EpiWatch - an open source observatory for early warning and risk analysis of biothreats and other hazards
MacIntyre, C Raina;
Joshi, Aditya
;
Adam, Dillon C;
Sparks, Ross;
Paris, Cecile
;
Karimi, Sarvnaz;
Veenstra, Patrick;
Lesmanawati, Shinta;
Suleiman, Feroza;
Heslop, David
2019-11-18
Conference Material
Chemical and Biological Defense Science & Technology (CBD S&T) Conference, Cinncinnati, Ohio, 18 November 2019
1
Background: EpiWatch, an automated epidemic observatory, has been developed between 2016-2019, utilising open source data such as social media and news feeds. Rapid epidemic intelligence can provide surveillance capability in countries with limited surveillance capacity, and can overcome lack of official reporting. A rapid signal from a low-income country (which may otherwise not be detected until very late) can forewarn deployed troops of potential biothreats. Aims: To develop an automated epidemic observatory and risk analysis tools using open source data, to improve surveillance capability, prioritise response, provide early signals for biological attacks and serious epidemics. Methods: An open source epidemic observatory, EpiWatch was developed as a management web application, enhanced by machine learning and has collected data since 2016. The database is curated, cleaned and enhanced by weekly review as new data is collected. Language-specific intelligence was developed for the Asia-Pacific region. We also developed EpiRisk, a tool that predicts risk of epidemics using automated data feeds to inform multiple risk parameters. Two case studies for rapid warning: the thunderstorm asthma outbreak in Melbourne in 2016, and the Ebola epidemic in West Africa in 2014. In the case of the former, we use an internal repository of tweets posted in Australia from 2014 to 2016. We devise a four-step computational architecture that applies natural language processing techniques and statistical monitoring. For the Ebola outbreak, we used tweets containing word algorithms from West Africa from 2011 to 2014. Results: EpiWatch has over 8,000 outbreak reports from 2016 onward that can be searched on disease, date, location and other key words. We found that outbreaks in countries speaking Malay and Indonesian languages are missed by engines such as HealthMap, because of specific local terminology. In the case of an acute asthma event where 8000 people were hospitalised in one night, we generated a disease signal nine hours before the official reported time. For the 2014 Ebola epidemic, we detected a disease signal three months before the official reports to WHO in March 2014. The EpiRisk tool provides multi-parameter risk forecasting for emerging disease events. Discussion: Open source data is a powerful source of rapid intelligence. It can be used to rapidly flag epidemics or biowarfare attacks and to identify clusters of key syndromes resulting from an emerging infection or biowarfare event. Open source data can provide more rapid signals than traditional disease surveillance, and can overcome the limitations of poor official reporting or lack of human resources in austere environments. Even in West Africa, with lower internet penetration than the global average we were able to identify a signal for the epidemic prior to official recognition. The methodology also captures acute, non-infectious threats such as sudden respiratory distress. EpiRisk can avert catastrophic epidemics by providing early risk analysis. A range of tools for early warning, risk prediction, forecasting and decision support can also be built on our platform, and the outputs can be transferred to mobile applications for use in the field by deployed troops.
Defense Threat Reduction Agency (DTRA)
Health Informatics; Natural Language Processing
EP194432
Conference Abstract
English
MacIntyre, C Raina; Joshi, Aditya; Adam, Dillon C; Sparks, Ross; Paris, Cecile; Karimi, Sarvnaz; Veenstra, Patrick; Lesmanawati, Shinta; Suleiman, Feroza; Heslop, David. EpiWatch - an open source observatory for early warning and risk analysis of biothreats and other hazards. In: Chemical and Biological Defense Science & Technology (CBD S&T) Conference; 18 November 2019; Cinncinnati, Ohio. Defense Threat Reduction Agency (DTRA); 2019. 1. csiro:EP194432. http://hdl.handle.net/102.100.100/251242?index=1
Loading citation data...
Citation counts
(Requires subscription to view)