Group-Based Nutrition Interventions to Promote Health and Mobility in Community-Dwelling Older Adults: A Systematic Review

Objective: To identify the ecacy of group-based nutrition interventions to increase healthy eating, reduce nutrition risk, improve nutritional status, and improve physical mobility among community-dwelling older adults. Design: Systematic review. Electronic databases MEDLINE, CINAHL, EMBASE, PsycINFO, and Sociological Abstracts were searched on July 15, 2020, for studies published in English since January 2010. Study selection, critical appraisal (using the Joanna Briggs Institute’s critical appraisal tools), and data extraction were performed in duplicate by two independent reviewers. Setting: Nutrition interventions delivered to groups in community-based settings were eligible. Studies delivered in acute or long-term care settings were excluded. Participants: Community-dwelling older adults aged 55+ years. Studies targeting specic disease populations or promoting weight loss were excluded. Results: Thirty-one experimental and quasi-experimental studies with generally unclear-high risk of bias were included. A broad range of interventions were identied, including nutrition education with behaviour change techniques (e.g., goal setting, interactive cooking demonstrations) (n=21), didactic nutrition education (n=4), interactive nutrition education (n=2), food access (n=2), and nutrition education with behaviour change techniques and food access (n=2). Group-based nutrition education with behaviour change techniques demonstrated the most promise in improving food and uid intake, nutritional status, and healthy eating knowledge compared to baseline or control. The impact on mobility outcomes was unclear. Conclusions: Our ndings should be interpreted with caution related to generally low certainty, unclear-high risk of bias, and high heterogeneity across interventions and outcomes in this body of literature. Quality research in group-based nutrition education for older adults is needed.


Introduction
Older adults are the fastest-growing age group, and the number of adults aged 65 years and older worldwide is expected to more than double from 727 million in 2020 to over 1.5 billion in 2050 (1) . As the population ages, the prevalence of chronic diseases, multimorbidity, and frailty will also increase (1)(2)(3) . Several modi able risk factors are associated with increased risk of disability and disease with aging, one of which is poor diet quality (4,5) . Unfortunately, many older adults do not meet current age-speci c nutrition guidelines (6,7) concerning both diet quality and quantity (8,9) . As individuals age, many decrease their total food intake (10) , in part due to reduced appetite, sensory impairment, hormonal imbalance, and changes in the gastrointestinal tract and dentition (11) . Agerelated changes in living situations, retirement, social isolation, and loss of relationships can also negatively impact food intake and diet quality (12,13) . The intersection of nancial, psychosocial, environmental, physical, cognitive, gender, and cultural factors are known to in uence eating behaviour (13) , food access (14) , and mobility (15) among older adults.
The relationship between mobility (the ability to move oneself within the immediate environment and broader community) (15) and nutrition has been shown to be bidirectional in older adults. One's mobility can impact food access (e.g., ability to transport oneself to locations with high quality food sources) (16) and may also be in uenced by dietary quality. Reduced intake of both micronutrients and macronutrients may lead to sarcopenia (17)(18)(19) , and the loss of muscle mass in aging may result in mobility limitations and impaired quality of life (20) . Proper nutrition also plays an important role in maintaining skeletal strength and preventing falls and chronic diseases among older adults (20)(21)(22) . Given this, promoting healthier eating and reducing nutrition risk is necessary to maintain and improve health and mobility among community-dwelling older adults. However, many older adults perceive functional decline as an inevitable part of aging and may experience di culties accessing available programs and services (23) .
Group-based nutrition interventions, including education, interactive discussion, and hands-on activities, have demonstrated bene t in supporting older adults to learn from each other's knowledge and experiences, overcome psychosocial and environmental barriers to healthy eating, enhance motivation, and promote dietary behavioural change (24)(25)(26) . Group-based interventions among older adults also foster a sense of group cohesion (27) , allowing individuals to feel acknowledged and form bonds with others who understand their experiences rsthand. Although many group-based nutrition interventions exist, some of which have been formally evaluated for effectiveness (28,29) , these interventions vary widely, and optimal design features remain unclear.
The Enhancing physical and community MoBility in OLDEr adults with health inequities using commuNity co-design (EMBOLDEN) trial is a multi-year research program in Hamilton, Ontario, Canada that aims to promote mobility among older adults residing in areas of high health inequity (30) . EMBOLDEN uses a community co-design approach, partnering with older adults and service providers to combine the best available research evidence with local knowledge to develop an integrated service delivery model. To inform the co-design of a novel, group-based intervention targeting physical activity, nutrition, and system navigation, the EMBOLDEN team conducted an umbrella review (31) to identify existing synthesized evidence regarding group-based physical activity and/or nutrition interventions for community-dwelling older adults. The team identi ed 54 systematic reviews; 46 were reviews of physical activity interventions alone, and eight combined physical activity with nutrition (namely protein supplementation). No systematic reviews of group-based nutrition interventions alone were identi ed, highlighting a lack of synthesized evidence to identify intervention effectiveness. This understanding is key to informing the development and implementation of evidence-informed community programs to promote health and mobility among older adults.
To address this gap, our team initiated a systematic review of single studies focused on group-based nutrition interventions targeting healthy eating in community-dwelling older adults. We speci cally aimed to address whether group-based interventions targeting healthy eating in community-dwelling older adults (≥ 55 years) improved access to nutrition, affected nutritional intake, or changed markers of physical mobility.

