Greater frequency of cooking dinner at home and more time spent cooking are inversely associated with ultra-processed food consumption among US adults

Abstract

Background:

Cooking at home has been promoted as a strategy to improve diet quality, however, the association between cooking behavior and ultra-processed food intake is unknown.

Objective:

The objective of this study was to examine associations between frequency of cooking dinner at home and time spent cooking dinner with ultra-processed food intake.

Design:

Cross sectional, nationally representative data from the 2007–2010 National Health and Nutrition Examination Survey were analyzed.

Participants/setting:

Participants were 9,491 adults (aged ≥20 years) in the United States.

Main outcome measures:

The main outcome measure was the proportion of energy intake (averaged from two 24-hour dietary recalls) from the four Nova food processing groups: 1) unprocessed or minimally processed foods; 2) processed culinary ingredients; 3) processed foods, and 4) ultra-processed foods.

Statistical analyses performed:

Separate linear regression models examined associations between cooking frequency and time spent cooking dinner and proportion of energy intake from the four Nova food processing groups, adjusting for socio-demographic characteristics and total energy intake.

Results:

Ultra-processed foods comprised >50% of energy consumed independent of cooking frequency or time spent cooking. Higher household frequency of cooking dinner and greater time spent cooking dinner were both associated with lower intake of ultra-processed foods (p-trends<0.001) and higher intake of unprocessed or minimally processed foods (p-trends<0.001) in a dose response manner. Compared to cooking 0–2 times/week, adults who cooked dinner 7 times/week consumed on average 6.30% (95% CI: −7.96%, −4.64%, p<0.001) less energy from ultra-processed foods. Adults who spent >90 minutes cooking dinner consumed 4.28% less energy from ultra-processed foods (95% CI: −6.08%, −2.49%, p<0.001) compared to those who spent 0–45 minutes cooking dinner.

Conclusions:

Cooking at home is associated with lower consumption of ultra-processed foods and higher consumption of unprocessed or minimally processed foods. However, ultra-processed food intake is high among US adults regardless of cooking frequency.

INTRODUCTION

In the last century, numerous societal and food system shifts have changed the way people both cook and eat in the United States (US).^1–3 Though food processing has long been part of many culinary traditions, ultra-processed foods, have only recently been developed, and have rapidly become ubiquitous on US supermarket shelves and in the food system broadly.⁴ Among American adults, ultra-processed foods⁵ contribute more than half of total energy consumed on a typical day across socio-economic and demographic subgroups while unprocessed/minimally processed foods contribute about 30%.⁶ Ultra-processed foods are convenient, quick and easy to prepare, widely consumed,⁶ and, thus, may also partially explain shifts to how Americans cook.¹

The degree to which ultra-processed foods are included in home prepared meals has not been systematically examined. Americans report cooking frequently;⁷ on average, Americans live in households in which dinner is cooked 5 days/week. Despite the growing consumption of foods away from home (e.g. fast food and other restaurant foods),^8,9 foods prepared and consumed at home comprise nearly 2/3 of daily energy intake in the US.¹⁰ A growing body of evidence has shown that cooking at home more frequently is associated with better diet quality among US adults based on a variety of indicators including fruit and vegetable consumption and Healthy Eating Index-2015 (HEI-2015) scores.^7,11–16 Measures of cooking frequency do not account for time spent cooking, the ingredients used for cooking, and the definition of what it means ‘to cook’. Additionally, whether processed or ‘convenience’ foods are included in definitions of home cooking varies across individuals.^17,18

Time spent cooking has also declined over the last 50 years⁹, likely due to advances in technology (e.g. dishwashers, microwaves), societal changes (e.g. women working outside the home), and the development and promotion of ultra-processed foods used as ingredients or components of home prepared meals.^19–22 Despite the growing body of evidence regarding the adverse effects of ultra-processed foods on diet and health,^23,24 and the concurrent growth in research about the role of cooking and cooking skills for healthy eating,^{7,12,16,25–27} the relationship between frequency of cooking at home and levels of processing of consumed foods, including ultra-processed food consumption, is currently unknown. Additionally, no studies have examined associations between usual time spent cooking (which may be a more sensitive measure than cooking frequency related to use and intake of ultra-processed foods) and the level of processing of foods consumed.

In this study nationally representative data from US adults was utilized to examine how household frequency of cooking dinner and time spent cooking were associated with the proportion of energy intake from the four Nova food processing groups. It was hypothesized that greater household cooking frequency and greater time spent cooking would be associated with lower ultra-processed food consumption and greater consumption of unprocessed/minimally processed foods.

METHODS

Data and Design

Data from the National Health and Examination Survey (NHANES) 2007–2008 and 2009–2010 waves were used.²⁸ NHANES is a cross-sectional, nationally representative, population-based survey designed to collect demographic, dietary intake, and health information about the non-institutionalized US population.²⁸ NHANES was approved by the National Center for Health Statistics (NCHS) Research Ethics Review Board.²⁹

Participants are selected based on a multistage, clustered, probability-based sampling strategy. NHANES participants provide written consent at the time of data collection.³⁰ The present study used data from 2007–2010 because those were the only survey waves in which NHANES included a question about household frequency of cooking dinner.^31,32 As part of NHANES data collection, participants answer questions about their household behaviors and characteristics and complete two 24-hour dietary recalls.²⁸ The first dietary recall was collected in-person in the mobile examination center (MEC) and the second was collected by telephone 3–10 days after the first but never on the same day of the week as the MEC interview.^30,33 Dietary recalls were conducted by trained interviewers using the validated USDA Automated Multiple-Pass Method.³⁴ This study was exempt from Institutional Review Board review because the data was solely de-identified secondary data.

