A model for managing quality control for a network of clinical chemistry instruments measuring the same analyte.

OBJECTIVES: Monitoring quality control for a laboratory or network with multiple instruments measuring the same analyte is challenging. We present a retrospective assessment of a method to detect medically significant out-of-control error conditions across a group of instruments measuring the same analyte. The purpose of the model was to ensure that results from any of several instruments measuring the same analytes in a laboratory or a network of laboratories provide comparable results and reduce patient risk. Limited literature has described how to manage QC in these very common situations. METHODS: Single Levey-Jennings control charts were designed using peer group target mean and control limits for five common clinical chemistry analytes in a network of eight analyzers in two different geographical sites. The QC rules used were 13s/22s/R4s, with the mean being a peer group mean derived from a large population of the same instrument and the same QC batch mean and a group CV. The peer group data used to set the target means and limits were from a quality assurance program supplied by the instrument supplier. Both statistical and clinical assessments of significance were used to evaluate QC failure. Instrument bias was continually monitored. RESULTS: It was demonstrated that the biases of each instrument were not statistically or clinically different compared to the peer group's average over six months from February 2023 until July 2023. Over this period, the error rate determined by the QC model was consistent with statistical expectations for the 13s/22s/R4s rule. There were no external quality assurance failures, and no detected error exceeded the TEa (medical impact). Thus, the combined statistical/clinical assessment reduced unnecessary recalibrations and the need to amend results. CONCLUSIONS: This paper describes the successful implementation of a quality control model for monitoring a network of instruments, measuring the same analytes and using externally provided quality control targets. The model continually assesses individual instrument bias and imprecision while ensuring all instruments in the network meet clinical goals for quality. The focus of this approach is on detecting medically significant out-of-control error conditions.

Reference-mean-centered statistical quality control.

Background: Statistical quality control (SQC) procedures generally use rejection limits centered on the stable mean of the results obtained for a control material by the analyzing instrument. However, for instruments with significant bias, re-centering the limits on a different value could improve the control procedures from the viewpoint of patient safety. Methods: A statistical model was used to assess the effect of shifting the rejection limits of the control procedure relative to the instrument mean on the number of erroneous results reported as a result of an increase in the systematic error of the measurement procedure due to an out-of-control condition. The behaviors of control procedures of type 1ks (k = 2, 2.5, 3) were studied when applied to analytical processes with different capabilities (σ = 3, 4, 6). Results: For measuring instruments with bias, shifting the rejection limits in the direction opposite to the bias improves the ability of the quality control procedure to limit the risk posed to patients in a systematic out-of-control condition. The maximum benefit is obtained when the displacement is equal to the bias of the instrument, that is, when the rejection limits are centered on the reference mean of the control material. The strategy is sensitive to error in estimating the bias. Shifting the limits more than the instrument's bias disproportionately increases the risk to patients. This effect should be considered in SQC planning for systems running the same test on multiple instruments. Conclusions: Centering the control rule on the reference mean is a potentially useful strategy for SQC planning based on risk management for measuring instruments with significant and stable uncorrected bias. Low uncertainty in estimating bias is necessary for this approach not to be counterproductive.