Rasch Analysis of the Power as Knowing Participation in Change Tool

Monday, July 11, 2011: 10:35 AM

Erika de Souza Guedes, RN1
Mariana Alvina dos Santos, RN, MNSc1
Fabiana Gonçalves Matos, RN, MNSc2
Michelle Mendes Trindade, RN1
Luis Carlos Orozco Vargas, MD, MEpi3
Dina de Almeida Lopes Monteiro Cruz, RN, PhD4
(1)School of Nursing, University of São Paulo, São Paulo, Brazil
(2)Collegiate Nursing, State University of West Paraná, Cascavel, Brazil
(3)Escuela de Enfermería, Industrial University of Santander, Bucaramanga, Colombia
(4)School of Nursing, University of Sao Paulo, Sao Paulo, Brazil

Learning Objective 1: • The learner will be able to describe the Power as Knowing Participation in Change Tool as a measure of power.

Learning Objective 2: • The learner will be able to contrast Rasch model with classical test theory.

Purpose: The 48 item Power as Knowing Participation in Change Tool (PKPCT) has been used to assess power. Modern psychometric theory has established new quality standards for measurement tools going beyond reliability and validity emphasized by classical test theory. The aim of this study was to examine the psychometric properties of the PKPCT – Brazilian Version by means of the Rasch analysis.

Methods:  A questionnaire, including the PKPCT adapted to Portuguese Brazilian language by means of back translation, was answered by 458 nurses or baccalaureate students. Responses were assessed in terms of Cronbach’s alpha, factor analysis, and Rasch analysis.

Results: The four domains of the PKPCT (12 items each) as also as the overall tool showed good internal consistency (Cronbach’s alpha ranged from 0.87 to 0.96). Unrotated factor analysis (principal components analysis) of the 48 items showed nine eigenvalues greater than 1 and explained 66.6% of the variance. A reduction from seven to four rating categories was needed to comply with criteria for correct category function in Rasch analysis. The factor analysis in Rasch showed 4.9 eigenvalues greater than 1.0. Factor analysis of each domain confirmed sufficient unidimensionality, so each domain was analyzed separately. Person reliability ranged between 0.86 and 0.90. Item reliability ranged between 0.98 and 0.99. The mean infit and outfit of the four domains ranged between 1.00 and 1,04 meaning the response patterns fit the expected pattern measurement. The precision of measurement depends on how well the items of an instrument separate individuals of different ability levels. The person-separation index of the four domains ranged from 2,43 to 3,03. 

Conclusion: Rasch analysis enabled us to confirm the validity of PKPCT – Brazilian Version as a four-scale tool to measure power. Summing scores of all 48 items is not recommended because unidimensionality was not supported by Rasch Analysis.