Study sample

The study sample included adults aged ≥20 years with two days of complete and reliable 24-hour recalls (as determined by NHANES staff). Individuals with only 1 day of dietary recall data were excluded. Individuals with missing data for cooking frequency (n=108) or cooking frequency >7 days/week (n=11) were excluded as were those with missing data for other study covariates (n=29). Participants with implausibly low (<500 kcal; n=32) or high (>5,000 kcal; n=48) values for energy intake (mean value across 2 24-hour recalls) were also excluded. The flow chart describing participants included in the analytic sample is available in Supplementary Figure 1. The final sample included 9,491 US adults with two days of complete and reliable 24-hour recalls and complete data for cooking frequency, and 8,947 adults with complete data for time spent cooking.

Measures

Household cooking frequency:

Household frequency of cooking dinner was measured by the following survey question: “During the past seven days, how many times did you or someone else in your family cook food for dinner or supper at home?” Following previous literature,^11,12,35 responses were recoded into the following categories: 0–2 times/week; 3–4 times/week; 5–6 times/week; 7 times/week.

Time spent cooking dinner:

Time spent cooking dinner was measured by the following survey question: “How much time do you or someone else in your family usually spend on cooking dinner or supper and cleaning up after the cooking? Please do not include times spent eating.” Based on logical cut points in the data, open responses were recoded into the following categories: 0–45 minutes/meal; 46–60 minutes/meal; 61–90 minutes/meal; >90 minutes/meal.

Nova food group classification:

Foods and beverages consumed by NHANES participants were categorized according to the Nova classification system which classifies foods and beverages according to the level and purpose of industrial processing.^5,36 Nova classifies foods and beverages into four groups: 1) unprocessed or minimally processed foods; 2) processed culinary ingredients; 3) processed foods, and 4) ultra-processed foods.⁵ Detailed explanations of the Nova classification system and rational are described elsewhere.^5,36 Briefly, group 1 foods include unprocessed foods such as fruits, vegetables, grains, fish, meat, and minimally processed items that are altered (such as by drying, freezing or pasteurizing) without adding salt, sugar, oils or fats. Group 2 foods are products such as oils, fats, sugar, salt and other products derived directly from group 1 foods or from nature and are used to make homemade or artisanal dishes. Group 3 foods are industrial products made by adding salt, sugar or other substances to group 1 foods (such as by canning or bottling or fermentation) such as bacon, smoked fish, simple breads, cheese, and canned vegetables. Processing for Group 3 foods is used to prolong the durability of foods and modify their palatability. Ultra-processed foods (Group 4) contain little or no whole foods and are highly palatable, and often include numerous ingredients (including emulsifiers, sweeteners, artificial flavors) and additives with cosmetic function of no or rare culinary use. Ultra-processed foods are often highly convenient, are ready-to-eat or only require re-heating. Examples of ultra-processed foods include hot dogs, French fries, breakfast cereals, chips, sugar-sweetened beverages, packaged soups, and ready-to heat frozen meals.

All recorded foods and beverages (Food Codes) were linked to underlying Standard Reference (SR) Codes obtained from the United States Department of Agriculture (USDA) Food and Nutrient Database for Dietary Studies (FNDDS) 4.0 and 5.0.^37,38 Each Food Code was then classified into one of the four Nova groups, taking into account “Main Food descriptions” (primary complete description identified by a unique 8-digit food code that may include the form, preparation method, and source of item), “Additional Food Description” (while more than one additional description may be associated with a food code these are not available for all food codes; many times these are brand names), which qualitatively describes food codes, and underlying “SR code description” which qualitatively describes each of the underlying SR codes. For each food code, a decision was made regarding whether food codes or underlying SR codes would be used to estimate Nova group energy contributions. Food codes that were likely to be homemade or artisanal and linked to a list of scratch ingredient SR codes, such as ‘beef stroganoff’ were classified at the SR code level. Conversely, foods likely purchased as ready-to-eat/heat/drink items, such as ‘Lasagna with meat, canned’ were classified at the food code level. Thereafter, the Food Code classification was modified, if necessary, taking into account participant-specific variables ”Source of food” (reflecting where the food or most of its ingredients were acquired) and “Combination Food Type” (capturing whether the food was consumed in combination with other foods).³⁶ For instance, some food codes (mainly mixed dishes, including sauces and cakes, cookies and pies) initially classified at the SR code level were reclassified as ultra-processed foods at the food code level for participants who reported the food code as ‘frozen meals’ or ‘Lunchables’ (combination food types) or from ‘Restaurant: fast food/pizza’ or ‘vending machine’ as the food source. The classification of most food items, however, did not change (e.g. raw apple from a fast-food restaurant or vending machine remained classified as an unprocessed/minimally processed food). Further details have been published elsewhere.³⁹

The daily proportion of energy intake from each Nova group was determined by summing the energy intake (in kilocalories (kcals)) for all foods in each group for each day of recall and dividing by the participant’s total energy intake of that day. Then, each proportion was averaged across the two days of dietary recall.³⁶ The primary outcome for this study is the proportion of energy intake from ultra-processed foods based on the average intake across both 24-hour dietary recalls. The proportion of energy intake from the remaining three Nova groups are also presented to examine patterns in consumption based on processing level. Additionally, absolute energy intake (kcals) from the Nova processing groups and patterns in absolute and relative energy intake from the mutually exclusive food subgroups that comprise the four Nova processing categories were examined.

