Development Of A Scoring System Using A Nomogram To Predict The Poor Visual Outcome In Culture-Proven Bacterial Keratitis: A 14-Year Retrospective Data Analysis
Published 2026 - 44th Congress of the ESCRS
Reference: PP20.04 | Type: Free paper | DOI: 10.82333/sbxh-nq75
Authors: Prof. Dr Chan-Ho Cho* 1 , Dr Ju Hong Park 1 , Prof. Dr Sang-Bumm Lee 1 , Dr Miri Hyun 2
1Ophthalmology,Yeungnam University College of Medicine,Daegu,Korea, Republic Of, 2Department of Infectious Diseases, Keimyung University Dongsan Hospital, Keimyung University School of Medicine and Institute for Medical Science,Daegu,Korea, Republic Of
Purpose
This study aimed to develop a score system that can predict the poor visual outcome in culture-proven bacterial keratitis using nomogram.
Setting
Retrospective, consecutive study. All of study cases were taken by Department of Ophthalmology, Yeungnam University Hospital, Daegu, South Korea. Study subjects included patients with microbiologically-proven bacterial keratitis who were hospitalized from 2007 to 2020. Totally 228 eyes were enrolled in this study.
Methods
The epidemiology factors, predisposing factors and initial clinical characteristics were analyzed. Poor visual outcome (PVO) was defined as final BCVA <0.1. The entire group was divided into training and testing data subgroups at a ratio of 8:2. The training data were used to develop the nomogram and the testing data to validate the nomogram. Multivariate logistic regression analysis was performed on the training data to select significant factors for the PVO group. The score was calculated using the coefficient values of the finally selected significant factors, and the nomogram was created using the sum of these scores for each patient. The created nomogram was validated through receiver operating characteristics (ROC) curve.
Results
Totally 59 eyes (25.9%) belonged to the PVO group. In training cohort, the scores for each selected variable for PVO were epithelial defect size (>16 mm2: 100 pts, <4 and ≤16 mm2: 68.32 pts, ≤4 mm2: 0 pt), initial VA <0.1 (94.95 pts), diabetes mellitus (88.73 pts), previous ocular surgery (66.14 pts), hypopyon (54.84 pts), previous ocular surface disease (50.78 pts), and contact lens wear (-115.58 pts). In the constructed nomogram, the area under the ROC curve was 0.898 in training data and 0.895 in testing data. Based on the training data, optimal cut-off point for predicting PVO was set at 291.51 points. At this point, the accuracy, sensitivity, and specificity were 84.6%, 66.0%, and 91.1%, respectively.
Conclusions
The model using the nomogram prediction for poor visual outcome of culture-proven bacterial keratitis showed good predictive power and high accuracy. In clinical practice, this scoring system can help predict the poor visual outcome in bacterial keratitis.