Validation of a home cooking quality assessment tool using community science and crowdsourcing approaches

Abstract

Objective:

To refine a measure of home cooking quality (defined as the usage level of practices with the potential to influence the nutrient content of prepared foods) and conduct a construct validation of the revised tool, the Healthy Cooking Questionnaire 2(HCQ2).

Design:

Two validation approaches are described: 1) a community science approach used to refine and validate HCQ constructs, and 2) responses to the revised HCQ (HCQ2) in a sample of Amazon MTurk workers to determine questionnaire comprehension.

Setting:

The Community Scientist Program at the University of Texas MD Anderson Cancer Center facilitated discussion groups to refine the HCQ questions and validate constructs. MTurk “workers” were subsequently recruited to complete the refined survey so that comprehension and associations with demographic variables could be explored.

Participants:

Ten community scientists participated in the refinement of the HCQ. The revised tool (HCQ2) was completed by 267 adult US-based MTurk workers.

Variables Measured:

Demographics, HCQ concepts, HCQ2, Self-Reported Questionnaire Comprehension.

Analysis:

Comprehension items were examined using descriptive statistics. Exploratory analysis the relationships between cooking quality and demographic characteristics, meal type, cooking frequency, as well as patterns of food preparation behavior was conducted on the Mturk sample

Results:

The HCQ was refined through activities and consensus building. MTurk responses to the HCQ2 indicated high comprehension and significant differences in cooking quality scores by demographic factors.

Conclusions and Implications:

Community scientist groups and online crowdsourcing are feasible approaches to behavioral assessment refinement and validation. Cooking quality measures offer critical evaluation methods for culinary programs.

Introduction

The high prevalence of diet-related disease, including obesity, diabetes, and heart disease, is a major public health issue. Although the United States Department of Agriculture (USDA) has offered nutritional guidelines for Americans since 1916, adherence across the population is low^1,2. More practical nutrition education, namely cooking skill development, has been touted as a potential way to promote dietary behavior change^3–6. Higher frequency of eating home cooked meals, as opposed to eating at restaurants or getting take-out, has been associated with higher diet quality, lower food costs and more normalized body weight^7–11. However, the relationship between cooking frequency and diet quality is attenuated by income^{12 11}, suggesting cooking frequency alone may be an insufficient metric to understanding the role of home food preparation in diet and health.

Cooking is a complex activity including multiple behaviors that contribute to the biochemical composition of a prepared meal¹³. Cooking quality is the degree to which food preparation practices with the potential to positively or negatively influence the nutrient content of prepared foods are utilized. Nineteen specific practices relevant to cooking quality were identified through a systematic review of 59 observational and experimental studies, in which constructs including cooking techniques and methods, strategic ingredient replacements/additions/minimizations, and flavoring usage were consistently associated with dietary exposures and chronic disease risk¹³. The emergent constructs were structured into an evidence-based conceptual framework of healthy cooking, and face validity was explored with an expert panel¹³. The conceptual framework was operationalized into a behavioral index to measure cooking quality on a −9 to +10 scale, the Healthy Cooking Index (HCI)¹⁴. HCI measures cooking quality according to the demonstration of behaviors, scored +1 or −1 based on their potential to improve or worsen the nutritional quality of prepared meals. HCI scores based on observations of home cooking events (n=40) found higher HCI scores and component scores were associated with lower saturated fat, and higher fiber, fruit and whole grain content of prepared meals¹⁴, suggesting these food preparation practices influence dietary exposures related to health and that cooking quality can be estimated through an examination of these key behaviors.

Existing nutrition interventions do not typically examine cooking quality improvement, nor associate improvement in food preparation practices with downstream correlates of nutrition and health such as weight or circulating markers of inflammation¹⁵. One of the biggest challenges is the lack of validated tools¹⁶. Easy to use, culturally flexible assessment tools that dynamically measure the degree of cooking quality are needed to help understand nuances of modern food preparation behaviors in the population and support the development of best practices for culinary education interventions. The objective of this study was to refine and validate the previously developed HCI-based Healthy Cooking Questionnaire (HCQ). Two validation approaches are described here: first, a community science approach was used to refine and validate constructs in the questionnaire, and second, responses to the revised HCQ (HCQ2) in a sample of Amazon Mturk workers were examined to determine questionnaire comprehension and relevance. Exploratory analysis of the relationships between cooking quality and demographic characteristics, meal type, cooking frequency, as well as patterns of food preparation behavior were also presented.