Socio-demographic covariates:

Covariates of interest, self-reported during the NHANES interview, included age in years, sex (male, female), race and ethnicity (Hispanic, non-Hispanic Black, non-Hispanic White, other), education (<high school, high school or General Education Development (GED), at least some college), employment status (not working, employed), marital status (not married, married or living with a partner), income to poverty ratio (<1.0, 1.0–1.99, 2.0–2.99, 3.0–4.99, ≥5.0, missing (n=758)),⁴⁰ household size (1–3 people, ≥4 people), participation in the Supplemental Nutrition Assistance Program (SNAP),⁴¹ and household food security status (high, marginal, low, very low; assessed via the USDA 18-item household food security screener module).⁴²

Analyses

To produce nationally representative results, all analyses used dietary 2-day sample weights, strata, and psu, survey weights provided by NHANES staff to account for unequal probability of being selected to the sample, the complex sampling strategy used by NHANES, non-response, and the day of the week and whether or not dietary recalls took place on a weekend or weekday. Survey weights were applied using the SVYSET command and svy: prefix in Stata. Taylor-linearized variance estimation was used to calculate standard errors. First, cross tabulations and chi-squared tests were used to examine the characteristics of the study sample by household frequency of cooking dinner. Then, simple linear regressions and multivariable regressions adjusted for all study covariates described above as well as total energy intake (mean total kcal from 2 days of dietary recall) were conducted to examine associations between household cooking frequency and proportion of energy intake from ultra-processed foods and each of the three remaining Nova processing Groups. These models were repeated to examine associations between time spent cooking and the proportion of energy intake from the four Nova groups. Tests for linear trend across the categories of cooking dinner frequency and time spent cooking were conducted by re-estimating the multivariable models with the ordinal cooking variable included as a continuous variable.⁴³ The “margins” command after the regression models was used to calculate predicted mean proportion of energy intake from the four Nova processing groups at different levels of cooking frequency while holding other variables included in the models at their means. ⁴⁴ Generalized linear models with a gamma family and log link⁴⁵ (due to the skewed distribution of the dependent variables) and adjusted for study covariates listed above were used to examine absolute energy intake (kcals) from the four Nova processing groups by cooking frequency and time spent cooking. Finally, generalized linear models with a gamma family and log link (due to the skewed nature of the food subgroup dependent variables) adjusted for all study covariates were used to examine differences in consumption (proportion of total energy intake) of the food subgroups that comprise the four Nova food processing categories by frequency of cooking dinner and by time spent cooking. In generalized linear models the link function (g) is the function of mean, (i.e., g(mean) = beta0+ beta1), therefore, the zeros are not dropped. All analyses were conducted using Stata, version 17.0,⁴⁶ all tests were two-sided, and significance was considered at p<0.05.

RESULTS

Characteristics of the study sample are presented in Table 1. Overall, 13% of American adults lived in households in which dinner was cooked 0–2 times/week, 21% lived in households in which dinner was cooked 3–4 times/week, 30% lived in households in which dinner was cooked 5–6 times/week and 36% lived in households in which dinner was cooked 7 times/week. Individuals in the cooking 7 times/week group were more likely to be older age, Hispanic, lower education, not working, married or living with a partner, lower household income. They were also more likely to have ≥4 people in the household, be receiving SNAP benefits, and have low or very low food security status (p’s≤0.001). Adults who reported greater cooking frequency were also more likely to spend more time cooking dinner. Among those in households that cooked dinner 0–2 times/week, 34% typically spent 0–45 minutes/meal cooking and cleaning up and 16% spent >90 minutes cooking and cleaning up. In contrast, among those who cooked dinner 7 times/week, 16% spent 0–45 minutes cooking and cleaning up and 31% spent >90 minutes cooking and cleaning up. Most respondents spent 45–60 minutes cooking and cleaning up, and across cooking frequency categories, the proportion of respondents who spent 45–60 minutes cooking and cleaning up was similar (38%−45%). Household cooking frequency did not differ based on the sex of the respondent. There were significant trends in the unadjusted differences in the percent of energy intake from Nova processing groups 1, 2, and 4 by household cooking frequency (all p-trends<0.001) with greater proportion of energy from group 1 and 2 foods and lower proportion of energy from group 4 when dinner was cooked at home more frequently. The proportion of energy from group 3 foods did not differ by household frequency of cooking dinner (p-trend=0.177).

Table 1. Characteristics of the study sample by frequency of household cooking dinner/supper frequency, NHANES 2007–2010 (N=9,491).

