Item response models for child perceptions questionnaire

Abstract

Background Oral health-related quality of life (OHRQoL) is an abstract concept which cannot be directly observed. Measurement instruments have been developed and constructed to assess it. Conventional methods in evaluating psychometric properties are common yet involve unrealistic assumptions. Item response theory (IRT) relaxes these assumptions and enables assessment of additional measurement properties. This study aimed to assess additional measurement properties of two 8-item short forms of Child Perceptions Questionnaire (CPQ11-14 RSF:8 and ISF:8) using IRT. The study also explored factors associated with OHRQoL in the 12 years old children and the change over 3 years by IRT explanatory analysis.

Methods In the baseline survey, 668 children aged at 12 were recruited and 436 (65.3%) were followed up successfully after 3 years. Data on the CPQ11-14 RSF:8 and ISF:8, demographics, oral health behavior and status were collected. Before assessing the measurement properties, the assumptions of unidimensionality and local dependency were tested. The data were then fitted using the graded response model (GRM) to obtain the estimates of the discriminatory parameters, threshold parameters and information functions. Thereafter, the equality of discriminatory parameters across items and the equality of test information functions between the 2 short forms were tested. Differential item functioning (DIF) across gender and the longitudinal invariance were also assessed. Explanatory GRM was fitted to the CPQ11-14 data. The Bayesian estimation method – MCMC with Gibbs sampling was adopted for parameter estimation and the credible intervals were obtained.

Results After confirmation of unidimensionality, both short forms were found to distinguish people with poor OHRQoL well, but not for people with good OHRQoL. Items related to oral symptoms were not informative to OHRQoL. Discriminatory parameters were found unequal across items and RSF:8 did not provide a significantly more precise OHRQoL estimate than ISF:8. Three items had DIF across gender but the bias induced to the scale was minimal. Four items lacked longitudinal invariance: three items from the domains of emotional and social well-being and the other one from the domain of functional limitation. Since items concerning oral symptoms were not informative, these items were removed, resulting in RSF:6 and ISF:6. The 6-item versions and 8-item versions were almost identical in terms of measurement precision of OHRQoL. For the 12 years old children, healthier oral health behavior, better gum status, families with both parents employed, higher father’s education level and lower mother’s education level were found to be associated with better OHRQoL. With statistical adjustment of longitudinal invariance, OHRQoL level in general has improved over 3 years but no predictor was associated with OHRQoL in follow-up. For those with decreased family income, their OHRQoL had worsened over 3 years.

Conclusions This study illustrates the use of IRT validation on the 8-item short forms of CPQ11-14 and a better understanding of the psychometric properties of CQP11-14 has been gained. In addition, the use of the IRT explanatory analysis enables the identification of the factors associated with OHRQoL and its changes over time provides important information to oral healthcare researchers and policymakers.