Clinical evaluation of circulating blood volume in critically ill patients--contribution of a clinical scoring system

Abstract

The circulating blood volume (CBV) of critically ill patients may be difficult to estimate on the basis of history and physical examination. The aim of this study was to evaluate the ability of seven clinical signs and central venous pressure (CVP) to predict CBV in critically ill patients; CBV was evaluated with the [125I]human serum albumin technique. A scoring system was constructed using a combination of independence Bayes method and logistic regression. Sixty-eight patients constituted a 'model development' sample and 30 patients a validation sample. Thirty-six patients (53%) in the model development sample were found to have a low CBV (measured CBV at least 10% lower than the predicted mean normal CBV). Neither the haemodynamic variables monitored in ICU, nor the spot urinary sodium concentrations were different between patients with and without a low CBV. Individually, none of the clinical signs tested have a good positive or negative predictive value. For CVP, only extreme values seem to have clinical significance. To construct the score, the signs tested were ranked according to their discriminating efficacy. The probability of a low CBV was obtained by adding the weights of each sign tested and converting the score obtained into a probability. On a validation sample of 30 patients, the predictions are reliable as assessed by Z statistics ranging between -2 and +2. Our results suggest that: (1) individually, no clinical sign presented a clinical useful predictive value; and (2) a clinical scoring system may be helpful for the evaluation of CBV in critically ill patients.