# Establishment and Temporal Validation of Next-Generation Reference Intervals for Routine Hematological Parameters Using Large-Scale Data

## TL;DR

This study creates dynamic reference intervals for blood tests that change with age and sex, using large health data and showing they remain stable over time.

## Contribution

The study introduces a reproducible method for constructing stable, age-continuous next-generation reference intervals for hematological parameters.

## Key findings

- Age- and sex-dependent patterns were observed in RBC, HGB, and HCT.
- Percentile curves showed gradual age-related changes, especially in females after midlife.
- Temporal validation showed consistent performance with less than 10% outside reference intervals.

## Abstract

Background: Conventional reference intervals (RIs) are typically expressed as fixed limits and may not adequately reflect continuous biological variation across age and sex. Next-generation reference intervals (NGRIs) allow dynamic modeling of laboratory parameters across the lifespan. This study aimed to establish age- and sex-specific NGRIs for routine hematological parameters using large-scale health examination data and to evaluate their temporal stability. Methods: Health examination records were linked with laboratory data, and a relatively healthy reference population was defined based on age (18–80 years), normal body mass index, normal blood pressure, and absence of documented disease history. NGRIs were constructed using generalized additive models for location, scale, and shape (GAMLSS) with the Box–Cox Cole and Green distribution. Age-dependent percentile curves (2.5th–97.5th) were generated using bootstrap resampling (100 iterations). Temporal external validation was conducted in five independent annual cohorts (2019–2023). Results: Age- and sex-dependent distributional patterns were observed across multiple hematological parameters, particularly RBC, HGB, and HCT. Continuous percentile curves demonstrated gradual age-related trends, with more pronounced changes in females after midlife. In temporal validation cohorts, the proportion of individuals classified outside the reference intervals remained consistently below 10% across years and sexes, indicating stable performance. Conclusions: Using large-scale real-world health examination data and a flexible distributional modeling framework, we established stable age-continuous next-generation reference intervals for routine hematological parameters. The proposed approach provides a reproducible strategy for modernizing laboratory reference interval construction and supports broader implementation in population-based laboratory medicine.