Background The (Development and validation of risk-adjusted outcomes for systems of emergency care) DAVROS study aims to develop a method to predict the probability of mortality for patients admitted to hospital as a medical emergency. This would facilitate evaluation of the performance of emergency care systems through comparison of actual to predicted mortality.

Methods We collected data on presenting physiology, demographic data, co-morbidities, blood tests and mortality from the ambulance patient report form, emergency department notes and hospital computer records of 11 528 emergency medical admissions to three hospitals in South Yorkshire. These data are used to develop risk-adjustment models to predict 7-day and 30-day mortality.

Results Mortality was 630/11 528 (5.5%) at 7 days and 1068/11 528 (9.3%) at 30 days. Univariate analysis has identified predictor variables associated with 7 and 30-day mortality and multivariable analysis has identified independent predictor variables. Preliminary analysis suggests a model based on demographics, co-morbidities and International Classification of Diseases code has a c-statistic of 0.766 for 7-day mortality and 0.776 for 30-day mortality. The addition of physiological parameters increases these values to 0.852 and 0.835, respectively. We will present the DAVROS risk-prediction model for 7 and 30-day mortality, derived on half the cohort and validated on the other half, with estimates of discriminant value (c-statistic) and calibration (Hosmer-Lemeshow).

Conclusion A risk-adjustment model based on demographic data, co-morbidities and baseline physiology can predict 7 and 30-day mortality in emergency medical admissions with acceptable discriminatory power.