Behavior Change Factors and Retention in Dietary Interventions for Older Adults: A Scoping Review

Abstract

Background and Objectives

Although poor diet is a major driver of morbidity and mortality in people aged 60 and older, few dietary interventions are widely implemented for this population. We mapped behavior change theories, agents, and techniques in dietary interventions for adults aged 60 and older and explored relationships between these factors and ability to retain at least 80% of the study participants.

Research Design and Methods

We conducted a scoping review using MEDLINE, CINAHL, and Web of Science through April 2021 for dietary interventions in adults aged 60 and older. We collated, summarized, and calculated frequency distributions of behavior change theories, behavior change agents, and behavior change techniques (BCTs) using BCTv1 taxonomy with regard to participant retention across 43 studies.

Results

Only 49% and 30% of the studies reported behavior theory and change agents, respectively. Of the studies reporting on theory and agents, the most common were social cognitive theory and the related mechanism of self-efficacy. The most common BCTv1 clusters were “shaping knowledge” and “goals and planning.” Several BCTv1 clusters such as “antecedents” and “reward and threat” and evidence for concordance between BCTs and change agents were more common in interventions with higher retention rates.

Discussion and Implications

Mechanistically concordant studies with BCTs that involve resource allocation and positive reinforcement through rewards may be advantageous for retention in dietary intervention for older adults. Future studies should continue developing theory and mechanism-oriented research. Furthermore, future studies should consider diversifying the portfolio of currently deployed BCTs and strengthening a concordance between BCTs and mechanisms of change.

Poor diet is responsible for more deaths globally than high blood pressure, tobacco, air pollution, or any other health risks. Specifically, low consumption of healthy foods, such as whole grains, fruits, and vegetables, accounts for one in every five deaths globally (Afshin et al., 2019). Poor nutrition is a global problem, with only a small proportion (12%) of adult Americans and Europeans consuming the recommended daily amount of fruits and vegetables (Lee-Kwan, 2017; Organization for Economic Co-operation and Development & European Union, 2016). The harmful effects of a poor diet are especially pronounced in people aged 60 and older who are at increased risk for poor health by virtue of age, cumulative disease processes, and lifestyle behavior factors. Observational and intervention studies consistently show that both quality and quantity of diet are important in shaping health in older populations, including those with multimorbidity and frailty. For example, our (Zaslavsky et al., 2017) and other studies (Ford et al., 2013) showed that a higher intake of vegetables, whole grains, and nuts was strongly associated with better survival in older adults, even after accounting for overall calorie and protein intake. Experimental studies also supported these findings by showing that increased intake of healthy foods had positive effects on older adults’ health and life span (Bandayrel & Wong, 2011). Thus, interventions aimed at promoting a healthier diet have the potential to improve health in later life, but to date, few dietary interventions are widely implemented for this population.

To better understand this implementation gap, several studies have reviewed the effectiveness of behavior change techniques (BCTs) for dietary behavior change in people aged 60 years and older (Bandayrel & Wong, 2011; Lara, Evans et al., 2014; Lara, Hobbs et al., 2014; Zhou et al., 2018). Given the emerging breadth of this literature, a scoping review is well poised to complement this body of knowledge and evaluate other psychological constructs involved in behavior change—specifically, evaluating the use of behavior change theory and change agents in addition to BCTs. Briefly, behavior change theories are an abstract representation of interrelated concepts, definition, and propositions that explain behavior change (Glanz et al., 2008). A behavior change agent is a putative mechanism or process that is measurable and modifiable and is hypothesized to play a causal role in producing behavior change (Nielsen et al., 2018). Finally, BCTs are observable, replicable, irreducible, and active ingredients within the intervention designed to change behavior (Michie et al., 2011). To date, none of the prior reviews have used the most recent comprehensive taxonomy (Michie et al., 2013) to map BCTs. Moreover, to the best of our knowledge, no report to date has explored the relationship between these constructs and retention as another practical and salient concern. Failing to retain ample participants in behavior interventions may not only lead to uncertainty about intervention effectiveness and pose a threat to the external validity of the results, but may also imply potential implementation challenges, such as increased burden and low engagement, concerning the proposed intervention (Coday et al., 2005). As older adults may be more likely to drop out of studies (Strotmeyer et al., 2010), special attention to this population is warranted. Because BCTs are the foundation of behavior interventions and, in turn, influence engagement (Michie et al., 2013), it is imperative to map the use of such techniques among older adults and understand how this relates to retention in dietary intervention trials. This improved understanding may facilitate the acceleration of scientific discovery and development of “stickier” dietary interventions that will have higher retention and ultimately higher potential for implementation and dissemination. As such, in the current review, we examined the integration of behavior change theories, techniques, and agents in dietary interventions targeting older adults and explored relationships between these factors and the ability to successfully retain at least 80% of the study participants.

Method

Framework

We followed Arksey and O’Malley’s scoping review framework, which has five distinct stages: (a) identifying the research question, (b) identifying relevant studies, (c) selecting relevant studies, (d) charting data from the selected studies, and (e) summarizing and reporting the results (Arksey & O’Malley, 2005). Unlike systematic reviews, scoping reviews do not evaluate the quality of evidence, but rather collate and descriptively summarize them by mapping literature and examining the extent, breadth, nature, and characteristics of the available knowledge (Arksey & O’Malley, 2005).

Stage 1: Identifying the Research Question

Two main research objectives were defined based on gaps in the literature: (a) synthesize the evidence on the integration of behavior change theories, agents, and BCTs in dietary interventions targeting older adults; (b) describe the relationships between behavior change theories, agents, and BCTs and ability to successfully retain at least 80% of participants in the dietary intervention studies.

Stage 2: Identifying Relevant Studies/Search Strategy

Medical Literature and Retrieval System Online (MEDLINE), Cumulative Index to Nursing and Allied Health Literature (CINAHL), and Web of Science were searched for relevant studies. The literature search was performed by Y. Su and an expert librarian. The keywords or combination of Medical Subject Headings was modified to optimize search strategies in each database. A keyword of (intervention* OR program* OR experiment* OR trial* OR “pilot study” OR “quasi-experimental” OR “prospective cohort study”) AND (behavior*[tiab] OR behavior*[tiab] AND change*[tiab]) AND (diet*[tiab] OR nutrition*[tiab] OR “healthy eating”[tiab] OR “healthy diet”[tiab] OR “healthy nutrition”[tiab]) was searched in each of the databases. Y. Su and the librarian independently verified search terms and discussed initial results to confirm that the strategies were performed as expected. No historical time limits were set, but the search was conducted in April 2021. Other search strategies, such as reference checking, were also deployed to capture potentially relevant studies. The detailed search terms are presented in Supplementary Table 1.

Stage 3: Study Selection

All peer-reviewed original articles retrieved from the three databases were considered eligible for further review.

