This is the second Labs for Life report from Ethiopia (read Labs for Life: Ethiopia – Day 1).

Most of the hard work in GIS has nothing to do with making maps. The heart of the matter, from a functional point of view at least, is not the pretty pictures but the quality, sourcing, content and structure of the data which feeds those images. Without quality data, there can be no quality GIS.

This lesson has been driven home with a vengeance over the past couple of days, as we’ve struggled to formulate a prototype that is capable of visualizing Ethiopia’s viral load specimen referral network. From where exactly will the data come? In what shape is it? How much work is required to clean it up? Do we even have sufficient permissions to proceed?

The team huddles on Wednesday afternoon after the initial conceptual GIS presentation and concludes that the only viable short-term strategy is to zero in on EPHI itself. We’re going to base ourselves at the prototype stage on historical data from EPHI’s laboratory information system (LIS). The LIS integrates several testing systems, including biology, chemistry, hematology, CD4 and viral load, as well as TB sequencing. Since it’s the national reference laboratory, EPHI receives at least a few specimens regularly from a landscape of clinical sites scattered throughout the country. From referral locations and key indicators such as testing totals, turnaround time, and test results, their LIS should contain enough information to illustrate at least one slice of Ethiopia’s viral load specimen referral network.

Once we’ve decided on a course of action, Adam and Anmol run over to track down Tigist, EPHI’s IT director. Explaining our plan, they’re able to secure permission from her to utilize the past five years of viral load data for our initial experiment in specimen referral mapping. We’re in business.

By the morning, it’s clear once again though that even modest projects with sufficient clearance face enormous challenges. The LIS was not designed with spatial analysis in mind, so it doesn’t yield easily to our goals. Like many datasets over which one has no authorship or control, there is a wide, muddy field to cross in terms of data cleaning and organization before it can become even moderately readable in GIS.

On Thursday, while one of our team members recovers from illness, the remaining group digs into the data with a local advisor, assigned to train with us, from Heal TB. The stark reality is that there is no standard in place for location descriptions, nor a matching set of IDs to leverage into viable data integration units. The geocoding problem, in particular, or the identification and attachment of coordinate points for mapping, beckons us to climb down deeper and deeper into the muck of tedious detail.

But the team is persistent. After several hours’ worth of careful code-matching, spreadsheet restructuring, field parsing and manual lookups our raw material starts to look like it contains the basis of cartographic form.  We have ourselves a working core dataset.  

Back at the hotel Thursday night, we rejoin our recovering colleague and manage to convince the concierge to open a quiet conference room up on the 12th floor for us to work. Up until the wee hours Friday morning we’re merging datasets, measuring indicators, and hammering out the first draft mapping applications in ArcGIS Online. Around 2 am East Africa time is the very first moment any of us sees this network on-screen. Even partially complete, it’s sort of exhilarating.

Friday morning, we’re back at EPHI to reconnect with primary stakeholders and review our progress. They’re a little bit stunned.  In just one week, we have pulled together a functioning model web application that enables spatial analysis of a laboratory network, in a way never before seen. Hotspot analytics are enabled in the desktop and the browser. Core indicators have been loaded into a geodatabase. By the light of our Powerpoint slides, feedback comes fast and furious from all corners of the CDC conference room.

As the first week comes to a close, the BD-PEPFAR group has its marching orders: widen the sphere of activity from EPHI to the regions, refine the data model and expand from the initial prototype phase towards viable and shareable GIS tools. 

Starting Monday, the team is heading from EPHI to the regional labs, beginning at the Addis Ababa Regional Lab and moving the following week to the Adama Regional Lab. Ethiopia’s laboratory landscape is taking shape.