Methods

Development of the HCQ2

The HCQ2 was developed using a multi-step development process for a metric of cooking quality (Figure 1).

Figure 1:

Healthy Cooking Questionnaire Development Process. This scheme shows the four step process in developing the HCQ.

Earlier phases (steps 1 and 2) have been detailed elsewhere¹⁴; this report focuses on revision and validation of cooking quality metric. Briefly, the Healthy Cooking Questionnaire I (HCQ) is a self-report tool mapped to an evidence-based conceptural model, the HCI described above¹³ (Figure 1, Step 1). The grounding framework was used to guide participatory cooking classes in diverse communities to ensure relevance of included framework items across racial/ethnic and socioeconomic strata^{17 18}. An observational study of 40 families was undertaken to examine the accuracy of the HCQ self-report tool compared to in person and body camera observations of a singular cooking event¹⁴ (Figure 1, Step 2).

Significant differences between observed and self-reported cooking behaviors were found. HCQ items with 20% or more error (compared to observed behavior) were extracted for review and revision. As the HCQ was administered directly after a cooking event, it is unlikely that these errors were due to recall bias. Further, errors were centralized around a small number of items, suggesting certain concepts were not clearly operationalized by the HCQ in the context of modern family cooking. Three main concepts were reported with the lowest accuracy including: 1) using processed food vs cooking entirely from basic ingredients, 2) cooking with animal fats, and 3) measuring fat and/or salt. These concepts and their corresponding items were the main focus of the community scientist review (described below), during which the HCQ was explored through activities, discussion, and consensus building with a view to revising and updating the HCQ.

Community science approach:

Community science broadly refers to research in which non-scientists support the development of various aspects of scientific research. The Community Scientist Program at the University of Texas MD Anderson Cancer Center (MD Anderson) is a coordinated resource for investigators that brings together groups of non-scientist individuals from multi-cultural backgrounds who have been briefly trained in research principles. The goal of the Community Scientist Program at MD Anderson is to facilitate community input into the research process to enhance research quality and ultimately, the translation of scientific research into community benefit. This approach is well-suited to the development of a home cooking quality metric as cooking is a familiar yet complex behavior that is potentially influenced by social factors including cultural norms, personal preferences, and familial structure, as well as more upstream factors such as individual behavioral capacity and food access. Little formative research has been done on modern home cooking practices, limiting understanding of home cooking in the twenty-first century. As members of the general population, community scientists facilitate the development of questionnaire items that are more relevant to the lived experience of home food preparation.

For this project, community science groups were enlisted to help refine the HCQ prior to testing the tool in a sample of online respondents (HCQ2 study). Two, 1-hour community scientist feedback sessions were undertaken to review and revise the original HCQ. Ten community scientists attended both sessions, approximately 1 month apart.

The first session included an introduction to cooking and nutrition research as well as an explanation of the 3 project goals: 1) to understand how different people define key cooking terms, 2) to identify logical ways to group ingredients and 3) to use this information to build an assessment tool of cooking quality. This discussion was followed by 2 activities to understand group perceptions of processed foods. The activities utilized drawing methods, which offer a more in-depth understanding of a participant’s point of view¹⁹. This approach highlighted perceptions of foods participants typically had at home, as opposed to theoretical classifications of products with which they may or may not be familiar.

In order to improve animal fats reporting, a structured pile sort technique was used, which allowed for the identification of perceived relations between the items^{20 21}. This technique clarified the questionnaire items on animal fats by breaking the concept of “animal fats” up into more logical sub-categories. Participants were given a stack of 16 cards, each featuring a different type of fat or oil. The participants were instructed to sort the cards into 2 or more piles that made sense to them. Piles were then banded together and kept separate for analysis. A matrix showing each type of fat and the number of times it was matched with each other type of fat was developed. Fat types were classified according to how often they were sorted into the same pile by participants, and a separate question developed for each classification. Following this activity, community scientists then reviewed the two existing HCQ questions regarding the measurement of fat and salt, and group consensus was used to determine modifications to the questions.

In the second community scientist session, all revised HCQ questions were reviewed with the group. Community scientists had an opportunity to comment on all concepts and related items during the sessions either verbally or through anonymous comment cards. Final versions of questions were determined as a group.

