[Validity evidence of the Hospital Birth Satisfaction Scale using the Item Response Theory]

With the objective of generating evidence of validity for a measure of the latent trait "level of satisfaction with childbirth", Item Response Theory (IRT) was used in a sample of 15,965 women who had hospital births in Brazil. Telephone interviews were conducted up to six months after childbirth with participants of the Birth in Brazil: National Survey into Labor and Birth in the five regions of the country. An 11-item questionnaire denominated the Hospital Childbirth Care Satisfaction Scale was used. The questions were incorporated into the IRT models, with Samejima's graded response model providing the best fit. The sample was randomly divided into dataset 1 (n = 7,982), adjusted using the graded response model, and dataset 2 (n = 7,983), for validation of the results. Reliability measures were Cronbach's alpha, McDonald's omega, and Spearman-Brown coefficient. Regarding construct validity, exploratory factor analysis of dataset 1 demonstrated a latent factor structure with three dimensions. The resulting structure was tested in dataset 2 using confirmatory factor analysis, obtaining a good fit of the model. Two items of the questionnaire were excluded from the measure for not discriminating the latent trait of interest. Thus, thefinal scale consisted of nine items. Based on IRT, a new version of the instrument with adequate psychometric properties was obtained. The associations between satisfaction with childbirth and external variables strengthen the validity of the scale.