Report From the Models of Infectious Disease Agent Study (MIDAS) Steering Committee

 

On May 1, 2004, the Models of Infectious Disease Agent Study was officially launched when NIGMS awarded four cooperative agreement grants to three research groups and one informatics group.  These four projects, called the MIDAS Network, are charged with conducting research on computational and mathematical models of emerging infectious diseases and creating models that will assist policymakers and public health officials in making decisions about the management of outbreaks.  The effort is guided by the MIDAS Steering Committee, whose membership consists of the four principal investigators, NIH representatives, and several outside experts.   In keeping with the sense of urgency and excitement that MIDAS engenders, the Steering Committee held its first meeting two days after the awards were made.

 

The goals of the meeting were threefold:

 

Ultimately, MIDAS will gain credibility by building long-term relationships with other agencies and with the public health community.  Involving the public health sector in planning and modeling will produce better products.  As the Steering Committee pointed out, these links remain to be developed.

 

 

MIDAS is a partnership between the National Institutes of Health and the scientific community to develop a valuable resource for policymakers and public health officials.  MIDAS is a multidisciplinary network of scientists conducting computational and mathematical research to improve the ability to detect, control, and prevent emerging infectious diseases.  It does so by developing  computational tools and models of emerging infectious diseases, building models in response to real or anticipated outbreaks, and making those tools and models widely available to policymakers, public health professionals, and researchers.

 

MIDAS uses existing or simulated data, which must, except under exceptional circumstances, be widely available through the MIDAS portal.  All MIDAS products will be available to the scientific community.

 

A unique provision of MIDAS is that in the event of an infectious disease outbreak, NIGMS may redirect the MIDAS network to conduct research to address questions, scenarios, or policy decisions for that specific outbreak. 

 

 

The computing and informatics component of MIDAS is critical to success.  RTI partners with IBM to provide MIDAS infrastructure support at its e-hosting facility and with SAS to provide current statistical tools for model analysis and verification.  Consultants from Duke University and Emory University contribute expertise in infectious disease modeling and bioinformatics.  The informatics group will increase the computational capacity of the MIDAS research groups; identify, incorporate and document data to be used in the development and evaluation of MIDAS models; develop and use informatics tools to identify novel patterns and facilitate parameter estimation for models; identify, incorporate, enhance, and document relevant models from MIDAS research groups and other sources; and develop simulated data sets and analytic tools to evaluate the models.  RTI will also build the MIDAS portal, providing access to databases, analytic and statistical tools, simulation models and datasets, and access to pre-set models. 

 

The team consists of multidisciplinary scientists from the Brookings Institution, the University of Maryland, the University of Pittsburgh, NASA, and the Imperial College of London.  The collaborating investigators have expertise in infectious diseases, epidemiology, ecology, biostatistics, time series analysis, non-linear dynamics, network theory, and decision theory.  Using an agent-based approach, the team will develop and evaluate software for analysis and modeling of epidemiological data.  The group also has experience collecting historical datasets  and working with simulated data and is interested in how to dock real data with simulations.  The JHU team has had some experience in modeling intervention scenarios (e.g., the impact of mass vaccination versus vaccination of first responders). 

 

This project will design, build, and validate models of disease epidemiology and response systems based on individuals' roles in a social network. Its basis is a program called EpiSims, built on a program for studying transportation networks.  EpiSims takes advantage of a large amount of information about the interactions of individuals in a large urban setting  to develop mathematical and computational methods for modeling transmission dynamics, mode of introduction, initial health state of the population, and response strategies.  Research will address what features of social networks best characterize their response to epidemics and mitigation of effects and how to estimate parameters and their variation in populations of different sizes.

 

 

The Emory group is developing stochastic epidemic simulation models for structured American communities in which individuals mix in households, neighborhoods, preschool groups, schools, work places, hospitals, and other settings. The models provide a means to evaluate the effectiveness of interventions  such as surveillance and containment, vaccination, antimicrobials, and closing of key institutions.  Using  stochastic optimization methods permits the researchers to integrate models of the use scarce resources.  Epidemic simulation models will help to determine the important parameters for infection transmission.  This information will be applied to designing field studies and intervention studies for infectious diseases, including the appropriate use and of statistical methods.

 

A second research area is the development of models of within-host dynamics of pathogens which cause acute infections in vertebrates. The goal is construction of general exploratory models of the interplay between pathogen and host immune response during infection.

 

 

 

The Steering Committee discussed the importance of the Network’s beginning immediately to study a problem as a group.  Doing so would quickly develop resources, strengthen collaboration within the Network, and bring to light issues that need to be resolved.  The research group’s expertise lies in developing agent-based models, recently applied to dynamics of smallpox outbreaks and intervention strategies. 

 

Smallpox is a frightening disease, especially when considered as a bioterrorist threat;  however, influenza can cause more deaths worldwide over a very short time.  The 1918 influenza pandemic killed tens of millions of people worldwide and at least 675,000 people in the United States.  One-fifth of the world’s population was infected.  Of the U.S. soldiers who died in World War I in Europe, over half succumbed to influenza, rather than to combat.  The 1918 influenza pandemic has been cited as the most devastating epidemic in recorded history.

