Rural adults’ perceptions of nutrition recommendations for cancer prevention: Contradictory and conflicting messages

Abstract

Despite robust evidence linking alcohol, processed meat, and red meat to colorectal cancer (CRC), public awareness of nutrition recommendations for CRC prevention is low. Marginalized populations, including those in rural areas, experience high CRC burden and may benefit from culturally tailored health information technologies. This study explored perceptions of web-based health messages iteratively in focus groups and interviews with 48 adults as part of a CRC prevention intervention. We analyzed transcripts for message perceptions and identified three main themes with subthemes: (1) Contradictory recommendations, between the intervention’s nutrition risk messages and recommendations for other health conditions, from other sources, or based on cultural or personal diets; (2) reactions to nutrition risk messages, ranging from aversion (e.g. “avoid alcohol” considered “preachy”) to appreciation, with suggestions for improving messages; and (3) information gaps. We discuss these themes, translational impact, and considerations for future research and communication strategies for delivering web-based cancer prevention messages.

Trial registration

ClinicalTrials.gov identifier: NCT04192071..

KEYWORDS:

• Colorectal cancer (CRC) prevention
• Nutrition risk factors
• Digital health interventions
• Risk messages
• Rural adults

Introduction

Robust scientific evidence links dietary intake to several cancers, including breast, colorectal, and prostate cancers. However, large segments of the U.S. adult population do not know that alcohol, red meat, and processed meats can increase cancer risk [1]. For instance, 1 in 8 breast cancer cases in women can be attributed to alcohol intake [2] and the risk of colorectal cancer (CRC) rises by 7% for each additional 10 grams a day of ethanol intake (which is the carcinogenic substance in alcoholic beverages) [3], yet most U.S adults (nearly 60%) are not aware that alcohol can increase cancer risk [4,5]. Similarly, only 43% and 53% of people are aware that diets high in red and processed meat, respectively, increase cancer risk [6]. Knowledge gaps about the links between nutrition and cancer present a challenge to promoting prevention measures that support population health. These knowledge gaps may be particularly felt among geographically rural populations, who have unique risk profiles and barriers to cancer screening such as longer distances to health care services. Rural populations’ higher cancer burden compared to non-rural populations, in terms of cancer incidence, late-stage diagnosis, and mortality, as well as lower dietary quality, make this a particularly important population to engage [7,8].

Communicating the nutrition – cancer link

In 2020, an expert panel convened by the U.S. National Cancer Institute proposed increasing focus on communication strategies along the cancer continuum as a strategy to increase awareness of the alcohol – cancer link [9]. However, the complexity of accomplishing a seemingly simple communication task should not be ignored. For one, nutrition recommendations may vary depending on the health goal or disease under consideration, thus disease context is important. For example, some cancer prevention recommendations may differ from recommendations promoting heart health. Second, message recipients are tasked with making health decisions in the context of multiple health conditions and comorbidities. Finally, cultural norms can affect people’s engagement with health messages [10].

Disease-specific nutrition communication

For cancer prevention, it is recommended to avoid alcohol [11]. However, for overall health, the USDA Dietary Guidelines for Americans say that adults who choose to drink alcohol should limit their intake to two drinks or less per day for men and one drink or less per day for women [12]. And despite the lack of formal recommendations on alcohol consumption for heart health, the scientific evidence linking light wine consumption to cardioprotective effects may shape the public’s belief that alcohol use, in general, promotes cardiovascular health [13–15]. Similarly, there is scientific evidence that some dietary patterns (e.g. ketogenic diets with high meat intake) show efficacy for the management of obesity and type 2 diabetes, a message that directly conflicts with nutrition recommendations for CRC prevention (i.e. limiting red and processed meat consumption).

Source-specific nutrition communication

People’s sources of information may also affect overall cancer prevention behaviors. Adults obtain health information from many sources (e.g. physicians, the internet, media, friends/family members, and coworkers), and the internet is used widely and frequently as the first source of health information [16]. Healthcare professionals are a trusted source of health information for many, yet they have a range of knowledge and self-efficacy regarding counseling patients on alcohol or nutrition recommendations [17,18]. For example, while the American Heart Association advises patients to discuss alcohol use with their physician [19], few adult cancer survivors recall receiving information about alcohol use from a healthcare provider [20].

Contradictory information, decisional conflict, and nutrition confusion

Obtaining health information from multiple sources or trying to incorporate information related to co-morbidities may lead to encounters with contradictory information (also called “conflicting” information [21,22]). Carpenter et al. [23] define contradictory information as two-sided coverage where “two or more health-related propositions are logically inconsistent with one another” (p. 3). Contradictory information is often related to a single health issue; a more expansive conceptualization offered by Nagler [22] considers contradictions between multiple health issues, such as when messages indicate that a single behavior can produce two different outcomes This may lead to uncertainty about what behavior to choose, negatively affect the credibility of the message source, and reduce behavior change intentions [24,25]. For example, conflicting recommendations about mammography decreased the uptake of breast cancer screening and future intentions to screen [26]. Understanding how users perceive contradictory information can inform the effective delivery of health messages.

When it comes to nutrition communication, confusion is not new. The 1980 Dietary Guidelines’ advice to limit butter prompted a shift towards margarine, yet was followed in the 2000s by another shift away from margarine because of trans fats [27]. Nutrition confusion is equally easily found in the context of complex findings around alcohol use and health. It becomes unclear how best to manage health when balancing potential cancer outcomes (e.g. heavy drinking related to cancer mortality) vs. heart health (e.g. some positive effects of light to moderate alcohol use for cardiovascular outcomes) [28–30].

