Background
In the United States [
1] India [
2] and elsewhere there is a growing emphasis on understanding the impact of medical interventions on the lives of patients to assess and improve quality of care. In the case of cataract surgery, success is often equated with an improvement in an individual’s best-corrected visual acuity (BCVA). However, it is also meaningful to judge the success of surgery in relation to vision-related quality of life (QOL) and vision-dependent functioning, as these factors likely drive patients to pursue surgery and affect their satisfaction.
Many patient-reported outcome (PRO) measures have been used to assess the impact of cataract on vision-related QOL and functioning [
3‐
9] and a number of these have been employed among Indian patients undergoing cataract surgery [
8‐
11]. However, existing instruments may not be culturally relevant to patients in India [
3,
4]; have not been adequately validated [
9]; or are too lengthy to be deployed as point-of-care tests in a high-volume clinical setting [
7].
Among existing PRO measures, the Indian Visual Functioning Questionnaire (IND-VFQ) was developed in India using rigorous qualitative methodology [
12]; was validated with Rasch analysis [
13]; and has been shown to detect a change in vision-related QOL after cataract surgery [
10]. Nonetheless, with 28 items, the Rasch-modified version of the IND-VFQ was too long for use at the point-of-care in a busy clinical practice at Aravind Eye Care System (AECS) in Madurai, India. Therefore, we sought to develop a rapidly administered short-form version of the IND-VFQ (SF-IND-VFQ) and to test its psychometric properties for patients undergoing cataract surgery at a tertiary eye hospital in south India. Our purpose in validating a rapid point-of-care PRO for cataract surgery is to facilitate patient-centered measurement for quality improvement, monitoring of surgical results, and use in future research studies.
Discussion
We found that the 11-item SF-IND-VFQ is a unidimensional instrument capable of detecting changes in visual functioning among Tamil-speaking patients undergoing cataract surgery in India. The culturally-relevant questionnaire was administered rapidly at the point-of-care and had good reliability, content validity and was largely unaffected by DIF.
The purpose of this study was to design a short-form PRO measure that is appropriate as a point-of-care instrument for use in high volume clinical settings in order to monitor outcomes, promote quality improvement, and facilitate outcome measurement in future pragmatic clinical trials and clinical research studies. Traditionally, PRO measures have been employed as research tools and are often too time-consuming to be deployed in a busy clinical setting. However, in more recent years, with the growth of the National Institutes of Health’s Patient-Reported Outcomes Measurement Information System (PROMIS) and the National Health System’s quality improvement initiatives in the United Kingdom, the use of PRO measures in clinical practice has become more common [
24].
The SF-IND-VFQ meets most criteria that Kroenke and colleagues proposed for using PRO measures in clinical practice [
25]. Briefly, the survey is setting-appropriate based on its cultural relevance and low respondent burden; can be self-administered and quickly completed; has items that focus on a single symptom; evaluates only one dimension (symptom severity), with no more than four response categories per item; and is widely accessible at no cost. Additional criteria such as actionability (translating scores into concrete actions) and universality (usefulness in other conditions) will be tested in future studies as the SF-IND-VFQ is used more broadly throughout AECS.
One item present in the initial SF-IND-VFQ, “Does bright light hurt your eyes?,” was removed since it did not fit the same unidimensional construct as other items and had poor fit with the Rasch model, as denoted by its MNSQ values. Notably, this was the only item that assessed pain symptoms and this may explain why it did not form part of a unidimensional scale. All other items had goodness-of-fit statistics within the acceptable range, which confirmed the instrument’s construct validity.
Two items retained in the SF-IND-VFQ were affected by DIF, though the magnitude of DIF was negligible for both of these. The first of these items (item 2) asked, “because of your vision how much trouble do you have making out the bumps and holes in the road when walking?” Patients undergoing MSICS, but not phacoemulsification, were less likely to endorse response categories associated with high visual functioning. This may be the case since MSICS is frequently the preferred surgical technique for very advanced cataracts that tend to cause more severe visual dysfunction. Item 8, “because of your eye problem are you ashamed that you can't see?” had non-uniform DIF in which patients who received free surgery showed less discrimination than those who paid for surgery. The reason for this finding is less clear but may be related to understanding or interpretation of the question.
