Moving Averages: Real-Time Data Monitoring in the Clinical Chemistry Laboratory

Abstract

Laboratorians are well accustomed to the practice of quality control (QC) with Levy-Jennings plots, both prior to and following analytical runs. One limitation of Levy-Jennings plots is that they represent QC at a single moment in time and are insensitive to error between QC runs. Currently, laboratory information systems, middleware systems, and laboratory instruments themselves offer limited options for real-time data monitoring (RTDM) and data exploration in clinical chemistry. In addition, these solutions are often proprietary, difficult to customize, structurally isolated to single vendors, and do not extend throughout the breadth of the clinical laboratory. Our purpose was to design and implement a technology-agnostic quality assurance framework, developed in our institution, to allow RTDM via moving average calculations. The moving averages platform was based on nonrela-tional (NoSQL), or document-oriented, database technology (elasticsearch, Los Altos, CA). The framework was hosted on a Windows-based machine (Microsoft, Redmond, WA). Deidentified laboratory data for testing were obtained via static data export from the Epic Clarity database and will soon be implemented with an HL7 interface for live streaming data. These data were stored within the NoSQL database to allow for rapid scaling and analytics. This approach allowed for successful implementation of a web-based application with a dashboard of moving average charts of individual chemistry assay components. Testing demonstrated a minute-wise moving average calculation for >400,000 chemistry results in a 30-day time period could be completed in less than 2 milliseconds. Moving average parameters are adjustable by the end user depending on the component being monitored. These parameters include moving-average calculation methods, chemistry component, expected normal ranges, number of samples used to calculate moving average, and time resolution. As a result, this has provided our institution with the capability of RTDM for chemistry assays. RTDM and quality assurance tools are widely used in other industries and historically have seen a slow migration into health care and the clinical laboratory practices. While Levy-Jennings plots have shown the ability to detect systematic and random errors, they do so only at isolated moments in time. RTDM with AJCP / MEETING ABSTRACTS