Participants were Amazon Mechanical Turk (MTurk) workers who were recruited to complete human intelligence tasks (HITs) through the MTurk site in 7 batches between November 2020 and January 2021. Participants were required to be aged 20 years or older, read and write in English as their primary language, live in the United States, and have at least a 98% task approval rating from completing previous tasks. If eligible, participants were provided the survey link and were required to reach the end of the survey in order to receive a completion code. The code was the product of their Research Electronic Data Capture (REDCap; Vanderbilt University)–generated participant ID multiplied by a random number (ie, 117) and was thus unique to each participant. In addition, precautions were taken to ensure that the same worker did not take the survey multiple times, whereby workers were marked with a custom qualification through the Amazon platform, which then restricted potential workers in future batches to only those who had not previously completed the survey. Once the survey was completed, participants were further screened through a 2-level process for response inattention, thoroughness, English language fluency, and automated web robot (bot) responses [35,36].

Participants were grouped according to self-reported experience with gardening based on the following question: “Have you ever participated in vegetable gardening in your lifetime? Yes or no.” Those who responded “no” were labeled as NGs, and all others were considered EGs. Participants completed basic demographic questions including age, race (self-selected), Hispanic or Latino ethnicity (yes or no), sex (self-selected, male or female), and education level. Participants were asked if they had received Supplemental Nutrition Assistance in the past year (yes or no) and answered the 6-item US Department of Agriculture (USDA) food security screening questionnaire [37]. Food security responses were scored according to USDA guidelines and dichotomized as low or very low security (score: 2-6) versus food security (score: 0-1) [37]. The neighborhood of residence had self-reported options for urban, suburban, and rural.

Health-related characteristics were reported using questions from the Behavioral Risk Factor Surveillance System Survey (BRFSS) [38]. General health was assessed with a single item from the BRFSS (“Would you say that in general your health is…”), and responses were dichotomized as good or better (excellent, very good, or good) versus all others (fair or poor) [38]. BMI was calculated with self-reported height and weight (weight in kg divided by the square of the height in meters) and dichotomized using standard cut points for normal weight (BMI 18.5-24.9 kg/m²) versus overweight (BMI≥25). Current smokers reported smoking at least 100 cigarettes in their lifetime and currently smoking every day or some days. Participants were identified as having any CVD risk factors (yes or no) if they reported being a current smoker, had a BMI≥25, or responded positively to any of the following individual BRFSS questions: “Has a doctor, nurse or other health professional ever told you that you had any of the following? [CVD or Angina, Stroke, Heart Attack, High cholesterol, high blood pressure, Diabetes].”

We asked 6 questions about participant perceptions of homegrown fruits and vegetables and garden space availability. Perceptions of homegrown foods were assessed with 4 items: “Do you think homegrown fruits and vegetables [taste different; differ in cost; differ in quality; differ in appearance] compared to store bought fruits and vegetables? [Yes or no].” If yes, participants were further asked to choose one of 2 options: “I think homegrown fruits and vegetables [taste better or taste worse; cost less or cost more; are higher quality or are lower quality; have a better appearance or have a worse appearance] than store-bought fruits and vegetables.” Questions about garden space included a root question asking: “Where you live now, do you have any space outside where you can grow fruits, vegetables, and herbs? [yes or no]” and “Is there space for you to grow plants in pots? [yes or no]” with responses to describe this outdoor space as “in pots on a patio or deck, in pots on a balcony, or in pots on the ground.” Responses were categorized as no outside space versus any.

