From ghost maps to model maps

Online attempts such as Flu Tracker's to map the swine flu outbreak globally are not only worth examining for their informational content (or lack of it), but also for what they tell us about the fast changing character of disease mapping and epidemiological surveillance in the age of the internet.  Ranging from the official WHO global atlas tool, to unofficial online newspaper  time series maps, to Flu Tracker's Google-Earth enabled hybrid map mash-ups of both, the global scale of the mapping and the consequent depiction of H1N1 as a global pandemic is as striking as the ease with which one can access all this geographic information in the first place.  So what do these global visualizations of the disease tell us about the changing character of disease mapping?

 In The Ghost Map, a widely acclaimed book on the epidemiological revolution represented by John Snow's mapping of cholera deaths in nineteenth century London, author Stephen Johnston  explains that the cartography was crucial. It served to debunk the miasmatic theory of cholera's etiology, and it also represented a revisioning of urban space itself as a community shaped by a geographically defined vulnerability to contagion.  This same geographic revisioning also served in turn as a geographic guide for public health interventions  (such as improvements in sanitation) at a distinctively urban scale.  Johnston concludes his book with some provocative reflections on how the geography of the urban scale may again be critical to public health today but this time in a twenty-first century of mega-slums and mega-cities tied together globally by the speedy cross-border movement of people and pathogens.  In this regard, the emerging maps of swine flu index some of the related changes in how global space is being conceptualized and represented in line with increasing awareness of both global vulnerabilities and the need for global action.  Moreover, by effectively replacing John Snow's hand-drawn map of London with an array of new, often internet-enabled or otherwise computer-algorithm-based mappings, today's maps illustrate the eclipse of the nineteenth century ghost map by a twenty-first century rise of model-mediated disease mapping.

The front-page of the New York Times for May 3rd 2009, for example, featured a story entitled "Predicting Flu with the Aid of (George) Washington".  The article is all about the ways in which computer models using spatial data  related to air traffic, cell-phone calls and even the movement of dollar bills can effectively generate predictive maps of the movement of the flu. Such predictive spatial modeling is similar in some respects to another approach used to model seasonal influenza that actually uses monetized incentives to create so-called 'prediction markets'.  These are designed to work like financial futures markets to predict outbreaks and their intensity based on the assumption that speculative activity in a market is a good way to pool heterogeneous data in a way that rewards good data and good predictions with a financial incentive. Given the widespread anxiety about the market failures of real financial derivatives in recent months, it seems that such a marketized approach to epidemiological planning is unlikely to find many followers.  Moreover, whereas prediction markets are used chiefly to generate temporal predictions (about the timing and intensity of disease outbreaks), it is the geographic patterning of vulnerability that tends to be of wider interest to the public as well as to public health officials interested in the geographic organization of pandemic preparedness.  In this respect, other model methodologies that generate actual maps look better set to set the new norms.

One such model-based mapping innovation that has been developed of late that does provide geographic information about seasonal influenza is Google's Flu Trends tool.  This tool uses internet query entry data matched with traditional CDC surveillance data to generate geographically predictive algorithms. Based on locational clusters of heightened use of key query entry terms (terms that tend statistically to be associated with heightened incidence of people experiencing flu symptoms), the algorithms are able to produce a state-by-state mapping of flu activity (and hence flu risk levels) across the US. The result is a form of what is called "syndromic surveillance" because it is based on data that reflect symptoms (in this case the use of certain tell-tale query terms that reflect google users' symptoms) rather than the actual clinically-proven experience of disease.  Freed thus from the practical complexities of on the ground clinical reporting and virological lab work, its big advantage is that it works much more speedily than traditional clinical surveillance.  However, as its developers have reported in Nature they do not see the tool as a replacement for clinical surveillance.  Instead,  working  10 to 14 days faster than the normal CDC surveillance methods, its speed provides an extra aid for making time-sensitive public health policy decisions, including possibly about where to move and stockpile anti-viral medicines such as Tamiflu.  Quite how well the Google Flu Trends tool might translate to the swine flu case is not yet clear, although the  site does now provide a link to a new "experimental" flu trends map of swine flu incidence in Mexico.  Questions linger, however, about how the digital divide between computer-using and non-using communities may skew such predictive model maps, as well as about how they might remain vulnerable to the sorts of false positives created by online panics and rumor-sharing cascades of bad data.

