Establishment and evaluation of a prediction model for levosimendan induced arrhythmia based on logistic regression and ROC curve
AIM To explore the clinical characteristics of levosimendan-induced arrhythmia,analyze the possible risk factors,and establish a prediction model.METHODS Clinical data of 643 discharged patients treated with levosimendan in a 3rd Grade Class A hospital from October 2020 to June 2022 were retrospectively collected and the time and type of arrhythmia were analyzed.Logistic regression analysis and ROC curve analysis were used to construct and evaluate the prediction model.Clinical data from another 69 patients were collected to verify the accuracy of the model.RESULTS Among 643 discharged patients,94 cases(14.6%)had arrhythmias,of which 82 cases(87%)occurred during infusion and 2 days after infusion,and atrial fibrillation(32%)was the most common.Multivariate logistic regression analysis showed that the lowest blood potassium within 5 days after medication,left ventricular end-diastolic diameter(LV),and micropump velocity>0.2 μg·kg-1·min-1 were independent risk factors for levosimendan-induced arrhythmia(all P<0.05),with the prediction model formula logit(P)=2.821-0.873 x lowest serum potassium within 5 days after medication(0:>3.89mmol·L-1;1:≤ 3.89 mmol·L-1)+0.697 x LV(0:≤ 55 mm;1:>55 mm)-1.054 x micropump speed>0.2 μg·kg-1·min-1(0:no;1:yes).When the cutoff value of logit(P)was 0.91,the area under the ROC curve was 0.672(95%confidence interval:0.627 to 0.715,P<0.001).The accuracy of the internal verification prediction model was 83%.CONCLUSION The incidence of arrhythmia induced by levosimendan is high,and ECG monitoring should be strengthened within 2 days after infusion.Blood potassium and micropump speed within 5 days after medication are controllable factors,and intervention should be strengthened to reduce the incidence of arrhythmia.