An early warning system for overcrowding in the emergency department.

摘要: Overcrowding of emergency departments impedes health care access and quality nationwide. A real-time early warning system for overcrowding may allow administrators to alleviate the problem before reaching a crisis state. Two original probabilistic models - logistic regression recurrent neural network were created predict crises one hour in future. The two pre-existing validated at 8,496 observation points from January 1, 2006 February 28, 2006. All showed high discriminatory ability terms area under receiver operating characteristic curve (logistic = .954; .957; EDWIN .879; NEDOCS .924). At comparable rates false alarms, gave more advance notice than other 62 min; 13 0 min). These results demonstrate feasibility using based on key operational variables anticipate real time.