Methods
This systematic review was registered with PROSPERO (CRD42020205045). The reporting of this review is based on PRISMA guidelines (32) .

Search Strategy
The electronic databases MEDLINE, CINAHL, EMBASE, PsycINFO, and Sociological Abstracts were searched to July 15, 2020, by a research librarian trained in building search strategies for systematic reviews (Supplementary Tables 1-5). To focus on interventions germane to the current context and nutrition guidelines, database searches were limited to studies published from January 2010. Only English language studies were eligible due to the research team's capacity. Reference lists of all identi ed systematic reviews were screened for potentially relevant and eligible studies; experts in the eld were contacted to locate any additional studies not identi ed in our search.

Study Selection
Citations were uploaded into Covidence (Vertitas Health Innovation Ltd., Melbourne, Australia), and duplicates were removed. Following a pilot test, titles and abstracts were screened by two independent reviewers against predetermined eligibility criteria. Full texts of potentially relevant studies were retrieved and screened for eligibility by two independent reviewers. Disagreements were resolved through discussion or with the input of a third reviewer.

Types of Studies
This review included experimental and quasi-experimental study designs, including randomized controlled trials (RCTs), non-randomized controlled trials, before and after studies, and interrupted time-series studies. Mixed methods studies with quantitative designs cited above were also included, although only quantitative data were extracted and analyzed. Theses and dissertations were eligible; publication status was not a criteria for inclusion. Conference abstracts, reviews, observational designs, and qualitative studies were excluded.

Participants
Eligible studies must have included community-dwelling older adults ≥ 55 years old or reported a mean age of participants as ≥ 55. Studies focused on disease-speci c populations were excluded, although included participants could report risk factors for or the presence of chronic diseases.

Interventions
Studies that evaluated group-based interventions targeting healthy eating were eligible. Examples of modes of delivery included interventions based on nutrition, education, gardening, and congregate dining. If studies reported on interventions with multiple delivery modes, only group-based interventions were extracted and analyzed. Programs focused on weight management or weight loss were excluded. Interventions delivered in any community-based setting were eligible, including seniors' and community centers. Studies that took place in acute or long-term care settings were excluded.

Comparators
Studies that compared an intervention to any comparison group (including single group pre-test/post-test) were eligible. Examples of comparator groups included pre-intervention, other intervention, or non-exposed control groups.

