This study represents a cross-sectional analysis of dietary surveillance data from the UK National Diet and Nutrition Survey (NDNS) 2008–16 (May 2018 release) . It is reported according to the STROBE-nut recommendations .
NDNS is an annual cross-sectional survey which collects information on food consumption and nutritional and health status of free-living individuals in the UK. Sampling, recruitment and data collection are carried out in a consistent manner, allowing data from different survey years to be combined for cross-sectional analysis.
A detailed account of the NDNS recruitment and sampling protocol has been published elsewhere [39,40,41]. In short, private addresses were randomly selected from postcode sectors across the UK. Within each household, a maximum of one adult and one child were randomly selected for inclusion in the study. These individuals were asked to complete a four-day food and drink diary, and to participate in an interview concerning more general dietary habits, socio-demographic status, lifestyle and physical activity, and receive a nurse visit which included measurement of height and weight.
NDNS was approved by the Oxfordshire Research Ethics Committee and written informed consent was obtained from all participants.
Individuals aged ≥19 years at the time of participation, who completed three or four days of the food diary, were included in the analyses. In order to compare those who achieved a relatively high dietary quality with and without a relatively high proportion of energy from home-cooked foods, only a sub-sample of the NDNS sample (the analytic sample) was included in this analysis: those in the top tertiles of both proportion of energy from home-prepared food and dietary quality (hereafter the high home preparation group), and those in the top tertile of dietary quality and the bottom tertile of energy from home-prepared food (the low home preparation group). This resulted in an analytic sample with universally high dietary quality, allowing inter-group differences to be associated with consumption of home-prepared food as opposed to dietary quality.
Participants completed unweighed food diaries, including all food and beverages consumed both inside and outside the home for three or four consecutive days. This process is described in detail elsewhere . Participants also recorded where the food was eaten, for example at home, in a restaurant or café, or at work. This variable included a specific category for food eaten at work but brought from home.
Characterisation of food-related variables
Food-related variables – proportion of home-prepared food and dietary quality, as well as other aspects of diet such as daily intake of food groups, energy and macro- and micronutrients – were derived from food diaries. The first two variables were derived in order to classify participants as being either in the high or low home preparation group. Further food-related variables were derived in order to characterise dietary intake in greater detail.
Proportion of energy from home-prepared food
Food items listed in food diaries were classified by the authors as either requiring or not requiring home preparation. All foods were classified as home-prepared except those listed in Table 1. Foods which should not be classified as being home-prepared were decided by the authors a priori.
Definitions of ‘cooking’ have been discussed extensively and remain contested [21, 43, 44], with many definitions not deeming the application of heat to be a necessary part of this process [44, 45]. As a result ‘home food preparation’ and ‘home-prepared food’ seem more accurate and are the concepts deployed here. Different, but related, conceptualisations exist, such as food ‘prepared from scratch’ , or food that is not ‘from outside the home’ . The conceptualisation of home-prepared food used here reflects several conceptions of ‘cooking’, or home food preparation, drawn from qualitative studies [47, 48] as well as behaviours which are habitually enquired about in studies of ‘cooking’, such as blending, mixing, boiling, chopping, roasting and pan frying . From this conceptualisation of home food preparation, a set of behaviours, we defined foods which we would deem to be home-prepared as being the products of these behaviours.
Food classification was carried out using food diary variables as illustrated in Fig. 1, with foods which were not classified as home-prepared being successively removed until only food included in home-prepared dishes remained. The proportion of energy from home-prepared food was then calculated for each participant by summing the energetic content of foods classified as home-prepared and dividing them by the participant’s total energy intake. Participants were then separated into tertiles based on this proportion. Individuals in the highest tertile for proportion of energy from home-prepared foods were categorised as belonging to the high home preparation group, while those in the lowest tertile were categorised as belonging to the low home preparation group.