Eligibility criteria

Inclusion criteria were as follows: (a) human participants with an average age 60 and older; (b) noninstitutionalized population; (c) experimental (including randomized controlled trials [RCTs]) or quasi-experimental design (i.e., nonequivalent control and pre–post, nonequivalent control and post only, one group pre–post, and time-series designs); (d) dietary interventions aimed at specific food groups (e.g., fruit/vegetables) and/or dietary patterns (e.g., Mediterranean diet); and (e) full text available in English. Exclusion criteria were as follows: (a) study protocols, (b) studies lacking sufficient information on the intervention protocol, (c) studies that blended physical activity and nutritional intervention, and (d) studies of dietary supplements and nutrition formulas.

Selection of studies for inclusion

Two authors (Y. Su and I. Roopsawang and later K.-C. Wu) independently examined the titles and abstracts to assess their relevance for the review. Next, Y. Su, B. Kim, and K.-C. Wu independently retrieved and reviewed full texts such that discrepancies in the selection could be mapped and summarized. The disagreement among the initial reviewers was arbitraged by the third reviewer (O. Zaslavsky). Specifically, O. Zaslavsky reviewed discrepant studies (n = 6) and independently determined their eligibility. The workflow was shown via Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline (Supplementary Figure 1).

Stage 4: Charting the Data

Rayyan software was used to check for duplicates and to allow an independent review process by the assessors (Y. Su, B. Kim, and K.-C. Wu) and the third arbiter (O. Zaslavsky), as described earlier. Next, we generated a data-reporting table shell to guide the data abstraction process and display a summary of these study features: citation, location, health context, study type, sample, intervention characteristics, retention, behavior change theory, BCT, behavior change agents, concordance between BCTs and behavior change agents, and main endpoints.

Stage 5: Assembling, Summarizing, and Reporting Results

We used the conceptual definitions provided in the introduction and the following procedures to guide our reporting. Behavior change theories were extracted if they were explicitly mentioned as guiding interventions. Behavior change agents were extracted if the constructs were consistent with behavior change theories that informed the intervention and were measured before and after the intervention. BCTs were coded according to the BCTv1 taxonomy (Michie et al., 2013) comprised of 93 individual BCTs grouped in 16 hierarchical clusters (e.g., scheduled consequences, reward and threat, repetition and substitution, antecedents, and associations). For each study, concordance between change agents and BCTs was determined by considering a conceptual alignment between the change agent and corresponding BCTs. For example, if self-efficacy behavior change agent was considered, we were looking for evidence of BCTs from the self-belief BCTv1 hierarchical cluster (e.g., focus on past successes, self-talk). Main endpoints were abstracted if they were listed as primary outcomes, were defined in study objectives, and were measured before and after interventions. Retention was defined as a percentage of study participants who completed procedures as specified in a study protocol. Retentions were abstracted if they were explicitly reported or could be calculated from flowcharts or other comparable information sources.

Results

Study Selection and Characteristics

Keyword searches in MEDLINE, Web of Science, and CINAHL resulted in 948, 1,390, and 646 hits, respectively (Supplementary Table 1), with a total of 2,984 search results. An additional 22 articles were added from reviewing references in candidate studies. After removing duplicates and title/abstract screenings, we retained 653 studies. Subsequent full-article eligibility criteria assessment led to a final list of 43 articles (see Supplementary Figure 1 for the PRISMA flowchart). Table 1 provides the characteristics of the included studies.

Table 1.

The Characteristics of Dietary Intervention Studies in Adults Aged 60 and Older (N = 43) in Chronological Order