Although the evidence for the impact of nutrition confusion is mixed, nutrition confusion is not without consequences. Both nutrition confusion (e.g. perceived ambiguity about nutrition information) and related nutrition backlash (e.g. negative feelings toward nutrition recommendations and research) are negatively associated with intentions to follow nutrition recommendations [22,31,32]. Yet other findings indicate that contradictory dietary information does not affect self-efficacy or motivation to make changes to dietary intake [33]. Thus, exploring perceptions of nutrition recommendations remains important for researchers working to raise awareness of the nutrition – cancer link. Furthermore, the importance of taking a translational approach to understand how the nuances involved in the way information presented may impact perceptions is paramount. As Krieger and colleagues [34] indicate “communication has a long tradition of generating theory, rigorously testing theoretically-informed ideas, and then translating findings into information that can be used to solve practical problems” (p. 33).

Study purpose and research question

The purpose of this study was to explore how nutrition risk factors for CRC prevention were perceived by rural adults participating in the brief intervention and message design stage of a larger study aiming to promote CRC screening. The brief intervention focuses solely on activating nutrition-based risk perceptions as we were interested in providing specific, clear, and actionable messages about one type of behavior (avoid/reduce consumption) to activate cancer risk perceptions and promote a primary type of protective behavior (CRC screening). This study contributes to understanding how nutrition risk messages in the context of colorectal cancer are presented and perceived. We highlight how to plan for context when delivering messages – in this case, potentially contradictory messages. Given the low awareness of the nutrition – cancer link among the general public and the fact that nutrition guidelines vary by health condition, there is a need to understand how message recipients perceive and respond to context-specific nutrition messages. Our study explored the following research question: What are rural adults’ perceptions of nutrition risk messages for CRC prevention delivered by a virtual health assistant (VHA)?

Methods

This analysis was part of a larger study that engaged rural adults in a community-engaged process to design and develop a web-based nutrition risk module centered around CRC prevention. The goal of the overall module was to communicate information about nutritional risks related to CRC and to promote CRC screening options, including a home-based stool test. During the development of this risk module, researchers focused largely on understanding cultural expectations and preferences related to communication surrounding food and cancer prevention. Researchers also sought feedback and suggestions from users regarding how to present nutrition risk information to be understandable, acceptable, and actionable (ClinicalTrials.gov Identifier: NCT04192071). During this initial process, insights related to contradictory information and message resistance among users testing the module prompted us to conduct an exploratory analysis of how users were responding to CRC avoidance guidelines and how those guidelines may conflict and compete with other sources of information.

Recruitment

Participants were recruited through several approaches. We began by using the University of Florida’s HealthStreet community engagement program to identify eligible participants; most of those recruited in this way participated in the first two focus groups, held in-person at the start of the study period. We used another university resource, the Integrated Data Repository, to identify and contact potentially eligible participants who had previously asked to be notified of relevant research studies. Flyers were placed in the surrounding community, but because the research protocol was moved from in-person to remote (e.g. Zoom) data collection in March 2020 due to the COVID-19 pandemic, flyers were not the main recruitment strategy. For inclusion in the study, participants had to be between 45 and 73 years old, be able to read or write in English, and self-identify as Black/African American or White. After the switch to remote data collection, these initial inclusion criteria were expanded to include having access to virtual video conferencing software (Zoom). A member of the research team contacted eligible participants by phone and/or email using standardized messages, approved by the institutional review board (IRB), to assess eligibility and schedule data collection.

Data collection and script development

All protocols were approved by an IRB and a Safety Review and Monitoring Committee. Data collection occurred between January 2020 and November 2021. Participants provided informed consent before participating. Two focus groups (in-person) were followed by several online focus groups, and then a series of online individual think-aloud interviews [35] (Figure 1). Each session lasted 60–90 min and was audio-recorded. During focus groups 1 and 2, participants completed photo sorting and storyboard activities to brainstorm preferred images and messages. Participants then viewed a brief clip of a VHA describing how to screen for CRC (created for the parent study), then participated in a moderated discussion. All subsequent participants interacted with the current version of the VHA prototype then participated in a moderated discussion of their experience and completed a questionnaire [35].

Figure 1. The virtual health assistant prototype. Can be customized to user preferences on features like race, gender, voice and appearance.

Moderator notes facilitated an iterative process of updating the VHA scripted content, including nutrition messages (Appendix). Informed by Construal Level Theory (CLT), the script was primarily developed with considerations for temporal framing (e.g. short-term vs. long-term or day vs. year). However, the multiple messages delivered in the approximately 12-minute script employed several message features informed by existing literature (gain vs. loss frame, repetition, intention formation prompt, and optional message branches). Based on CLT, and existing research, we conceptualized that among adults consuming any of the nutrition risk foods discussed by the VHA, messages about dietary patterns linked to cancer would invoke the immediacy of colorectal cancer risk (proximal threat), making individuals less likely to discount a theoretical future cancer risk. A meta-analysis of the persuasive impact of temporal frames found that present-oriented health messages were significantly more persuasive than future-oriented health messages, showing a positive effect on intentions [36].

