Working Paper Cnr-Ceris, N° 12/2014

Artificial Neural Networks and risk stratification in Emergency department

Casagranda I.[1], Costantino G.[2], Falavigna G.[3]*, Furlan R.[4], Ippoliti R.[5]

^* Corresponding  author:  Indirizzo e-mail protetto dal bots spam , deve abilitare Javascript per vederlo

                                                        011-6824.941

Abstract: The primary goal of the Emergency Department physician is to discriminate individuals at low risk, who can be safely discharged, from patients at high risk, who deserve prompt hospitalization for monitoring and/or appropriate treatment. Obviously, the problem of a correct classification of patients, and the successive hospital admission, is not only a clinical issue but also a management one since ameliorating the rate of admission of patients in the emergency departments could dramatically reduce costs and create a better health resource use.

Considering patients at the emergency departments after an event of syncope, this work propose a comparative analysis between multivariate logistic regression model and Artificial Neural Networks (ANNs), highlighting the difference in correct classification of severe outcome at 10 days and 1 year. According to results, ANNs can be very effective in classifying the risk of severe outcomes and it might be adopted to support the physician decision making process reducing, at least theoretically, the inappropriate admission of patients after syncope event.

keywords: Artificial Neural Networks (ANNs); Syncope; Emergency Departments; Risk stratification; Area Under the Curve, referring to the Receiver Operating Characteristics (ROC) Curve; correct classification;

JEL Codes: I12; D81;

The preliminary results of this paper have been presented at the workshop “Progettare per innovare: le reti” (Azienda Ospedaliera “SS. Antonio e Biagio e Cesare Arrigo” of Alessandria – 18^th July, 2014).

download