Partitioning reference values of several Gaussian subpopulations with unequal prevalence--a procedure with computer...

Area of Science:

• Clinical Chemistry
• Biostatistics
• Medical Informatics

Background:

• Interpreting laboratory values requires accurate population-based reference intervals.
• Existing methods for partitioning reference intervals are limited to two subpopulations.
• There is a need for methods applicable to multiple subpopulations with varying prevalences.

Purpose of the Study:

• To propose a procedure for partitioning reference intervals into several Gaussian subpopulations.
• To develop a computer program to support the calculations for this procedure.
• To generalize partitioning methods beyond the two-subpopulation limitation.

Main Methods:

• The core method involves partitioning reference intervals when subpopulation proportions outside combined limits deviate from the nominal 0.025.
• An equation solver algorithm is employed to determine the combined reference interval.
• The procedure accommodates subpopulations with unequal prevalences and differing sample-to-prevalence ratios.

Main Results:

• The equation solver algorithm successfully identified combined reference intervals for two or more subpopulations.
• The method demonstrated consistency with existing approaches when tested with similar data.
• A user-friendly computer program was developed and made available online.

Conclusions:

• The proposed procedure offers a generalized and complementary approach to existing reference interval partitioning methods.
• It effectively calculates combined reference intervals even when sample fractions do not mirror prevalence fractions.
• The procedure's applicability to multiple Gaussian subpopulations with unequal prevalences is a significant advantage.