Better than 'normal': Describing the distribution of and outliers in RBC usage

Background/Case Studies Optimizing a hospital blood inventory requires understanding the distribution of blood usage at the institution. Standard methods for describing red blood cell (RBC) usage like mean and standard deviation assume a normal distribution. Other distributions may fit the distribution of daily RBC usage better and provide more accurate insights into blood usage and in6ventory management. Study Design/Methods RBC usage data from 2014 through 2020 were queried from the laboratory information system. Theoretical distributions were fit against this empirical distribution of daily RBC usage. Goodness-of-fit was assessed visually using Pearson/Cullen and Frey plots as well as quantitatively using Kolomogorov-Smirnov (K-S) distance and the Akaike Information Criterion (AIC). Results/Findings 241,957 units of RBCs were transfused over 2557 days in the study period. The log-normal and gamma distributions had superior K-S distances and AICs for daily RBC usage. The top percentile of days for RBC usage was significantly more likely to have at least one ultramassively transfused patient (p < .001, Fisher's Exact Test). Conclusions The log-normal and gamma distributions better describe the right-skewed and entirely positive empirical distribution of daily RBC usage as compared to the normal distribution. This should broadly inform inventory management efforts and future descriptions of RBC usage at the institutional level.