			Household Cooking Dinner/Supper Frequencya (times/week)
	Overall (N=9,491)		0–2 (N=1,227)		3–4 (N=1,730)		5–6 (N=2,480)		7 (N=4,054)		p-valueb
	%	se	%	se	%	se	%	se	%	se
Totalc	100		13	0.6	21	1.0	30	1.0	36	1.3
	mean	se	mean	se	mean	se	mean	se	mean	se
Age in years	46.9	0.4	46.9	0.8	44.2	0.6	46.7	0.5	48.7	0.6	<0.001
	%	se	%	se	%	se	%	se	%	se
Sex
Male	47	0.5	49	1.7	47	1.2	45	1.0	47	0.6	0.142
Female	53	0.5	51	1.7	53	1.2	55	1.0	53	0.6
Race and Ethnicity
Hispanic	13	1.7	13	2.6	7.4	1.1	8.1	1.2	21	2.8	<0.001
Non-Hispanic Black	11	1.1	17	2.1	15	1.6	7.9	1.0	9.8	1.1
Non-Hispanic White	70	2.4	65	2.8	73	2.1	80	2.0	61	3.7
Otherd	5.5	0.7	4.8	1.1	4.4	0.9	4	0.8	7.7	1.2
Education
<High school	19	0.9	16	1.1	13	1.4	12	1.3	29	1.5	<0.001
High school or GEDe	23	0.8	24	1.8	24	1.5	22	1.3	24	1.4
> High school	58	1.4	59	2.3	63	2.0	66	1.9	47	1.9
Employment status
Not working	39	1.0	36	2.2	30	1.5	35	1.7	48	1.3	<0.001
Employed	61	1.1	64	2.2	70	1.5	65	1.7	52	1.3
Marital status
Not married	37	1.2	55	3.2	41	1.6	29	1.5	35	1.7	<0.001
Married or living with a partner	63	1.2	45	3.2	59	1.6	71	1.5	65	1.7
Income to poverty ratio
<1.0	14	0.8	15	1.6	9.5	1.2	8.8	0.9	19	1.3	<0.001
1.0–1.99	19	0.8	19	1.4	15	1.4	14	1.0	25	1.1
2.0–2.99	14	0.8	12	1.6	12	1.5	15	1.3	15	1.4
3.0–4.99	23	0.9	21	1.8	30	1.5	26	1.5	17	1.7
5.0+	24	1.4	27	2.8	29	2.4	31	2.0	15	1.7
missing	6.4	0.6	5.8	0.9	4.7	0.8	5.2	0.9	8.7	0.9
Household size
1–3 people	66	1.4	80	2.6	69	2.4	64	2.1	60	2.1	<0.001
4+ people	34	1.4	20	2.6	31	2.4	36	2.1	40	2.1
Received SNAPf in past 12 months
No	88	0.9	89	1.3	91	1.2	92	0.9	82	1.5	<0.001
Yes	12	0.9	11	1.3	9.4	1.2	8.4	0.9	18	1.5
Household food security status
High	80	1.0	82	1.9	84	1.7	86	1.2	73	2.0	<0.001
Marginal	7	0.5	5.2	1.0	6.1	1.2	6.3	0.8	8.7	0.7
Low	8.4	0.6	7.5	1.3	6.6	1.1	5.3	0.6	12	1.3
Very low	4.4	0.4	5.4	1.0	3.5	0.6	2.8	0.5	5.8	0.7
Time spent cooking dinnerg
0–45 minutes	21	0.8	34	2.9	26	1.9	19	1.1	16	1.2	<0.001
46–60 minutes	41	1.2	41	3.7	40	2.3	45	2.1	38	1.7
61–90 minutes	14	1.0	9	1.9	14	2.0	16	1.5	14	1.2
90+ minutes	24	1.0	16	2.0	20	2.0	19	1.4	31	1.8
	mean	se	mean	se	mean	se	mean	se	mean	se	p-trendh
Total energy intake Novai	2,064	16.1	2,099	36.6	2,103	36.9	2,110	18.8	1,992	23.5	0.003
% kcal/day from Group 1	31.89	0.47	28.66	0.67	30.18	0.56	30.73	0.60	34.95	0.72	<0.001
% kcal/day from Group 2	3.81	0.08	3.46	0.15	3.46	0.15	3.78	0.14	4.15	0.13	<0.001
% kcal/day from Group 3	9.45	0.21	9.40	0.42	9.55	0.30	10.11	0.36	8.85	0.32	0.177
% kcal/day from Group 4	54.86	0.57	58.48	0.86	56.82	0.68	55.38	0.75	52.05	0.86	<0.001

^aHousehold cooking frequency is based on frequency of cooking dinner in the past 7 days. The full question text is “During the past 7 days, how many times did you or someone else in your family cook food for dinner or supper at home?”

^bP-value from chi-squared test of association between two categorical variables for categorical measures and univariate linear regression model for the continuous measure (age).

^cTotal row presents weighted row percentages for the cooking frequency groups. The other rows present weighted column percentages.

^d“Other” refers to other non-Hispanic race including non-Hispanic multiracial. Based on responses by participants to questions about race and Hispanic origin.

^eGED= General Equivalency Diploma

^fSNAP= Supplemental Nutrition Assistance Program

^gTime spent cooking is based on time usually spent cooking and cleaning up for dinner. The full question text is “How much time do you or someone else in your family usually spend on cooking dinner or supper and cleaning up after the cooking? Please do not include time spent eating.”

^hP-trend from simple linear regression with cooking frequency as a continuous variable.

ⁱGroup 1 includes unprocessed or minimally processed foods; Group 2 includes processed culinary ingredients; Group 3 includes processed foods; and Group 4 includes ultra-processed foods.

Table 2 presents associations between household frequency of cooking dinner and the proportion of energy intake from the four Nova processing categories from multivariable models. Compared to cooking 0–2 times/week, having someone in the household cook dinner 3–4 times/week was associated with consuming 2.49% less energy from ultra-processed foods (95% CI: −4.54%, −0.43%, p=0.019) whereas cooking 5–6 times/week was associated with consuming 3.68% (95% CI: −5.49%, −1.86; p<0.001) less energy from ultra-processed foods. Someone in the household cooking dinner 7 times/week was associated with consuming 6.30% (95% CI: −7.96%, −4.64%, p<0.001) less energy from ultra-processed foods, compared to cooking 0–2 times/week and a similar increase (5.42%) in energy from group 1 foods (95% CI: 3.94%, 6.89%, p<0.001). There was a dose response relationship between greater household cooking frequency and higher consumption of Group 1 (unprocessed or minimally processed) foods (p-trend<0.001) and Group 2 (processed culinary ingredients) foods (p-trend=0.001) and lower consumption of Group 4 (ultra-processed) foods (p-trend<0.001).

Table 2. Associations between household cooking dinner/supper frequency with percent of energy intake^a from the four Nova processing categories among adults (NHANES, 2007–2010).

	Group 1: Unprocessed or minimally processed foods			Group 2: Processed culinary ingredients			Group 3: Processed foods			Group 4: Ultra-processed foods
	Beta	95% CI	p-value	Beta	95% CI	p-value	Beta	95% CI	p-value	Beta	95% CI	p-value
Household cooking frequencyb (N=9,491)
0–2 times/week	Ref			Ref			Ref			Ref
3–4 times/week	2.38	0.74, 4.01	0.006	0.07	−0.35, 0.48	0.749	0.04	−0.84, 0.93	0.918	−2.49	−4.54, −0.43	0.019
5–6 times/week	2.82	1.45, 4.19	<0.001	0.29	−0.07, 0.66	0.114	0.56	−0.34, 1.46	0.214	−3.68	−5.49, −1.86	<0.001
7 times/week	5.42	3.94, 6.89	<0.001	0.55	0.18, 0.93	0.005	0.33	−0.55, 1.20	0.451	−6.30	−7.96, −4.64	<0.001
p-trend			<0.001			0.001			0.342			<0.001
Time spent cooking dinnerc (N=8,947)
0–45 minutes	Ref			Ref			Ref			Ref
46–60 minutes	1.11	−0.01, 2.23	0.051	−0.01	−0.34, 0.32	0.962	0.78	−0.10, 1.66	0.081	−1.88	−3.29, −0.48	0.010
61–90 minutes	1.83	−0.00, 3.66	0.050	0.04	−0.40, 0.48	0.861	−0.12	−1.15, 0.90	0.808	−1.74	−3.97, 0.49	0.121
>90 minutes	4.10	2.53, 5.68	<0.001	0.59	0.07, 1.11	0.027	−0.41	−1.27, 0.44	0.330	−4.28	−6.08, −2.49	<0.001
p-trend			<0.001			0.015			0.017			<0.001

