Acute hemorrhage in dogs: construction and validation of models to quantify blood loss

We examined the ability of commonly used clinical parameters to quantify acute hemorrhage in dogs. Eight animals were bled 40 ml/kg body wt over 100 min. Ten hemodynamic and 20 blood laboratory parameters were obtained every 10 min to construct, with use of linear regression analysis, models that quantify blood loss. During model construction, the best indicator of quantity of hemorrhage was arterial base deficit [ABD; coefficient of variation (CV) 35%]. This model was more accurate (P < 0.05) than 27 others (CV range 43 to 63%) and similar to systolic (CV 40%) and mean (CV 40%) arterial pressures. In validation studies in 10 additional animals, our best models based on ABD and systolic and mean arterial pressures each unexpectedly showed a significant (P < 0.05) decrease in accuracy (CV 86, 57, and 60%, respectively) attributable to large baseline (before hemorrhage) variability among animals. To eliminate this variability, models based on changes from baseline measurements were investigated. The best predictor of change in blood volume was change in ABD (CV 27%). This model was significantly (P < 0.05) more accurate than any of 27 others (CV range 36 to 65%) and similar to change in venous base deficit and venous pH (each CV 31%). When validated, acid-base models such as ABD, venous pH, and arterial bicarbonate were the best predictors of volume change (CV range 28 to 40%). With the use of multivariate analysis, pairwise combinations of single parameter models (n = 465) improved prediction errors only minimally. In summary, most commonly used hemodynamic and blood indexes could not be validated as accurate measurements in quantifying hemorrhage. In contrast, changes in acid-base parameters were validated as moderately accurate predictors of blood volume changes and therefore may have utility in the assessment of patients with ongoing hemorrhage.