Objective: To evaluate the reliability of the image-based sample check module on the cobas 8100 system in detecting haemolysis, icterus and lipaemia, using the quantitative serum index assay as the reference standard.
Methods: A cross-sectional study of 334 serum and plasma samples analysed on the cobas 8100 system at the Department of Biochemistry, Bach Mai Hospital. Each sample was simultaneously classified by colour on the acquired image and quantified for the three serum indices; sensitivity, specificity, predictive values, likelihood ratios and the kappa coefficient were calculated at three sets of reference thresholds.
Results: For haemolysis, both specificity and positive predictive value reached 100%, and sensitivity rose from 45.2% at an H index threshold of 20 to 100% at an H index threshold of 80. For lipaemia, 29.0% of samples were flagged, yet the L index did not differ from that of unflagged samples (p = 0.317), with sensitivity of 50.0% and 64.3% at L index thresholds of 30 and 50, respectively; for icterus, sensitivity was only 2.5%. Using the manufacturer thresholds, any colour alert achieved 84.2% sensitivity, 98.6% negative predictive value, a negative likelihood ratio of 0.23 and missed 3 of 334 samples. Overall agreement between the two methods was 67.1%, with a kappa of 0.194.
Conclusion: Image-based sample classification is highly specific for haemolysis and detected all haemolysed samples above the interference threshold, but it is not sufficiently reliable for icterus and lipaemia.