Rasch development and validation of the “Medellín Dry Eye Inventory” (ME·Dry), an instrument to measure multidimensional quality of life compromise in patients with dry eye disease

Abstract

Purpose:

Dry eye disease (DED) is a common condition that affects the quality of life. There is a great need for better-developed scales that comply with Rasch model requirements.

Methods:

Prospective study including patients with DED. A series of focus groups were performed to determine the best items to be included. A Rasch modeling methodology was used to validate the Medellín Dry Eye Inventory (ME·Dry). After iterative analysis and scale modification, a final version of the scale was attained which complied with the Rasch analysis expectations. Correlation between the different subscales of the ME·Dry and the Ocular Surface Disease Index (OSDI) was evaluated through Spearman correlation.

Results:

A total of 166 patients with DED were included. Rasch modeling demonstrated an excellent behavior for the ME·Dry, including four subscales: Symptoms, Triggers, Activity Limitation, and Emotional Compromise. Infit and Outfit parameters were all between 0.50 and 1.50, with excellent category utilization. Person and item separation and reliability were excellent for all subscales. There was a need for a category collapsing for the Emotional Compromise subscale. There was a strong correlation between the different subscales of the ME·Dry except for the Emotional Compromise subscale, which seems to be independent.

Conclusion:

The ME·Dry is a reliable scale, complying with the Rasch model expectations, that allows for a reliable measurement of quality of life compromise in patients with DED. Emotional compromise secondary to DED does not seem to correlate with disease severity as assessed by the other quality-of-life subscales.

Dry eye disease (DED) is an extremely common condition, affecting millions of people worldwide, and generating an enormous burden on quality of life (QoL) compromise in affected patients.[1] Therefore, measuring subjective compromise in these patients is of paramount importance.

So far, an important number of Patient Reported Outcome Measures (PROM) scales have been developed specifically for patients with DED. Nevertheless, recent research has risen concerns regarding their reliability and their ability to act as good-behaving instruments. Therefore, there has been a need for more modern, better-built PROMs (ideally using Rasch modeling methodology) to be created and validated in the DED population. This will provide instruments for better capturing the day-to-day continuum of having the ocular surface disease. Besides, it would be important for these new scales to evaluate the emotional aspect of patients, as DED has been demonstrated to correlate with depression and anxiety.

The present study describes the process of creating and validating a new scale for use in DED: the Medellín Dry Eye Inventory (ME·Dry), which complies with the Rasch model expectations, has strong psychometric characteristics and has been demonstrated to correlate well with the current gold standard in DED QoL measurement while also providing the ability to evaluate the emotional compromise of patients.

Methods

This is a prospective, analytical study, using Rasch modeling methodology, which sought to develop and validate a questionnaire for measuring multidimensional QoL compromise in patients with DED.

The development of the instrument has been careful, taking into account the recommendations by Trakman et al.[2] for this type of process. As will be discussed later, a series of successive steps have been taken to obtain a valid and applicable instrument for the population of DED.

This research adhered to the tenets of Helsinki’s Declaration, and proper approval was obtained at the Clínica de Oftalmología Sandiego Ethical Research Committee. All patients signed an informed consent after being given a full explanation of the details and objectives of the study.

Definition of the construct and obtaining of potential items

A very detailed explanation of the lengthy process of defining the potential items for inclusion in a scale has been previously explained in full depth by our group[3] and by Trakman et al.[2] A short description is included:

The idea behind the development of the ME·Dry was to create a scale that would provide a measurement of different dimensions of QoL affected by DED. Most previous scales rely heavily in measuring symptoms per se while QoL has been demonstrated to include a much broader group of domains.

To develop an item pool, two of the authors (K.B. and T.H.C.) carried out four focus groups with five patients with a proven diagnosis of DED in every group, for a total of 20 patients. During focus groups, all patients were interviewed and later asked to write all the situations in which they felt that the DED affected their daily QoL. Later, all ideas were homogenized and turned into questions evaluating the degree of compromise in each aspect. From a visual inspection, four different dimensions of QoL compromise were obtained: Ocular Symptoms, Environmental Triggers, Activity Limitation and Emotional Compromise. Every domain was therefore built, treated, and analyzed as a different subscale. Items in every subscale were purposely written in simple and very similar language, being different only in the specific aspect evaluated by every item. Double negatives and two-edged questions were avoided to improve understanding of the instrument.[2]

