Other Processing
Data about participation
A total of 2086 persons initially participated (participation rate: 15%). Please also consult the chapter “Weighting”
A) Data from the dietary behavior and physical activity questionnaire
Missing questionnaires
Out of the 2086 participants 5 participants never gave back their questionnaire. One of them could finally be reached by phone and answered the questions about socio-economic determinants only. This results in a total of 2081 complete questionnaires.
Data cleaning
Data cleaning was done with Stata Statistical Software (Release 13. College Station, TX: StataCorp LP). Most variables were only slightly modified: they were only re-coded with numerical values and labeled in English to facilitate data analysis. For instance, the variable “gender” was initially coded “homme” and “femme” and was re-coded as such: 1 for male and 2 for female. Others variables had to be adapted prior to analysis. These adaptations and explanations of large missing values are described below.
Physical activity (questions 14-20)
Many participants (n=525) claiming they could not estimate how much time they had spent on average doing any of the three activity types of physical activity. In addition, 11 participants had missing or illogical (and deleted) data.
Occupation and socioeconomic background (questions 41-49)
An important limitation of the questionnaire was the instruction for question 46, stating that only people without remunerated professional activity at all should complete questions 46 and next. To comply with this statement, field dietitians were instructed to enter only questions 42 to 45 (and not question 46 and following) if the participant was a student or a housewife, etc. and had part-time paid-work, even if it was only 1 hour per week. Unfortunately, with the statement in the questionnaire plus the systematic correction of field dietitians, many students or retired people, etc. who worked part-time were recorded in the database only as active worker (and not additionally student, retired, etc.). A consequence of this limitation is that the information on occupation is not optimal.
Before data cleaning, the proportion of students was 4%, housewives / househusbands 6% and retired 16%, respectively. In comparison, in the 2012 Swiss Health Survey, for which data were collected by phone among 21'597 people aged 15 years old and over, there was about 10% of students, 20% of housewives/househusbands and 23% of retired people, taking into consideration the ones working. Based on that observation, we went back to all 2081 paper questionnaires and captured answers to question 46 that were not taken into consideration by field dietitians during data entry. Going back to the questionnaires allowed us to recapture participants who had ticked an answer to this question. We created then a new variable for question 46, which was considered of better quality. The initial variable for question 46 was left as such (not cleaned). Furthermore going through the 2081 questionnaires allowed us to identify a few inconsistencies in the questions 41-49, which were also cleaned. Finally, age limit was set for housewives / househusbands at 64 years for women and 65 years for men. Participants above this age were re-coded only as retired.
Although we did our best to capture as much information as possible for variables about occupation and question 46, all data should be interpreted with caution due to limitations of the questionnaire in questions 41-49. For example, we highly suspect that information about housewives / househusbands were partially lost as their percentage is still relatively low compared to the Swiss Health Survey. In addition, 115 participants (almost all women) declared working less than 20 hours/week and did not answer question 46. We may suppose they may be potential housewives / househusbands. Information about retired people are expected to be of good quality after the paper copy checks: more than 96% of people older than 65 years were classified as retired in question 46. Same for students: 7% of students was a very reasonable proportion compared to the Swiss Health Survey results (i.e., 10%).
Age
Participants' age was calculated from their birth date and estimated questionnaire completion date, which corresponds for most to face-to-face appointment date.
B) Data from anthropometry
Missing data
Out of the 2086 participants, 34 weight measures were missing. Following the study protocol, 27 pregnant or lactating women, 6 handicapped participants (e.g. in a wheel chair) were not measured. Only one participant refused to be weighted. For height, there were 7 missing values because height was again impossible to be measured in these 6 handicapped participants plus the same participant who refused weighing.
For waist and hip measures, there were again 34 missing values. The reasons were identical: 27 pregnant or lactating women, 6 handicapped participants and 1 refusal.
Data cleaning
Data were controlled for consistency (e.g., comparison of self-reported vs. measured weight and height) or incorrect rounding. Data were cleaned when necessary, going back to written data on the paper sheet.
