Further Philosophical Considerations about Covid-19: Why We Need Transparency (guest post by Stefano Canali)
The following is a guest post* by Stefano Canali, a postdoctoral fellow at Leibniz University Hannover who works in philosophy of medicine, with a focus on epidemiology and the epistemic role of data.
It follows up on the piece by Alex Broadbent (Johannesburg) published earlier this week, “Thinking Rationally About Coronavirus COVID-19.”
[David Spriggs, “Vision Series”]
Further Philosophical Considerations about Covid-19:
Why We Need Transparency
by Stefano Canali
I was very glad to see Alex Broadbent’s piece on March 9, which analyses and unpacks research and discourse on the current coronavirus pandemic[1]. Broadbent provides an analysis of various aspects and issues related to the virus and in particular discusses key dimensions and elements at the centre of public health reactions to the outbreak, including the effectiveness of the measures, effects on the economy, changes to quality and quantity of life, etc. This analysis helps us understand why and how certain measures are currently being taken by (some) governments around the world, thus also providing a rational basis for their assessment and critique. This is something we desperately need in an age where pseudoscientific information is constantly shared and used as justification of political positions, if not actions. In more general, meta-philosophical terms, seeing philosophers stepping into the public arena, discussing current scientific issues of key societal important and providing analyses that are helpful to both academic and non-academic audiences is exciting and reminds us of the potential value of philosophical research for science and beyond.
In this piece, I want to pick up from where Broadbent left and expand his analysis, discussing a few other aspects of the coronavirus pandemic and the surrounding discourse, which I think can profit from philosophical considerations. In particular, I will focus on the ways in which we are measuring and counting COVID-19, why this matters and why we need more transparency.
How Are We Counting COVID-19?
In the last few weeks, we have become increasingly familiar with some of the measurements and rates that epidemiologists use in their research and form the basis of many public health interventions. The history of epidemiology is closely related to the history of statistics and the development of statistical measures and tools, the contemporary nation state and the central collection of population data and more generally changes to counting and data practices[2]. The close relation between counting, sorting and memory practices and scientific developments and changes is of course present in many (if not all) scientific disciplines and has been recognised by many historians and philosophers of science[3]. Still, the role of counting is particularly evident as one of the main common elements of epidemiology as a discipline, since—unlike other disciplines—for the most part there are no longstanding theories that form the core of epidemiology, to the point that some have argued that epidemiology is rather based on a few core principles and measurement methods[4]. Measurements, rates and data are thus particularly important in epidemiological research and significant to discuss from a philosophical perspective. Rates such as case-specific mortality and fatality, which are constantly present in studies and policy interventions about the current pandemic, offer a case in point.
Case-specific mortality rate is the measure of deaths from a specific condition in the overall population. For instance, in order to get the mortality rate of COVID-19 in the Italian population, the number of deaths by COVID-19 is divided by the total size of the Italian population. By contrast, case-specific fatality rate is the measure of deaths from a specific condition in the limited population of individuals with the condition. For example, in order to get the fatality rate of COVID-19 for the Italian cases, the number of deaths by COVID-19 is divided by the total number of infected cases. Fatality rates are proportions and will as such result in numbers of a bigger size compared to mortality rates. On the other hand, mortality rates are what epidemiologists call “true rates” and can be more informative on the risks of a condition for the overall population.
What I want to highlight as an important feature of these measurements is that they are dependent on factors that are not directly related to the disease itself and their understanding and use depends on the context of application as well other rates.
Fatality rates of COVID-19 can change relatively rapidly throughout an epidemic and are dependent on the specific healthcare system, for instance on its capacity to contain the virus or deaths[5]. Furthermore, rates are dependent on the overall scientific and political context where the virus is spreading and its particular counting practices. For example, epidemiologists are currently suspicious of many figures for fatality rates of COVID-19. There is strong uncertainty about how many people have actually been infected worldwide, since many infected individuals have not been tested (yet), even after seeing their doctor or because of gaps screen and testing[6]. Changes to the total number of infections will affect the denominator of fatality rates calculations, thus substantially changing the figure.
The situation is similar for mortality rates. For example, in Europe the counting practices involving the number of deaths from the virus vary substantially and the decision of whether to count deaths as COVID-19 is highly contextual. While the Coronavirus Study Group of the World Health Organization classified the coronavirus-related condition as COVID-19 in early February and discussed it as a “severe acute respiratory syndrome”[7], in practice we are seeing substantial differences in the numbers of deaths attributed to the COVID-19. This is partly due to the fact that infected individuals who died tend to have critical pre-existing conditions[8] and thus the cause of their deaths is difficult to find. At the same time though, deaths that in some healthcare systems would be reported as ‘COVID-19 deaths’ may currently be reported as ‘pneumonia deaths’ in others and therefore not counted in mortality and fatality rates.
