Validação da versão em português do banco de itens função física do PROMIS® para a população brasileira

Abstract

Introduction: Instruments named Patient Reported Outcomes (PROs) are increasingly used to obtain and analyze data on health status, including the assessment of physical function. Instruments constructed according to item response theory (IRT) are available, allowing administration of the whole bank or groups of items, according to the need of the study, and the score obtained by a single individual is considered appropriate. The Patient-Reported Outcomes Measurement Information System project (PROMIS®) created a number of banks of items calibrated by TRI. To assess physical function, PROMIS® provides a bank with 121 items. This bank has recently been translated and culturally adapted to Portuguese and is available for validation. Purpose: To validate the Patient-Reported Outcomes Measurement Information System (PROMIS®) Physical Function Item Bank for the Brazilian population. Methods: The self-report item bank was administered on 1,000 participants. Data quality, reliability, and construct validity were assessed, in addition to determining the assumptions of one-dimensionality, local independence, invariance, and additivity, and the application of the Rasch model. For data quality were assessed floor and ceiling effects. Reliability was determined by the Person Separation Index (PSI). Construct validity was determined by confirmatory factor analysis (CFA), the one-dimensionality for the evaluation of the fit residuals of the Rasch analysis setting, the local independence by correlation CFA fit residuals, and invariance by analysis of the Differential Item Functioning (DIF) for sex and age. Results: The population is composed mainly for young (82% with age 18-45 yearsold), educated (86.1% with > 8 years) and healthy (85% without chronic diseases) people. Item bank presented excessive floor effect (116 items), and adequate reliability (PSI=0.95 to component 1 and 0.92 to component 2). MCA can be applied (KMO=0.811; sphericity=0.00). Two main components were identified (27.343 and 24.135), with explained variance of 22.8 and 20.1%, respectively. Both components were adequate for CFA (RMSEA=0.027, CFI=0.977, TLI= 0.978, χ²/gl = 1.31 for component 1 and RMSEA=0.064, CFI=0.934, TLI= 0.932, χ²/gl = 3.02 for component 2) and eight items with low construct validity (factor load < 0.4), fourteen with possible local dependency, and five with extreme scores were detected. However, none of them negatively influenced the model adopted. In conclusion, the Portuguese version of the PROMIS® Physical Function Item Bank proved to be valid and reliable, and adequate for use with the Brazilian population.