Yet another important online disease mapping development that has had a swine flu map added is Harvard's Health Maps project.  This is not a form of syndromic surveillance that uses models built around query entry data on the internet, but represents instead an attempt simply to use webcrawling technology to provide up-to-date global maps of news stories on particular diseases based on their place of publication.  The Swine Flu version of this project reveals both the extraordinary global scope of the resulting mapping, but also some of its limits insofar as stories linked to places in one part of the world (e.g. Mali) may actually be reporting on events elsewhere (e.g. Mexico).  These problems aside, what is telling about such mapping is the way in which it aspires to represent the whole of global space as a community that shares - albeit very unevenly - a common vulnerability to contagion.  

In such globalized 21st century revisions of John Snow's ghost map we can surely also see a revolution in public health cartography and epidemiological interpretation.  Gone are the ghosts of Londoners, and in their place are millions of stories, reports and algorithmic relays of reports from around the whole world.  Perhaps we can also see in these transformations one basis for more effective global planning and more globally redistributive public health infrastructure funding.  

Certainly the visualization of the whole globe as a community of contagion serves as a useful spur to reflect on how poor public health infrastructures and inadequate  epidemiological surveillance in poorer parts of the world translate into heightened and shared dangers for the whole planet.  Back in the age of John Snow's ghost map another physician -  William Budd - commenting on another disease - typhoid - made what are now world famous remarks about this kind of shared vulnerability.  Like today's flus, typhoid was much more lethal for the poor, and the point was not that everyone was equally doomed by the disease.  However, for Budd the ties between rich and poor still made typhoid a shared threat to all.

“The disease not seldom attacks the rich, but it thrives among the poor.  But by the reason of our common humanity we are all, whether rich or poor, more nearly related here than we are apt to think. The members of the great human family are, in fact, bound together by a thousand secret ties, of whose existence the world in general little dreams.  And he that was never yet connected with his poorer neighbor, by deeds of charity or love, may one day find, when it is too late, that he is connected with him by a bond which may bring them both, at once, to a common grave.” 

As histories of public health such as George Rosen's remind us, it was this sort of awareness of shared vulnerability that in turn inspired the development of public health planning and community-wide interventions, first at an urban scale in the nineteenth century, and then more and more at a national scale in the twentieth century. However, as Paul Farmer argues in his powerful preface to Dying for Growth, today in the era of neoliberal globalization we need to revise and rescale our geographic vision of community anew.

“The ‘thousand secret ties’ still connect the poor and the wealthy, despite all the barriers our age has set up to separate them.  They are, in fact, less secret (though for many, effectively censored into invisibility) in the age of telecommunications.” 

The global mapping of H1N1 is yet another reminder of such now-not-so-secret-ties; and while pointing to more danger in poor communities (in Mexico), it has also underlined that all the barriers, border fences and gated communities of today's world are easily transcended by jet traveling pathogens.  The global maps we see today of a global pandemic therefore really could inspire efforts to develop more globally-shared public health planning based on a heightened awareness of transnational, transclass and transcultural vulnerability.  They could in this way serve as model maps that also model a new more global approach to investing in global health as a collective planetary good. However, as the post on 'viral globalism' (below) highlighted, such awareness of shared global vulnerabilities does not always lead to shared global consensus on the need for globally organized and globally redistributive responses that address the needs of the poor.   Instead, talk of subsidiarity as the future for global public health planning  seems to be haunted by a different specter altogether, a specter, one might even argue, of  social justice haunting the world!