For many nutrition researchers, our analyses begin with data that describe what people are eating. But as we know, there are enormous gaps in the availability of high-quality dietary datasets. As a result, we rely on less-than-ideal measures, like average per capita intake based on food supply data, single-day recalls, or indices, which can obscure our results and conclusions. In our recent paper published in The American Journal of Clinical Nutrition, we attempted to address this challenge by consolidating dietary data on people’s usual nutrient intake in 31 countries, estimating what intake distributions look like for different age and sex groups, and making these distributions publicly available for nutrition researchers. I also recommend reading this great editorial by Jessica Fanzo accompanying the paper.
This is an important step forward because the analyses that nutrition researchers do often require data on more than just the average level of consumption in a population. We need to understand the full spectrum of a population’s nutrient intake—in other words, the shape of the distribution around the mean or median. We may need to estimate how much of a population is below or above the recommended nutrient intake, design appropriate nutrition interventions and predict their impacts, or determine which segments of a population are most susceptible to chronic diseases. If we get the shape of the intake distribution wrong, we might grossly overestimate or underestimate what proportion of a population is at risk. Historically, researchers have had to make an educated guess about what the shape looks like. With this study, we provide an empirical and open-source basis for making this decision. We based our distributions on actual dietary data, which can allow researchers to estimate population-level nutrient adequacy with greater precision.
Our paper also disaggregates these distributions for different age and sex groups. When we are forced to rely on nutrient supply data, we often have to make difficult assumptions about how nutrients are distributed among subpopulations. We might misunderstand entirely how inadequacy differs for children, women of reproductive age, seniors, or adolescent boys or girls. Such information is critical for targeting interventions based on different nutrients for different populations at different stages of life.
My personal favorite contribution of this paper is that all of these distributions are now available for you to use in an R package. You can look at our code and apply these methods in your next analyses. It allows the calculation of nutrient inadequacy in a population (with the Dietary Reference Intakes already built in), the calculation of variance and skewness, comparison of distributions across populations, the shift in distributions in response to an intervention, and fitting distributions around the means in any dataset. It’s also fun to browse the distributions for different countries and nutrients in this interactive R Shiny web application.
While this research is an important advancement, no method of estimating dietary intake is perfect. We could only calculate approximately precise distributions for the populations for which we had two days of dietary intake data. Many datasets are still not publicly available. And of course, the underlying surveys and food composition tables carry their challenges. But we hope that making our methods transparent and accessible will help researchers estimate population-level indicators and impacts with a greater degree of nuance.
There has been a great deal of progress in recent years in developing more precise estimates of what people eat around the world, like the GENuS Database, Global Dietary Database, and Global Nutrient Database. With our new paper, our aim is to contribute to this growing body of work—and support other researchers engaged in this vital field.