Undergraduates, Ph.D. candidate and professor team up to combat contagion

In a pandemic such as the virus that spreads in the movie Contagion the focus is on understanding the virus so as to develop a vaccine.  But there is a lot more to dealing with a pandemic than finding a vaccine, as hard as the movie shows that to be. 

ORIE Professor Jack Muckstadt, with Nathaniel Hupert, Associate Professor of Public Health at Weill Cornell Medical College, has spearheaded an effort to assure that public health officials pay attention to the proper design of supply chains for fighting contagion.  Their work is relevant to distribution and dispensing not only of vaccines but of antiviral medications that, even before a vaccine has been formulated, can be used to treat those already attacked by the virus.  Muckstadt, who serves on the Advisory Committee on Immunization Practices of the Centers for Disease Control and Prevention (CDC)  involves undergraduates and Ph.D. students closely in this work.

While Contagion establishes early (probably to simplify an already complex plot) that no antiviral drug is effective against the fictitious "MEV-1" virus, for actual seasonal flu and epidemics such as H1N1, antivirals can be positioned early and can ameliorate symptoms and reduce the death toll.  But their effectiveness depends on logistics. 

Muckstadt points out that "almost all states reported that there were problems with distribution and dispensing of antivirals and shortages of all types of antivirals in some geographic locations, at some points of time during the [2009 H1N1] pandemic."   As a result, the CDC is embarking on a new effort to redesign the way antiviral medications are distributed.   The CDC has awarded Muckstadt and ORIE M.Eng. student Christine Barnett '11 a grant to provide their expertise to this effort.

In the film [warning: plot spoiler] a vaccine is eventually found, but it takes 90 days to get into production and a year to distribute.  The National Guard, the Red Cross, FEMA and industry cooperate on the supply chain to distribute and dispense the vaccine.  Along the way, the characters played by Jennifer Ehle and Kate Winslet introduce concepts such as "fomites," "index case," and "R0" that are essential ingredients in epidemiological modeling, an application of stochastic processes.   

Muckstadt recently attended a meeting of the Advisory Committee at the CDC in Atlanta, and reports that the film, partly filmed on site, was much discussed there, mostly in a positive way. 

Microworlds

Here in ORIE, Barnett and others have been developing simulated environments, or "microworlds," that enable public health officials to see the impact of alternative designs to distribute antivirals and vaccines.   Using one of them, the Emergency Supply Chain Operations Evaluator (ESCOE), public health administrators test various logistics structures with their many constraints - demands, capacities, and service and refill rates - to predict how well each logistics design delivers countermeasures during a simulated pandemic, and at what cost.  The administrators can specify the number and location of dispensing points, distribution centers, and warehouses, as well as the policies that govern dispensing, such timing of startup. 

A paper describing ESCOE, by Muckstadt and Barnett with Kenneth Chu '11 and Ph.D. student Kathleen King, was a finalist in the annual INFORMS Undergraduate Operations Research Prize competition in Austin, Texas last fall.

ESCOE is one of several models developed for different aspects of planning to deal with contagion.  Another, intended for teaching purposes, is called the Pandemic Influenza Logistics Simulator (PILS).  It is designed to "establish the importance of designing and operating a coordinated command and control system for managing resources in a dynamic, uncertain and constrained supply chain" to confront a pandemic influenza, according to Muckstadt. 

In a hands-on class, students (or administrators) use PILS to work through a variety of policy approaches, for example moving stock to the hospital where it is as close to the patient as possible, setting system-wide target inventory levels, and deriving how much must be produced the factory each day.  PILS demonstrates that "this simplistic policy is not always the best," according to Muckstadt, partly due to the high degree of uncertainty and variation in demand as the pandemic progresses.  Users of PILS "learn by doing, carrying out experiments to learn the future and distant consequences of current decisions," he said.  

