Limited Prognostic Utility of a Simplified Vital Sign Based Risk Calculator in Acute Medical Admissions

Abstract

Background: Accurate efficient prognostication in acute medical admissions remains challenging.Methods: We constructed a Vital Sign based Risk Calculator using vital parameters and Major Disease Categories to predict 30-day in-hospital mortality using a multivariable fractional polynomial model. Results: We evaluated 113,807 admissions in 58,126 patients. The Vital Sign based Risk Calculator predicted 30-day inhospital mortality to increase from 2 points – 3.6% (95%CI 3.4, 3.7) to 12 points – 14.8% (95%CI 14.0, 15.7). AUROC was 0.74 (95%CI 0.72, 0.74). The addition of illness severity and comorbidity data improved AUROC to 0.90 (95%CI 0.89, 0.90). Conclusion: The Vital Sign based Risk Calculator is limited by its simplicity; inclusion of illness severity and comorbidity data improve prediction.