Characteristics of Smartphone Applications for Nutrition Improvement in Community Settings: A Scoping Review^,,,

Abstract

Smartphone applications are increasingly being used to support nutrition improvement in community settings. However, there is a scarcity of practical literature to support researchers and practitioners in choosing or developing health applications. This work maps the features, key content, theoretical approaches, and methods of consumer testing of applications intended for nutrition improvement in community settings. A systematic, scoping review methodology was used to map published, peer-reviewed literature reporting on applications with a specific nutrition-improvement focus intended for use in the community setting. After screening, articles were grouped into 4 categories: dietary self-monitoring trials, nutrition improvement trials, application description articles, and qualitative application development studies. For mapping, studies were also grouped into categories based on the target population and aim of the application or program. Of the 4818 titles identified from the database search, 64 articles were included. The broad categories of features found to be included in applications generally corresponded to different behavior change support strategies common to many classic behavioral change models. Key content of applications generally focused on food composition, with tailored feedback most commonly used to deliver educational content. Consumer testing before application deployment was reported in just over half of the studies. Collaboration between practitioners and application developers promotes an appropriate balance of evidence-based content and functionality. This work provides a unique resource for program development teams and practitioners seeking to use an application for nutrition improvement in community settings.

Introduction

The importance of improving nutritional intake to address the global burden of preventable, noncommunicable disease is well known (1). Nutrition improvement is here defined as any attempt to guide an individual’s diet toward semblance with national healthy eating guidelines. The contributors to poor dietary intake extend from individual-level factors through to social, physical, and macrolevel environmental factors (2). Public health nutrition improvement programs in community settings (i.e., for individuals living independently within a community) are often used to intervene at these different levels. However, such programs must overcome a number of unique challenges: for instance, large numbers of people must typically be reached to achieve a population benefit; thus, less-intensive health personnel involvement is essential to balance finite resources. Often the populations with the greatest need are geographically dispersed and hard to reach; because of interventions typically being disseminated across sociodemographic groups, tailoring information to best support behavior change may be difficult. More recently there has been a growing interest in the use of mobile health, in particular the use of smartphone software applications (3), to support nutrition improvement in community settings and overcome these unique challenges (4, 5).

Mobile phone ownership is prevalent and continues to increase in both developed and less-developed nations and across all socioeconomic groups (6, 7). Smartphones are now ubiquitous and increasingly intertwined with daily life (4, 8), primarily owing to rapid advances in the capabilities of applications (8). Fostering sustained user engagement with applications, however, is a constant battle for application developers, and therefore in the context of applications for health behavior change improvements can be short-lived (3, 4, 9). Hingle and Patrick (4) propose that an appropriate benchmark for heath application success might not be frequency or consistency of use, but rather that the application characteristics (that is, application features and content) provide such an engaging user experience that users return to using the application when it is most needed during the iterative process of nutrition improvement. However, guidelines regarding how to achieve this are lacking (9).

An overabundance of applications combined with a lack of subsequent scientific evaluation inhibits researchers and practitioners from determining the best approaches for choosing or developing health applications (4). In particular there is limited information regarding the best application characteristics for disadvantaged populations (5). There have been a number of early systematic reviews that examine the efficacy of health application interventions; however, these either focus on weight loss rather than nutrition improvement (10–14), exclude qualitative and descriptive studies (15, 16), and/or include text messaging and Web-based interventions (10, 12, 13, 16–18), with none specifically relating to community settings. Similarly, scoping reviews on the topic exclusively focus on weight loss (19), include text messaging interventions (8, 20), and have mapped the literature rather than application characteristics. Application content reviews examine features, content, and incorporation of behavioral change theory in commercially available applications but not applications that have been developed or used specifically for nutrition improvement programs (21–29). As such, a map of application characteristics and consumer evaluation of them, particularly incorporating a comparison of different population groups, will be an informative guide for future application development projects and practitioners seeking to choose appropriate applications for use in community settings.

