Development and Validation of a Clinical Score to Predict Neurological Outcomes in Patients With Out-of-Hospital Cardiac Arrest Treated With Extracorporeal Cardiopulmonary Resuscitation

Abstract

Extracorporeal cardiopulmonary resuscitation (ECPR) is expected to improve the neurological outcomes of patients with refractory cardiac arrest; however, it is invasive, expensive, and requires substantial human resources. The ability to predict neurological outcomes would assist in patient selection for ECPR. To develop and validate a prediction model for neurological outcomes of patients with out-of-hospital cardiac arrest with shockable rhythm treated with ECPR. This prognostic study analyzed data from the Japanese Association for Acute Medicine Out-of-Hospital Cardiac Arrest registry, a multi-institutional nationwide cohort study that included 87 emergency departments in Japan. All adult patients with out-of-hospital cardiac arrest and shockable rhythm who were treated with ECPR between June 2014 and December 2017 were included. Patients were randomly assigned to the development and validation cohorts based on the institutions. The analysis was conducted between November 2019 and August 2020. Age (<65 years), time from call to hospital arrival (≤25 minutes), initial cardiac rhythm on hospital arrival (shockable), and initial pH value (≥7.0). The primary outcome was 1-month survival with favorable neurological outcome, defined by Cerebral Performance Category 1 or 2. In the development cohort, a simple scoring system was developed to predict this outcome using a logistic regression model. The diagnostic ability and calibration of the scoring system were assessed in the validation cohort. A total of 916 patients were included, 458 in the development cohort (median [interquartile range {IQR}] age, 61 [47-69] years, 377 [82.3%] men) and 458 in the validation cohort (median [IQR] age, 60 [49-68] years; 393 [85.8%] men). The cohorts had the same proportion of favorable neurological outcome (57 patients [12.4%]). The prediction scoring system was developed, attributing a score of 1 for each clinical predictor. Patients were divided into 4 groups, corresponding to their scores on the prediction model, as follows: very low probability (score 0), low probability (score 1), middle probability (score 2), and high probability (score 3-4) of good neurological outcome. The mean predicted probabilities in the groups stratified by score were as follows: very low, 1.6% (95% CI, 1.6%-1.6%); low, 4.4% (95% CI, 4.2%-4.6%); middle, 12.5% (95% CI, 12.1%-12.8%); and high, 30.8% (95% CI, 29.1%-32.5%). In the validation cohort, the C statistic of the scoring system was 0.724 (95% CI, 0.652-0.786). The predicted probability was evaluated as well calibrated to the observed favorable outcome in both cohorts by visual assessment of the calibration plot. In this study, the scoring system had good discrimination and calibration performance to predict favorable neurological outcomes of patients with out-of-hospital cardiac arrest and shockable rhythm who were treated with ECPR.