The hypoxic burden algorithm: an integrated index for assessing intermittent hypoxia in pediatric obstructive sleep apnea-a retrospective cohort study.

BACKGROUND: Pediatric obstructive sleep apnea (OSA) involves recurrent upper-airway obstruction during sleep, which leads to intermittent hypoxia (IH) and contributes to neurocognitive, cardiovascular, and metabolic complications. Polysomnography (PSG) and the Obstructive Apnea-Hypopnea Index (OAHI) are commonly used for diagnosis; however, conventional nocturnal hypoxic metrics such as lowest oxygen saturation (LSpO2), mean oxygen saturation (MSpO2), total time with oxygen saturation below 90% (T90), and oxygen desaturation index ≥3% (ODI3) fail to capture the cumulative burden. Hypoxic burden (HB) has therefore gained attention as an integrated measure that reflects the frequency, depth, and duration of desaturation events; however, its use and clinical value in pediatric OSA remain uncertain. This study aims to evaluate the clinical value of HB in diagnosing pediatric OSA and compare its diagnostic performance with conventional nocturnal hypoxic metrics. METHODS: We conducted a retrospective analysis of PSG data from 115 children with suspected OSA at Tianjin Children's Hospital between December 2020 and November 2025. Data were analyzed using internally developed Matlab software. The differential performance of five nocturnal IH parameters (LSpO2, MSpO2, T90, ODI3, and HB) was compared across four groups: normal, mild, moderate, and severe OSA. All data were manually scored according to the AASM Scoring Manual (version 2.6). RESULTS: The cohort (72 males, 43 females; mean age 6.44±2.86 years) was distributed into normal (n=15), mild (n=63), moderate (n=23), and severe OSA (n=14) groups based on OAHI. Significant differences were observed among severity groups for ODI3, T90, and HB. Correlation analysis showed LSpO2 and MSpO2 had weak-to-moderate negative correlations with OAHI (r=-0.490 and r=-0.543, respectively). ODI3 (r=0.788) and T90 (r=0.700) showed moderate positive correlations. Importantly, HB demonstrated a strong correlation with OAHI (r=0.801). ROC curve analysis indicated weak diagnostic capability for LSpO2 (AUC =0.689) and MSpO2 (AUC =0.614). ODI3 (AUC =0.857) and T90 (AUC =0.701) showed moderate value, while HB (AUC =0.888) demonstrated the best diagnostic performance. CONCLUSIONS: We successfully developed an HB algorithm, demonstrating its significant advantage in pediatric OSA diagnosis. Compared to traditional indices that reflect only a single dimension of oxygen desaturation, HB integrates the frequency, depth, and duration of hypoxia. This provides a robust, objective alternative for the early screening and clinical assessment of children with OSA.