The prediction value of Glasgow coma scale-pupils score in neurocritical patients: a retrospective study.

BACKGROUND: External validation is necessary before  its clinical recommendation in new setting. The aim is to externally validate Glasgow Coma Scale-pupils score (GCS-P) in neurocritical patients and to compare its performances with Glasgow Coma Scale (GCS) and its derivatives. METHODS: GCS-P at admission was calculated for individual based on the model developed by Brennan et al. Area under the receiver operating characteristic curves (AUCs), Nagelkerke's R2 and Brier scores were used to assess external validity of GCS-P to predict mortality in neurocritical patients and to compare predictive performance with GCS and its derivatives. SUBJECTS: 4372 neurocritical patients from intensive care units of Beth Israel Deaconess Medical Center, United States between 2001 and 2012. RESULTS: GCS-P showed good discrimination (AUC 0.847 for in-hospital mortality and 0.774 for ninety-day mortality), modest calibration (Nagelkerke's R2 33.1% for in-hospital mortality and 23.3% for ninety-day mortality). Predictive performances of GCS and its derivatives was inferior to GCS-P. CONCLUSIONS: GCP-P discriminated well in between death in neurocritical patients. GCP-P improved predictive performance for short-term mortality over GCS and its derivatives in neurocritical patients. It would be a simple, early and reasonable daily routine option for prognosis assessment in neurocritical setting.