ANALYSIS | Applying chaos theory to epidemic crisis management
The COVID-19 pandemic demonstrated how difficult it can be to predict the spread of an infectious disease in a society where human behaviour, government decisions and the capacity of healthcare systems are closely interconnected. The course of an epidemic is determined not only by the biological characteristics of the pathogen, but also by the environment in which it spreads and the responses of individuals and institutions. Consequently, the spread of an infectious disease can become a rapidly evolving process for which simplified mathematical models may not always provide adequate predictions.
What is chaos theory?
Chaos theory is a branch of mathematics concerned with the behaviour of complex dynamical systems. One of its central concepts is deterministic chaos – a phenomenon in which the evolution of a system is governed by mathematical rules, yet its long-term behaviour can be extremely sensitive to initial conditions. The American meteorologist Edward Lorenz is regarded as one of the pioneers of chaos theory. In a study published in 1963, he demonstrated that even a simplified weather model could behave in such a way that small differences in initial conditions would lead to markedly different outcomes over time.1 This subsequently gave rise to the familiar metaphor of the butterfly effect. However, this does not mean that the flap of a butterfly’s wings will inevitably cause a storm or that every minor error will lead to a catastrophe. Chaos theory describes situations in which the evolution of a system is so sensitive to its initial conditions that the accuracy of long-term predictions progressively diminishes owing to limitations in measurement. Apparent randomness, therefore, does not necessarily imply an absence of underlying rules, but may instead reflect complex dynamics that are difficult to predict accurately in practice.
Chaos theory has been applied in meteorology, ecology, astronomy and the study of population dynamics, but its principles are also relevant to epidemiology. The spread of infectious diseases is influenced simultaneously by the characteristics of the virus, interpersonal contact, population immunity, behavioural patterns and disease control measures implemented by governments. These factors do not operate independently but can interact in ways that make the course of an epidemic difficult to predict. For instance, government-imposed restrictions may alter human behaviour, which in turn affects the spread of infection and the pressure placed on healthcare systems. Growing demand for hospital treatment may compel governments to introduce further measures, which subsequently affect the economy, education and everyday life. In examining these relationships, approaches informed by chaos theory may complement conventional epidemiological models by helping to explain the potential non-linear behaviour of the system. Nevertheless, it is important to distinguish mathematical chaos from broader forms of complexity, uncertainty and randomness, as not all processes that are difficult to predict are necessarily chaotic.
How can small differences alter the course of an epidemic?
One of the key indicators used in epidemic crisis management is the reproduction number, or R number, which describes the average number of people to whom an infected individual passes the infection. When the reproduction number is above one, infections may increase, whereas a sustained value below one generally leads to a decline in new infections. In practice, the reproduction number varies over time and depends on population immunity, patterns of social contact and the measures in place. Consider a simplified example in which an epidemic begins with 100 new infections. If each infected individual passes the infection to an average of 0.95 people, there would be approximately 46 new infections after 15 successive generations of transmission. If the corresponding figure were 1.05, there would be approximately 208 new infections by the same point. Although the difference between the reproduction numbers is only 0.1, the two scenarios produce markedly different outcomes. This is a mathematical illustration rather than a forecast of an actual epidemic, as it does not account for changes in immunity, human behaviour or other epidemiological factors. The example primarily illustrates the sensitivity of exponential growth rather than deterministic chaos itself, but it helps explain why relatively small differences in disease transmission can become highly significant for crisis management over time.
The application of chaos theory to epidemiology makes it possible to investigate even more complex non-linear relationships. Mangiarotti et al. (2020), for example, employed a modelling method based on chaos theory to analyse the spread of COVID-19 in China, Japan, South Korea and Italy.2 Their research demonstrated that epidemiological data could be used to construct alternative mathematical models capable of describing transmission dynamics and possible future scenarios. Postavaru et al. (2021) also investigated non-linear and chaotic components in the modelling of COVID-19 transmission, drawing on epidemiological data from countries including China, South Korea, Italy and the United States.3 The value of such approaches lies in their ability to explore patterns of disease transmission even when comprehensive information about all the factors influencing the system is unavailable. However, the mathematical sophistication of a model does not in itself guarantee that its predictions will be more accurate than those produced by conventional epidemiological models. The purpose of these approaches is not to predict the future with absolute certainty, but to improve our understanding of possible trajectories and support decision-making under conditions of uncertainty.
