Clinical risk assessment model for pediatric obstructive sleep apnea.

Abstract

To develop a clinical risk prediction model that identifies children with obstructive sleep apnea (OSA) in a clinical setting by examining the symptoms, physical status, and OSA-18 questionnaire results. Single institutional, cross-sectional study. Children aged 2 to 18 years with symptoms of OSA were enrolled. Pediatric OSA was diagnosed through full-night polysomnography. Clinical data, namely demographics, symptoms, OSA-18 survey results, tonsil and adenoid sizes, and the weight of each child, were examined for constructing a simple point-based clinical model for OSA prediction. Variables for the risk model were selected using multivariable logistic regression analyses. Of the 310 participants (mean age, 7.6 ± 3.7 years; boys, 67%), 170 (55%) experienced OSA. Modeling variables were determined using several univariate logistic regression analyses, followed by multivariable logistic regression analyses. A point-based clinical model incorporating the age, tonsil size (5 points maximum), adenoid size (5 and 20 points for age > 6 years and < 6 years, respectively), obesity (5 points for age > 6 years), and breathing pauses (5 points) was developed (area under the curve = 0.832). Moreover, the optimal cutoff points for predicting the apnea-hypopnea index of > 1 and > 5 were 10 (sensitivity, 72.9%; specificity, 65.0%) and 12 (sensitivity, 77.5%; specificity, 56.9%), respectively. Internal validation using the bootstrap method revealed no apparent overfitting problem. A novel clinical prediction model was developed for determining the risk of pediatric OSA; the model can be useful in identifying high-risk patients among those with sleep disturbances. 4. Laryngoscope, 126:2403-2409, 2016.