IPBN: alerts management in intravenous electromedical devices using bayesian networks

The incidence of false alerts in the hospital environment compromises the tasks of healthcare professionals, since they: (i) stress the teams of caregivers, and the patients themselves; (ii) may lead to an increase in the length of hospitalization; (iii) may cause risky situations, some of which with severe health implications for the patients. Considering this scenario, the objectives of this research are to discuss the occurrence of alerts in intravenous systems and to contribute to the reduction of the emission of false alerts in electromedical devices, more particularly in infusion pumps. The proposed system is called Infusion Pumps alerts management exploring Bayesian Networks (IPBN) which explores the use of Bayesian Networks to minimize the occurrence of false alerts when there is a change in the flow rate caused by occlusions in infusion pumps. Infusion line occlusion is the procedure with the highest rate of false alerts in this type of device. The results achieved with the IPBN are promising, reaching 85 % of accuracy, based on data from actual infusion pumps. This means that the false alerts rate is reduce to 15 % which is very competitive with commertial solutions where this rate can be as higher as 90 %.