Three cooking-related constructs were assessed, including food agency, cooking skills, and food skills [39,40]. These constructs are associated with higher consumption of fruits and vegetables in adults [41] and were included because they are a natural prerequisite for consuming homegrown produce. Food agency was assessed using the Cooking and Food Provisioning Action Scale and scored using recommended analytical methods, yielding a total score and 3 subscale scores: cooking self-efficacy, attitudes, and structural barriers [40]. Cooking skills, related to cooking methods and preparation techniques, were assessed with 14 questions on a 7-point scale (very poor=1 to very good=7) and a total score was calculated by summing the points from each question (score range 14-98). Similarly, food skills related to meal planning, shopping, budgeting, and cooking resourcefulness were assessed with 19 questions on a 7-point scale (very poor=1 to very good=7), and a total score was calculated by summing the points from each question (score range 19-133) [39].

To assess interest in participating in a gardening program, participants were asked the following question using a 3-option choice format used in previous clinical studies to gauge interest in future programs [42,43]: “Would you consider participating in any of the following programs if offered for free? (1) A series of 1-hour web-based sessions offered weekly where experts would lecture on a variety of gardening topics (yes or no), (2) a series of brief YouTube videos (less than 5 minutes) that provide “how-to” instructions on specific gardening topics (yes or no), and (3) a community gardening program where you and other people would meet in-person a few times a month to learn how to garden (yes or no).” Options 1 and 3 were chosen based on the educational formats commonly offered through national cooperative extension programs [44,45], and option 2 was based on formats used in popular social media video platforms (eg, YouTube and TikTok). An affirmative response to any of these 3 options was coded as “any interest” in a gardening program. For delivery method preference, participants were also asked, “Of these programs, which do you think is the most interesting?” and could choose one of the 3 options.

Participants were asked to complete 3 open-ended questions to understand what information they would consider important in a gardening program, including: (1) “If we were going to create a gardening program, what kinds of things would you like to see included?”; (2) “What information do you think you (or someone who is new to gardening) would need to start a vegetable garden on your (their) own?”; and (3) “What types of information sources would you most want [were helpful for you] to learn about how to start a vegetable garden?” One additional question also asked what sources of information participants would use to learn about gardening.

We began the analysis of the qualitative open-ended questions with 2 investigators (SV and MWZ) independently reviewing all the participant responses. Using the constant comparative method, we identified words or phrases with like meaning and unitized these phrases with codes [46,47]. We then met to establish a formal codebook with illustrative participant phrases for each code. Disagreements were discussed, and the codebook was modified as needed. Both investigators coded an initial sample of the data and achieved a high level of interrater reliability (Cohen κ=.82) [48]. Once all the data were coded, we organized the codes into intervention function categories based on Michie et al’s Behavior Change Wheel [32]. To assist with visualizing the relative importance of each intervention function, we presented counts of how many times each intervention function was mentioned and then calculated the percent of the total comments. Comments related to intervention delivery modalities (eg, “provide videos”) rather than intervention function were removed from the intervention function counts but were retained, coded, and reported separately. The quantitative data were summarized, and differences between NG and EG were assessed using t tests and chi-square tests, as appropriate.

This study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Penn State University Institutional Review Board (protocol #15352, approved June 14, 2020). All participants were provided a written summary explanation of the research and completed informed consent prior to providing study data. All study data were collected using REDCap [49] to ensure participant privacy and confidentiality. All study data were collected anonymously, and participants are not individually identifiable. Participants were compensated US $3 via the Amazon MTurk platform.

This ORBIT phase Ia study solicited input from potential web-based learners to inform the development of a future digitally delivered, MHBC gardening intervention. The majority of participants expressed interest in the intervention topic area and indicated that a web-based format using a series of brief videos would be preferable over webinar-based lectures or in-person models. Participants also suggested the inclusion of 6 different intervention functions based on the COM-B framework from Michie et al [32] including education, training, environmental restructuring (eg, social support), persuasion, enablement, and modeling. Below we will consider the results of this study through the lens of COM-B to examine the specifics of how the identified intervention functions may be linked to this model in the context of a gardening intervention.

