ABC | Volume 114, Nº5, May 2020

Original Article Silva et al. Florida Shock Anxiety Scale Arq Bras Cardiol. 2020; 114(5):764-772 from the association of a bootstrap with sample extrapolation to 5,000. Factor extraction was done initially with Robust Unweighted Least Squares (RULS), which reduces the matrix of residuals. 29 Item Response Theory Item discrimination index was used (a), which measures the association strength between the item and the latent variable, and whose interpretation is similar to factor loading in the exploratory factor analysis. Quality parameters of the translated and adapted versions of the FSAS To adequate the items and the models, the following criteria were taken into account: the explained variance of the model (60 to 70%), factor loading values (> 0.50), communalities (> 0.40) and item discrimination, and collinearity and multicollinearity problems (factor loads ranging from 0.80 to 0.85). Indices of adjustment obtained in the Confirmatory Factor Analysis The model adjustment indices and their respective expected values were: Robust Mean-Scaled Chi Square/df NNFI (Non-Normed Fit Index > 0.93), CFI (Comparative Fit Index > 0.94), GFI (Goodness Fit Index > 0.95), AGFI (Adjusted Goodness Fit Index > 0.93), RMSEA (Root Mean Square Error of Approximation < 0.07) and RMSR (Root Mean Square of Residuals < 0.08). 29-31 Reliability Three indicators were adopted to assess the reliability of the Brazilian version of the FSAS questionnaire: Coefficient Alpha (“Cronbach’s Alpha”), Omega and the Greatest Lower Bound (GLB). Results The final version of the FSAS The stages of the translation and cross-cultural adaptation resulted in similar versions of the FSAS instrument. The synthesis of the translations was quite concise and combined the most coherent elements of each translation. The back-translations confirmed the good quality of the translations and the synthesis process carried out in the initial stages. A total of 20 ICD patients, with a mean of age 55.6 ± 6.8 years, took part in the pretest. Of these, 50% were female, 50% were white and 30% had studied up to High School. All participants reported that the items were relevant, easy to understand and that the response options were clear. No modifications in the instrument were required. Table 1 shows the instrument items in its English and Portuguese versions. Psychometric properties of the FSAS Population composition In this stage of the study, 151 ICD patients, with a mean of 55.7±14.1 years (range, 19- 80 years), were included. There was a male sex predominance, which corresponded to 64% of the cases. Most patients were white (85.4%) and 49% had attended Middle School (Table 2). Among the cardiac diseases, there was a predominance of Chagas disease, which was present in 30.5% of the cases, followed by ischemic cardiomyopathy in 25.2%. Brugada syndrome and congenital long-QT syndrome (LQTS) were identified in 4.6 and 3.3% of patients, respectively. Baseline assessment showed that most patients were in the New York Heart Association (NYHA) functional classes I (37.1%) and II (47.7%). Left ventricular function was determined by bidimensional transthoracic echocardiography and ranged from 18 to 77%, with a median of 35%. Only 29.1% of the patients did not present any associated comorbidities. Dyslipidemia and arterial hypertension were the most frequent comorbidities, being present in 51.4% and 49.5% of patients, respectively. Atrial fibrillation was present in 27.1% of the individuals studied (Table 2). As expected, 80.1% of the indications for ICD were r secondary prophylaxis of sudden cardiac death. In Brazil, due to lack of resources, ICD implantation is still underused for primary prophylaxis of sudden cardiac death. Descriptive analysis of the FSAS items Through descriptive analysis of the instrument items, it was possible to identify that normality of distribution was violated, indicating, therefore, the need for polychoric correlation, instead of Pearson’s correlation coefficient. The means of the instrument items ranged from 1.5 to 2.9. The FSAS average score was 22.8 ± 11.1, with a median of 20 points and variation of 10 to 50 points. There was no impact of extreme values on the mean (Table 3). Construct validity and dimensionality of the FSAS The values obtained from the Kaiser-Meyer-Olkin index (KMO= 0.88), the Bartlett’s sphericity test (X²= 565.5, df= 45; p<0.001) and the matrix determinant (0.0206 (p<0.0001)) revealed a significant correlation between the items, which confirmed the adequacy of the EFA. The parallel analysis indicated the existence of only one dimension for the instrument. Moreover, this item set can explain 64.4% of latent variable (above the values recommended in the literature). 29-31 The eigenvalue criteria also indicated only one dimension, with a an eigenvalue of of 6.08. The fact that the instrument was unidimensional waived requirements for methods of matrix factor rotation. Unidimensionality indicated the use of the normal-ogive graded response IRT model, which is more adequate for a unidimensional polytomous model. 31 767