Select Page

Modeling an Asymptomatic Epidemic

By Rob Mitchum // October 16, 2017

On the World Health Organization’s target list for eradicating disease, hepatitis C is currently among the most wanted. An estimated 71 million people live with the viral liver disease globally, and 1.75 million new infections occur every year. Yet there is reason for optimism, as new treatments, preventative measures, and, perhaps soon, vaccines create novel strategies for driving down infection. By 2030, the WHO hopes to cut new infections by 90%, and decrease viral hepatitis mortality by two-thirds.

Meeting these ambitious global goals will require finding effective interventions at the local level — a difficult task given a hard-to-reach high-risk population (people who inject drugs), the asymptomatic early stages of the disease, and the multitude of interventions available. An innovative new partnership between the Computation Institute/Argonne National Laboratory, Loyola University Medical Center (LUMC), and the University of Illinois at Chicago (UIC) will combine rich epidemiological and biological data with agent-based modeling to test prevention and treatment strategies in silico, simulating the activity of 32,000 people who inject drugs in the Chicagoland area.

Agent-based modeling can simulate complex social behavior by creating thousands of individual “agents” representing people, programmed to reflect actual census and research data and behave realistically. The Social and Behavioral Systems Group at Argonne National Laboratory built a model of Chicago and its citizens called chiSIM, developed with RepastHPC, part of Argonne’s open-source Repast ABM suite, to conduct large-scale studies of MRSA, the spread of healthcare information, and even zombie invasions. In the new study, funded by a $2 million grant from the National Institutes of Health, they’ll adapt that model to study the demographics, geography, and social networks of the drug injecting population at risk of hepatitis C and how best to stop the disease.

“There are many hypotheses that could be tested via pilot programs or randomized controlled trials on this disease, but they are restricted by ethics, cost, and other factors,” said Jonathan Ozik, CI senior fellow and computational scientist at Argonne. “The intervention policy space can be very large, and yet we can strategically characterize this space using our model exploration techniques and our big computers. The models we produce can then be used to guide the design of trials that are more likely to be successful; you can fine tune them before taking them to the field.”

Co- principal investigators Harel Dahari, assistant professor and co-direct the Program for Experimental & Theoretical Modeling, department of medicine, division of hepatology at LUMC, and Basmattee Boodram, research assistant professor of epidemiology and biostatistics at UIC, previously constructed a smaller agent-based model of Chicago’s drug-injecting population also using Repast. Their study, published in 2015, predicted that hepatitis C infections would rise most sharply among younger, suburban people who inject drugs (PWID) and those that are not enrolled in harm reduction programs.

The new model, called HepCEP, will expand this work by simulating the full population of 32,000 injected-drug users in Chicago and the suburbs, including factors such as daily activities that drive risk, the geography of where drug users and networks interact, and characteristics of hepatitis C such as disease progression and transmission probability. This multi-scale approach, based on data from several large-scale epidemiological studies of the Chicago PWID population, will provide the most realistic model ever constructed of how hepatitis C spreads.

“We will try to actually model the interactions in space and time, where people get together and spread infection,” said Dahari. “We will also account for the state of their disease, because if you have a person with a high viral load and they share needles with someone, there’s a high probability it will transmit the virus. We want to examine how it’s important to take care and control the viral load in each person in order to avoid transmission.”

Different versions of this model will be compared with past data to find parameterizations that best reflect real world dynamics, calibrated using the Extreme-scale Model Exploration with Swift (EMEWS) framework. Researchers can then use the best models to forecast future changes in hepatitis C infections among different PWID populations, study how different networks of injected-drug users form and evolve, and identify potential targets for intervention. Then, those interventions themselves — such as placing new needle exchange programs in areas of high drug activity, or focusing treatment on individuals that raise infection risk for an entire network — can be simulated to compare their efficacy and cost-effectiveness.

“From a network perspective, individuals in the middle, the crossovers, are bridges between areas of high (urban) and low (suburban) HCV prevalence rates,” Boodram said. “The empirical work shows that these individuals should be a particular focus for intervention, but the question is, who are they? How can we use the entire model to predict a little bit more about them? Then we can evaluate, if we just address this small population, what impact an intervention would have on transmission?”

Beyond these questions, the HepCEP team also hopes to expand the model to study other areas and populations in the United States, and use it to help researchers developing hepatitis C vaccines to design and evaluate clinical trials. The model could also be useful for studying the spread of other bloodborne diseases, or for testing addiction treatment and overdose prevention strategies.