To the Editor:—
The format proposed by Rushmer (206:836, 1968) for display of clinical laboratory data is an excellent one for those variables which are known to be approximately normally distributed. However, some variables may have a distribution which is far from normal. As Mainland1,2 and others have pointed out, using the properties of the normal distribution when they do not obtain can result in misleading conclusions. Therefore, for any data which do not have a normal distribution, we suggest that a scale centered at the population median and graduated in percentile units would be more appropriate than a scale centered at the population mean and graduated in standard deviations.
Markowitz E, Leaverton PE. Statistical Properties of Normal Distribution. JAMA. 1969;207(1):153. doi:10.1001/jama.1969.03150140105034