Validation Testing of HCQ2 Using a Crowdsourcing Platform

Once revised, the modified tool (HCQ2) was examined using Amazon’s Mechanical Turk (MTurk), an online crowdsourcing platform²². US-based, English-speaking individuals over the age of 18, were recruited for participation in an anonymous HCQ2 survey study. Upon accepting participation in the survey study, MTurk workers were directed to an informed consent webpage through the online platform requesting indication of consent. Participants were then instructed to provide basic demographic information, complete the HCQ2 (Supplemental Data 1), and indicate overall questionnaire comprehension and issues/feedback. The HCQ2 was estimated to take approximately 8 minutes to complete. Participants were compensated $0.65 for the study, based on compensation ranges provided in other studies^{23 24}. To help ensure quality and validity of responses, a question instructing respondents to select “b” as the answer was randomly interspersed as a quality check. Respondents who missed the quality check question were removed from analysis. Respondents who indicated they last cooked their main meal at home 3 or more days ago were also removed to mitigate recall bias²⁵. The IRB of The University of Texas MD Anderson Cancer Center reviewed all elements of this study and deemed it exempt (PA19-0145).

Data Analysis

Six questions of the original HCQ were revised during community scientist sessions using group activities and consensus building (detailed below), resulting in a refined metric, the HCQ2. Data obtained from MTurk participants was coded and analyzed using a variety of statistical methods. Cooking quality scores were calculated as summative score of all items on the HCQ2 (range = −9 to +10). The HCQ2 relies on a high percentage of respondents indicating: 1) cooking a main meal at home within the last 2 days and 2) full comprehension of all survey questions in order to be a useful metric of cooking quality. Sample demographic and cooking frequency characteristics and questionnaire comprehension items were assessed with descriptive statistics. Differences in HCQ2 scores by demographic variables were explored using ANOVA or independent sample t-tests as appropriate. Correlations between cooking quality score and continuous variables (age, number of children) were examined with Pearson correlation coefficients. The goal of these analyses was to explore the potential influence of demographic factors on cooking quality. All inferential analyses used a two-sided 0.05 significance level; analysis was completed with statistical software [IBM SPSS Statistics for Windows, Version 25.0, Armonk, New York, USA].

Results

HCQ Refinement: Community Science Feedback Sessions

Revisions to the HCQ are detailed in Table 1. As the original questionnaire was administered directly after an observation of individuals preparing dinner, the participants first discussed optimal approaches to defining an individual’s main (or largest) meal and the last time it was prepared.

Table 1:

