Hierarchy of reference interval models: advancing laboratory data interpretation

Accurate interpretation of laboratory data is a critical step in clinical decision-making. This requires the availability of reliable reference data for comparison. Reference data can be derived from various sources, including hospital or laboratory databases, groups of reference individuals, or an individual's own data, and can be estimated using different statistical approaches. In addition to the possible lack of standardization of measurement methods this diversity results in the availability of multiple reference intervals for a given measurand. However, selecting the most appropriate reference data is challenging and requires a systematic approach to identify the best available option for each measurand. In this opinion paper, we aim to develop a systematic approach for constructing a hierarchical structure encompassing all known reference interval (RI) models, to discuss the advantages and disadvantages of each, and to provide a framework for selecting the most appropriate RI for routine clinical practice. To illustrate the model visually, we constructed a hierarchical pyramid with the less reliable reference intervals positioned at the base, gradually increasing in reliability toward the top. Based on the data sources and the statistical approaches used to estimate RIs, we conclude that, at least from a theoretical perspective, the currently widespread model - discrete population-based RIs derived from hospital or laboratory data - occupies the lowest level, that is, it represents the ground of the hierarchical pyramid, whereas multivariate continuous personalized RIs reside at the top.