Optimal surveillance network design: a value of information model

Abstract

Abstract Purpose Infectious diseases are the second leading cause of deaths worldwide, accounting for 15 million deaths – that is more than 25% of all deaths – each year. Food plays a crucial role, contributing to 1.5 million deaths, most of which are children, through foodborne diarrheal disease alone. Thus, the ability to timely detect outbreak pathways via high-efficiency surveillance system is essential to the physical and social well being of populations. For this purpose, a traceability model inspired by wavepattern recognition models to detect “zero-patient” areas based on outbreak spread is proposed. Methods Model effectiveness is assessed for data from the 2010 Cholera epidemic in Cameroon, the 2012 foodborne Salmonella epidemic in USA, and the 2004-2007 H5N1 avian influenza pandemic. Previous models are complemented by the introduction of an optimal selection algorithm of surveillance networks based on the Value of Information (VoI) of reporting nodes that are subnetworks of mobility networks in which people, food, and species move. The surveillance network is considered the response variable to be determined in maximizing the accuracy of outbreak source detections while minimizing detection error. Surveillance network topologies are selected by considering their integrated network resilience expressing the rewiring probability that is related to the ability to report outbreak information even in case of network destruction or missing information. Results Independently of the outbreak epidemiology, the maximization of the VoI leads to a minimum increase in accuracy of 40% compared to the random surveillance model. Such accuracy is accompanied by an average reduction of 25% in required surveillance nodes with respect to random surveillance. Accuracy in systems diagnosis increases when system syndromic signs are the most informative in a way they reveal linkages between outbreak patterns and network transmission processes. Conclusions The model developed is extremely useful for the optimization of surveillance networks to drastically reduce the burden of food-borne and other infectious diseases. The model can be the framework of a cyber-technology that governments and industries can utilize in a real-time manner to avoid catastrophic and costly health and economic outcomes. Further applications are envisioned for chronic diseases, socially communicable diseases, biodefense and other detection related problems at different scales.