Food and nutrient dataset

The challenge

The ability to accurately estimate dietary intake is fundamental to nutritional epidemiology. Traditionally, dietary consumption data is collected by paper based methods such as food frequency questionnaires, food diaries and interviewer administered 24-hr recalls.

Using a paper based method, the researcher is required to ‘code’ dietary intakes in order to provide data on energy and nutrient intake. The coding process is aided by a food composition database (FCDB) which usually contains country-specific, detailed information on the nutritional composition of foods.

The current edition of the UK food composition tables contains ~3,300 mostly generic food and drink items. The tables are also available in electronic format as an integrated dataset.

To better reflect the range and diversity of food products available in the UK and to meet the needs of the myfood24 dietary assessment tool, a new comprehensive electronic FCDB was developed by the Nutritional Epidemiology Group at the University of Leeds.

Building the dataset

A mapping exercise was undertaken to match ‘Back of Pack’ data (energy (kcal), protein (g), fat (g), saturated fat (g), carbohydrate (g), total sugars (g), fibre (g) and sodium (mg)) from branded food products to nutrient data from the generic food codes to create the complete nutrient dataset in myfood24. In addition to the mapping process, a full quality control activity was performed to ensure the accuracy of the data.

Additional database formatting was then carried out in order to enhance the searchability of the myfood24 dataset. These included; 1) Amending food descriptions to ensure search results displayed more popular foods first. For example, a search for milk would return ‘milk pudding’ near the top of the list so the descriptor was changed to ‘pudding, milk,’ (is this helpful or too much detail?) 2) Common synonyms (e.g. ‘coke’ appended to ‘coca cola’) and potential misspellings added to the database to aid searching, 3) Serving size description added where necessary to clarify serving unit (e.g. for powdered foods servings might be ‘as made up’ or ‘as powder’) and 4) Common accompaniment foods (i.e. milk with tea) added in a series of reminder prompts.

The myfood24 FCDB development process, including sourcing the food composition data and cleaning and mapping of micronutrients to back of pack (BOP) data, took approximately 18 months, involving a small team of nutritionists, a dietitian, data entry assistants and a database manager. The result is a high quality dataset that can meet the rigorous standards demanded by academic research both in the UK and internationally.

Future proof

But things do not stand still. The food and drink industry in the UK is the largest manufacturing sector and invests substantially in research and development. An estimated 10,000 new food and drink products are introduced each year and other products are discontinued as retailers react to changes in demand. Many of these products might only be available for a very short time and sometimes just for a matter of days if they are for seasonal occasions. Nutrient values in established products also change due to product reformulation. So, the management, quality checking and updating of the database never stops!