Multivariate anomaly detection models enhance identification of errors in routine clinical chemistry testing.

OBJECTIVES: Conventional autoverification rules evaluate analytes independently, potentially missing unusual patterns of results indicative of errors such as serum contamination by collection tube additives. This study assessed whether multivariate anomaly detection algorithms could enhance the detection of such errors. METHODS: Multivariate Gaussian, k-nearest neighbours (KNN) distance, and one-class support vector machine (SVM) anomaly detection models, along with conventional limit checks, were developed using a training dataset of 127,451 electrolyte, urea, and creatinine (EUC) results, with a 5 % flagging rate targeted for all approaches. The models were compared with limit checks for their ability to detect atypical EUC results from samples spiked with additives from collection tubes: EDTA, fluoride, sodium citrate, or acid citrate dextrose (n=200 per contaminant). The study additionally assessed the ability of the models to identify 127,449 single-analyte errors, a potential weakness of multivariate models. RESULTS: The KNN distance and SVM models outperformed limit checks for detecting all contaminants (p-values <0.05). The multivariate Gaussian model did not surpass limit checks for detecting EDTA contamination but was superior for detecting the other additives. All models surpassed limit checks for identifying single-analyte errors, with the KNN distance model demonstrating the highest overall sensitivity. CONCLUSIONS: Multivariate anomaly detection models, particularly the KNN distance model, were superior to the conventional approach for detecting serum contamination and single-analyte errors. Developing multivariate approaches to autoverification is warranted to optimise error detection and improve patient safety.

Sharing reference intervals and monitoring patients across laboratories - findings from a likely commutable external quality assurance program.

OBJECTIVES: Laboratory results are increasingly interpreted against common reference intervals (CRIs), published clinical decision limits, or previous results for the same patient performed at different laboratories. However, there are no established systems to determine whether current analytical performance justifies these interpretations. We analysed data from a likely commutable external quality assurance program (EQA) to assess these interpretations. METHODS: The use of CRIs was assessed by evaluating instrument group medians against minimum specifications for bias. The use of clinical decision limits was assessed using specifications from professional bodies, and the monitoring of patients by testing at different laboratories was assessed by comparing all-laboratory imprecision to within-subject biological variation. RESULTS: Five of the 18 analytes with Australasian CRIs did not meet specification for all instrument groups. Among these, calcium and magnesium failed for one instrument group out of seven, while bicarbonate, chloride, and lipase failed for two instrument groups. Of the 18 analytes reviewed currently without CRIs in Australasia, 10 candidates were identified. Among analytes with clinical decision limits, i.e. lipids, glucose, and vitamin D, only triglycerides met both bias and imprecision specifications, while vitamin D met the imprecision specification. Monitoring patients by testing at different laboratories was supported for 15 of the 46 (33 %) analyte-method principles groups that met minimum imprecision specifications. CONCLUSIONS: Analysis of data from commutable EQA programs can provide a mechanism for monitoring whether analytical performance justifies the interpretations made in contemporary laboratory practice. EQA providers should establish systems for routinely providing this information to the laboratory community.