Modelling in Public Health Research

How Mathematical Techniques Keep Us Healthy

E. Mansnerus

• Publisher: Palgrave Pivot
• Year: 2014
• ISBN:
  9781137298829
• Topic: Medicine
• Subtopic: Public Health, Administration & Care

This book analyses the development and use of mathematical models in public health research and policy. By introducing a life cycle metaphor, the author provides a unique perspective on how mathematical modelling techniques have increased our understanding of the governance of infectious risks in society.

1

Introduction: Life-Cycles of Models

Abstract: The use of mathematical models has increased and become more common in public health, recently. Along with monitoring activities, public health officials work with modellers in order to organise and analyse surveillance data, for example. In order to learn of the severity and spread of a pandemic outbreak, we can study different predictive scenarios. Mansnerus introduces us to the lives of models in both public health and other fields. She discusses how modelling has gained a dominance in current scientific research and policy-making.

Mansnerus, Erika. Modelling in Public Health Research: How Mathematical Techniques Keep Us Healthy. Basingstoke: Palgrave Macmillan, 2015. DOI: 10.1057/9781137298829.0005.

This book is about us. It tells a story of how our lives are affected by infectious risks and what we can do in order to manage them. However unwanted and unpredictable they are, we are not at their mercy. Mathematical models and simulation techniques have been developed and used to predict, prevent, and study infectious risks. They form invisible machinery behind convincing quantitative evidence needed for decision-making. Their use allows us to anticipate a pandemic outbreak, calculate how to best revise a vaccination programme, and understand how disease transmission happens in a population, for example.

Computer-based modelling and simulation techniques have gained increasing importance in knowledge production and use. They function as measuring instruments, experimental devices, and surrogate systems. They accommodate uncertainties in risk assessment processes and help overcome ethical and financial restrictions of population-wide experiments. Their predictive capacities are valued in climate research and infectious disease studies, for example. Yet, the limitations of model-based evidence have not been fully explored. We remain impressed by their fluency and forget to examine how the uncertain becomes certain through this computational machinery.

Climate research provides us with one of the most important examples in the development and use of modelling techniques. Already in the 1950s and 1960s, large-scale differential equation models were developed. The very same models form the core of the current climate models that are responsible for the pessimistic estimates of rising temperatures and melting glaciers. However, climate research is, par excellence, a case for the contested nature of model-based evidence. Research has shown that, partly because the facts of the anthropogenic cause of climate change were produced by models and simulations, they became contested, lobbied, ignored and rejected for a good number of years – which, unfortunately, delayed political interventions to reverse the undesirable development (Oreskes, 2011).

Modelling methods can be developed to produce knowledge in basic research, as in physics, but they may well reach into the interdisciplinary fields of environmental policy. This flexibility comes from two sources: from their manipulability, which means that as research tools they provide a way to question and inquire about the world, and from the fact that these techniques are developed in multidisciplinary teams and can accommodate the nuances of the different fields of enquiry. The development of modelling techniques in infectious disease epidemiology addresses both these characteristics. Driven by policy-relevant questions, such as ‘How do we best protect people against a pandemic outbreak?’ or ‘What would be the optimal vaccination schedule?’, development of modelling techniques is a way to produce new evidence in public health decision-making processes. Elementary for this development is the increase in computational power: from the early mathematical expressions of epidemic outbreaks to current simulation models on infectious diseases, computational capacity represents a key for the usability of modelling techniques and their adaptation as preventive and predictive tools in infectious disease epidemiology. This development is characterised as the life-cycle of models and simulation techniques.

Instead of providing a merely internal description of how models are built and how they function in epidemiological research, I will tell a story of their life-cycle – a story which combines the production of epidemiological evidence and its use in predicting infectious outbreaks. The metaphor of the ‘lives of models’ invites us to explore the emergence of modelling techniques in epidemiology, their maturation or ‘growing up’ through time, the ‘working life of models,’ and their seniority and potential passing away. These developmental stages in the life-cycle of the models form the core structure of the book. The metaphor itself binds together a developmental story that happens through time. It is an epistemic history that evolves from an ethnographer’s perspective on models and shows how different modelling methods relate to each other and form ‘families of models’ within epidemiological research. The metaphor is inspired by recent work that analyses various roles of computer models in environmental policy cycle (van Daalen et al., 2002). Contributions from history and sociology of science in the studies of biographies (Daston, 2002) or the lives of scientific objects (Creager, 2002) have been elemental for developing my understanding of the life-cycles of models.

