustained (activation and avoidance), accounting for 50.6 and 54% of the total variance in Brazil and Peru, respectively. Cronbach’s alpha in Brazil was 0.55 and 0.66 in Peru for the overall scale. KMO was 0.769 and 0.790 for Brazil and Peru, respectively. Bartlett’s test of Sphericity had signicance of 0.000 for both samples. Conclusion: Both studied versions of the BAD-SF showed coherent structure and internal consistency. We recommend different distribution of the items into the subscales.

Depression is associated with poverty, low education, and social exclusion, signi cant factors in lowmiddle income countries (LMICs) [8], such as Brazil and Peru. These countries also face major challenges regarding access to treatment, particularly speci c groups such as indigenous, low income, and vulnerable populations meaning 80% of people cannot access any mental health treatment. The treatment gap exacerbates mental health burden and social impact [9].
Behavioral Activation (BA) is an evidence-based treatment for depression that seeks to increase opportunities for the depressed individual to get in contact with reinforcement contingencies that oppose the situation of con nement and reduction of social activities/contacts that contribute to the establishment and worsening of depression [10]. BA is often referred to as an effective and easy to disseminate treatment and holds promise to be highly adaptable for different settings and cultures, standing as a useful approach to address the gap in mental health care in LMICs [11,12].
To evaluate the effectiveness of BA several measures were used, but none of the preexistent scales focused on behavior functions and its idiographic nature [13]. The Behavioral Activation for Depression Scale Short Form (BADS-SF) comprises nine items distributed in two subscales, activation, and avoidance, both pillars of the treatment rationale [14]. The BADS-SF has been successfully adapted and validated in French, German, and Japanese.
This study aimed to adapt and evaluate the psychometric properties of the BADS-SF in its Brazilian and Peruvian versions.

Methods
The Brazilian and Peruvian versions of the BADS-SF were administered to participants as part of a randomized clinical trial (RCT) which evaluated the effectiveness of a BA intervention in the treatment of depressive symptoms through a mobile application. The study was developed and carried out by the Participants in Brazil (n = 880) were recruited from 20 Family Health Units (FHU), cluster-randomized into 10 intervention-units and 10 control-units with equal number of participants per unit (n = 44). In Peru, participants (n = 432) were recruited from three tertiary hospitals and four Primary Health Care Centers for the elderly (PHC) and individually randomized into a control group (n = 215) and an intervention group (n = 217). All participants had diabetes and/or hypertension, were 21 years or older, literate, and had a score of 10 or more on the Patient Health Questionnaire 9 (PHQ-9). Those who showed medium-high or high suicidal ideation in the PHQ-9 assessment were excluded from the study [15].
The process of cross-cultural adaptation of the original scale to Brazilian Portuguese was based on the procedures outlined by Beaton et. al. [16]. Two translators a liated to the RCT, one psychologist and one nurse, both uent in English, created two independent versions of the scale. Both versions were discussed by the translators and after resolving differences and ambiguities, were synthesized in a 3rd version. This version was back-translated into English by a third translator, native in English and uent in Portuguese, not related to the study, to verify the translation accuracy. Finally, a committee composed of ve health experts, Brazilian and uent in English, were requested to, individually, compare the adapted version of the scale to the original one, make suggestions and comment on its appropriateness, and compute two separate grades in a Likert scale ranging from 1 to 5 (1 -very bad; 2 -bad; 3 -regular; 4 -good and 5 -very good), concerning its idiomatic and conceptual equivalence [17].
Regarding the Peruvian process, As the BAD-SF is composed of 7 items taken from the BADS and one additional item [18], translation was needed for this additional item. The other items underwent idiomatic adaptations to expressions used with the Peruvian vocabulary. Two psychologists conducted this process, and the result was submitted to six judges, experts in mental health, for the same idiomatic and conceptual equivalence analysis that occurred in the Brazilian adaptation.

Statistical Analysis Assessment of the Content Validity
After the process of cross-cultural adaptation, the content validity of both versions of the BADS-SF was measured using the Content Validity Index (CVI) calculation. This index measures the judges' level of agreement on how similar the adapted version is regarding comprehensiveness and representativeness when compared to the original version [19,20].
First, the proportion of 4 (good) or 5 (very good) in each item was calculated for both categories, idiomatic and conceptual equivalence, a total of 18 grades per expert. After that, the CVI was calculated by the mean of this proportion. The CVIs were considered acceptable if the values are superior to 0.80 [21]. Assessment of the construct validity The construct validity of the scale was assessed by the Principal Component Analysis (PCA) method. This analysis allows us to con rm the factor analysis of the original scale or to propose a new model of factors for the Brazilian and Peruvian contexts. The Promax method was used for factor rotation.
Bartlett's test of Sphericity and KMO (Kaiser-Meyer-Olkin) measure was extracted to con rm sampling adequacy. Reliability assessment (Cronbach's alpha) Cronbach's alpha coe cient was calculated to assess the internal consistency of the items for each version of the scale. Values from 0.60 were considered acceptable, as observed in the literature [22].
Quality supervision of the study To ensure the quality of the study and data collection, the BADS-SF completion was facilitated by trained researchers. Participants completed the BAD-SF at the time of inclusion, after 3 months (1st follow-up assessment), and again at 6 months (2nd follow-up assessment).
The data were electronically collected and protected by encryption. Missing data were excluded using the pairwise deletion analysis approach [23].

