Data fusion for identification of serious illness in children
Children presenting to an emergency care provider, such as an out-of-hours GP or Emergency Department, frequently have minor, self-limiting infections. However, a few will have a serious infection or complication, requiring urgent investigation and treatment to reduce the risk of long-term disability or death. Vital signs can be measured quickly and non-invasively on these children. However, no single sign is sufficiently predictive to accurately identify the children at risk. This is because signs may be physiologically coupled (e.g. heart rate and temperature), and also because the normal range of some signs, such as the breathing rate and heart rate, change with the age of the child. We propose a data fusion technique to identify children with serious illness from a combination of non-invasive vital signs and the age of the child. The technique has been applied to a dataset collected in primary care situations, and has shown that sensitivities and specificities of over 70% can be achieved.