By developing a life-cycle of models, I will discuss how mathematical methods, prior to the time when computer-intensive modelling methods entered epidemiological research, and current modelling methods, especially probabilistic transmission models, developed. I will construct a ‘family tree’ of transmission models built in Helsinki 1995–2001. This family tree not only introduces us to the interconnectedness of modelling approaches, their distribution across different research groups, but also shows how model-based evidence is disseminated across various models. This will lead us to study the ‘working life’ of models.

The working life of models means studying their uses in the governance of infectious risks. The predictive capacities of models are a significant way to anticipate risks and examine the applicability and effectiveness of mitigation strategies during a pandemic outbreak. I will identify two types of predictions: explanation-based and scenario-building ones. These predictions serve especially in pandemic risk assessment as tools to quantify, that is, to express through numerical representations how a pandemic potentially progresses. When they function as tools that can fluently express the uncertainties related to an infectious threat, we will see the limitations of models. These limitations, for example, manifest in models’ capacity to represent the social aspects of a situation. But the availability of data can also present limitations, as most models we discuss in this book depend on data. These shortcomings, I will suggest, are to be taken into account when models are used as technical aids in the governance of public health risks.

The successful working life of models draws on the development of modelling methods to predict and prevent a measles outbreak in the United Kingdom. Serving along with other sources of evidence, models interpret surveillance data and confirm a potential outbreak that can be prevented by a booster vaccination campaign. Their success may not always last. However useful they are in optimising strategies of antiviral distribution during a pandemic, they at best predict the future by modelling the past. During an outbreak, models encounter silent, absent evidence. Data may not be available, yet as techniques, they try to bridge the gaps in order to alleviate our ignorance. These limitations, although acknowledged by the modellers, may not be communicated to users. Current tendencies to trust in numerical evidence invite us to increase model use and neglect their restrictions.

Following the metaphor according to which models ‘live their lives and grow old,’ we will look at models as senior experts. With this focus, the seniority of models turns into technical rationality not only to predict public health risks, such as an outbreak of pandemic influenza, but also to plan and optimise preventive actions, such as vaccinations. The ‘analytics of governance,’ which takes into account a broad set of factors, is addressed as a way to balance the benefits of these techniques without neglecting their limitations. This broader framework, adopted from Michel Foucault, is a way to reverse the development that leads us to be governed by the numbers. I will especially focus on the technical rationality of governance that models represent, and extend the question to power relations between those who govern and those who are governed, which is of special interest when designing preventive interventions, such as vaccination strategies in public health policy-making. This framework of analytics of governance, which takes into account the ways in which knowledge is formed through expertise, and how social identities are shaped in conjunction with technical development, is a way to broaden our perspective beyond the fascination of the technicalities of models and critically assess what lies underneath their authority.

Through the lives of models framework, we will learn how mathematical techniques grow in the service of public health. Their successes and failures or shortcomings can be seen, after all, as aspects of life. But the lives of models are intertwined with our lives. Model-based predictions of outbreaks do not remain as scientific findings in journals; rather, they turn into action and mobilise vaccination programmes. When models enter as experts into the policy domain, they are welcomed as well as criticised. They are welcomed for their capacity to tell stories that inform the next steps in the decision-making process, but criticised for their limitations. In conclusion, this book reflects the tension that remains between the increased use of models as senior experts and the potential shortcomings if their advice is accepted uncritically.

Structure of the book

The structure of this book is as follows. Chapter 2 introduces a narrative framework that shows how models are developed to address policy-motivated questions. Storytelling is a helpful metaphor in order to understand how models are developed historically, what kind of elements are involved in building a model and how these historical accounts can be relevant for our current understanding of models.

Chapter 3 highlights the relatedness of models and how they form ‘kinship’ relations, as I call the organic development of models that is informed by achievements in collaborating research groups. By establishing the relatedness of models, this chapter will analyse how model-based evidence is disseminated across research groups and to the policy domain. Through this analysis, we will learn about the nature of model-based evidence.

In the fourth chapter, the focus is on the working life of models. The key is to analyse how models operate in the context of public health policy. Through the example of revising and implementing the MMR-vaccination strategy in the United Kingdom, this chapter contributes to the debate on the performativity of models.

