Kari Auranen and Tuija Leino
The Institute for Health and Welfare (THL), Helsinki, Finland
From modellers’ point of view: How to communicate with the policy community?
The challenge of communicating expert opinion and recommendations to public health decision makers has remained. In Finland, decision making regarding the implementation and the surveillance of the effectiveness of the vaccination programme is a convoluted process, including several levels of decision making by committees with expertise ranging from epidemiology and health economic evaluation to general health policy at the government level. When considered necessary, the expert opinions provided by vaccine-specific working groups at THL have included model-based, quantitative predictions of the likely effectiveness and cost-effectiveness of vaccination. In this talk, we review three recent cases in which such model-based predictions have been used to support policy recommendations for vaccination against varicella, pneumococcal infections, and the human papillomavirus. Although the recommendations had restricted uncertainty into scenarios (e.g. by providing worst-case scenarios), rather than presenting full statistical variability in the predictions, decision makers often seem unwilling to act in face of uncertain future. More generally, the long chain of decision-making means that even a well-informed quantitative synthesis of the impact of a health intervention does not necessarily turn into policy.
Robert Dingwall
Dingwall Enteprises and Nottingham Trent University
How do numbers govern the world?
A problem for all human societies is that actions in the present have implications for the future which is necessarily uncertain. Those actions can be highly consequential. At best, they consume resources that could be used in ways that would be more efficient, effective, equitable or humane. At worst, they can lead to catastrophe. Consequently, there is much interest in generating visions of the future that provide a stable basis for present decisions. This aspiration is part of the founding impetus for the social sciences – to discover laws that would predict human behaviour at either an individual or a social level. From this have emerged fields like epidemiology, demography and economics, which share the assumption that models constructed from measurements of variables in the present, and descriptions of the relationships between them, can be projected forward and used to constrain the possible pathways to be followed by collectivities at all levels from organizations and markets to entire societies. This process has established a high degree of legitimacy as a fundamental tool for contemporary governance, whether of states, markets or corporations. However, it has also come to be recognized that assurance may come at the expense of resilience because of its failure to embrace critical sources of uncertainty: the quality of the input data; the diversity of human behaviour, which often deviates from modelling’s techno-rationalist assumptions; the risk of high-impact, low-probability events; the disruptive innovation that destroys all previous assumptions about the future. As a result, significant social actors have been increasingly interested in exploring qualitative ways of thinking about possible futures. These vary from scenario planning and gaming through to creative works of speculative fiction, art and film. These investigations draw attention to the normative assumptions embedded in the modelling programme and the way in which it contributes to the depoliticization of contemporary politics. Modelling must be complemented by other ways of generating plural visions of the future. There is a particular challenge to the qualitative social sciences to develop methods for achieving this in a systematic fashion.
Nigel Gibbens
UK CVO, Defra
Governance of risk in public policy
Epidemiological modelling may be used, alongside other evidence, to aid the development of animal health policy: to support disease surveillance activities; to evaluate existing or new intervention strategies to control disease or reduce the risk of incursion or spread. In order to use model outputs effectively, the decision maker has to consider a range of issues.
Critical evaluation of models and their outputs is essential; not to prove that a model is ‘true’, or an exact replica of reality, but to give confidence that it may be used to inform decision making. We need to understand the assumptions made and the uncertainties around the data used. If there is a complete lack of data, how has the gap been filled and what does that mean for the reliability of the outputs? And how do we address the impact of uncertainty, and ensure models support further analysis, such as economic impact assessment?
Finally, how should model outputs be communicated so that they give clear, relevant and timely insights that can be used in the decision-making process? How can we explain uncertainties, like wide confidence intervals, effectively to Ministers and stakeholders?
Gabriele Gramelsberger
Freie Universität Berlin
Simulation in climate research and policy
In 1979 meteorologist Jule Charney and colleagues published a globally recognized report on Carbon Dioxide and Climate: A Scientific Assessment. The authors finished the report with the conclusions that “our best estimate is that changes in global temperature on the order of 3C will occur [until 2100] and that these will be accompanied by significant changes in regional climatic patterns”. The estimates of the so-called Charney report were based on two, at that time state-of the art, general circulation models of the atmosphere that carried out numerical studies on the impact of doubling carbon dioxide on the global mean temperature. The 1979 Charney report marked a watershed that transformed climate change into a public policy issue—interlinking climate science and politics. This interlinking has led to an outstanding development in both domains: policy and climate modeling. The paper explores the outcome of this development, which has created a unique infrastructure for ‘modelling policy’ (IPCC Intergovernmental Panel on Climate Change, Earth System Models, Downscaling, Model Intercomparison, etc.).