Outcomes
Studies that reported on a change in nutrition outcomes from pre-to post-intervention were eligible for inclusion. Nutrition outcomes were grouped retrospectively into three categories based on study ndings: 1) food and uid intake (e.g., vegetables and fruit, whole grain foods and protein), 2) nutrition risk, de ned as factors that impact food intake (33) (e.g., dietary habits, food access), and 3) healthy eating knowledge (e.g., nutrient functions, recommended servings). Physical mobility outcomes were considered secondary outcomes and were retrospectively grouped into two categories: 1) physical activity, and 2) functional outcomes (e.g., Timed Up and Go (TUG) test, gait speed).

Assessment of Methodological Quality
Two independent reviewers critically appraised the eligible studies for methodological quality using the Joanna Briggs Institute critical appraisal instruments for experimental or quasi-experimental studies (34) . Overall scores for each study were calculated by responses to the questions. Any disagreements between reviewers were resolved through discussion or input from a third reviewer.

Data Extraction
Two independent reviewers performed data extraction using a pre-developed and tested data extraction form. This form included general study information (i.e., study aim, design, country, start/end dates), population (i.e., age, sex, number of participants, ethnicity, socioeconomic status), intervention details (including duration, frequency, who delivered, how it was delivered, where it was delivered, and theoretical framework, with questions framed according to the Template for Intervention Description and Replication (TIDieR) checklist and guide (35) ), comparison groups, limitations, and conclusions reported by study authors. Relevant nutrition and mobility outcomes were also extracted for all time points reported in the individual studies. When measures of overall food and uid intake were reported (e.g., Food Frequency Score, Dietary Variety Score), these were extracted over speci c food group intake results. Any disagreements between reviewers were resolved through discussion or by a third reviewer. Data collection forms and extracted data used for analyses are available upon request.
A meta-analysis was not possible given the variation in intervention types and outcomes across included studies. A narrative approach was used to synthesize included studies (36) , with data summarized and presented in supporting tables. Results tables with effect size measures, including mean differences, odds ratios, effect sizes, and proportional changes, were structured by intervention category and outcome measures to explore variation and possible sources of heterogeneity. When only pre-test/post-test means or percentages were reported, mean or percent differences between groups were calculated. When missing, mean differences, con dence intervals, and/or standard deviations of the changes were calculated using accepted equations (37) and RevMan software (38) . A correlation coe cient of 0.5 was estimated for both food and uid intake outcomes (39)(40)(41)(42)(43) and physical activity outcomes (44)(45)(46) , based on available literature. Reporting bias was not explored as most studies did not cite a protocol or trial registration. Sensitivity analyses were not performed. A comprehensive approach to assess the overall certainty of the evidence for each outcome was not used due to high heterogeneity across interventions and outcomes.
Five studies showed improvements in some but not all aspects of food and uid intake following nutrition education with BCT, as ndings were inconsistent across outcomes (52,53,60,76,77) . Physical activity and nutrition education with goal setting and skill-building components increased the percentage of participants meeting recommended fruit intake, but not other food groups and macronutrients, as compared to no intervention (low risk of bias) (60) . Nutrition education and culturally tailored lifestyle program incorporating goal setting, action planning, and hands-on activities increased the number of participants consuming ≥ 3 meals/day and decreased fried food consumption, but also decreased vegetable intake and found no change in fruit intake as compared to baseline (low risk of bias) (52) . Eight weeks of bi-weekly drama-style lectures, food tasting, and group discussion improved dietary variety compared to control, but inconsistent ndings were noted for macronutrient consumption (low risk of bias) (76) . Two studies evaluated the effects of the Texercise Select intervention (10-weeks of twice-weekly education, physical activity, goal setting and action planning). In the rst study, Texercise Select increased the likelihood of vegetable and fruit consumption and decreased the likelihood of fast-food intake but did not change soda or water consumption compared to a non-randomized waitlist control (moderate risk of bias) (77) ; improvements were not sustained at six month follow up. Texercise Select improved vegetable, fruit, and water consumption but not soda and fast food consumption compared to baseline in the second study (moderate risk of bias) (53) .
Although heterogeneity across interventions was evident, similar nutrition education with BCT interventions were used in four studies (low to high risk of bias) that found no signi cant changes in food and uid intake (57,59,66,70) . Didactic nutrition education (56,68,74) , interactive nutrition education (48,62) , and food access (47,54) interventions alone did not appear to change food and uid intake for the better. Only one study evaluated a nutrition education with BCT and food access (take-home meal portions) intervention and found improved consumption of recommended portions of all food groups (moderate risk of bias) (50) .