Scoping reviews are well placed to examine research questions that go beyond intervention effectiveness (30) typically used to map the parameters of a particular body of knowledge (31). They are distinct from systematic reviews in that, rather than seeking to answer a specific research question through examination of the outcomes and quality of a narrow selection of studies, scoping reviews explore how a topic has been studied or approached (32). This information is particularly useful in new and rapidly developing research areas where there has been a proliferation of ideas and approaches, and the evidence is complex and heterogeneous (30, 33), as is the case with the topic under investigation here. This work used a systematic scoping review methodology to examine and map the characteristics of smartphone applications intended for nutrition improvement in community settings. The subobjectives were 1) to identify the features, key educational content, and strategic approach or behavioral change theory utilized in smartphone applications to improve nutrition in community settings, 2) to report any information relating to user evaluation of these characteristics, and 3) to map the methods used in consumer testing of these applications.

This work provides a unique resource for program development teams and practitioners seeking to use an application for nutrition improvement in community settings. We present an overview of the methods used for the systematic search and review process, describe the studies that were identified and how their findings relate to each of the above subobjectives, and discuss the key implications for application development teams, practitioners and researchers.

Methods

A scoping review was conducted consistent with the methods described by Arksey and O’Malley (32) and further elaborated by Levac et al. (30) and the Joanna Briggs Institute (33). The research question addressed by the search was “what features, key content, strategic and theoretical approaches, and methods of consumer testing have been used in smartphone applications intended for nutrition improvement in community settings?” Detailed inclusion and exclusion criteria are listed in Table 1. Briefly, any published, peer-reviewed literature reporting on a smartphone application with a specific nutrition improvement focus intended for use in the community setting was considered for inclusion. Reviews and commercial application content analyses were however excluded. Smartphones were defined in line with others (34, 35) and thus excluded personal digital assistants. However, reports not specifying the use of either a smartphone or personal digital assistants, generically using “mobile phone” or “mobile app,” were included.

TABLE 1.

Review inclusion and exclusion criteria

	Inclusion criteria	Exclusion criteria
All reports	Reports including a smartphone application with a nutrition-improvement focus	Reports including an application with no specific nutrition-improvement focus [e.g., fitness applications or dietary tracking only with no within-application feedback or nutrition education delivery (controlled self-monitoring trials excluded)]
	Reports providing data regarding a relevant smartphone application characteristic as identified by the subobjectives	Reports with inadequate descriptions of all application characteristics
	Participants and target users living independently within a community	Participants and target users that are preschool-aged children (aged <12 y) or clinical populations with dietary treatments that extend beyond national healthy eating guidelines (for example, coeliac disease, allergy) but including overweight and obese, hypertensive, cardiovascular disease, and diabetic populations
	Applications intended for use in the community setting	Applications intended for use in a primary, clinical, or acute care setting
	Any published peer-reviewed literature	Review articles, commentaries, and theses or dissertations
	Full reports published in English language after 2008 in which data were collected after 2008	Abstracts or brief reports, those not in English, or those with pre-2008 data collection or publication date
Additional criteria for reports of controlled trials	Reports describing programs using a smartphone application as a primary or complementary program delivery tool	Reports of social media, eHealth, text messaging, or other online programs not specifying the use of a smartphone application
	Reports describing programs with nutrition content delivery via a smartphone application as a primary or complementary component, or studies assessing the impact of dietary self-monitoring by using a smartphone application on dietary improvement	Reports of programs with no nutrition content delivered via the application, in which applications were used to record dietary intake without responsive in-application–generated feedback/education, or in which applications were used only for research data collection without an explicit participant self-monitoring objective
	Reports describing programs delivered in the community setting	Reports describing interventions or programs delivered in a primary, clinical, or acute care setting

