Data analytics: Lessons learned from Ebola

Big data can help with rescue efforts after natural disasters. The challenge is to use data smartly to gain insight. This then improves the ability of institutions to respond effectively to future health emergencies.

This entry discusses three challenges that will help manage unprecedented health emergencies better:

1. Detecting an outbreak of a deadly disease sometimes beats predicting when and how it might happen (e.g., ebola outbreak versus this winter’s flu spread);
2. Getting regulators to collaborate across borders quickly – instead of possibly chiding each other – requires them to take action; and
3. Guiding interventions effectively necessitates that resources arrive promptly at the right place – logistics.

Ebola already drains weak health systems in West African countries, such as:

  • Liberia, population 4.2 million: 51 doctors; 978 nurses and midwives; 269 pharmacists; AND
  • Sierra Leone, population 6 million: 136 doctors; 1,017 nurses and midwives; 114 pharmacists.

By the way, some claim that the UK’s NHS (National Health System) employs 10 percent or more of Sierra Leone’s trained doctors.

Subscribe to this blog’s content via email for your mobile: DrKPI – the blog for insiders

So how does this relate to KPIs (key performance indicators), big data, measurement and benchmarking? Glad you asked – I explain below.
Ebola Outbreak in West Africa illuminated the significant threat posed by infectious diseases to human lives and society.
West Africa’s ebola outbreak illuminated the significant threat posed by infectious diseases to human lives and society.

Is the answer predicting an outbreak?

Google Flu Trends (GFT), which tries to predict likely flu outbreaks based on how often people use key search terms, has been shown to be inaccurate. Other methods that make use of a much wider range of data sets are enjoying more success.

For example, business consultancy Accenture, big data specialist SAS and the University of North Carolina say they predicted the US 2012-13 flu season three months before the US Centers for Disease Control (CDC) issued its official warning. That is impressive.

Analyzing social media including blog posts and tweets may give an indication of where people worry about a disease. Unfortunately, this method is far from accurate. More often than not, the number of tweets in a region are the result of news coverage (i.e. similar to Google Flu Trends using search data).

We still do not know if such warning signs (i.e. more tweets or Facebook comments about the flu) are accurate reflections of what is actually happening. We do know, however, that posts to services like Flickr is not that helpful for learning more about the people affected by the disaster (e.g., pictures of Hurricane Sandy posted on Flickr or tweets during the LAX airport incident).

As Facebook (see Facebook mood study: Why we should be worried!) has taught us, online users can be significantly influenced. For instance, more positive items in someone’s newsstream results in more positive posts by that user, though the effect is small.

Given the above, it seems more useful to focus on how we can detect an outbreak faster. In turn, we can put the necessary resources in place faster and more intelligently, thereby saving more lives.

Step 1: Detect the outbreak

Besides prediction, one can also use big data to try and detect an outbreak. Healthmap is a website founded in 2006. It crawls news articles, social media, and other online sources for indications of public health threats. For instance, a timeline published by Healthmap in early March, found evidence of an unusual febrile illness in Guinea, before the World Health Organization announced the outbreak.

Without a system like Healthmap, epidemiologists must rely on hospitals, clinicians, or schools to identify and report outbreaks. This means we need more systems and databases that are built and tested to provide us with such data. Researchers can then access this and provide valuable insights about things like natural disasters and health epidemics.

Rivers, Caitlin M. (October 24, 2014). We could’ve stopped ebola if we’d listened to the data. Retrieved October 26, 2014 from

Rivers, Caitlin M., Lofgren, E.T., Marathe, M., Eubank, S., Lewis, B.L. Modeling the Impact of Interventions on an Epidemic of Ebola in Sierra Leone and Liberia. PLOS Currents Outbreaks. 2014 Oct 16. Edition 1. doi: 10.1371/currents.outbreaks.fd38dd85078565450b0be3fcd78f5ccf Retrieved October 26, 2014 from

Step 2: Regulators must move their buds

Putting various systems such as Healthmap and others in place will give us more time to identify public health threats. In turn, possible epidemics can be fought faster than was the case with ebola. It will also allow rescue services to respond quicker. The unprecedented ebola crisis has exposed failings in the ability of international and local institutions to respond swiftly.