Dietary quality was determined by quantifying accordance to the Dietary Approaches to Stopping Hypertension (DASH) dietary pattern using a method adapted for use with NDNS  from an existing index . The DASH diet has been shown to lower blood pressure  and reduce low-density lipoprotein cholesterol levels,  as well as being associated with a lower risk of stroke and coronary heart disease . This score is based on food and nutrients emphasised or minimised in the DASH diet, and has eight components: high intake of fruits, vegetables, nuts and legumes, low-fat dairy products, and whole grains; and low intake of sodium, red and processed meats, and non-extrinsic milk sugars; all adjusted for total energy intake. The score is adjusted for overall energy intake. Components are evenly weighted, and three components (sodium, sugar, and red and processed meats) are reverse-scored, so that higher consumption would lower an individual’s DASH score.
Participants were separated into tertiles by DASH score. Participants in the highest tertile were categorised as high-DASH.
Intake of energy, macronutrients and micronutrients
Mean daily intake of energy was estimated by the NDNS team using food diaries, as were daily intakes of several macro- and micronutrients: fat, saturated fat, protein, carbohydrate, sugar, and sodium, a process described in detail elsewhere . These nutrients make up mandatory nutrition labelling in the UK . Intake was categorised as meeting or not meeting relevant UK dietary guidelines, [54, 55] except in the case of carbohydrates. Current UK recommendations suggest a population mean of approximately 50% of total energy from carbohydrate, but note that total carbohydrate intake does not appear to be associated with health outcomes, as it is composed of different nutrients such as fibre, sugar and starches, which may have a variety of impacts . Therefore, carbohydrate intake was described in all groups, but adherence to a particular recommendation was not defined.
Daily intakes of other nutrients were also estimated by the NDNS team using food diaries . Where UK guidelines existed,  adherence to these guidelines was also determined. Nutrients included were: fibre, thiamine, riboflavin, niacin, vitamin B6, vitamin B12, folate, vitamin C, vitamin A, calcium, phosphorus, magnesium, zinc, selenium, iodine, iron, chloride, vitamin E, copper, manganese, biotin, and pantothenic acid. Nutrients derived from supplements were not included in the data presented here.
Intake of food groups
Daily intakes of the main food groups determined by NDNS were calculated using food diaries. Where possible similar groups of food were collapsed (e.g. 1% fat milk, skimmed milk and semi-skimmed milk).
Prevalence of overweight and obesity
Interviewers collected measurements of height and weight from NDNS participants using standard protocol. These measures were used to calculate BMI, which was categorised as overweight/obese (BMI ≥25 kg/m2), or not.
Socio-demographic variables considered include age, sex and ethnicity (categorised as white or not due to the high proportion of white participants in NDNS) were determined using self-reported survey responses, as were the presence of a child under 16 years of age in participant households. Socioeconomic position was also assessed using self-reported survey responses, and was characterised using a range of markers: occupation (professional/other), education (degree level or above/other), and annual income equivalised for household composition (above or below £35,000).
The demographic and socioeconomic characteristics of individuals in the high-home preparation and low home-preparation groups were described. The statistical significance of differences between groups was tested using either a linear or logistic regression as appropriate, mutually adjusted for all other socio-demographic variables.
Overall dietary characteristics were examined in two ways. First, the high home preparation and low home preparation groups were compared in terms of DASH score, proportion of energy from home-prepared food, energy intake, and adherence to dietary guidelines for macro- and micronutrients. Prevalence of overweight or obesity was also compared across groups. Second, the groups were compared in terms of their intake of each of the food groups or nutrients that make up each of the eight components of the DASH diet and index: low-fat dairy, whole grain, fruit, vegetables, nuts and legumes, sodium, sugars, and red and processed meats. In both cases, the statistical significance of differences between groups was tested using either a linear or logistic regression as appropriate, adjusted for all socio-demographic variables.
In addition, food-level differences between home preparation groups were then assessed through an examination of the food group codes provided by NDNS. Due to the high proportion of individuals who did not consume many of the food groups over the course of the recorded days, this was done in two steps. First, the proportion of individuals consuming any amount from each food groups was calculated for both the high home preparation and the low home preparation groups. Differences in these proportions were tested using logistic regression. Second, the median quantity of each food group consumed by consumers of those food groups was determined. Differences between home preparation groups in these quantities were tested using linear regression. All regressions in food-level analyses were adjusted for all socio-demographic variables.
All analyses were conducted using Stata (version 14; Stata Corp.). Alpha-level of 0.05 was used throughout.