Long before the emergence of COVID-19, policy analysts described pandemics as a “neglected dimension” of global security, because fundamental aspects of prevention and preparedness were (and continue to be) persistently under financed. Several high-level panels convened in the midst of the COVID-19 pandemic have called for large increases in global spending on health system strengthening, surveillance, and preparedness. These recommendations are vital, but contend with an entrenched pattern of panic and neglect—a strong tendency for sporadic health emergencies to spark short-term attention and investment, which tails off all too rapidly once the crisis has passed.
The slide from panic to neglect happens in part because policymakers are operating under uncertainty; they lack estimates of the probability of epidemics—including pandemics—that would enable them to prioritize preparedness for such events relative to other needs. As a result, epidemics and pandemics tend to be treated as inevitable and unpredictable phenomena rather than events for which decision makers can perform rigorous analysis, estimate costs, and prioritize investments.
Policymakers often choose to prioritize preparedness for high-probability events rather than rare ones, as the benefits of preparing for infrequent events are often invisible. To grapple with the threat posed by infrequent, severe epidemics, analysts should adopt a risk-based approach for decision-making, more akin to methods used for other natural catastrophes. Emergency planners and policymakers are accustomed to thinking about natural catastrophes, such as floods and earthquakes, in terms of their frequency and severity.
Adopting the most cost-effective strategies to prevent, prepare for, and respond to epidemics requires an understanding of their anticipated frequency and severity—that is, the level of risk that they pose. Interventions such as strengthening disease surveillance systems, investing in lab and diagnostic capacity, and developing new vaccine platforms, production systems and supply chains might have a modest benefit-cost ratio if severe pandemics are merely a “once in a century” risk, but—if the risk is substantially greater—might be very cost-effective strategies to protect global health and reduce mortality. Our risk estimates allow decision makers to approximate the necessary level of epidemic preparedness measures, and to identify which measures would be most cost-effective, both during and between epidemic periods.
While many pathogens are capable of sparking large infectious disease events (e.g., pandemic influenza viruses, Zika virus, coronaviruses, HIV, cholera, dengue virus, and more), we need to focus for framework for assessing the controllability of epidemics caused by different pathogenic threats. In that framework, risk of an uncontrollable epidemic increases with human-to-human transmission efficiency and decreases with detection probability. Therefore, we need to focus on a subset of pathogens that meet these criteria and comprise the majority of risk: respiratory diseases, notably those caused by pandemic influenza viruses and epidemic/novel coronaviruses. We also develop estimates for viral hemorrhagic fevers (VHFs), encompassing filoviruses (e.g., Ebola and Marburg viruses), and Nipah virus. These pathogens are of global concern as they have shown the potential for causing asymptomatic infection.
We also do not model risk from bio-terror (deliberate release of infectious agents) or bio-error (accidental release of infectious agents, for example from laboratory accidents), as this would require additional modeling efforts incorporating, for example, the characteristics, capabilities, and strategies of terrorist organizations, and biosafety protocols and practices within specific laboratories. These factors can be explicitly modeled and linked with the broader risk modeling framework that we present here, but are beyond the scope of our present analysis.
The epidemics can lead to many adverse outcomes—including infections, hospitalizations, deaths, societal disruption, educational delays, and economic shocks. The welfare losses caused by epidemics and pandemics—including economic damages as well as losses to education, livelihoods, and trauma and psychological damages—are considerable, but require distinct modeling techniques. We need to focus on deaths as they are the most readily measurable, observable, and reported metric, and therefore provide a less biased indicator of epidemic severity than other metrics such as infections or hospitalizations.
Risk modeling is not simply an exercise in mathematics —it must appropriately represent real world processes, and modelers should have familiarity with the complex web of underlying factors shaping the risk. Our modeling framework therefore explicitly incorporates several critical drivers of epidemic risk, including zoonotic spillover, global travel patterns, and governance challenges.
Nearly all modern pandemics have sparked when zoonotic pathogens have jumped from animals to humans, often through activities such as hunting, habitat encroachment, and intensive livestock farming. Multiple studies have shown that epidemics, especially those caused by zoonotic spillover events, are increasing in both frequency and severity. For a subset of high priority viruses, this trend is exponential, meaning that not only are epidemics becoming more frequent and more severe but that spillover-driven epidemics are occurring at an accelerating rate. Climate change and other forms of anthropogenic environmental change, such as deforestation and habitat fragmentation, are predicted to increase the frequency of zoonotic spillover events because they increase the frequency of contact between humans and animal reservoir species.
