Ebola Mapping

Perhaps the earliest map of Ebola was included in a March 24 report issued by the Republic of Guinea’s Health Ministry.¹¹ The map identified 4 adjacent prefectures, distinguished by red hatching, in which Ebola had been confirmed: Guéckédou, where the majority of deaths had occurred; Macenta; Kissidougou; and N’Zérékoré. In the map the names of neighboring countries (Liberia, Mali, Senegal, and Sierra Leone) were lettered but otherwise not distinguished. The map thus argued a limited outbreak in 4 provinces of 1 country. The potential for disease diffusion to neighboring countries was implicitly denied in this manner. The map conformed to the assumption that, like previous outbreaks beginning in the 1970s, this one would be spatially contained in nonmetropolitan, interior areas.

In its March 27, 2014, report, the World Health Organization (WHO) similarly mapped all national prefectural boundaries, coloring red those reporting active Ebola cases (Figure 1).¹² That map included the coastal prefecture of Conakry and its port city of the same name where 4 cases were confirmed and a fifth was suspected. That brought the total number of confirmed Ebola cases to 103. The possibility of other new infections was being investigated, at that time, in only 2 adjacent provinces, which were colored pink.

Figure 1 March 27, 2014, World Health Organization Map of Ebola in Guinea. The map presented a localized outbreak in 4 provinces in the Republic of Guinea and, separately, the port area of Conakry. Source: World Health Organization Regional Office for Africa website. http://www.afro.who.int/en/clusters-a-programmes/dpc/epidemic-a-pandemic-alert-and-response/outbreak-news/4069-ebola-virus-disease-guinea-27-march-2014.html.¹²

For a medical geographer, the WHO map, but not the Ministry of Health’s map, would have raised a warning flag. A slow diffusion from Guéckédou to neighboring districts would be expected and could potentially be managed with local containment and treatment programs. But a hotspot on the nation’s capital, Conakry, promised the likelihood of an expansive epidemic. Localized in central provinces with sparse populations, the Ebola outbreak would, presumably, be containable and short lived. But an outbreak in Conakry, the nation’s capital with an estimated population of nearly 2 million persons,¹³ was something else. Here was the possibility of a large outbreak in an urban population in a city with maritime and other connections to cities in the region and the world.

The questions the first Conakry cases raised were thus exigent. Were they an independent, unrelated outbreak or was the source the Guéckédou outbreak? If the latter, as seemed likely, were the Conakry cases the result of person-to-person transmission or something else (perhaps a food source)? If the outbreak in Conakry was expansive rather than independent, if Ebola was traveling, where else might it soon appear?

If, in retrospect, the epidemic that resulted was an “avoidable crisis,” as some have argued,Reference Jeremy, Farrar and Piot¹⁴ here was where it might have been stopped. Early symptoms of epidemic expansion posted in national prefectural maps were ignored, or at the least, not sufficiently credited. Experts assumed reasonably, but in the end inaccurately, what the March WHO map denied: Ebola was always a highly localized, short-term, typically rural event. A complex of factors ranging from regional poverty, a minimal health infrastructure, poor communication between attending agencies, and an environment conducive to viral expansion all contributed to the outbreak. But the failure to see in this and subsequent early disease maps the epidemiological signs of expansion delayed a series of potentially effective interventions.

Epidemic Expansion

As the epidemic matured, the scale of the maps expanded beyond the Republic of Guinea to permit inclusion of cases confirmed in adjacent countries. By the summer of 2014, the resolution of these maps increased to permit the mapping of relative incidence in various cities in the region (Figure 2). In these “dot maps” the size of the circle was based on the number of reported cases. In some maps, pie charts were used to distinguish between “confirmed,” “probable,” and “suspected” cases in a city or town. In official reports by the WHO and others, graphs of the increasing epidemic curve—in cities and across the region—were included based on the same data as the maps. Eventually, maps of the regional epidemic included a third event class in which hospitals and medical facilities were included. These did not, however, include institutional capacity (number of beds, doctors, and nurses, etc) and patient loads. Had those data been included, the inadequacy of existing medical resources—local and international—would have been immediately evident.

