COMPARISON OF UNIVARIATE AND MULTIVARIATE REFERENCE INTERVAL METHODS

Abstract

BackgroundIn clinical practice, reference intervals play a pivotal role in interpreting laboratory test results. Yet, when several tests are taken into consideration simultaneously, the traditional univariate intervals might not suffice due to the elevated risk of Type 1 errors. MethodsThis study introduces and evaluates two multivariate reference interval techniques: one based on Mahalanobis distance and the other an adaptation of the multivariate confidence interval. Using Monte Carlo simulations, we focused our assessments on the interplay between "Serum Ferritin and Transferrin Saturation" values. ResultsUpon evaluation, it became evident that the multivariate methods significantly reduced false positives. They presented enhanced accuracy over traditional univariate intervals. Notably, the method involving Mahalanobis distance stood out in terms of efficacy. ContributionsBeyond presenting novel techniques, our research underscores the importance and potential of using multivariate approaches in clinical lab settings. The findings can guide better medical decision-making, ensuring optimized allocation of healthcare resources. HighlightsO_LIWe present two alternative multivariate reference interval methods that consider the relationship between two analytes simultaneously. C_LIO_LIThrough comprehensive simulation studies, we compare the proposed methods with the conventional univariate reference interval method. C_LIO_LIOur results show the superior performance of our proposed methods in terms of patient rate, sensitivity, specificity, and overall accuracy. C_LIO_LIWe discuss the practical relevance of these new methods. C_LI