^aEnergy intake based on mean intake from day 1 and day 2 dietary recalls. Results reported are from linear regression models adjusted for age, sex, race and ethnicity, marital status, household size, employment status, education, income to poverty ratio, food security status, SNAP receipt, and total energy intake.

^bHousehold cooking frequency is based on frequency of cooking dinner in the past 7 days. The full question text is “During the past 7 days, how many times did you or someone else in your family cook food for dinner or supper at home?”

^cTime spent cooking is based on time usually spent cooking and cleaning up for dinner. The full question text is “How much time do you or someone else in your family usually spend on cooking dinner or supper and cleaning up after the cooking? Please do not include time spent eating.”

Associations between time spent cooking dinner and the proportion of energy intake from the four Nova processing categories from multivariable models are also presented in Table 2. In multivariable models spending 90+ minutes cooking dinner was associated with 4.28% less energy from ultra-processed foods (95% CI: −6.08%, −2.49%, p<0.001) compared to spending 0–45 minutes cooking dinner and a similar increase (4.10%) in energy from group 1 foods (95% CI: 2.53%, 5.68%, p<0.001). Trends for all four Nova processing groups were significant; with a dose response relationship between greater time spent cooking and greater intake of Group 1 (unprocessed/minimally processed) foods (p-trends<0.001) and lower intake of Group 4 (ultra-processed foods) (p-trends<0.001).

Figure 2 shows the predicted mean energy intake from all four Nova groups by household frequency of cooking dinner and time spent cooking and cleaning up from dinner based on post-estimation margins from the multivariable models in Table 2. Daily energy intake from ultra-processed foods was 58.78% among adults in households cooking dinner 0–2 times/week; 56.29% among those cooking 3–4 times/week; 55.10% among those cooking 5–6 times/week and 52.48% among those cooking 7 times/week. Among adults living in households spending the least amount of time cooking dinner (0–45 minutes/meal) 56.67% of total energy intake came from ultra-processed foods whereas among those spending the most time cooking (> 90 minutes/meal), 52.38% of energy intake came from ultra-processed foods. These results demonstrate that reductions in the proportion of energy intake from Group 4 ultra-processed foods based on greater cooking frequency and more time spent cooking correspond with higher energy intake from unprocessed/minimally processed Group 1 foods. Across cooking frequency and time spent cooking groups ultra-processed foods form the majority of energy intake followed by Group 1 foods and Group 3 foods. Group 2 foods comprise the smallest contribution to energy.

Figure 2. Proportion of energy intake across the four Nova classification categories by cooking frequency and time spent cooking, NHANES 2007–2010.^a.

^a Predicted mean energy intake (as a proportion of total energy intake) from each Nova processing group after linear regression models adjusted for age, sex, race and ethnicity, education, employment status, marital status, income to poverty ratio, household size, SNAP receipt, food security status, and total energy intake. Household cooking frequency is based on frequency of cooking dinner in the past 7 days. The full question text is “During the past 7 days, how many times did you or someone else in your family cook food for dinner or supper at home?” Time spent cooking is based on time usually spent cooking and cleaning up for dinner. The full question text is “How much time do you or someone else in your family usually spend on cooking dinner or supper and cleaning up after the cooking? Please do not include time spent eating.”

Figure 3 presents the predicted mean absolute energy intake (kcal/day) from all four Nova groups by household frequency of cooking dinner and time spent cooking and cleaning up from dinner based on multivariable generalized linear models. Energy intake from Group 1 foods increased 97 kcals from 571 kcals when cooking dinner 0–2 times/week to 668 kcals when cooking dinner 7 times/week (p<0.001). Mean energy intake from Group 4 ultra-processed foods was 138 kcals lower among those living in households where dinner was cooked 7 times/week compared to 0–2 times/week (1105 kcals vs. 1243 kcals, p<0.001). Similarly, energy intake from Group 1 foods was 77 kcals higher and energy intake from Group 4 ultra-processed foods was 94 kcals lower among those spending > 90 minutes/meal cooking and cleaning up compared to those spending 0–45 minutes/meal (p’s<0.001). Energy intake from Group 2 and Group 3 foods was similar by cooking frequency and time spent cooking.

Figure 3. Energy intake (kcal/day) across the four Nova classification categories by cooking frequency and time spent cooking, NHANES 2007–2010.^a.

^a Predicted mean energy intake (kcal/day) from each Nova processing group after generalized linear models with gamma family and log link regression models adjusted for age, sex, race and ethnicity, education, employment status, marital status, income to poverty ratio, household size, SNAP receipt, food security status, and total energy intake. Household cooking frequency is based on frequency of cooking dinner in the past 7 days. The full question text is “During the past 7 days, how many times did you or someone else in your family cook food for dinner or supper at home?” Time spent cooking is based on time usually spent cooking and cleaning up for dinner. The full question text is “How much time do you or someone else in your family usually spend on cooking dinner or supper and cleaning up after the cooking? Please do not include time spent eating.”

