Bayesian explanatory item response analysis for CPQ11-14 RSF:8

Abstract

OBJECTIVES: In contrast to summing up the score of each item in CPQ11-14, latent score generated by item response theory (IRT) does not assume all items to be contributed equally and allows heterogeneity of variance in the explanatory analysis. Under the IRT framework, this study investigated the associated factors of oral health-related quality of life (OHRQoL) of junior secondary students in Hong Kong using the CPQ11-14 RSF:8 questionnaire. MetHods: The questionnaire was administered to 649 12-year-old school children. An explanatory graded response model using the Bayesian software WinBUGS was fitted. Contribution of each item was assessed by the discriminatory item parameter. Heterogeneity of the estimated latent score was examined by plotting the standard error against the OHRQoL scale. Demographic information (gender, place-of-birth, residing in Hong Kong for >7 years), family background (parents’ education attainment, employment status, family income) and oral health behaviors (snacking frequency, tooth-brushing frequency, use of fluoride toothpaste, participation in student dental care service) were considered as the explanatory variables. Factors were considered to be significant if the 95% credible intervals (CI) of the regression coefficients (β) exclude zero. RESULTS: 5 out of 8 items had discriminatory parameters significantly differed from 1. Standard errors of the estimated latent scores increased with OHRQoL. Brushing twice or more per day (median β=-0.23, [CI:-0.43,-0.04]), lower daily snacking frequency (>twice vs. none: 0.47[0.11,0.85]; >twice vs. once: 0.53[0.20,0.87]; twice vs. once: 0.31[0.07,0.57]), higher father’s education level (junior secondary vs. < primary: -0.65[-0.99,-0.33]; senior secondary vs. < primary: -0.56[-0.89,-0.21]) and higher mother's education level (senior secondary vs. junior secondary: -0.25[-0.49,-0.003]) were shown to be significantly associated with better OHRQoL. CONCLUSIONS: After adjusting for the heterogeneity of variance and allowing items to contribute differently, it was found that oral health behavior and parents’ education played a role in junior secondary students’ OHRQoL in Hong Kong.