Classical test theory versus RASCH analysis in reducing CPQ's items

Abstract

BACKGROUND: To reduce burden of response shortening (or reducing) OHQoL questionnaires is routinely conducted. To date, reducing OHQoL questionnaires has been based on classic test theory (CTT). It is now recognised that the CTT approach to item reduction has major conceptual limitations – the resultant short form produced lacks an explicit ordered continuum of items that represents a unidimensional construct. RASCH analysis provides an alternate method to reduce items in questionnaires and overcomes this major conceptual limitation of CTT. OBJECTIVES: This study aimed to develop a short form of the Child Perception Questionnaire (CPQ) using RASCH Analyses and compare its psychometric properties with short forms of the CPQ developed through CTT – CPQ-16 and CPQ-8. METHODS: Data on child OHQoL was collected using the original 37-item CPQ in an oral health survey among 506 12-year-olds in Hong Kong. RASCH analysis using Winsteps software was used to create a short form of CPQ. The three short forms were compared for CTT and RASCH requirements. RESULTS: RASCH analysis resulted in a 23-item short form of CPQ (RASCH-23). All short forms showed similar characteristics under CTT requirements: item-total correlation ranged from 0.29-0.87 for ISF-16, 0.25-0.63 for ISF-8 and 0.22-0.77 for RASCH-23, while reliability ranged 0.52-0.78, 0.39-0.63 and 0.53-0.79 respectively. However, the CPQ-16 and CPQ-8 failed to meet the goodness-of-fit criteria of RASH modelling. CONCLUSIONS: A short form of CPQ developed through RASCH analyses (RASCH-23) satisfied the criteria of classic test theory similar to CPQ-16 and better than CPQ-8. However, the CPQ-16 and CPQ-8 failed to meet the requirements of RASCH modelling.