Results
This study had 880 participants in Brazil and 432 in Peru. In both localities, participants were mainly female (86% in Brazil and 81% in Peru), aged between 41 and 60 years old (53% and 45%, respectively). Regarding depressive symptoms, 42% of Brazilian and 63% of Peruvian participants were classi ed as having moderate symptoms.
The majority did not present suicide risk (68% and 56%) and were not receiving treatment for depressive symptoms at the time of the baseline assessment (83% and 91%). Table 1 presents other baseline characteristics of the participants. ³Missing data represents responses not included in analysis due to the lack of essential information in the system.
The Content Validity Index in Brazil was 0.91 and in Peru 0.86, both considered acceptable. The CVI for the whole scale and items are presented in Table 2.  Table 3 shows the communalities of the BADS-SF items for both adapted versions. Item 3 (I engaged in many different activities) of the Brazilian version is the only item presenting a communality value lower than 0.4. The results of the factorial analysis calculated by PCA con rmed the existence of two primary factors for the BADS-SF in Brazil and Peru. In Table 4 we present initial and rotated eigenvalues for both versions of the BADS-SF. The naming of each factor followed the names for the original version of the scale. The rst factor, named "Activation", explained 30% of the variance of the Brazilian version and 32% of the Peruvian version. The second factor, "Avoidance", explained 20.6% of the variance in Brazil and 22% in Peru. Factor load scree plots are shown in Fig. 1 and Fig. 2.
The factor load of the items indicating their distribution around the extracted factors is demonstrated in Table 5. All items were higher than the adopted cutoff point of 0.4 [24].  The construct validity of the Brazilian and Peruvian versions of the BADS-SF was assessed using the Principal Component Analysis method (PCA), which resulted in a two-factor solution. This result echoes the ndings of the English version of the BADS-SF, and the naming of both factors, Avoidance and Activation, is sustained. However, the factor load of items demonstrated a different distribution of the items among the factors, placing item 1 (There were certain things I needed to do that I didn't do) in the avoidance subscale, which also occurred in one of the studies in the original validation studies and other languages validations [14]. This different organization of items into factors also increases Cronbach's alpha in both studied versions.
Cronbach's Alpha for the BADS-SF was 0.55 in Brazil, not a high internal consistency index according to the literature [22]. In Peru, Cronbach's Alpha was 0.66, within the range of reliability. However, very high measurements for the alpha do not necessarily represent a more consistent scale but may indicate redundancy in the construction of the items or an unnecessarily long scale, since the alpha value is affected according to the number of items on the scale [30]. The original validation study of the BADS-SF found =0.82 [14]. Other studies have found different values, ranging from 0.46 to 0.89, showing this measure might be controversial.
Hence, although the removal of item 8 (I engaged in activities that would distract me from feeling bad) could improve the overall alpha of the Brazilian scale, we decided to keep this item due to the size of the scale, lack of agreement in the literature of what would be a good alpha cut-off point, and because some items have a clinically relevant application and can be valuable in a clinical context to help follow up patient progress.
The same rationale was applied to the communality value identi ed for item 3 (I engaged in many different activities) of the Brazilian version of the BADS-SF, as literature points to communalities cut-off values ranging from 0.4 to 0.5 [20] and the item behaved as expected on the PCA.
In the present study and all cited studies on BADS-SF psychometric validity, Cronbach's Alpha was higher for the activation subscale than for the complete scale and lower for the avoidance subscale. Very often the avoidance subscale did not reach desirable reliability or was limitrophe, which complies with the original version authors' recommendations of using other tools if the objective is to measure avoidance.
Although Brazil and Peru have very distinct historical backgrounds, languages, and cultures, we found similar results in our psychometric analysis for both versions, indicating the adequacy of the scale adaptation. This study corroborates previous data showing the BADS-SF is a consistent measure to assess both activation and avoidance in research and clinical contexts in different cultures and settings [14,25,27,29,[31][32][33]. It is of utmost relevance and importance to present current results of using this scale with people with depressive symptoms in two different countries in Latin America.
Cronbach's alpha results and factor analysis allowed us to conclude we have an internally consistent version of the scales in both versions, however, with factor organization different from the original version. This also happened in other adapted versions of the scale [27,31].

Limitations Of The Study
Our ndings face some limitations. Both adapted versions were not submitted to face validity assessment with research participants, which hindered us from knowing if further adaptations and corrections were needed.
The population sample in this study is composed of chronic patients (those with diabetes and/or hypertension) in treatment for this condition in public health services in both countries. In Brazil, the study concentrates on patients of the eastern region of Sao Paulo, known for its low-income rate.
It may not represent the entire population of those localities.

Conclusion
Our psychometric results are acceptable and show both Peruvian and Brazilian versions of BADS-SF have a coherent structure and internal consistency. Considering both sites presented the same pattern of item adherence we suggest the following distribution of items into subscales: items 2 to 5 and item 9 -Activation subscale; item 1 and items 6 to 8 -Avoidance subscale.
This scale will allow clinicians to assess and follow-up on patients' activation and avoidance, as an evidence-based practice to treat depression in the general population.
Further psychometric studies should focus on assessing item by item contribution to the scale and test possible distortion of the measurement using psychometric tests such as the Rasch model approach or Written informed consent for participation was taken from all the study participants, which included authorization for participation in both the matrix and present study. The objectives of the study were described and explained to all participants. Participants were informed about the con dentiality of the data and about their right to withdraw from the study at any time.

Consent for publication
Not applicable.
Availability of data and materials The corresponding author agrees to make datasets used in this study available upon reasonable request.  Factor load scree plot demonstrating extracted factors of the Peruvian version of the BADS-SF.