After the 2010 Haiti earthquake, a research team analysed calling data from two million mobile phones on the Digicel Haiti network (see below).

Figure 2. Estimated distribution of persons who were in PaP on the day of the earthquake but outside PaP 19 days after the earthquake. Circles are shown for communes that received at least 500 persons.
Figure 2. Estimated distribution of persons who were in PaP on the day of the earthquake but outside PaP 19 days after the earthquake.
Circles are shown for communes that received at least 500 persons.

This work enabled the United Nations and other humanitarian agencies to understand population movements during relief operations. It also helped improve our understanding of how people moved during the subsequent cholera outbreak. In turn, these agencies were able to allocate resources more efficiently, and they were now empowered to identify areas at increased risk of new outbreaks.

The crux of the matter is, one must get access to these data rather quickly, or preferably in real-time, and regulators must move quickly. One challenge is that researchers are often elsewhere, so regulators need to find a way to give them access. Of course, neither violating local regulation nor mobile phone users’ rights, such as privacy, is an option. This is further discussed below.

The health emergency in Western Africa has revealed weaknesses, particularly how such things are addressed by the UN, non-governmental organisations (NGOs), and in particular, local regulators.

Most helpful is if these agencies and regulators, including the International Telecommunications Union (ITU) develop a template quickly. The template or checklist should address the steps that must be taken next time, so that necessary approvals are more promptly forthcoming, data is made available for analysis sooner.

A template can make it easier for people to follow an accepted path. It also helps those averse to risk make decisions that help saving lives. This helps move things along, contrary to what we have experienced with the ebola disaster.

Bengtsson, Linus; Lu, Xin; Thorson, Anna; Garfield, Richard; von Schreeb, Johan (August 30, 2011). Improved Response to Disasters and Outbreaks by Tracking Population Movements with Mobile Phone Network Data: A Post-Earthquake Geospatial Study in Haiti. PLoS Med 8(8): e1001083. doi: 10.1371/journal.pmed.1001083 Retrieved October 25, 2014 from

No author (October 25, 2014). Ebola and big data. Waiting on hold. The Economist, p. 73-74. Retrieved October 27, 2014 from

Step 3: Guide interventions fast

As previously suggested, regulators need to work with a template for giving researchers access to data, such as that collected by mobile networks, which helps access important data quickly. In turn, insights gained from these data can be passed on faster to guide interventions accordingly. However, we also need better ways to coordinate the fight internationally, meaning that researchers and rescue staff need to use big data records to coordinate efforts.

For instance, the level of activity at each mobile phone mast also gives a kind of heatmap of where people are. It also reveals where and how far they are moving from the epicenter of a quake, for instance.

Telecom operators use call-data-records (CDRS) to manage their networks and bill their clients. The records include:

– caller identity,
– call timestamp,
– phone tower location, and
– number dialed.

Of course, as long as the mobile phone is turned on, phone operators can identify where the phone is, even though the user may not be on the line. The reason is that mobile phones, if turned on, constantly send out signals, which are picked up by the closest tower. This information is needed to allow the person to receive a phone call (e.g., think roaming abroad).

Figure 5. (Top) Number of individuals that visited the university campus during the second alert period (in blue) and its baseline (in red) aggregated daily. (Bottom) The same data aggregated hourly.
Figure 5. (Top) Number of individuals that visited the university campus during the second alert period (in blue) and its baseline (in red) aggregated daily.
(Bottom) The same data aggregated hourly.

Frías-Martínez, Vanessa, Rubio, Alberto, Frias-Martinez, Enrique (not dated). Measuring the impact of epidemic alerts on human mobility. Retrieved October 26, 2014 from

Tracking human mobility in disaster areas is vital. It tells us where resources are most needed to help victims. It also reveals how a disease may spread.

For instance, in the study by Buckee et al (see below for reference), accumulating evidence reveals a strong link between human mobility and the spread of epidemics.

Big Data: Applying human mobility data to understand malaria transmission.
Big Data: Applying human mobility data to understand malaria transmission

In the case above, the intention was to alert medics to go where infected people might carry the disease. As Buckee et al have shown, it worked very well!

