Statistical Data Analysis

Sierra Leone: IBM's New Ebola Insights

Beloved aunty, Mammy Kumba, died from a stroke at her home in Barthurst, a mountainous village about six miles west of Freetown, Sierra Leone, at the start of October. Like any death this was a painful and traumatic experience for the family, but due to the timing it also put her relatives in a serious quandary.

The government has directed that bodies cannot be touched until they are 100% confirmed to be Ebola-free. Yet the weather is hot, sticky, humid – 25 degrees centigrade today according to Google. And as her doting nephew, Alhassan Fouard Kanu, explained in the Sierra Express Media:  “There have been instances where bodies nearly decomposed in homes during the waiting time for the laboratory result.”

The international media has talked extensively about how local customs have caused the Ebola virus to run rampage through this West African country. Yet new data patterns from IBM have also identified some surprising logistical issues, which have certainly seen the situation worsen.

“Believe it or not, most people are dying from other diseases outside of Ebola [in Sierra Leone],” explains Uyi Stewart, Chief Scientist at IBM’s Africa lab. There is also huge stigma associated with Ebola and “living family want to clear the name of their dead relative” as soon as possible. [Excellent BBC World Service, documentary.]

However, Stewart has discovered via local data patterns, that the real problem could be that the testing apparatus is completely overwhelmed. “We found out that only 50 tests can be completed per day and this creates a backlog of people who are waiting to clear the name of their dead relatives. This long backlog sometimes takes a week, sometimes 10 days to clear and what happens is that families become exasperated and take the law into their own hands.”

“[This means the families] eventually do what they are not meant to do and bury their dead…. and in some cases become infected. It is not only the cultural issue,” clarifies Stewart “it is not that they’re not listening to the messages. They set out trying not to bury the dead but it [just] takes so long to show that the relatives did not die from Ebola.”

This makes perfect sense on a human level. As Kanu wrote, albeit in a slightly different context: “It was a painful brooding moment for us as we wallowed in the fear that our beloved aunty will be buried in a very disrespectful manner; where men in overall suit will tend her; where her body will be thrown at the back of a pick-up van. We were worried over rumours that our aunty will be sharing a grave with another unfortunate Ebola infected body (ies). It was damn devastating.”

Not surprisingly, it is hard for those embroiled on the frontline of the Ebola crisis to remain dispassionate. And in the midst of panic and pandemonium it is extremely difficult tell what is really going on. This is exactly why the cold, hard, empirical evidence delivered by data could be a serious asset in the fight.  

IBM announced its humanitarian initiatives to help contain the Ebola outbreak today. But this has been running live for five weeks and has already highlighted some clear, actionable patterns such as the need for faster testing.

In practice, IBM works in conjunction with the highly respected community radio network to take phone calls and SMS messages in order to gather new intelligence. The calls are transcribed and added to the database along with the text messages received. From there, machine learning is applied to place the data into categories and identify the frequency with which each issue is raised.  This information can then be turned into a chart to map how much certain topics get talked about.

“We found out that the majority of people’s concerns were about corpses on the street,” explains Stewart. Yet he is also keen to stress this is not a “shallow analysis” because once big picture patterns have been identified, individuals from Sierra Leone’s Open Government Initiative physically verify the data. This in turn builds a wider, more concrete overview of what is really going on.

“Whether the person is in the streets, or at home,” the directive is not to touch the bodies, says Stewart. This means IBM gets “all sorts of calls”. However, they tend to fall into two categories. There is something horrific about cadavers rotting in the open which is destined to cause mass uproar. Yet there is something personally devastating about a dead family member rotting in the home.  This means corpses in the streets may get the highest volume of calls, but those with dead relatives in their houses are tired of “waiting [for the body] to be tested”.

IBM has an extensive reach through Africa and has been involved in numerous Big Data initiatives. Many of these have tackled infrastructure and health on the ground – the African cancer registry is a particularly interesting example.  It has also been involved in various crisis situations. However, the important point here is that while this presents clear short term benefits there are also crucial medium and long term benefits, which may have even more significance.

“Compare the situation in Sierra Leone with that in Nigeria,” says Stewart. “In Nigeria we were able to identify patient number one. Sierra Leone has lost the ability to identify patient number one. It is a mess because a lot of people have become infected.”  The “next phase” is to take a step back and look at the really big picture. This can be achieved by blending existing geotagged data with movement data from Telco partners to map the progression of the disease. “But in order to do that we need massive amounts of data.”

Stewart is “optimistic” if cautious and believes IBM could get something underway in “a few weeks”. In fact, over the last few days he has put together 10 simple person-to-person interview questions which will be put out to civil society leaders. These will be conducted with the aid of a GPS system to instantly location-tag the data.

“The reality on the ground is people want to leverage the technology in the fight against Ebola. The agencies have done some great work, but if we can generate this type of data and create models that allow us to map the progression it will allow us to do better containment,” adds Stewart.

Ebola has been around for four decades unchecked. And despite some hyped warnings in the 90s, it has taken this recent crisis to make the world sit up and take notice. Yet even as this vile disease rampages through west Africa, any evidence being gathered is both vital for getting to grips with what is going on now… and in helping to prevent problems like this from happening in future.

Imagine if these tools had been available at the start of the various Cholera pandemics, the Spanish flu outbreak in 1918 or even the Black Death. And think how much more we’d know today. This means that while this data analytics solution may not provide immediate relief in the extreme short term, it does create important, empirical information which could be very useful indeed, for many years to come.


Kathryn Cave is Editor at IDG Connect


« Indian Telecoms Industry: The Inevitable Consolidation


South Wales: Cyber Security in the Once Ignored Valleys »


Do you think your smartphone is making you a workaholic?