Abstract / Summary
Background: The clinical interpretation of myopia progression depends on the measurement interval, refractive protocol, and the population from which the analytic sample was selected.
Methods: We performed a secondary analysis of 200 myopic schoolchildren from 10 schools who had been selected for the intervention stage of a parent study after rapid progression (>=1.0 D/year). Parent-study documentation indicates a 12-month follow-up and progression calculated from the change in spherical-equivalent refraction (SE). The de-identified analytic archive retained only three 12-month progression values (1.25, 1.50, and 1.75 D); it did not retain paired eye-level refractions or sufficient metadata to independently verify the exact discretization mechanism or cycloplegic status of both examinations. We therefore contrasted a study-specific higher progression category (>=1.50 D; 99/200, 49.5%) with 1.25 D. A parsimonious logistic model used school-clustered robust standard errors (n = 199 complete cases; 99 outcomes; 10 predictor coefficients), and prediction was evaluated by leave-one-school-out cross-validation.
Results: None of the four questionnaire-derived behavioural indices was independently associated with the higher progression category with stable precision: device-time aOR 1.074 (95% CI 0.993-1.161), outdoor aOR 1.015 (0.929-1.109), near-work aOR 0.972 (0.825-1.145), and reading-risk aOR 1.091 (0.799-1.490). Grade level showed an association (aOR 1.210 per grade, 95% CI 1.099-1.332). Cross-school prediction was poor (AUC 0.540 for the core model and 0.570 after behavioural augmentation; Brier scores 0.252 and 0.254, respectively).
Conclusions: In this selected high-risk cohort, questionnaire-derived behavioural scores did not provide robust adjusted associations or clinically useful cross-school prediction. The findings should not be used for individualized risk stratification or causal behavioural claims. Future studies require prospectively documented cycloplegic refraction, axial length, treatment data, objective exposure measures, and a larger number of schools.