Evaluation of current patient-based real-time quality control in clinical chemistry testing.

Abstract

Introduction: To perform simulation studies on patient-based real-time quality control (PBRTQC) for aspartate aminotransferase (AST), iron (Fe), potassium (K), and thyrotropin (thyroid stimulating hormone, TSH) analytes, focusing on optimizing systematic error detection while minimizing data loss.

Methods: Clinical laboratory data for the four analytes were analyzed using various truncation methods. Among these methods, truncation limits corresponding to fixed percentiles (e.g., 1st-99th percentiles), reference change value based on between-individual biological variation (RCVg), and truncation limits derived from ± 3 standard deviations from the mean were included. These exclusion methods were applied using trimming or winsorization techniques, and transformation methods (logarithmic, square root, and Yeo-Johnson transformations) were employed to fit the data to a normal or near-normal distribution. Moving average techniques, such as exponentially weighted moving average (EWMA), were used with various block sizes to evaluate systematic error detection performance.

Results: Truncation based on RCVg improved performance for analytes with lower individuality indices-AST, potassium, and TSH-by enabling faster error detection. In contrast, methods either without truncation or with winsorization applied proved to be more effective for Fe. Among the moving average methods, EWMA with smaller block sizes (20 and 30) generally showed superior performance by detecting systematic errors more quickly.

Conclusion: RCVg-based truncation improves error detection for analytes with low individuality when combined with PBRTQC methods like EWMA, minimizing data loss. Tailored strategies considering analyte-specific individuality and distribution are essential for optimal error monitoring, warranting further validation in diverse clinical settings.