The SF-IND-VFQ was able to discriminate effectively amongst individuals with high levels of visual dysfunction. However, overall the instrument demonstrated suboptimal targeting. The mean difference between person ability and item difficulty was 3.41 logits, suggesting that the amount of visual functioning required by SF-IND-VFQ items was not matched to the visual ability of our subjects. Other cataract surgery PRO measures have also suffered from poor targeting because of items being too easy for patients with cataract [
26,
27]. One exception is Catquest-9SF, which showed excellent targeting in a Swedish population (preoperatively, − 0.34 logits) [
4]. In our current study, since patients had poorer vision before surgery, SF-IND-VFQ targeting was considerably better preoperatively, though it remained suboptimal (− 1.92 logits).
One reason for suboptimal targeting in this study was that the long-form IND-VFQ was also not well-targeted to patients with cataract [
13]. The IND-VFQ was developed through focus groups of patients with cataract, glaucoma, diabetic retinopathy, macular degeneration, and “mixed low vision.” Among the 308 participants from three regions of India, about half (51.9%) had cataracts. In a subsequent study, Finger and colleagues surveyed patients with cataract in south India with the initial 33-item IND-VFQ [
13]. Using principal components analysis, they found that the survey contained five subscales (general functioning, mobility, activity limitation, psychosocial impact, and visual symptoms). When Rasch analysis was performed, the visual symptoms subscale had the best targeting (− 0.93 logits), while general functioning (− 2.32), mobility (− 2.94), activity limitation (− 1.93) and psychosocial impact (− 1.39) were not as well targeted. We had similar results when we assessed preoperative targeting (− 1.92 logits), which is comparable to the cross-sectional analysis in Finger et al.’s study of patients with cataract [
13]. Since subjects in our study had relatively preserved preoperative visual function (mean BVCA of about 6/18) the instrument may be better targeted to cataract patient populations in India with more advanced disease and this could be explored in future studies.
Given the strong psychometric properties of Catquest-9SF, we had considered employing this instrument for our study rather than validating a new short-form questionnaire. However, although Catquest was well-targeted to patients in Sweden [
4] and several other settings, targeting had not been optimal in all populations [
27]. Additionally, we did not believe that all Catquest items were culturally relevant to our patients at AECS. Specifically, items related to reading newspaper text, doing handicrafts, reading text on television, and carrying out a preferred hobby were not thought to be pertinent.
The SF-IND-VFQ was extremely sensitive to changes in visual function after cataract surgery, with the average score improving by 2.98 logits. Additionally, the effect size in this study, measured by Cohen’s
d, was 1.80, which is considered
very large based on accepted criteria [
14]. This was due in part to including only IND-VFQ items that were known to be sensitive to the impact of cataract surgery. The effect size for the SF-IND-VFQ was also high in comparison to other cataract PRO measures. McAlinden and colleagues conducted a study in which 16 PROs were administered to patients having cataract surgery in Australia [
28]. Out of this battery, they found that Catquest-9SF had the largest effect size with a Cohen’s
d of 1.45.
Limitations
Although our findings indicate that the SF-IND-VFQ is a valid, reliable and quickly administered PRO, there were several limitations to our study. First, our findings may not be generalizable beyond a tertiary eye center in south India; however, future work should be done to test the measurement properties of the SF-IND-VFQ in different geographic and eye care delivery settings, such as in a lower resource field-based setting in India. Second, we intentionally selected items from the parent IND-VFQ that were most sensitive to the effect of cataract surgery, though we acknowledge that targeting might have been better had we chosen items using different criteria.
All analyses used combined pre-operative and post-operative data. We acknowledge that this violates the Rasch model’s assumption of independence. However, in pooling data we are able to provide item calibrations that can be applied to patients both before and after undergoing cataract surgery, a feature that makes this approach clinically useful. To determine the impact of this violation of the assumption of independence, we analyzed pretest data alone and compared this to results from our pooled analyses. We found no differences in the results for the EFA or CFA, while Rasch person reliability was 0.78 for pretest data alone compared to 0.80 for the pooled data and the 11 retained items continued to show good fit with the model. Of note, we were unable to fit a separate model for the post-operative data since there were few responses indicating severe visual dysfunction, as one would expect due to patients’ improved vision after undergoing surgery.