Study; location Study design	Sample	Intervention Interventionists	Retention	Behavior change theory	Behavior change technique BCTv1 category Individual BCTs	BCA	Concordance between BCTs and BCA	Main endpoints
Zacharia et al. (2020); Australia Pilot one group pre–post design	Robust; N = 17, mean age = 71, 29% male	2-week intervention where participants received a group counseling session, food sample, educational materials, and text messages Researchers	88%	Behavior Change Wheel and COM-B model (utilizes capability, opportunity, and motivation to elicit behavior change)	4. Shaping knowledge   4.1 Instruction on how to perform the behavior 4.2 Information about antecedents 6. Comparison of behavior 6.1 Demonstration of the behavior 7. Associations 7.1 Prompts/cues 12. Antecedents 12.5 Adding objects to the environment	No	N/A	Dietary behavior
Smith et al. (2020); USA Quasi-experimental design	Many have cardiovascular risk; N = 430, mean age = 74.5, 23% male	12-week intervention where participants received small group educational sessions focused on healthy eating Unclear	77%	Socioecological framework	1. Goals and planning   1.1 Goal setting (behavior) 1.2 Problem solving 1.4 Action plan 1.9 Commitment 2. Feedback and monitoring 2.3 Self-monitoring of behavior 4. Shaping knowledge 4.1 Instruction on how to perform the behavior	Self-efficacy Social support	No	Dietary behaviors Self-efficacy
van Doorn-van Atten et al. (2019); The Netherlands One group pre–post design	Home care older adults; N = 20, mean age = 81; 25% male	3-month intervention where participants received e-health intervention consisting of nutritional monitoring, messaging, and dietary advice Researchers and health care professionals	55%	No	1. Goals and planning   1.2 Problem solving 2. Feedback and monitoring 2.2 Feedback on behavior 2.4 Self-monitoring of outcome(s) of behavior 4. Shaping knowledge 4.1 Instruction on how to perform the behavior 7. Associations 7.1 Prompts/cues 8. Repetition and substitution 8.1 Behavior practice/rehearsal 15. Self-belief 15.1 Verbal persuasion about capability	No	N/A	Feasibility Dietary behavior Physiological status Clinical outcomes
van Doorn-van Atten et al. (2018); The Netherlands Quasi-experimental	Home care and/or living in shelter; N = 214, mean age = 80, 23%–35% male	6-month telemonitoring intervention where participants received access to nutritional monitoring and dietary advice Nurses	87%	Control theory	1. Goals and planning   1.1 Goal setting (behavior) 2. Feedback and monitoring 2.2 Feedback on behavior 2.3 Self-monitoring of behavior 4. Shaping knowledge 4.1 Instruction on how to perform the behavior	Goal setting Self-monitoring Feedback Knowledge	Yes	Knowledge and attitudes
Downes et al. (2019); USA One group pre–post design	91% had ≥1 chronic diseases; N = 47, mean age = 67, 26% male	6- and 8-week intervention where participants received group educational sessions on healthy eating Researchers	47%	Modified Motivators and Barriers of Health Behaviors Model	4. Shaping knowledge   4.1 Instruction on how to perform the behavior 4.2 Information about antecedents 5. Natural consequences 5.1. Information about health Consequences	Healthy lifestyle barriers; No		Dietary behavior Other behavior Physiological status
Ahn et al. (2018); South Korea Quasi-experimental two-group	Older adults living alone; N = 71, mean age = 77.6, 18% male	8-week intervention where participants received individualized nutritional education and follow-up calls Nurses or dietitian	N/A	No	4. Shaping knowledge   4.1 Instruction on how to perform the behavior	No	N/A	Dietary behavior Knowledge
Demark-Wahnefried et al. (2018); USA RCT	Cancer survivors; N = 46, mean age = 70.1; 30% male	12-month intervention where participants received one-on-one mentoring on how to plant and maintain three vegetable gardens with gardening supplies support Gardeners	91%	SCT, Social–Ecological Model	3. Social support   3.1 Social support(unspecified) 4. Shaping knowledge 4.1 Instruction on how to perform the behavior 12. Antecedents 12.5 Adding objects to the environment 15. Self-belief 15.1 Verbal persuasion about capability	Self-efficacy Social support	Yes	Feasibility
Schlaff et al. (2018); USA RCT	Inactive older adults; N = 72, mean age = 64; 28% male	12-week intervention where participants received a healthy eating workbook and group discussion sessions Investigator	69%	SCT, TTM	1. Goals and planning   1.1 Goal setting (behavior) 1.2 Problem solving 2. Feedback and monitoring 2.3 Self-monitoring of behavior 3. Social support 3.1 Social support(unspecified) 4. Shaping knowledge 4.1 Instruction on how to perform the behavior	No	N/A	Dietary behavior
Chiu et al. (2019); Taiwan One group pre–post design	Robust; N = 21, mean age = 65, 48% male	6-week intervention where participants received technology-enhanced educational sessions on healthy eating Dieticians	60%	Scaffolding Theory	4. Shaping knowledge   4.1 Instruction on how to perform the behavior 6. Comparison of behavior 6.1 Demonstration of the behavior	Self-efficacy	No	Knowledge Self-efficacy Dietary behavior
Chen et al. (2017); Taiwan Quasi-experimental two-group	Robust; N = 129, mean age = 73; 36% male	12-week intervention where participants received a healthy eating workbook and individual counseling sessions Research team	93%	SCT	1. Goals and planning   1.1 Goal setting 1.2 Problem solving 8. Repetition and substitution 8.1 Behavior practice/rehearsal 15. Self-belief 15.1 Verbal persuasion about capabilities	Self-efficacy	Yes	Knowledge and attitudes Physiological status
Bird and McClelland (2017); USA RCT	Older adults with limited resources; N = 453, mean age = 74; 21% male	4-week intervention where participants received educational and experiential group sessions Extension agents, educators	70%	Adult learning theory, TPB	1. Goals and planning   1.1 Goal setting (behavior) 1.4 Action planning 1.8 Behavior contract 3. Social support 3.1 Social support (unspecified) 4. Shaping knowledge 4.1 Instruction on how to perform the behavior 6. Comparison of behavior 6.1 Demonstration of the behavior 6.2 Social comparison	No	N/A	Knowledge and attitudes
Anderson et al. (2015); USA RCT	Robust; N = 40, mean age = 61; 35% male	20-week intervention where participants received a workbook and individual and group counseling sessions Unclear	100%	No	1. Goals and planning   1.1 Goal setting (behavior) 1.2 Problem solving 2. Feedback and monitoring 2.3 Self-monitoring of behavior 4. Shaping knowledge 4.1 Instruction on how to perform the behavior 6. Comparison of behavior 6.1 Demonstration of the behavior 8. Repetition and substitution 8.2 Behavior substitution 8.3 Habit formation 10. Reward and threat 10.6 Nonspecific incentive	No	N/A	Salt intake
Lara et al. (2015); UK One group pre–post design	Robust; N = 23, mean age = 66; 30% male	3-week intervention where participants attended educational group sessions and received individual support by phone Nutritionist	100%	No	3. Social support   3.2 Social support (practical) 4. Shaping knowledge 4.1 Instructions on how to perform the behavior 5. Natural consequences 5.1. Information about health consequences 12. Antecedents 12.5 Adding objects to the environments	No	N/A	Feasibility Dietary behavior
Moreau et al. (2015); Canada One group pre–post design	Robust; N = 144, mean age 60+, 12.5% male	8-week intervention where participants attended educational sessions and practiced food preparation Dietitian	N/A	SCT, TPB, TTM	1. Goals and planning   1.1 Goal setting (behavior) 1.2 Problem solving 4. Shaping knowledge 4.1 Instructions on how to perform the behavior 5. Natural consequences 5.1. Information about health	Confidence	No	Knowledge and attitudes Dietary behavior
Appleton (2013); UK RCT	Robust; N = 95, mean age = 74, 29% male	5-week intervention where participants were repeatedly exposed to and received fruits Researcher	82%	No	12. Antecedent   12.5 Adding objects to the environments	No	N/A	Dietary behavior Attitude
Archuleta et al. (2012); USA One group pre–post design	Type 2 diabetes; N = 117, mean age = 63; 22% male	An intervention where participants took part in group educational sessions that involved hands-on experience Educators, dietitians, and extension agents	N/A	SCT	1. Goals and planning   1.1 Goal setting (behavior) 3. Social support 3.3 Social support (emotional) 4. Shaping knowledge 4.1 Instruction on how to perform the behavior 8. Repetition and substitution 8.1 Behavior practice/rehearsal 15. Self-belief 15.1 Verbal persuasion about capability	No	N/A	Dietary behavior
Racine et al. (2012); USA RCT	Hypertension or hyperlipidemia; N = 298, mean age = 72.4, 16.1% male	One-year trial where participants received home-delivered foods Dietitian	93%	No	12. Antecedents   12.5 Adding objects to the environments	No	N/A	Physiological status
Gibson et al. (2012); UK RCT	Robust; N = 83, mean age = 71, 35% male	16-week intervention where participants received personal nutritional counseling and complementary foods Researchers	99%	No	2. Feedback and monitoring   2.1 Monitoring of behavior by others without feedback 2.5 Monitoring of outcomes of behavior without feedback 4. Shaping knowledge 4.1 Instruction on how to perform the behavior 12. Antecedents 12.5 Adding objects to the environments	No	N/A	Physiological status