Helen Lambert
University of Bristol
Evidentiary persuasions: Policy advocacy and plural forms of evidence
This paper considers the political use of models for policy advocacy and public health priority-setting in relation to evidentiary salience, drawing on examples from UK (obesity, screening) and South Asia (HIV, NCDs). As forms of (simulated) ‘evidence’, modelled projections of disease spread or intervention success may be peculiarly persuasive as rapid advocacy tools. This rests in part on the kinds of social assumptions required for modelling. Inaccuracies arise not just from missing information but from the nature of assumptions, and priorities are not always reappraised when projections are revised on the basis of more accurate information. The predominance of biological and behavioural models (of disease) over sociological models (of human groups) affects the transferability of epidemiological scenarios across settings. I consider the potential implications for generating reliable inferences of incorporating other forms of evidence (qualitative, ethnographic) into public health decision-making.
Melissa Leach
University of Sussex
Outbreak narratives and silent voices: The politics and anti-politics of modelling in dealing with pandemic threats
Around any actual or threatened disease outbreak, multiple narratives can be constructed – stories about the origins of the problem, why it matters and to whom, and what should be done about it. More than just stories though, narratives often underwrite and justify particular policy responses whose real, material effects involve winners and losers. There is a politics around which – and whose – pandemic narratives come to dominate, and whose remain silent and excluded. Drawing especially on the examples of influenza and haemorrhagic fevers, I will highlight contrasts between global ‘outbreak narratives’ that portray emerging viruses leading to global pandemics and requiring rapid, international response measures, with alternative and often marginalised narratives – including those of people living with disease in located African settings. Such selective narratives call upon particular forms of knowledge – whether global or local, biomedical or ‘traditional’, epidemiological, clinical or social-ecological. They label particular people and populations in contrasting ways – as heroes or victims, disease-spreaders or controllers, or those whose livelihoods and lifestyles influence disease transmission, for good or ill. And they draw on very different approaches to modelling. Whereas large-scale epidemiological models are pivotal to the construction and legitimacy of global outbreak narratives, alternative narratives imply – even if implicitly – other sorts of ‘models’ – of local ecology-health interactions, of cultural understandings and responses, or of wider socio-economic and environmental drivers, for instance. I will reflect on the political, as well as technical, challenges of engaging such diverse forms of modelling in attempts to understand and respond to pandemic threats in more inclusive, rounded and socially just ways. And I will also explore how modelling can act as a kind of anti-politics, erasing the inherent social and political in pandemic policy challenges by re-casting them in technical terms.
Erika Mansnerus
LSE
Working life of infectious disease models: A case study of predicting a measles outbreak in the UK
When models live their lives they grow up and enter working life. They leave behind the sheltered world of research where they serve as scientific instruments, measuring devices, virtual experiments or representations of the world. They enter a new domain of use, where they are no longer close to the modelers, researchers, or instrument makers. Rather they stand on their own to disseminate reliable and usable evidence across research communities and policy domains.
This presentation discusses the working life of infectious disease models through a case study of predicting a measles outbreak in 1994 in the UK. The study shows how modeling was developed to predict a measles outbreak and how this prediction supported a decision on a ‘booster’ vaccination campaign in order to prevent the outbreak. The analysis is based on archived documents and informal conversations on measles modeling at the Health Protection Agency, UK.
Angela McLean
University of Oxford
Understanding the Evolution of Emerging Infections
Recent decades have seen several events where new emerging infections have arisen and spread around the globe. HIV, SARS and H1N1 influenza are the most prominent examples. Their arrival raises questions about what other new infections lie over the horizon, whether they will cause us severe problems and how we might best detect and control them. In this talk I will discuss a range of different uses of mathematical models for understanding the evolution and spread of novel emerging infections. Using data from HIV and H1N1 influenza I will draw examples of how I think models are useful for extracting the maximum amount of information from diverse data sources. I will go on to use an abstract model of an imaginary infection to illustrate how models can be useful for thinking through what new information may (or may not) be helpful. In a final example I will discuss a model about the spatial distribution of infection that I believe helps thinking about the surveillance and control of emerging infections. I will not be presenting models that claim to predict the future course of events. Instead I will be presenting examples where I believe mathematical models can help us understand the data we already have and think clearly about what other data we need.
Mary S. Morgan
London School of Economics and University of Amsterdam
From the world in the model to the model in the world
Scientific models offer small, abstract, accounts of the world, within which scientists explore their ideas about the world, and hope to learn something about the real world in the process. Scientists both enquire into the world of the model, and enquire with the model into the real world. While these are very different aims, the mode of enquiry can be understood as a form of experiment. Such an account immediately highlights the problem of making inferences from model experiments, particularly to the world that the model represents. This in turn focusses on the representing qualities of models that might make them useful objects for working with in the world, where comparisons with other forms of succinct representation suggest that both cognitive skills and imagination are needed for scientists to use models to remake the world through policy interventions.
