(1478) Comparing Statistical Models for PaO2 Prediction From SpO2: Insights from Hemoglobin Parameters

Sunday, February 23, 2025

2:30pm – 3:30pm Eastern Time

Location: Connections Central - RST 10

First Author(s)

Samuel Huang, MD

Icahn School of Medicine at Mount Sinai South Nassau, United States

Co-Author(s)

AH

Alexander Huang, MD

Physician
Northwestern University Feinberg School of Medicine, United States

Introduction: Accurate prediction of arterial oxygen partial pressure (PaO2) from peripheral oxygen saturation (SpO2) is crucial in clinical settings. This study evaluates and compares the performance of Hill's equation, Severinghaus-Ellis equation, Gadrey’s model, and an imputed equation in estimating PaO2 from SpO2. We also investigate how these models are influenced by hemoglobin cooperativity and partial pressures in acidotic versus non-acidotic conditions.

Methods: We analyzed ABG data from two cohorts: the Prospective Inpatient Acutely Ill (PIA) cohort, including ICU and ER patients, and the Retrospective Outpatients Pulmonary (ROP) cohort, with patients undergoing pulmonary function tests. We fitted Hill's equation, Severinghaus-Ellis equation, Gadrey’s model, and an imputed equation to the data. The mean absolute error (MAE) and 95% confidence intervals (CIs) were calculated for each model. The analysis included comparisons between acidotic (pH < 7.35) and non-acidotic (pH ≥ 7.35) conditions, with t-tests used to assess differences.

Results: MAE values were as follows: Hill's equation 0.0202 (CI: 0.0166-0.0237), Severinghaus-Ellis 0.0194 (CI: 0.0159-0.0230), Gadrey’s model 0.0188 (CI: 0.0152-0.0225), and the imputed equation 0.0188 (CI: 0.0152-0.0225). For acidotic conditions, MAE values were: Hill 0.0233 (CI: 0.0131-0.0335), Severinghaus-Ellis 0.0234 (CI: 0.0131-0.0337), Gadrey 0.0264 (CI: 0.0159-0.0370), and imputed 0.0264 (CI: 0.0159-0.0370). For non-acidotic conditions, MAE values were: Hill 0.0188 (CI: 0.0165-0.0210), Severinghaus-Ellis 0.0176 (CI: 0.0153-0.0199), Gadrey 0.0152 (CI: 0.0128-0.0176), and imputed 0.0152 (CI: 0.0128-0.0176). Significant differences were observed between acidotic and non-acidotic groups (p < 0.05).

Conclusions: Gadrey’s model and the imputed equation provided slightly better accuracy compared to Hill's and Severinghaus-Ellis equations. Acidotic conditions led to higher MAEs, indicating increased variability in PaO2 predictions. This highlights the need for considering hemoglobin parameters in predicting PaO2 from SpO2. Future research should aim to refine these models for improved clinical accuracy and better understanding of physiological variations.