The authors defined to keep all items “closed” with a Likert-like system, meaning that only a set of predefined potential answers were given to the patient. The different response options along with their respective coding were: “never” = 0, “rarely” = 1, “sometimes” = 2, and “most or all of the time” = 3. Patients were also given the option of responding “not applicable” in case they felt that the specific item did not correlate to any aspect of their daily life. Increasingly “worse” answers were given an increasing value, therefore, a higher score in the ME·Dry means a greater compromise by disease, keeping in line with the usage in most current DED scales.[4]

A Content Validity Index (CVI) using three Ophthalmologists with experience in managing DED was performed as has been previously described by Trakman et al.[2] Through this process, an initial pool of 95 potential items was decreased to a total of 43 questions (17 for Ocular Symptoms, 6 for Environmental Triggers and 11 to Activity Limitation and nine to Emotional Compromise). These 43 questions were included into the Rasch modeling process.

Rasch modeling

To soundly evaluate the psychometric characteristics of the ME·Dry scale, a Rasch modeling process was performed. Rasch is currently regarded as the gold standard of modern psychometric evaluation of outcome scales,[5] with significant advantages over previous methodological approaches, now collectively known as “Classical Test Theory”.[6]

A detailed explanation on the characteristics of Rasch modeling and its advantages is well outside the scope of the present paper. The interested reader is forwarded to the introductory article by Vanhoutte et al.[7] and the excellent works by Boone et al.[8,9] In short, Rasch modeling provides a mathematical framework to study the behavior of a scale, and to determine whether the items actually provide a valid measurement of a latent trait of interest, in this case, QoL compromise secondary to DED. It is much more stringent than Classical Test Theory, so scales complying with the Rasch Model expectations are expected to have excellent behavior in real-life measurement scenarios. Besides, probably the most important characteristic of Rasch modeling is that it provides for the calculation of a final score in an interval-level form, amenable to parametric statistical analysis.[7] This is a great improvement over the mere ordinal-level data obtained from scales created using Classical Test Theory, such as most previously described DED questionnaires.[10]

Therefore, Rasch modeling was used as a way of guaranteeing the excellent psychometric behavior of the final ME·Dry scale. First, nonfunctioning items were excluded until a final, well- functioning version of every subscale was achieved. Then, a final, complete Rasch modeling analysis was performed on every subscale and all its psychometric properties were studied and described.

Data analysis

First, data was introduced into an ad-hoc database created for the project in Numbers version 10.3.5 (Apple Inc; Cupertino, CA, United States). Then, it was exported into a comma-separated values (.csv) file and imported into Winsteps version 4.7.1 (J. M. Linacre; Beaverton, OR, United States) where four control files were created, one for every individual subscale. First, item polarity through Point-Measure correlation was checked as a means of diagnosing potential mistakes in the database. Then scale quality parameters were determined, including a Person Separation (expected to be over 2.0 for functioning scales) and Person Reliability (expected to be over 0.80). Category function was evaluated by inspection of Andrich’s Thresholds, expecting a monotonic increase in category usage as global Person Measure increases. In case of category disordering, an iterative process of category-collapsing was undertaken until a well-functioning scale was achieved. Visual inspection of thresholds was also performed by a Category Probability Plot graphed through IRT Illustrator version 2017.1 (Psychomeasurement Systems LLC; Charlottesville, VA, United States).

Item fitting to the Rasch model was measured by both the Infit Mean Square (MNSQ) and Outfit MNSQ. Keeping up with Wright and Linacre’s recommendations,[11] a fit value between 0.50 and 1.50 was considered of good clinical use, whereas values between 1.51 and 2.00 were considered poorly useful but not necessarily degrading of the score. Fit levels over 2.01 were considered as degrading of the scale and eliminated right away. A Principal Component Analysis (PCA) of standardized residuals was used to evaluate whether the scale was or not unidimensional. An eigenvalue over 2.00 for the first or second contrast was considered as strong evidence of a secondary dimension spanning at least two items. Raw variance explained by measures was also evaluated and expected to be over 50%.[3]

After initial analysis, poorly functioning items (misfitting or loading strongly to a secondary dimension) were excluded one by one until a final version of the scale was achieved under iterative analyses. This final version was expected to comply with all of the Rasch model’s expectations and a full analysis was performed.

