Harvard School of Public HealthKresge Building, Room 213677 Huntington AvenueBoston, MA
Michael MersonFarzad Mostashari
James AndersonMarcelo Azevedo CostaPhil CooleyDerek CummingsJosh EpsteinMoshe FederHelen RayDoug RobertsJohn StellingWilliam ThompsonJen VillaniWilliam WheatonAlcia Williams
The next steering committee meeting will be May 5, 2008. The steering committee will meet annually.
The January 2006 minutes erred in stating that Gary Smith is from the Bolger Center of Pennsylvania State University. The minutes will be corrected to show that he is from the University of Pennsylvania.
As its first order of business, the steering committee voted to meet once a year in May. The next meeting will be May 5, 2008.
In order to provide oversight in the interim, the steering committee will meet by videoconference or conference call in six months (July 2007).
The steering committee reviewed progress on the priorities set in 2006 and listed below.
The MIDAS PIs noted that the pressure has been to increase, rather than decrease, complexity. Indeed, given sufficient computational resources and appropriate algorithms, one might question whether simplification is an appropriate goal. A more suitable goal might be to create a hierarchy of models, from simple to complex, that can address a variety of research and policy questions.
All projects have contributed to research on modeling influenza. Some connections have formed naturally among groups, largely through more informal discussion and common research interests. This is a good time to expand MIDAS's research interests. Although influenza remains a high priority, especially with respect to integrating field studies, the program should work on models of other endemic diseases such as vector-borne and enteric diseases. The committee noted that integrating behavioral, evolutionary, and ecological information continues to be an important research direction.
The data group has focused primarily on historical data, mostly documents. The steering committee suggested looking into collecting international data, possibly from Japan or Europe.
MIDAS projects must be in compliance with the approved policy for deposition of information. The steering committee restated the importance of depositing model code and results in the MIDAS model repository.
Two groups (Longini and Lipsitch) are working on this problem.
At least two groups (Smith and Longini) have done some work on consequences of poorly matched vaccines. This is a continuing area of research.
This collaboration has not materialized; however, MIDAS is establishing collaborations with the Centers for Disease Control and Prevention. In addition, MIDAS should look into working with the CDC Centers of Excellence and Bioinformatics.
After discussion, the steering committee endorsed the goal of a 5-year plan to move toward an industrialized model(s); however, this is probably not within the purview of MIDAS. The executive committee should systematically approach identifying gaps in the models and ways to address these gaps. The steering committee also endorsed developing materials for the public and for policymakers in order to explain the role and successes of MIDAS modeling. CDC would be a very good collaborator.
The recommendations of the IOM panel may be a useful foundation for developing the field of infectious disease modeling. The report clearly points out the need for better training, more surveillance, better understanding of underlying biology, and more effective communications with policymakers. Models are only as good as what we already know; better science is necessary for better models. Models can help surveillance by identifying what information needs to be collected and how it should be collected as well as the cost effectiveness of surveillance and interventions.
Education about the value and role of modeling is an ongoing effort at every level. Some people may expect modeling to do the impossible; others believe that modelers are promoting a self-serving agenda. Part of the challenge of assisting policymakers is remaining independent of their biases, both in developing questions and reporting results.
Members of the steering committee visited the programs at University of Pittsburgh, Virginia Bioinformatics Institute, and University of Washington. The PIs felt that the visits were very useful and generated very useful ideas. The site visitors were very pleased with the groups' progress and the quality of their collaborators. All three groups have benefited from their moves to new institutions and are in various degrees of consolidating their research programs.
Dr. Wagener reported on progress in computational support, data identification acquisition, analytic support, and other projects.
Nearly unlimited support is available through the IG Linux cluster and NSF Supercomputing Centers. Several groups (UMd, Imperial, RTI, and VBI) have developmental allocations (DACs) of up to 300,000 hours. VBI is using a 200,000 medium resource allocation (MRAC) to develop a grid-based version of an agent-based simulation.
The informatics group provides support on code development and optimization. A collaboration with Josh Epstein has focused on a global transportation model. In collaboration with Gary Smith, the IG has acquired licenses for MatLab to enhance the use of large data sets. Finally, a collaboration with Stephen Eubank and the San Diego Supercomputing Center focuses on visualization tools.
The MIDAS model repository (MREP) archives, documents, and allows for retrieval of model results. Six projects (16 models) are archived and available for retrieval.
The North Carolina use case is a prototype to develop tools, acquire data, and test computation options for agent-based models. Work on the use case enables the informatics groups to develop visualization and analytical tools, to evaluate alternative data sources, and to create template for documentation of model versions.
The historical data document catalog contains more than 420 documents with nearly 2000 annotated data sources. These can be searched for outcome, time period, location, and other parameters. Over 40 GIS databases are also available, and the informatics group will work with any MIDAS researcher to produce data files for specific model inputs.
With Gary Smith, the informatics group has developed a synthetic poultry population using multiple GIS layers to create a probability surface for synthetic poultry farms in Lancaster County, PA. A synthetic human population is based on TransSim and includes for each household the age distribution, income, size, household owner age, and vehicles. The data set is being expanded to include schools and bus ridership as well as group quarters.
Other available data include the New York City annual survey (a 2006 survey of commuting data including demographic information) and data on the Yancey County, NC influenza B outbreak.
In collaboration with Josh Epstein, the global model has provided opportunity to develop the AnyLogic model, conduct experiments on granularity, and review Java code for parallel processing model. With the Platt research group, the informatics group is developing capability in using SatScan. The community containment project, involving three research groups and the Homeland Security Council, called on the informatics group to compile results and provide analyses.
The informatics group has worked with the Homeland Security Council to collect data for economic analyses, with AHRQ on tool development, and with CDC on community containment strategies. The Toronto SARS community response is summarized in a white paper distributed to the research groups and available on the portal.
The CDC reports which were provided by Robin Bush contain very detailed information on influenza outbreaks as they were happening. Scanning these reports is labor intensive, so it is important to know if people are actually going to use them.
Building data sets is the highest priority, and MIDAS must address the problem of determining what is available and how to access it. The steering committee recommended that people who send in data give short talks to MIDAS, possibly on the monthly conference calls. Users should send out information on data when it is deposited.
Although many interactions between the research groups and the informatics group are excellent, there is concern about the timeliness and quality of response in some cases. The steering committee advised the informatics group to develop procedures to provide more timely response to problems and concerns of the research groups. The big iron subcommittee will follow up at the site visit to the informatics group this spring.
The steering committee also reiterated the importance of depositing results and source code.
Jen Villani reported that the MIDAS evaluation is nearly ready to start. The evaluator will look at infrastructure, policies, resources, mechanism, and value of products. The steering committee suggested that the evaluation include the role of the steering committee on the impact and development of the project.
Marc Lipsitch, Ira Longini, Don Burke, Stephen Eubank, and Neil Ferguson gave presentations about their scientific work. These presentations are available on the MIDAS Portal.
Connect With Us: