BACKGROUND: The utility of risk stratification following an emergency medical admission has been debated. We have examined the predictability of outcomes, from a database of all emergency admissions to St James' Hospital, Dublin, over a six year period (2005-2010). METHODS: Analysis was performed using the hospital in-patient enquiry system, linked to the patient administration system and laboratory data. The utility of a fractional polynomial laboratory only model to predict 30-day in-hospital mortality was determined. RESULTS: The AUROC for the laboratory parameters to predict a 30 day death was 0.90 ( 95% CI 0.89, 0.90) in the 2002 - 2010 derivation dataset and was 0.88 (95% CI 0.86, 0.90) in the 2011 validation set. The addition of co-morbidity measures did not improve the model prediction (0.89 : 95% CI 0.88 - 0.89). CONCLUSION: A fractional polynomial laboratory only model can reliably predict 30-day hospital mortality following an emergency medical admission, potentially allowing resources to be risk focused and patients to be prioritised.
BACKGROUND: The utility of risk stratification following an emergency medical admission has been debated. We have examined the predictability of outcomes, from a database of all emergency admissions to St James' Hospital, Dublin, over a six year period (2005-2010). METHODS: Analysis was performed using the hospital in-patient enquiry system, linked to the patient administration system and laboratory data. The utility of a fractional polynomial laboratory only model to predict 30-day in-hospital mortality was determined. RESULTS: The AUROC for the laboratory parameters to predict a 30 day death was 0.90 ( 95% CI 0.89, 0.90) in the 2002 - 2010 derivation dataset and was 0.88 (95% CI 0.86, 0.90) in the 2011 validation set. The addition of co-morbidity measures did not improve the model prediction (0.89 : 95% CI 0.88 - 0.89). CONCLUSION: A fractional polynomial laboratory only model can reliably predict 30-day hospital mortality following an emergency medical admission, potentially allowing resources to be risk focused and patients to be prioritised.
Authors: Mohammed A Mohammed; Gavin Rudge; Duncan Watson; Gordon Wood; Gary B Smith; David R Prytherch; Alan Girling; Andrew Stevens Journal: PLoS One Date: 2013-05-29 Impact factor: 3.240