Abstract:AIM: To develop and externally validate a risk prediction model for adolescents wearing orthokeratology (OK) lenses to predict the probability of suboptimal control effects. METHODS: A retrospective analysis of myopic children who used OK lenses consistently for at least 18mo between January 2021 and March 2023 were conducted. Various ophthalmic examinations and recorded baseline parameters were performed. A risk prediction model was developed using multifactorial logistic regression analysis, and the model’s discrimination and calibration were assessed using the area under the receiver operating characteristic curve (AUC) and Brier score. RESULTS: The modeling group included 228 eyes (mean age, 10.46±1.99y; 55.7% males), with 120 right eyes (52.6%) and 108 left eyes (47.4%), and the validation group included 100 eyes. Anisometropia was absent in 174 eyes (76.3%) and present in 54 eyes (23.7%). The risk prediction model identified age, anisometropia, and corneal diameter as independent predictors for the control effect of OK lenses. The model showed good discrimination [AUC 0.781, 95% confidence interval (CI) 0.721–0.841] and acceptable calibration in derivation (Brier score 0.186) in the validation group. CONCLUSION: The study constructs a risk prediction model to forecast the likelihood that myopic adolescents will not achieve the ideal control effect after wearing OK lenses, offering proactive and effective guidance for frontline clinical practitioners in devising diagnostic and therapeutic strategies.
