Burden of Pandemics

Burden of Pandemics

Quantifying the morbidity and mortality burden from pandemics poses a significant challenge. Although estimates are available from historical events, the historical record is sparse and incomplete. To overcome these gaps in estimating the frequency and severity of pandemics, probabilistic modeling techniques can augment the historical record with a large catalog of hypothetical, scientifically plausible, simulated pandemics that represent a wide range of possible scenarios. Modeling can also better account for changes that have occurred since historical times, such as medical advances, changing demographics, and shifting travel patterns.

Scenario modeling of epidemics and pandemics can be achieved through large-scale computer simulations of global spread, dynamics, and illness outcomes of disease (Colizza and others 2007; Tizzoni and others 2012). These models allow for specification of parameters that may drive the likelihood of a spark (for example, location and frequency) and determinants of severity (for example, transmissibility and virulence). The models then simulate at a daily time step the spread of disease from person to person via disease transmission dynamics and from place to place via incorporation of long-range and short-range population movements. The models also can incorporate mitigation measures, seasonality, stochastic processes, and other factors that can vary during an epidemic. Millions of these simulations can be run with wide variation in the initial conditions and final outcomes.

These millions of simulations can be used to quantify the burden of pandemics through a class of probabilistic modeling called catastrophe modeling, which the insurance industry uses to understand risks posed by infrequent natural disasters such as hurricanes and earthquakes (Fullam and Madhav 2015; Kozlowski and Mathewson 1997). When applied to pandemics, this approach requires statistically fitting distributions of the parameters. These parameter distributions provide weightings of the likelihood of the different events. Through correlated statistical sampling based on the parameter weights, scenarios are selected for inclusion in an event catalog of simulated pandemic events. A schematic diagram shows how the catastrophe modeling process is used to develop the event catalog.

Analysis of the event catalog yields annual EP curves (for example, as shown in figure 17.2), which provide a metric of the likelihood that an event of a given severity, or worse, begins in any given year. The EP curve is a visualization of the event catalog, in which the number of estimated deaths for each event is ranked in descending order. Because the event catalog includes scenarios incorporating spark probabilities and estimates of disease propagation, the EP curve includes the combined impacts of both spark risk and spread risk. Although a global curve is shown in figure 17.2, EP curves can be estimated for other geographic resolutions, such as a country or province.

 

Estimated Annual Exceedance Probability Curve for Global Pneumonia and Influenza Deaths Caused by Influenza Pandemics, 2017.

The EP curve is a powerful tool that yields several key findings regarding the frequency and severity of potential pandemics. Applied to influenza pandemics, we find the following:

  • An influenza pandemic having the global mortality rate observed during the 2009 Swine flu pandemic (0.2–0.8 deaths per 10,000 persons) or worse has about a 3 percent probability of occurring in any given year.
  • In any given year, the probability of an influenza pandemic causing nearly 6 million pneumonia and influenza deaths (8 deaths per 10,000 persons) or more globally is 1 percent.
  • The annual probability of an influenza pandemic’s meeting or exceeding the global mortality rate of the 1918 Spanish flu pandemic (111–555 deaths per 10,000 persons) is less than 0.02 percent.
  • As indicated by the heavy tail of the EP curve, most of the potential burden from influenza pandemics comes from the most severe pandemics.

Table 17.4 shows select EPs for influenza pandemics in low-, middle-, and high-income countries, based on further analysis of the event catalog. For example, in any given year, all LICs combined have a 3 percent probability of experiencing at least 140,000 deaths attributable to an influenza pandemic and a 0.1 percent chance of experiencing at least 8.3 million deaths. LICs bear a substantial burden of mortality risk from influenza pandemics. Strikingly, LICs contain only about 9 percent of the global population, yet they would contribute nearly 25 percent of deaths during an influenza pandemic.

Based on the event catalog, the average estimated global mortality from pneumonia and influenza during an influenza pandemic is more than 7.3 million deaths. However, because influenza pandemics occur on average once every 25–30 years, the average annual pneumonia and influenza mortality from influenza pandemics is a little more than 230,000 deaths. This is comparable to seasonal influenza, which worldwide causes at least 250,000 deaths annually (WHO 2016b). Although both numbers reflect an annual average, they differ in the combination of frequency and severity. Seasonal influenza deaths occur every year, but pandemic influenza deaths occur much less frequently, are concentrated in larger spikes, and affect a younger demographic.

When pandemics cause large morbidity and mortality spikes, they are much more likely to overwhelm health systems. Overwhelmed health systems and other indirect effects may contribute to a 2.3-fold increase in all-cause mortality during pandemics, although attribution of the causative agent is difficult (Simonsen and others 2013). If indirect deaths are taken into account, the average annual global deaths from influenza pandemics could be greater than 520,000, although there is a significant uncertainty in the estimate.

Pandemics caused by pathogens other than influenza also must be considered. Novel coronaviruses (such as SARS-CoV), filoviruses (such as Ebola virus), and flaviviruses (such as Zika virus) have caused large epidemics and pandemics. These viruses, like influenza, are ribonucleic acid viruses that have high mutation rates. Noninfluenza viruses typically cause more frequent, smaller epidemics but also an overall lower burden of morbidity and mortality than pandemic influenza. For diseases caused by coronaviruses and filoviruses, the lower burden stems from the mode of transmission, which often requires closer and more sustained contact than influenza does to spread.