The Quality of Life Impact Refractive Correction (QIRC) questionnaire: validation of the Malay-translated version of the QIRC using Rasch analysis

The QIRC was designed to assess the impact of refractive correction on QoL either through spectacles, contact lenses or refractive surgery. It consists of five domains; visual function (Item 1), visual symptom (Item 2), visual convenience (Items 3 to 7), visual concern (Items 8 to 13), and emotional well-being (Items 14 to 20) [7]. Five response categories were used to score each item of visual function and visual convenience (Category 5 = extremely to Category 1 = not at all) and the other five response categories for each item of visual symptom, visual concern, and emotional (Category 5 = always to Category 1 = never). One additional response category (Category 0 = don’t know/not applicable) was included for all items and was considered as missing data [18].

The Malay QIRC went through the standard protocol for forward-backward translation. First, a professional translator (AAA) and an optometrist (NSS), who were Malay-English bilinguals, independently translated the original QIRC into Malay. A panel of experts analysed the content equivalence of the translation with the original version [19] and the suitability of the translated phrases to the culture of the target population. The experts then resolved the discrepancies between the two Malay translations [20]. The panel consisted of two experts in the QIRC measures (MMMMS, HAM), an expert in the questionnaire validation process (NAH) and an expert teacher of the Malay language (AHJ), and all were proficient in the Malay-English languages.

The other two blinded bilingual Malay-English translators, an optometrist (AFF) and a family medicine specialist (MSES), translated the Malay version back into English. Three of the researchers (MMMMS, HAM, NAH) assessed the consistency of the backward translations with the original QIRC and achieved a consensus translation of the Pilot Malay QIRC.

The pre-test was conducted by one researcher (NSM) on 105 participants who were bilingual Malay-English speakers. The participants self-administered the 20-item Pilot Malay QIRC and commented on any ambiguous items. The time taken to administer the questionnaire was recorded (mean, 9.76 ± 0.66 min). Subsequently, the pre-test output was reviewed by all researchers (MMMMS, HAM, NAH, NSM) to enhance the item comprehension and cross-cultural adaptation.

This cross-sectional study was conducted at an institutional optometry clinic in Kuantan, Pahang, Malaysia, and at a general private optometry clinic in Gombak, Kuala Lumpur, Malaysia, from 4 February 2019 to 28 February 2020. Participants were randomly selected from patients attending their optometric appointment using Research Randomizer version 4.0 [21, 22]. The Malay QIRC was self-administered by non-presbyopic participants who were able to understand the Malay language. Only participants with refractive error of − 3.00 D and above in spherical equivalent refraction (SER) and astigmatism correction of less than 2.00 D were included. All participants wore their current spectacle or contact lens correction prescribed by optometrists for less than 1 year. Patients with ocular or systemic disease, ophthalmic surgical history, or trauma noted in the previous medical record or optometric examination were excluded from this study.

This study design was granted approval by the International Islamic University Malaysia Research Ethics Committee (IIUM/504/14/11/2/IREC2019-KAHS[U]). Written consent was gained from all eligible participants. The study protocols were operated in compliance with the tenets of the Declaration of Helsinki, the International Conference of Harmonisation Good Clinical Practice Guideline (ICH-GCP) and the Council for International Organisations of Medical Sciences (CIOMS) International Ethical Guidelines.

Sample size determination was calculated for the 20-item QIRC with five response categories at a power of 0.80 and a probability level of 0.05 using Free Statistics Calculators version 4.0: A Priori Sample Size for Structural Equation Models [23]. The minimum sample size required for the model structure was 100. The suggested sample size was adequate for Rasch analysis based on a 95% confidence of item calibrations or person measures stable within ±0.5 logits [24]. Furthermore, it corroborated the guidelines for respondent-item ratios, ranging from 4:1 to 15:1 [25, 26].

Psychometric validation of the items and response categories was performed using Rasch analysis and test-retest reliability analysis. Rasch analysis was employed using the Andrich rating scale model (Winsteps software version 4.5.1). While test-retest reliability analysis was conducted using Statistical Package for the Social Sciences (SPSS) software for Windows version 25.0.

First, Rasch principal components analysis of residuals (PCAR) was employed to identify the unidimensionality of the scale. Unidimensionality is crucial as it demonstrates that the instrument measures the underlying trait for which it was designed [27]. In this context, PCAR is used to determine if the residuals exhibit patterns after accounting for the observed variance explained by the Rasch measure. PCAR examines contrasts in the correlation matrix of the residuals [28]. The first contrast refers to the component that accounts for the greatest amount of variance in the residuals. The Rasch (first) dimension should achieve a variance explained by measures of 50% [28]. Unexplained variance in the first contrast of 3.0 Eigenvalue units or greater indicates the presence of an additional (secondary) dimension that is not considered random noise [28]. The PCAR result revealed multidimensionality in the Malay QIRC version. The variance explained by measures was less than 50%, with unexplained variance in the first contrast being more than 3.0 Eigenvalue units. Hence, the Malay QIRC was divided into the functional (Items 1 to 13) and emotional (Items 14 to 20) scales, and subsequent analyses were conducted separately.

