Rasch analysis of visual function and quality of life questionnaires.

Purpose: To investigate the psychometric properties of the Visual Function (VF) and Quality of Life (QOL) questionnaires, using the Rasch measurement model.

Methods: Four hundred eighty-nine patients with cataract awaiting surgery, drawn from the surgical waiting list of the Flinders Medical Centre, Adelaide, South Australia, were mailed the VF and QOL questionnaires for self-administration. Rasch analysis was undertaken to assess the following properties: measurement of a single construct (i.e., unidimensionality by fit statistics and principal components analysis), discrimination between strata of participant ability (person separation, recommended minimum value 2.0), targeting of item difficulty to participant ability, and differential item functioning (DIF, whether items function similarly for different subgroups and notable DIF defined as >1.0 logit).

Results: The VF questionnaire fit the Rasch model showing good metric properties, including unidimensionality, and was free of notable DIF. The QOL questionnaire showed inadequate person separation, two misfitting items and significant multidimensionality (Eigenvalue of 2.3 for the first contrast), suggesting the presence of a separate subscale, self-care, but this was not valid because of inadequate person separation. Deleting these contrasting items from the QOL questionnaire removed multidimensionality and item misfit but did not improve person separation, indicating that it could not effectively discriminate between the participant's QOL. Both VF and QOL questionnaires demonstrated poor targeting (-2.22 and -3.26 logits, respectively), indicating that items were too easy and participants had higher QOL than that the items allowed for. Three items in the QOL questionnaire demonstrated notable DIF by age and systemic comorbidity. None of the subscales functioned optimally.

Conclusions: The QOL questionnaire is not a valid measure of QOL. However, the VF questionnaire is a reliable and valid measure of visual disability in patients with cataract, and although targeting was suboptimal in a developed country, it may be optimal in a developing country as was originally intended.