Newly developed risk model using the interaction between age and low-flow time for neurological outcomes in patients who undergo extracorporeal cardiopulmonary resuscitation

Background: To develop a risk prediction model for neurological outcomes in patients who underwent extracorporeal cardiopulmonary resuscitation (ECPR). Methods: Between May 2004 and April 2016, a total of 274 patients who underwent ECPR were included in this analysis. To develop a new predictive scoring system, backward stepwise elimination and a z-score-based scoring scheme were used based on logistic regression analyses. The leave-one-out cross-validation (LOOCV) method was used for validation. Results: Ninety-five patients (34.7%) survived until discharge. Of these, 78 patients (28.5%) had good neurological outcomes (CPC scores of 1 or 2). In the multivariable logistic regression analysis, significant predictors of poor neurological outcome included age>65 years, initial Sequential Organ Failure Assessment score>13 points, first monitored arrest rhythm, cardiopulmonary resuscitation (CPR) to extracorporeal membrane oxygenation (ECMO) pump-on time>30 min, initial pulse pressure<25 mmHg, initial mean arterial pressure<70 mmHg, and serum glucose level>300 mg/dL. There was also a significant interaction between age and CPR to ECMO pump-on time. The newly developed neurological outcome score after ECPR (nECPR) more effectively predicted poor neurological outcome (C-statistic 0.867, 95% confidence interval 0.823-0.912) than the former ECPR score (P = .019 for nECPR vs. ECPR) and the survival after veno-arterial ECMO (SAVE) score (P < .001 for nECPR vs. SAVE). LOOCV analysis confirmed the low misclassification rate of the nECPR score compared with ECPR and SAVE scores (13% vs. 22% and 38%, respectively). Conclusions: The nECPR score can better predict early neurological prognosis in patients who undergo ECPR than currently used methods.