Next, the category probability curve was analysed to determine the order of response categories. The category probability curves show that the response categories for the functional scale were unordered (Fig. 1), while the response categories for the emotional scale were ordered. Thus, the unordered category for the functional scale was collapsed into an adjacent category. The new four response categories for the functional scale were applied in the Final Malay QIRC.

Then, the items fit were analysed to determine any misfits. The following item removal criteria were set: Items infit and outfit were outside 0.50 to 1.50 mean square (MnSq) or were equal to/greater than 2.0 z-standardised (z-std) [28], missing data were greater than 50%, floor and ceiling effects were higher than 50%, and skewness and kurtosis were out of − 2.00 to + 2.00 [7]. Analysis of the item fit statistics identified one misfit item, indicating removal of the item. As a result, a version called Final Malay QIRC was prepared with 19 items.

A Rasch analysis was performed to evaluate the psychometric properties of the Final Malay QIRC, consisting of 12-item functional and 7-item emotional scales. The validity and reliability of the psychometric properties were analysed using category threshold, separation indices, fit statistics, targeting precision, differential item functioning (DIF), and test-retest reliability. A new sample of 304 participants was recruited for the psychometric validation of the Final Malay QIRC. The time spent on the questionnaire was documented (mean, 8.22 ± 0.41 min).

The category threshold was examined to assess the functionality of the response categories. The modified four response categories (after collapsing) for the functional scale were evaluated as to whether they were ordered or needed further modification. The original five response categories for the emotional scale were also re-evaluated.

The person reliability and separation index were used to evaluate the capability of the 19-item Malay QIRC to distinguish the participants’ strata. The coefficient of person reliability ranges from 0 to 1 designates the item measures precision. A higher coefficient represents a better performance of the item measures in distinguishing participant ability between different strata. In this study, a minimum of 0.80 and 2.00 were set for the person reliability and separation index, respectively [28].

The overall fit of the Final Malay QIRC to the Rasch model was analysed using the fit statistics. In this study, the researchers set the criteria for the overall fit to be 0.50 to 1.50 MnSq or less than 2.0 z-std of the average infit and outfit. Therefore, the fit statistics determined whether the Final Malay QIRC has a good fit to the Rasch model [28].

Targeting precision was determined by analysing the distribution of item difficulty to participant ability in the person-item map. The difference between the means of the person and item measures (mean difference) should be lower than 0.50 logits to signify that the QIRC items were properly targeted to the refractive correction group [29]. The person-item map also illustrates the item difficulty order for refractive correction wearers, from easiest to most difficult.

DIF was evaluated between two subgroups of participants classified by gender (male versus female) and SER (moderate myopia: ≤ − 3.00 D to > − 6.00 D versus high myopia: ≤ − 6.00 D). A noticeable DIF was considered present if the DIF contrast was greater than 0.50 logits and the probability was meaningful (p < 0.05) [28].

Test-retest reliability was conducted on 59 participants (out of 304 participants) during their optometric follow-up appointments at an interval of 2 to 4 weeks (median, 2.7 weeks; mean, 2.7 ± 0.6 weeks). The interval was sufficient to mask the participants’ memory on the construct being measured and its stability over time [30, 31]. In addition, the time spent completing the questionnaires at the first (mean, 8.20 ± 0.39 min) and second visits (mean, 8.08 ± 0.53 min) was tracked. Test-retest reliability analysis included the intraclass coefficient (ICC), Cronbach’s alpha (α) and coefficient of repeatability (CoR).

The ICC of greater than 0.90 reflects excellent reliability for clinical purposes. The ICC test of the two-way mixed model and consistency type was performed as described elsewhere [32, 33]. Cronbach’s α is a reliability coefficient used to measure a set of items that are consistently related as one group. It was set at 0.90 to consider that the questionnaire items had excellent internal consistency. The CoR is a standard deviation of the test-retest difference multiplied by two. Thus, a smaller CoR indicates higher repeatability.

One hundred and twenty-three eligible participants were randomly selected to take part in the pre-test. The response rate was 85.4%, with 18 participants unwilling to participate. Thus, a total of 105 participants self-administered the Pilot Malay QIRC. Majority of the participants were Malays who came from various work statuses. Most of them were spectacle wearers, with a current correction at around 6 months of age, and had more than 10 years of experience wearing corrections (Table 1).