Item calibration (“difficulty”) was measured for all remaining items, as well as a final Person Measure for every subscale in every patient. As a means of making final Rasch scores easier to interpret and also to keep similarity with the OSDI, final scores were linearly transformed into a 0 to 100 scale, with a higher value meaning more disability.

Final score of the OSDI was calculated as previously described in the Dry Eye WorkShop (DEWS),[4] and patients were categorized as having a normal ocular surface (0–12 points) or as having mild (13–22 points), moderate (23–32 points), or severe (33–100 points) ocular surface disease.[12] A Spearman’s Rank Correlation Coefficient was used to evaluate the correlation between the OSDI scores and the final transformed scored of every one of ME·Dry subscales.

All non-Rasch analyzes were performed on IBM SPSS Statistics version 23 (International Business Machines Corporation; Armonk, NY, United States).

Bioethics

There are no potential conflicts of interests related to this article. This research adhered to the tenets of the Helsinki’s Declaration and proper ethical approval was obtained at the Ethical Committee of the Clínica de Oftalmología Sandiego. All patients provided a written informed consent for their participation in this study.

Results

Studied sample

A total of 166 patients with a confirmed diagnosis of DED were included. Most of the patients (n = 136; 81.9%) were female, and the median age was 46.50 years (Interquartile range: 26.00; Kolmogorov–Smirnov, P < 0.001). Five patients (3.0%) had a history of Sjögren’s syndrome. A total of 146 patients (87.9%) were currently under at least one lubricant eye drop. The most commonly used lubricating compounds were 4% sodium hyaluronate (n = 52; 31.3%) followed by 0.5% carboxymethylcellulose (n = 49; 29.5%).

Symptoms Subscale

An initial Rasch analysis was performed including all Symptoms items (S1–S17). Nevertheless, two items (S15 and S17) demonstrated to be significantly misfitting and disordered. As they were considered to distort measurement, they were dropped from final analysis.

A final analysis including all Symptoms items except for S15 and S17 was performed. It proved to be a viable scale with adequate person and item separation and reliability [Table 1]. Raw variance explained by measures was 58.1%, and the scale proved to be unidimensional (Eigenvalue for the first contrast 1.93). Fit values were adequate for all items. All categories were used by a significant number of patients (range: 21% to 31%) in an ordered fashion [Fig. 1]. The scale was well targeted with a mean nontransformed person score of 0.18 ± 1.32 logit. A Wright map including both person and item measures was created [Supplementary Material 1 ^{(708.4KB, tif)} ].

Table 1.

Scale Rasch Characteristics

Item	Measure	Standard Error	Infit MNSQ	Outfit
S1	−1.49	0.12	0.84	0.95
S2	−0.56	0.11	0.97	1.16
S3	−1.06	0.11	0.66	0.68
S4	0.07	0.10	1.19	1.14
S5	−1.14	0.12	0.77	0.84
S6	0.02	0.10	0.88	0.94
S7	0.89	0.10	0.87	0.78
S8	0.02	0.10	1.39	1.34
S9	1.04	0.10	1.11	0.92
S10	−0.17	0.10	1.11	1.25
S11	1.25	0.11	1.34	1.36
S12	−0.47	0.10	0.90	0.80
S13	0.65	0.10	0.98	0.87
S14	−0.22	0.10	0.87	0.82
S16	1.17	0.11	1.07	0.93
Person	Separation 2.99		Reliability 0.90
Item	Separation 7.54		Reliability 0.98
Triggers Subscale
Item	Measure	Standard Error	Infit MNSQ	Outfit
T1	−0.17	0.12	0.92	1.07
T2	−0.32	0.13	0.64	0.69
T3	0.21	0.12	1.07	1.14
T4	−0.37	0.12	1.16	1.09
T5	−0.85	0.13	0.90	0.96
T6	1.51	0.14	1.22	1.04
Person	Separation 2.06		Reliability 0.83
Item	Separation 5.51		Reliability 0.97
Activity Limitation Subscale
Item	Measure	Standard Error	Infit MNSQ	Outfit
L1	0.78	0.15	0.97	0.94
L2	−0.58	0.14	1.23	1.19
L3	0.02	0.11	1.25	1.23
L5	0.71	0.12	0.87	0.78
L6	−0.85	0.11	1.07	1.00
L7	−0.74	0.12	0.97	1.09
L8	−0.20	0.11	0.90	0.89
L9	0.52	0.12	0.76	0.81
L10	0.86	0.12	0.88	0.73
L11	−0.52	0.11	0.92	0.94
Person	Separation 2.36		Reliability 0.85
Item	Separation 4.81		Reliability 0.96
Emotional Compromise Subscale
Item	Measure	Standard Error	Infit MNSQ	Outfit
E1	−0.85	0.17	1.12	1.08
E2	−1.56	0.17	0.98	1.30
E3	1.13	0.19	1.11	1.36
E4	−0.44	0.17	0.89	0.84
Emotional Compromise Subscale
Item	Measure	Standard Error	Infit MNSQ	Outfit
E5	1.48	0.20	0.87	0.74
E6	0.45	0.18	0.83	0.96
E7	0.97	0.19	0.88	0.80
E8	−0.81	0.17	1.20	1.25
E9	−0.37	0.18	0.93	0.89
Person	Separation 2.25		Reliability 0.83
Item	Separation 5.19		Reliability 0.96

