A revised approach to data modelling for patient classification to determine analytical performance specifications...

Area of Science:

• Laboratory medicine
• Clinical diagnostics
• Biostatistics

Background:

• The Milan models, particularly Model 1b, are standard for setting Analytical Performance Specifications (APS) in laboratory medicine.
• These models often assess assay performance by its impact on patient classification, using original data as a baseline.
• Existing methods may not fully account for inherent biological and analytical variation in repeated classifications.

Purpose of the Study:

• To introduce an alternative approach for developing APS in laboratory medicine.
• To establish a method that quantifies baseline expected alternate classification rates.
• To evaluate changes in assay performance relative to this baseline variation.

Main Methods:

• Quantifying the baseline rate of alternate patient classification due to inherent biological and analytical variation.
• Modeling the impact of assay imprecision and bias on this baseline classification rate.
• Comparing changes in classification rates against the established baseline variation.

Main Results:

• The proposed method provides a baseline for expected classification variability.
• It allows for assessment of assay performance changes relative to inherent study variations.
• This approach offers a more robust framework for setting APS.

Conclusions:

• The presented method offers a novel perspective on APS development by incorporating baseline variation.
• This approach enhances the assessment of analytical method performance in laboratory medicine.
• It provides a more accurate and reliable foundation for clinical decision-making based on diagnostic tests.