Search method

The 3-stage approach to searching for scoping reviews as outlined by the Joanna Briggs Institute was used (33). First-stage, limited searches to define keywords were conducted in MEDLINE, Ovid, and the Association for Computing Machinery Digital Library. Second-stage, fully computerized searches were conducted in 9 databases (listed in Figure 1) up until 9 May 2016. Two concepts were used to structure search queries including smartphone applications (example keywords “mobile apps,” “mobile applications,” “smartphone,” “cell phone,” “mHealth”) and nutrition improvement (example keywords “nutrition,” “diet, food, and nutrition,” “eating,” “food habits,” “health promotion,” “public health”). Search terms were comprehensive and broad to achieve high sensitivity as outlined in scoping review protocols (30, 32, 33). Database-specific terms (for example, MeSH terms) were used where relevant, complemented by keyword searching of the concept terms and synonyms. Pearling strategies (using attributes of relevant articles to develop the search query in a recursive process) were used on retrieved articles to ensure all relevant studies were included. Third-stage searching involved searching reference lists of retrieved articles and review articles for further reports (33). Searches were limited to the years 2008 to present because the Apple App Store was initially released in July 2008 (Google Play in 2012) and to English language only because of practical constraints.

FIGURE 1.

Search process and results. ACM, Association for Computing Machinery; app, application; CINAHL, Cumulative Index to Nursing and Allied Health Literature; IEEE, Institute of Electronics and Electrical Engineers; PDA, personal digital assistant.

Data extraction

Citations were exported to Endnote (Thomson Reuters) for organization. Duplicates were removed and titles and abstracts screened for consistency with inclusion criteria in first-pass screening. Full-text reports were retrieved and screened for inclusion by one reviewer (E Tonkin), with ambiguous reports being sent to 2 other authors (J Brimblecombe and TP Wycherley) for a consensus decision. Articles were initially grouped according to report type: reports of controlled trials and other articles. These categories were further subdivided into the following 4 groups (hereafter referred to as study types) for data extraction: 1) controlled dietary self-monitoring trials (no nutrition education provided in the application), 2) controlled nutrition-improvement trials (nutrition content and education provided in the application), 3) application description papers (with and without uncontrolled consumer usability testing), and 4) qualitative studies informing application development with consumer groups.

Relevant data for each study type were extracted into summary tables by one reviewer (E Tonkin) (Supplemental Tables 1–4). Data extraction was structured around the review subobjectives. Only a brief summary of results relating to study outcomes (for example, weight loss) is reported in summary tables because many included studies were multicomponent, and therefore identifying the sole contribution of applications to these outcomes is impossible (5). Results relating to the evaluation of application characteristics were extracted. Because this is a scoping review, the risk of bias in studies, heterogeneity, and publication bias were not assessed (30, 32, 33).

Multiple reports of the same trials or projects were grouped. Articles from the same or similar research group reporting on application development and then controlled testing (for example, references 36 and 37) were grouped within the relevant controlled trials study type (category 1 or 2 above). For mapping, studies were also grouped into categories based on the target population and aim of the application/program.

Results

Overview

The results of the search and screening processes are presented in Figure 1. Of the 4818 titles identified from the database search, 64 articles were included: 9 reporting dietary self-monitoring trials (38–46), 18 reporting nutrition-improvement trials (36, 47–63), 30 reporting application-development projects (37, 64–92), and 7 reporting qualitative studies with consumers (93–99). Five reports were published in 2016, 16 each in 2015 and 2014, 11 in 2013, and 16 between 2008 and 2012. Once organized according to our study types, these articles reported on 7 controlled dietary self-monitoring trials, 13 controlled nutrition-improvement trials, 21 discreet application-development projects, and 6 qualitative studies with consumers (Supplemental Tables 1–4) (47 studies). Fifteen studies targeted the general adult population, 15 overweight and obese populations, 8 young adults, 6 parents and low-socioeconomic-status (SES) families, and 3 adolescents. Twenty-four studies were aimed at general nutrition improvement, 11 weight loss, 5 grocery-shopping support, 4 food-access support, and 3 parenting support. Five of the applications described were commercial applications (LoseIt!, FatSecret’s Calorie Counter, and The Eatery) (38, 40, 41, 43, 78), whereas the remainder were developed by research teams (hereafter “made to order”). Twenty-three studies were conducted in the United States, 8 in Australia, 6 in the United Kingdom, 2 in Norway, and 1 each in Austria, Canada, Finland, Korea, Malaysia, the Netherlands, New Zealand, and Portugal.