Data analysis

Transcripts were managed using NVivo version 12. An initial transcript review prompted an exploratory analysis of comments about the intervention’s nutrition messages. Three members of the research team drafted a codebook to systematically explore this data. Based on transcript review and existing literature on competing and contradictory nutrition messages, we created operational and conceptual definitions and updated these via iterative coding and review of inter-rater reliability. The final codebook included themes related to perceptions of the nutrition messages. All transcripts were coded by one or more of the three-member coding team and reviewed weekly until all transcripts were coded. When necessary, subthemes were defined and previously coded data from the parent theme was re-coded.

Results

Participants

Overall, 48 men and women between 50 and 73 years old participated in a one-time focus group or think-aloud interview. Participants self-identified as non-Hispanic Black (65%) or non-Hispanic White (35%) in addition to self-identifying as men or women and participated in race and gender-stratified focus groups or were asked to participate in the individual think-aloud interviews. All of them self-reported having health insurance. Participant characteristics are presented in Table 1.

Table 1. Participant demographics.

Variable	n	%
Gender
Men	24	50%
Women	24	50%
Race
Black	31	65%
White	17	35%
Primary language
English	47	100%
Marital status
Single	12	25%
Married	23	48%
Divorced/separated	9	19%
Widowed	3	6%
Missing data	1	2%
Education completed
Grade 1–8	1	2%
High school graduate, technical college, some college	24	50%
College	11	23%
Postgraduate training	11	23%
Missing data	1	2%
Household Income
<$11,000	7	14%
$11,000–$30,000	9	19%
$30,001–$75,000	14	29%
$75,001+	10	21%
Missing data/prefer not to answer	8	17%
Health insurance coverage
Yes	48	100%
No	0	0%
Total	48	100

Inter-rater reliability

We conducted an iterative process to establish acceptable coding schemes to reduce coding errors. Establishing acceptable inter-rater reliability consisted of multiple rounds of coding, which included reviewing coding discrepancies on 20% of transcripts and pulling out codes for negative case analysis (codes that indicate the opposite of the theme as a process of refining and confirming the patterns emerging from analysis; see [37]). All themes had a percent agreement higher than 97% and a weighted kappa score greater than .75, indicating sufficient agreement between coders [38]. Percent agreement and Cohen’s kappa statistic scores indicate high agreement between coders and that the codebook was functioning well as a tool to aid data coding (Table 2).

Table 2. Inter-rater reliability of all themes coded.

Theme	Kappa	% Agreement
Contradictory recommendations	1.00	100.00
Reactions to nutrition messages	0.81	99.14
Information gaps	0.78	97.67

Overview

Three themes were coded about how participants perceived the tailored messages on nutrition risk and CRC prevention: contradictory recommendations, reactions to nutrition messages, and information gaps. These themes and related subthemes are defined in Table 3. Examples of the subthemes are detailed in Table 4.

Table 3. Themes, subthemes, and definitions.

Theme	Definition	Subtheme
Contradictory recommendations	Users perceive differences between colorectal cancer nutrition guidelines and nutrition information related to other health needs or conditions.	Contradicting health conditions. Nutrition recommendations for one health condition (e.g. heart disease) may contradict recommendations for a different condition (e.g. colorectal cancer).
		Contradicting sources. One source (e.g. a health care provider) provides nutrition information that is different from the information provided by a different source (e.g. a trusted relative).
		Contradicting cultural or dietary preferences. Nutrition recommendations contradict cultural ways of being and/or cultural norms or dietary intake preferences.
Reactions to nutrition messages	Comments show discomfort with messages prompting avoidance of certain behaviors (drinking alcohol, eating red/processed meat) or alternatively endorsing this framing.	Aversion to prohibitive recommendations. Reactance to cancer prevention nutrition messages prohibiting intake of specific foods or beverages, regarding potential negative impacts on credibility and/or behavior.
		Appreciation for guidance. Stated preferences for avoidance messages and/or expressed appreciation for a message that clearly said what behavior they should take (e.g. avoid alcohol, avoid processed meat, limit red meat) for colorectal cancer prevention.
		Desire for a balance of risk messages and positive messages. Stated preferences for positive messaging (e.g. risk-reduction or health promotion vs. avoidance) for cancer prevention or suggestions to insert positive messages.
Information gaps	Comments on missing information or information needed to understand colorectal cancer – specific risks, understanding nutritional health, or engaging in promoted behaviors.	Gaps in nutrition information. Requests for more nutrition information beyond the three colorectal cancer risk factors discussed by the virtual agent.
		Gaps in statistics and numbers. Requests for statistics related to colorectal cancer risk.
		Gaps in screening information. Requests for more information on colorectal cancer screening, risks, and modalities.

Table 4. Expansion of major themes and subthemes.

Theme	Subtheme	Exemplar
Contradictory recommendations	Contradicting health conditions	“I have different health issues … I’m not supposed to do a lot of seafood because I had breast cancer and I had chemo and radiation. I’m not supposed to eat a lot of things that I can’t peel, whether they’re fruits or vegetables. And then I may have a flare-up with that. I have lymphedema. So, there’s other things that I can’t eat that are really salty. My diet is limited to a lot of different things, so I don’t have many choices. So, like they [my doctors] said, eat the lunch meat, but don’t get it too salty.” (P2118, TA1, BF, prototype 3) “Just thinking about the alcohol, the other health issues are saying, like a glass of wine or beer a day is like good for like cardiac issues or something. So, it’s like. What do you want to focus on improving?” (P9916, WF, TA7, Prototype 3)
	Contradicting sources	“I love to eat liver because they say that’s high in iron, so I’m like, how could they say liver is high in iron and is red meat and you can get colon cancer from red meat?” … [This] made me question my, my inner self of eating liver now because it’s red, and I don’t want to get colon cancer. But I know my doctor says liver is good for you to eat, so I want to learn more about the liver. Why do they say that liver is red meat and why is it bad for us to eat it because of the colon, and why do they say it’s good for us to eat because it’s high in iron? (P2948, BF, FG12, prototype 2)
	Contradicting cultural or dietary preferences	“You know, I didn’t see the point about the wine being a problem, but I don’t see it as a problem. It’s biblical, Jesus drinks wine and it’s, it’s just, if you don’t do it in excess, then it’s supposed to be good for you. It’s good for your blood circulation. So, I don’t see it really as a problem, I would have to really research that and find out for myself is it, that doesn’t sound like correct, like correct information … You know, I’m probably wrong, but not as far as biblical, I’m, I’m not wrong.” (P7930, WW, FG3, Prototype 1)
Reactions to nutrition messages	Aversion	“It came across as, if you want to avoid colorectal cancer, you better not ever have another drink in your life. And you probably should never have another piece of red meat or processed meat. It seemed very one or the other … for people that do enjoy a cocktail from time to time, I think it may be a little off-putting that they’re gonna say, wow. Not me. I’m not going to go get screened. If one drink is going to do for me, I don’t want to know.” (P5533, WM, FG7, prototype 1) “When I heard him say don’t drink it at all … You’ve got to enjoy life. And yes, you can drink it. Drinking wine has nothing to do with getting colorectal cancer. Drinking a large amount of wine increases the risk, but it can’t give you colorectal cancer. It’s like saying don’t go outside because something might fall on your head. It’s extreme advice given that people have to enjoy their lives and yet try to live as long as they can.” (P3963, WM, TA5, prototype 3)
	Appreciation	“[limit red meat to] three servings a week? Hmm. Okay. Well, that’s nice to know. Thanks, Alex, that’s good information. That’s nice and encouraging. Four ounces of meat, huh. A deck of cards. Okay.” (TA13, BM, Prototype 4)
	Balance with positive	“Me personally, anything for my health. I don’t care who I talk to as long as it’s positive. It doesn’t bother me. I’m used to stuff like that. I don’t really care as long as it’s benefiting me because nobody can judge me.” (P464, WM, TA2, prototype 3)
Information gaps	Nutrition information	“I thought there would be some kind of, … pathway to resources for a dietary sample of a meal plan or something, what things you could [eat]. We know what things we couldn’t [eat], shouldn’t [eat], but it wasn’t explained, ‘eat more vegetables, have more fiber,’ … that type of thing, I thought would have been included.” (P9163, BF, FG13, prototype 2) “There are things that I might look it up on my own now that the issue has been raised or the question has been raised. Like, I didn’t know about pork. And again, I wondered, you know about, we don’t eat it, that much, but I wonder, you know if there was a way to eat it without the danger of it being contaminated by, or preserved, you know. Obviously, ham has a lot of salt in it, but I’m aware of preservatives.” (P6736, WF, TA9, prototype 3)
	Statistics and numbers	“Well, from my perspective, it might be interesting to see something along the lines of the relation between red meat or alcohol and the number of cases. Whether it’s a line graph or a bar graph, or something of that nature, just showing the relationship or interrelationship between the foods, the alcohol, and the number of cases that are diagnosed.” (P5533, WM, FG7, prototype 1)
	Screening information	“What I, what I think is missing is showing where colon cancer can start … is it the large intestine, the colon is at the small intestine? Might be useful to show where they go with a colonoscopy.” (P5533, WM, FG7, prototype 1) “So, I’m asking, oh, how often should I discuss this [screening] with my doctor?” (P7466, WM, TA14, prototype 4) “I would mention that there’s a change [in the screening age] because you’re more likely to get the person to listen to what you’re saying.” (P7466, WM, TA14, prototype 4)

Theme 1: Contradictory nutrition recommendations

Participants commented when they perceived that a specific recommendation made by the VHA contradicted a recommendation they had heard previously or had previous knowledge of. This general theme of contradictory information manifested itself in three distinct subthemes: contradicting health conditions, contradicting sources, and contradicting cultural or dietary preferences.

Contradicting health conditions

The first subtheme included comments referencing contradictory nutritional guidelines for different health conditions. Here, participants discussed the differences between nutrition recommendations for CRC prevention compared to nutrition recommendations for other health conditions. For example, “I’m aware that you have to be careful, but there’s a lot of reading that says a beer is good for the heart” (P651, BF, FG14, Prototype 2). Another participant mentioned, “Especially for women who are still having their menstrual cycles. They need heme iron, they need meat” (P8556, WW, FG4, Prototype 1).

When the recommendations for a health condition other than CRC conflicted with messages to avoid alcohol and processed meat and limit red meat, participants asked for clarification on how the recommendations fit into a broader context of health behaviors. One participant, P2948, described a conflict between red meat consumption in two health concerns (low iron and CRC) and identified her doctor as a message source. She specified that her doctor’s advice that liver consumption is good for low iron contradicted the VHA’s message to avoid red meat for CRC prevention. This overlap with conflicting health concerns leads to our next identified theme: contradicting sources.

