Recommended changes of the current version of the German Rili-BAEK

Abstract

<jats:title>Abstract</jats:title> <jats:p>A number of improvement proposals and corrections of the German Rili-BAEK (Guideline of the German Medical Association on Quality Assurance in Medical Laboratory Examinations) are discussed with special focus on the internal and external quality assurance (IQA/EQA) as well as reference intervals for quantitative results. Particular attention is paid to reconsider the retrospective analysis of control measurements. Such an analysis can be very useful to monitor establishing errors of measurement even before they become critical. The present method “Quadratischer Mittelwert der Messabweichung (QMMA)” has proved to be ineffective. Furthermore, the current idea of a common limit for single control measures and the retrospective statistics must be revised. As a more sophisticated concept, the novel Adaptive Retrospective Monitoring (ARM) has been developed. ARM is recommended as the new minimum requirement for the entire internal quality assurance. Further proposals to enhance clarity are given concerning the release decisions of medical devices and the EQA. Individualized medicine begins with a patient-specific interpretation of analytic results. This requires standardized subgroup-specific reference intervals with smooth age-related adaptations. Only large laboratories are able to ensure the desired specificity and a sufficient statistical significance of self-developed in-laboratory reference intervals. Hence, the need of a central database for harmonized reference intervals is discussed and recommended. Suitable and consistent reference intervals are also an essential prerequisite for unitless laboratory values like the zlog value.</jats:p>

The Statistical Monitoring by Adaptive RMSTD Tests: an efficient, informative, and customizable method for the complete internal quality control intended for low-frequent sampling of control measures

“…This "multi-source" data set consists of 150 remaining entries. Except slightly less populated margin areas, the histogram of the multi-source data is similar to the full-data histogram (compared in [16] [18,19] were derived from biological variations (see [20,21]). The 95% confidence intervals of λ based on the multi-source dataset are 2.23±0.5 (for κ=2.33) and 2.0±0.4 (for κ=1.96).…”

“…(4) with t=3 may provide a still too restrictive limit for SMC (and probably even for RA) in several situations: (i) underestimated seval, δeval due to poor statistics during evaluation period, (ii) measuring techniques with small imprecision but significant inaccuracy, and/or (iii) a high Sigma Metric of the method (usable room for extended tolerance). As revealed in [16], the implication of the present Rili-BAEK to use the same limit for both the single SMC value and the RA is clearly problematic. In fact, the limits should optimally differ by a factor of about two, where LSMC>LΔ (see [16] and Chapter 2.1).…”

“…As revealed in [16], the implication of the present Rili-BAEK to use the same limit for both the single SMC value and the RA is clearly problematic. In fact, the limits should optimally differ by a factor of about two, where LSMC>LΔ (see [16] and Chapter 2.1). This finding and the given statements to handle outliers require a complete revision of the currently prescribed RA of the RiliBAEK [15].…”

“…The presented SMART method is intended to provide a common approach for the entire IQC, considering only one prescribed maximum permissible limit. It is based on the finding that an approximately constant relation exists between the optimal maximum permissible limits for one SMC value (LSMC) and for the true RMSTD of a (hypothetically) very large retrospective set of SMC data (LΔ) [16]. Hence, the general parameter λ is defined as the relation factor…”

“…Such limits are often the only statutorily prescribed information in guidelines. For example, the limits given in the German guideline (Rili-BAEK, table B1a-c column 3, in relative values [15]) can be considered as sole LSMC data [16]. In contrast to the determination of LΔ, the definition of LSMC is usually done empirically at a certain significance level and/or related to a fraction of the biological variation of the analyte [20,21].…”