Healthy Cooking Questionnaire Concepts and Opportunities for Refinement

Concept	HCQ Original Item / Issue	Revision Approach	Resolution	Revised question
Main meal preparation	Original testing occurred directly after a home cooking event of an evening meal. All questions were prefaced “During the observation session…”	Discussion of how to operationalize the concept of cooking main or largest meals in diverse food environments and group consensus	The term “main meal” was combined with “largest” meal for clarity. Added question to indicate main meal time to avoid presumptions of “dinner” or “supper” being the largest meal. Added time estimation for last cooked main meal to be able to remove those that cooked several days ago from future analysis to reduce recall bias.	Q1. What do you normally consider to be your main (largest) meal of the day?: Breakfast/Brunch/Lunch/Dinner Q2. When was the last time you prepared your main meal (selected above) at home?: today/1 day ago//2 days ago/3 days ago/4 days ago/More than 4 days ago
Processed Food / Basic Ingredients	Qa. During the observation session, did you cook with processed foods (such as ready to heat meals, frozen pizza, bottled salad dressing, hamburger helper, pre-made dips?: Yes/No/ I don’t know Qb: During the observation session, did you cook “completely from basic ingredients”? (all basic, raw ingredients such as fresh, dry or frozen fruits or vegetables, grains, legumes, meat, fish and/or milk, salt, spices, unflavored oils): Yes / No/ I don’t know	Label fridge/pantry and rank processed foods from most to least processed. Pile sort processed foods into groups	Offer series of processed foods from each group and prompt respondent to add related in “Other” open text box. Group processed foods that are spice mixes with other spices. Group sauces/dressings together, ask about something jarred, boxed and frozen to promote reporting of other packaged products that come in various forms. Avoid questions about “basic ingredients” or “from scratch” cooking due to high variability in definition	Qa.1 The last time you prepared your main meal at home, did you use any of the following herbs / spices / flavorings? Select Yes or No: Salt/Pepper/Oregano/Parsley/Cumin Garlic or Garlic powder/ Garlic salt or Onion salt/ Cayenne/Chili powder/Cajun seasoning (Tony Chachere’s / Season All / etc)/Taco Seasoning/Adobo Seasoning/Lemon juice or zest/Orange juice or zest/Other, if yes please explain [open text] Qa.2 The last time you prepared your main meal, did you use any of the following convenience foods? Select Yes or No: Jarred Pasta Sauce/Canned Goods/Bottled Salad Dressing/Soy Sauce/Teriyaki Sauce/Ketchup | Mustard/Boxed Mixes/Frozen Meals/Frozen Vegetables/Other, if yes please explain [open text] Qb: deleted
Animal Fat	Qc. During the observation session, did you use any of the following fats (check yes or no for each option): Lard /Chicken fat/Butter/Bacon Grease/Full Fat Cheese/Cream/Olive Oil/Canola Oil/Other [open text]	Pile sort animal fats into groups	Break up fats to true animal fats, dairy, and vegetable-based oils	Qc. The last time you prepared your main meal at home, did you use any of the following (please check yes or no for each option): 1.1 Animal fats Lard/Bacon Grease/Tallow/Chicken Fat/ Other [open text] 1.2 Dairy: Butter/Cheese/Low Fat Cheese/Sour Cream/Cream/Half&Half/Other [open text] 1.2 Oils: Olive Oil/Canola Oil/Vegetable Oil (corn, soybean, blends)/Coconut/Palm Oil/Avocado Oil/Other [open text]
Measured Salt / Fat or Oil	Qd. During the observation session, do you know approximately how much oil or fat you used?: Yes, I used about [open text] (amount, please note teaspoons| tablespoons or cups in measurement) of [open text] (type of fat)/No/I don’t know/ I did not use oil or fat Qe. During the observation session, do you know approximately how much salt or salty seasoning (fish sauce, soy sauce, seasoning mix) you used?: Yes, I used about [open text] teaspoon(s) of [open text] (specify salt | type of salty seasoning)/No/I don’t know/I did not use salt or salty seasoning	Group discussion or how to operationalize the concept of intuiting measurement in the home food environment and group consensus	Increase answer options to account for intuitive measurement as opposed to formal measuring spoons/cups etc. and add emphasis to improve readability	Qd. The last time you prepared your main meal at home, did you measure the amount of oil or fat you used?: Yes, I measured the amount of oil or fat used/No, I did not measure the amount of oil or fat used BUT I can accurately guess the amount/No, I did not measure the amount of oil or fat used and I could not accurately guess the amount/I did not use oil or fat Qe. The last time you prepared your main meal at home, did you measure the amount of salt you used? Yes, I measured the amount of salt used/No, I did not measure the amount of salt used BUT I can accurately guess the amount/No, I did not measure the amount of salt used and I could not accurately guess the amount/I did not use salt

In the first drawing activity, participants were given an outline of a blank refrigerator and asked to draw a representation of the foods they currently or typically have in their home refrigerators. Participants were then asked to circle the foods that they considered to be “processed” in red and foods that they considered to be “unprocessed” in blue, and then rank the red-circled foods from least to most processed. This process was repeated, focusing on the home pantry. Commonly reported processed food items and their ranks were compared across the 10 participants. Less processed foods reported by participants included oatmeal, mixed seasonings and canned vegetables. More heavily processed foods included canned soups/stock/sauces, peanut butter, and chips. This approach clarified how individuals conceptualize processed foods and their level of processing, which was often a function of the food’s packaging as opposed to ingredients. As a result, some foods that are minimally processed but packaged, such as frozen vegetables, were listed as a selection but not coded as a processed food during scoring. The final question was structured to ask about canned, bottled, boxed, and frozen foods, with sauces/condiments listed together. Processed foods that are used as spices, such as “taco seasoning”, were recategorized with other herbs and spices. An “other” open-text item was added to the end of the new questions to prompt reporting of similar foods for each item. Other items were reviewed individually during scoring for categorization into minimally processed or processed/ultra-processed foods. Only processed/ultra-processed food usage, as defined by the NOVA classification system²⁶, resulted in a score of −1.

This was followed by the pile-sort activity examining fats. Fat types were classified according to how often they were sorted into the same pile by participants, and a separate question developed for each classification. Participants most commonly grouped fats together based on fat origin: dairy-based (butter, cream, cheese, etc.), non-dairy animal-based (beef tallow, chicken fat, etc.), and vegetable-based (olive oil, canola oil, avocado oil). This resulted in separate questions about dairy fats, animal (non-dairy) fats, and vegetable-based fats.

