B-252 Prospective real-time implementation and clinical validation of patient-based real-time quality control using average of normals in a tertiary hospital laborator

Abstract

Practical implementation of patient-based real-time quality control (PBRTQC) requires analyte- and institution-specific optimization, particularly in tertiary hospital settings characterized by substantial patient heterogeneity and case-mix variability. This study evaluated the operational feasibility of average-of-normals (AoN)-based PBRTQC through a structured selection process and prospective event-based validation. PBRTQC monitoring was performed using the Average of Normals module in navify® Lab Operations (Roche Diagnostics, Mannheim, Germany), which implements a simple moving average (SMA) algorithm. Candidate analytes were selected using retrospective routine patient results generated on cobas 8000 series analyzers (Roche Diagnostics, Germany). Analyte selection was guided by a predefined, institution-specific algorithm incorporating distributional stability, biological variability, and operational characteristics. Outpatient (O) and inpatient (I) groups were analyzed separately when the difference in coefficient of variation exceeded 0.5%. PBRTQC parameters were optimized using a retrospective training dataset and subsequently verified in an independent dataset. Prospective validation was conducted using an event-based framework incorporating patient sample re-measurement and contemporaneous IQC evaluation. Based on the selection algorithm, albumin, sodium, chloride, potassium, calcium, and free thyroxine were initially selected, with AFP, AST, ALT included for comparative evaluation. Albumin was subsequently excluded after retrospective simulation and exploratory visualization demonstrated frequent false-positive alarms and pronounced heterogeneity incompatible with stable SMA-based PBRTQC monitoring. Across final 14 selected PBRTQC parameter sets, 217,372 AoN points generated 1,599 alarms that were consolidated into 300 evaluable events. ALT(O), AST(O), and Cr(I), did not generate any AoN alarms, whereas analytes that did generate alarms showed event-to-alarm ratios ranging from approximately 13% to 52%. Analytically meaningful events, defined as confirmed bias on patient re-measurement with either failing or passing IQC, were observed primarily in analytes with stable distributions and acceptable analytical variability. Calcium, creatinine (O), and sodium showed substantial proportions of meaningful events, accounting for approximately 38% to 82% of total events. In contrast, analytes with high variability or non-Gaussian distributions—including AST, ALT, AFP, and potassium (O) —generated predominantly false-positive alarms or failed to generate confirmable events. Comparison with conventional IQC demonstrated limited overlap, with only 4% of IQC reject events occurring concurrently with AoN alarms. AoN-based PBRTQC can complement conventional IQC by enabling detection of analytical deviations not identified by control materials for selected analytes. Its performance is strongly dependent on analyte characteristics and institution-specific patient result distributions, underscoring the need for ongoing monitoring and periodic re-evaluation to ensure sustained performance in routine laboratory practice.