Do item weights matter? An assessment using the oral health impact profile

To determine whether or not item weights contribute to the performance of the Oral Health Impact Profile (OHIP), a comprehensive measure of the functional, social and psychological outcomes of oral disorders.

Data were obtained as part of an oral health survey of older adults living in Ontario, Canada. Subjects completed a personal interview, clinical examination and a self-complete version of the OHIP. OHIP scores were calculated in three ways: a simple count method, an additive method and a method incorporating item weights derived from the Thurstone paired comparison technique. These scores were calculated for the full 49-item version of the measure and for a short form consisting of 14 selected items. The discriminant, concurrent and predictive validity of these scores for the two versions of the measure were ascertained.

Complete data were obtained for 522 subjects. Just over half were female (56 per cent) and their mean age was 66 years.

The OHIP discriminated between groups based on dental status (dentate/edentulous), presence of dry mouth (yes/no) and, for the dentate, according to the number of remaining teeth (less than 20/20 or more) irrespective of scoring method or the version of the questionnaire used. All scores showed significant associations with self-rated oral health, self-perceived need for dental care and dissatisfaction with oral health status. There was evidence to suggest that weighted scores were better at discriminating between groups than the simple count method but no better than the additive method. Similar findings emerged with respect to the ability of the scores to predict prosthodontic, surgical and restorative treatment needs.

Although the data suggested that item weights did improve the performance of the OHIP, the fact that simple scoring methods were as good as more sophisticated ones might mean that the OHIP could be used in contexts, such as patient assessment for clinical care, where the calculation of weighted scores was not feasible.