The ongoing Ebola outbreak is taking place in one of the most highly connected and densely populated regions of Africa (Figure 1A). Accurate information on population movements is valuable for monitoring the progression of the outbreak and predicting its future spread, facilitating the prioritization of interventions and designing surveillance and containment strategies. Vital questions include how the affected regions are connected by population flows, which areas are major mobility hubs, what types of movement typologies exist in the region, and how all of these factors are changing as people react to the outbreak and movement restrictions are put in place. Just a decade ago, obtaining detailed and comprehensive data to answer such questions over this huge region would have been impossible. Today, such valuable data exist and are collected in real-time, but largely remain unused for public health purposes – stored on the servers of mobile phone operators. In this commentary, we outline the utility of CDRs for understanding human mobility in the context of the Ebola, and highlight the need to develop protocols for rapid sharing of operator data in response to public health emergencies.

Fig. 1: Mobility patterns and connectivity in West Africa.

A) Map showing the location of Ebola outbreaks in humans since 1976 (black dots) overlaid on a map of strength of connectivity measured by travel time to the nearest settlement of population 500,000 or more, with dense areas of low travel time indicative of high connectivity. No previously recorded Ebola outbreak has ever occurred in such a densely populated and large area of high connectivity as the ongoing outbreak that began in Guinea; B) Visualization of the flows of 500,000 mobile phone users between the (population-weighted) centres of sous-préfectures in Cote d’Ivoire. The inset highlights the mobility in the western border region (main figure: flows above 20 km with more than 10 average movements per day included, inset figure: flows above 20 km with at least one movement on average per day included); C) Outputs of a within-country mobility model for West Africa built on mobile phone CDRs. The lines show the flows predicted to be greater than 75-95% of the estimated flows per country between settlements for the average number of trips per week and are overlaid on a map of population density (

The rise of mobile phone usage across the past decade, even in the most remote low-income settings, has been astonishing. The global mobile phone penetration rate (i.e. the ratio of active subscriptions to the population) reached 96% in 2014.1 In developed countries, the number of subscribers has surpassed the total population, with penetration rates now reaching 121%, while in developing countries it is as high as 90%, and continuing to rise.1 Mobile phone networks, also called cellular networks, are composed of cells, i.e. geographic zones around a phone tower. Each communication can be located by identifying the geographic coordinates of its transmitting tower and the associated cell. Mobile call data records (CDRs) detailing the time and associated cell tower of calls and text messages from anonymous users therefore provide a valuable indicator of human presence, and sequences of these data can be used to measure population movements over time, especially when existing mobility data is poor (see Figure S1).2,3

With network operators serving substantial proportions of the population across entire nations, the movements of millions of people at fine spatial and temporal scales can be measured in near real-time and across seasons. Although such data inevitably contain biases due to phone ownership and usage patterns, evidence suggests that these have limited impacts on general estimates of population movement patterns and the relative importance of different travel routes.4 The engagement of network operators has resulted in population movement analyses based on CDRs that have been particularly promising for improving responses to disasters 5,6,9 and for planning malaria elimination strategies.3,7,8,19

The benefits of CDRs in the context of the current Ebola outbreak are clear. The rapid spread of the virus within Guinea, Sierra Leone and Liberia, and to Nigeria and Senegal, has been driven by local and regional travel.18 Epidemiological models of the spatial spread of Ebola, both retrospectively and for the purposes of prediction, rely on estimates of the volumes and flows of traffic between populations. This allows modelers to assess the likely routes of infected individuals between populations, with imported cases sparking new outbreaks or augmenting local transmission. Since mobility is not only a major driver of the epidemic, but is also likely to shift dramatically in response to the outbreak and be directly targeted by control policies, these estimates are critical. Although the epidemiological data are still highly uncertain and CDRs cannot currently capture cross-border movements, understanding the potential routes of spread of the virus within a country are critical to national containment policies, and will strongly influence more regional spread across borders. Further, the benefits of information on population distributions and mobility for assessment of the implementation of movement restrictions and efficient delivery of interventions, including possible drugs and vaccines, are clear. Detailed aggregated mobility patterns of half a million anonymous phone users on the Orange network 10 in Cote d’Ivoire are shown in Figure 1B, information that would be unobtainable through any other means.

In the absence of operator data from the currently affected countries, we have produced spatial interaction models of national mobility patterns parameterized using CDRs from Cote d’Ivoire, Senegal (made available by Orange in response to the Ebola epidemic10), and Kenya (Appendix 1). Figure 1C shows a visualization of the outputs from this initial set of freely available mobility models (for download links and regular updates follow: These models estimate the amount of travel between predefined locations using data on the size of population of the locations and the distances between them (Appendix 1), but do not take into account behavioural changes as the outbreak has progressed and the impact of travel restrictions on population mobility. The models are likely to capture the most important routes of travel and the relative volume of traffic between different populations in the region, but substantial uncertainty remains due to lack of contemporary operator data from different countries in the region and the potential for regional variations in mobile phone usage and ownership patterns.4 While the spatial scales of mobility estimates are defined by tower coverage, ranging from hundreds of meters in urban centers to several kilometers in rural areas, understanding travel between population centers is likely to be most critical for planning national and regional containment strategies. Further, the integration of census-derived migration data enables estimation of a wide range of regional cross-border movement patterns, bounded below by international migration data and above assuming borders do not hinder movement (Appendix 1).

Of particular concern, this regional overview of national mobility patterns shows that large areas of West Africa are likely to exhibit much higher population flows than the currently affected areas. Both the mobile operator data from Cote d’Ivoire and the modelled mobility patterns across the region highlight the dominant influence of large population centers, which serve as hubs of national mobility. Several countries in the region are now suspending flights from affected countries, reducing the flow of travel between national hubs. However, rural areas near porous borders remain vulnerable to Ebola importation, and could undermine containment strategies since many of these border areas are likely to be well connected to population centers within their borders. The border between Liberia and Cote d’Ivoire highlights this vulnerability (Fig. 1B).

Despite the value of CDRs in the face of the Ebola emergency, mobile network data is generally very difficult to access due to commercial and privacy concerns. The data contain detailed information on mobile operators’ system designs, their customers, as well as detailed information about individuals’ locations and mobility.11 However, for the purposes of responding to epidemics or other public health emergencies, operators may not need to provide access to their complete databases. Since travel between, rather than within, populations is likely to be the most critical information – particularly in the absence of highly spatially resolved epidemiological information – aggregated data on mobility between populations will often be sufficient. Such connectivity matrices are relatively easy for operators to produce themselves, and are much less sensitive than the raw data with regards to regulator requirements for personal privacy as well as commercial competition between operators. Such matrices could potentially be generated routinely for preparedness planning or in near real-time in response to an epidemic.

Careful interpretation of local contexts and data biases are required to generate robust mobility models from mobile phone data, and on-going efforts to validate and improve estimates are crucial.3,12,13 However, the value of these data in the context of a public health emergency like the ongoing Ebola outbreak is undeniable, particularly when integrated with other datasets, as has been done previously for other infectious diseases.3,7,8,14 Mobile operators such as Orange, Safaricom, Digicel, MTC, and Telenor/Grameenphone, who have previously released anonymous network data for public health purposes, deserve credit for actively engaging with researchers to build partnerships to leverage CDRs for public health and development. The continuing spread of Ebola highlights the reality of emerging infections in our increasingly connected world, and we hope that these partnerships can serve as models for operators, researchers, governments, and agencies globally, and in West Africa in particular.

Competing Interest

The authors have declared that no competing interests exist.