The Problem with the Data Revolution for Global Health

My first job in public health was with a company that shared the same name as a main character in Game of Thrones. I went to painstaking lengths to explain to my friends, immigration officials, and even other public health professionals that the story of their favorite TV character actually paled in comparison to that of the real John Snow. Often called the father of epidemiology, Snow determined cholera was a waterborne disease in the late 1800’s by marking the residences of affected individuals on a map of London. He subsequently observed that cases were clustered around Broad Street, an area that shared a public water pump.

Today, the very same techniques that Snow used — combining geographic information with disease incidence — are a hallmark of information systems for public health, and are often the premise behind the dialogue surrounding big data for development. With the advent of technology and the plethora of information available to us today, Snow’s conclusion would have taken far less time to develop had an outbreak occurred in 2017.

John Snow’s map of cholera cases in London. (Source: Wikipedia)

Over time, the public health community has fiercely advocated for data-driven development, issuing calls for improving the evidence base and utilizing this to improve best practices. This approach has seen global initiatives such as the Sustainable Development Goals (SDGs), and its predecessor, the Millennium Development Goals (MDGs), carefully lay out internationally recognized indicators to measure progress toward a brighter future for all. These global indicators help to fill data gaps, provide standardization for cross comparison on development and health across countries, guide international funding priorities, and increase accountability of governments and practitioners.

The Sustainable Development Goals (Source: United Nations)

In a 2015 needs assessment of the resources required to support the collection and analysis of SDG indicators, a consortium of experts suggested that 77 low-income countries collectively required $1 billion annually to operationalize proper data systems. However, this report fails to acknowledge a very important stakeholder in development data: country-level end users. Often, strengthening country health information and statistical systems requires significant resources to support long-term efforts to build capacity for data collection and analysis. Cost-saving measures ultimately leave end users at the country level out of the process of determining global health estimates. As Carla Abou-Zahr and colleagues suggest, more often than not, this information is most useful for the individuals that calculate the indicators — academia and international development agencies. So this begs the question:

How can we support developing countries to collect data that is useful for them?

At Akros in Lusaka, Zambia, where I’ve worked as a Global Health Corps fellow and now as a full time staff member, this question lies at the heart of our programming. Akros believes that tangible changes in health can occur when individuals are empowered to understand information and to use that information for decision-making purposes, from the national level down to the community level. International efforts to advance technologies for information systems often don’t take into account that data systems alone do not elicit change. A data system that is not used, is used incorrectly, or is filled with “bad” data, will prove ineffective. We’ve found that when use of data is minimal, its quality (i.e. accuracy and timeliness) is also poor, resulting in a vicious data cycle.

The virtuous and vicious data cycle demonstrating the relationship between data quality and data use.

Alternatively, the virtuous data cycle model proposes that as data is understood and used for decision-making, its quality inherently improves. If the data cycles continue, the increased speed of feedback to users of data instills a greater sense of ownership of data, which also drives improvement of data quality and use of data to make real decisions.

Community-led Total Sanitation

A hallmark example of this is Akros’ work with UNICEF to support implementation of a community-led total sanitation (CLTS) approach in rural Zambia. CLTS is a behavior-change intervention focused on mobilizing communities to eliminate open defecation through construction and use of latrines. UNICEF approached Akros to build an information system to monitor sanitation access, and Akros delivered a simple and easy to use mobile-to-web platform for village level data collection from community volunteers. While the data was simple, easy to understand for those collecting it, and paired with incentives for timely reporting, Akros discovered that the data did not translate into behavior change, and sanitation access remained stagnant. Akros and UNICEF then had to re-evaluate who the opinion leaders and decision-makers were in these communities. As a result, we realized that we were leaving out traditional leaders in our data-driven approach.

Chief Singani of Cooma Chiefdom in Zambia’s Southern Province displays the tablet dashboard he uses to track sanitation progress in the many villages he oversees. Involving traditional leaders like Chief Singani in the community-led total sanitation (CLTS) process has prompted a distinct spike in hygiene practice uptake. (Photo Credit: Andy Prinsen, Akros)

Working with developers, Akros created a widget to pull data from the information system onto an Android interface, accessible on a smart phone or tablet. The graphs display sanitation access within each chiefdom and those around it. Now chiefs are able to monitor their chiefdom’s progress toward open defecation-free status, and use the information to make decisions about encouraging their communities to build toilets. After this intervention, Akros and UNICEF saw a steep increase in sanitation access, leading to the first open defecation-free district in Zambia. Today, the Akros supported Water, Sanitation and Hygiene Management Information System under the Ministry of Water Development, Sanitation, and Environmental Protection in Zambia is the largest known community-level surveillance system in Africa, covering over 60 districts and 40,000 villages.


