Meta-analysis of diagnostic test accuracy of pediatric sleep questionnaire and the OSA-18 questionnaire in the screening of pediatric obstructive sleep apnea.

OBJECTIVE

To meta-analyze diagnostic test accuracy (DTA) data in predicting obstructive sleep apnea (OSA) in children for the Pediatric Sleep Questionnaire (PSQ) and OSA-18.

STUDY

DESIGN

Studies were included if they reported data analyzable for DTA or correlation between apnea-hypopnea index (AHI) and PSQ or OSA-18. Random effects DTA estimates and their 95% confidence intervals were computed with AHI cut-offs ≥1 and ≥5. Spearman's correlation coefficients (ρ) were pooled using a random effects model.

RESULTS

Sensitivity and specificity of PSQ and OSA-18 with AHI cut-off ≥1 were 0.70 (0.63, 0.76), 0.43 (0.33, 0.54), from 14 studies, 0.63 (0.51, 0.73), 0.74 (0.54, 0.87), from 5 studies, respectively, and with AHI cut-off ≥5 were 0.75 (0.71, 0.79), 0.41 (0.29, 0.55) from 11 studies and 0.73 (0.45, 0.90), 0.34 (0.18, 0.54), based on 3 studies, respectively. After excluding children with cranio-facial anomalies and select chronic medical conditions, mean sensitivity and specificity of PSQ and OSA-18 with AHI cut-off ≥1 were 0.74, 0.42 and 0.65, 0.73, respectively, and with AHI cut-off ≥5 were 0.80, 0.44, and 0.80, 0.27, respectively. Pooled ρ between AHIs and PSQ and OSA-18 were 0.28 (0.15; 0.40), and 0.34 (-0.30; 0.77), respectively.

CONCLUSIONS

As screening questionnaires, PSQ and OSA-18 are neither sufficiently sensitive to detect OSA nor sufficiently specific to determine the absence of OSA and cannot substitute a sleep study. The questionnaires are particularly poor at predicting presence or absence of OSA in children with cranio-facial anomalies and select chronic medical conditions.