A main challenge for local healthcare authorities in the ongoing COVID-19pandemic is tracking and tracing infections and high risk exposure contacts . Accurate analysis ofthe corresponding data can improve decisions on adapting social restrictions and declaring quarantinesto stop the pandemic from spreading. Early detection of outbreaks are vital for a comprehensivepublic health strategy, especially with the median serial interval for COVID-19 being shorter thanthe incubation period , i.e. people being infectious before developing symptoms in contrast tomost other infectious diseases . Contact tracing data can be represented as a temporal graph. Wepropose a framework consisting of temporal graph analysis methods detecting outbreak patterns inthese networks to help the local healthcare authorities manage the situation by identifying personsmost at risk of spreading the virus further. We assess our framework on a real world dataset oncontact tracing including more than 10k persons.
We collaborate with Germany’s largest local healthcare authority inCologne, covering a population of over a million. The data providedresults in a temporal graph consisting of nodes (positively testedcases and their contacts) and temporal edges (infection and exposureevents) which changes over time as more people get infected andreport contacts, that had a high risk exposure as seen in Fig. 1.