Indirect Estimation of Clinical Chemical Reference Intervals from Total Hospital Patient Data: Application of a Modified Bhattacharya Procedure

Abstract

Summary:Indirect methods for the determination of reference intervals utilise äs input the total bulk of test results from all admitted hospital patients. One of the strongest arguments in favour of the use of these indirect methods is that one automatically obtains a one to one relationship between the selected population sample and the category of people for which the test results are intended. A study was conducted of the suitability of a inodification of the indirect Bhattacharya method, which also permits the evaluation of heavily skewed distributions. The robustness of the method was evaluated by studying the homogeneity, stability and reproducibility of the hospital population in terms of the statistical parameters of the input distributions and of the resulting reference intervals for creatinine, uric acid and alanine aminotransferase. Finally, the * age-and sex dependency for 15 clinical chemical parameters was compared with reported literature data. Indirekte Bestimmung klinisch-chemischer Referenzintervalle aus den gesamten Daten von IntroductionTherefore we agree with Schütz (1) that it is often It is now generally accepted that the term "Normal easier to diagnose illness than health. Values" should be abolished when dealing with the criterion for the intuitive Interpretation of test results. Considerations regarding the semantics of the referThe application of a normal ränge can only be justi-ence interval concept has been discussed at length by fied when a representative and homogeneous selec-several authors (2 -6). As advocated by the IFCC tion of the healthy population is present for the Expert Panel on Theory of Reference Values (7) it is pripr study of its "healthy" .parameters. The problems good practice to speak of reference values that result associated with ascertaining state of health are great. in a reference interval after appropriate assessment

Determination of reference limits: statistical concepts and tools for sample size calculation

“…Unfortunately, such an indirect approach (cf. [17]) necessarily requires to rely on model assumptions which are not open to direct empirical validation. Concerning the question of whether or not it is admissible to tolerate in a study to be evaluated by means of direct methods a certain amount of values measured in diseased individuals rather than healthy controls, no general recommendations can be given.…”

Age- and gender-specific brain natriuretic peptide (BNP) reference ranges in primary care

“…For establishing the reference ranges, we used the Bhattacharya method, after logarithmic data transformation [32] (courtesy of Dr. Graham Jones, Department of Chemical Pathology, St. Vincent's Hospital, Sydney, Australia). This deconvolution technique is able to identify an underlying normal distribution in a dataset containing abnormal results from patients [33].…”

“…This technique assumes that the majority of data is from "normal", healthy subjects and our study cohort met this requirement. The Bhattacharya technique has been shown to be useful for estimating reference ranges in large cohorts of patients, even in a hospital setting referring to much more severely ill patients [33].…”

Establishing reference intervals in the coagulation laboratory

“…The more defined the subjects are, the better the evaluation of health status. Recruiting subjects by indirect methods include data mining based on selection of healthy individuals from general hospital populations based on statistical methods . Using an indirect method can be problematic as there is little or no control of pre‐analytical conditions which are very relevant in coagulation testing.…”