Differences in the proportion of total energy intake from the food subgroups that comprise the four Nova processing categories by frequency of cooking dinner is presented in Table 3. Greater intake of Group 1 unprocessed or minimally processed foods associated with higher cooking frequency was driven by legumes, roots and tubers, vegetables, fruit and juices, fish, grains, and other Group 1 foods. Lower intake of Group 4 ultra-processed foods associated with higher cooking frequency was driven by cakes, cookies, pies, pancakes and pastries, sweet snacks, frozen and shelf stable plate meals, pizza, French fries and potato products, sauces, dressings and gravies, and carbonated soft drinks. There were few notable differences in food groups comprising Group 2 processed culinary ingredients and Group 3 processed foods by frequency of cooking dinner. However, the p-trend for differences by cooking frequency were significant for table sugar (Group 2 (p=0.041)) and plant foods in brine (Group 3 (p=0.014)).

Table 3. Distribution of proportion of total energy intake^a from food subgroups within Nova categories by frequency of household cooking dinner/supper frequency, NHANES 2007–2010 (N=9,491).

	Household Cooking Dinner/Supper Frequencyb (times/week)				p-trend
	0–2	3–4	5–6	7
	% of total energy intake	% of total energy intake	% of total energy intake	% of total energy intake
Group 1: unprocessed or minimally processed foods
Legumes	0.74	0.87	0.85	1.11c	<0.001
Roots and tubers	1.78	1.97	1.98	2.14c	0.011
Vegetables	0.72	0.92c	0.93c	1.01c	0.001
Fruit and juices	3.89	4.73c	4.83c	5.49c	<0.001
Meat and poultry	8.68	9.41c	9.50c	9.42c	0.101
Fish	0.82	0.97	0.80	1.12c	0.039
Eggs	1.59	1.52	1.62	1.75	0.072
Milk and yogurt	4.17	3.78	3.76	4.09	0.777
Grains	2.57	3.18	3.02	3.41c	0.005
Pasta	1.57	1.73	1.77	1.68	0.675
Other	2.05	2.29	2.42c	2.49c	0.007
Group 2: processed culinary ingredients
Table sugar	1.40	1.24	1.33	1.57	0.041
Oils	0.92	1.18c	1.20c	1.14c	0.073
Animal fats	1.14	1.06	1.22	1.23	0.146
Other	0.07	0.07	0.09	0.09	0.224
Group 3: processed foods
Cheese	2.83	3.16	3.03	2.83	0.510
Ham and salted meats	1.07	1.09	1.08	1.16	0.247
Plant foods in brine	0.74	0.68	0.81	0.87	0.014
Other	4.41	4.42	4.95	4.53	0.612
Group 4: ultra-processed foods
Reconstituted Meat & Fish Products	3.32	3.17	3.22	3.12	0.496
Bread	10.96	11.10	10.73	11.03	0.949
Cakes, cookies, pies, pancakes and pastries	4.78	4.99	4.23	3.79c	<0.001
Ice cream, ice pops, and frozen yogurts	2.34	2.11	2.04	1.95	0.079
Desserts and other sugary products	0.88	0.91	0.91	0.89	0.930
Sugared breakfast cereals	2.82	2.99	3.03	2.47	0.051
Salty snacks	4.22	4.27	4.10	3.91	0.116
Sweet snacks	2.76	2.56	2.41	2.02c	0.001
Frozen and shelf stable plate meals	2.56	2.38	2.40	2.02c	0.016
Pizza	3.04	2.68	3.50	2.26c	0.042
Sandwiches and hamburgers on buns	1.31	1.07	1.17	1.07	0.481
French fries and potato products	2.05	1.44c	1.45c	1.43c	0.012
Instant and canned soups	0.83	0.69	0.75	0.83	0.581
Sauces, dressings and gravies	3.32	3.41	3.27	2.97	0.026
Sugared milk drinks	1.43	1.45	1.43	1.38	0.758
Soft drinks, carbonated	4.73	4.52	4.01	3.61c	<0.001
Other sweetened drinks	3.02	2.81	2.54	2.72	0.323
Other UPFs	4.36	4.65	4.25	4.61	0.591

^aBased on generalized linear model with gamma family and log link adjusted for age, sex, race and ethnicity, marital status, household size, employment status, education, income to poverty ratio, food security status, SNAP receipt, and total energy intake. Energy intake based on mean intake from day 1 and day 2 dietary recalls.

^bHousehold cooking frequency is based on frequency of cooking dinner in the past 7 days. The full question text is “During the past 7 days, how many times did you or someone else in your family cook food for dinner or supper at home?”

^cDifference from cooking 0–2 times/week significant at p<0.05.

Table 4 presents differences in the proportion of total energy intake from the food subgroups that comprise the four Nova processing categories by time spent cooking dinner. More time spent cooking dinner was associated with higher intake of the following Group 1 food subgroups: legumes, roots and tubers, vegetables, meat and poultry, and grains. More time spent cooking was associated with lower intake of cheese but higher intake of plant foods in brine (Group 3 processed foods). More time spent cooking was also associated with lower intake of salty snacks and pizza (Group 4 ultra-processed foods). More time spent cooking was also associated with higher intake of table sugar (Group 2 processed culinary ingredients).

Table 4. Distribution of proportion of total energy intake^a from food subgroups within Nova categories by time spent cooking dinner, NHANES 2007–2010 (N=8,947).

