Refining screening questionnaires for prediction of sleep apnea severity in children.
Journal: 2019/November - Sleep and Breathing
ISSN: 1522-1709
Abstract:
Screening instruments are poor predictors of the severity of pediatric obstructive sleep apnea (OSA). We hypothesized that their performance could be improved by identifying and eliminating redundant features.Baseline scores from three screening questionnaires for pediatric OSA were obtained from the Childhood Adenotonsillectomy Trial (CHAT). The questionnaires included the (i) modified Epworth sleepiness scale (ESS), (ii) the sleep-related breathing disorders subscale of the pediatric sleep questionnaire (PSQ), and the (iii) obstructive sleep apnea-18 (OSA-18) scale. Key features from each questionnaire were identified using variable selection methods. These selected features (SF) were then assessed for their ability to predict the severity of OSA, measured by the apnea-hypopnea index (AHI) and oxygen desaturation index (ODI). In addition, prediction performance of SF was also calculated for AHI > 5 and > 10 and ODI > 5 and > 10, respectively.Four hundred fifty-three children aged 5-10 years were included. The majority of the pairwise correlations among the items within the 3 screening questionnaires were statistically significant. The prediction of AHI and ODI by overall questionnaire scores was poor. Four-item SF, comprising apneic pauses, growth problems, mouth breathing, and obesity predicted AHI and ODI significantly better than each of the individual questionnaires. Furthermore, SF also predicted AHI > 5 and > 10, as well as ODI > 5 and > 10 significantly better than the original questionnaires.Elimination of redundant items in screening questionnaires improves their prediction performance for OSA severity in children with high pre-test probability for the condition.
Relations:
Diseases
(5)
Conditions
(4)
Chemicals
(1)
Processes
(1)
Affiliates
(2)
Similar articles
Articles by the same authors
Discussion board
Collaboration tool especially designed for Life Science professionals.Drag-and-drop any entity to your messages.