One choir rehearsal, dozens of infections: how does a superspreading event occur?
A striking example of the complexity of COVID-19 transmission comes from a choir rehearsal held in the US state of Washington in March 2020. On 10 March, 61 people gathered in Skagit County for a rehearsal lasting approximately two and a half hours. Among those attending was an individual who was already experiencing symptoms of the disease. An investigation by the US Centers for Disease Control and Prevention (CDC) identified 52 secondary cases associated with the rehearsal, of which 32 were laboratory-confirmed and 20 were classified as probable cases. Three people were admitted to hospital and two died.4 The incident highlighted the importance of the setting in which transmission occurs, the duration of interpersonal contact and the nature of the activity involved. A relatively small group of people spent an extended period in close proximity, whilst singing may have facilitated the spread of virus-containing aerosols. Although the choir rehearsal does not demonstrate deterministic chaos, it provides a documented example of transmission heterogeneity – a situation in which a single event can result in a disproportionately large number of infections.
From a crisis management perspective, such an incident raises the question of how events with a high potential for transmission can be identified and contained before their consequences extend into the wider community. A single cluster of infections may increase demand for testing, contact tracing, isolation and healthcare services. If several superspreading events occur simultaneously, their combined effects may place additional pressure on the entire healthcare system and necessitate further interventions. Average epidemiological indicators alone may not always reveal the settings in which transmission is most intense. Attention must therefore also be paid to where transmission occurs, the conditions under which it takes place and the activities that may amplify it disproportionately. The experience of the Skagit County choir rehearsal demonstrates that understanding the risks associated with particular activities and environments can be just as important for crisis management as monitoring overall levels of disease transmission.
What can the 1918 influenza pandemic teach us?
The 1918–1919 influenza pandemic also offers valuable insights into the importance of timing in crisis management. Markel et al. (2007) examined the experiences of 43 US cities, focusing on the measures introduced to limit the spread of infection and the relationship between the timing and duration of those measures and the consequences of the epidemic.5 The study identified an association between the early introduction of multiple, sustained non-pharmaceutical interventions and lower mortality. St Louis, for example, introduced measures to limit transmission relatively promptly, and its peak mortality during the first wave remained considerably lower than that of Philadelphia. In Philadelphia, interventions were introduced later, and the first wave proved more severe. The study does not allow all differences between the cities to be attributed solely to the timing of restrictions, as demographic characteristics, local circumstances and other factors may also have played a role. Nevertheless, the comparison illustrates an important principle of crisis management: the potential effectiveness of an intervention depends not only on what is done, but also on when and how consistently it is implemented.
The experience of the 1918 influenza pandemic is particularly noteworthy because it illustrates the relationship between crisis management and epidemiological dynamics over time. An intervention introduced during the early stages of an outbreak may influence the development of transmission chains before the number of infections reaches a critical level for the healthcare system. Intervening later may require considerably greater resources, and the effects of those measures may take longer to materialise. Historical findings cannot, however, be directly extrapolated to contemporary infectious diseases, given differences in pathogens, healthcare systems, population structures and social conditions. Such comparative studies nevertheless reinforce the need to assess crisis management measures in relation to their timing, duration and local context. From the perspective of chaos theory, this provides a broader illustration of how changing conditions can influence the evolution of a complex system, although the historical comparison itself does not establish the presence of mathematical chaos.
The Ebola epidemic: chaos theory as a modelling tool
The application of chaos theory to infectious disease research did not begin with the COVID-19 pandemic. Mangiarotti et al. (2016) examined the transmission dynamics of the West African Ebola epidemic between 2013 and 2016,6 drawing on data published by the World Health Organization and the health authorities of the countries concerned, with a particular focus on recorded infections and deaths. The researchers constructed a four-dimensional dynamical model that enabled them to analyse the evolution of the epidemic, distinguish between its different phases and investigate the chaotic properties of the system. One of the most noteworthy aspects of this approach was its ability to derive mathematical relationships from observed time-series data without requiring prior knowledge of every relationship governing the system. Such an approach may be particularly valuable in epidemiology, as decision-makers confronted with an emerging infectious disease rarely have comprehensive information about its characteristics, transmission mechanisms or future patterns of human behaviour.