Increasing human population density and connectivity through global travel and trade facilitate the spread of the outbreaks. The accessibility of global air travel makes effective containment of emerging outbreaks increasingly difficult because infected individuals can disperse over large geographic distances before cases are detected and reported to public health officials. For example, rapid geographical spread was well-documented in the severe acute respiratory syndrome (SARS; caused by SARS-CoV-1) outbreak of 2003. One individual infected ten people in a Hong Kong hotel, six of whom took international flights to Australia, Canada, Singapore, the Philippines, and Vietnam. These traveling secondary cases subsequently led to SARS outbreaks in Hanoi, Singapore, and Toronto within a few days of the first reported case in Hong Kong (Cherry,2004). Similarly, during the COVID-19 pandemic, early detection of SARS-CoV-2 variants occurred in airline passengers. Spread by air travel also occurred during the 2014 West Africa Ebola epidemic.
Experience in infectious disease crises, such as COVID-19 and the 2018 North Kivu Ebola virusepidemic, has provided a clear reminder that governance and human behaviour play important roles in shaping infectious disease transmission. Research on the relationship between governance and epidemic risk suggests that political factors also play an important and under appreciated role in both frequency and severity of epidemics. Armed conflict and political instability can degrade disease surveillance systems, creating “blind spots” and lengthening the period during which disease transmission can occur before it is detected and mitigation measures are put in place. These same factors can also increase the risk of disease spread by facilitating population displacement. Public distrust of government institutions can also impede disease control measures, potentially leading to increased morbidity and mortality.
In particular the focus on losses in terms of deaths—also clearly represents a lower-bound estimate of total potential impact, since it does not include other sources of loss to human health and livelihoods (e.g., infections, hospitalizations, long-term sequelae, economic shocks, impacts on education, and societal disruption), nor does it include all sources of epidemic risk, such as vector-borne pathogens, bacterial infections, and viral threats presently unknown to science. As per research results, it’s also suggest that, among the diseases, respiratory diseases are the dominant driver of epidemic risk, with Viral hemorrhagic fevers (VHFs) representing a relatively modest global risk in terms of expected deaths. VHFs are deadlier on an individual level, but less prone to spread than respiratory diseases. However, the risk is not negligible, especially in Sub-Saharan Africa, and merits attention based on the direct and indirect impacts of these events on lives and livelihoods. Effective priority setting in global health requires the comparison of disparate burdens and risks, some of which operate on different timescales. As such, it may be helpful to understand how our expected mortality estimates compare to other risks. It might seem intuitive to compare epidemic Average Annual Loss(AAL) to the annual mortality burden caused by endemic diseases, as these both represent average deaths per year. For example, the average annual deaths we estimate for respiratory epidemics is comparable in magnitude to the annual number of deaths caused by routinely-occurring endemic lower respiratory infections—approximately 2.4 million deaths. However, when comparing such estimates, it is important to keep in mind that the underlying patterns leading to these averages are very different. While the average annual deaths from endemic diseases is made up of moderate levels of loss that occur regularly, the epidemic AAL represents much larger spikes in losses that occur sporadically, punctuating stretches of non-epidemic years. Mortality spikes caused by low frequency, high severity events are potentially more economically disruptive than regularly-occurring endemic disease, suggesting that, even when AALs may be similar between both types of diseases, planning efforts towards high-impact epidemics should at least be equal to, if not greater than, endemic diseases. Our model results also show that tail risk cannot be ignored. Low frequency, high severity events—the tail in our results—heavily drive expected deaths. It is all too easy to unconsciously discount the risk that the tail represents. The under representation of extreme events in small sample sizes can lead policymakers to underweight their probability, especially when relying on a limited and biased historical dataset. Moreover, the low annual probability of such extreme events tends to cause policy makers to round this probability down towards zero, due to cognitive biases that draw attention towards the frequency component of risk rather than the joint product of frequency and severity.
Over a 10-year view, an event on the scale of COVID-19 has a roughly 25% probability of occurrence; over the next 25 years, such an event has a likelihood roughly equivalent to a coin toss. These estimates demonstrate that future epidemic risk is more substantial than commonly believed, and that severe events are likely to occur much more frequently than “once in a century”.