Nutrition Risk
Nine studies evaluated the effectiveness of nutrition education with BCT (n = 7, 78%) (49,57,58,63,64,72,73) , food access (n = 1, 11%) (54) , and nutrition education with BCT and food access (n = 1, 11%) (69) for decreasing nutrition risk. Measures such as the Mini Nutritional Assessment (n = 3, 33%) (54,64,69) , Dietary Screening Tool (n = 2, 22%) (58, 73) , problematic dietary habits (n = 2, 22%) (49,72) , and food security (n = 1, 11%) (58) were used (Table 3). Heterogeneous interventions and outcomes and inconsistent results were found. Note: Bold text indicates statistical signi cance. C = comparator group; CCAA = Canadian Center for Activity and Aging; CI = con dence interval; FV = fruits and vegetables; I = intervention group; MD = mean difference; NIA = National Institute of Aging; NR = not reported; NS = not statistically signi cant; PA = physical activity; SD = standard deviation; wk = week Among seven studies that combined nutrition education with BCT, ve demonstrated consistently positive effects (49,58,63,64,73) . The "I am Active" intervention (twice weekly nutrition sessions including meal planning and goal setting for two months) increased the percentage of participants with "normal" nutritional status (as de ned by the Mini Nutritional Assessment) and decreased the number at risk for malnutrition compared to waitlist control (moderate risk of bias) (64) . Compared to didactic education alone, six-monthly nutrition and health education sessions incorporating goal setting and taste testing decreased nutrition risk (moderate risk of bias) (58) . People who frequently attended Supplemental Nutrition Assistant Program-Education (SNAP-Ed) nutrition education sessions with goal setting and recipe tasting decreased their nutrition risk status as compared to control (moderate risk of bias) (73) ; however, it is important to note that only those attending four or more sessions were included in the analysis. Compared to usual care, weekly nutrition education that incorporated counselling, food preparation, goal setting, and behavioural monitoring improved overall healthy eating scores after three months (moderate risk of bias) (63) .
Finally, interactive nutrition education and skill-building activities also improved personal eating habits as compared to baseline (moderate risk of bias) (49) .
Conversely, two additional studies that combined nutrition education with BCT did not improve nutritional status (moderate to high risk of bias) (57,72) . Two studies (low to moderate risk of bias) evaluated food access or nutrition education with BCT and food access (54,69) ; these did not consistently reduce nutrition risk.

Healthy Eating Knowledge
Five studies reported changes in healthy eating knowledge, generally using study-speci c single-item questions (e.g., roles of nutrients, recommended servings) following nutrition education with BCT (n = 3, 60%) (51,75,76) , nutrition education with BCT and food access (n = 1, 20%) (50) , and didactic nutrition education (n = 1, 20%) (55) . Nutrition education with BCT may improve healthy eating knowledge, as found in four studies (low to moderate risk of bias) that incorporated skill-building activities into nutrition education interventions (50,51,75,76) (Table 4).  Study ID Description of Intervention/ Comparator Data Collection Tool Outcome Effect size (95% con dence interval or standard deviation, pvalue)

Risk of Bias
Note: Bold text indicates statistical signi cance. C = comparator group; CI = con dence interval; I = intervention group; MD = mean difference; PA = physical activity; SD = standard deviation Physical Mobility Outcomes