Two questions about measurement were reviewed, one regarding salt/salty seasoning usage and another about fat/oil usage. In the home cooking environment, formal measuring utensils are not always used. However, some individuals have other ways of estimating salt or oil usage such as first placing salt in the palm of the hand, or counting seconds of pouring oil. These more informal measurements are difficult to capture but represent some level of awareness of fat and salt usage. This concept was reviewed through group discussion centered around the following questions: 1) Do you measure salt or oil when you cook? 2) What do you use? (A measuring utensil/ your hand / count it out / “just know”?) 3) Using those tools, could you tell me about how much salt/oil you used with confidence? The discussion generated wording to capture both formal and informal measurement types, as well as lack of measurement.

In the second community scientist session, participants were given a stack of cards, each with a different HCQ revised item. Each of the 5 revised concepts had between one and three revision options. Participants were encouraged to use the cards for reference and write any comments they did not wish to share during the group discussion. Each concept and related items were then presented to the group and discussed. Group consensus (majority of votes) was used to identify the optimal revised questions for each concept. Comments were noted and further revisions were incorporated as appropriate. The revision resulted in a new time frame for respondents and six new or revised questions. The final questionnaire, the HCQ2, underwent pilot testing among research lab members in order to estimate completion time and correct minor issues. The HCQ2 was then distributed to an adult sample via Amazon’s MTurk crowd sourcing program.

Construct Validation: HCQ2 Study Results

Participant Characteristics.

A total of 267 US-based MTurk workers provided responses to the HCQ2. Six failed to pass the quality check question. Twenty-four participants reported cooking their main meal three or more days ago and were removed to reduce potential recall bias. Six participants were removed due to implausible demographic responses. A total of 231 respondents were included in the analysis. Participant characteristics are shown in Table 2. Participant ages ranged from 19 – 71 years old with most (78.3%) under 40 years old. The majority (81.8%) of participants indicated lunch or dinner as their main meal of the day.

Table 2:

Respondent characteristics

Item	n	%
Sex
Male	117	50.6
Female	114	49.4
Race / Ethnicity
Non Hispanic White	148	64.1
Non Hispanic Black	26	11.3
Hispanic Black	21	9.1
Hispanic White	15	6.5
Asian	11	4.8
Other	6	2.6
American Indian / Alaskan Native	4	1.7
Education
Less than College Degree	83	35.9
Bachelor’s degree	113	48.9
Post graduate degree	35	15.2
Employment
Employed full time	182	78.8
Employed part time	25	10.8
Other (homemaker, unemployed, retired, student)	24	10.4
Marital Status
Married / Living with Partner	143	61.9
Single	72	31.2
Other (divorced, widowed, separated)	16	6.9
Main Meal
Dinner	143	61.9
Lunch	46	19.9
Breakfast	36	15.6
Brunch	5	2.2
Last Time Cooked Main Meal
Today	48	20.8
1 day ago	153	66.2
2 days ago	30	13
Mean Age (years)	35	SD=10.5
Mean # children at home	0.8991	SD=1.2

Questionnaire Comprehension and Relevance.

With regard to survey comprehension, 99.6% indicated that they understood all questions in HCQ2. Regarding reported improvements, 3 respondents reported structural issues with the online delivery (e.g., font size, advancing options), and 2 requested changing the number of questions (shorter (n=1) or longer (n=1). The HCQ2 was relevant to the population, with (91%) reporting cooking their main meal at home within the last 2 days. Participants reported a range of individual behaviors, both positive and negative, suggesting the items are applicable across different food environments (Figure 2). The most common cooking practice was using low fat cooking methods (baking, steaming, boiling, etc.) (93.1%) and the least common cooking practice was marinating meat (4.3%).

Figure 2:

Response to Healthy Cooking Questionnaire Items. Bar graph depicting percentage of respondents indicating using each food preparation behavior the last time they prepared their main meal at home.

Relationships between HCQ2 cooking quality scores and demographic variables.

There were no significant differences in HCQ2 cooking quality score based on sex (t(229)= −.551, p= 0.58), marital status (F(2,228) = 1.29, p = .28), or educational attainment (F(2,228) = .816, p=0.44). Age was positively associated with HCQ2 score (Pearson correlation =.179, p = .006), and number of children was negatively associated with HCQ2 score (Pearson correlation = −0.158, p = 0.017). There were no significant differences in cooking scores by meal reported (F(3, 226) = 2.11, p = 0.10) or time since last cooked at home (F(2,228) = 0.318, p= 0.73). HCQ2 scores differed by employment status (F(2,228) – 3.03, p= 0.05) and race/ethnicity (F(6,224) = 3.44, p=0.003). Those that reported being employed full time demonstrated significantly lower HCI scores than those that reported employment as other (homemaker, unemployed, retired, student) (mean difference= −0.86 +/− 0.39, p =0.03). Hispanic Black participants demonstrated significantly lower HCI scores than Non-Hispanic White (−1.49 +/− 0.38, p=0.01), and Non-Hispanic Black (−1.84 +/−0.44, p=0.003); however it should be noted that this group only included 26 individuals, or 11.3% of the total sample.