	Time spent cooking dinnerb				p-trend
	0–45 minutes	45–60 minutes	61–90 minutes	>90 minutes
	% of total energy intake	% of total energy intake	% of total energy intake	% of total energy intake
Group 1: unprocessed or minimally processed foods
Legumes	0.85	0.90	0.86	1.17c	0.022
Roots and tubers	1.86	1.99	2.10	2.26c	0.010
Vegetables	0.88	0.91	0.97	1.05c	0.028
Fruit and juices	5.03	4.77	5.19	5.10	0.506
Meat and poultry	8.86	9.24	9.88c	9.84	0.034
Fish	0.86	0.93	0.88	1.14	0.067
Eggs	1.61	1.64	1.49	1.74	0.552
Milk and yogurt	3.88	3.90	3.70	4.25	0.220
Grains	2.95	2.97	3.33	3.56	0.019
Pasta	1.59	1.78	1.61	1.69	0.914
Other	2.29	2.44	2.22	2.55	0.239
Group 2: processed culinary ingredients
Table sugar	1.40	1.28	1.39	1.68	0.029
Oils	1.08	1.15	1.04	1.26	0.197
Animal fats	1.15	1.14	1.27	1.28	0.070
Other	0.08	0.08	0.08	0.09	0.395
Group 3: processed foods
Cheese	3.17	3.14	2.84	2.59c	0.005
Ham and salted meats	1.14	1.16	1.08	1.02	0.237
Plant foods in brine	0.75	0.75	0.86	0.90c	0.023
Other	4.37	4.98	4.45	4.29	0.226
Group 4: ultra-processed foods
Reconstituted Meat & Fish Products	3.24	3.28	3.17	2.98	0.113
Bread	11.38	10.74	10.95	10.94	0.529
Cakes, cookies, pies, pancakes and pastries	4.44	4.19	4.76	3.95	0.288
Ice cream, ice pops, and frozen yogurts	2.00	2.08	2.24	1.98	0.965
Desserts and other sugary products	0.92	0.90	1.06	0.90	0.817
Sugared breakfast cereals	2.95	2.77	2.53	2.80	0.406
Salty snacks	4.43	4.22	3.94c	3.71c	0.001
Sweet snacks	2.77	2.13c	2.90	2.02c	0.103
Frozen and shelf stable plate meals	2.30	2.39	2.03	2.07	0.882
Pizza	3.09	3.18	2.38	2.41	0.019
Sandwiches and hamburgers on buns	0.94	1.16	1.47	0.98	0.729
French fries and potato products	1.47	1.53	1.63	1.33	0.494
Instant and canned soups	0.83	0.78	0.59	0.80	0.557
Sauces, dressings and gravies	3.24	3.18	3.19	3.18	0.793
Sugared milk drinks	1.46	1.37	1.52	1.24	0.339
Soft drinks, carbonated	4.22	4.11	3.68	3.86	0.195
Other sweetened drinks	2.79	2.42	2.85	2.87	0.318
Other UPFs	4.40	4.66	4.55	4.22	0.200

^aBased on generalized linear model model with gamma family and log link adjusted for age, sex, race and ethnicity, marital status, household size, employment status, education, income to poverty ratio, food security status, SNAP receipt, and total energy intake. Energy intake based on mean intake from day 1 and day 2 dietary recalls.

^bTime spent cooking is based on time usually spent cooking and cleaning up for dinner. The full question text is “How much time do you or someone else in your family usually spend on cooking dinner or supper and cleaning up after the cooking? Please do not include time spent eating.”

^cDifference from cooking 0–45 minutes significant at p<0.05.

DISCUSSION

In this nationally representative study, associations between household frequency of cooking dinner and time spent cooking dinner with the proportion of energy intake from foods with different levels of processing as measured by the Nova classification was examined. Despite ultra-processed food intake being high (>50% of energy intake) overall in the study population, both greater frequency of cooking dinner and more time spent cooking dinner were associated with higher intake of unprocessed or minimally processed foods and correspondingly lower energy intake from ultra-processed foods. Given the lower nutritional value of many ultra-processed foods,^23,24,47,48 these findings suggest the potential for programs and policies to improve home cooking skills and practices to facilitate shifts away from ultra-processed foods in favor of unprocessed or minimally processed foods.

Results from the present study are consistent with other evidence showing that ultra-processed food consumption is high (consistently >50% of total daily energy intake) among Americans.^6,48 Ultra-processed foods are associated with poor dietary quality and have adverse effects on health, including obesity, diabetes, cardiovascular disease, and some cancers.^4,23,24 Emerging evidence suggests that ultra-processed foods are associated with food addiction⁴⁹ and that their consumption may have intergenerational outcomes in which maternal consumption of ultra-processed foods during pregnancy⁵⁰ and childrearing⁵¹ is associated with the offspring’s verbal cognitive development and weight status respectively. Ultra-processed foods also contribute to food systems related climate change due to environmental degradation associated with their production and consumption.⁵² Thus, while ultra-processed foods can also contain beneficial nutrients,^53,54 the growing role of ultra-processed foods in the diets of US adults, and their proliferation across the globe,⁵⁵ present an important issue for individuals’ dietary quality,^6,24,56 risk of long-term chronic diseases and other health problems,^4,23,57,58 and the health and sustainability of the planet.⁵²

The results from this study are also consistent with other evidence showing that cooking at home is associated with multiple measures of better diet quality.^{12,14,15,26,59–62} Cooking more meals at home could be a strategy to consume fewer ultra-processed foods.⁶³ However, this depends on what foods and ingredients are being cooked, and, in the current study, ultra-processed food consumption remained high even among those living in households in which dinner was cooked most frequently and substantial time was typically spent cooking dinner. Of note, when examining differences in food subgroups that comprise the four Nova processing categories by cooking frequency and time spent cooking, there were few significant differences in consumption of the Group 4 ultra-processed food subgroups based on time spent cooking and cleaning up (only salty snacks and pizza were significantly lower with more time spent cooking). In contrast, cooking more frequently at home was significantly associated with higher intake of Group 1 minimally processed food subgroups, particularly legumes, roots and tubers, vegetables and grains and lower intake of Group 4 ultra-processed food subgroups, particularly cakes, cookies, pies, pancakes and pastries, sweet snacks, frozen and shelf stable meals, pizza, French fries and potato products, sauces, dressings and gravies, and carbonated soft drinks. This reflects the large role that ultra-processed foods play in US adults’ diets.⁶ However, while frequency of cooking dinner at home was associated with lower intake of specific ultra-processed food subgroups, results cannot speak to whether or how much ultra-processed foods were used in meals cooked at home. Other evidence suggests that, for many people, ultra-processed foods are often used in meals prepared at home,^17,64,65 and the frequency of preparing meals with ‘scratch or fresh’ ingredients (Group 1 unprocessed or minimally processed foods) is consistently lower than measures of cooking frequency without that qualification.^11,18

