Rotational thromboelastometry (ROTEM) is widely used for point-of-care coagulation testing to reduce blood transfusions. Accurate interpretation of ROTEM data is crucial and requires substantial training. This study investigates the inter- and intrarater reliability of ROTEM interpretation among experts and compares their interpretations with a ROTEM-guided algorithm. This study was conducted at Amsterdam University Medical Center and included 90 cardiac surgery patients. ROTEM data were collected at 4 surgical stages: before induction, after aortic declamping, postcoagulation correction, and within 2 hours of intensive care unit (ICU) admission. An international panel of 7 cardiovascular anesthesiologists and one intensivist interpreted the data. Interrater reliability was assessed using Fleiss' kappa for binary decisions and the simple matching coefficient (SMC) for multiple-choice questions. Intrarater reliability with the ROTEM-guided algorithm was also evaluated. Three hundred forty-three ROTEM measurements were analyzed. The interrater reliability for binary decisions was substantial to almost perfect, except after declamping (Fleiss' kappa = 0.34). The SMC for determining type of abnormality and interventions ranged from good to excellent across all ROTEM measuring moments (SMC ≥0.75). Intrarater reliability was almost perfect for binary questions (intraclass correlation coefficient [ICC] ≥0.81) and showed excellent agreement for multiple-choice questions. Comparing expert recommendations with the algorithm resulted in an average SMC of 0.70 indicating differences in intervention recommendations, with experts frequently recommending fibrinogen and protamine over the algorithm's suggestions of plasma and PCC. This study demonstrates high inter- and intrarater reliability in ROTEM interpretation among trained professionals in cardiac surgery, with almost perfect agreement on abnormalities and interventions. However, differences between expert evaluations and the ROTEM-guided algorithm underscore the need for advanced clinical decision-making tools. Future efforts should focus on developing personalized, data-driven algorithms without predefined cutoff values to improve accuracy and patient outcomes.
Read full abstract