A total of 386 patients were included. Mean age was 29.41 ± 9.98 years, and most patients (n = 234, 60.62%) were female. Upon evaluating their previous surgery history, 53 patients (13.73%) had had a keratoplasty in at least one of their eyes, while 189 patients (48.96%) had a history of corneal rings in at least one eye. Regarding history of corneal crosslinking and phakic intraocular lens implantation, the proportions were 33.94% (n = 131) and 4.92% (n = 19), respectively.

Mean maximum keratometry (Kmax) was 52.63 ± 7.18 D, while mean thinnest pachymetry was 474.58 ± 49.54 µm. Mean asphericity (Q) was − 0.57 ± 0.47.

A total of 386 unique measurements on the KEPAQ-E were obtained, yielding adequate scale measures, with a person separation of 2.60 and a person reliability of 0.87. Active datapoints were 2687, accounting for 99.4% of potential responses. Only 0.6% of potential responses were answered as N/A. Only 22 patients obtained an extreme score, accounting for a floor and ceiling effect of 5.69%.

Item polarity ranged from 0.77 (Q_E07 fear about future) to 0.84 (Q_E04 place to another). Andrich’s threshold was shown to be well ordered (− 2.34; − 0.10; 2.45) and the category probability plot demonstrated well-functioning categories (Fig. 1). No category collapsing was deemed necessary. Item calibration ranged from − 1.14 logit (Q_E02 leave the house) to 1.73 logit (Q_E07 fear about future). No item was called to be either misfitting or overfitting (Table 3). Median person measure was 1.25 logit (interquartile range 3.65; skewness − 0.66; kurtosis 0.34; Kolmogorov-Smirnov P < 0.001).

Category probability plot for the emotional compromise sub-scale of the Keratoconus End-Points Assessment Questionnaire (KEPAQ-E)

PCA of the standardized residuals demonstrated that raw variance explained by measures was 66.9%. Eigenvalue for the first contrast was 2.57, explaining a variance of 12.2%. Two questions had a significant positive loading into a potential secondary dimension, Q_E06 confidence about future (loading 0.90) and Q_E07 fear about future (loading 0.88). A pilot separate analysis was performed by analyzing the KEPAQ-E but excluding Q_E06 and Q_E07, but the results obtained were not significantly different from what was obtained by analyzing all questions of the sub-scale together.

A DIF analysis was performed for sex (male vs. female), presence of keratoplasty (yes vs. no) and presence of corneal rings (yes vs. no), without finding any significant DIF in any item. DIF for presence or absence of corneal crosslinking, Q_E05 self-esteem, seemed to be slightly easier for the “presence of crosslinking” group (effect size 0.15; 95% confidence interval 0.02 to 0.28). All other items demonstrated a non-significant DIF behavior.

A Wright map was constructed for the emotional sub-scale (Fig. 2).

Wright map for the emotional compromise sub-scale of the Keratoconus End-Points Assessment Questionnaire (KEPAQ-E)

A total of 386 unique measurements on the KEPAQ-F were obtained, yielding adequate scale measures, with a person separation of 2.95 and a person reliability of 0.90. Active datapoints were 3441, accounting for 99.1% of potential responses. Only 0.9% of potential responses were answered as N/A. Only 36 patients obtained an extreme score, accounting for a floor and ceiling effect of 9.32%.

Item polarity ranged from 0.69 (Q_F01 play sports) to 0.90 (Q_E04 watch television). Andrich’s threshold demonstrated to be well ordered (− 2.13; − 0.44; 2.57) and the category probability plot demonstrated well-functioning categories (Fig. 3). No category collapsing was deemed necessary. Item calibration ranged from − 1.03 logit (Q_F02 objects near) to 1.63 logit (Q_E09 objects faraway). Regarding fitting, Q_F01 play sports was found to have an infit MNSQ and outfit MNSQ over 2.0, therefore deemed to be somewhat misfitting and potentially degrading of the overall score. No other item was called to be either misfitting or overfitting (Table 4). A new analysis was performed for the KEPAQ-F excluding question Q_F01, but results were not different from questions that were already analyzed. Therefore, this question was not deemed to degrade measurement. Median person measure was 0.82 logit (interquartile range 3.88; skewness 0.09; kurtosis − 0.45; Kolmogorov-Smirnov P < 0.001).