Contradicting sources

A second way this theme manifested was in participants expressing uncertainty about managing several information sources providing contradictory nutritional information. This included instances when participants explicitly stated the source of the contradictory information. This source was often a clinician, who gave nutritional information that conflicted with the cancer prevention message delivered by the VHA.

I totally disagree with that because I’m aware of a doctor – Doctor [name], I don’t know if you ever heard of him or not, but … some of that information that the avatar was saying was not accurate … I don’t agree with it because I know better. (BM, TA10, Prototype 4)

Participants also expressed caution and savvy when encountering multiple information sources. One participant described searching for multiple sources and deferring to the most reputable source:

Before you fall in behind [an online] article, you need to confirm it and see it in more than one written form or by somebody who’s very reputable. You can’t just believe what you’re reading when you go on the computer and it’s there. (P3267, BW, TA3, Prototype 3)

Contradicting cultural or dietary preferences

Finally, participants discussed how dietary patterns linked to cultural or normative ways of eating and drinking conflicted with cancer prevention recommendations. These instances occurred most often in the context of the message that “to prevent CRC it is best to avoid alcohol.” One participant commented, “Part of it is because I’m French, so I was raised on wine from a small child down, so I was a little surprised that drinking any wine, you know, would increase my chances of colorectal cancer” (P7930, WW, FG3, Prototype 1).

There was also concern that incorporating new dietary restrictions would be difficult, especially if they conflicted with current dietary intake patterns. Dietary modifications to meet the recommended guidelines were sometimes discussed as unattainable or in direct conflict with personal needs, like this participant’s description of always being hungry:

I have no choice actually, these people that want to cut out meat in our diets, they’ll kill me literally … because protein from plants, doesn’t be my nutrient. It doesn’t fill me up and I need to eat a lot of vegetables, but the eggs and the beef fill me up. Like I eat chicken, even chicken doesn’t fill me up. (P2729, WW, FG3, Prototype 1)

While some participants questioned the feasibility and compatibility with existing consumption patterns, subsequent themes include comments that manifested as stronger emotional reactions to the messages presented by the VHA.

Theme 2: Reactions to nutrition messages

Participants had a diverse spectrum of reactions to the VHA’s nutrition risk messages, including aversion to prohibitive recommendations, appreciation, and desire for balance in the framing.

Aversion to prohibitive recommendations

In some cases, participants seemed to experience discomfort or react negatively when the VHA suggests avoiding alcoholic beverages to reduce the risk of CRC. Although the most pronounced reactions were in response to the “avoid alcohol” message, participants also indicated that the “limit red meat” and “avoid processed meat” recommendations were not what they were expecting. These recommendations to avoid or limit specific food/beverage items were described as “preachy” or “unrealistic.” In response to being told that the amount of alcohol they consumed was more than the amount recommended for CRC prevention, one participant responded:

The assistant seems rather harsh … when you go, “You shouldn’t drink at all, and you shouldn’t have any red meat” … I’m really not a big beer drinker. And it’s when I play tennis I have a beer, and now you’re making me feel guilty about this. I’m going to get colon cancer because I had one beer [a] week socializing with people? That’s a little bit harsh. Especially when later on … the assistant goes on to say, “Well, we really don’t know how many beers it takes for you to get cancer.” (P464, WM, TA2, prototype 3)

Participants also reacted negatively to the tailored message that for CRC prevention, it’s best to avoid processed meat and limit red meat: “If I’m not dead by now, then hey, I’m going to keep eating pork. More power to it” (P7896, BM, FG10, Prototype 2).

Participants expressed their discomfort in various ways, including shutting themselves off from new information. In one instance, when the moderator asked if any additional foods should be discussed in the intervention, one participant responded: “No. I don’t want to learn about any more foods that I like, and you want to remove them out of my life. I don’t want to know about it” (P7896, BM, FG10, Prototype 2). Additionally, some participants communicated doubts about message validity and negative impact on personal motivation to act on the advice:

To tell people you can’t consume any red meat or any processed meat, or you’re at high risk for cancer, for me, I would just discount [the advice]. If you’re telling me I have to eliminate it, okay, well I’ll just accept the fact that I’m a candidate or at risk for colon cancer. If there are grades of recommendations … it would be considered a more valid measure than just … “You can’t eat red meat, you can’t have processed foods, or you’re just at high risk for colon cancer.” (P5520, WM, FG8, Prototype 1)

Appreciation for guidance

Receiving the messages with skepticism, however, was not a universal reaction. Some participants expressed appreciation for a clear and consistent message that indicated the precise behavior they should take: “She complimented me on … not doing beef … but the alcohol she did come down. What I like is she didn’t play. I’m an absolute person. Okay so, she was definite that the alcohol needed to be stopped” (P651, BW, FG14, Prototype 2).

Another participant mentioned feeling okay, even positive, at the message to reduce alcohol:

Hmm. Now, that’s interesting, and then numbers that [the virtual agent] was giving me, when it came to the meat, I didn’t care what the numbers were, and what I’m supposed to eat. But when it came to the alcohol, and he said, reduce your level of drinks by one a week, that will help. There was some happiness in that, which is kind of odd. But I saw that more positive than reduce your meat intake. I don’t get it, but that’s where we’re at. (P2474, BM, TA13, Prototype 4)

Additionally, given that the intervention’s nutrition recommendations focused on risk factors and specific foods to limit (rather than other dietary recommendations) some participants suggested that the messages could be more balanced.

