A myriad of patient-reported outcomes (PROs, commonly referred to as questionnaires or instruments) have been developed in the field of ophthalmology over the past 20 to 30 years.
1 The National Eye Institute Visual Functioning Questionnaire
2 (NEI-VFQ) is among the most popular PROs. Like the majority of the PROs, the NEI-VFQ has been developed predominantly for a Western population and so the content may be perceived as irrelevant by those in developing countries, such as India. By contrast, the Indian Vision Function Questionnaire (IND-VFQ) was specifically developed for relevance to an Indian population. Using CTT, both the NEI-VFQ and the IND-VFQ have been demonstrated to possess adequate psychometric properties.
2–4 However, Rasch analysis conducted recently on the NEI-VFQ
39 5 and IND-VFQ
6,7 (albeit in different populations) reported both instruments to be flawed. Nonetheless, re-engineering resulted in the formation of a valid unidimensional measure of visual functioning (VF) for both instruments. Consequently, the corresponding scales have been labeled as the Long form Visual Functioning scale
39 (LFVFS
39) and Visual Functioning Scale of the IND-VFQ (VFS of the IND-VFQ). The resemblance in the names (VFS) of the two instruments—the LFVFS
39 and VFS of the IND-VFQ—encourages the assumption that there is also similarity in the content and in the underlying measurement construct (visual functioning, VF). However, this assumption hasn't been tested as yet, so the question still remains as to what extent these instruments capture the same construct (if they do).
It is important that similar-appearing instruments (at least name wise) actually do measure the same construct because only under such conditions can instruments
communicate with each other. Such a concurrence in the measurement construct will facilitate the comparison of results across studies and valid conclusions can be drawn from existing data despite use of different instruments to measure the same construct.
8 For example, studies reporting people who have taken the LFVFS
39 can be compared with those that have taken the VFS of the IND-VFQ.
In addition to providing detailed insight into performance of instruments, Rasch analysis offers the benefits of possibly cocalibrating instruments after they have been demonstrated to measure the same construct. To the best of our knowledge, except for the few studies by Massof, there is paucity of literature regarding the cocalibration of ophthalmic instruments. Massof compared four VFQs (Activities of Daily Vision Scale, Visual Function Index-14, NEI-VFQ, and Visual Activities Questionnaire) administered to a low vision sample and demonstrated that all these instruments measure a single construct; however, the instruments only differed in measurement precision and accuracy.
8,9 Massof reported that all the four VFQs can be calibrated to a common measurement scale and that the measurement scale of each of the four instruments was a linear transformation of the measurement scale estimated from the merged responses of all four instruments.
8,9
Against the above background and encouraged by the work of Massof, we were interested in investigating if the LFVFS39 and VFS of the IND-VFQ measure the same construct, VF. Given that both instruments have been propagated to measure VF, users of these instruments would expect to obtain similar estimates of patient visual functional ability from both. A comparison of these instruments in the same patient population should help prove or disprove the validity of this expectation. An opportunity thus arose for such a comparison in the present study wherein both these instruments were administered concurrently to common persons. Therefore, our primary aim was to cocalibrate the items from both the LFVFS39 and VFS of the IND-VFQ onto a single metric to determine if they did measure the same construct in a cohort of Indian visually impaired (VI) patients. If it is valid to cocalibrate the two instruments, then item measures estimated from individual Rasch assessments of the two instruments should be a linear transformation of item measures estimated from the analysis of the pooled data. So to check the validity of pooling the data, we tested the LFVFS39 and VFS of the IND-VFQ separately against Rasch model assumptions, specifically unidimensionality, in the same cohort. Given that cocalibration enables mapping the VF from both the instruments onto a single metric, our secondary aim was to compare the measurement range of VF of LFVFS39 with that of VFS of the IND-VFQ (i.e., rates of ceiling and floor effects [percentage of cases with maximum and minimum scores]). We envisage that such comparisons would assist an investigator who is interested in evaluating the VF in VI patients, in deciding which of the two instruments could be used for his or her study.