Penn State University
Outbreaks of emerging pathogens are among the most unpredictable threats to health and personal welfare. The causes of pathogen emergence are variable and seemingly idiosyncratic. Although forecasting infectious disease emergence (and re-emergence) therefore seems intractable, new developments in epidemic modeling show that emergence events are united by patterns common among dynamical systems that cross tipping points. Particularly, it has recently been shown that tipping points in the transmission of infectious diseases influence epidemiological patterns in a way that leaves signatures" of their proximity. For instance, a statistical phenomenon called critical slowing down influences the variance and autocorrelation of data streams in characteristic ways. The long term objective of this study is to develop a theory for forecasting epidemic transitions before they occur, focusing on critical slowing down and other near-critical statistical patterns. Its specific aims are: (1) T develop theories of forecastability for emerging and eliminable infectious diseases; (2) to develop model-independent statistical methods for forecasting disease emergence based on this theory; (3) to document critical slowing down and other near-critical phenomena by comparing data on re-emerging and non-emerging pathogens; and (4) to produce software packages implementing these methods. The research strategy adopted by this project combines theoretical studies with data analysis in an empirically- driven feedback loop. The research design requires first the development of mathematical models of intermediate complexity allowing both mathematical solution and dynamic realism to represent contemporary emerging and re-emerging infectious diseases. Solutions of these models provide the basis for statistical algorithms for detecting early warning patterns in epidemiological data. These algorithms will be stress-tested" using computer simulations of emergence events, including both simple and complex scenarios for the conditions under which emergence occurs and the imperfect surveillance and reporting systems that generate data. The algorithms will then be applied to data on recent emergence and re-emergence of measles, pertussis, and rubella in populations from which these pathogens had been nearly eliminated; lessons learned from this empirical study will inform subsequent rounds of model development, simulation and validation. Finally, the results of these studies will be brought together in a range of software applications for use i research, policy making, and education.