4

The Economics and Modeling of Emerging Infectious Diseases and Biological Risks

During session I, part B, of the workshop, speakers explored the economics and modeling of emerging infectious diseases and biological risks. The session, moderated by Rebecca Katz, associate professor of global health at Georgetown University, opened with an overview of the cost of pandemic influenza by Martin Meltzer, senior economist and distinguished consultant for the U.S. Centers for Disease Control and Prevention (CDC). Anas El Turabi, Frank Knox fellow in health policy at Harvard University, followed with a discussion on assessing economic vulnerability to emerging infectious disease outbreaks. Carlos Castillo-Chavez, professor of mathematical biology at Arizona State University, then presented on an epidemiological-economic model that explicitly incorporates human behavioral responses influenced by infectious disease outbreaks. Thomas Inglesby, director of the Center for Health Security of the Johns Hopkins Bloomberg School of Public Health, concluded the session with a presentation on infections that have the potential to cause significant harm to the global economy and international security.

COST OF PANDEMIC INFLUENZA

Martin Meltzer, senior economist and distinguished consultant for CDC, discussed the economics of planning and preparing for influenza pandemics. Influenza pandemics are inevitable, but they vary greatly in terms of timing, severity, and populations affected. Influenza pandemics can occur anywhere from every 10 to 50 years (Potter, 2001). The timeline for influenza pandemics complicates communication with policy makers,

who are significantly more motivated by immediate problems than a potential problem in the next decades, he said. Influenza pandemics also vary in terms of mortality risk. The 1918 pandemic resulted in an estimated 675,000 deaths in the United States and 50 million deaths worldwide, while the 2009 H1N1 pandemic resulted in an estimated 12,500 deaths in the United States and 285,000 deaths worldwide (Taubenberger, 2006; Shrestha et al., 2011; Dawood et al., 2012). Estimates of macroeconomic impact are also important to consider as influenza pandemics, even if short in duration, can cause billions of dollars in economic loss and affect gross domestic product (GDP) (Meltzer et al., 1999; McKibbin and Sidorenko, 2007; Fan et al., 2016).

Economic Modeling for Influenza Pandemic

Meltzer highlighted the potential for economic modeling to guide preparedness efforts against pandemic influenza, as it can provide information that can be useful when planning for rationing, shortages, and prioritization of interventions during an epidemic. He added that unless things change drastically in terms of technologies, come the next pandemic, there are likely to be shortages in medical countermeasures (at least initially). so the question is who gets to receive the care first. He reiterated that plans must be flexible and nimble, and respond to the unique characteristics of the outbreak as it happens. As an example, he noted that in 2009 many people over the age of 65 had a degree of unexpected immunity to H1N1. He argued:

Meltzer described models that provide information about mortality from pandemics and the effect of vaccines. It is difficult to make precise mortality estimates about future pandemics, even with good economic and epidemiologic models, he said. Death rate estimates are in the form of a range of potential outcomes depending on gross clinical attack rates, which are not a precise prediction.

Meltzer said that economic models can help evaluate the cost-effectiveness of vaccination programs by age group and risk, including comorbid conditions and pregnancy. Economic analysis of vaccination has produced positive rates of return for every age and risk group but at varying magnitudes (Meltzer et al., 1999). Categorizing people into age and risk groups can help determine who to vaccinate first in the case of a supply shortage

during a pandemic. For example, population groups can be evaluated either by their risk of death or by their potential future returns to society, the latter of which would favor vaccinating working-age people before the elderly. Meltzer noted that determining the best approach to this type of valuation is beyond an economic problem and is up for debate for society. He added that the best way to carry out an economic analysis, from his experience, is by using simple models that are transparent, take account of uncertainty about the severity and size of impact of the pandemic, and are readily accessible to the public.

