Rasch analysis of Impact of Vision Impairment for Children questionnaire.

PURPOSE

To validate the Indian-translated Impact of Vision Impairment for Children (IVI_C), a vision-related quality of life (VRQoL) instrument, and to investigate the effect of sociodemographic and ocular characteristics on VRQoL.

METHODS

The 24-item IVI_C was administered face to face to 221 children with vision impairment (VI) (mean age, 12.6 years; male, 68%) referred to the Vision Rehabilitation Centres. Rasch analysis was used to investigate the IVI_C for key indices such as measurement precision as measured by person separation (minimum recommended value, 2.0), unidimensionality (i.e., whether all the items contribute toward measurement of a single underlying construct as assessed by item fit and principal component analysis of residuals), targeting of items to participants' VRQoL (i.e., matching of item difficulty to participant ability; ideal targeting, <1.0 logits). Univariate analysis was performed to determine if the person scores of IVI_C were significantly different across participant subgroups stratified by sociodemographic and ocular characteristics, including the severity of VI.

RESULTS

The IVI_C showed misfit to the model and lacked unidimensionality. Principal component analysis confirmed the presence of an additional construct, mobility, but it did not possess adequate person separation when assessed individually. Deleting the mobility-related items restored unidimensionality, but additional items misfit, necessitating item reduction. Finally, a 17-item IVI_C possessed good measurement precision (person separation, 2.04), was unidimensional (albeit not purely), and consisted of items that fit the Rasch model and were well targeted to the participants' VRQoL (targeting 0.76 logits). There was no statistically significant difference in the VRQoL across subgroups stratified by age, sex, type of school, cause of VI, duration of vision loss, and severity of VI (p > 0.05 for all).

CONCLUSIONS

The revised 17-item IVI_C is shorter and has better psychometric properties than the original version in school-aged children with VI in India, and the responses are unrelated to sociodemographic and clinical variables. Although it has the potential for use in cross-sectional and outcomes research in children with VI, caution should be exercised while interpreting the 17-item IVI_C scores given the presence of slight multidimensionality.