The search for optimal oxygen saturation targets in critically ill: Patients observational data from large ICU databases

• Main Authors: VAN DEN BOOM, Willem, HOY, Michael, SANKARAN, Jagadish, LIU, Mengru, CHAHED, Haroun, FENG, Mengling, SEE, Kay Choong
• Format: text
• Language: English
• Published: Institutional Knowledge at Singapore Management University 2020
• Subjects: blood oxygen saturation; electronic medical records; hyperoxemia; ICU; oxygen therapy; pulse oximetry; Databases and Information Systems; Medicine and Health Sciences; Numerical Analysis and Scientific Computing
• Online Access: https://ink.library.smu.edu.sg/sis_research/5318; https://ink.library.smu.edu.sg/context/sis_research/article/6321/viewcontent/VanDenBoom_Chest_av_2019.pdf
• Institution: Singapore Management University

Background: Although low oxygen saturations are generally regarded as deleterious, recent studies in ICU patients have shown that a liberal oxygen strategy increases mortality. However, the optimal oxygen saturation target remains unclear. The goal of this study was to determine the optimal range by using real-world data. Methods: Replicate retrospective analyses were conducted of two electronic medical record databases: the eICU Collaborative Research Database (eICU-CRD) and the Medical Information Mart for Intensive Care III database (MIMIC). Only patients with at least 48 h of oxygen therapy were included. Nonlinear regression was used to analyze the association between median pulse oximetry-derived oxygen saturation (SpO2) and hospital mortality. We derived an optimal range of SpO2 and analyzed the association between the percentage of time within the optimal range of SpO2 and hospital mortality. All models adjusted for age, BMI, sex, and Sequential Organ Failure Assessment score. Subgroup analyses included ICU types, main diagnosis, and comorbidities. Results: The analysis identified 26,723 patients from eICU-CRD and 8,564 patients from MIMIC. The optimal range of SpO2 was 94% to 98% in both databases. The percentage of time patients were within the optimal range of SpO2 was associated with decreased hospital mortality (OR of 80% vs 40% of the measurements within the optimal range, 0.42 [95% CI, 0.40-0.43] for eICU-CRD and 0.53 [95% CI, 0.50-0.55] for MIMIC). This association was consistent across subgroup analyses. Conclusions: The optimal range of SpO2 was 94% to 98% and should inform future trials of oxygen therapy.