Informatics methods in Infection and Syndromic Surveillance

Karin Verspoor1,2

1School of Computing and Information Systems, The University of Melbourne, Melbourne VIC 3010;

2Health and Biomedical Informatics Centre, The University of Melbourne, Melbourne VIC 3010


The availability of electronic health records creates opportunities to use computational algorithms for analysis, prediction, and automation of tasks in the healthcare context. In this presentation, I will introduce two examples of applications for surveillance that hinge on computational analysis of health texts, including monitoring of emergency department triage notes for outbreaks and classification of radiology reports as positive for invasive fungal disease. Health text data poses a challenge to traditional analytical approaches due to its unstructured nature, variability, and noisiness; I will describe how we use Natural Language Processing (NLP) techniques to extract actionable information and build effective predictive models.

In addition, I will describe the application of machine learning techniques to predict probability of Invasive Aspergillosis in high-risk haematology patients. Large amounts of data are collected during the treatment of these patients which can be leveraged to produce more accurate predictions of Invasive Aspergillosis diagnosis. The inferred model demonstrates high negative predictive value, ruling out IA infection for some patients. In these cases, antifungal treatment may be safely avoided, minimising over-treatment and drug toxicity, and reducing associated costs. This method can therefore enhance the interpretation of biomarker results.

Developing methods in the context of real-time surveillance tasks has value for public health and health services performance monitoring, but also has the potential to impact clinical decision making and strengthen antimicrobial or antifungal stewardship.