Weighted Hypoxemia Index: An adaptable method for quantifying hypoxemia severity

Abstract

Objective To quantitate hypoxemia severity. Methods We developed the Weighted Hypoxemia Index to be adapted to different clinical settings by applying 5 steps to the oxygen saturation curve: (1) Identify desaturation/resaturation event i by setting the upper threshold; (2) Exclude events as artifact by setting a lower threshold; (3) Calculate weighted area for each i, as (Δi × Φi); (4) Calculate a normalization factor Ω for each subject; (5) Calculate the Weighted Hypoxemia Index as the summation of all weighted areas multiplied by Ω. We assessed the Weighted Hypoxemia Index predictive value for all-cause mortality and cardiovascular mortality using the Sleep Heart Health Study (enrollment 1995–1998, 11.1 years mean follow-up). Results We set varying upper thresholds at 92%, 90%, 88%, and 86%, a lower threshold of 50%, calculated area under the curve and area above the curve, with and without a linear weighted factor (duration of each event i), and used the same normalization factor of total sleep time <90% divided by total sleep time. After excluding subjects with missing data, we analyzed 4,509 participants (Alive: N = 3,769; All-cause mortality: N = 1,071; cardiovascular mortality: N = 330). Since the Weighted Hypoxemia Index-Area Under the Curve set at upper threshold of 90% (WHI-AUC90) had the best results in predicting all-cause mortality, we then compared it to the Apnea-Hypopnea Index and Total Sleep Time <90%. WHI-AUC90 showed statistical significance across quintiles for all-cause mortality, but not cardiovascular mortality, in adjusted Cox regression models. Conclusion The Weighted Hypoxemia Index offers a versatile and clinically relevant method for quantifying hypoxemia severity, with potential applications to evaluate mechanisms and outcomes across various patient populations.