Lammes et al. (2012); Sweden RCT	Frail older adults; N = 93, mean age = 83, 40% male	3-month intervention where participants received individual and group dietary counseling Nutritionist or dietitian	85%	No	4. Shaping knowledge   4.1 Instruction on how to perform the behavior	No	N/A	Dietary behavior Physiological status
Salehi et al. (2011); Iran Quasi-experimental two-group	Robust; N = 400, mean age = 64; 26% male	4-week intervention where participants took part in educational sessions Researcher	N/A	TTM, SCT	1. Goals and planning   1.1 Goal setting (behavior) 1.4 Action planning 1.9 Commitment 4. Shaping knowledge 4.1 Instruction on how to perform the behavior 5. Natural consequences 5.1 Information about health consequences 9. Comparison of outcomes 9.3 Comparative imagining of future outcomes 9.1 Credible source 10. Reward and threat 10.3 Nonspecific reward 13. Identity 13.2 Framing/reframing	Self-efficacy Stage of change questionnaire	No	Dietary behavior Attitude
Locher et al. (2011); USA RCT	At risk for malnutrition; N/A	2-month intervention where participants received individualized nutritional counseling and phone calls Dietitian	N/A	Ecological Model, SCT	3. Social support   3.1 Social support (unspecified) 4. Shaping knowledge 4.1 Instruction on how to perform the behavior 12. Antecedents 12.1 Reconstructing the physical environment	Self-efficacy Goal setting	No	Efficacy and feasibility
Abusabha et al. (2011); USA One group pre–post design	Low-income senior; N = 4, mean age = 69; 14% male	3-month interventions where participants received weekly access to a vendor selling inexpensive produce Researcher	54%	No	12. Antecedents   12.1 Restructuring the physical environment	No	N/A	Dietary behavior
Endevelt et al. (2011); Israel RCT	Malnourished; N = 127, mean age = 84, 36%–40% male	6-month interventions where participants received individual counseling sessions Dietitian	N/A	No	1. Goals and planning   1.2 Problem solving 1.3 Goal setting (outcome) 2. Feedback and monitoring 2.1 Monitoring of outcomes of behaviors without feedback 3. Social support 3.3 Social support (emotional) 4. Shaping knowledge 4.1 Instruction on how to perform the behavior 12. Antecedents 12.5 Adding objects to the environments	N	N/A	Dietary behavior Cost of service Clinical outcome
Wunderlich et al. (2011); USA Quasi-experimental two-group	Robust; N = 476, mean age = 74, 32%–35% male	2-year intervention where participants received group or individual counseling sessions Nutritionists	75%	No	3. Social support   3.1 Social support (unspecified) 4. Shaping knowledge 4.1 Instruction on how to perform the behavior 6. Comparison of behavior 6.1 Demonstration of the behavior	No	N/A	Dietary behavior Physiological status
Babatunde et al. (2011); USA RCT	Robust; N = 110, mean age = 70; 10% male	6-week intervention where participants attended group educational sessions Investigator	85%	HBM	4. Shaping knowledge   4.1 Instruction on how to perform the behavior 4.2 Information about antecedents 6. Comparison of behavior 6.1 Demonstration of the behavior 13. Identity 13.3 Incompatible beliefs 15. Self-belief 15.1 Verbal persuasion about capability	Self-efficacy Susceptibility Seriousness Benefits of and barriers to increasing calcium Health motivation	Yes	Dietary behavior Knowledge and attitudes
Bradbury et al. (2006); UK RCT	Edentulous; N = 66, mean age = 66, 43% male	6-week intervention where participants received individual counseling sessions Nutritionist	88%	TTM, Optimistic bias	1. Goals and planning   1.4 Action planning 2. Feedback and monitoring 2.3 Self-monitoring of behavior 2.2 Feedback on behavior 4. Shaping knowledge 4.1 Instruction on how to perform the behavior 5. Natural consequences 5.1. Information about health	Stage of change	Yes	Dietary behavior Attitude
Manios et al. (2006); Greece RCT	Self-dependent; N = 82, mean age = 60, 0% male	5-month intervention where participants received educational sessions and food samples Unclear	91%	No	1. Goals and planning   1.4 Action planning 4. Shaping knowledge 4.2 Information about antecedents 5. Natural consequences 5.1. Information about health 8. Repetition and substitution 8.2 Behavior substitution 12. Antecedents 12.5 Adding objects to the environments	No	N/A	Dietary behavior Physiological status
Mitchell et al. (2006); USA RCT	Low income; N = 1,006, mean age = 77, 22% male	9-week intervention where participants attended weekly education sessions Research team	70%	SCT	1. Goals and planning   1.4 Action planning 4. Shaping knowledge 4.1 Instruction on how to perform the behavior 5. Natural consequences 5.1 Information about health consequences 7. Associations 7.1 Prompts/cues 15. Self-belief 15.1 Verbal persuasion about capability	No	N/A	Dietary behavior Knowledge and attitudes
Chapman-Novakofsk & Karduck (2005); USA One group pre–post design	Diabetes; N = 239, mean age = 63, 27% male	This intervention involved monthly education and experiential session Unclear	N/A	SCT, TTM	4. Shaping knowledge   4.1 Instruction on how to perform the behavior 6. Comparison of behavior 6.1 Demonstration of the behavior 15. Self-belief 15.1 Verbal persuasion about capability	Self-efficacy Stage of change	Yes	Knowledge and attitudes
Verheijden et al. (2004); The Netherlands RCT	CVD risk; N = 143, mean age = 60, 27% male	This intervention involved group and individual counseling sessions Family physician	91%	TTM	1. Goals and planning   1.2 Problem solving 2. Feedback and monitoring 2.2 Feedback on behaviors 2.7 Feedback on outcomes of behaviors 4. Shaping knowledge 4.1 Instruction on how to perform the behavior 9. Comparison of outcomes 9.2 Pros and cons 13. Identity 13.3 Incompatible beliefs	Stage of change	Yes	Dietary behavior Attitude
Miller et al. (2002); USA RCT	Type 2 diabetes mellitus, N = 98, mean age = 72.5, 47% male	10-week intervention where participants received educational sessions on healthy eating, diabetes, and glucose control Dietitian	94%	Theory of Meaningful Learning; Information Processing Model; SCT	1. Goals and planning   1.1 Goal setting (behavior) 1.5 Review behavior goal(s) 1.6 Discrepancy between current behavior and goal 2. Feedback and monitoring 2.2 Feedback on behavior 2.3 Self-monitoring of behavior Self-monitoring of outcome(s) of behavior 4. Shaping knowledge 4.1 Instruction on how to perform a behavior 4.2 Information about antecedents 6. Comparison of behavior 6.2 Social comparison 8. Repetition and substitution 8.1 Behavioral practice/rehearsal 10. Reward and threat 10.3 Nonspecific reward	No	N/A	Physiological status
Bernstein et al. (2002); USA RCT	Low physical function; N = 70, mean age = 78, 20% male	6-month intervention where participants received individual counseling sessions, phone calls, and mailed written materials Dietitian	100%	No	1. Goals and planning   1.1 Goal setting (behavior) 1.2 Problem solving 2. Feedback and monitoring 2.3. Self-monitoring of behavior 4. Shaping knowledge 4.1 Instruction on how to perform the behavior 5. Natural consequences 5.1. Information about health 6. Comparison of behavior 6.1 Demonstration of the behavior 10. Reward and threat 10.6 Nonspecific incentives	No	N/A	Dietary behavior
Glasgow et al. (1996); USA RCT	Diabetes; N = 200–206, mean age = 62.4, 39% male	This intervention involved individual counseling sessions and phone calls Research staff	87%	SCT, systems approach	1. Goals and planning   1.1 Goal setting (behavior) 1.2 Problem solving 2. Feedback and morning 2.2 Feedback on behaviors 4. Shaping knowledge 4.1 Instruction on how to perform the behavior 15. Self-belief 15.1 Verbal persuasion about capabilities	No	N/A	Dietary behavior Physiological measure Clinical outcome
Kupka-Schutt and Mitchell (1993); USA RCT	Robust; N = 104, mean age = 72.2, 23% male	4-week intervention where participants attended group educational sessions Dietitian	80%	Nutrition Instruction Model	1. Goals and planning   1.1 Goal setting (behavior) 1.4 Action planning 1.2 Problem solving 3. Social support 3.1 Social support (unspecified) 4. Shaping knowledge 4.1 Instruction on how to perform the behavior	No	N/A	Dietary behavior
Salas-Salvadó et al. (2008, 2014), Fernández-Real et al. (2012), Sánchez-Villegas et al. (2013), Murie-Fernandez et al. (2011), Zazpe et al. (2008); Spain RCT	High CVD risks, N = from 127 to 4,557, mean age = from 66 to 68, 38%–100% male	A 12-month, multiyear intervention where participants received group educational sessions and complementary olive oil and nuts Dietitian, nurse	From 88% to 97%, or N/A	No	4. Shaping knowledge   4.1 Instruction on how to perform the behavior 12. Antecedents 12.5 Adding objects to the environments	No	N/A	Clinical outcome Physiological status Dietary behavior
Sarma et al. (2019), Prentice et al. (2017, 2006); USA RCT	Robust N = 20,380–48,835; mean age = N/A or 61–66, 0% male	Multiyear intervention where participants received a self-controlled eating plan with occasional feedback on their performance during group sessions Nutritionist	N/A or 42%	No	1. Goals and planning   1.1 Goal setting (behavior) 2. Feedback and monitoring 2.3 Self-monitoring of behavior 2.7 Feedback on outcome(s) of behavior 3. Social support 3.1Social support (unspecified) 4. Shaping knowledge 4.1 Instruction on how to perform the behavior	No	N/A	Dietary behavior Other behavior Clinical outcome