Figure 1.

Category utilization for every subscale of the final version of the ME·Dry. Categories 1 and 2 have been collapsed for the Emotional Compromise subscale

Final transformed score of the Symptoms subscale was 51.7 ± 13.0 (range: 11.9–80.9).

Triggers subscale

A Rasch analysis was performed including all Triggers items (T1–T6) proving a viable scale with adequate person and item separation and reliability [Table 1]. Raw variance explained by measures was 57.7% and the scale proved to be unidimensional (Eigenvalue for the first contrast 1.65). Fit values were adequate for all items. All categories were used by a significant number of patients (range: 14% to 40%) in an ordered fashion [Fig. 1]. The scale was well targeted with a mean nontransformed person score of 0.76 ± 1.82 logit. A Wright map including both person and item measures was created [Supplementary Material 1 ^{(708.4KB, tif)} ].

Final transformed score of the Triggers subscale was 58.3 ± 20.1 (range: 0.0–99.9).

Activity limitation

An initial Rasch analysis was performed including all Activity Limitation items (L1–L11). Nevertheless, one item (L4) demonstrated to be significantly misfitting and disordered (Infit 1.65, Outfit 1.72, disordered thresholds). As it was considered to distort measurement, it was dropped from final analysis.

A final analysis was performed including all Activity Limitations items except for L4, proving a viable scale with adequate person and item separation and reliability [Table 1]. Raw variance explained by measures was 50.5% and the scale proved to be unidimensional (Eigenvalue for the first contrast 1.65). Fit values were adequate for all items. All categories were used by a significant number of patients (range 20% to 37%) in an ordered fashion [Fig. 1]. The scale was well targeted with a mean nontransformed person score of –0.43 ± 1.93 logit. A Wright map including both person and item measures was created [Supplementary Material 1 ^{(708.4KB, tif)} ].

Final transformed score of the Activity Limitation subscale was 44.6 ± 20.5 (range: 0.0–100.0).

Emotional compromise

An initial Rasch analysis was performed including all Emotional Compromise items (E1–E9), finding a poorly functioning scale (person separation 1.86, person reliability 0.78). The lowest category (0) was used by most patients (53.0%) and Andrich Thresholds were nonoptimal [Fig. 2]. Item fit was good, and there was no individual item to blame for the poor functioning of the scale. Therefore, a decision was made to collapse categories 1 and 2.

Figure 2.

Category utilization for the Emotional Compromise subscale before category collapsing. Please note the poor category utilization for the category 1

A collapsed version of the scale proved to behave much better regarding person separation and reliability [Table 1]. Category utilization also improved significantly [Fig. 1]. Raw variance explained by measures was 60.0% and the scale proved to be unidimensional (Eigenvalue for the first contrast 1.78). Fit values were adequate for all items. The scale was well targeted with a mean nontransformed person score of –0.92 ± 2.43 logit. A Wright map including both person and item measures was created [Supplementary Material 1 ^{(708.4KB, tif)} ].

Final transformed score of the Emotional Compromise subscale was 38.3 ± 23.5 (range: 0.0–100.0).