This contextual status is not a problematic aspect of rates and measures, but can become problematic if we strip context out of the communication and use of data. Differences between measurements like mortality and fatality rates are not immediately clear, at least to non-experts, and in the last few weeks we have seen confusion in the public debate, as many have taken one from the other and argued for or against containment measures on the basis of the wrong rate. For example, when the World Health Organization announced that 3.4% of infected cases had died globally on March 3, commentators and politicians including the US president Donald Trump expressed their doubts about the mortality rate of COVID-19 being more than a fraction of 1%. However, 3.4%—the rate shared by the World Health Organization—was obtained by deaths by the total number of infected individuals, which as we have seen is the fatality rate. More generally, the ways in which rates are calculated, as well as the counting and data collection practices they are based on, matter when it comes to the reliability and quality of data. Yet, in the context of both scientific and public debates, these contextual features often go unmentioned and data tends to be taken and communicated as objective mirrors of the phenomena we are dealing with—which is why I think we need more transparency.
The Need for Transparent Data and Decisions
Data and measurements of COVID-19 are thus influential on how we interpret phenomena, but are also highly contextual and their role is often not considered. Far from giving us necessarily objective representations or mirrors of reality, data depends on assumptions that are based on scientific as much as political and social considerations and, in turn, can influence the scientific research, policy decisions and public perceptions of phenomena.
In line with recent philosophical work on the epistemic role of scientific data, the counting practices in the COVID-19 case show that data and measurements are not a “given” and should not just be taken at face value[9]. The value of data as representations of the COVID-19 pandemic lies in the relations established between measurements and the questions asked, the tools used to interpret them, and the context of application[10]. For example, fatality rates are not necessarily representative or informative in themselves. Whether they are a good measurements of the current situation of the pandemic depends on considerations including how infections test were performed, how deaths were counted on the basis of specific assumptions, and ultimately on what we want to know and for which policy interventions we want to use the rate for.
The COVID-19 case shows how the extraction of value from measurements and data is not a neutral process, as interpretations of data are based on specific interpretations, assumptions and values, and has always potential social, ethical and political implications. This suggests that, when using scientific measurements and data, especially in high-stake and uncertain situations like a pandemic and in the context of extreme policy interventions, we should always closely consider the provenance, manipulations, assumptions and interpretations of data and critically analyse the relations between data, representations, interventions.
This is why I argue that we need something we have only partly seen in the COVID-19 case so far: transparency. At the research level, we need to make sure that the collection, manipulation and use of data is well-documented and open to other researchers and the public. In addition, the set of assumptions and tools used in counting practices should be discussed in the community and standards need to be agreed upon at a relatively general and universal level. Health institutions in Europe are increasingly moving in this direction when reporting data about the pandemic, but we need to more: in a relatively localised context like the European Union, it seems reasonable to have common standards when counting cases, collecting data and communicating rates to the public and policy makers. Finally, the reasons and rationale for the application of policy interventions need to be made transparent: here, the role of the scientific experts in influencing and giving evidence for these decisions is paramount and equally needs to be made explicit.
Extensive research in philosophy of science, social epistemology and science and technology studies[11] has shown how both epistemic and non-epistemic values enter decision making in the sciences and how these need to be balanced, especially in contexts where uncertainties are abundant, like the current COVID-19 pandemic. As argued by Broadbent, many factors and considerations need to be weighted in cost-benefit analyses for policy interventions. I want to add that, in turn, these factors are based on assumptions and values both at the point where we collect data as well as when we apply policies. In democratic societies, these processes need to be made as transparent as possible—data should empower us to go behind the scenes, rather than creating further friction[12].
NOTES
[1] Broadbent A. Thinking Rationally About Coronavirus COVID-19 (guest post by Alex Broadbent). http://dailynous.com/2020/03/09/thinking-rationally-coronavirus-covid-19-guest-post-alex-broadbent/. Published 2020. Accessed March 12, 2020.
[2] Broadbent A. Philosophy of Epidemiology. London: Palgrave Macmillan UK; 2013.
[3] Bowker, G. C. Memory Practices in the Sciences. Cambridge, MA: The MIT Press, 2006.
