A Citizen Observatory Approach for Developing a Disease Outbreak Early Warning System

Abstract

In health matters, early warning systems are timely surveillance systems that collect information on epidemic-prone diseases to trigger prompt public health interventions. However, these systems rarely apply statistical methods to detect changes in trends or sentinel events that would require intervention. Often, they rely on an in-depth review done by epidemiologists of the data coming in, which is rarely done systematically. This research introduced the use of ICT for collecting and analyzing citizen observations on disease trends and outbreaks. A citizen observatory ICT tool, which utilizes mobile and web features was developed. Data was collected on symptoms observed from diseases in four locations within Nairobi city. The system made use of mathematical models and outlier detection techniques to detect observations that deviated from the expected pattern in the dataset. New clusters were considered as outliers and the system flagged them as potential outbreaks. We clustered data using a K-Means algorithm and the Euclidean distance of each object from its corresponding cluster centre was obtained. From the results, the developed prototype was able to detect an outbreak of Flu and URTI diseases for the period of study. The proposed tool can therefore enhance the management of risks associated with disease outbreaks.

 Keywords: Early Warning Systems, citizen observatory, health surveillance, Outlier Detection, Modeling disease outbreak,