Predicting the post-dilution hematocrit is an important tool to avoid preventable anemia or unnecessary transfusion. Simplified empirical formulas currently used for such a prediction may lead to large errors. We aimed to improve the accuracy of these formulas by a better estimation of the dilution volume and the patient circulatory blood volume.Methods:
We compared the estimation accuracy of two formulas, using fixed (formula A) versus estimated (formula D) dilution volume and patient circulatory blood volume for 100 cardiac interventions. The difference between predicted and measured HctT1 was considered as “good” if less than 0.5%, “moderate” between 0.5 and 2% and “poor” if higher than 2%. The influence of the body mass index (BMI) on patient blood volume estimation was explored by categorized groups' comparison.Results:
The mean difference between predicted and measured HctT1 differed significantly between formulas A and D. Formula A didn't differ from HctT1 (p=0.19, IC95% [-0.89-0.18]), but a significant and higher underestimation was observed in the BMI≤25 group compared to the other BMI groups (p<0.001). Formula D overestimated HctT1 (p<0.001, IC95% [1.01-1.93]) without a difference between the BMI groups. No difference was observed in their overall proportions of good (11 vs 10%), moderate (44 vs 46%) and poor predictions (47 vs 44%) (p=0.117).Conclusions:
Formulas used for post-dilution hematocrit prediction lead to major estimation errors and a risk of inadequate transfusion practices. Estimations performed by experienced clinicians could not minimize these biases in all clinical cases as significant errors remain, with potential clinical impact. No estimation formula should be used as a hard tool for transfusing patients, but rather as a guide to predicting the probability of transfusion requirement.