Comparative Performance of Oxygenation Indices for Mortality Prediction in Patients With Acute Respiratory Distress Syndrome at a Tertiary Care Intensive Care Unit in Bangladesh

BackgroundAcute respiratory distress syndrome (ARDS) is a syndrome of acute inflammatory lung injury and impaired gas exchange, not hypoxemia alone, and carries high mortality.This study compared the discrimination of the partial pressure of arterial oxygen (PaO₂)/fraction of inspired oxygen (FiO₂) (PF) ratio, oxygenation index (OI), and oxygen saturation index (OSI), measured on the day of ARDS diagnosis, for intensive care unit (ICU) mortality in a Bangladeshi tertiary ICU. MethodsThis single-center prospective observational study enrolled 80 consecutive adults meeting the 2012 Berlin definition of ARDS and requiring invasive mechanical ventilation at Bangladesh Medical University between June 2024 and May 2025.Time zero (day 1) was the day the Berlin criteria were first met.Indices were calculated on days 1, 3, and 7 from simultaneous blood gas, oximetry, and ventilator data.Groups were compared by independent t-and chi-square tests, serial change by paired t-tests in survivors only, and discrimination by receiver operating characteristic (ROC) curve analysis with Youden cutoffs.Correlated areas under the curve (AUCs) were not formally compared, and no multivariable model was fitted. ResultsForty-five of 80 patients (56.3%) died in the ICU, and 35 (43.8%) survived.Non-survivors had lower blood pressure, higher heart and respiratory rates, and greater ventilatory requirements (p<0.05).OI and OSI were higher and the PF ratio was lower among non-survivors (p<0.05).Discrimination was numerically highest for OSI (AUC=0.879;95% CI: 0.786-0.977),followed by OI (AUC=0.867;95% CI: 0.788-0.963)and the PF ratio (AUC=0.731;95% CI: 0.612-0.881);as the intervals overlapped and no paired test was applied, this ordering is descriptive only.Specificity was 91.4% for both OSI and OI, and accuracy was 86.3% and 83.8% versus 70% for the PF ratio.Serial improvement among survivors was not statistically significant. ConclusionsOI and OSI were higher among non-survivors and showed numerically higher discrimination for ICU mortality than the PF ratio, with OSI performing best.As the analysis was univariable and cutoffs were internally derived, these findings indicate discrimination and association rather than independent prognostic value, and the thresholds are exploratory.Larger multicenter studies with formal paired comparison of ROC curves, multivariable adjustment, and independent validation are warranted.