ABSTRACT 571

Abstract

Background and aims: Clinical, bedside estimates of cardiac output are typically inaccurate. Aims: We wished to evaluate whether a structured combination of routine, haemodynamic data could diagnose a low stroke volume (SV). Methods: IRB approval and informed consent was obtained. SV was measured using transpulmonary ultrasound dilution before and after clinical interventions on 398 occasions for 100 critically ill children, median (IQR) weight 10 kg (5.6 to 15.2). A low SV was defined as <20 ml/m2. Analysis was via mixed effects logistic regression, with discrimination quantified via the area under the ROC curve (AUC). Results: Overall 34% of measurements demonstrated a low SV. Discrimination was excellent using a model combining age, heart rate (HR), pulse pressure (PP), lactate and CVO2 (AUC = 0.895). Discrimination was still very good when only the continuous haemodynamic variables HR and PP were used with age (AUC = 0.874). Discrimination worsened when systolic blood pressure was substituted for PP (AUC = 0.857). Mixed-effects adjustment to account for repeated measurements on patients produced minor AUC change across models; however, of note, lactate and CVO2 now became non significant predictors. The optimal cut-point for the simplified model with age, HR and PP yielded: sensitivity 83.0%, specificity 79.6%, with a correct prediction rate of 80.7%.FigureConclusions: The combination of HR, PP and age provide a reasonable diagnostic tool for a low flow state.