The experience of modelling the Ebola epidemic highlights the potential of methods informed by chaos theory to complement epidemiological analysis, although their applicability must be assessed critically. The ability of a particular mathematical model to describe the transmission dynamics of one epidemic does not automatically mean that it will be suitable for another infectious disease, country or period. The usefulness of a model is influenced by factors including data quality, the characteristics of the pathogen and the assumptions made during the modelling process. From a crisis management perspective, the value of such methods may lie in their ability to identify patterns of change in data that would be more difficult to detect using conventional analytical tools. Nevertheless, mathematical models cannot replace epidemiological expertise, the practical knowledge of institutions or political accountability for decisions. Models based on chaos theory should instead be regarded as an additional tool within a broader analytical framework supporting decision-making.
Models cannot answer every question
At the beginning of the COVID-19 pandemic, epidemiological modelling became an important crisis management tool. Models were used to assess the potential spread of infection, demand for hospital treatment, pressure on healthcare systems and the possible effects of different disease control measures. It is nevertheless important to recognise their inherent limitations, as their results invariably depend on the underlying data, mathematical assumptions and purpose of the modelling exercise. According to Stuart et al. (2020), those using epidemiological models must understand their purpose, assumptions and the quality of the data on which they are based.7 For instance, one model may address the number of hospital beds likely to be required in the coming weeks, whilst another may assess possible transmission patterns in the absence of further restrictions, and a third may compare the theoretical effects of different interventions. These are distinct questions requiring different assumptions. Problems arise when a conditional projection is treated as a definitive prediction of the future, without recognising that actual circumstances may differ from those on which the projection was based.
The reliability of models is also significantly affected by shortcomings in the underlying data. Schneider (2020) drew attention to deficiencies in US COVID-19 testing data and the resulting difficulties in assessing the true scale of the pandemic.8 Where testing capacity, data collection methods or case reporting practices differ between regions, comparisons of epidemiological data may prove misleading. Delays in publishing data may also result in crisis managers making decisions based on assessments of an epidemiological situation that has already changed. Even a mathematically precise model constructed from inadequate data may therefore fail to provide a reliable picture of actual circumstances. Data quality can be particularly problematic during the early stages of a crisis, when reliable information is scarce but the need to make decisions is pressing. Crisis management arrangements must therefore allow projections to be updated promptly as new data become available and previous decisions to be reconsidered where necessary.
Why do the same solutions not work in every country?
One of the important lessons of the COVID-19 pandemic was that the outcomes of measures introduced in different countries are not easily compared. Gibney (2020), for example, examined scientists’ efforts to establish which measures had been most effective in containing the pandemic.9 Such assessments are complicated by differences in population characteristics, healthcare systems, testing practices and the implementation of restrictions. Several measures are often introduced simultaneously, making it difficult to isolate the effects of individual decisions. For instance, transmission may decline as a result of restrictions on movement, voluntary changes in behaviour, the adoption of remote working and other concurrent factors. If the number of infections subsequently falls, that observation alone cannot establish how much of the change can be attributed to a particular intervention. Mathematical modelling can help investigate these relationships, but it does not remove the need for critical interpretation, nor does it necessarily make it possible to disentangle every factor operating simultaneously.
McBryde et al. (2020) similarly emphasised that epidemiological models can support both healthcare resource planning and the assessment of public health interventions, but that decision-makers must consider their limitations and the wider societal consequences of their decisions.10 A measure that proves effective in one country may not produce the same results elsewhere, as social and institutional conditions can differ considerably. Public trust in government institutions, healthcare capacity and willingness to comply with restrictions, for example, may all influence how a particular measure works in practice. Crisis management cannot therefore rely solely on adopting solutions deemed successful in other countries; their suitability must also be assessed within the local context. From the perspective of chaos theory and the broader study of complex systems, it is important to recognise that similar interventions may not produce identical outcomes under different conditions and that the consequences of decisions often depend on the interaction of numerous interconnected factors.