[4] Morabia A. History of Epidemiologic Methods and Concepts. Basel: Birkhauser Verlag; 2004
[5] Wei-jie G., Zheng-yi N. Yu Hu, et al. Clinical Characteristics of Coronavirus Disease 2019 in China. The New England Journal of Medicine. http://doi.org/10.1056/NEJMoa2002032
[6] del Rio C, Malani PN. COVID-19—New Insights on a Rapidly Changing Epidemic. JAMA. Published online February 28, 2020. doi:10.1001/jama.2020.3072
[7] Gorbalenya, A.E., Baker, S.C., Baric, R.S. et al. The species Severe acute respiratory syndrome-related coronavirus: classifying 2019-nCoV and naming it SARS-CoV-2. Nat Microbiol (2020). https://doi.org/10.1038/s41564-020-0695-z
[8] Wu Z, McGoogan JM. Characteristics of and Important Lessons From the Coronavirus Disease 2019 (COVID-19) Outbreak in China: Summary of a Report of 72 314 Cases From the Chinese Center for Disease Control and Prevention. JAMA. Published online February 24, 2020. doi:10.1001/jama.2020.2648
[9] Leonelli, S. Data-Centric Biology: A Philosophical Study. Chicago: The University of Chicago Press, 2016.
[11] Longino, H.E. Science as Social Knowledge: Values and Objectivity in Scientific Inquiry, Princeton, NJ: Princeton University Press. 1990.
[12] Leonelli, S., Rappert, B., & Davies, G. Data Shadows: Knowledge, Openness, and Absence. Science, Technology, & Human Values, 42(2), 191–202. 2017. https://doi.org/10.1177/0162243916687039
To be perfectly honest, I think that making data sets and models available publicly to create transparency is a terrible idea. This piece contains many assertions to the contrary but not much of an argument.
Doing this would create an internet nonsense-fest of cranks producing confusing messages at a time when clarity and consistency are critical. I don’t know how it is in the US, but where I am in the UK it already feels like half the people in the country have become (very) amateur epidemiologists and are merrily spouting their opinions and conspiracy theories. These ideas would make this worse.
Transparency is not always a good idea.
In the absence of accurate numbers, won’t people just do the same thing with inaccurate numbers that someone has just made up?
Some will, but fewer will pay attention. Publicising data would be an invitation to more people to do this.
Note – I think that it is a very good idea for other qualified people to look at data sets etc. and challenge policy responses, but this should be an orderly ‘red team’ process conducted confidentially, not an online free for all.
How could transparency not be a good idea when confusing information is already spread even with the most ‘transparent’ possible scenarios?!?! Now the ‘confusion’ is “good”!?!?!? So do our comments are embedded with values inevitably, and from these we may infer the type of interests and preoccupations inhabiting the commentator minds… The “piece” seems to be especially concerned about the imprecision rates (not that unusual for statistical-type measurements by the way), and DESCRIBES how there can be implications on all social levels, since we’re dealing with interpretations of ‘realities’, thus making very perceivable the ARGUMENT that data should always be manipulated specifically depending on the interest at play… Indeed, the discourse may have very unperceivable levels!
This is a lovely piece, thank you, but I share the scepticism that the answer is transparency (but for slightly different reasons than those already articulated). The piece has several references to “we” and “the community” needing to have access to data. I guess that the “we” here is ambiguous – it seems plausible that scientists, policy-makers, and others directly involved in decision-making are helped by having as much data as possible. I can see the argument that philosophers are part of the “we” involved in those decisions, so “we” should have access.
My worry is that if the “we” is intended to refer to something like “all interested citizens”, then I’m not certain what the basis for the argument is: to a large extent, “our” personal decisions about what to do are guided by policy-makers’ responses – if the schools are closed, I can’t send my kids to school, regardless of how I interpret the data. To the extent that I so have scope for personal choices, I’m not obviously helped by being presented with a mass of uninterpreted data – I lack the skills to decide what to make of it. A similar problem arises if we think of the data as relevant to my “political” decisions – what to agitate for or demand from the State, say: even if value judgments play a role in measurement, it’s not as if raw data translates itself into compound stats just by the addition of some values. Lacking statistical skills I might lack the ability to use the data to promote my sense of proper response.
Of course, these aren’t arguments against transparency per se, but against the idea that transparency is of any benefit. The worry, however, is that calls for transparency can easily seem to imply that epidemiologists and others should give up on presenting “interpreted” results entirely. That strikes me as a complete dereliction of duty – akin to a museum curator who just dumps the exhibits in the hall, with no attempt to contextualise or explain. As the post nicely explains, there are huge and important epistemic and ethical challenges in constructing measures, but what that implies is that we need to construct them well, and thoughtfully, rather than that we should not construct them at all. I don’t think the author would disagree, but I do worry that calls for transparency about decisions can easily divert us from this task, because they can slide into a sense that we shouldn’t construct at all, or, even worse, that we can construct any old way as long as we were “transparent” about how we did it.