Effects of Using Different Indirect Techniques on the Calculation of Reference Intervals: Observational Study

Area of Science:

• Clinical Laboratory Science
• Biostatistics
• Medical Diagnostics

Background:

• Reference intervals (RIs) are crucial for clinical decision-making.
• Establishing RIs directly is costly and time-consuming.
• Indirect methods using existing laboratory data are gaining traction.

Purpose of the Study:

• To evaluate the accuracy of various indirect methods for establishing complete blood count (CBC) reference intervals (RIs).
• To assess the impact of data transformation and outlier removal techniques on RI accuracy.
• To compare indirect method results against direct method standards.

Main Methods:

• Utilized large-scale CBC data from adult physical examinations (2010-2011).
• Applied multiple indirect methods (e.g., parametric, nonparametric, Hoffmann, Bhattacharya, kosmic) with log/BoxCox transformations.
• Employed outlier removal techniques (Reed-Dixon, Tukey, iterative mean (3SD)) and calculated bias ratios (BRs).

Main Results:

• Data transformation and outlier removal significantly influenced RIs, especially for skewed data.
• Tukey and iterative mean (3SD) methods demonstrated comparable outlier removal efficiency.
• Indirect method performance varied by analyte and sex; Bhattacharya, Hoffmann, and parametric methods were superior for certain CBC parameters in males.

Conclusions:

• Outlier removal and indirect techniques markedly affect final RIs; data transformation has minimal impact unless data is highly skewed.
• The optimal indirect technique is analyte-dependent.
• This study offers evidence supporting clinical laboratories in establishing RIs using historical data.