Based on the pre-test feedback, two items had to be modified to improve comprehension and adaptation to the local culture. Item 6 ‘swimming in the sea’ was reworded to ‘swimming at the beach’, and Item 7 ‘when using a gym/doing keep-fit classes/circuit training’ was paraphrased to ‘during exercise or sports activities’. Items 4, 8, 9, and 12 were amended to include ‘LASIK’ as an example of ‘refractive surgery’ to enhance the ability of the items to be self-administered.

The unidimensionality of the QIRC was investigated prior to further analysis of the validity and reliability of its psychometric properties. PCAR showed that the variance explained by the measures and the unexplained variance in the first contrast were 48.7% and 5.1 Eigenvalue units (13.1%), respectively. Further analysis of the standardised residual loadings revealed that all emotional items (Items 14 to 20) had residual loadings of greater than 0.60. These findings indicated that the Malay QIRC was a multidimensional scale. It should therefore be considered separately. It also highlighted the critical nature of Rasch analysis in translation versions, even when the original instrument was developed and validated using Rasch analysis.

As illustrated in the category probability curve, the five response categories for the 13-item functional scale were unordered (Fig. 1). The sunken category (Category 3 = a moderate amount/fairly often) was collapsed into the adjacent category (Category 2 = a little bit/occasionally). The modified four response categories were later used for the functional scale of the Final Malay QIRC.

The item fit statistics showed that infit and outfit MnSq (z-std) for Item 3 were 1.69 (4.2) and 1.78 (4.5), respectively. It reflected that Item 3 did not fit the Rasch model when MnSq was above 1.50 and z-std was greater than 2.0. In contrast, the other items fulfilled the criteria for item fit. Therefore, Item 3 was removed, and the following items were re-numbered one step up in sequence. For instance, Item 4 was re-numbered as Item 3, and the next items followed. The 19-item Final Malay QIRC was set for the further validation process.

Out of 366 eligible participants who were randomly selected, a total of 304 participants (response rate, 83.1%) agreed to be recruited for the psychometric validation of the Final Malay QIRC. Participants were age-matched to the pre-test group (mean difference, 0.26 years; t = 0.42, p = 0.67). Most participants were Malays from various work statuses with more than 10 years of experience wearing corrections (Table 1). The time spent administering the questionnaire was significantly reduced for the Final Malay QIRC compared to the pilot version (mean difference, 1.53 min; t = 22.2, p < 0.001).

The category probability curve illustrates that the modified four response categories for the functional scale were ordered (Fig. 2). The original five response categories for the emotional scale were also ordered.

After modifying the items and the response categories, the higher person reliability and separation index of the 12-item functional and the 7-item emotional scales were found. These person separation indices demonstrated that the 19-item Final Malay QIRC could secede the participants into distinct strata following their ability. The item reliability and separation index also showed that participants were able to differentiate items according to their difficulty hierarchy (Table 2).

The fit statistics exposed that the average infit and outfit of the person-item on both the functional and emotional scales were approximately 1.0 MnSq (Table 2). Furthermore, the infit and outfit for all items were within 0.66 to 1.45 MnSq (Table 3). Overall, the 19-item Final Malay QIRC showed a satisfactory fit to the Rasch model.

Targeting precision assessment was illustrated by the item difficulty and participant ability distribution and their mean difference in the person-item map (Figs. 3 and 4). The mean difference (targeting) on the functional and emotional scales was 0.23 and 0.08 logits, respectively. The differences were considerably small. On the functional scale, Items 2 and 1 (in sequential order) were the two least difficult items, whereas Items 9 and 10 were the most difficult for participants to answer (Fig. 3). On the emotional scale, Item 15 was the least difficult item, while Item 17 was the most difficult item (Fig. 4).

Analysis of the DIF revealed that all items of the functional and emotional scales had a DIF contrast of less than 0.50 logits. However, Item 18 ‘able to do the things you want to do’ showed a significant probability between genders (DIF contrast, 0.40 logits; t = 2.12, p = 0.04). It indicated that Item 18 was 0.40 logits tougher for females than for males.

There was no significant difference in QIRC scores between the test and retest (mean difference, 1.09 ± 4.07 units; t = 2.05, p = 0.05). Test-retest reliability analysis showed a high ICC (single measures, 0.94) and Cronbach’s α (0.97) with a CoR of ±8.14 units. In addition, the time required to complete the questionnaire between the first and second visits was insignificant (mean difference, 0.12 ± 0.70 min; t = 1.33, p = 0.19). These findings confirmed that the 19-item Final Malay QIRC had excellent repeatability and internal consistency.