Stockpiling and Nonpharmaceutical Interventions

Many economic models have suggested the benefit of stockpiling vaccines, antiviral drugs, and mechanical ventilators to prepare for an influenza pandemic, but the effect of this strategy is limited and attacks only part of the crisis, according to Meltzer. The problem is not merely a shortage of material supplies, but also of human resources, he said. As an example, he described the limitations of stockpiling mechanical ventilators, which require trained critical care nurses and respiratory therapists to effectively operate (Ajao et al., 2015) (see Figure 4-1). This illustrates the need for flexible planning and consideration of a system’s maximum capacity to use a commodity when deciding on the amount to stockpile, he concluded.

Additionally, nonpharmaceutical interventions, such as school closures to limit the spread of the virus, can be considered as a response strategy to an outbreak. Meltzer noted that these strategies work in some situations, albeit with limitations. He described a natural experiment in Texas during

the 2009 H1N1 influenza pandemic where a school district that closed public schools during the outbreak was compared to neighboring school districts that mostly remained open. The population near the school district that closed reported a significant reduction in visits to emergency rooms from influenza-like illness during the closure period (Copeland et al., 2013). After those schools reopened, however, rates of illness rose once again. According to Meltzer, the lesson here is that during a pandemic, schools must close early and close for a long time until vaccines are available. Both of these approaches can be difficult political decisions to make, he said—a fact that underscores the need for the support of businesses and communities when implementing pandemic response strategies.

ASSESSING ECONOMIC VULNERABILITY TO EMERGING INFECTIOUS DISEASE OUTBREAKS

Anas El Turabi, Frank Knox fellow in health policy at Harvard University, stated that economic analysis can take two forms: the “snow-globe” approach and the “empiricist” approach. The snow-globe method attempts to build mathematical models of the world in its current state, which are then “shaken” to hypothesize the consequence of a given scenario. Simulations can be repeated to create a dataset of potential outcomes, but the results depend on model inputs and assumptions. The empiricist approach, on the other hand, attempts to measure effects after a real-world event has happened. According to El Turabi, not enough of the latter is happening for infectious disease outbreaks. He stated that there is a “need to move from a modeled world to a measured world.”

Role of Economic Analysis for Outbreaks

El Turabi presented a framework of the economic impacts of an infectious disease that includes three components: transmission dynamics, economic impact, and disease dynamics (see Figure 4-2). According to the framework, the pathogen in a reservoir infects humans (often through vectors), which can lead to an outbreak. This infection causes a biological response, including illness and death, which subsequently affects consumption and productivity in both the short and long term. In addition, there is a social response to the outbreak from individuals (e.g., change travel patterns to avoid disease prone area), organizations including private companies and nongovernmental actors (e.g., rescind investment commitments), and governments (e.g., impose regulations and a cordon sanitaire).¹ Taken together, these responses lead to significant economic damages associated

___________________

¹ A blockade enacted to prevent the spread of individuals afflicted by an infectious disease.

with their effect on consumption and productivity. According to El Turabi, this economic impact, particularly from social responses, is often the forgotten dimension of analysis related to emerging outbreaks.

El Turabi noted that different factors might affect a country’s economic vulnerability to an infectious disease event. Intrinsic vulnerability is defined as the likelihood an infectious outbreak will occur in a given country or context (Sands et al., 2016). Strengthening pandemic preparedness through strong health systems is a way to bolster intrinsic vulnerability. Intrinsic vulnerability and preparedness are evaluated through Joint External Evaluations, a voluntary and multisectoral process to assess country capacity to prevent, detect, and rapidly respond to public health risks. He added that while assessing the economic vulnerability is an intersectoral issue, the vulnerability of the industry sector has been not examined rigorously or consistently yet to understand fully the potential effect of major outbreaks on private enterprises.

