Development of an explainable prediction model for the risk of moderate-to-severe obstructive sleep apnea in children
摘要
Early identification of children at high risk for moderate-to-severe obstructive sleep apnea (OSA) is crucial for timely intervention, yet is often hindered by limited access to polysomnography (PSG). We aimed to develop an interpretable clinical prediction model using easily obtainable clinical and inflammatory biomarkers to distinguish moderate-to-severe from mild pediatric OSA. We conducted a retrospective study of 164 children diagnosed with OSA by PSG. From multiple biomarkers and clinical variables, least absolute shrinkage and selection operator (LASSO) regression was employed to select the most predictive features. A multivariable logistic regression model was built and presented as an interpretable nomogram. Model performance was evaluated via bootstrap validation assessing discrimination, calibration, and clinical utility. The LASSO algorithm identified eight core predictors: female, tonsil size grades 3 and 4, adenoid-to-nasopharynx ratio (A/N ratio), IgE, IL-4, IL-6, and IL-10. The final model demonstrated robust performance, with a bootstrap-corrected AUC of 0.763 (95%CI 0.690–0.836). Decision curve analysis confirmed the model’s clinical utility. Conclusion: We developed an explainable nomogram that integrates upper airway anatomy, allergic, sex, and specific inflammatory cytokines. This tool provides clinicians with a practical, non-invasive method for individualized risk assessment, facilitating the identification of children with moderate-to-severe OSA who may benefit from prioritized diagnosis and intervention.