JMIR Publications

ABSTRACT

Background:

Reference intervals (RIs) play an important role in clinical decision-making. However, due to the time, labor, and financial costs involved in establishing RIs by direct means, use of indirect methods, based on previously obtained “big data” from clinical laboratories is increasingly preferred. At present, there has been no systematic evaluation of diverse indirect techniques, combined with data transformation approaches and various outlier removals for calculating RIs.

Objective:

This study aims at evaluating the accuracy of combinations of different data transformation, outlier-removal, and indirect techniques on establishing RIs for large-scale data. Moreover, data derived from direct methods as reference standards were used.

Methods:

Data of apparently healthy individuals aged ≥ 18 years were retrieved from a hospital in northern China. We compared parametric, nonparametric, Hoffmann, Bhattacharya, and truncation points and Kolmogorov-Smirnov Distance (kosmic) indirect methods, combined with log or BoxCox transformation, and Dixon, Tukey, and mean±3SD with iteration-outlier removal approaches to derive RIs for eight complete blood count (CBC) parameters with those established directly and previously. Furthermore, bias ratios (BRs) were calculated to assess which combination of indirect technique, data transformation pattern, and outlier removal method preferrable.

Results:

Skewness of seven CBC parameters (platelets, red blood cells, HGB, MCH, MCHC, MCV, and HCT) was greater after BoxCox than after log transformation. Tukey-based outlier removal yielded the maximum number of outliers. Lower limit (LL) bias of WBC (Male), PLT (Male), HGB (Male), MCH (Male/Female), and MCV (Female) was greater than that of the corresponding upper limit (UL) for more than half of 30 indirect methods. Among the top-10 methodologies for WBC, PLT, HGB, MCV, and MCHC with high-BR qualification rate among males, Bhattacharya, Hoffmann, and parametric methods were superior to the other two indirect methods, whereas kosmic method performed better for deriving MCHC RIs among females.

Conclusions:

Compared with results derived by direct method, outlier-removal methods and indirect techniques markedly influence the final RIs, whereas data transformation has negligible effects except for obviously-skewed data. This study provides scientific evidence for clinical clinicians to use their previous datasets to establish RIs.