Economic Effects of the Ebola and Zika Epidemics

For the last portion of his presentation, El Turabi presented case studies of the economic effects of the Ebola epidemic in West Africa and the Zika epidemic in Latin America, highlighting their differing disease dynamics. He described the disease dynamics of Ebola as a “raging forest fire” and Zika as a “slow burn,” though both significantly affected human health (see Table 4-1). Zika has infected more people than Ebola, but the clinical syndrome is much less severe in the acute stage. The difference in fear induced by the visible hemorrhagic condition of Ebola versus apathy from the flulike symptoms or neurological complications experienced with Zika is a key qualitative factor in comparing the two diseases, he said.

El Turabi noted that the different characteristics of Ebola and Zika explain the different behavioral responses and economic impacts generated by these diseases. Recent estimates on Ebola and Zika from 2016 and 2017 demonstrate these distinctions in both the short and long terms following the outbreaks. Short-term costs of the West African Ebola epidemic are estimated to be approximately $3 billion, while a long-term assessment of Ebola’s effect on population distributions, migration, and investment confidence remains incomplete (World Bank, 2016). Estimates of the short-

TABLE 4-1 Epidemiology of Public Health Emergencies of International Concern, Ebola Versus Zika

NOTE: PHEIC = public health emergency of international concern; WHO = World Health Organization.

SOURCE: El Turabi presentation, June 12, 2018.

term economic impact of Zika across the Latin American region range from $7–18 billion; long-term costs related to lifetime care of microcephaly patients may be as high as $11 billion (UNDP-IFRC, 2017).

According to El Turabi, there is a trend of large economic losses occurring even after a relatively modest event. He reiterated that this scenario is typically driven by human behavioral response—a phenomenon that El Turabi believes needs far more research. Economic effects can also long outlast the epidemiologic events themselves. Finally, El Turabi concluded that better postevent analysis and data collection are needed to calibrate and refine predictive models.

EPIDEMIC RISK MODELING: MEASURING THE EFFECT OF AVERSION BEHAVIOR AND CASCADING SOCIAL RESPONSES

Carlos Castillo-Chavez, professor of mathematical biology at Arizona State University, presented on epidemic risk models that incorporate human behavioral responses. He began by describing the “susceptible-infected-recovered” (SIR) model, a mathematical model that can be used to evaluate disease outbreaks and predict epidemiologic outcomes (Huppert and Katriel, 2013). The SIR model is able to make accurate short-term predictions when provided with the appropriate information inputs and has been successful in modeling the rapid spread of several recent epidemics such as the severe acute respiratory syndrome outbreak in Canada (Choi and Pak, 2003); however, it does not explicitly include behavioral responses to disease risk.

Dynamics of Human Behavior and Infectious Diseases

Castillo-Chavez proceeded in describing models that incorporate human behavior. He first pointed out an agent-based simulation of an influenza outbreak in Portland, Oregon, that examined disease transmission based on the physical contact patterns that result from movements of individuals between locations.² He showed a simulation of a scenario of an influenza epidemic where nobody changes their behavior (carrying out their normal activities), versus a scenario where 75 percent of the population avoids social contact. This model suggests that different types of human behavior and decisions affect the spread of infectious diseases, which have implications for public health interventions and policies (Eubank et al.,

___________________

² An agent-based model is a type of computational model that is used to study complex systems by examining the way individual agents of a system behave, as a function of individual characteristics and interactions with each other and the environment, according to predefined rules (IOM, 2015).

2004). However, while this kind of model explicitly assigns behavioral rules for all individuals, it often requires the modeler to specify ex ante how changing incentives modify behavior and therefore is limited in its ability to aid in designing incentives (Fenichel et al., 2011).

According to Castillo-Chavez, accounting for human behavior is challenging when modeling a complex adaptive system of human disease dynamics,³ as the model must consider human decision making, disease transmission, and disease prevalence (see Figure 4-3). Disease risks both affect and are affected by human decisions, which creates a feedback loop whereby infection levels drive behaviors and human decisions shape disease spread (Fenichel et al., 2011). These human decisions are determined by trade-offs when humans consider the scarcity of time, money, and other resources, he said. While people might value their own health status, they can also value family, relationships, work, and social activities that affect their decision making and exposure to infectious diseases. With these various sources of trade-offs, people can change their behaviors in response to changing circumstances and incentives. They make these decisions based on the best available information, but they may be missing key information, he added.