Figure 2 Map Included in the World Health Organization Situation Report in August 2014. By August, maps were describing a regional outbreak with varying mortality in various cities across West Africa. Source: World Health Organization. http://apps.who.int/iris/bitstream/10665/131974/1/roadmapsitrep1_eng.pdf?ua=1.¹⁵

These dot maps, which were widely reproduced in international newspapers, served as evidence of a serious, expansive epidemic. Expansion into Conakry and Freetown (Sierra Leone) and Monrovia (Liberia)—larger port cities—signaled potential pandemic expansion. The maps argued, simply, “this (epidemic) is here and it might soon be everywhere.” Increasingly, maps of potential pandemic occurrence, and then international cases involving former health workers, were published.

Local Mapping

When the first cases of Ebola hemorrhagic fever were reported in March 2014, Swiss-based Médicine sans Frontièrs (MSF) quickly dispatched an epidemiological team to Guéckédou, Guinea, the epicenter of what was to become a regional epidemic. The MSF team included a geographic information systems (GIS) specialist who, over an 8-week period, created 109 maps of local roads, landmarks, and villages, in Guéckédou prefecture.Reference Lüge¹⁶ In a region without existing maps of affected towns and villages, this was critical. Without these maps, tracking the incidence of Ebola across the villages and towns of a region was difficult. Approximately 60% of the GIS specialist’s time therefore was spent on producing way-finding maps of the roads and villages in the region; the remainder was used to create maps that addressed public health concerns over the location of water, hygiene, and sanitation teams. Only a few maps were made to serve epidemiological investigations.

In recent decades, mapping has been transformed from a manual to a digital skill with the use of computer-based GIS programs. Before the MSF team’s departure, a general map of the Guéckédou region was traced based on satellite images of the region. In this process, a satellite image of the target area is displayed in a program permitting streets, buildings (homes, hospitals, schools, etc), and biogeographic elements (lakes, streams, etc) to be constructed as separate digital event classes. The result is at once spatially precise and easily modified when new data become available (for example, patient location). These satellite-based computer maps are typically composed of lines (for streets), points (precise locations), and polygons (biogeographic and political boundaries) and can be easily printed or digitally distributed.

The resulting way-finding maps were critical for foreign field workers unfamiliar with the region. Without them, getting from medical sites to local villages would have been impossible. Also, the maps permitted a picture of the spatial spread of the epidemic to be ascertained. Within the target prefecture, for example, 14 separate villages were named “Bendou.” Identifying which Bendou was the home of a specific patient was potentially critical to those tracing the chain of disease transmission and to others attempting to understand the general structure of the district epidemic.Reference Lüge¹⁶ “The challenge is good information, because information helps tell us where the disease is, how it’s spreading and where we need to target our resources,” a United Nations emergency response official told reporters in October 2014.Reference Muhumuza¹⁷ “Unfortunately, we don’t have good data from a lot of areas. We don’t know exactly what is happening.”Reference Muhumuza¹⁷ Mapping was a critical medium in which data on “what is happening” could be collected, organized, analyzed, and presented. Because local and foreign workers often spoke different languages, maps served as a lingua franca promoting shared knowledge among those who otherwise had difficulty communicating.

HOT Mapping

As the epidemic expanded, detailed and accurate maps of newly infected towns and villages across the infected territory were similarly unavailable. Innovatively, MSF and Red Cross International decided to crowdsource the task of mapping these newly infected, previously unmapped towns and villages.¹⁸ The idea had been tried before but never at this scale. Hundreds of international volunteers were invited to join the “missing maps” project organized by Humanitarian OpenStreetmap (HOT). Participants registered, logged in online, and were directed to satellite maps of towns or villages where maps were urgently required by ground personnel. In some countries, “HOT parties” were organized for volunteers.¹⁹

“Firstly,” the website of the program instructed, “we need to trace the features of the target location. Using aerial photographs as a backdrop, we can literally trace road networks, buildings and landmarks to build a line drawing of the area.” Once a contributor’s work was saved to the website, it was integrated automatically into the cooperatively constructed evolving map. Local personnel then added the names of buildings, neighborhoods, lakes, roads, etc, to the digital construct. Those were then uploaded to volunteers who entered them into the map’s final version. As a result, what otherwise would have taken weeks of individual labor required only a few days per map.