Final Version of the ME·Dry

The final version of the ME·Dry can be found on Table 2 while data for transforming raw score into interval Rasch-derived score can be found on Table 3.

Table 2.

Final Version of the ME·Dry scale

During the last week…	Never	Rarely	Some- times	Most or all of the time	Not Applicable
My eyes have felt dry	0	1	2	3	X
My eyes have felt gritty	0	1	2	3	X
My eyes have felt burning	0	1	2	3	X
My vision has been worse during the night than during the rest of the day	0	1	2	3	X
My eyes have felt tired	0	1	2	3	X
I’ve felt the need to keep my eyes closed even though I was not sleepy	0	1	2	3	X
I’ve felt pain on my eyes upon waking up	0	1	2	3	X
My vision has been distorted	0	1	2	3	X
I’ve felt like my eyelids are sticking to my eye- balls and I can’t open them	0	1	2	3	X
My eyes have gotten red repeatedly	0	1	2	3	X
I’ve felt like I had cut my eyes with something	0	1	2	3	X
I’ve felt like I have a foreign body in my eyes	0	1	2	3	X
I’ve felt pain in my eyes upon looking sideways or up or down	0	1	2	3	X
My eyelids have felt heavy	0	1	2	3	X
I’ve felt like my eyelids are closing by them- selves	0	1	2	3	X
Environmental Triggers
During the last week my eye symptoms have gotten worse than usual after…	Never	Rarely	Some- times	Most or all of the time	Not Applicable
Reading	0	1	2	3	X
Using the computer	0	1	2	3	X
Being in a room with air conditioning	0	1	2	3	X
Being exposed to the sun	0	1	2	3	X
Being exposed to a dusty or smokey environ- ment	0	1	2	3	X
Driving	0	1	2	3	X
Activity Limitation
During the last week my eye symptoms have lim- ited my ability to…	Never	Rarely	Some- times	Most or all of the time	Not Applicable
Drive during the day	0	1	2	3	X
Drive during the night	0	1	2	3	X
Recognize the face of people	0	1	2	3	X
Walk up or down stairs	0	1	2	3	X
Read books or documents	0	1	2	3	X
Use the computer	0	1	2	3	X
Do my job	0	1	2	3	X
Play sports	0	1	2	3	X
Do my household chores	0	1	2	3	X
Walk in places with poor illumination	0	1	2	3	X
Emotional Compromise
During the last week my eye symptoms have…	Never	Rarely	Some- times	Most or all of the time	Not Applicable
Made me unhappy	0	1	1	2	X
Generated fear	0	1	1	2	X
Made me feel inferior to others	0	1	1	2	X
Taken away my confidence about the future	0	1	1	2	X
Made it hard relating to others	0	1	1	2	X
Made me feel different from others	0	1	1	2	X
Made me feel ugly	0	1	1	2	X
Made me feel older than I actually am	0	1	1	2	X
Taken away my confidence in myself	0	1	1	2	X

Table 3.

Raw score to Rasch measure conversion

Raw Score	Rasch Measure	Raw Score	Rasch Measure
0	0	23	50.2
1	11.9	24	51.2
2	18.9	25	52.1
3	23.1	26	53.1
4	26.1	27	54.0
5	28.5	28	55.0
6	30.5	29	56.0
7	32.3	30	57.1
8	33.9	31	58.1
9	35.4	32	59.2
10	36.7	33	60.4
11	38.0	34	61.6
12	39.2	35	62.8
13	40.3	36	64.2
14	41.4	37	65.6
15	42.4	38	67.2
16	43.5	39	69.0
17	44.5	40	71.1
18	45.5	41	73.5
19	46.4	42	76.6
20	47.4	43	80.8
21	48.3	44	87.9
22	49.3	45	100
Triggers Subscale
Raw Score	Rasch Measure	Raw Score	Rasch Measure
0	0	10	52.8
1	14.1	11	55.8
2	23.0	12	59.0
3	28.8	13	62.4
4	33.4	14	66.3
5	37.2	15	70.8
6	40.6	16	76.7
7	43.8	17	85.7
8	46.9	18	100
9	49.8
Activity Limitation Subscale
Raw Score	Rasch Measure	Raw Score	Rasch Measure
0	0	16	50.4
1	12.9	17	51.9
2	20.5	18	53.5
3	25.1	19	55.1
4	28.5	20	56.8
5	31.2	21	58.6
6	33.6	22	60.4
7	35.7	23	62.5
8	37.6	24	64.7
9	39.4	25	67.3
10	41.1	26	70.2
11	42.7	27	73.9
12	44.3	28	78.8
13	45.8	29	86.7
14	47.4	30	100
15	48.9
Emotional Compromise Subscale
Raw Score	Rasch Measure	Raw Score	Rasch Measure
0	0	10	53.2
1	12.8	11	56.6
2	21.3	12	60.1
3	27.0	13	63.9
4	31.7	14	68.0
5	35.8	15	72.7
6	39.6	16	78.6
7	43.1	17	87.1
8	46.6	18	100
9	49.9