The following sections report the findings for each of the review subobjectives. The reports of application characteristics (features, nutrition content, and strategic approach or theory) are structured as follows: description of characteristic, characteristic mapped by study type, target population and aim of the application or program, and finally relevant consumer evaluation data. A report of the approaches to consumer testing and a summary checklist for application development projects concludes the results section.

Application features

All 7 dietary self-monitoring trials (38–44), 13 nutrition-improvement trials (36, 47, 48, 50–53, 56–58, 60, 62, 63), 21 discreet application-development projects (64, 65, 68, 70, 72–80, 83, 85, 86, 88–92), and 6 qualitative application-development studies included applications with or a discussion of nutrition-relevant features (93, 95–99). Table 2 demonstrates the great diversity in the application features used and lists all features found within the reports. The broad types of features used in applications (hereafter referred to as feature domains) were dietary logging and tracking (for example, detailed dietary diaries), social connectivity (the incorporation of social interaction, for example, connecting with social networks, team participation, and forums), reminder, encouragement and prompt (for example, push notifications and rewards), goal setting and challenge (for example, dietary goal setting), game element (the use of games or game-type features, for example, use of avatars, narrative gaming, and quizzes), environmental support (for example, food and menu suggestions and locating local produce), and application-development features such as multiple interfaces (different application presentations and functions depending on the user) and crowdsourcing (for example, the collection of data, such as food product information from application users).

TABLE 2.