Data correction for clothing
Because weight was measured while participants were wearing light clothes (more than underwear), 1.2kg were deducted from measured weight for men, respectively 0.8kg for women, independent of season. This correction for light clothing was performed based on recommendations from the literature and in relation to what was done in other Swiss surveys.
Because waist circumference was taken directly on the skin, we did not apply any correction factor. By contrast, hip circumference was taken while participants were wearing pants or skirts. However, we decided not to correct hip measures for clothing as field dietitians were required to put more tension on the string when the pants/skirts were thick.
C) Food consumption (data from GloboDiet®, 24h dietary recalls)
Missing data
Of the 2086 initial participants 29 participants did not complete the second 24-hour dietary recall by phone. Additionally, one face-to-face (first) 24-hour dietary recall was deleted due to incompleteness. All other 24-hour dietary recalls (4142 in total, 2085 face-to-face and 2057 phone) were considered as valid.
Data cleaning
The 15,637 notes written in GloboDiet® by field dietitians were handled centrally by a senior registered dietitian following IARC guidelines. The latter also checked all 24HDR with extreme energy intakes (n=85). Furthermore, food consumption data were evaluated using all criteria recommended by IARC. Detailed quality control procedures were implemented and published.
Data linkage for energy and nutrient values
Food consumption data e.g. foods, recipes and ingredients (from GloboDiet®) were linked semi-automatically with the most appropriate item from the Swiss Food Composition Database (SFCDB, http://naehrwertdaten.ch) using a developed matching tool on the food information platform FoodCASE (Premotec GmbH). The version 37 of the SFCDB, which is not available on the SFCDB website, was used for this data linkage. The matching tool was programed to provide for each different consumed item a list of possible matches from the SFCDB ranked according to similarity. Whenever possible the exact generic item was selected out of the SFCDB. If neither was available the consumed item was linked with an item similar with regard to the most relevant nutrient. Exceptions are the mineral waters and some breakfast cereals which are registered as branded foods. For menuCH more than 32’000 different items from GloboDiet® were linked in this way.
The actual data subset named res_consumption_data_V05_2022 is matched with generic products with regard to the most relevant nutrient. Therefore, this data subset is suitable for micronutrients and macronutrients analyses. The few branded products, namely mineral water and breakfast cereals have precisely known micronutrients composition like the generic products.
A first data subset (version 2016-08-04) was established in 2016 for macronutrients analysis only. The consumed items were matched as much as possible with branded products from the SFCDB version 20. This data subset is no longer available on the menuCH data repository.
Unfortunately, no insights into micronutrients were possible with this old data subset. This is the reason why the SFCDB was updated. As a result, a new matching was performed with mainly generic products from the SFCDB and the actual res_consumption_data_V05_2022 was created.
For further information regarding the development of the matching tool, please consult the master thesis of Hochuli, Alexandra, 2014, from chapter 4 onward:
https://doi.org/10.3929/ethz-a-010129946.
When using the consumption data of this survey please be aware that:
1. Both recipes and their ingredients have been matched separately to the most appropriate food item and additionally to the most appropriate recipe from the SFCDB. Therefore summing up energy/nutrients of the ingredients may not result in exactly the same figures as are given by summing up the respective recipes.
To avoid duplication of nutrient intake you must either use recipes or ingredients but never both!
2. The Swiss Food Composition Database contains information on the composition of foods that are available in Switzerland. For all the foods contained in the database, complete information is presented on the macronutrients (carbohydrates, proteins, fats) as well as for dietary fibres, water, salt, alcohol and energy content. Fat values include amount of saturated, monounsaturated and polyunsaturated fats, and cholesterol. In addition, the micronutrients, vitamin A, B1, B2, B6, B12, C, D and E, niacin, folate, pantothenic acid as well as potassium, sodium, chloride, calcium, magnesium, phosphorus, iron, iodide and zinc contents are listed for the majority of generic foods (selenium could not be included in the dataset because there are still too many missing values to allow useful calculations). For branded products, however, information about micronutrients is not always available, only if provided or published by the manufacturer (e.g. on packaging or websites).
