Jenny Y. K. Cheng¹, Tommy Y. T. Cheung²

¹Department of Chemical Pathology, Prince of Wales Hospital, Sha Tin, Hong Kong

²Chemical Pathology Laboratory, Department of Pathology, Princess Margaret Hospital, Kowloon, Hong Kong

Reference intervals (RIs) are required for proper interpretation of laboratory results. However, establishing RIs with direct sampling is a resource-intensive practice. Moreover, specimen types requiring invasive collection are uncommonly obtained from healthy individuals. We propose a novel approach in estimating RIs from routine clinical data. The presented algorithm, named as ‘RECOVAR’, can produce continuous RIs adjusted for covariates such as age.

 

Routine clinical results of 16 common analytes were retrieved from laboratory information system of Prince of Wales Hospital and Princess Margaret Hospital. Data were analyzed with Python 3.11 and associated packages including Matplotlib v.3.8.0, NumPy v.1.26.4, Pandas v.2.2.2, SciPy v.1.13.1, and statsmodels v.0.14.4. A custom pipeline was developed for estimation of RIs.

 

Continuous RIs of the 16 analytes with respect to age (1 month-18 years) were produced with minimal manual selection. Results were in general agreement with RIs from Prince of Wales Hospital. Additionally, covariates apart from age were also tested. The estimator demonstrated a change in the span in RIs for plasma calcium across various levels of albumin. The algorithm would be available as a Google Colab Notebook.

 

A new algorithm for determining covariate-adjusted RIs was developed. It was robust against pathological data, allowing utilisation of routine laboratory data in establishing RIs.