Models can be built upon available data from previous pandemics, thus, modelling the past in order to predict the future. This will be the focus in Chapter 5. As we learned from the 2009 A/H1N1 (swine flu) outbreak, modelling and simulation techniques were widely used during and prior to the pandemic. Pre-pandemic modelling became a way to encounter possible pandemic risks and to assess effective mitigation strategies. This chapter defines two types of model-based predictions: explanation-based and scenario-building. By discussing their reliability, we learn what kind of limitations modelling techniques face.

During the 2009 pandemic outbreak, modelling took place in a reality that was under time pressure. The sense of urgency to make decisions was challenged by the lack of data. Evidence was silent and weak, and modellers did their best to bridge the gaps. In Chapter 6, the main interest is to analyse ‘known unknowns’, factors of which we have very limited understanding at the beginning of modelling. These factors can be related to the microbiology of the pathogen or to the safety of the pharmaceutical interventions, for example. This chapter looks at how modelling methods alleviate unknowing in the context of pandemic risk assessment.

How do numbers govern the world? How is the authority of computational techniques shaped? Model-based evidence, as shown in previous chapters, is an important part of the whole body of evidence upon which pandemic predictions or vaccination schemes are based. Models, when they function as an evidence base, turn into instruments of governance. Their authority is likely to make us believe in the numerical representations they produce. They act as senior experts guiding and governing health risks, as Chapter 7 will show through a case study on animal health modelling.

When models gain seniority as experts in public health, their lives become intertwined with ours. They not only provide estimates and predictions but also turn them into action. Vaccination policies are renewed, as we learned in the case of measles modelling in the United Kingdom. Yet we know that the modelling process can be described as fluctuation between simplicity and complexity. Chapter 8 discusses these issues. Policy needs are simplified in order to be successfully addressed in models, and model-based evidence faces a heterogeneity in which its recommendations may be resisted and disputed. This fluctuation is best captured in the life-cycle of models. This framework accommodates both the critical voices that warn us about overreliance on model-based evidence and the supportive ones that remind us of their beneficial use. After all, we use models to overcome ethical and financial restrictions we face when making sense of infectious risks that affect us all.

2

Models and the Stories They Tell Us

Abstract: Models tell stories that can answer vital questions. Models have become essential tools for epidemiological research. How well they address policy-driven questions in public health is explained. In epidemiological research, the mathematical predecessors for current computer-based modelling techniques have a long history. They aimed at policy advice by providing mathematical representations of infectious patterns in populations.

Mansnerus analyses the early development of probability theory, showing why modelling became beneficial in governing infectious risks. How models are tailored to meet policy needs is revealed through an analysis of interdisciplinary collaboration. Models and mathematical techniques become accessible for us when we see them through the metaphor of storytelling: the transmission of an infection is presented to exemplify this.

Mansnerus, Erika. Modelling in Public Health Research: How Mathematical Techniques Keep Us Healthy. Basingstoke: Palgrave Macmillan, 2015. DOI: 10.1057/9781137298829.0006.

‘A man is always a teller of tales, he lives surrounded by his stories and the stories of others.’

(Jean-Paul Sartre, 1938, in Bolton 2010)

2.1Introduction

Time and again we are challenged by new emerging infections and witness their victories amongst us. When the A/H1N1 (Swine flu) outbreak hit the headlines in 2009, we learnt of its rapid spread from country to country, how many people became severely ill, and how many died. Along with these loud narratives, the silent ones whisper worries about common childhood infections.¹ All these stories share something in common: they voice concerns, demand change, and respond to policy calls. They reach beyond the natural habitat of scientific journals, where their findings may be buried and forgotten. After all, these stories are told in order to keep us healthy.

We do not ‘live surrounded by stories’ passively; we use them to make sense of things.² How to improve vaccination uptake? What is a reliable estimate for the spread of a pandemic outbreak? Answers to these questions as told by both historical and current narratives³ help us understand what models are and how they are used. Their capacity to integrate evidence from various sources, interpret it and express it mathematically explains why modelling techniques became dominant in th...

Table of contents

• Cover
• Title
• 1  Introduction: Life-Cycles of Models
• 2  Models and the Stories They Tell Us
• 3  Kinship Relations of Models
• 4  Working Lives of Models
• 5  Encounters with Risks
• 6  When Evidence Is Silent
• 7  Governing by Numbers
• 8  Lives of Models in the World of Policy
• Glossary
• References
• Index