Application features used in included studies, organized by study type¹

Feature	Self-monitoring trials (n = 7)	Nutrition-improvement trials (n = 13)	Application-description articles (n = 21)	Qualitative studies (n = 6)
Dietary logging and tracking
Detailed meal/snack/drink logging by using a database (commercial/custom)	6 (38, 40–44)	5 (36, 50–52, 63)	3 (75, 77, 83)	3 (93, 98, 99)
Real-time/daily tracking
Calorie	5 (38, 40–43)	2 (52, 63)	2 (75, 83)	—
Macro-/micronutrient	1 (41)	—	1 (83)	1 (98)
Real-time/historical graphical report based on food logging	4 (38, 40, 42, 43)	2 (36, 60)	6 (77, 78, 83, 86, 88, 90)	5 (93, 95, 97–99)
Barcode scanner	4 (38, 40, 41, 43)	2 (53, 58)	2 (68, 83)	1 (98)
Simple food group/meal type logging	1 (39)	2 (52, 60)	3 (86, 88, 90)	2 (95, 96)
Store favorite meal combinations and recently used items	1 (42)	1 (51)	1 (83)	—
Diary of food photographs	1 (42)	2 (51, 62)	5 (72, 74, 78, 80, 86)	—
Chronological log of all of a day’s dietary events	—	1 (51)	2 (72, 91)	—
Snack tracking of family members	—	1 (50)	1 (77)	—
Mood and hunger logging	—	—	2 (72, 88)	1 (98)
Visual cues, such as “calories remaining” turning from green to red	—	—	—	2 (98, 99)
Social connectivity
Find and connect with friends	4 (38, 40, 41, 43)	—	—	—
Recipe sharing	4 (38, 40, 41, 43)	—	—	—
Join public groups	3 (38, 40, 43)	—	—	—
Connect with social networking sites	3 (38, 40, 43)	1 (53)	2 (75, 80)	2 (97, 98)
Interact with a professional of choice	1 (41)	—	—	2 (93, 96)
Team/buddy participation	1 (44)	4 (47, 48, 50, 52)	1 (77)	—
View of team members’ self-monitoring adherence	1 (44)	1 (50)	—	—
Peer-to-peer messaging	1 (44)	3 (48, 50, 52)	2 (77, 86)	—
Interactive team activity feed	—	1 (48)	—	—
Automatic upload of people who met daily challenges to private Facebook group	—	—	1 (80)	—
Option to follow other users, “like,” and comment	—	—	1 (78)	—
Optional forums/chat rooms	—	—	—	2 (93, 98)
Reminders, encouragement, and prompts
Noncustomizable push notifications (prompts/reminders) (hourly, twice daily, daily)	4 (38, 40, 42, 43)	1 (52)	6 (65, 68, 70, 78, 86, 88)	1 (98)
Motivational push notifications
Generic	1 (42)	1 (57)	—	2 (93, 98)
Tailored to self-reported behavior	1 (39)	4 (36, 52, 56, 60)	—	1 (96)
Virtual rewards (badges, stars)	—	3 (48, 56, 60)	1 (86)	1 (98)
Customizable automated reminders
Time based	—	3 (47, 56, 60)	2 (72, 80)	3 (93, 98, 99)
Location based	—	1 (47)	—
Prompting to resume engagement after periods of inactivity
Full-application-interruption	—	1 (52)	—	—
Peripheral	—	2 (52, 60)	—	—
Real rewards (cartoons, games, blogs)	—	1 (52)	—	—
Goal setting and challenges
Join public challenges	4 (38–40, 43)	—	—	—
Automated graphical and text-based reports of progress toward challenges/goals	3 (39, 41, 44)	4 (47, 48, 52, 56)	1 (74)	—
Dietary goal setting	2 (41, 44)	2 (52, 56)	3 (74, 86, 90)	2 (93, 98)
Start a dietary challenge
Predefined challenges	1 (39)	2 (47, 48)	1 (80)	—
Customizable challenges	—	1 (48)	1 (80)	—
Historical record of completed challenges	—	—	1 (80)	—
Game elements
Use of avatar or virtual pet	—	3 (50, 62, 63)	1 (89)	—
Narrative-based game using points	—	2 (50, 62)	1 (77)	—
Food games
Quiz	—	2 (52, 63)	4 (70, 79, 85, 86)	1 (95)
Word search/jumble	—	—	1 (74)	—
Competition aspect	—	2 (47, 50)	1 (77)	1 (93)
Virtual-reality short game	—	—	2 (79, 89)	—
General gamification	—	—	—	1 (98)
Environmental supports
Food/menu and recipe suggestions	—	2 (36, 60)	7 (65, 68, 73, 76, 86, 89, 91)	2 (95, 96)
Detailed food product information (price, nutritional, health claims, rating)	—	2 (53, 58)	4 (64, 73, 83, 92)	—
Budgeting supports (simple fresh-produce cost calculator/shopping list)	—	—	3 (68, 77, 91)	—
Locate inexpensive and fresh local produce	—	—	2 (65, 73)	1 (96)
Augmented-reality colored tagging of food products in shopping aisles	—	—	1 (92)	—
Video-based food recognition	—	—	1 (64)	—
Customizable nutritional combinations used for product information/tagging	—	—	1 (92)	—
Tailoring to local cuisine	—	—	1 (79)	1 (97)
Application development
Crowdsourcing function enabling users to assist with improving breadth of back-end database	—	2 (53, 58)	—	—
Multiple interfaces/modes (e.g., care giver, child, shopping, logging)	—	1 (50)	2 (77, 83)	—

¹Values are n (reference numbers).

Food logging, either detailed meal entry or simple food group logging, was the most frequently incorporated feature, used in just over half of the studies (24 of 47). Push notifications (messages sent to the user through the application), customizable and noncustomizable, were the second most common feature (used in 20 of 47 studies) and were often used to prompt food logging. Other common features were automated graphical reports (graphs or images of food intake) based on food logging (17 of 47), real-time calorie and nutrient tracking (10 of 47), dietary goal setting (9 of 47), barcode scanning (9 of 47), and the ability to connect with social networking sites (8 of 47).

Study type

Table 2 demonstrates that different feature domains were emphasized between study types. No dietary self-monitoring trials and only 4 nutrition-improvement trials used applications incorporating environmental support features, although these were frequently emphasized in application-development projects (12 of 21). Conversely, social connectivity features were emphasized in dietary self-monitoring trials and not commonly found in nutrition improvement trials or application-development projects. There were no major differences in the main feature domains used between single- and multicomponent trials (self-monitoring or nutrition-improvement), except that reminders, encouragement, and prompts were more likely to be features incorporated into applications used in multicomponent interventions, particularly in nutrition-improvement trials.

