Revising risk stratification in mechanically ventilated ICU patients: the longitudinal European BioVent study

Abstract

Background: The assessment of disease severity is an essential part of medical evaluation across all disciplines, allowing patient-tailored treatment regimens. Risk assessment is particularly important in intensive care, where improved therapy may translate to a gain in many life-years; however, intensive care can be a limited resource, has tremendous costs and may prolong suffering. Our goal was to assess and combine various tools to advance risk assessment in mechanically ventilated ICU patients.

Methods: BioVent (Biomarkers for mechanically ventilated patients) is a prospective investigator-initiated longitudinal cohort study including five centers across three European countries (Austria, France, Switzerland; ISRCTN59376582). In total 961 patients mechanically ventilated for at least 12 hours were included. Over 100 predictors were obtained at inclusion and throughout ICU stay, and assessed for predicting 28-day and 1-year survival. New prediction models were proposed and thoroughly validated.

Results: We revealed markers of renal disease, novel biomarkers as well as subjective survival estimates (from physicians and nurses) among the best single variables predicting 28-day survival. We then pooled all variables into three groups, namely patient characteristics, biomarkers and subjective survival estimates. Regression models revealed AUCs of 0.67, 0.77 and 0.80, respectively. Combining all markers increased model performance to an AUC of 0.85, outperforming SAPS2 and SOFA.

Conclusions: We conclude that combinations of novel biomarker and especially subjective survival estimates outperform traditional risk assessment tools in mechanically ventilated ICU patients.