The desire for a balance of risk messages and positive messages

Participants also suggested advising people to lessen, rather than eliminate items: “And maybe he could say, hey, what you know … try to reduce your alcohol consumption by one drink … without getting too prescriptive” (P8819, WM, FG9, Prototype 1). Perhaps related to participants’ sense that the guidance to eliminate alcohol, red meat, and processed meat was “extreme advice” (P3963, WM, TA5, Prototype 3), a recurring sentiment included wanting the VHA to give positive feedback. “We always want to hear positive things, you know … we always want positive reinforcement” (P7896, BM, FG10, Prototype 2). The fact that participants described the positive talk as “makes you feel good, right?” (P6000, BM, FG10, Prototype 2) suggests that positive reinforcement may help mitigate perceptions that the intervention’s nutrition messages are too negative or restrictive. “If somebody’s drinking one drink per week, they should be – you should be lauded” (P464, WM, TA2, prototype 3).

Theme 3: Information gaps

The overall theme of information gaps manifested in participants’ requests for additional information, specifically information not featured during their brief interaction with the VHA. Requests for additional information were grouped into three smaller groups by topic, which included requests for additional nutrition information, statistics and numbers, and screening information. (Although not all requests were incorporated into later iterations of the VHA intervention, the parent study’s research team did supplement the VHA-delivered intervention with optional access to a cancer prevention website created based on this feedback.)

Gaps in nutrition information

Participants expressed a desire for more information about the three nutrition risk factors for CRC introduced by the VHA. Specifically, when the VHA said that consumption of red meat, processed meat, and alcohol increased CRC risk, participants wanted details about variation in risk levels and “how much is too much” (P2, FG1, BF, pre-prototype).

How much less … less meat, less – I know it should be less – red meat, less alcohol as much is possible. But if you … if you choose to continue to eat red meat and drink alcohol, how much will be the lowest that would keep you … Yeah, well, lower the risk, the chance of getting cancer? (P9, FG1, BF, pre-prototype)

Some participants had questions about how food production methods (e.g. organic vs. non-organic) might affect red and processed meats’ role in CRC risk. “When she was talking about beef, I was wondering if there were kinds of beef that have … less preservatives … , like organic meat or grain-fed meat” (P6736, TA9, WF, Prototype 3).

Participants also requested information about foods and drinks they could consume in addition to foods and beverages they should avoid. These requests included suggestions for protective dietary items that reduce CRC risk, like fiber, fruit, vegetables, and water. For instance, one participant said,

The whole app seems to be focused on the bad stuff for you. What about the good stuff? Like you’re leaving out the fiber, you know, broccoli and foods of that nature that can be good and can counteract some of the issues. (P5533, FG7, WM, Prototype 1)

Furthermore, participants suggested including dietary recommendations to help replace red and processed meats. Participants said these recommendations could include alternative food items and specific guidance on preparing meals (e.g. meal plans, recipes).

Gaps in statistics and numbers

Some participants wanted to see statistical data describing the relationship between CRC risk and specific population demographics (e.g. gender, race). Participants also asked for visual representations of risk and graphs of the relationship between dietary intake and the number of diagnosed cancer cases. Details about how CRC risk might change at different levels of consumption and about dose-response relationships between consumption and incidence prompted questions about how much alcohol, red meat, and processed meat they could consume without increasing CRC risk. Participants indicated that statistics could help personalize the intervention: “It’s too general on the risk, you’re going to ask specifics of how often [you eat these items], I would suggest that you provide information on what is the increased risk for what I have provided” (P5520, WM, FG8, Prototype 1).

Gaps in screening information

Participants also expressed that the early scripts were too focused on nutrition to the exclusion of CRC screening information. Frequently requested information included the recommended screening age, the accuracy of various screening modalities, as well as guidance on how to choose a screening modality. “There’s a lot of [tests] floating around, and somebody could end up off in the weeds when they really should be on the highway to getting some help” (P6087, WM, FG6, Prototype 1). Visual representations of the colon were also discussed. In response to being asked if an image of a colon should be included in the intervention, one participant described how the use of visuals had facilitated their understanding during a recent healthcare visit. “Whenever the doctor would show us where the cancer was and explain to us … it helps, it helped [my dad] to have some idea of exactly what the doctor was talking about” (P8161, WW, FG4, Prototype 1). Although images were suggested, participants cautioned against elaborate visual aids and suggested using simple depictions in the intervention.

Discussion

This exploratory paper highlights considerations for delivering accessible and actionable cancer prevention messages via a novel message source (i.e. a virtual human/VHA). Rural adults shared their thoughts during real-time message testing of cancer risk messages. We describe three themes, and related subthemes, derived from a qualitative analysis of perceptions related to the nutrition – cancer link: (1) contradictory nutrition recommendations, (2) reactions to nutrition messages, and (3) information gaps. Supplementing existing research on contradictory nutrition messages [22,31,39], which includes randomized designs and quantitative assessments, the current study takes a slightly different approach. Here (1) messages were co-developed with the priority population, (2) message perceptions are assessed iteratively, and (3) we use qualitative, in-depth analysis of narratively described perceptions (vs. Likert scales).