Note: BCA = behavior change agents; BCT = behavior change technique; CVD = cardiovascular disease. HBM = health belief model; RCT = randomized controlled trial; SCT = social cognitive theory; TPB = theory of planned behavior; TTM = transtheoretical model; N/A = nonavailable.

Geographically, most of the included studies were conducted in the United States (n = 21), Europe (n = 15), followed by Asia (n = 5), Australia (n = 1), and Canada (n = 1). Twenty-eight studies were RCTs; six studies had pretest/post-test designs; and six studies were nonrandomized controlled interventions. The reported mean participant age ranged from 60 to 84, with a median of 69 years of age. The sample size ranged from 17 to 48,835, with a median sample size of 127 participants. The most common clinical health context was diabetes, high blood pressure, and cardiovascular disease (n = 13). Other less common clinical contexts included frailty (n = 2), cancer survivorship (n = 1), edentulousness (n = 1), and malnourishment (n = 2).

Interventions

Thirty-six out of 43 studies reported the duration of the intervention. The median intervention duration was 12 weeks (range = 3–192 weeks). Thirty-nine out of 43 studies provided information on the interventionist professional background. Specifically, 23 studies noted that the interventionists were dietitians/nutritionists. Thirteen studies reported that a researcher/investigator delivered interventions without specifying their professional background, and three studies reported that either a nurse (van Doorn-van Atten et al., 2018), family physician (Verheijden et al., 2004), or gardener (Demark-Wahnefried et al., 2018) were interventionists.

Theory

Only 21 out of 43 studies (49%) explicitly mentioned theories/models of behavior change that guided their interventions. Of these, 10 studies mentioned more than one model. The most commonly used model was social cognitive theory (n = 11). Other theories were transtheoretical model (n = 6; Bradbury et al., 2006; Chapman-Novakofski & Karduck, 2005; Moreau et al., 2015; Salehi et al., 2011; Schlaff et al., 2018; Verheijden et al., 2004), control theory (n = 1; van Doorn-van Atten et al., 2018), health belief model (n = 1; Babatunde et al., 2011), optimistic bias (n = 1; Bradbury et al., 2006), nutrition instruction model (n = 1; Kupka-Schutt & Mitchell, 1993), ecological model (n = 3; Demark-Wahnefried et al., 2018; Locher et al., 2011; Smith et al., 2020), adult learning theory (n = 1; Bird & McClelland, 2017), theory of planned behavior (n = 2; Bird & McClelland, 2017; Moreau et al., 2015), behavior change wheel (n = 1; Zacharia et al., 2020), a modified motivators and barriers of health behaviors model (n = 1; Downes et al., 2019), scaffolding theory (n = 1; Chiu et al., 2019), theory of meaningful learning (n = 1; Miller et al., 2002), information processing model (n = 1; Miller et al., 2002), and systems approach (n = 1; Glasgow et al., 1996).

Change Agent

Thirteen out of 43 studies (30%) measured behavior change agents. The most commonly reported agents were self-efficacy (n = 8; Babatunde et al., 2011; Chapman-Novakofski & Karduck, 2005; Chen et al., 2017; Chiu et al., 2019; Demark-Wahnefried et al., 2018; Locher et al., 2011; Salehi et al., 2011; Smith et al., 2020) and stage of change (n = 4; Bradbury et al., 2006; Chapman-Novakofski & Karduck, 2005; Salehi et al., 2011; Verheijden et al., 2004). Interestingly, social cognitive theory was the most commonly used in tandem with behavior change agents (n = 5; Chapman-Novakofski & Karduck, 2005; Chen et al., 2017; Demark-Wahnefried et al., 2018; Locher et al., 2011; Salehi et al., 2011).

Behavior Change Techniques

All of the 43 intervention protocols included at least one BCT, and most studies included multiple BCTs. Specifically, the total number of individual BCTs included in the studies ranged from 1 to 10, with a median of four techniques. The total number of BCTv1 clusters ranged from 1 to 6 with a median of three clusters. Approximately one third (n = 36, 39%) of the 93 potential BCTs were reported in these studies. The abstracted individual BCTs in each study are listed in Table 1. Of the 16 hierarchical clusters of the BCTv1, 13 (81%) were identified. Figure 1 illustrates the frequency distribution of BCTv1 clusters in the studies. Shaping knowledge was the most frequently deployed BCTv1 cluster (included in 39 of the 43 studies). The other common clusters were goals and planning (n = 22), antecedents (n = 16), feedback and monitoring (n = 15), and social support (n = 12).

Figure 1.

Frequency distribution of behavior change techniques (BCTv1) clusters in dietary intervention studies in adults aged 60 and older (N = 43).

