An integrated framework for universal coagulation test reference intervals: Cross-platform validation of eleven...

Area of Science:

• Clinical Chemistry
• Biostatistics
• Medical Informatics

Background:

• Conventional direct methods for coagulation reference intervals (RIs) face challenges with population recruitment, data distribution, and age effects.
• Indirect methods using real-world data (RWD) lack consensus on algorithms and cross-platform validation.

Purpose of the Study:

• To develop a multi-algorithm framework for evaluating outlier detection algorithms.
• To optimize coagulation RIs for the Chinese population using real-world data.
• To validate the cross-platform applicability of selected algorithms.

Main Methods:

• Retrospective analysis of 630,946 coagulation test records.
• Evaluation of 11 outlier detection algorithms for RIs derivation.
• Quantification of algorithm stability and sex/age effects using reference change value (RCV) and variance component models.
• Cross-platform validation on Mindray CX-9000 and Stago STA R Max.

Main Results:

• Z-Score and Tukey algorithms showed optimal performance for 8 coagulation parameters.
• Minimal sex-based differences observed, negating the need for sex stratification.
• Age stratification was necessary for D-Dimer in individuals aged 60 years and older.
• Cross-platform consistency confirmed algorithm generalizability.

Conclusions:

• The Z-Score and Tukey methods are optimal for deriving coagulation RIs.
• Age-specific D-Dimer RIs are recommended; sex stratification is unnecessary.
• This RCV-driven framework enhances standardization in coagulation testing.