Public Health Preparedness Modeling

SummitUp 2010: The Public Health Preparedness Summit’s Conference Blog has a post about our session on Using Computer Models for Vaccination Clinics. The post was written by Rachel Abbey (Montgomery County, Maryland); Rachel, Bill Stephens (Tarrant County, Texas) and I will be leading the session.

I received a note that the Division of Strategic National Stockpile (DSNS) will be having a webinar about the new and improved SNS TourSolver on Friday, Jan. 29, 2010 at 2 p.m.

According to the note, “This pre-released beta version of the new TourSolver was custom designed for use by DSNS’ state and local partners and is simplified for and focused on SNS routing and distribution needs. Please note: This new version does not require Citrix, and the previous Citrix-based TourSolver will no longer be available.”

Interested parties should contact Rick Pietz at CDC or Bob Roberts at C2Logix.

Some public health emergencies could require the quick and efficient distribution of medication and supplies to a large number of people. For instance, to respond to an aerosol anthrax attack, officials will set up and operate Points of Dispensing (PODs) to distribute antibiotics. The medication to be distributed at these PODs must be delivered quickly from a central depot as soon as it arrives.

Planners need a robust plan because there are many uncertainties in medication distribution (including the timing of shipments to the depot, the time needed to load and unload vehicles, travel times, and the demand for medication at each POD). In particular, it is better if the plan calls for delivering medication to PODs much earlier than it is needed. This improves the likelihood that the PODs will open on-time, will not run out of medication during operations, and will dispense medication to the largest number of people in a timely manner.

Our research group recently completed a study on fast algorithms to construct medication distribution plans with slack. Our approach divides the problem into multiple subproblems, including generating a single route, dividing that route into subroutes for each vehicle (”clustering”), scheduling the deliveries, and finding the best delivery quantities. We tested different algorithms on a variety of problem instances based on real-world data. The results show that clustering by route duration generates solutions with more slack than those created by clustering by demand and that these algorithms can generate near-optimal solutions. See the technical report for details.

This work is part of our collaboration with the Montgomery County, Maryland, Advanced Practice Center for Public Health Emergency Preparedness and Response, which is part of the NACCHO Advanced Practice Center program.

The November-December 2009 issue of Operations Research includes the article “Selfish Drug Allocation for Containing an International Influenza Pandemic at the Onset” by Peng Sun, Liu Yang, Francis de Véricourt. (The first two authors are at the Fuqua School of Business at Duke University; de Véricourt is at the European School of Management and Technology in Berlin.)

The article models the spread of an influenza pandemic and discusses how each nation (or a centralized authority like the World Health Organization) should best allocate its stockpile of antiviral drugs to slow down the epidemic. Of course, each nation is acting on its own (and in its own interests), so this can be modeled using game theory. The authors conduct a numerical study of various scenarios, in which sometimes each nations’ selfish decisions are the same as those that a centralized authority would make. In some cases, reducing the total number of infected persons would require a nation to give up its medication to another nation.

Citation: Peng Sun, Liu Yang, Francis de Véricourt, Selfish Drug Allocation for Containing an International Influenza Pandemic at the Onset, Operations Research, Volume 57, Number 6, pages 1320-1332, November-December 2009.

The 2010 Public Health Preparedness Summit will be in Atlanta next month. Among the many sessions on the schedule are the following that discuss modeling:

• Using Computer Models to Plan for the H1N1 Fall Event and Other Health Events, by Jeffrey W. Herrmann (University of Maryland), Rachel Abbey (Montgomery County, Maryland), and William Stephens (Tarrant County, Texas).
• Modeling the Processing of Mass Fatalities, by CM Wood (CDC).
• Simulation for Success: Using Models for Preparedness and Response, by Colleen Monahan (University of Illinois at Chicago) and colleagues from Chicago, Minneapolis, and Pittsburgh.
• Computational Models to Better Understand and Improve Vaccine Delivery, by Bruce Y. Lee, Tina-Marie Assi, and Rachel R. Bailey (University of Pittsburgh).

A report that the Trust for America’s Health (TFAH) and the Robert Wood Johnson Foundation (RWJF) released last month concludes that

the H1N1 flu outbreak has exposed serious underlying gaps in the nation’s ability to respond to public health emergencies and that the economic crisis is straining an already fragile public health system.

Read the Ready or Not? Protecting the Public’s Health from Diseases, Disasters, and Bioterrorism report and see how each state performed. See also the post by Richard Hamburg, deputy director of Trust for America’s Health.

Back in June, 2009, the TFAH and RWJF published Pandemic Flu Preparedness: Lessons from the Frontlines, which described the early lessons learned. The June report offered specific recommendations regarding the H1N1 flu pandemic, whereas the new report offers more general recommendations about the structure of public health preparedness.

The folks at DSNS have announced that a pre-release beta version of the new web-based SNS TourSolver is now available. Apparently, users will still create Excel spreadsheets with their data but will now upload the data to the web server. (The old Citrix-based version is no longer available.)

According to the SNS TourSolver web site:

The SNS TourSolver route optimization application is a web-based software for use by the Strategic National Stockpile for planning and managing distributions. This software is provided by the Centers for Disease Control and Prevention and developed by C2Logix – Route Optimization Solutions. The application is available for use by all participants of the Strategic National Stockpile.

As usual, contact Rick Pietz [eoj3@CDC.GOV] for assistance or information.

RealOpt-Regional is a new planning tool by the folks at Georgia Tech’s Center for Operations Research in Medicine and HealthCare, which is led by Associate Professor Eva K. Lee. It is designed to find the best locations for multiple points of dispensing (PODs).

To use RealOpt-Regional, visit the model web site and request a username and password. To schedule training or ask a question, contact Rick Pietz at eoj3@cdc.gov.

New recommendations by the U.S. Preventive Services Task Force about routine mammograms are getting lots of attention (see statement published today in Annals of Internal Medicine).

Although this topic is not directly related to public health preparedness, it is interesting that the justification for the new recommendations includes the results of six mathematical models. The modeling is discussed in Effects of Mammography Screening Under Different Screening Schedules: Model Estimates of Potential Benefits and Harms by Mandelblatt et al. (This article also cites the articles describing the details of each model.)

The mathematical models are similar and all estimate the impact of different screening strategies. The benefits are measured in life-years gained because of averted or delayed breast cancer death due to a screening strategy; the harms include false-positive mammograms, unnecessary biopsies, and overdiagnosis.

Mandelblatt et al. examined 20 different strategies and identified eight “non-dominated” strategies – that is, for each of these eight, there was no other strategy that gave more benefits with fewer overall mammograms. Among these eight, there is a tradeoff: more benefits will necessarily require more mammograms.

The models also estimate that false-positives occur more often in those strategies that include screening for women between the ages of 40 to 49 years.

The University of Minnesota Center for Infectious Disease Research & Policy maintains a web site called Promising Practices: Pandemic Preparedness Tools. One can download the practices directly from this web site.

The practices, developed by different states and local jurisdictions, are organized into the following topics: types of patient care, communication, community disease mitigation, and helping at-risk groups. The practices are also organized by the states that developed them.