Discussion
Given the wide heterogeneity and inconsistent ndings across this body of literature, our certainty in the effectiveness of group-based community nutrition interventions to improve food and uid intake, nutritional status, healthy eating knowledge, and measures of physical activity or physical function in older adults is low. The available evidence suggests that nutrition education with BCT may be the most promising approach to improving food and uid intake, nutritional status, and healthy eating knowledge. Given the variation across interventions and outcomes, it is unclear which intervention is optimal for implementation in community-based settings. Both intervention duration and frequency varied widely across studies, with no discernable patterns to suggest a minimally or optimally effective intervention "dose". While one would suspect that longer programs or more frequent sessions would have a greater impact, this did not appear to be the case in the studies included in this review. Overall, these conclusions should be interpreted with caution related to high variability among intervention components and outcome measurements, in addition to unclear to high risk of bias within the studies themselves.
Most of the interventions combined nutrition education with BCT. Although we broadly grouped interventions as either including BCT or not, we did not explicitly code these based on the BCT Taxonomy (78) to identify the discrete strategies used. The effectiveness of nutrition education with BCT, particularly concerning food and uid intake and nutrition risk, remains unclear; there is lack of evidence on which speci c BCT are required to elicit signi cant change. Given wide heterogeneity across intervention components, duration, frequency, interventionists, locations, and theoretical frameworks used, we could not distinguish any noticeable patterns among nutrition education with BCT interventions that were consistently effective versus those that were not. Interventions that described nutrition education with BCT appeared to be more intensive than interventions that focused on didactic or interactive nutrition education alone. However, it is conceivable that individuals who consent to participate in a more intensive program could perhaps be more committed to overall behavioural change. Appropriately selecting and evaluating the effectiveness of BCT remains an emerging area of inquiry (79) ; thus, understanding the most relevant and effective BCT to improve nutrition and mobility outcomes among community-dwelling older adults is an important next step. More fulsome reporting of intervention components following de nitions from the BCT Taxonomy (78) or using a recognized framework such as the TIDIeR checklist (35) would allow future exploration of key intervention components.
We explored physical activity and functional outcomes given the established link between adequate nutritional intake and mobility in older adults; however, all studies that explored mobility outcomes also included a physical activity co-intervention. The existence of a co-intervention made it di cult to determine which component(s) of these multifaceted interventions were driving change when observed. Although we hypothesized that comprehensive healthy lifestyle programs might have a greater impact on behavioural change overall, we did not observe any clear trends to indicate whether the interventions that included both nutrition and physical activity components were more effective than those focused on nutrition alone. There is limited available evidence regarding the effectiveness of single versus multiple health behaviour change interventions in older adults (80) , highlighting a potential area for further investigation (81) .
Given the complex factors (e.g., nancial, environmental, cultural) known to impact older adults' ability to maintain a healthy diet (14) , it is important to recognize that while nutrition education and skill-building may be effective at increasing healthy eating knowledge and intentions, they may be insu cient to change outcomes such as food and uid intake or nutrition risk. Using an equity lens, we assessed the nine studies included in this review that explicitly targeted populations with low socioeconomic status (e.g., recruitment from low-income housing). Overall, ndings were inconsistent, with improvements following education with BCT noted in some but not others. This may not be surprising if the primary barriers to quality food intake (e.g., vegetable and fruit consumption) are cost or ease of access (9) . Environmental support and policy-level public health interventions are likely needed to ensure equitable access to healthy food before nutrition education and skill-building can be expected to make a meaningful difference (82-84) .
To our knowledge, this review is the rst to systematically identify, appraise, and synthesize evidence regarding the effectiveness of nutrition-focused groupbased interventions targeting food and uid intake, nutrition risk, and mobility outcomes in community-dwelling older adults. However, our results are consistent with recommendations from a pair of evidence syntheses and an expert commentary published in 2003 that concluded nutrition education alone was insu cient to improve nutritional status among older adults (24,85,86) . In line with our ndings, the authors recommended that education be paired with behaviour change strategies and community participation to enhance program effectiveness. Similarly, a 2007 review of Canadian research highlighted successful components of community nutrition programs for older adults, including cooking classes, recipe exchanges, counselling, social support and engagement, motivation, and interactivity (87) . Consistent with our ndings, these strategies would also be considered techniques to support behaviour change.
Several important considerations should be made while interpreting the ndings from this review. Although our search strategy was comprehensive, it was limited to studies published in English since 2010. Given the consistency with ndings from older, related reviews, it is unlikely that literature published before 2010 would be of higher methodological quality or change our overall conclusions. Given that the aim of this review was to explore the effectiveness of groupbased interventions, it was appropriate to focus on intervention studies only. Qualitative data may highlight important insights into reasons for variable intervention effectiveness (e.g., implementation insights). While we did include two mixed-methods studies, only quantitative data were extracted. Further, although we did endeavour to integrate considerations about study quality, consistency, and directness throughout, the wide variability in outcomes across included studies limited us from applying a formal approach, such as GRADE (88) to assess certainty in this body of evidence.
Our conclusions are also limited by the nature of the primarily quasi-experimental single studies with incomplete follow-up included within the review. Lack of reliable outcome measurement tools may explain some of the inconsistency across studies. Challenges associated with measuring the impact of community nutrition programs have previously been documented (89) ; given the nature of self-reported data, outcomes such as food intake, dietary behaviour, and knowledge are notoriously complex constructs to measure accurately. Despite previous calls for community nutrition interventions for older adults based on behaviour change theories (86) , less than half of the studies in this review used a theoretical framework to inform intervention delivery; this might further explain some of the variability noted in our results. We also observed variability in the content of the nutrition education provided across interventions. It is unclear if recommendations were consistently based on current, evidence-based healthy eating guidelines for older adults, further explaining the inconsistent effectiveness observed.

