Covariate-Adjusted Adaptive Reference Ranges in Longitudinal Data Monitoring

Forough Pazhuheian, University of Galway

Co-authors: John Newell, University of Galway; Davood Roshan, University of Galway

Abstract: Clinical reference ranges are vital for interpreting laboratory test results, providing benchmarks for medical diagnostics. Recently, adaptive reference ranges have enhanced personalised monitoring by detecting abnormalities while accounting for individual biomarker variability over time. However, existing methods overlook the influence of clinical and physiological covariates on biomarker fluctuations over time. This paper proposes a Mixed Effects Modelling framework, using the Expectation-Maximization (EM) algorithm, to derive covariate-adjusted adaptive reference ranges. Applied to longitudinal data, the method demonstrated improved biomarker monitoring when covariates were strongly correlated with the biomarker. However, in the absence of such a strong correlation, the covariate-adjusted and regular adaptive reference ranges yielded similar results. While adaptive reference ranges enhance personalised monitoring by accounting for individual variability, their effectiveness may be limited if relevant covariates are not considered. Our proposed Mixed Effects Modelling framework addresses this limitation, offering a more robust tool for longitudinal biomarker monitoring and abnormality detection, particularly when strong correlations exist between covariates and biomarkers.