We reanalyzed the data from proficiency testing (PT) to assess the effect of the geometric mean in the statistical analysis of immunohematologic data.
Using the five most recent anti-blood group antibody titer participant summary results, the geometric mean (GM) ±2 × geometric standard deviation (GSD) was used as the comparative consensus criterion to mode ±2 titers.
Using the PT evaluation criterion of mode ±2 titers, the mean percentages of participants with acceptable results were 97.5% and 97.8% for anti-A and anti-D, respectively. When applying GM ±2 GSD, the mean percentages of acceptable results were 96.1% (anti-A) and 96.1% (anti-D). The percentages of responses included in each consensus criterion were lower using GM ±2 GSD, with a few exceptions.
Geometric means are more robust and precise in visualizing the central tendency. This method can improve the statistical robustness of PT evaluations.