Symptom-based stratification algorithm for heterogeneous symptoms of dry eye disease: a feasibility study

Abstract

Background/objective: To test the feasibility of a dry eye disease (DED) symptom stratification algorithm previously established for the general population among patients visiting ophthalmologists.

Subject/methods: This retrospective cross-sectional study was conducted between December 2015 and October 2021 at a university hospital in Japan; participants who underwent a comprehensive DED examination and completed the Japanese version of the Ocular Surface Disease Index (J-OSDI) were included. Patients diagnosed with DED were stratified into seven clusters using a previously established symptom-based stratification algorithm for DED. Characteristics of the patients in stratified clusters were compared.

Results: In total, 426 participants were included (median age [interquartile range]; 63 [48-72] years; 357 (83.8%) women). Among them, 291 (68.3%) participants were diagnosed with DED and successfully stratified into seven clusters. The J-OSDI total score was highest in cluster 1 (61.4 [52.2-75.0]), followed by cluster 5 (44.1 [38.8-47.9]). The tear film breakup time was the shortest in cluster 1 (1.5 [1.1-2.1]), followed by cluster 3 (1.6 [1.0-2.5]). The J-OSDI total scores from the stratified clusters in this study and those from the clusters identified in the previous study showed a significant correlation (r = 0.991, P < 0.001).

Conclusions: The patients with DED who visited ophthalmologists were successfully stratified by the previously established algorithm for the general population, uncovering patterns for their seemingly heterogeneous and variable clinical characteristics of DED. The results have important implications for promoting treatment interventions tailored to individual patients and implementing smartphone-based clinical data collection in the future.

Fig. 1. Hierarchical heat map of dry eye disease subjective symptoms determined by the Japanese version of the Ocular Surface Disease Index score (J-OSDI).

Stratified clusters identified by Uniform Manifold Approximation and Projection are shown on the left side of the heatmap as color bars. The dendrogram clustering for each of the J-OSDI items is shown at the top of the heatmap. Heatmap color intensity reflects the scores (0–4) from each J-OSDI item, with a color intensity scale shown at the top left angle.

Fig. 2. Heatmap of severity ranks of dry eye disease examination by clusters.

Heatmap showing the ranks of severity (0–7) determined from each dry eye disease examination test, discriminated according to the seven different clusters.

Fig. 3. Comparison of the Japanese version of the Ocular Surface Disease Index score (J-OSDI) between patients in this study and smartphone application users in a previous study by each stratified cluster.

a Paired box plot representing the J-OSDI total scores collected in the present study (red bars) and from users of a smartphone application in a previous study (blue bars). Equivalent clusters are juxtaposed in the plot. b Scatter plot showing the correlation between the median J-OSDI total score for each cluster in the present study (X-axis) and the median score for each cluster in the previous study (Y-axis).