Abstract
To develop a nomogram model to predict obstructive sleep apnea in children and perform an external validation. 864 children who underwent polysomnography at our hospital were randomly assigned to a training cohort and an internal validation cohort (7:3) and 292 children at another hospital were enrolled for external validation. Logistic regression analyses were performed to explore the risk factors for obstructive sleep apnea. Features selected by least absolute shrinkage and selection operator logistic regression were enrolled to develop a predictive nomogram, which was assessed by discrimination, calibration, and clinical benefit in three cohorts and six subgroups from the all development cohort: age, gender, nonobese children, and term children. Furthermore, the nomogram was compared with the Obstructive Sleep Apnea-18 questionnaire. Six features were selected to construct the nomogram: adenoid hypertrophy, tonsil size 2, tonsil size 3, birth weight, total sleep time, and the lowest oxygen saturation. The areas under the curve of the nomogram in the training cohort [0.784, 95%CI (0.746-0.821)], internal validation cohort [0.780 (0.721-0.840)], and external validation cohort [0.782 (0.726-0.838)] were higher than those of the OSA-18 and remained stable in subgroups. Calibration curves and decision curve analysis both exhibited favorable performance. CONCLUSION: The nomogram with satisfactory performance was constructed as a robust tool for evaluating the risk of obstructive sleep apnea in children. WHAT IS KNOWN: • Obstructive sleep apnea (OSA) is common in children and is closely related to multiple system dysfunction, while still remains underdiagnosed and undertreated because of the limited usage of polysomnography. • A prediction model for accurately identifying OSA in children is warranted in a subsequent recommendation of essential PSG to provide proper treatment and delay the occurrence of OSA complications. WHAT IS NEW: • Adenoid hypertrophy, tonsil size 2, tonsil size 3, birth weight, total sleep time, and the lowest oxygen saturation were risk factors of children with OSA. • This nomogram with satisfactory performance was a robust tool for evaluating the risk of OSA in children.