Purpose: Appropriately identifying and triaging patients with newly diagnosed acute respiratory distress syndrome (ARDS) who may progress to severe ARDS is a common clinical challenge without any existing tools for assistance.
Materials and methods: Using a retrospective cohort, a simple prediction score was developed to improve early identification of ARDS patients who were likely to progress to severe ARDS within 7 days. A broad array of comorbidities and physiologic variables were collected for the 12-hour period starting from intubation for ARDS. Extracorporeal membrane oxygenation (ECMO) eligibility was determined based on published criteria from recent ECMO guidelines and clinical trials. Separate data-driven and expert opinion approaches to prediction score creation were completed.
Results: The study included 767 patients with moderate or severe ARDS who were admitted to the intensive care unit between January 1, 2005, and December 31, 2010. In the data-driven approach, incorporating the ARDS index (a novel variable incorporating oxygenation index and estimated dead space), aspiration, and change of Pao2/fraction of inspired oxygen ratio into a simple prediction model yielded a c-statistic (area under the receiver operating characteristic curve) of 0.71 in the validation cohort. The expert opinion-based prediction score (including oxygenation index, change of Pao2/fraction of inspired oxygen ratio, obesity, aspiration, and immunocompromised state) yielded a c-statistic of 0.61 in the validation cohort.
Conclusions: The data-driven early prediction ECMO eligibility for severe ARDS score uses commonly measured variables of ARDS patients within 12 hours of intubation and could be used to identify those patients who may merit early transfer to an ECMO-capable medical center.
Keywords: Acute respiratory distress syndrome; Adult; Clinical prediction; Extracorporeal membrane oxygenation; Patient selection.
Copyright © 2016 Elsevier Inc. All rights reserved.