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

Area of Science:

• Transfusion Medicine
• Health Informatics
• Operations Research

Background:

• Optimizing hospital blood inventory necessitates understanding institutional red blood cell (RBC) usage patterns.
• Traditional methods often assume a normal distribution for RBC usage, which may not accurately reflect real-world data.
• Alternative statistical distributions could offer more precise insights for blood inventory management.

Purpose of the Study:

• To evaluate statistical distributions that best describe daily red blood cell (RBC) usage in a hospital setting.
• To compare the fit of theoretical distributions against empirical RBC usage data.
• To inform more accurate blood inventory management strategies.

Main Methods:

• Queried RBC transfusion data from 2014-2020 from the laboratory information system.
• Fit theoretical distributions (e.g., log-normal, gamma, normal) to the empirical daily RBC usage data.
• Assessed goodness-of-fit using visual plots (Pearson/Cullen and Frey) and quantitative metrics (Kolmogorov-Smirnov distance, Akaike Information Criterion).

Main Results:

• Over 241,957 RBC units were transfused across 2557 days.
• Log-normal and gamma distributions demonstrated superior fit (lower K-S distance and AIC) compared to the normal distribution for daily RBC usage.
• High RBC usage days were significantly more likely to involve ultramassive transfusions (p < .001).

Conclusions:

• Log-normal and gamma distributions provide a more accurate representation of the right-skewed, positive-valued empirical distribution of daily RBC usage than the normal distribution.
• These findings should guide institutional blood inventory management and reporting of RBC usage.
• Improved statistical modeling can lead to more efficient blood supply chain operations.