Validation of a Point-of-Care Prediction Model for Childhood Obstructive Sleep Apnoea - A Preliminary Analysis
Refereed conference paper presented and published in conference proceedings


Times Cited
Web of Science0WOS source URL (as at 10/04/2021) Click here for the latest count
Altmetrics Information
.

Other information
AbstractIntroduction: Childhood obstructive sleep apnoea (OSA) affects around 5% of school-aged children in Hong Kong. Early diagnosis is a prerequisite for timely treatment to reduce future morbidities. Current diagnostic gold standard, polysomnography (PSG), is limited by its cost and availability. Therefore, diagnostic alternatives that do not require hospitalisation and/or continuous monitoring are being sought. A point-of-care prediction model, combining OSA symptoms, anthropometric measurements, upper airway size and craniofacial parameters, was built based on a dataset of 90 prepubertal children aged 6-12 years suspected to have OSA using logistic regression. In our previous pilot study, this prediction model had an accuracy of 89% with 79% sensitivity and 93% specificity in predicting OSA, as defined by an obstructive apnoea hypopnoea index (OAHI) ≥1/h; and an accuracy of 92% with 95% sensitivity and 80% specificity in predicting moderate-to-severe OSA (OAHI ≥5/h).
Objective: This study aims to prospectively evaluate the validity of this prediction model.
Materials and methods: Prepubertal Chinese children aged 6-12 years old, suspected to have OSA were recruited. In addition to overnight PSG, anthropometric measurements, tonsil size evaluation, OSA-symptom questionnaire, x-ray cephalometry, and sonographic measurement of lateral parapharyngeal wall (LPW) thickness were done. In this preliminary analysis, Pearson’s correlation and Student’s t-test were done. A sample size of 225 was needed to validate the sensitivities and specificities with precisions within 8.0%. The performance of the prediction model was assessed by its discrimination ability, evaluated by both 2 x 2 contingency table analysis and the receiver operating characteristic (ROC) curve analysis.
Results:  This was a preliminary analysis of a prospective cohort.  At the time of writing, 97 subjects with a mean age of 8.19 (±1.64) years were recruited with complete data collection. From cephalometry, the ratio of the hyoid bone position to the length of mandible (MP-H/Go-Gn) was significantly correlated with OSA severity namely OAHI (r=0.302, p=0.003). From the sonographic measurement, the LPW thickness was significantly correlated with OSA severity in terms of OAHI (r=0.302, p=0.003). From the contingency table analysis, the diagnostic accuracy of the prediction model with current data was 64.9% and 65.6% for OAHI ≥1/h and OAHI ≥5/h respectively. The area under the ROC curve was 65.7% and 71.8% for OAHI ≥1/h and OAHI ≥5/h respectively.
Conclusion: In this preliminary analysis, cephalometric and sonographic measurements correlate with PSG results, which are consistent with our previous findings. The accuracy of the prediction model is comparatively lower than that of our pilot study which may due to subjects’ heterogeneity. Younger age and lower BMI Z-scores were observed among subjects from the current dataset compared to that from the pilot study.
Acknowledgements: This study was supported by the General Research Fund provided by the Research Grants Council of the Hong Kong Special Administrative Region, China [14123618].
All Author(s) ListH.M. Yuen, C.T. Au, K.C.C. Chan, A.M. Li
Name of ConferenceWorld Sleep 2019
Start Date of Conference20/09/2019
End Date of Conference25/09/2019
Place of ConferenceVancouver
Country/Region of ConferenceCanada
Proceedings TitleSleep Medicine
Year2019
Month12
Volume Number64
Issue NumberSuppl. 1
PagesS435 - S436
LanguagesEnglish-United Kingdom

Last updated on 2021-10-04 at 23:52