A multivariate hierarchical Bayesian approach to measuring agreement in repeated measurement method comparison...

Area of Science:

• Biostatistics
• Medical Statistics
• Data Analysis

Background:

• Method comparison studies require assessing both validity (between-method agreement) and reproducibility (within-method agreement).
• The Bland-Altman limits of agreement is favored for validity but less adopted for reproducibility due to limited statistical tools for repeated measurements.
• Lack of flexible, accessible statistical machinery hinders analysis of repeated measurement method comparison data.

Purpose of the Study:

• To present a statistical framework using Bayesian methods to extend the Bland-Altman approach for assessing both validity and reproducibility in method comparison studies with repeated measurements.
• To provide easily implementable multivariate hierarchical Bayesian models for analyzing complex method comparison data.

Main Methods:

• Developed two multivariate hierarchical Bayesian models based on the Bland-Altman framework.
• Model 1: Assumes static underlying values (exchangeable replicates).
• Model 2: Assumes dynamic underlying values (non-exchangeable replicates) for repeated measurements.

Main Results:

• Illustrated model advantages using two previously analyzed datasets.
• Demonstrated application of both models for static underlying values.
• Showcased the non-exchangeable replicate model for continually changing underlying values.

Conclusions:

• The presented Bayesian models offer a flexible and easily implemented solution for method comparison studies.
• Models allow for full parameter uncertainty, simultaneous method comparison, and handling of unbalanced or missing data.
• Freely available computer code in WinBUGS is provided for analysis.