Category probability plot for the functional compromise sub-scale of the Keratoconus End-Points Assessment Questionnaire (KEPAQ-F)

PCA of the standardized residuals demonstrated that raw variance explained by measures was 67.6%. Eigenvalue for the first contrast was 1.88, explaining a variance of 6.8%. Therefore, the scale was considered to be unidimensional.

A DIF analysis was performed for sex (male vs. female), presence of keratoplasty (yes vs. no) and presence of corneal rings (yes vs. no), without finding any significant DIF in any item. DIF for presence or absence of corneal crosslinking, Q_F07 use computer, seemed to be slightly easier for the “presence of crosslinking” group (effect size 0.18; 95% confidence interval 0.05 to 0.26). All other items demonstrated a non-significant DIF behavior.

A Wright map was constructed for the functional sub-scale (Fig. 4).

Wright map for the functional compromise sub-scale of the Keratoconus End-Points Assessment Questionnaire (KEPAQ-F)

Rasch analysis converts raw scores into a Rasch-derived score ranging from minus infinite to infinite logit, although most measurements fall between − 7.0 and 7.0 logit. A score of 0.0 logits corresponds to a score that is equal to the mean difficulty of the items. Nevertheless, the authors are well aware that this kind of scoring may be difficult to comprehend for authors not dedicated to Rasch studies, and some authors [7] have suggested transforming the score into an easier-to-handle scale. To make results easier to digest for the non-Rasch user, and to standardize the grading of the scale for future international use, the authors have decided to perform a linear transformation ranging from 0 to 100, with a higher score representing a better quality of life.

A table for manually grading the responses of the patient in the KEPAQ-E and KEPAQ-F sub-scales has been developed and are presented in Tables 5 and 6. Using these tables, assign the score for each item corresponding to the response category selected by the patient. Add these scores and divide by the number of questions answered to arrive at the final KEPAQ score in their both sub-scales. To keep standardization, all results from the KEPAQ should be expressed in a 0 to 100 scale. Even when the researchers have obtained a logit value at the first instance, it should be linearly transformed for easy interpretation.

So far, no studies have determined the correlation of different scores of the KEPAQ to actual levels of ableness or dis-ableness due to vision. Therefore, it is too early to assign a clear functional classification to different scores of the KEPAQ and this sort of determination is not possible for now.

Currently, the best way of classifying the results of the KEPAQ resides on determining Tukey’s Hinges of the score distribution, and assigning it a denominative ordinal number from 1 to 4 as used in the ABCD classification proposed by Belin for use in keratoconus [5]. For the emotional sub-scale, Tukey’s Hinges are 74.27, 59.15 and 43.90, while for the functional sub-scale they are 69.14, 54.71 and 36.63. Based on these values, a simple table was created for classification (Table 7). Emotional classification should be anteceded by an “E”, so a person within the first and the second Tukey’s Hinge (grade 2) in the emotional sub-scale should be expressed as E2. The same applies with the functional sub-scale but preceded by the letter “F”. We have not included a “0” grading as this denotes an absence of disease [5] and the KEPAQ should only be used in patients with a confirmed diagnosis of keratoconus.

To facilitate computing and classification of KEPAQ measures, we have developed an Excel file that allows the user to easily input the patient’s responses for the scale, obtaining an immediate real-time measure with the corresponding E&F Classification. It allows for a quick and streamlined implementation into everyday practice for researchers and clinicians alike. The Excel file can be freely downloaded from Harvard Dataverse at https://dataverse.harvard.edu/api/access/datafile/4289090.

Please note that we have used the very same system as Belin’s ABCD classification [5], as our ultimate objective was to provide a more complete classification of keratoconus, with categories starting from A to F (ABCDEF). This way, the first three categories (ABC) include anatomical aspects, the fourth category (D) includes functional vision, while the last two letters (EF) comprise emotional and functional quality of life. The first four letters of the classification come from Belin’s paper [5] while the last two letters (EF) come from the present paper. We hope this will provide for a much richer classification of keratoconus to include anatomical, funcional, and subjective aspects of disease.

Our group is currently performing a study using Rasch modeling and machine learning algorithms for better determining the best cut-off values for both sub-scales of the KEPAQ.