From Simulation to Optimization

The simulations developed by Barnett and others, including current ORIE M.Eng. Adam Schultz '11 and past ORIE undergraduates Chu, Cindie Wu '08, Caitlin Hawkins '08  and James Codella '07 M.Eng. '08, require the user to nominate the inventory approach to be evaluated.  The computations do not seek such a stocking and distribution approach on their own.   To supplement the simulation work, King is building a mathematical model that, if solved computationally, can find an optimal stocking and distribution policy based on the characteristics of a pandemic.  Her research takes the uncertainty and variability in the spread of the pandemic into account.  

However, despite the power of current computers, models like King's can not be solved to find a truly optimal solution, due to what is known as the "curse of dimensionality."  Nonetheless she has been able to use the model to find approximate solutions using various strategies for allocating resources.

One strategy (which public health officials have planned to use) is to deploy inventory proportional to population.  This strategy is called the "fair share" method.  Another strategy is to reduce the number of time periods the model looks ahead to see the implications of current allocations.  A third uses a technique, developed by ORIE Professor Huseyin Topaloglu and Sumit Kunnumkal, ORIE Ph.D. '07, that breaks the overall problem into a series of smaller problems whose solutions can then be integrated to arrive at an estimate of the optimal solution.  King has found that the latter two methods provide much better solutions than the "fair share" method.

And Back Again

Output from King's optimization model can be tested through the simulations, which incorporate more operational details.  "The simulation that Christine and Kenneth built (ESCOE) includes more detail than I can include in my optimization models," said King, "so I can use ESCOE to test my optimized logistics policies under more realistic scenarios." Equally important, the optimization results provide a benchmark against which other designs can be measured using simulations that report "these are the consequences and this is the best that you can achieve,"  King said.

"Christine has designed a great interface for ESCOE that has been really helpful when we are discussing our work with public health officials," King said.  In the process, she added, "I've learned that clear user interfaces and output reporting mechanisms are essential when you want to discuss models with policy-makers." 

The undergraduate students who were employed in ORIE during the summer months to work on public health projects were supported through generous gifts from Sarah Jacoby '96 M.Eng. '97 and Charles Tall '78. 

Educational Impact

 The close collaboration among faculty, graduate students and undergraduates working to contain contagion has a valuable educational and personal  impact.  King and Muckstadt have become mentors to the students involved in the activity.  

"Working with them has been a huge part of my decision to go on for a Ph.D.," said Barnett.  "I've talked to Kathy a lot about the process of applying for a Ph.D. and about what to look out for in doing so."  King said "I've told her a few things that I wished I 'd known at the beginning of graduate school—like the fact that it's okay if you feel like you're out of your depth sometimes—everyone feels that way!" 

"I was an undergraduate myself just a few years ago," said King, "so I can easily remember my first experiences with research—I didn't really understand what 'research' meant and I wasn't used to having extremely open-ended projects.   I'm happy to have the chance to help other undergraduates make the transition to research."    

Barnett said that in working with Muckstadt and King "I got to see the interaction between Ph.D. student and professor."    The project "opened my eyes to the possibility of a Ph.D." she said.

Inspired in part by her work with Muckstadt and King, Hawkins is now a Ph.D. student in industrial and systems engineering at the University of Southern California.  She has held internships relating to logistics at Weill Medical College, RAND Corporation, and the US Navy's Bureau of Medicine and Surgery.   At RAND she worked with Ed Chan '92, M.Eng. '93, Ph.D. '99, as did Wu and King. Wu, on leave from Stanford University's Ph.D. program, works at a supply chain consulting company in Palo Alto.

Codella is now a Ph.D. student in industrial engineering at the University of Wisconsin-Madison, where has written that he is "implementing operations research tools to complex decision-making problems in health-care with a primary focus on infectious disease control."  He is currently working on methods to mitigate infectious disease spread in hospitals.

Barnett will work full time on the CDC project after completing her M.Eng. in December, and has begun applying to Ph.D. programs. 

Researchers like Barnett, Chan, Chu, Codella, Hawkins, Hupert, King, Muckstadt, Schultz and Wu,  no less than the "virus hunters" celebrated on the CDC's web site about Contagion, are in the forefront of preparedness for the counterattack on future pandemics.