___________________

³ A complex adaptive system is a collection of individual agents with freedom to act in ways that are not always predictable and whose actions are interconnected (Plsek and Greenhalgh, 2001).

An Epidemiological-Economic Model to Account for Human Behavior

Taking these factors into account, Castillo-Chavez introduced a model that combines the SIR model and an economic behavioral model, also known as an epidemiological-economic model (Fenichel et al., 2011).⁴ It explicitly models the trade-offs that drive human-to-human contact decisions in response to disease risk. The model assumes that people make decisions to maximize utility (an index of well-being) based on their health status, their understanding of disease risk, and their evaluation of future potential scenarios. That is, the model recognizes that individuals, particularly those who are susceptible to the disease, may respond to disease risks by limiting contacts but may also derive utility from contacting others, which may lead to an increase in disease prevalence. The model further assumes that people have instantaneous access to information when they make decisions. He noted that unlike traditional nonlinear contact models, this model focuses on trade-offs not based explicitly on the basic reproductive number of the disease, R₀.⁵ To Castillo-Chavez, R₀ implicitly includes disease-free behavior and confounds biological aspects of the pathogen with social aspects of adaptive human response to disease risk; thus, it may not reliably guide postoutbreak disease management.

Castillo-Chavez highlighted how the results of this model reveal that adaptive human behavior can have a significant effect on disease dynamics (see Figure 4-4), and thus have critical implications for developing public health policies, such as social distancing policies that alter the incentive structure of humans contacting each other. Analyzing both behavior and disease dynamics, he argued, may shed light on how to develop incentives for individuals to change their behavior for an optimal, cost-effective disease response strategy. He concluded that this kind of work requires a better understanding of human behavior and collaboration of different disciplines.

IMPACT AND FUTURE OF GLOBAL CATASTROPHIC BIOLOGICAL RISKS

Thomas Inglesby, director of the Center for Health Security of the Johns Hopkins Bloomberg School of Public Health, highlighted the importance of understanding global biological risks that are acute, fast-moving, and consequential in terms of health and economic impact. He argued for the need to better communicate these type of risks among scientists, researchers, and policy makers. Recognizing this need, he and his team at

___________________

⁴ “Epi[demiological]-economic models merge economics and epidemiology by explicitly analyzing individual behavioral choices in response to disease risk” (Fenichel et al., 2011).

⁵ R₀ is defined as the number of secondary infections in an uninfected population that is generated from the initial introduction of a pathogen.

the Johns Hopkins Center for Health Security created a term that encapsulates these risks called global catastrophic biological risks (GCBRs) along with a working definition:

Inglesby argued that GCBRs warrant heightened attention because of their extraordinary consequences, including damages to governments, economy, and society. He also argued that they are potentially tractable problems and that categorizing these set of risks under a term like GCBRs could help the global community work in a concerted way to prevent colossal consequences. He added that scientists have helped drive global concern or action on other widely accepted global catastrophic risks, such as nuclear weapons, climate change, and artificial intelligence, and could see the same happening for GCBRs.

A Retrospective Look at Global Catastrophic Biological Risks

Inglesby provided three examples of GCBRs. He noted that the 1918 influenza pandemic is an archetypal example of a GCBR because of its extreme economic and social impact in addition to its large mortality toll—an estimated 50 million people. The case fatality rate was 1–2 percent (Taubenberger, 2006). Most notably, he said the disease’s short-term effect on governments and economies was undetectable over the years that followed; however, the effect of a similar pandemic today would likely be far more severe with the increased convenience of travel and globalization. Global consequences from this type of pandemic might include high absenteeism caused by fear, conflicts among countries over access to therapeutics, and disruptions to critical infrastructure, government, commerce, air traffic, and the military.