Capability involves an individual’s knowledge or skills and can be described as the psychological or physical capacity to engage in a particular behavior [32]. Capability is linked to 2 related, but distinct COM-B intervention functions including education (imparting knowledge) and training (achieving skills). It is also in alignment with social cognitive theory which posits that knowledge is an important element for behavior change [50]. Both NG and EG were fairly consistent in suggesting the need for gardening-related knowledge and skills and the topics they identified were broadly consistent with other programs that have offered gardening-based education [18,23,44,45].

The participants identified cooking knowledge and skills as areas of interest and this is encouraging since preparing garden produce is a necessary prerequisite to consuming it. Other investigators have found that both gardening and cooking are independently and positively associated with higher-quality diets [41,51]. In our sample, EG had significantly higher scores than NG on all included measures of cooking agency. To our knowledge, this is the first study to measure cooking skills in a sample of gardeners and nongardeners and it raises the question: “Are gardeners more likely to have strong cooking skills or are people with strong cooking skills more likely to garden?” It seems entirely possible that this relationship is bidirectional. Perhaps, among gardeners, the arrival of fresh garden produce encourages them to cook more leading to better cooking skills. Conversely, perhaps skilled cooks more highly value fresh fruits and vegetables and begin gardening as a way to procure high-quality produce. Either way, our data suggest a potential relationship between gardening and cooking that warrants further exploration in future work.

Opportunity in the COM-B model includes the factors that lie outside of the individual and can encompass both physical and social aspects [32]. In this study, a surprisingly low number of participants (3 NG and 0 EG) suggested that physical enablement, or the need for space, tools, seeds, and other supplies, be included in an intervention. It is possible that both NG and EG already have what they need to engage in gardening. However, for NG, it is equally plausible that they are unaware of the materials needed to engage in gardening. Given that the provision of gardening tools and supplies would add to the cost of delivering an intervention, further exploration is warranted to understand what participants may need to enable gardening activity. In addition, while gardening materials can be supplied in the context of a gardening intervention, helping participants identify outdoor space to garden either in the ground, in containers, or a community garden setting is a critical barrier to overcome.

When considering space limitations, it is tempting to assume that gardeners are less likely live in urban settings. However, in a representative sample of US adults, Kegler et al [15] found that 30% of US adults reported gardening and while the proportion of gardeners was lower in urban settings (28.5%), the difference was not as striking as might be expected (32.9% semiurban, 35.9% rural). One reason that gardening is feasible in urban settings is that a meaningful amount of produce can be grown in a relatively small space. Conk et al [52] found that in a small garden bed (4'×8'), a gardener can produce up to 320 servings of vegetables over a single growing season. This is sufficient to meet 100% of that individual’s USDA-recommended vegetable servings for 4 months [52]. Importantly, while the majority of gardening that occurs in the US is home-based (92%), there are many practical options for outdoor gardening in small spaces and urban settings. This includes growing in pots or containers on apartment balconies, in small front or back yards [53], or by accessing community gardens which are located in all 50 states [54,55], at schools [56], and on medical center campuses [23,57,58].

In this study, both NG and EG identified small-space gardening as a topic of interest. However, it was not clear from their responses how small these spaces might be (ie, a pot of thyme on a windowsill or a network of pots on a balcony). Given that our interest is in the potential for gardening to improve CVD outcomes, future work is needed to understand these comments and to examine if small-space gardening can provide meaningful amounts of fruits and vegetables or physical activity.

Finally, motivation in the COM-B model includes internal processes that energize and direct behavior [32]. This is in alignment with self-determination theory, which posits that activities that garner enjoyment, personal accomplishment, and excitement (eg, internal motivation), are better able to be sustained over long periods of time [59]. A unique aspect of gardening is that although it is both a means to increase household fruit and vegetable availability and physical activity, these are not typically the reasons that gardeners cite for engaging in it. Chalmin-Pui et al [60] surveyed gardeners in the United Kingdom (n=5776), to understand gardeners’ motivations and while participants did refer to the health benefits of gardening, there were many other motivations identified including pleasure, seeing plants grow, an expression of self-identity, relaxation, and environmental sustainability.

