Discrepancy between visual and spectrophotometric assessment of sample haemolysis

Background: This study assessed the agreement between visual grading of sample haemolysis with automated haemolysis index (H index) measurement on the Roche 917 clinical chemistry analyser. Methods: The H indices of 800 serum potassium samples and 800 EDTA plasma troponin I samples were compared with routine visual grading of sample haemolysis by technologists. Photographs of mild (haemoglobin 1 g/L), moderate (2·5 g/L) and severe (5 g/L) haemolysis were available for reference. Results: For serum samples, the weighted kappa coefficient was 0·42 with an 8·0% haemolysis rate determined by visual inspection and 3·4% by H index. For EDTA-plasma samples, the weighted kappa coefficient was 0·35 with a 4·5% haemolysis rate determined by visual inspection and 7·9% by H index. Conclusion: Visual assessment of sample haemolysis is unreliable and can give variable results depending on sample type. Laboratories should consider investigating sample haemolysis using automated haemolysis index measurement as an alternative to visual grading whenever it is available.