Inglesby presented smallpox as another example of a GCBR for its potential to reemerge and cause devastating global effects as seen prior to its eradication in 1977. The case fatality rate was 30 percent, killing more than 2 million people annually before the eradication campaign began (CGDEV, 2007; CDC, 2016b). Given the recent de novo synthesis of horsepox, a virus closely related to smallpox, there is increasing concern of an accidental or purposeful reintroduction of smallpox, which was extremely

transmissible among humans and had a high case fatality rate as noted earlier (Noyce et al., 2018). Inglesby emphasized that today’s population would be immunologically naïve to smallpox globally, and that the societal impact would depend on the efficacy of global governance, public health measures, medical countermeasures, and other responses.

Finally, he described a more recent example, between 2005 and 2007 related to the pandemic potential of H5N1 influenza, which was spreading fast among bird populations in Africa, Asia, and Europe. It had a high case fatality rate of more than 50 percent among humans, though no sustained human-to-human transmission developed (Neumann et al., 2010). If the virus evolves naturally to have a high case fatality rate and becomes readily human transmissible or if it is deliberately manipulated to become so and purposefully reintroduced, this in fact would constitute a GCBR according to Inglesby. A virus with these types of properties would be efficient and lead to sustained and perhaps permanent damage to society. Governments and the global community would have to take extraordinary action to swiftly develop and deploy medical countermeasures, he said.

Future Global Catastrophic Biological Risks

Looking into the future, Inglesby described the characteristics of potential pandemic pathogens. Based on a poll of experts in the field, the highest risk for a future pandemic will likely be from a respiratory RNA virus, with characteristics such as segmented genome, cytoplasmic replication, small genome host size, high host viremia, and zoonotic relationships (Adalja et al., 2018). Inglesby noted that these are the most probable attributes for the next major outbreak, but they are by no means definite: smallpox, for example, is a DNA virus that replicates in the nucleus. He also added that CDC considers H7N9 influenza as the greatest pandemic risk, which has a 40 percent case fatality rate (Xiang et al., 2016).

Inglesby noted that factors such as the presence of effective vaccines and therapeutics and the range of disease vectors determine the pandemic potential of the disease, but these limitations are malleable, as pathogens can evolve through deliberate or natural modifications. Some experts also raise the risk of fungi, whose thermal range limitations could be overcome by natural selection or biological engineering in the near future. Inglesby pointed out that biotechnology facilitates pathogen targeting against specific populations with shared genetic history. This kind of technology could also be applied to designing novel or artificial organisms harmful to existing life, as well as pathogens targeting livestock or plant food sources, which could have devastating consequences for the global food supply.

Economic Analysis of Biologic Threats

Inglesby stated that several economic analyses of these biologic threats have demonstrated the magnitude of their impact, although the estimates vary. One study concluded that pandemic mitigation programs could save as much as $360 billion over the next century (Pike et al., 2014), while another estimated that future pandemics could cost the global economy as much as $500 billion per year (Fan et al., 2018). Additionally, a World Bank study estimated that a 1918-style pandemic would result in a $1.5 trillion loss in global GDP (Burns et al., 2006).

Inglesby proposed further studies on the economic consequences of GCBRs. He argued that more studies need to examine the effect of pandemics with case fatality rates that are greater than that of the 1918 influenza pandemic. He also urged for more studies that take into consideration effects that are beyond lives lost and income lost to prolonged societal instability, prolonged interruption of international trade, and collapses of industries and governments. Finally, he pointed to the need for studies that consider the economic effects of pandemics with different dynamics than influenza such as smallpox and deliberately initiated events.

DISCUSSION

Katz summarized some of the key points raised during the presentations. In her view, she noted that there was a need for the following:

• Accounting for significant variability and uncertainty in pandemic preparedness planning;
• Undertaking post-hoc analyses of outbreaks and investing in understanding social responses to gain a more comprehensive view on the economic consequences of outbreaks;
• Explicitly incorporating adaptive human behavioral responses in economic and disease modeling as they can change the course of epidemics; and
• Performing more economic analyses on GCBRs because of their potential effect on governance, international relations, and society.

