Fighting Ebola with Elasticsearch: Cause Award Honoree eHealth Africa

This is the first in a series of blogs highlighting the Elastic Cause Award honorees: their projects, teams, and the causes that motivated them. This week we’re featuring Adam Thompson, the Executive Director of eHealth Africa.

eHealth Africa designs and implements data-driven solutions and technologies to improve health systems for and with local communities across West Africa. They currently employ nearly 800 staff in Nigeria, Sierra Leone, Liberia, Guinea, the United States, and Germany. Their technology works in low connectivity settings, and smartly uses data to drive decision-making by local governments and partner agencies to get optimum results.

The following excerpts are from Thompson during the closing keynote at Elastic{ON} 2017 and have been edited for length and clarity. Editorial commentary (in bold) is sprinkled in to add context. 

When Ebola struck West Africa late in 2014, eHealth Africa already had infrastructure in Nigeria and they quickly helped stop the outbreak. Their success there prompted the health ministries in Guinea, Liberia, and Sierra Leone to request their help.

"We hit the ground behind. Behind the virus. It was taking one, sometimes two weeks between a suspected case being identified and getting a laboratory confirmation to start contact traces. At the speed Ebola works, we were getting further and further behind all the time.

We needed to identify suspected cases, get them quarantined get lab results, and get contact tracing to figure out who the they had contact with over the last several days. That needed to happen in 24 hours. In order to do that we really needed visibility. We only saw the people showing up to the hospitals. We needed to broaden that picture."

They decided to expand the health call lines that existed in the country — and they needed expanding.

"What we were able to do was expand these health call lines that had mostly been dormant in the countries. They were reactivated early on in the response, but were really kind of limited to one or two people on phones, writing down stuff mostly on paper. And we still weren’t getting enough information. And we weren’t getting that true picture of the outbreak.

So we built a call center application. We have our own software team and we do a lot of work with open source tools. But we looked around and we needed a tool that was going to allow us to record data at a central location with all of those calls from around the region and we needed to distribute that data to the district centers. The centers that were spread around the country where they would get the data, respond to the alerts, and then add in the follow up alerts and the rest of the data. We needed to build a solution."

They started with one NoSQL database solution because of its powerful synchronization features for partially offline architecture, which was important for remote district offices in areas of low connectivity. But it had primitive indexing and reporting. They needed more capability.

"We stumbled upon Elasticsearch after a bunch of late night Google queries and realized it was something that we could try to dramatically improve the way that we were searching our data because the database was growing so quickly. We constantly had to refactor and rewrite the application because we had no idea the volume this response was going to take on."

In Sierra Leone, they were reaching 8,000-10,000 calls a day, and in Guinea, 10,000-20,000. People reporting symptoms or deaths. People asking questions.

"Elasticsearch allowed us to immediately index all the different facets of the call center. And as we were adding new data elements in, new parts of the response, we could quickly update the indexes and get new reports based on the new elements. Kibana let us set up these visualizations and snapshot in time reports so that every 24 hours our response teams met in the morning and evening, they’d get reports on all aspects of the response in real time so they could see a picture — the whole picture — of the country for the first time in terms of what was going on."

They quickly saw that the outbreak was much bigger than the number of people showing up to treatment centers indicated. Elasticsearch allowed district office staff to filter and search local cases and respond rapidly.

"Every single person that called this call center, 24 hours later they got a call back. 'Hey, you reported a death, you reported a sickness. Did anyone come? Did anyone show up? If they did show up, did they do these things?'

We created a feedback loop for the actual response [teams] themselves so they could quickly iterate through and improve the situation. We went from a two-week response window down to a 24-hour response window in two to three weeks just by constantly reiterating and running through the data and making sure the people making these decisions had access to the real time information."

And the result of tireless work of all the responding agencies informed by real time-data?

"Responding to public health emergencies is all about time. And every minute, every second, every day that goes by where you don’t have visibility on the response is time you lose and it’s lives lost. So, the call center application, the data team, all the people in the responding agencies are really hooked into this application and used this data to drive decision making and drive Ebola out of these three countries."

The call center played a crucial role in curbing the Ebola outbreak in 2014 and currently operates as a universal health emergency number in Sierra Leone. eHealth Africa is currently in the process of determining how the solution they built with Elasticsearch can best be modified to support health systems strengthening in the years to come.

Watch Thompson's full presentation from Elastic{ON} 2017 and learn more about the Elastic Cause Awards.

Banner image credit: CDC / Frederick A. Murphy.