Early identification of myopia risk in children through school-based vision screening: a longitudinal cohort study.

PURPOSE: To develop and validate a myopia onset prediction model and a simplified scoring tool for school-aged children using longitudinal data from a routine school‑based vision screening program, based on parameters obtainable on‑site. METHODS: This retrospective longitudinal cohort study included data from a school‑based vision screening program conducted between 2021 and 2025, which measured non‑cycloplegic spherical equivalent (SE) and axial length (AL). A total of 3,749 children with complete baseline measurements were included. Among them, 1,969 children who were non‑myopic at baseline formed the prediction model development cohort. A multivariable logistic regression model was developed using baseline SE, AL, age, and gender, with myopia onset as the outcome. The model was trained on data from 2021 to 2023 and temporally validated on an independent cohort from 2024. A simplified risk scoring system was then derived from the regression coefficients. RESULTS: Over a minimum follow‑up of 0.5 years (0.97 ± 0.69 years), the model achieved AUCs of 0.795 (training) and 0.754 (validation). The simplified scoring system showed moderate discrimination (AUC = 0.739). Risk stratification into low‑, medium‑, and high‑risk groups yielded significantly different myopia‑free survival (p < 0.001), with cumulative incidences of 8.7%, 25.8%, and 87.3%, respectively. Decision curve analysis indicated net clinical benefit when using the model to guide interventions. CONCLUSION: The prediction model and simplified scoring tool, based on non‑cycloplegic refraction and AL, may help stratify children into myopic risk categories for targeted prevention in school and primary care settings. Further calibration improvement is needed before widespread implementation.