In this talk, I will introduce the notion of transients in health and demonstrate its significance to early detection through different examples. I will then propose a mathematical methodology for modeling short term epidemiological trends in general and early detection of anomalous or rare events in particular. I will show how historical data can be transformed into a quantity representing the infected pool in a population in order to estimate vital epidemiologic parameters of infection transmission within the population. These parameters are then utilized to detect temporal irregularities in the surveillance data with high sensitivity, specificity and in a timely manner. Finally, I will illustrate the detection power of our methodology under different models of outbreaks of infectious disease applied to real historical surveillance data. Our proposed framework can be potentially applied to a wide spectrum of outbreaks of infectious disease, whether of malicious nature or otherwise (such as West Nile and SARS).
Mojdeh Mohtashemi is currently the director of biomedical informatics at the Boston Heart Foundation, an affiliate of the Harvard/MIT Division of Health Sciences and Technology. She is also a research affiliate of the MIT CSAIL.