Abstract
BACKGROUND: Myopia is a growing health concern, especially among children, with Orthokeratology (OK) lenses showing promising results in myopia control. However, treatment outcomes vary significantly among individuals, highlighting the need for personalized approaches. This study aimed to develop and validate a predictive model for OK therapy outcomes in myopic children. METHODS: This retrospective cohort study included 439 myopic patients fitted with OK lenses. Patients were randomly divided into training (n = 308) and test (n = 131) sets. Least absolute shrinkage and selection operator regression was used for variable selection, followed by logistic regression to construct the predictive model. A nomogram was developed to visualize individual risk predictions. Model performance was assessed using calibration plots, receiver operating characteristic (ROC) curves, and decision curve analysis (DCA). RESULTS: Four variables were identified as significant predictors: age, parental myopia, white-to-white distance, and spherical refraction. The model demonstrated good discriminatory ability with areas under the ROC curve of 0.831 (95% CI: 0.786-0.877) in the training set and 0.820 (95% CI: 0.742-0.899) in the test set. Sensitivity and specificity were 75.6 and 72.8% in the training set, and 79.3 and 75.0% in the test set. Calibration plots and DCA confirmed the model's potential clinical utility across a range of threshold probabilities. CONCLUSION: This study developed a predictive model for OK therapy outcomes in myopic children. The model demonstrated good discriminatory ability in both training and test datasets. This predictive approach might contribute to risk stratification in myopia management. Further validation through prospective studies across diverse populations is needed before such models could potentially inform clinical decision-making and resource allocation in myopia control practice.