Theme 1: Contradictory nutrition recommendations

Given that nutrition recommendations vary by disease context, it is not altogether unexpected that participants would point out instances when the nutrition guidelines for cancer prevention were contradictory. In this study, we captured three domains within which the VHA-delivered messages contradicted other information: contradictions with other health conditions or concerns, contradictions between information sources, and contradictions with dietary preferences or cultural norms. In context with existing literature, we discuss implications for how contradictions in risk communication for behavior change may affect message recipients’ reactions (e.g. confusion) and perceptions of the credibility of the message.

Contradictions in risk communication: Confusion, negative beliefs, reactance

Participants identified several ways nutrition messages were contradictory or conflicting. Research indicates that exposure to conflicting nutritional information is linked with increased nutrition confusion (e.g. confusion about what foods to eat) and nutrition backlash (e.g. negative beliefs about nutrition research) [31,40] as well as reduced intentions to engage in healthy behaviors [22]. For example, in the current study, some participants questioned if they should consume red meat, processed meat, and alcohol, despite having received the personally tailored, evidence-based nutrition guidelines.

Other studies have also documented nutrition confusion and backlash as negatively associated with cancer prevention behaviors like physical activity and dietary intake; however, the message source may be important. One study found that exposure to contradictory nutrition information via health websites or interpersonal communication produced lower backlash, while contradictory nutrition information received from healthcare professionals was associated with higher backlash [32]. VHAs, as a novel source of information, are a promising means of providing consistent information to influence perceptions. However future research is needed to evaluate user responses (e.g. like backlash) to information delivered by VHA’s.

Psychological reactance (i.e. engaging in an admonished behavior or resisting an advocated behavior) as a means of restoring a sense of freedom and choice is another possible response to conflicting messages or contradictory information [41]. Although we did not explicitly measure reactance in the current study, participant questions such as “How much meat can I eat and still be okay?” may signify an attempt to manage the threat of being told to avoid or limit certain foods. This insight is drawn in part from research findings that participants who receive threatening health messages may be motivated to manage the potential threat to freedom [42]. Authors de Lemus et al. [43] provide additional evidence that reactance can result from perceived threats to one’s groups, social identities, or cultural identities.

Contradictions in risk communication: Credibility

In the context of this project, establishing high message credibility was a key factor affecting motivation to implement the recommended behaviors. Source perceptions (e.g. how a message recipient perceives the source or communicator of the information) can shape opinions on credibility and trust as well as whether or not a recipient ultimately acts on a message [44]. Some participants in the current study expressed skepticism about the information’s accuracy and credibility when told to avoid specific foods. Similarly, past research [45] has found that perceptions of nutrition message accuracy increased perceptions of message credibility. Still, regarding conflicting health information, a longitudinal study found neither levels of trust nor literacy protected against carryover effects [46]. Although previous research has found that VHAs can be a credible source of health information, credibility can change with context and should always be assessed [47–49]. Thus, any intervention using nutrition recommendations to activate risk perceptions should be tested to ensure the messages are perceived as accurate and credible and that overall goals regarding screening or other risk-mitigating behaviors are not compromised.

Theme 2: Reactions to nutrition messages

A growing body of literature demonstrates the role of specific message cues and negative affect on successfully increasing risk perceptions [50]. In the current study, receiving the message to eliminate foods that increase CRC risk overwhelmed some participants, especially those who were approaching the dietary intake recommendations or who already followed restrictive dietary guidelines for other health conditions. Some research indicates that reducing the receiver’s defensive processing of threatening messages may be necessary to achieve desired outcomes [51,52]. However, in terms of behavior change and risk perception, it may also be effective to elicit emotional reactions to nutrition guidelines, if a reasonable plan of action is made available to the message recipients as emotional states have implications on attitudes [53]. Emotional arousal has been used as a strategy to motivate individuals to make and maintain behavior changes [54]. Thus, with transparency and access to acceptable risk-reduction options, it may be ethical and practical to ignite low levels of risk perception to persuade adults to modify their health behaviors [55].

Finally, participants suggested that the VHA should balance the nutrition risk messages with more positive messages. Rothman et al. [56] indicate that the role of message frames on outcomes is complex, with evidence that gain frames (e.g. focusing on the benefits of adopting a behavior) work better for prevention behaviors such as avoiding alcohol, whereas loss frames (e.g. focusing on the costs of not adopting a behavior) work better for detection behaviors such as engaging in cancer screening. However, there is limited support to date for the impact of nutrition-specific message framing on intentions or behavior [57,58].

The Transtheoretical Model or Stages of Change (SOC) Model is a behavior change model with implications for the theme “reaction to nutrition messages”. The distinct groups (i.e. aversion, request for balance, and appreciation) seem to map onto this model, which has demonstrated positive outcomes for understanding eating behavior change processes [59]. In the SOC, individuals move cyclically through six stages of change. Contemplation is the stage where individuals begin to recognize concerns about a behavior and start to weigh the pros and cons of making a change. This definition suggests participants requesting balanced messages may be in Contemplation and desire information that can help them decide what to do. DiClemete [60] further supports this interpretation when he says “[t]he task for individuals in Contemplation is to resolve their decisional balance consideration in favor of change.” p. 29.

