28 Aug 2015

Working with global nutrition data: chasing 193

Komal Bhatia Data Analyst for The Global Nutrition Report (2016)

Working with and assembling global nutrition data throws up unique challenges and insights, and engaging with these is an integral part of the process, as well as an outcome in itself. We share some experiences from data analysts’ perspectives.

The Global Nutrition Report’s dataset has 193 rows of data – one for each UN member country – which stretch out into more than half as many columns of nutrition-related indicators. This magic number is the basis for much of the Report’s content and the nutrition country profiles published online; we try to ensure that each indicator has data for as many of the 193 countries as possible before we include it in the Report (we have discussed inclusion criteria previously).

What is global nutrition data?

The field of nutrition in the global sense is emerging and expanding, reaching out to economics, demography, agriculture, health systems and food policy, water and sanitation, education, and development. While data on nutritional status assessments (prevalence of different forms of malnutrition in children and adults), behaviours (infant and young child feeding practices) and intervention coverage (treatment of severe acute malnutrition, vitamin A supplementation) are central to our work, casting a wider data net has enabled us to present a fuller picture of malnutrition and its determinants across countries and regions.

Assembling this data is easy when indicators are well-defined and data collection methodologies are consistent and internationally standardised. But such coherence is an ideal rather than the reality of gaping holes in the availability of essential, basic information about nutrition in so many countries. It can be frustrating when the differences are minor but significant; however, it also makes room for improvements and greater cooperation between countries and international agencies.

Using different definitions or indicators for each country would defeat the purpose of a global report, and so this is an option we do not have. Moreover, as we head towards the Sustainable Development Goals that will apply to all countries rather than just the developing countries that aimed for the Millennium Development Goals, achieving data consistency across the globe will be crucial to monitoring progress and accountability. And so the target score of 193 seems worthy after all.

Open, not closed

A fundamental premise of our work that hasn’t been emphasised enough is that the data we use must be open, i.e. freely available for public use, as far as possible. When datasets held by an organization aren’t publicly shared, we seek permission to make these available on our website and in the country profiles.

Achieving this goal has been fairly easy. Most data that meet other essential inclusion criteria (availability, consistent with theory of change, robust methodology, comparability across countries) are generally freely shared by UN agencies or other international agencies. In our experience, data that are desirable (e.g. data on consumer expenditure on food) but closed (e.g. only available by paid subscription to Euromonitor Passport) are often available for only a handful of large countries rather than all 193, and so easy to exclude. When updated datasets aren’t shared online (e.g. data on national implementation of the International Code of Marketing of Breast-milk Substitutes were last published by WHO in collaboration with UNICEF in 2011), we have found that organizations have been willing to share unpublished updated data with us, and are happy for us to publish this information (David Clark from UNICEF has shared updates with us in 2014 and 2015).

Surveys, modelled estimates, or both?

Some will call modelled estimates of global health indicators, such as obesity or mortality, a necessary evil – they are imperfect, but we can’t do without them. Others still will question the validity of survey results from a strife-torn country – how did data collectors go about measuring height and weight of children amid heavy warfare? These are valid concerns, but it is important to make the best use of what is available, constantly asses the methodological quality of data, and be honest and upfront about the limitations of nutrition data.

We include survey data as well as modelled estimates in the Global Nutrition Report; we prefer survey results to estimates, but in the complete absence of any global database of nationally representative surveys of robust methodology, we include estimates. Competing sources of estimates are reviewed for methodological soundness before including the better one. Transparency is an important part of our work, and we justify our decisions, such as the use of adult obesity estimates in 2014.

The role of gatekeepers

We are thankfully exempt from the impossible task of reviewing every single source of nutrition data. We rely heavily on the technical expertise of other gatekeepers of nutrition data, mainly our partner UN agencies – including UNICEF, WHO, and FAO – and other international bodies (World Bank, for example) that maintain databases in allied fields. Trust looms large in this relationship – in the trust that gatekeeper organizations inspire in us as a result of their technical integrity and commitment to nutrition, and in our readers’ trust that we hope to gain and maintain through our use of good quality data.

There are practical challenges to this: each organization has a different data cycle and calendar, and release dates are sometimes a few days after our data and publishing deadlines. Country classifications and data formats can vary a great deal, and it takes a lot of effort for us to curate all this data into a standardized dataset. Again, the order of 193 countries needs to be matched across all indicators.

But the experience of working with our partner organizations has been a very positive one – in 2014 and 2015. The sharing of data has gone beyond mere transfer of datasets, and we have relied on their technical input and guidance throughout the process. Our experience underlines the importance of a culture of sharing that makes it possible to work with global data on so many aspects of nutrition.

A day in the life of a nutrition data analyst

Someone at a Report roundtable event in 2014 asked us, “so do you have to go to all 193 countries to collect all this data?” Alas, putting together data for a global report doesn’t entail very much globe-trotting.

Collating and assembling a large amount of data takes time, extensive data management skills, painstaking attention to detail, and an infinite amount of patience as the datasets trickle in one by one. All of this can be achieved at a computer in an office building, and is more straightforward than the subsequent task of analysing the data and interpreting results.

We support the Independent Expert Group (IEG) in delivering the Global Nutrition Report, and most data analysis is guided by the narrative and framework agreed on through a collective, collaborative process. Initial, exploratory analysis serves as a starting point for some stages of this. We also provide technical input at all stages of the production process, ensuring that the findings, conclusions and messages are an accurate reflection of the data and are supported by the scientific evidence base on nutrition. This is a challenging task, because we sometimes don’t have the data to answer some big questions.

Assessing trends in malnutrition rates and country progress towards global nutrition targets can be tricky. Global numbers often paint a gloomy picture when many countries are making great strides in improving nutrition. National averages mask the granularity that comes with sub-national data at the state and district levels. Aggregating up to the global level is another challenge. The data tell us that childhood malnutrition is decreasing in many countries, but that adult obesity is rising everywhere. So is malnutrition increasing or decreasing globally? The answer is that we can’t be certain, because we don’t have the data to give us a reliable conclusion.

Data work can be exhilarating and disheartening in equal measure, but once you’ve built a structure using it, it’s difficult to imagine how you could have done it otherwise.

Komal Bhatia (Institute of Development Studies, Brighton, UK) , Kamilla Eriksen (MRC Human Nutrition Research, Cambridge, UK) and Natasha Ledlie (International Food Policy Research Institute, Washington D.C.) are Data Analysts for the Global Nutrition Report.