Target population

The feature domains used in applications varied by target population. Game elements were incorporated into applications targeting low-SES families (5 of 6) and young adults (4 of 8) more commonly than overweight and obese (3 of 15) or general adult (2 of 15) populations. The type of game also differed, with narrative style gaming with points only used in adolescent or low-SES populations and quiz type games used in adult populations. Environmental support features were more common in applications for low-SES (5 of 6), general (8 of 15), and young-adult (4 of 8) populations than for adolescent (0 of 3) and overweight and obese (2 of 15) populations. The remaining feature domains were used with similar frequency across target populations. Applications targeting low-SES families generally had the greatest variety of feature domains used, with those for adolescent populations the least.

Aim of the application or program

The feature domains emphasized in applications also varied by aim of the application or program. Applications used in programs to support general nutrition improvement or weight loss commonly used dietary logging and tracking features (20 of 24 and 10 of 11, respectively), although these features were not as commonly utilized in applications aimed at supporting grocery shopping (1 of 5) and food access (1 of 4). Conversely, all applications supporting grocery shopping, food access, and parenting practices used environmental support features (5 of 5, 4 of 4, and 3 of 3, respectively), although these features were less commonly found in general nutritional improvement (5 of 24) and weight-loss applications (2 of 11). Because applications to support grocery shopping and food access had very specific aims, they typically had the least variety in feature domains incorporated, whereas applications to support parenting and general nutrition improvement typically incorporated the greatest variety of domains.

Evaluation of features

Results relating to consumers’ evaluation of application features were found to be predominantly concerning application function, interface (application presentation), tailoring, food logging, games, and social connectivity.

Application function, interface, and tailoring.

Interactive features, an attractive user interface, and nonrepetitive images and colors were particularly important to low-SES and young-adult populations for maintaining application engagement (73, 74, 77, 95, 98). Separate interfaces for care givers and children were valued by care givers but not children (50, 77). The importance of the application being functional without an internet connection was expressed by all population groups (36, 40, 45, 68, 90, 95, 98), with data access particularly limited for adolescents (40). Excessive data entry for setup and logins was reportedly undesirable (36, 93, 98, 99). Crowdsourcing data from users was useful for providing a comprehensive back-end food database (53, 58) and tailoring feedback on meal healthiness (62). Within nutrition improvement trials it was found that staggered deployment of features (rolling out new features over time) helped maintain participant engagement with the application (52).

Tailoring and personalization of the application and its features were considered important or desirable across all population groups. Although push notifications were valued, consideration of their timing was important, with user customization ideal (56, 65, 74) particularly in adolescents (57). Personalized settings and the ability to turn off features were considered important across population groups (74, 90, 93). Personalization of goals and challenges, and feedback based on these, was considered important by overweight and general adult populations (39, 41, 74, 80, 88). Similarly, using culturally relevant and favorite foods within games was desirable for low-SES populations (50, 79, 89) and enabled transfer of learning to real life (85). Finally, information about local and seasonal foods was highly valued, although recipe suggestions were not (65).

Food logging, games, and social connectivity.

Achieving a balance of simplicity and detail in food logging was important for all populations, with manual entry considered boring and burdensome (51, 56, 90, 93, 97). Autocomplete functions (41, 93, 98), crowdsourcing-based semi-automated approaches (51), barcode scanners and drop-down menus (51, 56), and comprehensive (exact products and brands) databases improved logging (93, 98). Additionally, reports of progress based on logging commenting on overall diet quality, rather than just calorie tracking, were highly valued (38, 43, 93, 95, 97–99). Narrative game-style applications (typically games with a character and storyline) were more appealing to adolescents than adults, and emotional and social realism were important for motivation (50, 62, 77, 85, 89). Adults and overweight populations generally valued quiz-style games if they were quick to play and included incentives such as real rewards (70, 74, 85). Competition, team participation, and social interaction motivated engagement with applications across all population groups (48, 50, 77, 78, 80, 90, 99), but negative scoring of points was not well understood by low-SES populations (77). Young adults in particular emphasized that social networking and sharing must be voluntary and not automatic (65, 93, 98, 99).

