PRISM is a novel capability developed to predict the outbreak of infectious diseases. It has shown promising results indicating it is possible to accurately predict impending trends for important diseases such as dengue and malaria in certain parts of the world (e.g. Peru, Philippines, South Korea). PRISM utilizes automated data mining techniques to extract rules from clinical and environmental data and builds the prediction model. Outputs from the model predict high or low disease incidence in regions of interest; thereby, enhancing awareness and providing critical lead times for taking preemptive public health actions [Buczak et al., 2012].
» PRISM Press Release
NOTE: The PRISM graphic user interface (GUI) uses Silverlight. The Microsoft Silverlight plugin must be installed/enabled in order for the GUI to work.
The prediction method uses types of data selected from studies in the literature that have shown significant correlation with malaria incidence. Accordingly, data types include human incidence information and factors affecting vector proliferation. Of note is that while predictive capability is enhanced in the presence of strong disease indicator data sources, in locations where human disease data in real-time may be unavailable, the PRISM model is able to use less frequent incidence aggregates and leverage a variety of publicly available climate and environmental information to produce operationally meaningful results. Typical data types used by the model include disease incidence; remotely sensed meteorological data (i.e., land surface temperature, rainfall); remotely sensed vegetation data; climate data (sea surface temperature anomalies; Southern Oscillation Index); and socio-economic data (population, sanitation) [Buczak et al., 2012][Buczak et al., 2014].
In previous collaborative studies conducted by JHU/APL, the rule mining methodology used by PRISM demonstrated good performance in predicting dengue fever in Peru [Buczak et al., 2012]. For the Philippines provinces, the model predicted dengue four weeks in advance with high accuracy [Buczak et al., 2014].
United States Forces Korea (USFK) requested the predictions to be made with a higher geographical resolution than a province and further in time than 4 weeks. The present predictions are performed at a region level (64 regions that cover the four northern provinces of South Korea). The predictions are for a 2 week period which is week 7 and week 8 in the future.
The predictions for Korea have three levels: low (0-2 cases), medium (3-16) and high (17 or more). The number of cases predicted is the aggregate for the two week prediction period.
AL. Buczak, PT. Koshute, SM. Babin, BH. Feighner, SH. Lewis, "A Data-driven Epidemiological Prediction Method for Dengue Outbreaks Using Local and Remote Sensing Data", BMC Medical Informatics and Decision Making, 12(1):124 (2012).
AL. Buczak, B. Baugher, SM. Babin, LC. Ramac-Thomas, E. Guven, Y. Elbert, PT. Koshute, JMS. Velasco, VG. Roque, EA. Tayag, IK. Yoon, SH. Lewis, "Prediction of High Incidence of Dengue in the Philippines", PLOS Neglected Tropical Diseases, 8(4): e2771. doi: 10.1371/journal.pntd.0002771 (2014).
© 2014 The Johns Hopkins University Applied Physics Laboratory LLC. All rights reserved.