Correlation between subscales

There was a strong correlation in the scores of three of the subscales: Symptoms, Triggers, and Activity Limitation. The Emotional subscale score did not demonstrate to be correlated to any of the other subscales [Fig. 3].

Figure 3.

Pearson correlation between the different subscales of the ME·Dry. For the scatterplots, the solid line represents the correlation line while the gray area represents a 95% confidence interval of the correlation. Bell-shaped lines represent value densities for each subscale demonstrating a reasonably normally distributed set of scores. Corr = Correlation. ***represents P < 0.001

Relationship between OSDI and ME·dry

A total of 162 patients answered the OSDI, obtaining a median value of 38.00 (Interquartile range: 25.75; Kolmogorov–Smirnov, P = 0.023). A total of 14 patients (8.6%) were classified as “no DED” and “mild DED” each. “Moderate DED” was classified in 32 patients (19.8%) while 102 (63.0%) had an OSDI score over 32, accounting to a “severe DED” classification. Spearman correlation between OSDI score and ME·Dry Ocular Symptoms, Environmental Triggers, Activity Limitation and Emotional Compromise subscales were ρ = 0.671, ρ = 0.642, ρ = 0.755, and ρ = 0.631, respectively (all P < 0.001).

Discussion

DED is an extremely common eye condition, affecting millions of people worldwide, and is currently intensely studied.[13] DED has the potential of significantly affecting the patient’s daily functioning, and measuring QoL in this group is of paramount importance.[14]

A significant number of questionnaires have been developed and validated for evaluation of patients with DED. This can be very convenient for standardizing complaints, and facilitating research. They can also be used as outcome measures for clinical trials, including those registered at the Unites States Food and Drug Administration (FDA).[4] Taking into account that results derived from their use, these scales need to be thoroughly and consciously evaluated in order for the scientific community to be sure of their psychometric behavior. Despite of the great proliferation of scales for measuring QoL in DED patients, many of them have important psychometric flaws or measure a way-too-narrow spectra of complaints by the patient.

The OSDI is currently regarded as the gold standard in measurement of QoL in ocular surface disease subjects. Nevertheless, although some studies using CTT have suggested that the OSDI may be useful,[15] papers using Rasch methodology have found the OSDI to be multidimensional and poorly targeted.[16] McAlinden has previously found the OSDI to have a Person Separation of only 0.94, way below the minimum threshold of 2.0, therefore suggesting that the OSDI does not provide a “valid measurement” on the basis of Rasch analysis.[17] These discrepancies between studies using CTT and those using Rasch modeling reside precisely on Rasch being a much more stringent process, having the capability of demonstrating special psychometric flaws in scales previously (and erroneously) labelled as adequate by CTT methods.[18]

Rasch methodology evaluation of a number of other DED questionnaires has also demonstrated them to have misbehaving psychometric characteristics. Recent Rasch modeling studies evaluating the McMonnies questionnaire have demonstrated it to have poor psychometric properties,[19] and to not provide a valid scale for measurement on the basis of Rasch expectations.[17] Regarding the Standard Patient Evaluation of Eye Dryness (SPEED) questionnaire, research has suggested it to be multidimensional, and to require a modification to comply with the Rasch model expectations.[20,21] The regular version of the Contact Lens Dry Eye Questionnaire-8 (CLDEQ- 8) was proven to be multidimensional and misfitting, and removing 50% of the items was necessary for it to comply with the Rasch model expectations.[21] Many other DED questionnaires have never been evaluated through Rasch methodology so there is still doubt over their psychometric characteristics.