Concordance Between BCTs and Agents

Among the 13 studies that specified change agents, six were concordant with the BCTs (Babatunde et al., 2011; Bradbury et al., 2006; Chapman-Novakofski & Karduck, 2005; Chen et al., 2017; Demark-Wahnefried et al., 2018; van Doorn-van Atten et al., 2018). The most common concordance was between self-efficacy change agent and self-belief BCTv1 cluster (n = 4; Babatunde et al., 2011; Chapman-Novakofski & Karduck, 2005; Chen et al., 2017; Demark-Wahnefried et al., 2018).

Main Endpoints

From the included studies, we abstracted 36 distinct interventions because several studies deployed the same intervention protocol across different endpoints. The reported endpoints included dietary behaviors (studies n = 24, unique interventions n = 24), knowledge and attitudes (studies n = 11, unique interventions n = 11), clinical outcomes (studies n = 9, unique interventions n = 5; Endevelt et al., 2011; Gibson et al., 2012; Glasgow et al., 1996; Murie-Fernandez et al., 2011; Prentice et al., 2006, 2017; Salas-Salvadó et al., 2008, 2014; Sánchez-Villegas et al., 2013), physiological status (studies n = 9, unique interventions n = 9; Chen et al., 2017; Downes et al., 2019; Fernández-Real et al., 2012; Glasgow et al., 1996; Lammes et al., 2012; Manios et al., 2006; Miller et al., 2002; Racine et al., 2012; Wunderlich et al., 2011), feasibility (studies n = 4, unique interventions n = 4; Demark-Wahnefried et al., 2018; Lara et al., 2015; Locher et al., 2011; van Doorn-van Atten et al., 2019), cost (n = 1; Endevelt et al., 2011; Lara et al., 2015), and BCT frequency (n = 1; van Doorn-van Atten et al., 2018). Nineteen interventions targeted multiple endpoints. Dietary behaviors were captured by measures of adherence to dietary patterns and/or intake of specific foods and food groups. Knowledge and attitudes included measures of intention, feeling, beliefs, and readiness. Clinical outcomes included physical symptoms, quality of life, and functional limitations. A physiological status was measured using biological indexes such as immune response, biochemical indices, osteocalcin, fat mass, nutritional status, and bone mineral density. Feasibility was measured by retention and satisfaction with the intervention. Costs were measured using a cost of adherence to a specific diet or a cost of health care services. BCT frequency was measured by calculating the number of BCT occurrences during the study.

Supplementary Table 2 summarizes the frequency of BCTv1 clusters with regard to the main endpoints in the studies. Of 24 unique intervention protocols targeting dietary behavior, 21 (87.5%) shaping knowledge, 12 (50%) goals and planning, and seven (29%) feedback and monitoring BCTv1 clusters were included. Six protocols included all three of the BCTv1 clusters (Anderson et al., 2015; Bradbury et al., 2006; Endevelt et al., 2011; Glasgow et al., 1996; Schlaff et al., 2018; Verheijden et al., 2004). Among 11 unique intervention protocols targeting knowledge and attitudes, nine (82%) included shaping knowledge, six (55%) contained goals and planning, and four (36%) included comparison of behavior. Of nine unique protocols targeting physiological status, seven (78%) shaping knowledge, four (44%) goals and planning, three (33%) antecedents, and three (33%) repetition and substitution were included. Finally, among five unique intervention protocols targeting clinical outcomes, five (100%) shaping knowledge, three (60%) antecedents, and four (80%) feedback and monitoring were included. Two protocols included all three of these BCTv1 clusters (Endevelt et al., 2011; Gibson et al., 2012).

Retention

Thirty-two (29 unique studies) out of 43 studies (74%) reported retention/attrition information. Retention ranged from 42% to 100%, with a median of 88%. Table 2 summarizes the frequency of behavior change features in the studies by two retention groups. Specifically, studies were grouped into “high” (>80%, n = 21 studies, n = 18 unique studies) and “low” (≤80%, n = 11 studies, n = 11 unique studies) retention categories. For ease of interpretation, we flagged redundant studies that deployed the same intervention protocol across different endpoints. In the case of redundancy, only the first published study was used for calculations. We also indicated marked differences in the frequency of behavior change features between high- and low-retention studies. The marked differences were conceptualized as 15 or more percent points difference between proportions of features present in each group. As given in Table 2, the high-retention protocols were characterized by the use of antecedents (44% in high retention vs. 9% in low retention) and reward and threat (17% vs. 0%). On the other hand, low-retention studies often reported the use of goals and planning (64% vs. 44%) and social support (45% vs. 11%) BCTv1. Among the six studies that reported agents and in the high-retention group, all of them (100%) had concordance between BCTs and change agents, while in the low-retention group, none of the studies were concordant.

Table 2.

Frequency of Behavior Change Technique (BCTv1) Clusters, Theories, and Agents in Dietary Intervention Studies in Adults Aged 60 and Older (N = 32) by Studies With High (>80%) vs. Low (≤80%) Retention Rates