Akros used a similar approach to redesign structure tracking for indoor residual spraying (IRS) against mosquitoes in malaria endemic areas though the President’s Malaria Initiative-funded Africa Indoor Residual Spraying Project. While IRS has been a malaria prevention practice for many years, there are some shortcomings. IRS coverage is measured as a percentage — the number of houses sprayed relative to the number of houses visited. However, this assumes that 100 percent of the houses in a particular area are visited, and of these, at least 85 percent are sprayed in order to achieve protective effect — when in fact, many resting structures are missed altogether. This assumption leads to considerable over-estimations of IRS coverage, despite parasite prevalence remaining high.

Furthermore, the process of finding and counting dwellings — what we call enumeration — is difficult and costly. It requires deployment of enumeration teams to planned spray areas where they physically identify every structure in a community, and then mark these locations with GPS devices. As one can imagine, it is challenging to ensure that every structure is located using this process. After the enumeration, when spray teams begin the process of spraying homes, spray technicians often struggle to track which dwellings they have sprayed and to understand whether they are reaching enough homes for protective effect, nor do they have any tangible feedback on their performance due to paper-based reporting lag times

Akros’ approach to IRS management is encompassed in mSpray, a mobile data collection tool for in-field spatial management that feeds into a web-based dashboard to drive decision-making for IRS coverage. It consists of three components: enumeration, targeting, and spray operation.

Using DigitalGlobe satellite imagery and open source GIS software, teams of Akros enumerators located and traced dwellings in hard to reach areas from computers at our base of operations in Lusaka, a faster and more cost-effective process than traditional enumeration. More importantly, at 95% found coverage, satellite enumeration led to identification of significantly more structures in each community, meaning the enumeration was increasingly more accurate and complete.

Targeting is a step designed to help stakeholders, such as spray operators and spray managers, make the best evidence-based decisions regarding where IRS interventions should occur, when, and the supplies and staffing required to make each spray endeavor a success. Akros approached this component using decision-making principals, developing a tool to ensure decisions are made based on evidence and data with assistance from feedback mechanisms. A 2016 assessment reviewed the effects of of mSpray’s protocol on district decision-making to maximize spray coverage per community. The protocol consists of procedural instructions, backed up by spatial feedback and a daily decision-making form to determine if sprayers achieved an 85% threshold for IRS coverage. The protocol prompts spray operators and team leaders to determine why they did not achieve their targets and how to take remedial action, including returning to the spray area. Out of 541 spray areas in one spray season in seven districts of Luapula Province, district managers received 416 forms from each spray area that did not achieve an 85% target, indicating a high use of data for decision-making. 374 spray areas were revisited because of the decision-making protocol. As a result, spray effectiveness went from 47.97% to 78.71% in the areas that were revisited, an increase of 30.68%.

The final component of mSpray is spray operation. In 2013 Akros partnered with Ona, a social enterprise technology company, to create a web-based tool, accessible through portable tablets, that helps spray teams locate dwellings to be sprayed and tracks their spray progress in real-time. Lessons learned from implementation have informed several iterations of the mSpray platform. Where once before the tool simply provided an overlay of a spray area map on top of an enumeration map with static numbers of structures, Akros and Ona worked together to convert the enumeration files into Open Street Map layers to allow sprayers to collect information against the map. In this way, the number of eligible structures for IRS within a spray area can be verified and marked for spray completion by the spray operators themselves.

An IRS spray technician uses the mSpray application to find homes and mark them as complete after they have been sprayed. The mSpray approach makes IRS campaigns more effective and much more efficient. (Photo Credit: Andy Prinsen, Akros)

These examples demonstrate the powerful nature of data when it is placed in the hands of decision-makers at country-level. At Akros, we find that increasingly relevant data, combined with data use promotion and capacity building, result in measurable improvements in health outcomes — an approach the global health community would do well to adopt.

Aimee Edmondo was a 2016–2017 Global Health Corps fellow at Akros in Zambia.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.