Evaluated realms of QoL is another important aspect of evaluation in ocular surface disease patients. Most questionnaires focus too much on symptoms and activity limitations, and do not measure an important latent trait such as emotional compromise. The Symptoms Assessment in Dry Eye (SANDE) questionnaire only focuses on DED symptoms per se (frequency and severity), but fails to evaluate other aspects of QoL.[22] The 5-Item Dry Eye Questionnaire (DEQ-5) only measures frequency and severity of three ocular symptoms: eye discomfort, eye dryness and watery eyes.[4] The Ocular Confort Index (OCI) focuses mainly on ocular symptoms and has two questions on visual function.[23] The OSDI has three subscales: vision-related function, ocular symptoms and environmental triggers[4]; but does not measure emotional compromise due to DED. This is especially important due to the fact that depression and anxiety are both more prevalent and more severe in patients with DED.[24,25] Dry eye symptoms have also been linked to an increase in suicide ideation in a recent paper by Um et al.[26]

Having a questionnaire that comprehensively evaluates the different aspects of health-related QoL is of paramount importance, and would provide a good instrument for both clinical and research endeavors. One of the main advantages of the ME·Dry is precisely that it covers a wide array of aspects for comprehensively evaluating QoL while still being shorter than other comprehensive scales such as the Impact of Dry Eye in Everyday Life (IDEEL) questionnaire.[27] This will help adequately measure DED burden while still being compatible with busy clinics. Another of the clear advantages of the ME·Dry is that it was created and validated from the ground-up through Rasch methodology, which warranties it having excellent psychometric properties and provides the clinician with the confidence that the scale will actually provide an adequate measurement quality. The fact that it correlated closely with the OSDI provides evidence of a well-developed construct measurement, compatible with the current gold standard in DED QoL measurement.

Besides, the fact that the ME·Dry generates a Rasch scores is important, as this is of interval-level nature and amenable to parametric statistical analyzes.[8] This is not possible with scales residing in CTT background,[10] such as the OSDI.

The results from this paper confirm the ME·Dry to be a valid scale that allows for the multidimensional QoL evaluation of patients with DED. Scores from the Symptoms, Triggers, and Activity Limitation subscales are greatly correlated, as may be expected. The fact that there was no correlation between the Emotional Compromise subscale and the rest of the subscales may point toward the fact that emotional impact of DED is not necessarily correlated with the degree of disease severity per ser.

This paper provides evidence on the fact that the ME·Dry provides a very reliable scale for measuring multidimensional compromise in patients with DED, including their symptoms, the events that trigger them, how much it limits their everyday activities, and how much it affects them emotionally.

Scoring the ME·Dry is simple and straightforward using the included scoring tables. The patient answer the ME·Dry and then raw scores are converted and a person measure is obtained for every subscale. Please keep in mind that two of the categories of the Emotional Compromise subscale have been collapsed; therefore, two of them will provide the exact same raw score. Being a very new scale, no information exists regarding different thresholds for classification. The authors of this scale are currently performing a study to determine the ME·Dry values that best correlate with different disease severity stages.

Prospects for future use of the scale

The authors believe that the ME·Dry scale may be useful for both clinical and research scenarios and are conscious that unrestricted access to the scale is of benefit to the scientific community.; therefore no licensing is necessary for the academic or clinical use of the scale. Both researchers and clinicians can make free use of the scale as long as proper attribution is given and the scale is not modified. A short communication to Balparda (kb@kepabalparda.com) expressing the desire to use the scale is greatly appreciated but not necessary. Use of the scale for commercial use will require a formal approval by the authors, which can be sought via the same email address.

Conclusion

The ME·Dry is a psychometrically robust scale, developed from the ground-up using Rasch methodology. It provides an excellent and reliable measurement of multidimensional QoL compromise in patients with DED, including their symptoms, the events that trigger them, how much it limits their everyday activities, and how much it affects them emotionally.

Author contributions

All authors participated in the study concept and design. Data was collected by XC and FG. Data analysis was performed by KB and THC. Initial manuscript draft was written by KB and FG. All authors reviewed and approved the final version of the manuscript.

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the Ethics Committee at the Clínica de Oftalmología Sandiego and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.

Supplementary Material

Wright map for the subscales of the ME·Dry.