	High retention (>80%)		Low retention (≤80%)		Difference in frequency (%)
	Studies	n (%)	Studies	n (%)
BCTv1
Antecedent	Demark-Wahnefried et al. (2018), Lara et al. (2015) {Salas-Salvadó et al. (2014), Murie-Fernandez et al. (2011), Salas-Salvadó et al. (2008), Zazpe et al. (2008)}, Appleton (2013), Gibson et al. (2012), Manios et al. (2006), Racine et al. (2012), Zacharia et al. (2020)	8 (44)	Abusabha et al. (2011)	1 (9)	35
Associations	Zacharia et al. (2020)	1 (6)	van Doorn-van Atten et al. (2019), Mitchell et al. (2006)	2 (18)	−12
Comparison of behavior	Anderson et al. (2015), Babatunde et al. (2011), Bernstein et al. (2002), Zacharia et al. (2020), Miller et al. (2002)	5 (28)	Bird and McClelland (2017), Wunderlich et al. (2011), Chiu et al. (2019)	3 (27)	1
Comparison of outcome	Verheijden et al. (2004)	1 (6)		0	6
Feedback and monitoring	van Doorn-van Atten et al. (2018), Anderson et al. (2015), Gibson et al. (2012), Bradbury et al. (2006), Verheijden et al. (2004), Bernstein et al. (2002), Glasgow et al. (1996), Miller et al. (2002)	8 (44)	van Doorn-van Atten et al. (2019), Schlaff et al. (2018), Sarma et al. (2019), Smith et al. (2020)	4 (36)	8
Goals and planning	van Doorn-van Atten et al. (2018), Chen et al. (2017), Anderson et al. (2015), Bradbury et al. (2006), Verheijden et al. (2004), Bernstein et al. (2002), Glasgow et al. (1996), Miller et al. (2002)	8 (44)	van Doorn-van Atten et al. (2019), Schlaff et al. (2018), Bird and McClelland (2017), Mitchell et al. (2006), Kupka-Schutt and Mitchell (1993), Sarma et al. (2019), Smith et al. (2020)	7 (64)	−20
Identity	Babatunde et al. (2011), Verheijden et al. (2004)	2 (11)		0	11
Natural consequences	Lara et al. (2015), Bradbury et al. (2006), Bernstein et al. (2002)	3 (17)	Mitchell et al. (2006), Downes et al. (2019)	2 (18)	−1
Repetition and substitution	Chen et al. (2017), Anderson et al. (2015), Manios et al. (2006), Miller et al. (2002)	4 (22)	van Doorn-van Atten et al. (2019)	1 (9)	13
Reward and threat	Anderson et al. (2015), Bernstein et al. (2002), Miller et al. (2002)	3 (17)		0	17
Self-belief	Demark-Wahnefried et al. (2018), Chen et al. (2017), Babatunde et al. (2011), Glasgow et al. (1996)	4 (22)	van Doorn-van Atten et al. (2019), Mitchell et al. (2006)	2 (18)	4
Shaping knowledge	van Doorn-van Atten et al. (2018), Demark-Wahnefried et al. (2018), Anderson et al. (2015), Lara et al. (2015) {Salas-Salvadó et al. (2014), Murie-Fernandez et al. (2011), Salas-Salvadó et al. (2008), Zazpe et al. (2008)}, Gibson et al. (2012), Lammes et al. (2012), Babatunde et al. (2011), Bradbury et al. (2006), Verheijden et al. (2004), Bernstein et al. (2002), Glasgow et al. (1996), Zacharia et al. (2020), Manios et al. (2006), Miller et al. (2002)	15 (83)	van Doorn-van Atten et al. (2019), Schlaff et al. (2018), Bird and McClelland (2017), Wunderlich et al. (2011), Mitchell et al. (2006), Kupka-Schutt and Mitchell (1993), Sarma et al. (2019), Downes et al. (2019), Chiu et al. (2019), Smith et al. (2020)	10 (91)	−8
Social support	Demark-Wahnefried et al. (2018), Lara et al. (2015)	2 (11)	Schlaff et al. (2018), Bird and McClelland (2017), Wunderlich et al. (2011), Kupka-Schutt and Mitchell (1993), Sarma et al. (2019)	5 (45)	−34
Theory	van Doorn-van Atten et al. (2018), Demark-Wahnefried et al. (2018), Chen et al. (2017), Babatunde et al. (2011), Bradbury et al. (2006), Verheijden et al. (2004), Glasgow et al. (1996), Zacharia et al. (2020), Miller et al. (2002)	9 (50)	Schlaff et al. (2018), Bird and McClelland (2017), Mitchell et al. (2006), Kupka-Schutt and Mitchell (1993), Smith et al. (2020), Downes et al. (2019), Chiu et al. (2019)	7 (64)	−14
Change agent	van Doorn-van Atten et al. (2018), Demark-Wahnefried et al. (2018), Chen et al. (2017), Babatunde et al. (2011), Bradbury et al. (2006), Verheijden et al. (2004)	6 (33)	Smith et al. (2020), Downes et al. (2019), Chiu et al. (2019)	3 (27)	6
Concordance between BCTs and change agents	van Doorn-van Atten et al. (2018), Demark-Wahnefried et al. (2018), Chen et al. (2017), Babatunde et al. (2011), Bradbury et al. (2006)	6 (100)	Smith et al. (2020), Downes et al. (2019), Chiu et al. (2019)	3 (0)	100

Notes: n (%) = indicates an overall number (n) and proportion of BCTv1 features (%) out of total unique studies. In curly brackets are redundant studies with the same intervention protocol across different endpoints. In the case of redundancy, only one study was used for calculations. Differences in frequency (%) indicate the difference between high- and low-retention studies: positive number—higher proportion in high-retention studies; negative number—higher proportion in low-retention studies.

Discussion and Implications

The primary purpose of this scoping review was to map behavior change features such as theories, agents, and techniques used in dietary intervention studies targeting adults aged 60 and older and describe the extent to which these features might relate to the retention of study participants. Our findings showed that approximately half of the studies reported their theoretical basis, and slightly less than one third reported their behavior change agents. Of these studies reporting on theory and agents, the most common were social cognitive theory and the related mechanism of self-efficacy. The most common techniques were in the form of instructions on how to perform the behavior (“shaping knowledge” BCTv1 cluster) and helping participants to set and act upon their goals (“goals and planning” BCTv1 cluster). We observed some differences with respect to specific BCTv1 clusters which were more common in interventions with higher versus lower retention. Specifically, BCTv1 that focused on the physical environment and specific and nonspecific rewards was more common in higher retention protocols. We also found that studies with higher retention were more likely to have a good conceptual agreement between BCTs and change agents.

In our findings, theory was sparsely used to guide dietary interventions, which is consistent with other reports indicating insufficient theorization of behavior interventions. For example, another review that investigated the extent of theory use in physical activity and dietary interventions revealed that only half of the interventions reported their theory base (Prestwich et al., 2014). As the use of theory to inform behavior change interventions may be associated with stronger intervention effects (Taylor et al., 2012; Webb et al., 2010), albeit inconsistently (Prestwich et al., 2014), it is important to consider strengthening the theoretical foundation of studies. Furthermore, using and reporting theory should not occur in isolation and should explicitly inform the selection of behavior change mechanisms and techniques. Our observation that less than half of the studies that specified theories also reported theory-based behavior change agents is also alarming. Without a clear indication of how change agents are measured, basing an intervention only on theory may miss an opportunity to examine whether the intervention worked as intended by engaging its prespecified behavior targets (Connell et al., 2019). As the National Institute on Health currently calls for an increased focus on mechanisms of change (Riddle & Science of Behavior Change Working Group, 2015), future studies should explicitly specify behavior change agents and evaluate whether interventions activated their putative mechanistic targets.

We also found that only one third of potential individual BCTs were used across 43 included studies. Although these findings might suggest a relatively narrow scope of intervention strategies deployed in dietary intervention for older adults, the estimate is on par with other reports. For example, Ashton et al. (2019) indicated that slightly more than half of 93 potential BCTs were deployed in dietary interventions for young adults. Lara et al. (2014) reported that only 28 of the 40 BCTs listed in the CALO-RE taxonomy (predating the BCTv1) were identified in dietary intervention for adults of retirement age. Similar to our findings, Ashton et al. reported that “shaping knowledge” was the most common, and “goals and planning” was the second most common BCTv1 cluster. Ashton et al. found that “natural consequences” (BCTs that provide information about health consequences) ranked in third place, whereas, in our list, this technique was less common. This might suggest an intriguing insight to differences in how we prioritize providing information on health consequences to younger rather than older age groups, in spite of growing clinical evidence of health benefits of improved diets in older populations (Bandayrel & Wong, 2011).

