Modeling "The Perfect Storm": Natural Disasters and Infectious Disease

2011 saw an incredible range of natural disasters throughout the globe. Earthquakes shook the ground the world over, from minor (if unexpected) ones in the eastern states of the US, to the 8.9-magnitude quake and its subsequent after-shocks that ravaged Japan and the 7.6-quake in New Zealand along with several others in Southeast Asia, just barely more than a year after the 7.0-magnitude quake hit Haiti in 2010. Unusually strong storms took their toll as well, including Hurricane Irene which flooded parts of New England, and cyclones which inundated Queensland, Australia this time last year - some five years after Katrina made her mark on the southern US and seven years after the tsunami in Indonesia forever changed how the world looks at waves.

But while the biggest initial shock comes from the disaster itself and its destructive force, the impacts to public health are much longer-lasting and more far-reaching than what can be seen in videos and images on CNN. Natural disasters such as these create ideal conditions for the transmission of a whole host of infectious diseases. Pooled water and debris create ideal breeding-grounds for mosquitoes and other vectors. Unsanitary conditions and lack of clean water amplify the possibility of food- and water-borne disease. The intense clustering of large populations in small areas such as shelters and refugee camps drives effective contact rates through the roof. And the high demand for (and shortage of) adequate medical care - particularly when disasters occur in developing countries - means that more people who do become infected by these means will not be treated, will in turn infect other people, and will have a greater chance of dying.

From a mathematical perspective, natural disasters are difficult to model due to their obvious highly stochastic nature. How can one quantify the probability of a tsunami occurring during a relatively short simulation period? How can a modeler account for all the different known - and unknown - possible natural disasters that could occur? Each disaster is so unique, how can a model make justifiable assumptions regarding what will happen, and when, and how? Yet in spite of their uniqueness, from a mathematical perspective, natural disasters share a small handful of common characteristics:

  • The local population’s background (non-disease-related) death rate suddenly jumps (though the amount depends on the severity of the event, and the availability of recovery infrastructure);
  • Person-to-person direct contact rates within the local population also suddenly jump, and may remain at these higher levels for an extended time-frame - in Haiti, millions were still living in high-density tent cities more than a year after the earthquake ended;
  • Indirect contact rates (i.e. contact with environmental contamination and reservoirs for infection) jump as well, as do contact rates with vector populations; and
  • Disease-related death rates also rise as over-extended medical teams are unable to keep up with both the injured and the infected.

    Based on this commonality, a simplified (albeit imperfect) way of bringing natural disasters into mathematical models of disease transmission is to create a stochastic event that has some probability of occurring during the simulation period. During the duration of this event parameter values for background and disease-related death rates, and direct and/or indirect contact rates temporarily elevate, and then return to baseline levels at the termination of the event, possibly allowing for multiple independent or dependent events to occur over the entire simulation period.

    Within this approach, natural disaster event probabilities, durations and additional parameters can be derived based on historical records of natural disasters in the modeled region, or can be included as part of sensitivity analyses to determine the stability of disease eradication efforts, for example - since such incredibly dramatic and dynamic events as we’ve seen in just the past year can have a tremendous effect on the quantitative outcomes of infectious disease models, and on the health of human - and plant and animal - populations around the world.

    If your organization is struggling with complex public health, infectious disease, environmental or ecological questions, contact MathEcology to learn more about how mathematical modeling can help!

    Back to Table of Contents... >>