Qiao Yang, Kai Yang, Yan Liu, Jue Wang, Guoping Feng
Earlier treatment, lower severity, and less complex morphology were associated with successful correction. The internally validated model may support pretreatment counseling and follow-up planning, but external validation is required before routine clinical use.
BACKGROUND: Nonsurgical ear molding is effective for congenital auricular deformities, but treatment response varies by patient- and ear-specific characteristics. This study developed and internally validated an ear-level prediction model for successful correction.
METHODS: This single-center retrospective cohort included infants undergoing nonsurgical ear molding between September 2022 and December 2025. The individual ear was the analytical unit. Successful correction was defined as an excellent or good outcome based on blinded photographic assessment. Six prespecified pretreatment predictors were entered into a multivariable logistic generalized estimating equation model accounting for within-infant correlation between bilateral ears. Missing predictor data were handled using multilevel multiple imputation. Model performance was evaluated using discrimination, calibration, Brier score, decision curve analysis, and infant-level cluster bootstrap validation. An exploratory restricted cubic spline analysis assessed the association between treatment-initiation age and correction probability.
RESULTS: The cohort included 328 infants and 482 eligible ears. Successful correction was achieved in 426 ears (88.4%). Older age at treatment initiation (OR, 0.94 per day; 95% CI, 0.91-0.97), greater baseline severity (OR, 0.26 per grade; 95% CI, 0.15-0.45), and mixed or other anomalies versus helical-folding deformities (OR, 0.32; 95% CI, 0.14-0.74) were associated with lower success. The apparent and optimism-corrected AUCs were 0.812 and 0.795, respectively. The optimism-corrected calibration slope was 0.965, calibration intercept -0.021, and Brier score 0.088. The age-success association remained significant in spline analysis (overall P < 0.001), with no evidence of nonlinearity (P = 0.41) and no distinct data-supported age threshold.
CONCLUSIONS: Earlier treatment, lower severity, and less complex morphology were associated with successful correction. The internally validated model may support pretreatment counseling and follow-up planning, but external validation is required before routine clinical use.