(1463) Evaluating Hypoxemia Burden as a Mortality Predictor in Step-Down Units

Introduction: Routine pulse oximetry (SpO2) monitoring focuses on SpO2 drops below a clinician-determined threshold, often overlooking sustained oxygen deprivation impacts. This study introduces and evaluates the hypoxemia burden (HXB) – a measure of cumulative SpO2 desaturations – as a mortality predictor in step-down units (SDU) to enhance clinical decisions and patient management.

Methods: We assessed 2.9 million SpO2 alerts from 6,824 patients monitored in one surgical-trauma SDU, employing a classifier to filter artifacts, which achieved an AUC of 0.97. HXB was calculated by integrating the area under the curve where SpO2 readings fell below a fixed threshold (e.g., 90%), multiplied by the duration in minutes that SpO2 remained below this threshold, thereby quantifying the cumulative impact of each desaturation episode. HXB mitigates noise and allows for an integrated index of both the duration and severity of hypoxemia.

Results: Among all the 6,824 patients, we identified 3,684 patients (mean age 63±17, 43% female, 18% in-hospital mortality) with SpO2 alerts and established a logistic relationship between in-hospital mortality and log-transformed HXB. The logistic model fit a coefficient of 1.8±0.5, a constant of -4.7±1.5, and had an ROC AUC of 0.91. Selecting a HXB threshold of 151 (in unit of % × minutes), the model achieved a true positive rate (TPR) of 0.91 and a false positive rate (FPR) of 0.26. When applied to the same patient cohort (3,684) using the same threshold, the performance decreased, evidenced by a reduced AUC of 0.81, sensitivity of 0.63, and positive prediction value of 0.54 (reduction due primarily to lower hypoxemia burden in greater proportion of the total sample). Despite these variations, the odds ratio (OR) of 12.1 (95% CI 10.3 to 15.3) indicates a strong association between higher hypoxemia burden and an increased risk of mortality.

Conclusions: The HXB provides an advancement in predicting mortality in SDUs, contributing substantially to current methodologies. By quantifying critical thresholds of oxygen deprivation and duration, it offers a valuable metric for assessing risk. This study demonstrates the metric's potential effectiveness and lays a foundation for integration into clinical practice, and potentially improved outcomes through refined monitoring.