Reducing consumption of ultra-processed foods is increasingly a focus of public health authorities and policy makers given the growing body of evidence regarding their adverse health effects.^{4,23,57,58,66,67} During the White House Conference on Hunger, Nutrition and Health in September 2022,⁶⁸ multiple policymakers and public health leaders focused on ultra-processed foods as a key target for improving diet quality and health outcomes. Proposed strategies, however, mostly focused on increasing access to “healthy” food in communities with low access and improving the quality of food offered in institutional settings such as schools.⁶⁹ While those efforts are critical, given the ubiquity of ultra-processed foods in the food supply and on grocery store shelves, the path to reducing consumption of ultra-processed foods should also consider at home food environments and the skills, capacities, and resources necessary to reduce use of ultra-processed foods during home food preparation.⁷⁰ Targeted strategies or messages may be needed to observe meaningful reductions in the amount of UPFs consumed, even when people are cooking frequently or spending a lot of time cooking. Future research to develop and evaluate approaches to reducing UPF consumption is needed and it will be critical for those efforts to consider both food access and whether food and cooking skills and behaviors can play a role in meeting this goal. High quality food and cooking skill development programs may help reduce some barriers to cooking from home with more unprocessed or minimally processed foods, and could have benefits to dietary quality and diet related health outcomes over the long term. Well-designed evaluations with robust study designs of such programs are needed.^61,71,72

In addition to fostering food and cooking skills, it is critical that policies and public health programs consider the time, price and convenience tradeoffs that people make when making food choices.^17,73–76 Cooking interventions focused on building skills for scratch cooking may be promising and helpful for some, but given the ubiquity of ultra-processed foods, and challenges of modern-day schedules, creation of affordable “convenient” and less processed foods may also be needed to truly move the needle on ultra-processed foods to make consuming a less processed diet an easier choice for everyone regardless of how frequently they cook. Strategies to help make convenient pre-prepared, or ready-to-heat meals, made without ultra-processed foods, more widely available and affordable in a variety of venues would also help make it easier for more individuals to reduce their consumption of ultra-processed foods without having to necessarily cook more frequently or spend more time cooking at home. Importantly, more research is needed to understand barriers and facilitators for shifting towards less processed diets, and how cooking skills and structural factors are associated with ultra-processed food intake to inform policies and public health strategies to improve diet quality among US adults.

Strengths and limitations

Strengths of this study include a large, nationally representative sample that included two 24-hour diet recalls collected using a validated approach,^77,78 and two complementary measures of cooking behavior. To the authors’ knowledge, this is the first nationally representative study to examine how cooking frequency and behavior is associated with ultra-processed food consumption among US adults. This study has several limitations that should be considered. First, the data are cross-sectional and are from 2007–2010, the only years in which NHANES collected information on frequency of cooking dinner at home. It is possible that cooking frequency, and the degree to which ultra-processed foods are used in home cooked meals, have shifted in the years since the data for this study was collected. Second, although NHANES 24-hour recalls collect some information indicative of food processing, these data are not consistently determined for all food items, which could lead to inaccuracies in Nova estimates. Additionally, there could be inaccuracies in the Nova classification estimates because, 1) some potential handmade mixed dish food codes could not be disaggregated into constituent ingredients due to being matched to a single SR Code, or; 2) the constituent SR Codes obtained from standard recipes may not necessarily represent the consumed underlying ingredients.³⁹ On the other hand, social desirability bias,⁷⁹ may lead people to underreport ultra-processed food consumption. Third, frequency of cooking dinner is measured at the household level over the last seven days whereas dietary intake, and level of processing is an individual level measure measured using two 24-hour dietary recalls. It is unknown how frequently individual respondents consumed the home cooked dinners, or the level of processed foods used in the preparation of those meals. Prior evidence suggests that the interpretation of what it means to cook varies across individuals and a key point of variability is the extent to which different people include use of processed or convenience foods in their definition.^17,64 These differences could complicate the relationship between cooking frequency and ultra-processed food consumption. Relatedly, the measure of cooking frequency in this study only captures frequency of cooking dinner, as this is the only cooking frequency measure included in the NHANES. Finally, it is unknown how frequently other meals were cooked at home or how frequently home cooked meals were consumed. More research is needed, including studies with robust study designs such as randomized controlled trials, with detailed measures of foods prepared and consumed at different meals (including different sub-categories of foods such as snacks, beverages, frozen foods), to fully understand the relationship between cooking at home and ultra-processed food consumption, and other measures of nutritional profile/dietary quality (sodium, energy density, saturated fat), and whether the results of this study would be replicated with more recent samples and more detailed data related to cooking behaviors.

CONCLUSION

Among adults in the US, cooking dinner more frequently at home, and greater time spent cooking were both associated with greater consumption of unprocessed or minimally processed foods and lower consumption of ultra-processed foods. However, regardless of cooking frequency or time spent cooking, ultra-processed foods comprised, on average, more than half of all energy consumed. These findings suggest that, while cooking more frequently at home may enable some, particularly those able to spend time cooking dinner, to consume fewer ultra-processed foods, ultra-processed food consumption is high regardless of cooking frequency.

RESEARCH SNAPSHOT.

Research Question:

Are greater household frequency of cooking dinner and greater time spent cooking associated with lower intake of ultra-processed foods and higher intake of unprocessed or minimally processed foods?

Key Findings:

In this cross-sectional study of 9,491 adults in the US from the 2007–2010 National Health and Nutrition Examination Survey, greater household frequency of cooking dinner and greater usual time spent cooking dinner were associated with lower intake of ultra-processed foods and higher intake of unprocessed or minimally processed foods. However, consumption of ultra-processed foods as a percent of energy intake was high regardless of cooking frequency or time spent cooking.