Next-generation reference intervals for pediatric hematology

Abstract

Background Interpreting hematology analytes in children is challenging due to the extensive changes in hematopoiesis that accompany physiological development and lead to pronounced sex- and age-specific dynamics. Continuous percentile charts from birth to adulthood allow accurate consideration of these dynamics. However, the ethical and practical challenges unique to pediatric reference intervals have restricted the creation of such percentile charts, and limitations in current approaches to laboratory test result displays restrict their use when guiding clinical decisions. Methods We employed an improved data-driven approach to create percentile charts from laboratory data collected during patient care in 10 German centers (9,576,910 samples from 358,292 patients, 412,905–1,278,987 samples per analyte). We demonstrate visualization of hematology test results using percentile charts and z-scores (www.pedref.org/hematology) and assess the potential of percentiles and z-scores to support diagnosis of different hematological diseases. Results We created percentile charts for hemoglobin, hematocrit, red cell indices, red cell count, red cell distribution width, white cell count and platelet count in girls and boys from birth to 18 years of age. Comparison of pediatricians evaluating complex clinical scenarios using percentile charts versus conventional/tabular representations shows that percentile charts can enhance physician assessment in selected example cases. Age-specific percentiles and z-scores, compared with absolute test results, improve the identification of children with blood count abnormalities and the discrimination between different hematological diseases. Conclusions The provided reference intervals enable precise assessment of pediatric hematology test results. Representation of test results using percentiles and z-scores facilitates their interpretation and demonstrates the potential of digital approaches to improve clinical decision-making.

Blood counts in adult and elderly individuals: defining the norms over eight decades of life

“…In elderly patients, recruitment of ‘healthy’ individuals poses additional challenges due to the high prevalence of chronic diseases and prescription medication: exclusion of individuals according to these criteria leads to the selection of the ‘most healthy minority’, with questionable appropriateness of resulting reference intervals for the general population, and substantial practical obstacles when recruiting reference individuals. We circumvented these problems by using a data mining approach and applying an algorithm (Arzideh et al , ), that has been successfully used to evaluate paediatric (Zierk et al , , , ) and adult (Zierk et al , ) reference intervals in haematology, with the aim to define age‐ and sex‐specific reference intervals in the elderly. We did not collect any medical histories, but, starting with two very large datasets of routine clinical samples, we removed pathological samples in a five‐step process: (i) we excluded units and specialists caring for patients with a high expected fraction of haematologic abnormalities, (ii) we removed samples from patients with reduced renal function or acute‐phase reaction, (iii) we excluded samples with clearly abnormal test results in the same sample in other haematology analytes, (iv) we selected only one sample per patient, thereby reducing over‐representation of samples from chronically and critically ill patients, and (v) we calculated the distribution of physiologic measurements using an established algorithm (RLE), separating them from pathologic test results.…”

“…However, great care must be taken to avoid an unwarranted shift or widening of reference intervals in populations with a high prevalence of slightly abnormal test results. In previous work, we have shown that establishment of haematology reference intervals using a sophisticated data mining approach (the reference limit estimator, RLE) is feasible, even in the challenging setting of a tertiary care centre and a high proportion of abnormal test results (Zierk et al , , ). In short, this approach assumes a mixture of parametrically distributed samples from healthy individuals and pathologic samples.…”

Reference Interval Estimation from Mixed Distributions using Truncation Points and the Kolmogorov-Smirnov Distance (kosmic)

“…We employ an approach based on previous works by Arzideh et al 9,[16][17][18][19] and our experiences in their application to pediatric and adult datasets 8,[10][11][12][13] : This procedure is based on the assumption that the proportion of physiological samples in the input dataset can be modeled with a parametric distribution (so-called Power Normal distribution, a Gaussian distribution after Box-Cox transformation of the data, i.e. a distribution that can accommodate skewed data), and that a truncation interval T exists within the dataset, in which the proportion of abnormal test results is negligible.…”

“…As large numbers of test results are readily available from laboratory information systems, this enables the establishment of reference intervals specific to different populations, age-groups, analytical devices, and even batches and reagents. Extensive experience with these methods exists in children, where unique ethical challenges limit access to blood samples to create reference intervals 8,[10][11][12] and in challenging adult populations with a high proportion of patients with substantial morbidity and mortality 13 .…”

“…However, both of these methods assume a Gaussian distribution of physiological test results, and require visual identification of a proportion of purely physiological test results, a process which is prone to bias and prevents integration into automated pipelines. Recently, a method developed by Arzideh et al 9,[16][17][18][19] has been used to establish reference intervals for adults 13 and children 8,[10][11][12]20 . This method uses a truncation interval of the range of test results in the input dataset after Box-Cox transformation to estimate a distribution of supposedly physiological test results, and can therefore estimate non-Gaussian distributions.…”