The discussion with the audience began with a focus on human behavioral responses to infectious disease outbreaks. Peter Daszak, president of EcoHealth Alliance, asked Castillo-Chavez to expand on the relationship between incorporating human behaviors into modeling and the estimates of R₀ of an infectious disease outbreak. Castillo-Chavez said that R₀ is calculated based on underlying assumptions of a problem where biological aspects of the pathogen with social aspects of adaptive human responses

are confounded. However, it is important to parse this out, he noted, as problems like infectious diseases are dealt with in a heterogeneous manner, which can lead to multiple disease states. For example, a disease like HIV has multiple modes of transmission and does not have a single endemic state. Heterogeneity across individuals with varying social dynamics and behavior often leads to different decisions and different endemic states, particularly if long-term dynamics are considered. He noted that heterogeneity makes modeling complicated so it gets ignored as it can pose a challenge in finding policies that can be implemented easily. However, he argued that policies that ignore uncertainty and behavior miss opportunities to have a significant effect.

Katharina Hauck, senior lecturer in health economics from the Imperial College London, asked Castillo-Chavez about the extent to which adaptive human behavior can reduce the threat of pandemics during the eradication and elimination stages. She noted that individuals might demand less prevention as prevalence declines, making eradication difficult. Castillo-Chavez reiterated that modeling behavioral responses is critical but a great challenge. He mentioned a study that examined the HIV epidemic among a homosexually active population in San Francisco. The study assumed that even though homosexual people would want to move into a welcoming environment that accepts homosexual people such as the Castro neighborhood in San Francisco, they would also be dissuaded to go if they were aware of the high levels of HIV infection. In other words, “recruitment” to the homosexual population depended on levels of infection. However, Castillo-Chavez noted that while this population in San Francisco was well informed of the risks, when the bathhouses opened back up, the infections increased again. This led to oscillations in incidence of the disease, he said. As mortality from HIV/AIDS has dropped over the years, behaviors have changed in response to the trend. These modeling efforts are not always precise, he said, but they shed light on the importance of understanding how adaptive behaviors shape epidemics.

Jeffrey Duchin, health officer and chief of the Communicable Disease Epidemiology and Immunization Section for Public Health—Seattle and King County, Washington, asked panelists if they had seen modeling of pseudo-outbreaks, meaning an increase in the number of cases reported that is not associated with an actual increase in disease incidence but found to be an artifact. He noted that these often arise through social media-generated scares, but they can have real consequences in health and economics. He noted that the autism scare could be considered such a pseudo-outbreak as it affects conducting effective immunization programs. Inglesby responded that the focus on syndromic surveillance in public health can lead to tradeoffs and shortfalls in routine public health priorities. He explained that cities often chase false signals in the electronic surveillance systems, resulting

in huge costs, though there is a lack of economic analyses that calculate them.

Meltzer added that public health officials and policy makers often make decisions on outbreak response and resource allocation up front with incomplete information, including uncertainty about how people will respond to interventions. Modeling decision making during an outbreak requires making assumptions about a population’s compliance with interventions, such as self-isolation, and how their behaviors change over time. Meltzer emphasized there is no way to guarantee the results of these models to policy makers, but data from previous outbreaks can be helpful. He also cautioned that human behaviors change quickly and modeling human behavior produces great variability, so models may not be able to provide public health officials with an estimate of an outbreak’s magnitude without a great deal of variability.

Castillo-Chavez also commented on Duchin’s question, illustrating two examples of behavioral responses related to the contagion effect. He described a study evaluating Internet activity following the case of Ebola in the United States. The study suggested that television news segments on imported Ebola cases led to significant increases in Ebola-related Internet searches and Twitter activity (Towers et al., 2015). A similar behavioral correlation was observed related to the incidence of copycat events following school mass shootings, although the mechanisms in that scenario remain unclear (Towers et al., 2018). He reiterated that the media indeed influences behavior and any contagion effect.