In addition to trying to point health teams to where they are most needed, the cellphone trackers sent health advice to Haitians via text or voicemail. Examples were things like frequent hand-washing or oral rehydration for those who got sick. Mothers were advised about continuing to breastfeed infected babies.

While these data are updated every single second CDRS are structured. For instance, people who call an emergency number can now be tracked. Such data is helpful and allows researchers to gain insights about how a disease or epidemic spreads.

We must improve detection by using big data smartly. With the help of smoothed procedures (see templates / checklists), regulatory hurdles against sharing data can be removed. Using these data can then guide inverventions on the ground, saving many more lives.

Buckee, Caroline O.;  Wesolowski, Amy; Eagle, Nathan; Hansen, Elsa; Snow, Robert, W. (Jan-Feb, 2013). Mobile phones and malaria: modeling human and parasite travel you can find. Travel Med Infect Dis. 2013 Jan-Feb; 11(1): 15–22. Retrieved October 25, 2014 from doi: 10.1016/j.tmaid.2012.12.003

What is your opinion?

– Have you recently found / experienced a case in which big data helped disaster relief efforts?
– What other recommendations would you make?

I love to read your comments below and look forward to answering them. Merci.

Source: Data analytics: Lessons learned from Ebola

Get the latest news on your mobile by subscribing to our award-winning content: DrKPI – the blog for insiders

More resources

Aranka Anema, Sheryl Kluberg, Kumanan Wilson, Robert Hogg, Kamran Khan, Simon Hay, Andrew J Tatem and John Brownstein (November, 2014). Digital Surveillance for Enhanced Detection of Outbreaks. The Lancet Infectious Diseases, Volume 14, Issue 11, Pages 1035 – 1037, November 2014 doi: 10.1016/S1473-3099(14)70953-3 Retrieved October 24, 2014 from

Isaac Bogoch, Maria Creatore, Martin Cetron, John Brownstein, Nicki Pesik, Jennifer Miniota, Theresa Tam, Wei Hu, Adriano Nicolucci, Saad Ahmed, James W Yoon, Isha Berry, Simon Hay, Aranka Anema, Andrew J Tatem, Derek MacFadden, Matthew German and Kamran Khan (October, 2014). Assessment of the Potential for International Dissemination of Ebola Virus through Commercial Air Travel During the 2014 West African Outbreak (The Lancet, Oct, 2014) The Lancet, Early Online Publication, 21 October 2014 doi: 10.1016/S0140-6736(14)61828-6 Retrieved October 26, 2014 from

Informatics Resources for Ebola Epidemic Response (Resource page).

David Pigott, Nick Golding, Adrian Mylne, Zhi Huang, Andrew Henry, Daniel Weiss, Oliver Brady, Moritz Kraemer, David Smith, Catherine Moyes, Samir Bhatt, Peter Gething, Peter Horby, Isaac Bogoch, John Brownstein, Sumiko Mekaru, Andrew Tatem, Kamran Khan and Simon Hay. (September 2014). Mapping the Zoonotic Niche of Ebola Virus Disease in Africa (eLife). Retrieved October 22, 2014 from

Talbot, David (August 22, 2014). Cell-phone data might help predict ebola’s spread. Mobility data from an African mobile-phone carrier could help researchers recommend where to focus health-care efforts. MIT Technology Review. Retrieved October 26, 2014 from

The Guardian – Ebola funding tracker – interactive

Amy Wesolowski, Caroline Buckee, Linus Bengtsson, Xin Lu, Andy Tatem (September 2014). Containing the Ebola Outbreak – The Potential and Challenge of Mobile Data (PloS Current Outbreaks, Sep, 2014). Retrieved October 29, 2014 from

YouTube Preview Image

Urs E. Gattiker

Professor Urs E. Gattiker - DrKPI is corporate Europe's leading social media metrics expert (see his books). He continues to work with start-ups. Urs is CEO of CyTRAP Labs GmbH.

3 thoughts on “Data analytics: Lessons learned from Ebola

Leave a Reply

Your email address will not be published. Required fields are marked *