ABSTRACT
PURPOSE: To use classification tree analysis to identify risk factors for nonsurvival in a neurological patients with subarachnoid haemorrhage (SAH) and to propose a clinical model for predicting of mortality. METHODS: Prospective study of SAH admitted to a Critical Care Department and Stroke Unit over a 2-year period. Middle region of pro-ADM plasma levels (MR-proADM) was measured in EDTA plasma within the first 24 hours of hospital admission using the automatic immunofluorescence test. A regression tree was made to identify prognostic models for the development of mortality at 90 days. RESULTS: Ninety patients were included. The mean MR-proADM plasma value in the samples analysed was 0.78 ± 0.41 nmol/L. MR-proADM plasma levels were significantly associated with mortality at 90 days (1.05 ± 0.51 nmol/L vs 0.64 ± 0.25 nmol/L; P < .001). Regression tree analysis provided an algorithm based on the combined use of clinical variables and one biomarker allowing accurate mortality discrimination of three distinct subgroups with high risk of 90-day mortality ranged from 75% to 100% (AUC 0.9; 95% CI 0.83-0.98). CONCLUSIONS: The study established a model (APACHE II, MR-proADM and Hunt&Hess) to predict fatal outcomes in patients with SAH. The proposed decision-making algorithm may help identify patients with a high risk of mortality.
