MIDAS Steering Committee Meeting Minutes

January 19, 2005
Natcher Conference Center
Bethesda, Maryland


James Anderson

Daniel Hitchcock
Jeremy Berg Eric Jakobsson

Donald Burke

Paul Keim
Emily Carlson James LeDuc

Phil Cooley

Peter Lyster

Derek Cummings

Ellis McKenzie

Jamie Cuticchia

Farzad Mostashari

Irene Eckstrand

Simon Levin

Joshua Epstein

Steve Naron

Harold Garner

Ben Schwartz

Paul Glezen

Bruce Weir

Elizabeth Halloran

Diane Wagener

Peter Highnam


Action Items

  • Invite a network expert to join the Steering Committee
  • Develop a plan for rotation of Steering Committee membership
  • Develop contacts with the Regional Centers of Excellence
  • Plan for a fuller discussion of the MIDAS Vision
  • Delay public presentation of MIDAS models until the papers have been peer reviewed (Don and Ira will participate in the AAAS meeting as inPiduals)
  • Hold Big Iron Subcommittee meeting (Garner, Weir, Jakobsson, Highnam)
  • Hold meeting with Fred Hayden and Arno Monto to discuss parameter values and biological assumptions

Meeting Reports

Consultation on Pandemic Influenza (Oct 27-28, 2004) – Paul Glezen

The report of this meeting was distributed to participants.  Dr. Glezen discussed the implications of the meeting for the modeling effort, specifically noting that the next pandemic might not be avian influenza.  Some evidence at the influenza consultation suggested that an H7N1 strain might be as serious a threat.  The A Wellington strain, which has now spread from Australia to Europe, is also worth considering.  Dr. Glezen noted that the discussion at the meeting pointed out the need for much better surveillance and a better ability to identify strains.

Consultation on Social Networks (January 5-6, 2005) – Elizabeth Halloran

One goal of the consultation was to identify datasets and to find ways to get a network from data.  It was both satisfying and unsatisfying that none of the experts could adequately describe a system relevant to disease transmission.  The approaches to such a problem included analytical and data-based methods.  The experts also talked about the importance of dynamical networks and addressed the question of whether one can consider networks structure independent of the disease or question.  As a result of the meeting, the MIDAS investigators felt that having social network expertise on the steering committee and having experts available for consultation would be a help.  An action item was obtaining the Nang Rong dataset from Chapel Hill.  We are also considering a future meeting on behavioral psychology.  Simon Levin mentioned that there was a possibility of having a DIMACS meeting relevant to these issues.

Steering Committee Organization

There are several activities in which Steering Committee members can or should participate:

  • Steering Committee meetings (2/year)
  • Monthly conference calls for progress and planning
  • Specialty meetings
  • Network meetings (~4/year)
  • Emergency response planning
  • Setting priorities
  • Reviewing and commenting on policies

The committee would like to discuss scientific issues in more depth and see MIDAS develop a closer relationship with the Regional Centers of Excellence (RCE's).  These centers are having similar organizational and planning issues as MIDAS.

A larger issue is determining what the MIDAS vision is, as well as when and how to get there.  An important component is deciding who the MIDAS client is (e.g., NIGMS, CDC, DHHS, state and local health departments, WHO).  Knowing the client will help inform how we develop an emergency response plan.  Modeling will be relevant to decision making if it interacts constructively with CDC and DHHS processes.

MIDAS Request for Applications

The applications are due at the end of February, and new groups will need to demonstrate their complementarity with current work.  One new area to consider is agricultural applications.

International Collaborations

MIDAS is developing interactions with the European Union, Japan, Russia, Israel, and Thailand.  Some of these collaborations have potential to benefit MIDAS.  For example, INFTRANS (a European Union program) will start in a few weeks and has already expressed an interest in international collaboration.  This program also has an interest in pandemic influenza and will likely plan a meeting for June 2-3.  In addition, Ira Longini and Neil Ferguson will be presenting their results to the Thailand Ministry of Health in February, and Steve Eubank will be attending a meeting in Japan in February to talk about MIDAS.

Media Interactions

The Steering Committee discussed the option of presenting the MIDAS influenza models at the AAAS meeting in February.  The decision was to hold off on a MIDAS announcement until the work has been peer reviewed.  This may mean that Neil's and Ira's papers can be published jointly.

Surge Capacity

MIDAS needs to address the serious question of how to build surge capacity.  Options include

  • Buy more nodes for the current RTI machine so that it can handle more funs of larger programs
  • Pursue agreements with the Department of Energy national labs
  • Pursue agreements with the National Science Foundation supercomputing centers
  • Participate in a distributed system such as Condor

Although the cluster was designed to handle all the usage requirements estimated by the Research Groups last September, it is clear that regardless of surge capacity, MIDAS needs more computational resources since the current computer would not handle multiple copies of Ira's and Steve's models simultaneously.  The models easily run simultaneously in the past;  however, we are gaining experience in the level of computation required to simultaneously run multiple models multiple times.

Model Assessments

Assessment also is critical, including both calibration against data and validation to ensure that the models make sense.  Assessment applies to every step, from the structure of the models to the parameter values to the analytical methods.  Also important are strategies for metamodeling, for sensitivity analysis, and for comparisons of models.  MIDAS needs to know different models respond to changes in parameter values.

Agreed-upon scenarios and parameters are required so the models can be compared.  It is also essential that the models be well documented.  Documentation will require cooperation and communication between the code writers and the RTI staff.

We also must find ways to reconcile the presentation of results and make this a MIDAS effort.  To be useful, the results must be accessible to the audience and must consider the real questions of public health (How many doses?  What is the cost?  Where should the interventions occur?).

We need to keep in mind the utility of having a library of findings as a resource for emergencies.

Model Presentation

The committee was very impressed with the progress that has been made on the Emory and Imperial College models of a SE Asian outbreak.  These incorporate a variety of interventions and potential outcomes.  The committee agreed that these models need to be complementary when presented so that they truly inform our understanding of intervention effects and the dynamics of the disease.

Betz Halloran presented the current model of pandemic influenza from the Emory group for Ira Longini, Azhar Nizam, Shufu Xu, and Derek Cummings. The latter has been helping with information about Thailand. A Thai investigator is being added to the team.

The model is of a typical structured rural population of 500,000 people in southeast Asia. The group explores interventions based on using targeted antiviral prophylaxis, closing schools, workplaces, and/ or social groups, and combinations of the above. The research has investigated the timing and effectiveness of interventions at different patterns of transmission as well as different Ro values. The research is ongoing.