Angus Nicoll
European Centre for Disease Prevention and Control and the London School of Hygiene and Tropical Medicine
Using and Developing Models for Infectious Disease Policy – The Example of Influenza
The applications of modelling to infectious diseases are legion. By focusing on the influenzas – animal, seasonal and pandemic, it will be shown how various form of modelling makes many contributions to determining policy and practice in mitigating the impact of one of the most explosive of the infectious disease threats to man and animals. For more static disease control, annual influenza immunisation and use of antivirals, modelling including economic information can indicate where limited resources should be best applied. The application of modelling in preparing for pandemics is important. However it is in the initial assessment and determining optimal responses when pandemics emerge that modelling really comes into its own. At the same time it will be demonstrating how dependent modelling is on virological, behavioural, clinical and epidemiological data if it is to produce useful results. Results also have to be considered with careful reality checks and in relation to practicalities and feasibilities for proper policy development. Finally there is the exciting new area of viral risk assessment determining through multidisciplinary approaches which are the animal influenza for which diagnostic tests and vaccines should be prepared. As a postscript reference will be made to the importance of allowing the scientific method and peer-review to take place with mention of recent A(H5N1) papers.
Sabine Roeser
University of Delft
Philosophical dilemmas of modelling risks and uncertainties
Philosophical limits to modeling risks and uncertainties: the case of values and emotions
Quantitative approaches to modeling risk and uncertainties face severe philosophical limitations. Quantitative methods fail to address important ethical considerations such as justice, fairness, equity and autonomy. They fall prey to ‘complexity neglect’: it is difficult to capture ethical aspects of risk and uncertainty in models. This presentation argues that quantitative approaches have to be supplemented by emotional considerations, as they reveal ethical aspects of risk. This suggestion runs counter to the standard approaches to decision making under uncertainty, according to which emotions are a source of irrationality that need to be banned or at most be acknowledged for democratic reasons. This presentation will argue that emotions such as compassion, empathy and enthusiasm point out ethical aspects of risk that cannot be sufficiently be captured by formal models. Emotions can also play an important role in communicating about infectious diseases and vaccination. For example, one can entice sympathy and compassion with potential victims that can counteract potential free-rider behavior of people. Altruistic emotions can help to correct egoistic emotions and provide for commitment to collaborate with a vaccination program. Quantitative models and moral emotions should go hand in hand in dealing with risk and uncertainty.
Charlotte Watts
London School for Hygiene and Tropical Medicine
Using mathematical modelling to inform HIV policy in low and middle income countries: fact, fiction and future directions
Mathematical modelling of the spread of HIV infection in different settings is commonly used to help inform HIV policy and programmes globally. Early modelling analyses focuses on the potential demographic and economic impact of HIV, and sought to identify key opportunities for intervention. This included the development of the ‘core group theory’, that highlighted the central role of commercial sex and other high risk behaviours to HIV transmission, influencing not only the levels of infection among these groups, but the degree to which infection may spread more generally. These classic analyses helped motivate large scale investments in prevention programmes targeted at commercial sex, and have curbed the HIV epidemic in many settings. However, there is the risk that this approach to conceptualising and modelling HIV transmission among core groups is over-reductionist, and may miss key populations that are also central to HIV transmission.
More recent modelling has included a focus on the potential importance of different prevention options – commonly being used to assess the implications of clinical trial findings to different HIV epidemic settings, as well as to motivate for the value of increased global HIV investments. While important, there has been limited critical discussion about how best to parameterise these models, and the degree to which these analyses should be used to highlight the potential of new developments in HIV interventions (such as male circumcision or ART based HIV prevention options), versus to illustrate what may be the realistic impact of adding a new prevention option onto existing HIV programmes. Over-optimistic modelling can help mobilise funding, but has the danger of creating false expectations. Unduely pessimistic introduction scenarios may stifle ambition, and condemn advancements that may nevertheless be important.
The relevance and potential value of HIV modelling could be enhanced with the increased investment in ensuring that modelling analyses are conducted in collaboration with programme staff, and draw upon relevant insights from the behavioural sciences, economics and epidemiology. Important potential developments include further exploration of how different forms of social organisation and sexual networking influence expected patterns of HIV transmission, and the development of mathematical modelling techniques, to enable the influence of societal influences on patterns of HIV related risk behaviours and health service use to be explored. Although this would take mathematical modelling into less well charted territories, this has the potential to provide important new insights, and help support a stronger, more nuanced analyses of HIV prevention priorities.