Abstract 257: Clinical Decision Tree Model in Patients with Out-of-Hospital Cardiac Arrest for Predicting Neurological Outcome at the Emergency Department
Yoshikazu Goto, Tetsuo Maeda, Yumiko Nakatsu Goto, Keisuke Ohta · Circulation · 2012
Introduction: The ideal factors for predicting neurological outcomes in patients with out-of-hospital cardiac arrest (OHCA) remain elusive. Hypothesis: We hypothesized that a novel decision tree model using 4 pre-hospital variables can predict neurological outcome after OHCA at the emergency department. Methods: We analyzed 390226 adults in whom resuscitation was attempted after non-traumatic OHCAs in a prospectively recorded nationwide Utstein-style database in Japan over 5 years (2005-2009). The endpoint was 1-month survival with good neurological outcome (cerebral performance category scale, 1-2). We developed a prediction model using data from 4 years (2005-2008, n = 307896), with validation using external data from 2009 ( n = 82330). Results: Overall 1-month survival and good neurological outcome rates were 4.19% and 1.99%, respectively. Multivariate logistic regression analysis indicated the following pre-hospital factors to be significant: shockable initial rhythm (odds ratio [OR] 5.87; 95% confidence interval [CI] 5.23-6.60), emergency medical service (EMS) personnel witnessed (OR 3.40; 95%CI 3.15-3.69), witnessed arrest (OR 3.05; 95%CI 2.83-3.28), and age (OR 3.24; 95%CI 2.97-3.53). Using recursive partitioning analysis with these 4 predictors, we developed a clinical decision tree model: if a witnessed OHCA patient had shockable initial rhythm, the probability of good outcome was 20.0% (ageor=71); if an unwitnessed OHCA patient had shockable initial rhythm, the probability of good outcome was 6.8% (ageor=81); if a witnessed OHCA patient had unshockable initial rhythm, the probability of good outcome was 3.8% (EMS personnel witnessed arrest) or 1.1% (EMS personnel unwitnessed arrest); if an unwitnessed OHCA patient had unshockable initial rhythm, the probability of good outcome was 0.5% (ageor=71). The c-statistics for this model in the groups for development and validation were 0.867 (95%CI 0.861-0.873) and 0.887 (95%CI 0.877-0.896), respectively. Conclusions: Stratification of patients using a decision tree model accurately provides the severity of OHCA classification rules and can predict good neurological outcomes after OHCA at the emergency department.