Discussion

The measurement and quantification of home cooking quality can greatly enhance understanding of the mechanisms linking cooking and health and support the development of objective assessment tools and best practices in the growing field of culinary education. A multi-step process was used to develop and test the HCQ2, a self-report assessment tool of home cooking quality. After a thorough community scientist review, the HCQ2 was distributed through Amazon’s MTurk platform. The HCQ2 study found overall high comprehension among participants and significant differences in HCQ2 scores by race/ethnicity, age, and employment. Most participants cooked within the last two days and reported multiple HCQ2 behaviors. Taken together, these findings suggest the HCQ2 may be used to assess cooking quality in many settings and offer meaningful context to measures of overall diet quality.

One unique aspect of this study was the community scientist approach for refinement of the HCQ2. Cooking is influenced by taste preferences, cultural norms, family structure and other factors, and in-depth research on modern home cooking practices is limited. Further, it is unclear if conceptualizations of cooking have shifted with the advent of convenience foods and more advanced home cooking equipment. The community science approach deepens collective understanding of the modern cooking experience²⁷. The community science-informed method to item operationalization may be one reason for the high rates of self-reported comprehension of the HCQ2 during subsequent testing. Other studies successfully used community (also known as “citizen”) science in similar ways to develop evaluation plans and metrics for community-based health programs^{28 29}.

Other tools to measure cooking practices have been created in recent years, but they have many limitations. Existing metrics focus on confidence or technical ability to complete a set of specific food preparations or general cooking attitudes^7,30–34. Focusing on one’s ability to cook specific a priori dishes is challenging in the context of a culturally diverse population like the United States, which does not have universal culinary staples like Switzerland³⁵ or Japan³⁴. The validation of such metrics is also problematic, as commonly used metrics such as internal consistency may not be particularly meaningful when examining a cooking quality tool, given the complexity of cooking behavior. Previous work with the HCI demonstrated only moderate internal consistency¹⁴, which was expected given that the tool was developed to function within a culturally heterogenous food environment. The global cuisines (i.e., patterns of traditional food preparation practices) that inform the “melting pot” of modern American foodways typically include a combination of both negative and positive practices, all of which vary by cuisine origin and are unlikely to track with latent psychosocial constructs such as confidence or attitudes in the broader population. Measures of attitudes around cooking / willingness to cook and cooking frequency are much easier to administer and validate, however, without additional, culturally-flexible metrics of cooking quality these tools fail to acknowledge that cooking more doesn’t necessitate cooking well.

The use of MTurk may limit conclusions from this study; although the platform has been used for related assessment tool validation in the past³⁶, it should be noted that the MTurk work model may attract a specific subset of individuals that are not comparable to the general population. Also, the community scientist and mTurk samples were limited in size, further hindering generalizable conclusions. An alternative measure of diet, which would have allowed for a comparison between HCQ2 scores and diet quality or exposures, was not used in this study and was a limitation. The next steps of this work include validation of the HCQ2 against ground truth metrics of home cooking behavior, such as direct observation or body camera images.

Implications for Research and Practice.

Cooking education is an increasingly popular approach to nutrition education across multiple sectors including schools, hospitals, and community centers³⁷. The HCQ2 makes a unique contribution to the field by offering a robust self-report tool to assess home cooking quality. Cooking education programs are typically based on recipes, but participants are unlikely to exclusively use intervention recipes once the interventions are complete. The HCQ2 assesses nutrition optimizing cooking practices as opposed to one’s general confidence or capacity to cook a specific dish. As such, HCQ2 scores pre and post culinary intervention may help elucidate the degree to which knowledge is transferred from teaching kitchens or cooking interventions to home food environments³⁸. This is essential to understanding the effectiveness of cooking education programs in successfully communicating cooking skills with the potential to positively influence diet and biological correlates of disease. The HCQ2 is a brief self-report tool, making it an accessible metric for population-based studies that seek to examine associations between cooking and health. Future research may consider integrating the HCQ2 into larger representative surveys in order to better elucidate the role of cooking quality in dietary exposures and downstream correlates of health.