 

Most knowledgeable influenza experts have argued that another outbreak of pandemic influenza is almost certain to occur at some time.  When it happens, we need to know more about effective responses, distribution of resources, and long-term consequences of various strategies.   For these reasons, the MIDAS Steering Committee selected pandemic influenza resulting from antigenic shift as the first emerging disease to model.

 

The first step is to identify clear questions.  This must be done in collaboration with influenza experts who know the biology, demography, and public health issues in detail and can provide reality checks for the modelers.  The MIDAS Steering Committee requested that NIGMS organize a meeting of experts in the late summer or early fall.

 

The influenza project has two objectives – (1) to replicate local and global spread of known outbreaks (including the 1918 pandemic) in the absence of interventions and (2) to create scenarios exploring the efficacy of various interventions to slow or halt local and global spread of the outbreak.

 

To accomplish this task, the MIDAS Network will need the following:

 

The research groups have already generated logic models and software for single communities.  JHU has experimented with simple 2-community models, and Emory has published on global flu movements;  however, existing models will need to be validated and expanded to accommodate influenza.  The models must incorporate multiple years of data and antigenic shift/drift. They should include individuals’ previous exposures to the influenza virus and their vaccination status and the consequent effect on susceptibility and immunity.

 

The Network needs to estimate numerous parameters, such as duration of different stages of influenza illness,  household attack rates, and transmissibility parameters, from literature.  Research groups will need ongoing expert advice to ensure that the parameters make biological sense.

 

LANL has done considerable research on within-city social networks, and the Network will need more information on other social networks and contact frequencies. It will also need information on intercity transportation networks, airline schedules and policies, as well as customs and quarantine information. 

 

The Network will need retrospective outcome data (e.g., mortality, flu strain data, sentinel physician networks, syndromic surveillance data) by time and place to help develop and validate the models and to fine-tune parameter estimates. 

 

Considerable computational power (Big Iron) is needed to run models with many different combinations of parameters, with the same parameters many times, and for large networks.

 

In the process of developing influenza models, the Network will seek to identify general principles and create useful tools and models that can be extended to other situations.  The Network will also seek to understand how to validate models, analyze their statistical properties, and identify metrics for evaluating models.  The Steering Committee also recommended that the Network create materials to explain the project to public health officials and policymakers.

 

 

 

The Steering Committee set the following specific milestones for MIDAS.   

 

Deadline	Task	Responsible Group
May 2004	Complete social network of Portland, OR	LANL
May 2004	Create password-protected Web FTP site Post descriptions of research group models Post the computational needs for models	RTI
May 2004	Hold data liaison/programmers meeting	RTI
May 2004	Post hardware and storage capacity and list of supported software	RTI
June 2004	Organize flu meeting for late summer/fall	NIGMS
June 2004	Schedule next year's meetings and conference calls	NIGMS
July 2004	Hold Network meeting for all key personnel	All
August 2004	Establish research groups portal	RTI
August 2004	Port models to RTI	Research Groups
November 2004	Complete social network of Chicago, IL	LANL
November 2004	Create data standards and exchange formats	RTI
November 2004	Complete comparative analysis of models	RTI
November 2004	Create draft press kit	NIGMS
December 2004	Modify smallpox models for flu	Research Groups
February 2005	Scale models to large community size	RTI
May 2005	Create database of synthetic smallpox outbreak	RTI
August 2005	Run models (large N) times	RTI
August 2005	Complete parameter sweep for calibration	RTI
November 2005	Complete social network for small community	LANL
November 2005	Complete parameter sweep for optimization	RTI
May 2006	Complete identification of data sets for flu model	Emory
May 2006	Complete identification of data and create model transportation network	RTI
May 2006	Create database of synthetic flu outbreak	RTI
November 2006	Evaluate intervention strategies	Emory
May 2007	Reconstruct 1918 flu data	JHU
May 2007	Construct generic social network	LANL
May 2007	Create public portal	RTI

 

 

 

The Steering Committee sets milestones, monitors progress, develops policies and information transfer protocols, and may alert NIGMS to relevant research or concerns outside the Network. The Steering Committee established four subgroups to provide support and oversight to the MIDAS Network. The groups are available to the Network for consultation, and they will report to the Steering Committee on progress.

 

Big Iron Subgroup:

Skip Garner (chair), Bruce Hannon

 

Data Subgroup:

Paul Keim (chair), Jim LeDuc, Paul Glezen

 

Modeling Subgroup:

 Simon Levin (chair), Dyann Wirth

 

Surveillance:

Farzad Mostashari (chair), Bruce Weir

 

The Steering Committee decided to meet again in approximately six months, possibly by telephone conference call.  Members would like to be invited to participate in the monthly Network conference calls as a way of keeping informed about the project.