Achieving consistent data use in global health is hard work. A colleague once quipped that despite advances in better routine surveillance systems, 80% of our time is spent collecting and managing data, 10% is spent conducting analyses, and 10% is spent pulling our hair out!
Aside from the hair loss, flipping the ratio of data collection to data analysis is a personal goal of mine. Since 2015, I’ve gained clarity working on the Visualize No Malaria initiative, a partnership between PATH, the Zambian Ministry of Health, the Tableau Foundation, and incredible private-sector partners helping to eliminate malaria in Zambia. This coalition is flipping that data ratio, designing visualization and analysis tools that allows Zambia’s National Malaria Elimination Program to focus more time on understanding data and to applying their analytic insights toward stamping out malaria.
A portion of an interactive dashboard developed for the Visualize No Malaria initiative displays declining malaria incidence.
Just as Steve Jobs remarked that computers were “bicycles of the mind,” I see visual data analysis as a powerful catalyst for everyone who works with data. Now, three years into Visualize No Malaria, we can reflect on the introduction of data tools and four guiding principles that support this work:
1. Be tenacious in the pursuit to expose routine data at the “richest grain.”
Guide your teams to assemble disaggregated data with more opportunities to include gender, age, location, and time. This provides more ways to slice data, and in the process create increased familiarity with it. This process can prepare users to take a closer look at unexpected findings that might indicate data quality issues.
Raw data downloaded from DHIS2 software platform for reporting, analysis, and dissemination of data.
2. Be data pioneers and explore data, pushing the limits of “normal” reporting.
Visualization techniques such as “stoplight” charts, league tables, or funnel plots allow users to identify outliers and patterns in the data that trigger users to think differently about the data and then apply insights to program improvements. If your data typically show annual progress against targets, then shape the data to include elements of time or geography to enable consumers to ask more questions on the fly. Dynamic data discovery tools allow users to pivot instantaneously from aggregate to disaggregated data. Whatever your “normal” standard is, go beyond it!
Funnel plot of HIV incidence.
3. Be advocates of “why” and push users to explain potential reasons for outcomes.
Building on the explore principle, seeing variations or patterns moves us to pinpoint the potential drivers behind them. Look for visualization opportunities, such as scatterplots or bubble charts, which can tell us more about the context or drivers of the health outcomes in the data. Push for analyses to find positive, negative, or no correlations between variables.
This map of Zambia’s network of water channels is used in predicting malaria incidence.
4. Support all data users and empower stakeholders to build their analysis and interpretation skills.
In global health, we expect data to drive better decisions, yet we still consider analysis as highly specialized. Instead, encourage everyone to be an analyst! Create data products that can help users improve analytical skills as they go, provide opportunities for them to add new insights in our tools, and ultimately embed mechanisms so they can readily share feedback and recommendations with others who need them.
SMS alerts can improve reporting rates and free up resources for other tasks.
These four principles—expose, explore, explain, empower—will not generate fancier reports or prettier dashboards on their own. But they can guide us toward better, more insightful data products that teams and counterparts can use to improve health and save lives. In future blog posts, we’ll expand on these principles and show them in action.