Epidemic Mapping

Inclusion of a GIS expert resulted from an earlier MSF study describing epidemiology as “the domain where GIS can bring the most positive evolution.”Reference Lüge¹⁶ Yet, the potential contribution of mapping to the medical effort was hampered by ignorance about the uses of maps in an epidemic environment. “Most MSF staff know very little about GIS in general, let alone how a GIS officer could support them.”Reference Lüge¹⁶ As a result, the GIS team member spent relatively little time contributing to epidemiological studies.

That MSF team members knew little about the use of maps in epidemic studies and public health initiatives was unfortunate. Beginning with the mapping of yellow fever in late 18th century cities, maps have served as a vehicle through which an infectious chain of transmission can be traced in an effort to uncover the source of an outbreak.Reference Koch³Figure 3 is a typical illustration of this kind of mapping by W. H. Frost, a US epidemiologist investigating a poliomyelitis outbreak in Mason City, Iowa, in 1912.Reference Frost²⁰ The mapped approach was based on techniques developed for single-source, bacterial disease outbreaks such as cholera and typhoid fever in the 19th century.

Figure 3 A 1912 Map Attempting to Identify the Chain of Diffusion in a Polio Outbreak in Mason City, Iowa.Source: Frost WH. Hygienic Laboratory Bulletin. 2012;90:9-105; 234-52.Reference Frost²⁰

In the current Ebola outbreak, transmission mapping at the beginning of the outbreak might have alerted field medical personnel that the outbreak that they thought was contained in May was in fact expanding. Mapping of affected persons and their travels during the period of disease incubation almost certainly would have described what was only later recognized. Citizens of Meliandou, where the epidemic’s 2013 index case was eventually located, and Guéckédou in Guinea lived in a porous border region across which Kissi-speaking persons regularly traveled into Liberia and Sierra Leone. The effort to contain the disease in Guinea was therefore doomed without a cooperative effort by all 3 countries to contain an outbreak in a shared region. “The most tragically missed opportunities stemmed from the poor flow of information about who was infected and whom they might have exposed.”Reference Sacks, Fink and Belluck²¹ That failure of information was more geographic than medical, and thus potentially reportable via mapping of local travel patterns.

Investigatory Mapping

Medical mapping has served as a means to both develop and test theories about the environmental origins of infectious disease outbreaks. In this case, increasing attention was paid to the possibility that the outbreak was promoted not only by person-to-person transmission but also through the food chain in which wild animals (“bush meat”) served as a local protein source.Reference Funk and Piot²² Of special interest was the prevalence of fruit bats, a primary protein source in local diets.Reference Pigot, Golding and Mylne²³ Bats have been previously implicated in the transmission of several diseases, including Ebola, MERS, SARS, and Marburg hemorrhagic fever.Reference Aberlin²⁴ Thus, by the autumn of 2014 some disease ecologists were mapping the natural territory of bats both as a way of considering their engagement in this epidemic and in a first assessment of the potential of the Ebola virus to spread to other areas in Africa with robust, shared bat populations.

Figure 4 by Pigot et alReference Pigot, Golding and Mylne²³ is one of a series of maps of bat species suspected as potential West African Ebola reservoirs. The dots on the map locate observation points used to identify the species’ spatial range. Coloration, yellow to green, describes the relative probability of the presence of the target species, in this case the little colored fruit bat (Myonycteris torquata).

Figure 4 Territorial Range of Fruit Bats Implicated in the Ebola Epidemic of 2014. The extent of the map is roughly congruent with the area most affected by the virus. Source: Pigot DM, Golding N, Mylne A, et al. ELife. 2014;10.7554/eLife.04395. http://elifesciences.org/content/early/2014/09/05/eLife.04395.Reference Pigot, Golding and Mylne²³