TITLE : A Data-Driven Approach to Patient Risk Stratification for Acute Respiratory Distress Syndrome ( ARDS )

INTRODUCTION Acute Respiratory Distress Syndrome (ARDS) is a condition in which fluid and inflammatory cells permeate the lungs preventing effective gas transport. ARDS is linked to complications such as nosocomial infections and mortality. Over 70% of patients with ARDS are diagnosed late or not at all. In the US, ARDS is estimated to affect 34 per 100,000 patients per year, and studies show it has a 28-day mortality rate of 20-40%. These facts show the importance of making predictions as early and accurately as possible. 1 Previous studies, such as the LIPS study, have created models that identify patients who are at high risk for developing ARDS based on information available during the first six hours of a patient’s hospital stay. We test the generalizability of LIPS in our study population and improve upon this model by leveraging the structured contents of the electronic health record (EHR) to identify patients who will develop ARDS during their hospital stay as early and accurately as possible.