In this study, EG commented on the personal satisfaction that arises from gardening by stating “how nice it is” when you “get to pick the vegetables.” They also valued garden produce and were significantly more likely to rate garden produce higher than store-bought produce for characteristics such as cost, flavor, and quality. Taken together, these findings suggest that a future gardening intervention could be positioned to appeal to non-health–related values; an approach that has been called a stealth intervention [61,62]. Stealth interventions are designed to engage people in a target activity (eg, gardening) that indirectly promotes the desired behavior (eg, physical activity and improved diet) by leveraging peoples’ interests and values (eg, pleasure, enjoyment, agency, cost-effectiveness, and sustainability). It seems plausible that a future gardening intervention could achieve positive health outcomes by explicitly targeting internal motivational aspects of enjoyment and personal satisfaction rather than health benefits.

After education and training, the next largest proportion of participant suggestions was related to environmental restructuring of the social environment including social support, group interaction, discussions, and feedback. These comments raise the question as to whether a digitally delivered intervention needs to include in-person components or whether digital person-to-person interactions are adequate. Santarossa et al [63] take up this question in a recent narrative review and suggest that investigators should take a new “Web 2.0” perspective to eHealth intervention development. This perspective considers not just asynchronous delivery models but also person-to-person synchronous intervention components [63]. The authors suggest that high-quality, in-person connections can be facilitated through digital person-to-person components (eg, videoconferencing) and that these interactions are likely to be as effective as in-person social support. Though published prior to the COVID-19 pandemic, the Santarossa et al [63] commentary provides a timely option for integrating person-to-person interactions into a post-pandemic intervention.

This study has limitations and strengths. Previous investigators have noted some concerns about the use of the MTurk platform for survey distribution [64]. In the context of this study, relevant concerns are focused on three main issues including (1) the representativeness of the MTurk worker pool, (2) issues of attentiveness, and (3) the presence of bots. Consistent with ORBIT model phase 1a, this study was not designed to be representative but rather our goal was to solicit input that can provide direction for the development of future interventions. Thus, the results of this study provide useful insights that can be used for intervention development in addition to generating new ideas that can be tested in future studies. A significant strength of our study is that we used previously published strategies to ensure high-quality MTurk samples including the use of written open-ended comments to screen for English language fluency, screening for bot responses taken from other locations on the internet, and removing participants with nonsensical responses [35,36,64]. This, in addition to the inclusion criteria we used (eg, 98% HIT approval rating, requiring a unique completion code, and limiting duplicate responses with the use of the custom qualification through the Amazon platform), provides reassurance that responses from our participants are valid and can be used to draw the conclusions presented in this report.

Finally, our results are based on the acceptability of a hypothetical intervention. It is possible that what participants want in an intervention may not be practically feasible to deliver, may not fit what is known by behavioral scientists to optimize behavior change, or may not confer cardiovascular benefits (ie, small space gardening). In addition, gardening is complex and relies not only on individual behavior change, but is also dependent on the external environment (eg, soil conditions, weather, and pests) that may be outside the control of an individual or what an intervention can address. Future work is needed to develop an intervention that balances what can practically be delivered with fidelity in a digital environment, the recommendations made by the participants in this report, and behavior change strategies known to be effective.

In this ORBIT Phase 1a study, potential web-based learners were interested in a digitally delivered gardening program, they preferred brief videos for content delivery (<5 minutes), and they suggested content topics that encompassed how to garden from planting to harvesting and cooking. Participant comments and health behavior theory support incorporating opportunities for gardening knowledge and skill development while also fostering social support and highlighting the enjoyable and personally satisfying aspects of gardening. The next step in this line of work is to identify target behavior change techniques, develop an intervention delivery strategy that engages these targets, and conduct pilot testing to assess participant acceptability and preliminary efficacy of the newly developed intervention.