Fieldset
“All data has a story, and it’s an epidemiologist’s job to tell it”

Is it a few isolated cases? Or the beginning of a mass outbreak? When MSF needs answers like this, we turn to our epidemiologists. But how do they get them? Julia introduces a new tool that’s made getting those answers much faster, with potentially life-saving results…

An MSF car on a dusty track in Chad

Within a week of arriving in Chad I found myself bumping along in an MSF Land Cruiser on a three-day journey to the south of the country.

My travelling companions, members of the Chad Emergency Response Unit (CERU), pointed out sights to me along the way: nomads with their herds of camels, hippos submerged in the river (followed by a long and heated discussion about whether they were hippos or rocks), and other landmarks.

Over three dusty days we made our way south, reaching our final destination – Béboto, Chad – near the country’s border with Central African Republic.

Our reason for travelling was simple: measles.

 

A jump in cases

CERU had been monitoring the number of measles cases in Béboto with data provided weekly by Chad’s Ministry of Health. As January progressed, CERU became increasingly concerned. Shortly after I arrived, cases jumped from 23 to 41 from one week to the next.

We decided MSF needed to get on the ground, assess the situation first-hand, and decide whether these were isolated cases, or signs of something more serious.

We were a small team of seven to start: a project coordinator, nurse supervisor, data manager, health promoter, logistician, two drivers, and me, the epidemiologist. Epidemiologists use data to understand how diseases spread and potential strategies for controlling them.

 

Like little ovens

We set up our base the evening we arrived – an old concrete structure with six rooms. For the next few weeks we worked there when we weren’t out visiting health centres or the local hospital.

We set up our laptops on long tables and benches outside, moving frequently as the sun passed overhead, hiding from its heat in the shade of a large mango tree.

base_1200.png

The MSF base, with Julia's tent out the front

I pitched my tent just outside the building. There was some incredulity among the team that I would prefer a tent to a room indoors, but as someone who was used to camping, this felt more like home than the airless rooms with their concrete walls and empty metal bed frame.

Not to mention it was already a breathless 40-something degrees during the day and I expected the rooms would become little ovens; everyone was quick to warn me that the temperatures would only continue to climb: “just wait until it’s 50 degrees in the shade!”, was a favourite refrain, followed by chuckles.

 

From handwriting to high-tech

Our first stop after setting up the base was to speak with local authorities, where we explained the reason for our presence and asked for permission to travel within the region. With their approval, we set off to investigate the potential outbreak.

Each day, we travelled in a convoy of two Land Cruisers to one or two health centres (there were thirteen health centres to visit in total).

The team would split up upon arrival and speak to different people – the head(s) of village, health centre managers – depending on our roles. I was fortunate to work alongside a data manager, so he and I asked to see the handwritten health centre registers containing patient records.

We perched on borrowed chairs in the shade and used a laptop to note the details of any measles patients who had come to the health centre in the last few months.

 

Asking the right questions

As an epidemiologist, part of my role during an outbreak is to determine the size of the outbreak and who is most at risk. The medical team needs this information quickly to make informed decisions on where, when, and how best to respond.

To do this, MSF’s data manager and I collected data from the health centres and conducted analyses, focusing on three categories: person, place, and time.

There are a multitude of questions you could aim to answer within these three categories.

How many cases are there?

What are the demographics of those affected?

How many have died and what are their characteristics?

How might the disease be spreading?

Which villages are most affected?

Which health centres are seeing the most children?

When did cases first start?

How quickly are cases increasing?

How long are children staying in hospital? when was the peak of the outbreak?

Those are just some of the many questions we seek to answer – and answer quickly – with data. The faster we can do this, the more quickly the medical team can take action, and the more lives can be saved.

 

Finding the right answers

To do my job, I use statistical software and code. This allows me to turn data from a collection of numbers and letters into a story about the population as a whole.

There are several software options available, but I am most familiar with “R”.

R is used for data manipulation, calculation, and graphics. Luckily for me, MSF created “R4epis”, an initiative that offers tools for epidemiologists who use R.

One of those tools is a template for outbreak situations, which includes sample code to clean and analyze your data. This template helped me overcome a few obstacles in the coding and analysis process while in Béboto.

For one, coding requires a lot of Googling. If you can dream up a question to ask, then chances are you can answer it with the right code. But it’s an entire language, and most people don’t have it all memorized. So often a good deal of this process involves online searching and trial and error.

 

The challenge

This was where things got challenging for me. Béboto is quite rural and our internet connection was extremely limited, to say the least.

Our team brought a portable modem that we could use as a hotspot.

However, not only were we in a zone with only 2G internet, but with the whole team connected to one modem, it barely worked. Sending one email took hours, and sometimes you’d check the next morning to find it still in your outbox, taunting you. Quickly searching online for the answer to a coding question was just not happening.

 

The computer cowboy

In this environment, the R4epis template code was extremely helpful. I still needed to do a fair amount of “data wrangling” (I love this term, it makes me feel like a computer cowboy) to get my data in the right shape to run the code without errors, and I needed to write my own code for certain portions.

But it gave me an outline to work with, questions to answer that I hadn’t thought of, and ideas to modify existing code I’d already written.

 

The highs and lows

Of course, it didn’t always run smoothly. If there was an error in the code or the data wasn’t in the right format, being unable to google solutions left me pretty frustrated until I found a solution on my own.

Co-workers would frequently witness me swearing under my breath followed moments later by shouts of glee – these are the highs and lows of coding.

Overall, having the R4epis template saved me a lot of time, and time is at a premium in an outbreak. Once I had the code written and ready to run, doing analyses each week as I received updated data was a lot quicker.

 

What we needed to know

I could produce graphs and tables with just a few clicks. And these graphs and tables showed us early on that Béboto was indeed in the midst of a measles outbreak, and needed MSF’s support.

Our team set to work, supporting the local health centre managers and hospital staff on best practices for identification and treatment of measles, providing free medication, and, ultimately, conducting a vaccination campaign for all children aged six months to nine years.

In all, 25,000 children were vaccinated.

tree_1100.png

Collecting data under a tree

 

Solid data

Throughout all these activities – determining if there’s an outbreak, providing treatment, and conducting a vaccination campaign – gathering solid data and conducting analyses were essential.

For example, during the vaccination campaign, MSF’s data manager tracked and reported the vaccination numbers each day to the medical team.

That way we knew how many children we’d vaccinated in each health zone, and if there were areas with low coverage where we’d need to return.

 

Still more questions…

Even after the vaccination campaign, there were more questions to answer…

Did MSF vaccinate all the children we needed to in order to protect that age group from future outbreaks? (We did!)

What lessons could we learn about the reasons why children hadn’t been vaccinated during the MSF campaign or previous campaigns run by Chad’s Ministry of Health?

 

And more answers!

Answering these questions are part of what’s called a “vaccination coverage survey”, which the Data Manager and I conducted directly following the vaccination campaign.

And fortunately, R4epis has a template for those analyses too!

I’m glad I was able to contribute to MSF’s measles outbreak response in Béboto. To me, data is much more than just a spreadsheet filled with information. All data has a story, and it’s an epidemiologist’s job to tell it.

--

Read more: Stories about measles

Fighting an epidemic in a pandemic

How Merveille beat measles (with some help from MSF)