Predicting Sudden Decrease in Oxygenation in Mechanically Ventilated ICU Patients as a Surrogate Marker for ARDS

papers.ssrn.com February 5, 2025

Acute Respiratory Distress Syndrome (ARDS) is a life-threatening form of respiratory failure characterized by widespread lung inflammation that severely impairs oxygenation.

Affecting millions of patients worldwide, ARDS is associated with high morbidity and mortality, particularly in Intensive Care Unit settings. Early detection of ARDS is critical, as delays in diagnosis and treatment may be linked to an increased risk of mortality.

Despite this, ARDS is frequently underdiagnosed due to its nonspecific symptoms and overlap with other conditions.

This study presents a novel machine learning-based Clinical Decision Support System designed to predict rapid oxygenation loss, serving as a surrogate marker for ARDS in mechanically ventilated ICU patients.

Unlike traditional ARDS predictors that rely on static thresholds from the Berlin criteria, our model focuses on dynamic changes in the Horowitz index (PaO2/FiO2 ratio), allowing for earlier and more precise detection of acute deterioration in lung function.