Prediction rules for estimating neurologic outcome following out-of-hospital cardiac arrest.

BACKGROUND: No valid model has been developed to predict survival following out-of-hospital cardiac arrest. The purpose of this study was to develop a prediction model for meaningful survival following out-of-hospital cardiac arrest using variables available during resuscitation.
METHODS: This was a retrospective cohort study. Consecutive adult cardiac arrest patients were studied between 1994 and 2001. Variables included age, sex, race/ethnicity, arrest location, whether the arrest was witnessed, initial rhythm, whether CPR was performed, patient downtime, paramedic response time, survival to hospital discharge, and Glasgow Coma Score (GCS) at hospital discharge. Classification and Regression Tree analysis was used to develop decision rules to predict meaningful survival, as defined by the patient's discharge GCS.
RESULTS: Of the 754 patients, 16 (2%) survived with a GCS > or =13, 15 (2%) survived with a GCS = 14, and 5 (0.7%) survived with a GCS = 15. The decision rule for survival with a GCS > or = 13 incorporated whether the arrest was witnessed and the patient's age, resulting in a negative predictive value (NPV) of 99.8%. The rule for survival with a GCS > or = 14 incorporated the initial arrest rhythm, whether the arrest was witnessed, and the patient's age, resulting in a NPV of 99.6%. The rule for survival with a GCS = 15 incorporated only the interval between collapse and the initiation of life support, resulting in a NPV of 99.8%.
CONCLUSIONS: This study reports decision rules for potential meaningful survival following out-of-hospital cardiac arrest with high NPVs for each. Future studies need to be performed to prospectively validate these models.