Stochastic methods for epidemiology and biochemical reaction networks

eSMB2020 eSMB2020 Follow Wednesday at 9:30am EDT
Share this


Hye-Won Kang and Wasiur KhudaBukhsh


Stochastic modelling is becoming increasingly popular in biological sciences. The ability to account for intrinsic fluctuations and uncertainty in experimental outcomes has been an advantage of stochastic methods. The application of stochastic tools has proven to be tremendously useful in analyzing biological data. In particular, stochastic methods have found usefulness in studying the spread of infectious diseases, in understanding the biophysics of enzyme kinetics, metabolism, immune-response mechanisms, in constructing phylogenetic trees etc. The objective of this two-part mini-symposium is to highlight some of the recent advances in the closely related areas of stochastic epidemiology and biochemical reaction networks - both at the molecular as well as the ecological scale. The first session will focus on epidemiology, while the second will discuss biochemical reaction networks in broad generality. Both sessions will cover a wide range of themes (including applications and techniques) giving a broad overview of the two fields. Specific topics include new asymptotic results/approximations, multiscale methods and statistical inference algorithms. Network-based approaches to epidemic modelling will also receive attention. With the recent Ebola and COVID-19 outbreaks in West Africa and China respectively, these methods are not only interesting from a theoretical standpoint, but also potentially important for timely public health interventions.

Eben Kenah

The Ohio State University, United States,
"Pairwise survival analysis for measuring and controlling risk in epidemics"
When a disease is transmitted from person to person, infections in different individuals are not independent. These “dependent happenings” cause fundamental problems for standard principles of epidemiologic study design and data analysis in infectious disease epidemiology. Pairwise survival analysis is an extension of standard survival analysis to the transmission of infectious diseases through households, hospitals, or other situations where there is a clearly-defined population at risk of infection. We show how this framework can be used to assess and control for confounding and selection bias. We then show how these methods can be used to extend traditional derivations of case-control and case-cohort designs to obtain novel study designs for outbreak investigations and public health intervention trials for emerging infections. Finally, we consider the possibilities for incorporating pathogen genetic sequences.

Jessica Stockdale

Simon Fraser University, Canada, jessica
"How long does it take to detect a change in COVID-19 control measures?"
Countries around the world have implemented population-wide interventions in efforts to control COVID-19, with varying extent and success. Many jurisdictions are moving to relax measures, while others are re-intensifying them to curb growing spread. But uncertainty remains around the length of time between a population-level change in control measures and its observable impact on detected cases. I will describe our recent work in estimating the time frame for a substantial difference between the cases that occur following a change in control and those that would have occurred under continued strategy, under a compartmental model for disease transmission incorporating physical distancing. Using a likelihood-based approach and data from British Columbia, Canada, we examine how long it takes to detect such a difference given delays and noise in reported cases. We find that these time frames are long: longer than the mean incubation period and the often-used 14 days.

Forrest Crawford

Yale University, United States,
"Causal evaluation of infectious disease interventions using stochastic transmission models"
Deterministic and stochastic models of infectious disease transmission are widely used to understand the dynamics of epidemics, and project the impact of control measures, in human populations. However, most clinical evaluations of vaccines and other interventions designed to prevent infection do not use these models. Instead, clinical infectious disease epidemiologists use randomized trials and statistical regression models to evaluate interventions. Recent work has shown that these approaches may deliver erroneous estimates of the susceptibility effect of the vaccine, even when treatment is randomized or all baseline confounders are measured. In this presentation, we develop approaches to causal evaluation of interventions in networked populations in randomized and observational trials using a flexible semi-parametric class of stochastic transmission models. We show analytically and by simulation that causal susceptibility and infectiousness effects are identified, and that researchers do not need to specify the functional form of some model components in order to make useful inferences. The approach illustrated in an application to evaluation of risk factors for tuberculosis in a large-scale cluster cohort study.

Boseung Choi

Korea University, South Korea,
"Statistical inference for epidemic models using the Survival dynamical system based on the Bayesian approach"
I introduce new methods for Bayesian Markov Chain Monte Carlo-based in references in certain of a stochastic model for epidemic data. SIR (Susceptible-Infected-Removed) model is the classical method for modeling infectious disease spread. In this research, we applied solutions of ordinary differential equations describing the large-population limits of Markovian stochastic epidemic models to individual-level SIR model by introducing survival or cumulative hazard functions derived from population-level equations. We call the method as survival dynamical system (SDS). In this research, we also construct an additional estimation step for initial number of susceptible by utilizing a hierarchical Bayesian approach for inference of the number of trials in the Binomial distribution. we applied the SDS approach to data from a 2009 influenza A(H1N1) outbreak at Washington State University.

Hosted by eSMB2020 Follow
Virtual conference of the Society for Mathematical Biology, 2020.