El Turabi commended the empirical evidence from such studies and urged for more social science research for infectious disease outbreaks. He cited the large amounts of research and development funds directed toward developing new vaccines in contrast to relatively low amounts spent to understand vaccine uptake behavior. El Turabi argued for a dramatic shift to quadruple the funding to build capacity in this area, and noted that current efforts from Wellcome Trust and the U.K. Department for International Development, who are building a platform for rapid social science research in the context of infectious disease outbreaks, is a starting point (DFID-Wellcome Trust, 2018).

The discussion then focused on technical aspects of modeling, particularly on incorporating data. Jennifer Gardy, associate professor at the University of British Columbia’s School of Population and Public Health, asked how to inform advanced parameters of utility functions that modify the models described in the session, considering there are already challenges with setting basic transmission and recovery rate parameters. She was interested in the types of data necessary for retrospective analysis, future model development, and real-time efforts when dealing with the next pandemic. Castillo-Chavez said that models can easily be fit to available data. What

is more important about models, however, is incorporating the influence of human decision making into these models, he said. As an example, he described the uncertainty around estimating social contacts during an influenza outbreak. He argued for a new approach that evaluates how people make decisions in relation to the evolving epidemiology of an outbreak and pointed out that people often change their behaviors in such a way as to mitigate the epidemic. He concluded that behavior is being ignored by current models, and that moving forward it should be incorporated into analysis related to a variety of relevant situations.

Also in response to Gardy’s question, Meltzer referred back to the example he highlighted in his presentation about naturally occurring experiments related to school closures in Texas during the 2009 H1N1 influenza pandemic (Copeland et al., 2013). He argued that this historical experience can be used to model and plan for future disease outbreaks. If quick answers are needed in the face of a new outbreak, he said assumptions must be made. The key is for these assumptions and their implications to be clear and straightforward when presenting the model to policy makers. He reiterated that models are not meant to provide accurate predictions of the future, but rather to describe the relationships and “levers,” or potential response actions, that influence the disease and human behaviors.

El Turabi highlighted methodological practice from CDC and the United Nations Children’s Fund that provides near real-time opinion polling in emergency outbreak scenarios for making decisions in the field. He noted that these preagreed frameworks and rapid assessment tools could feed back into emerging disease models. He also noted the need to build capacity to do these rapid polls in an ethically robust and reactive manner.

Anna Vassall, professor of health economics from the London School of Hygiene & Tropical Medicine, asked El Turabi about the potential for preepidemic data collection and experimental evidence, rather than relying on real-time measurements for post hoc models. El Turabi said that the tools for modeling are not restricted and should include preemptive evidence on behavior gathered prior to outbreaks, with the understanding that behaviors may change in the face of threats and uncertainty. He stated that he is skeptical regarding preferences being consistent during an outbreak. He shared the example of decision making for cancer treatments, stating that people who do not have cancer, when asked, want to be very involved in the treatment choice. This preference changes with the greater threat and uncertainty of an actual cancer diagnosis. He maintained that live analysis, during disease outbreaks, is also important.

Meltzer cautioned that there are limited resources available for real-time measurements particularly during large-scale outbreaks. In 2009, CDC measured the uptake of influenza vaccine during the pandemic through a cumbersome telephone interview as there was no data available on the char-

acteristics of those receiving vaccination. School closure analysis is similarly limited by the lack of a central registry of such events at the national or even state levels, so researchers rely on social media for data. He added that it is also difficult to collect data on when the schools reopen. At a certain point, he said, it takes immense person power to track and measure such data. Therefore, Meltzer said that selection of data to be measured should be judicious and prioritized, since there are insufficient resources to measure every possible variable, and not every parameter is equally valuable.

This page intentionally left blank.