It’s a wet and chilly Monday morning outside the Addis Ababa Regional Laboratory (AARL). The thick metal gates open wide and a motorcycle with knobby tires and a tall yellow container attached to its rear roars up to the front reception area. In large block letters the container reads, “Ethiopia Mail.” Inside the yellow container is a battered blue cooler with peeling biohazard stickers affixed to both sides. Specimens in Ethiopia move through an agreement with the postal service.
The driver dismounts, removes the blue cooler, and carries it over to the receptionist’s desk. She extracts a pair of translucent plastic packs, which in turn contain a set of thin, stoppered vials full of dark red blood. The vials each have barcode stickers, which are read by a hand scanner. The receptionist also enters several columns of information manually into a spreadsheet. This includes information on the specimen’s origin and its departure and arrival time, which is connected to the laboratory information system.
We’re witnessing the intake process for potentially HIV positive and high-viral-load blood specimens.
Notebooks and pens in hand, we pepper the staff with questions: Is there a regular schedule for the referral of testing samples from healthcare facilities to the regional lab? Yes. Can we have a copy for mapping? Of course, but it’s all on paper. How often does any particular facility send samples to the lab? Once per week. Do you know how long it takes to return the results? It should be about one week, but it could be longer. Does part of the intake process always involve manual data entry? Yes, unfortunately, but some diagnostic machines such as CD4 and hematology are linked directly to the LIS. Is there anyone checking on this data in terms of structure and quality? The lab has a data quality team of 5 persons. Are all of your machines functioning right now? There’s one that’s been out for a little while, but it should be on the repair list.
Later, when we’re back at EPHI, I check my freshly built map. It tracks the national inventory of CD4 counters, and, sure enough, there’s a point for the AARL indicating that a machine is in need of repair. All of this process graphing points to positive signs that our spatial data integrations could work.
One of the LIS staff from EPHI has accompanied us to AARL. We brief the lab director about the BD-PEPFAR program and our GIS project for viral load testing and equipment maintenance tracking. Meanwhile, the LIS staff downloads five years of viral load specimen data for us onto a USB stick. We’ll be able to pair this dataset with the one from EPHI.
Our assessment takes about an hour and a half. Anmol has a conference call to make, so the rest of us decide to walk the ½km back to the hotel to work. Before we’re even past the laboratory gates a man walks up beside me and spits on the ground, getting a few flecks of it on my pants. I try to explain that everything’s fine and not to worry while I walk, but he insists on wiping the side of my leg with a cloth.
Abruptly the man turns to leave. I quickly realize that my Android is no longer in my pocket. I stop him just before he darts into the street. He returns my phone with a sheepish shrug. I suppose no harm has been done, but from now on I’ll keep my phone in my zippered pocket.
Back in our hotel conference room, Adam loads the AARL data into Tableau while Jessica and Anmol take over on Excel. We grind through another marathon geocoding and data cleaning session. By late-afternoon Monday, our vision is blurry, and the team is in dire need of food and caffeine.
But there’s good news: the data model derived from our EPHI work last week turns out to work for us once again, with only minor variations. We can see age and gender distributions sprouting, along with the spatial distribution of the specimen network for Addis subdivided by scheduled day, testing frequencies and viral load results. Now that we’ve completed data cleanup and formatting on a second major lab, there’s every reason to believe that most every LIS system in Ethiopia — at least those from the same vendor — ought to allow us to back out a spatial understanding of the specimen referral network.
My long-deferred idea from four years ago is one small step closer to reality.