A novel scale for anxiety-related fixation instability during laser in situ keratomileusis

Abstract

This cohort study aimed to investigate the correlation between the severity of anxiety during laser in situ keratomileusis (LASIK) and fixation instability, as measured and plotted by the eye tracker during photo-ablation, and to develop a novel quantitative scale for anxiety-related fixation instability. LASIK was performed to correct myopia and hypermetropia with and without astigmatism in 2435 eyes of 2435 patients. Participants fulfilled the seven-item anxiety sub-score of the Hospital Anxiety and Depression scale questionnaire for scaling patients' anxiety levels before LASIK into normal, borderline, and anxiety case groups. The eye-tracking pupil center position plots, patient's heart rate, and surgeon-reported level of patient's cooperation during the procedure were analyzed. An anxiety-related fixation instability score (FIS) was calculated, for which the best cut-off points to differentiate between normal, borderline, and anxiety case groups were defined. The FIS showed a high performance in separating participants into normal, borderline, and anxiety case groups and when used as a scale (0-90), values from 0 to 12 are considered normal, from 13 to 36 are considered borderline, and from 37 to 90 are considered anxiety cases. The FIS and scale are useful objective tools to quantify anxiety-related fixation instability during LASIK.

Keywords: Anxiety; Eye-tracker; Fixation instability; Hospital Anxiety and Depression Scale; Laser in situ keratomileusis.

Fig. 1

Eye tracker’s pupil center plot. Samples of eye tracker’s pupil center plots exported from the WaveLight EX500.

Fig. 2

Centroids in eye-tracker’s pupil center plots. The same plot samples as in Fig. 1, after defining centroids using a weighted average algorithm. Centroids of isolated islands formed by confluent pupillary center position marks are marked with yellow rhombi, and the calculated weighted average centroid position of multiple islands is marked with a red rhombus.

Fig. 3

Spearman rank correlation of the studied parameters. Heatmaps displaying Spearman rank correlation coefficients of all parameters before (A) and after (B) filtering for statistically significant correlations with p value ≤ 0.05. HADS = Hospital Anxiety and Depression Scale.

Fig. 4

Relationship between fixation instability score, HADS anxiety sub-score with age, spherical equivalent and treatment duration. Combined categorical (male/female) and facet grid plots showing the relationship between fixation instability score, Hospital Anxiety and Depression Scale (HADS) anxiety sub-score with age (A), spherical equivalent (B), and treatment duration (C), respectively. The dashed lines indicate the observed regression.

Fig. 5

Fixation instability scores among the 3 groups of HADS anxiety sub-score Legend: The distribution of fixation instability scores between normal, borderline, and case groups (classes) of the Hospital Anxiety and Depression Scale anxiety subscale.

Fig. 6

Receiver operating characteristic (ROC) diagrams. Extension of receiver operating characteristic (ROC) to one-vs-one multiclass for 3 Hospital Anxiety and Depression Scale anxiety subscale groups (normal, borderline, and case) in the test subset. (A) ROC results for support vector machine (SVM) using the Sum Area parameter, (B) ROC results for SVM using the Sum perimeter parameter, (C): ROC results for SVM using the Average Centroid deviation parameter, (D) ROC results for SVM using the Fixation instability score.

Fig. 7

Support vector machine (SVM) classifiers. (A) Results for binary (normal vs. borderline) classification task by a trained support vector machine (SVM) on the test subset using the Fixation Instability score (FIS). A-1: Receiver operating characteristic (ROC) curve with an area under the curve (AUC) of 0.98. The ROC curve is marked by a black dot at the site closest to the perfect classification point. A-2: Plot showing probability score at the cut-off point of 12 and the Youden Index. A-3: The corresponding confusion matrix. (B) Results for binary (borderline vs. case) classification task by a trained SVM on the test subset using the FIS. B-1: ROC curve with an AUC of 0.94. The ROC curve is marked by a black dot at the site closest to the perfect classification point. B-2: Plot showing probability score at the cut-off point of 36 and the Youden Index. B-3: The corresponding confusion matrix.

Fig. 8

Average maps of pupil center tracking plots. Average maps of pupil center tracking plots for all right and left eyes in each Hospital Anxiety and Depression Scale anxiety subscale group (surgeon’s view). The center of the crosshair corresponds to the central fixation target. Lighter areas indicate the more commonly spotted locations of the pupil center during tracking.