Statistical analysis

A descriptive analysis was used for the sociodemographic and clinical characterization of the participants. The floor and ceiling effects of PASE were calculated as the proportion of patients with the lowest and highest scores on the scale, respectively.

The reliability of the instrument was assessed by means of the correlation between items and the corresponding subscale and Cronbach’s alpha and intraclass correlation coefficient (ICC). Values over 0.7 and 0.75, respectively, were considered satisfactory [14,15].

The construct validity was evaluated by the transcultural translation and adaptation (transcultural validity), validity of known groups and structural validity. The structural validity was evaluated using confirmatory factorial analysis. For this purpose, the database of respondents was randomly divided into two equal groups (www.randomizer.com). In one group, a CFA was performed assuming that all PASE items had a factorial load of only one factor, which represents a structural equation modeling confirmatory strategy used to explain the relationship between items and the general factor or construct associated with PsA [16]. Weighted least squares with adjustments for the mean and variance (WLSMV) was the estimation method used. The fit indices analyzed were as follows: comparative fit index (CFI; ideal when > 0.90), root mean square error approximation (RMSEA; < 0.08 adequate fit, < 0.06 satisfactory fit), and the Tucker-Lewis index (TLI; ideal when > 0.90) [17,18].

The data corresponding to the second group were used for Exploratory Factor Analysis (EFA) by means of parallel analysis (PA). We applied polychoric correlations and unweighted least squares as a method to extract ordinal categorical data, with oblimin rotation for factor rotation and interpretation of factor loading.

Spearman’s correlation coefficient was used to calculate the item discriminant validity to establish whether the items in each factor exhibited satisfactory correlation. Patients with Pso with or without PsA were compared in the analysis of known-groups validity by means of the Mann-Whitney U test.

A ROC (receiver operating characteristic) curve was plotted to establish the cutoff point of the total PASE scores and its sensitivity and specificity, which indicate higher odds of arthritis associated with Pso.

The data were analyzed using statistical software SPSS 20.0, Factor 8.0, MPLUS 6 and MedCalc. The significance level was set to 5%.