The complexity of human behaviors creates a challenge for health educators. An individual’s decision is generally influenced by a network of factors [2]. Disentangling the more important factors in the decision-making process can help educators develop more successful behavior change programs. Conjoint analysis is a technique that can be used to help determine which factors of a particular human experience are most important to a study participant [3].

The value of conjoint analysis in public health education programs was demonstrated several decades ago. Previous studies have used this method to evaluate specific aspects of health programs, to identify what type of program would be most successful for the target population. In 1989, conjoint analysis was used to evaluate attributes of a smoking cessation program [4]. Later, in 2006, this technique was used to evaluate preferences for cost and intervention strategies for diabetes prevention programs [5]. In 2009, conjoint analysis was used to help develop walking programs for older adults, to ascertain the preferred program duration, frequency, incentives and setting, to maximize acceptability and participation [6]. Despite its potential, few recent applications of this method to health promotion program development have been located.

Mindset segmentation takes conjoint analysis a step further by identifying response differences among the population as opposed to treating participants as a single group. It has been suggested that segmentation by individuals’ attitudes and behaviors plays an essential role in health communication and disease prevention efforts, and this has been explored as a general strategy [1] [7] [8]. However, the combination of conjoint and cluster analysis through mindset segmentation appears to be primarily applied to product development.

Mindset segmentation has been recognized as a relatively quick way to describe individuals’ preferences and how they vary among a given population. Recent studies have demonstrated the principle of conjoint analysis and mindset segmentation in food product development. This has included consumer responses to the appeal, preferences, and consumption of meat-free alternatives, rice products, and conventional dairy products [9] [10] [11]. These same principles have extensive potential for creating or marketing health education programs. Educators can use mindsets to inform a wide range of program development features (e.g., cost, setting, delivery mechanism, content) as a way of tailoring interventions to the unique preferences and values of specific segments of target populations. As such, mindset segmentation is a way of integrating the science of consumer behavior with the science of behavior change to magnify the impact of public health initiatives across communities [4].

The purpose of this study is to demonstrate how mindset segmentation can be used to create and/or market health programs. This is achieved through exploration of an example study to determine if willingness to participate in health education surveys varies with different consumer mindsets.

Questionnaires and surveys are a useful tool in health research and commonly used to inform program development and evaluation. A survey application can help assess customer expectation and satisfaction, justify need for program and/or budget (accountability), measure behavior change or intent to change behavior. Accuracy of results generated from questionnaires is often limited by quantity or quality of responses. Finding a way to improve the survey experience is one way to overcome these limitations.

To address this challenge, a study was developed using mindset segmentation to determine how mindsets among the public can be used to improve survey participation. This type of information can be used to inform health educators on how to appeal to the primary mindsets, improve participation in the survey process, and ultimately use this data to increase the impact and reach of their programs.

This study utilized the Mind Genomics^® (MG) system, a cognitive science that applies a combination of conjoint and cluster analysis to identify patterns of thoughts (mindsets) within a population [12]. This system uses responses to a set of specifically constructed survey elements to address the overarching question, what factors motivate decision making?, as a means of highlighting the factors that can help predict behaviors.

The MG survey design can produce meaningful statistics with a relatively small number of participants (approximately 50 - 100). The technique categorizes a population into subsets, based on the importance ascribed to the survey elements. The BimiLeap (https://www.bimileap.com) online software application was used in this study to facilitate the segmentation of primary mindsets within our study population. This program handles inputting of the MG survey design and calculates basic statistics for segmentation analysis.

The steps involved in designing the study and survey questions are consistent regardless of the topic; they are outlined below [12].

1) Clearly state the purpose of the study.

2) Formulate the topic/question of interest and the rating scale.

3) Identify four main attributes that describe the topic of interest.

4) Describe distinct elements within the main attributes.

5) Distribute to target audience.

A clear purpose statement is fundamental to formulating the overarching question for survey design. The goal of this study was to identify motivations for survey participation i.e., what motivates people to participate in the survey process (step 1). By understanding the motivations, community health educators can design surveys to better engage their target audience.

Conjoint analysis involves asking a single question multiple times while varying key features and asking the participant to assign a rating to each variation. In this study, the root question of interest was: “How likely are you to take this questionnaire?” with a rating of 1 = NOT very likely to 5 = VERY likely (step 2). This technique is demonstrated in Figure 1.

For conjoint analysis to be effective in predicting human behavior, the study must include attributes relevant to the human experience and the decision making process (step 3). This involves identifying overarching themes that impact the question of interest. For this study, a literature search was conducted using the following key words: survey, participation, willingness, determinants, perspective. Informal interviews among colleagues were also conducted to create a list of survey qualities that are salient to those debating survey participation. An iterative process of discussion, prioritization and comparison to literature was used to reduce the list to four specific topics or “silos” that are appropriate across a wide range of questionnaire types and styles. The silos, as labeled in Table 1, are referred to as Question A, B, C, and D.

The fourth step in study design is to identify distinct elements within each silo, to capture a range of scenarios that reflect the real-world experience. Elements are stand-alone phrases that elicit an emotion or feeling while “painting a word picture” [13] for the participant. Ultimately, participants are asked to rank these elements; the more succinct but descriptive the element, the more constructive the response. The final list of elements developed for this study are presented under each silo in Table 1, labeled as A1, A2, … D4.

The BimiLeap software captures basic demographic information such as age and gender, and provides an option for the researcher to include an additional, preliminary multiple-choice question and open-ended question to help further identify the audience. Once the foundational design is complete, the software creates a series of distinct permutations of the design, known as “vignettes”, by combining 2 to 4 elements from different silos [13]. A total of 24 vignettes are presented to each participant, who rates them using the rating scale identified in step 2. This design ensures that all elements are equally represented while no two participants see the same combination of questions. Examples of vignettes that a participant would have encountered as part of this study are shown in Figure 1.

Once the survey is ready for distribution, the software generates a link which allows the researcher to directly source participants and deliver the survey via email, social media, or other forms of outreach. Local participants were recruited for this study via the Nextdoor community social platform and by e-mail to previous community program participants. The Nextdoor filters were used to target local participants, preferred for this study to more accurately inform community educators in the immediate region.

A participation cap of 150 was set as part of the design. This number exceeds