Statistical analysis

The participants’ demographic data were descriptively analyzed. The CFS, TBUT, and Schirmer test from the worse eye was used for data analysis. For numerical data, the mean (SD) was used for data with normal distribution, while the median (range) was used for non-normally distributed data. The internal consistency was calculated to evaluate the reliability of the questionnaire; Cronbach’s alpha coefficient ≥ 0.7 was considered acceptable. Intraclass correlation coefficient (ICC) was calculated to determine the temporal relationship in test-retest reliability. Concurrent validity was evaluated using Pearson’s correlation coefficient to evaluate the correlations between the DEQS-Th scores and other measurements including the OSDI and EQ-5D-5L index scores.

Convergent validity is denoted by the level of correlation between constructs and instruments. These relations may be strong or weak correlations depending on the relationship expected between the constructs or instruments compared [28]. We created a correlation matrix between the QOL assessed by the subscale Impact on Daily Life scores of the DEQS-Th and scores from EQ-5D. Correlations would be expected to be high if the similar domain of impact of daily life and EQ-5D were assessed, thereby demonstrating convergent validity.

Discriminative validity was conducted to assess whether a measure can discriminate between the groups [28]. The total scores of DEQS-Th were analyzed to compare normal and clinical samples to indicate its discriminatory ability.

Responsiveness is defined as an ability of a measurement to detect clinically significant changes over time [29]. It was evaluated by comparing the DEQS-Th scores at baseline and follow-up periods after treatment using the standardized response mean (SRM). SRM values of 0.8, 0.5, and 0.2 were considered to be large, moderate, and small, respectively [30].

Floor or ceiling effects indicate the limitation of content validity, and reliability of the questionnaire. Floor or ceiling effects were suggested to be no more than 15% [31]; otherwise, it may affect responsiveness as the participants’ changes cannot be assessed [29].

To find the optimal cut-off score of DEQS-Th for suspected DED, the gold standard diagnosis for dry eye was made by using the OSDI score of ≥ 13 and the TBUT of ≤ 5 seconds. The receiver operating characteristic (ROC) curve was generated and the area under the ROC (AUC) was analyzed to determine the accuracy of the DEQS-Th. Sensitivity, specificity, positive predictive value, negative predictive value, and estimated cost were calculated. To simply apply in real-life practice, we also evaluated the Short Form DEQS-Th (SF DEQS-Th) for DED screening by using a sum of frequency scores of subscale Bothersome Ocular Symptoms. Our previous published data from non-DED participants were served as a control in some parts of the analysis [19]. A p-value < 0.05 was used to determine the significant level. SPSS program (version 22.0, SPSS Inc., Chicago, IL, USA) was used for data analysis.