Novel sources of data for monitoring food and nutrition

Monitoring food and nutrition is of significant public health importance. Overall, dietary risks were the third-leading risk factor contributing to the total burden of disease and injury in Australia for 2015. Monitoring requires timely, reliable, consistent and accessible data related to food supply, food purchasing and acquisition, food and physical activity behaviours, and nutritional status.

There are, however, limitations to many of the data sources traditionally used in food and nutrition monitoring. Several of the traditional data sources are infrequently and/or irregularly collected, are subject to various biases, are expensive to collect, and/or have a high participant burden. The increasing creation and collection of data via less traditional means provides opportunities for novel sources that can potentially fill some of these gaps.

This report provides information about novel data sources that could be used for monitoring food and nutrition to guide future use and strengthen the evidence available for policy makers. It includes example analyses of 3 data sources—purchase data, electronic payment data and location data—to highlight the potential uses of, and considerations when using, such data sources.

What are novel data sources?

Novel data sources—in the context of food and nutrition monitoring—are sources of data that were not collected for statistical purposes and are yet to have been extensively used for these purposes. Novel data sources that could be used in food and nutrition monitoring include market share data, purchase data, electronic payment data, location data, and app and wearable device data.

What are the potential uses of novel data sources?

The potential uses of novel data sources depend on the data source, with different data sources better suited to different uses. Market share data and location data could contribute to monitoring food supply, while purchase data and electronic payment data could be used in monitoring food purchasing and acquisition. App and wearable device data could be used in both monitoring food and physical behaviours and monitoring nutritional status.

What did the example analyses show?

Comparing novel data sources against traditional data sources can be useful in broadly validating data. Findings from the Australian Bureau of Statistics’ Apparent Consumption of Selected Foodstuffs 2018–19 showed that:

• the percentage of energy purchased from discretionary foods (38% in 2018–19) was similar to that reported as consumed in a population nutrition survey (35% in 2011–12)
• there were differences in the relative contributions of macronutrients between energy purchased and consumed which potentially reflect the different scope of the 2 data sources (supermarket purchasing compared with overall diet).

Electronic payment data from 2 of Australia’s largest banks were separately analysed to investigate average spending on dining out or fast food, average frequency of dining out or fast food transactions, and average dining out or fast food transaction value. The results of these analyses were difficult to interpret, given differences in use of payment methods over time and between sociodemographic groups, and require careful consideration.

The location of food services (cafes, restaurants, and takeaway outlets), from the Australian Business Register, was used to investigate associations between the density of food services and overweight and obesity. Broadly, Statistical Areas Level 3 (SA3s) closer to capital cities had higher densities of food services than rural areas, although there were exceptions.

What are some of the key considerations when using novel data sources?

Novel data sources can complement, but not replace, more traditional data sources. While some novel data sources offer advantages over more traditional data sources—such as continuous data collection, large sample sizes, prospective and objective data collection, and frequent, regular and timely data provision—there are also limitations. These include issues with data coverage and representativeness, consistency and comparability of definitions, transparency of data collection and analysis methods, and commercial sensitivities.

Given the above, the quality of novel data sources should be assessed for specific research questions and, where possible, the novel data sources should be validated against more traditional data sources. Opportunities to work with the data providers and owners to further develop and standardise the data should also be explored.