Implications for Research
More studies using RCT designs are needed to increase con dence in the impact of group-based community nutrition interventions. Although blinding of participants and interventionists is nearly impossible given the nature of the interventions, future studies should strive to blind outcome assessors and data analysts to enhance internal validity. Authors using quasi-experimental approaches should include control groups to facilitate stronger comparisons. Given that community-based nutrition programming tends to be delivered via public health initiatives and not always through funded programs of research, challenges noted with intervention design, outcome assessment, study quality, and inappropriate statistical analyses might be attributed to the probable lack of resources available to support community program development and evaluation. Prioritizing research funding to support the development and evaluation of community-based nutrition programs for older adults is necessary to improve the quality of the evidence base.

Implications for Practice
For organizations looking to design and implement community-based nutrition programming for older adults, nutrition education with embedded BCT (e.g., goal setting, hands-on skill-building activities, taste testing) demonstrated the most promise to improve healthy eating outcomes. However, there is wide heterogeneity in the available evidence, including program length and session frequency. The discrete techniques and intervention components that might be most important to include have yet to be determined. These will likely need to be tailored based on the needs and preferences of the community and local context. Future program design should be based on recognized theories of behaviour change. There is a potential to draw upon signi cant recent advancements in behaviour change theory (90,91) , which have been applied in developing complex interventions for healthy eating (92,93) .

Conclusion
Group-based nutrition education with BCT demonstrated the most promise in improving food and uid intake, nutritional status, and healthy eating knowledge among community-dwelling older adults. The impact of these programs on mobility outcomes is less clear. These ndings should be interpreted with caution, given the generally unclear to high risk of bias and low quality, heterogeneous evidence base. We have highlighted several key takeaways regarding how the quality of this body of literature could be improved. Future group and community-based programs should use recognized behavioural change theories to develop and implement evidence-based nutrition education with skill-building activities to improve healthy eating among older adults.
Declarations Figure 2 Summary of Risk of Bias in Randomized Controlled Trials Assessed using JBI Critical Appraisal Checklist for Randomized Controlled Trials. Figure 3