Abstract
BACKGROUND: The Modified Fresno Test has been used to evaluate the use of the Evidence-Based Physiotherapy (EBP). So far, none of the versions of the Fresno Test were subjected to analysis of the factorial structure. The objective of the study was to describe the exploratory and confirmatory factor structure of the Modified Fresno Test adapted to the Portuguese-Brazilian and analyze the statistical feasibility for the elaboration of a short version. METHODS: The questionnaire was applied with a convenience sample of 57 physiotherapists, being 36 professionals (13 of these also professors) and 21 students from the last semester of the physiotherapy course. Exploratory Factor Analysis (EFA) was performed by the method of principal components. Confirmatory Factor Analysis (CFA) was performed by the method of maximum likelihood. The total score of the answers in the test and retest was evaluated, totalling 228 observations. Reliability was assessed by means of internal consistency, using Cronbach's alpha coefficient. RESULTS: Reliability was satisfactory (α 0.81) for all questions of the instrument. The coefficient α calculated for the corrected item-total showed values higher than 0.20 except for item 9. Preliminary tests for Exploratory Factor Analysis showed acceptable values with Kaiser-Meyer-Olkin (KMO = 0.80) and Bartlett's test of sphericity [chi-square (78) = 1149.615, p < 0.001], indicating that the correlations were sufficient for analysis. The analysis revealed the presence of 3 factors (eigenvalues> 1), which explains 60.9% of the instrument's total variance. In Confirmatory Factor Analysis, none of the indices came close to an acceptable level (≥ 0.90), however, the second model which tested a three-factor structure provided a better fit to the data. From the results of this study the Modified Fresno Test short version was drawn. CONCLUSION: The analysis showed good factor validity and adequate internal consistency for the use of the instrument consisting of 13 questions and 3 factors. This model proved to be better than the original model. The short version consisting of 9 questions may be an appropriate alternative for use in the population of interest.