The consequences of an epidemic extend beyond healthcare
One of the complexities of epidemic crisis management is that infectious diseases affect not only healthcare systems but can also cause widespread disruption throughout society. Illness may reduce workforce availability and business productivity, whilst restrictions on movement may affect education, transport, commerce and social services. The reorganisation of healthcare services may also affect the ability of patients with other medical conditions to receive timely treatment. Boon et al. (2021), for instance, examined the impact of the COVID-19 pandemic on radiation oncology and the need to reorganise treatment, adopt remote services and adapt existing working practices.11 This example illustrates how changes introduced to address one health crisis may also affect people who have not contracted the infectious disease in question. If, for example, hospitals restrict elective treatment to reduce the spread of infection, this may free up healthcare professionals and hospital capacity to address the immediate crisis, whilst delaying treatment for other patients. Should such delays lead to a deterioration in patients’ health, their subsequent treatment needs may increase, placing additional pressure on the healthcare system.
Such chains of events illustrate the systemic interactions associated with crisis management, the potential scale and duration of which may be difficult to assess fully at the outset of a crisis. This does not, however, mean that every reorganisation of healthcare services will result in the adverse consequences described, as the actual effects depend on the scope, duration and circumstances of the intervention. The value of chaos theory and broader approaches to understanding complex systems lies here primarily in recognising that crisis management decisions can have both direct and indirect consequences across different areas of society. When assessing the potential effects of decisions, attention must therefore be paid not only to the immediate objective, such as reducing disease transmission, but also to how the chosen intervention may affect the provision of other public services, the well-being of the population and society’s long-term capacity to recover. Such an approach requires cross-sectoral co-operation and the ability to assess the evolution of a crisis as an interconnected process.
What can crisis management learn from chaos theory?
Chaos theory does not provide crisis managers with a definitive answer as to which intervention should be implemented next, but it offers an additional perspective on understanding complex and rapidly evolving situations. First, it is necessary to recognise that decision-making often begins under conditions of incomplete information, meaning that crisis management cannot rely exclusively on a single projection or an unchanging plan of action. Secondly, the timing of decisions must be considered, as an intervention during the early stages of an epidemic may alter transmission dynamics before more resource-intensive measures become necessary. Thirdly, it is important to understand that different effects within complex systems can reinforce or counteract one another, and that disease transmission, pressure on healthcare systems, human behaviour and economic consequences are not independent phenomena. Crisis management arrangements must therefore allow decisions to be revised in the light of new information, as one of the principal benefits of modelling lies in its ability to explore different scenarios and compare their potential consequences.
Post-crisis learning is equally important and should not be confined to determining whether the final outcome was the one intended. It is also necessary to examine the assumptions that prevailed at the time decisions were made, the information available to decision-makers and how alternative courses of action might have influenced the evolution of the crisis. Such analysis can help identify shortcomings in both pre-crisis preparedness and decision-making arrangements, thereby improving organisations’ capacity to prepare for future crises. The objective of crisis management cannot be to eliminate uncertainty altogether, but rather to develop the capacity to make informed decisions even when uncertainty persists.
A perspective informed by chaos theory helps to illuminate the limitations of long-term predictions, and its integration with epidemiological modelling, systemic risk analysis and adaptive crisis management may support better-informed decisions. The COVID-19 pandemic and previous infectious disease outbreaks demonstrate that the evolution of crises is influenced by biological factors as well as societal and institutional choices. The better we understand the interactions between these factors, the greater our capacity to identify emerging crises, respond to changing circumstances and mitigate their impact on society.
Suggested citation:
🟠 Nagel, A.-M., & Kostytska, S. 2022. ANALYSIS | Applying chaos theory to epidemic crisis management. 22.04.2022, Crisis Research Centre.
🔺The use, distribution, modification or publication of the text and images is permitted only with the prior written agreement of NGO Crisis Research Centre; any unauthorised use is prohibited.
Sources
2 Mangiarotti, S., Peyre, M., Zhang, Y., Huc, M., Roger, F., & Kerr, Y. 2020. Chaos theory applied to the outbreak of COVID-19: An ancillary approach to decision making in pandemic context. Epidemiology & Infection, 148(e95).
Share post: