top of page

METHODOLOGY

The Swiss Nutrition Atlas takes a consumption-based approach to assessing the population’s dietary intake. Consumption is estimated based on household shopping data. The methodology of the Swiss Nutrition Atlas is designed to be fully transparent and replicable and can, in subsequent years, be collected by the IRM-HSG or other research bodies and compared with the initial survey.

Data collection shopping data

Screening questionnaire

A screening questionnaire was used to randomly select households from the three largest Swiss language regions. In addition to the demographic household characteristics, the households were also asked which persons living in the household could provide which shopping data. An expected number of receipts could be calculated based on the frequency of shopping and consumption.

Uploading the receipts

Over a two-week period (21 February 2022 to 6 March 2022), households photographed their receipts and uploaded them via an online portal. For each receipt, households had to indicate, using a constant sum technique, the proportion of the receipt that (1) the person completing the questionnaire, (2) other persons in the household (in a second step, again divided into the number of persons) and (3) other persons outside the household had consumed or would consume.

Final questionnaire

The final questionnaire consisted of three parts. In the first part, the receipt uploads were evaluated. The second part consisted of requesting nutrition-related data (e.g. how important nutrition is to the participant or what healthy nutrition means to them). In the third and final part, additional demographic data of the members of the participating households were collected (e.g. food intolerances, sporting activities, weight and height, etc.).

Collection of nutritional data

Nutritional data from retailers

The receipt abbreviations were sent to the respective retailers with the request to provide nutrition declarations and ingredients as well as relevant master data (e.g. net weight) for the corresponding products. Three retailers complied with this request.

Nutritional databases

Since it was impossible to assign nutritional data from retailers to all products, other databases and data sources were used:

  • Non-public database of ETH Zurich with around 50,000 food products;

  • Swiss nutrition database;

  • Product information directly from manufacturers and dealers (e.g. website or via enquiries from Xyxle);

  • Databases from other countries (e.g. “WebAppendix” by Louie et al. (2015), USDA)

  • Modelling consumption.
     

Modeling consumption

After linking the nutritional values per 100 grams or 100 millilitres and calculating the free and added sugars per 100 grams or per 100 millilitres, effective consumption is modelled. For this purpose, the total content of the nutritional values, as well as the free and added sugars, are calculated for each receipt item:

Intake_EN.png

This is followed by the distribution of the nutritional value per item with the value of the constant sum estimate to the individual members of the household.

This is the basis for consumption per person modelling. It must be considered here that the sum of all item positions per person does not correspond to the actual intake. A simple summation of the values merely reflects the sum of the products purchased and shipped. To approximate the actual intake, the sum of the “purchased and shipped” nutritional values per person is corrected by the following four factors:

  • Correction for the share of foods for which no nutritional values could be calculated;

  • Correction for the share of foods for which no total amount of nutritional values could be calculated;

  • Correction for the share of foods for which no receipts were sent in;

  • Correction for expected food waste.
     

Data cleaning

To clean the data, the following steps were taken at the household or individual level:

  • Two persons were removed from the random sample due to implausible weight information (older than 18 but below 30 kilograms).

  • Since 28 people consumed zero kilocalories according to the modelling, they were removed from the sample.

  • Because it was indicated that only the receipts of person 1 of a household (person of the household who was recruited via the panel) could be provided, 361 people were removed from the random sample. 

  • The final random sample consisted of 456 people from 371 households.


The Swiss Nutrition Atlas uses purchasing data instead of consumption data as a data basis. This leads to a large dispersion of the modeled consumption data in the sample, because the sample naturally contains households that have purchased significantly more than they consume in the two weeks. At the same time, however, the sample also contains households that purchased significantly less than they consumed during the two weeks because there are still some stocks in the household. This large variance in the data has different implications for our analyses. For example, while the sample is well suited for mean analyses, without further adjustment it is not suitable for drawing conclusions about extreme values or single individuals, as these are biased by above- and below-average purchase volumes. For analyses that go beyond mean comparisons, it is imperative to perform further data cleaning. 
The first step is to adjust for daily energy intake. Assuming a weight stable condition, a plausible daily energy intake corresponds to the daily energy consumption. Daily energy expenditure is the basal metabolic rate multiplied by an activity factor (UNU et al., 2004). Basal metabolic rate is determined primarily by gender, height, build, and age (UNU et al., 2004). For example, the activity factor for office work ranges from 1.40 to 1.69 (UNU et al., 2004). To adjust the daily energy intake of our data set, we recommend setting a lower limit at basal metabolic rate (activity factor = 1.00) and an upper limit at an activity factor of 2.40. An activity level corresponding to a factor greater than 2.40 is difficult to maintain for an average person (UNU et al., 2004).
In a second step, a plausibility check of certain macronutrients has to be performed. The intake of macronutrients differs for people based on age, dietary habits and body weight. The recommended daily protein intake for a healthy adult person under 65 years of age is 0.83 grams per kilo of body weight per day (BLV, 2022a). In principle, a protein proportion of the energy intake of more than 20% is referred to as a high-protein diet. The intake of fat, carbohydrates and salt cannot be plausibilized with a formula. A diet in which fats account for more than 50% of the daily energy intake seems too high in our opinion, since this is significantly above the recommended guideline values (BLV, 2022a). For carbohydrates, we believe a plausible range is between 10% and 70% of daily energy intake. A person following a strict low carb diet may well have a low intake of 10%, while 60% is the upper limit of the emfphlen daily amount for a healthy adult (BLV, 2022a). Salt, in turn, cannot be plausibilized in relation to daily energy intake. Here, we recommend setting an upper limit of 30 grams per day. Salt consumption becomes lethal at a level of 0.5 - 1.0 grams per kilo of body weight (Strazzullo & Leclercq, 2014).

bottom of page