Furthermore, in the current study, participants seemed to prefer risk-reduction language combined with positive reinforcement for nutrition messages, further suggesting that a request for balanced messages is consistent with a strategy used for individuals in Contemplation. Also, self-efficacy is an important construct linked to longer-term dietary change [61,62], thus is a potentially important construct to test in the context of contradictory messaging delivered by a novel source. Future studies should consider delivering messages that successfully leverage emotional arousal, and consider how to effectively incorporate theoretically-informed processes of change into brief, digital interactions.

Theme 3: Information gaps

Participants identified several content areas with inadequate information, which we named information gaps. These gaps broadly covered the desire for more content on nutrition, colorectal screening, and statistics. However, the Ottawa Decision Support Framework indicates that both inadequate and excessive information can negatively affect decision-making in health contexts [25,63]. Thus, asking users about the “right amount” of information necessary to fill information gaps is important.

Surprisingly, none of the participants indicated that the length of the interaction was too long. While we took care to provide essential information, some messages, such as the importance of getting screened for colorectal cancer and remaining within guidelines were repeated in several portions of the interaction. Repetition as a strategy is consistent with previous research indicating that persuasive messages when repeated are more effective [64].

Perhaps in an attempt to assuage negative feelings when advised to avoid specific foods, study participants requested more risk statistics and numerical representations of risk information. This study’s participants’ requests for additional statistics on “levels of risk” based on personal dietary intake may indicate a desire to reduce rather than eliminate intake of certain items. While an important part of the community-engaged and user-centered approach prioritizes meeting user needs, we proceeded with caution when adding statistics on nutrition risk outcomes. Previous research indicates that understanding numerical presentations of risk is a complex skill; nearly one-third of U.S. adults cannot perform basic math functions or process simple statistics or graphs [65]. Thus, thoughtful approaches to effectively communicate numerical risk for health promotion are needed [66].

Furthermore, communicators must be mindful when delivering cancer-related risk statistics for several reasons. First, numeracy, the ability to process and understand numerical health and risk information, can influence how people process and understand cancer risk information due to its impact on information seeking, and the ability to interpret and accept population risk data [67,68]. Alternatively, Han et al. [69] indicate it may be helpful to visually represent risk, but some formats like dynamic, animated representations of risk may function better than other formats. They presented the dynamic, animated format as “9 individuals scattered randomly through an icon array of 100 persons, employing web animation to randomly change the placement of these 9 individuals every 2 s.” (p. 107). Finally, verbal sources of information (e.g. their physicians) may still be more trustworthy than numbers even when numeracy is high [70]. Thus, when patients say, “My doctor told me it was okay,” it may be a clue that certain verbal advice is more persuasive and trustworthy than numbers.

Study limitations and future research

This study has several limitations. Due to the exploratory nature of the study, there was no opportunity to include probes to elicit more in-depth responses related to contradictory sources of nutrition information. Due to the iterative study design, participants responded to evolving messages adapted for each new round of data collection. Future research should further assess the relationship between personal dietary intake and message preferences for or against contradictory nutrition risk messages. Future studies should also seek to understand the role of temporal framing of nutrition messages delivered in virtual settings. Specifically, how do they function to promote cancer prevention behaviors across a range of audience types (variations in stage of change, self-efficacy, dietary patterns)? Using an audience segmentation approach to tailor information may be helpful, especially for those in “contemplation” who may need more information to decide to change.

Lessons learned from studies on correcting health misinformation contextualize the current study findings and might help guide message developers when health messages are potentially contradictory. Based on previous work [71] correcting misinformation follows recommendations such as (1) correcting misinformation after exposure rather than as a warning before, and (2) focusing on why information is wrong and where it comes from (rather than why the new information is right and credible). Message design for contradictory nutrition information may also benefit from applying and testing these strategies. From a practical implication, as healthcare providers feel urgency and motivation to correct misinformation but lack professional, interpersonal, and institutional support in doing so, health information technologies may become essential to support and supplement challenges found in the healthcare system [72].

Conclusion

Appropriate use of health information technologies could help bridge the gap between awareness of the nutrition – cancer link and engaging in cancer prevention behaviors for older rural adults at risk for CRC. We recommend that cancer prevention interventions bundled with nutrition recommendations consider the potential effects of contradictory and conflicting nutrition guidelines. We also suggest that when “filling” or “satisfying” information gaps, audience characteristics like the receiver’s ability to manage the information (e.g. numeracy skills) or readiness (e.g. stage of change) be considered. We suggest testing perceptions of message contradictions in the context of actual behavior change, intentions to change behavior, and self-efficacy to change behavior. Message testing and community-engaged approaches may help identify unanticipated outcomes of engagement with cancer prevention and nutrition risk messages in novel web-based contexts.

Ethics approval

All procedures were approved by the University of Florida institutional review board (IRB#201902537) and Scientific Review and Monitoring Committee (SRMC) OCR27722.

Acknowledgements

MJV, EB, TP, collected data, analyzed data, and wrote the first draft of the manuscript. EJC and GM contributed meaningful contributions to the manuscript and critical comments to revised drafts. MZ and BCL advised on virtual human script development and code and contributed meaningful edits/comments to revised drafts of the manuscript. JLK contributed to data analysis and interpretation and provided key insights and critical comments and edits to revised drafts. JKL and MJV conceptualized the study.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the National Cancer Institute, National Institutes of Health, Award #3R01CA207689-03S2. Dr. Cooks was supported by a grant from the National Institutes of Health award (T32CA257923). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.