Another important insight is that there was no evidence for the use of several BCTv1 clusters, such as “regulation,” “scheduled consequences,” and “covert learning,” even though they appear to be relevant for dietary interventions. For example, BCTs such as “rewarding completion” (an aspect of “scheduled consequences”) entail shaping behavior by rewarding progressive steps of behavior change and might be well poised to target and reinforce desired dietary behavior. In such instances, a participant identifies the desired reward for dietary behavior change, such as watching a favorite television show. This reward is first linked to a smaller goal, such as eating one fruit or vegetable per day; the reward is then made contingent on more complex or larger goals, such as eating three servings of vegetables a day or replacing one’s usual hamburger with a salad and lean protein for lunch. Likewise, techniques to improve “regulation” such as “conserving mental resources” may help promote dietary change by minimizing the cognitive burden of behavior change, such as in the case of using an electronic food diary or app for calorie counting. Finally, “imaginary reward and punishment” (included in “covert learning” BCTv1) might be used to simulate scenarios concerning wanted or unwanted behaviors. These BCTs have individuals imagine either a desired or undesired behavior, followed by a reward or adverse consequence, respectively. For example, an intervention using imaginary reward might have individuals imagine themselves replacing their usual high-sugar colas with water during their daily lunch and dinner and then visualize themselves checking their blood glucose levels after a week or month of this behavior change and being “rewarded” with lower, more consistent readings and feeling proud of their efforts. This underutilization of these BCTs may be at least partially explained by the largely low-tech nature of dietary interventions that historically rarely involved digital solutions. In fact, a recent scoping review of mobile health (mHealth) behavior interventions for older adults noted that just a small fraction of the reviewed studies targeted diet (Zaslavsky et al., 2020). Contemporary mHealth interventions that feature just-in-time feedback, quick access to a vast array of information, personalization, and interactivity might be leveraged to diversify currently deployed BCTs. We suggest that future studies enrich the portfolio of BCTs applied in dietary interventions so as to engage other behavior change mechanisms.

With regard to participant retention, there is an emerging consensus that retention could be shaped by behavioral factors (Duncan et al., 2020). As such, retention could be investigated through the behavior change perspective that involves theory, agents, and BCTs. Our results showed studies with high retention seemed to report more frequent use of “antecedent” and “reward and threat” and less frequent use of “goals and planning” and “social support” (mainly unspecified social support) BCTv1 clusters. Our finding that high retention studies often used “adding objects to the environments” (a type of antecedent, in which participants are given a tool or type of food to promote behavior change) or specific or nonspecific rewards was supported by a previous study that demonstrated monetary incentives were the only behavior technique linked to improved retention (Duncan et al., 2020). Even though these BCTs differ—monetary incentives reward behavior, while antecedents prompt behavior—both emphasize behavioral determinants or consequences of participation. Another intriguing finding is that unspecified social support was more common in studies with low versus high retention. This observation is in sync with Coday et al., which showed that retention benefits from instrumental but not general social support (Coday et al., 2005). Finally, having a good conceptual alignment between BCTs and change agents accounted for the largest positive difference in frequency between high- and low-retention studies. Pending further investigation, one cautious interpretation is that mechanistically concordant studies with BCTs that involve resource allocation and positive reinforcement through specific and nonspecific rewards may be more advantageous for retention in this age group than BCTs that involve general goal planning and social support. This assertion, however, needs further empirical validation.

Limitations

Our scoping review should be interpreted in light of limitations. First of all, as is the case of a scoping rather than a systematic review, some of the included studies had lower methodological quality. That is because the scoping review aimed to map rather than evaluate the strength of evidence from a diverse body of literature (Arksey & O’Malley, 2005). In some cases, it was challenging to identify BCTs for poorly described interventions. For instance, although interventions described providing education, some failed to specify the exact educational content. Coding the insufficiently described interventions was challenging. In other instances, techniques were imperfectly matched to the BCTv1 taxonomy. Hence, we had to approximate the most relevant BCTs. The limitations of the BCT mapping were previously reported (Duncan et al., 2020; Kebede et al., 2017). Another limitation is that, in case of multiarm trials, we abstracted BCT from intervention arms. Furthermore, studies might have deployed strategies other than BCTs to optimize retention. For example, Zacharia et al. (2020) evidenced 88% retention in their trial and appeared to have used a user-centered approach to designing and refining their psychoeducational materials. This may have improved the usability of such patient-facing materials, but given the study design (single-arm trial), we are unable to infer the exact role this played in retention. Conversely, van Doorn-van Atten et al. (2019) cited poor usability and low acceptance of their technology-based intervention as a prime driver of participant attrition (55% retention). Participant incentive may also contribute to retention but was not explicitly addressed in many reviewed studies; when it was addressed, we coded it as a BCT accordingly. As we aimed to comprehensively capture any observable, active ingredients within the interventions designed to change behavior, all retention techniques were mapped into the BCTv1 taxonomy. The studies included in this review varied considerably regarding participant characteristics (e.g., robust community-dwelling vs. frail or home-bound older adults) and design, which limits our ability to analyze specific retention techniques above and beyond the contribution of BCTv1 taxonomy. Finally, the differences in BCTs between high- and low-retention studies were based on a cutoff of ≥15% difference, which suggests a nominal but not necessarily statistically significant difference between studies. Nevertheless, a strength of this review is the focus on a wide variety of behavior change features that transcend previous boundaries, inclusion of a new concept of concordance between BCT and change agents, and the study retention-specific context.

Implication for Research and Practice

This is the first study to comprehensively describe behavior change features used in dietary interventions for older adults and link these features to the extent to which the studies retained more than 80% of their participants. We found that slightly less than half of the studies reported their theoretical basis, and only a third reported their behavior change agents. While few BCTs were reported frequently, only one third of 93 total BCTs were identified in the included studies. We also noted that several BCTv1 clusters, such as “antecedents” and “reward and threat,” and conceptual concordance between BCTs and change agents were more common in interventions with higher as compared with lower participant retention rates. On the other hand, BCTv1 clusters such as “goals and planning” and “social support” (mainly unspecified social support) were more common in the studies with participant retention rates of equal or less than 80%.

Based on these observations, we made several calls for action for future experimental, theory-driven, and mechanistically concordant dietary interventions. First and most important, clear reporting of theoretical basis, behavior change agents, and BCTs will serve to enhance replication, strengthen the research base, and assist in the translation of successful interventions into clinical practice. Second, interventions should consider resource allocation and positive reinforcement BCTs to enhance retention. Third, diversifying the portfolio of currently deployed BCT through more extensive use of mobile technologies may enhance dietary interventions.

In summary, this review of behavior change theories, agents, and techniques is a step toward improving the development and implementation of dietary interventions targeting older adults. Specific BCTv1 clusters might be best positioned to enhance not only intervention effectiveness but also retention. Future studies should consider mechanistically concordant BCTs and more thoroughly address drivers of participation retention and how these might extend beyond research settings to the community or clinical practice to ultimately have an impact on behavior and health.

Funding

This work was supported by the National Institute on Aging Mentored Patient-Oriented Research Career Development Award (NIH, NIA K23AG059912-01A1). The National Institute on Aging had no role in the design, analysis, or writing of this article.

Conflict of Interest

The authors declare that this writing was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.