Linear time-dependent reference intervals where there is measurement error in the time variable-a parametric approach

Area of Science:

• Biostatistics
• Clinical Chemistry
• Measurement Error Analysis

Background:

• Parametric methods for time-specific reference intervals are crucial in clinical diagnostics.
• Existing methods often overlook or incorrectly handle measurement error in the time covariate.
• This can lead to biased interval calculations and unequal patient treatment.

Purpose of the Study:

• To re-examine parametric methods for calculating time-specific reference intervals in the presence of measurement error in the time covariate.
• To demonstrate the inadequacy of standard ordinary least squares (OLS) approaches when such errors exist.
• To propose and validate a method that ensures equitable patient referral patterns.

Main Methods:

• Application of measurement error models to address inaccuracies in the time covariate.
• Comparison of results with traditional weighted ordinary least squares (WLS) methods.
• Analysis of referral patterns generated by different methodologies.

Main Results:

• Ordinary least squares (OLS) methods, even when weighted, are inappropriate when measurement error is present in the time covariate.
• The use of incorrect methods can result in referral patterns that vary with covariate values, violating the principle of equal treatment.
• Measurement error models produce reference intervals that adhere to the principle of equal treatment for all subjects.

Conclusions:

• Measurement error models provide a statistically sound approach for calculating time-specific reference intervals.
• Adopting these models ensures adherence to the International Federation for Clinical Chemistry's principle of equal treatment.
• Accurate interval calculation is essential for equitable clinical decision-making and patient management.