An automated method for constructing continuous reference intervals based on real-world data-driven approaches and forming intelligent agents.

INTRODUCTION: Deriving reference intervals (RIs) for clinical laboratory tests requires age- and gender-stratified processing with strict exclusion of outliers. However, most laboratories still rely on manufacturer-defined, single, non-continuous reference intervals. This study aims to develop and validate a fully automated, open-source analytical workflow that simultaneously addresses these limitations and enables operation without programming expertise through an agent-based AI interface. METHODS: We developed an automated open-source workflow applied to 222,010 routine test records. These included alpha-fetoprotein (AFP), albumin(Alb), alkaline phosphatase (ALP), pro-gastrin-releasing peptide (proGRP), total protein (TP), and Urea. The framework integrates GAMLSS-based age trend modelling, a novel iterative automated age stratification algorithm, five parallel outlier removal methods (Tukey, Gaussian mixture models , Z-score, Isolated Forests, and Local Outlier Factor), bias ratio consistency assessment, and internal time series validation. The complete workflow is encapsulated as "agent skills" deployable via a large language model conversational interface. RESULTS: The automated age stratification algorithm identified 46 gender-age subgroups across six analyte-gender combinations in both males and females. Of 230 estimated RIs, 229 (99.6%) met the CLSI EP28-A3c time validation threshold (≥90%), with an average validation consistency of 94.6% (standard deviation 2.1%). Analysis of the bias ratio across five algorithms confirmed excellent consistency between methods across all subgroups. CONCLUSION: This workflow generates statistically rigorous, biologically plausible, and algorithmically consistent reference intervals without requiring predefined age grouping, parametric distribution assumptions, or programming expertise. This agent skills deployment architecture enables equitable access to state-of-the-art reference interval construction methods for clinical laboratories worldwide.