Using Bayesian Networks to Assist Decision-Making in Syndromic Surveillance

The decision as to whether an alarm (excess activity in syndromic surveillance indicators) leads to an alert (a public health response) is often based on expert knowledge. Expert-based approaches may produce faster results than automated approaches but could be difficult to replicate. Moreover, the effectiveness of a syndromic surveillance system could be compromised in the absence of such experts. Bayesian network structural learning provides a mechanism to identify and represent relations between syndromic indicators, and between these indicators and alerts. Their outputs have the potential to assist decision-makers determine more effectively which alarms are most likely to lead to alerts.