Nutrition content

No dietary self-monitoring trials, all 13 nutrition improvement trials (36, 47, 48, 50–53, 56–58, 60, 62, 63), all 21 discreet application development projects (64, 65, 68, 70, 72–80, 83, 85, 86, 88–92), and 4 of the 6 qualitative application-development studies included applications with or a discussion of educational nutrition content to be delivered via an application (93, 95, 97, 98). Table 3 lists all the nutrition content areas targeted in the applications described in the reports and the ways in which this information was delivered. The broad nutrition content delivered by applications covered food composition (targeted in 32 of 38 studies), diet-health links (13 of 38), and feeding practices (5 of 38). The broad education delivery strategies were feedback on activity (22 of 38), general information (5 of 38), nutrition tools (6 of 38), and environmental supports (11 of 38).

TABLE 3.

Nutrition content delivered through applications in included studies, organized by study type¹

Nutrition content	Nutrition-improvement trials (n = 13)	Application-description papers (n = 21)	Qualitative studies (n = 4)
Content areas
Food composition
Decreasing undesirable nutrients (sugar, saturated fat, sodium)	7 (36, 50, 53, 56–58, 62)	4 (77, 80, 83, 92)	—
Specific macro-/micronutrients	4 (36, 51, 53, 58)	7 (74, 76, 80, 83, 85, 91, 92)	—
General healthy eating/nutritional knowledge	4 (47, 52, 60, 63)	6 (68, 70, 73, 78, 85, 90)	1 (93)
Food groups general	2 (36, 48)	2 (80, 89)	—
Food group balance	—	2 (85, 86)	1 (97)
Portion sizing	—	2 (68, 86)	—
Glycemic index	—	1 (80)	—
Relative healthiness of meals (ingredients, preparation)	—	1 (79)	—
Diet-health links
Diet-disease link	2 (47, 57)	4 (65, 74, 75, 80)	1 (98)
Eating for life stage	—	2 (68, 76)	1 (95)
Information about specific diets (e.g., Mediterranean)	—	2 (64, 72)	—
Consequences (negative and positive) of dietary choices	—	1 (65)	—
Dietary strategies for enhancing nutrient absorption	—	1 (74)	—
General importance of healthy eating	—	1 (73)	—
Feeding practices
Meal planning	1 (48)	1 (75)	—
Eating slowly and mindfully	1 (48)	1 (88)	—
Eating breakfast	1 (62)	—	—
Time-management tips	—	1 (65)	—
Feedback
Tailored feedback [healthiness (e.g., star rating)/hints/tips/strategies] based on food logging/scanning ± anthropometrics/life stage	7 (36, 50–53, 58, 62)	5 (75, 77, 83, 86, 88)	4 (93, 95, 97, 98)
Alternative, healthier food suggestions based on logging/scanning	5 (50, 51, 53, 58, 62)	1 (77)	—
Questions posed based on food logging	2 (56, 60)	—	—
Within-game feedback/education on answers	1 (63)	4 (70, 79, 85, 86)	—
Comparison of healthiness of food logging with other users	1 (50)	1 (78)	—
Information about affective and/or physiological consequences of dietary choices based on food logging	—	1 (77)	1 (98)
Crowdsourced feedback on healthiness of food logging	—	1 (78)	—
Game characters react to food logging	—	1 (77)	—
General nutritional information
Links to additional information	2 (48, 60)	1 (65)	1 (95)
Answers to frequently asked nutritional questions	1 (60)	—	1 (95)
Dietary guidelines (tailored to life stage)	1 (36)	—	1 (95)
Nutritional tools
Practical dietary examples of nutrients	2 (48, 56)	3 (74, 76, 85)	—
Portion-sizing guide with photographs	—	2 (76, 86)	—
Meal balance wheel	—	1 (86)	—
Environmental supports
Food product recommendations based on dietary guidelines/user-defined criteria	2 (53, 58)	4 (64, 65, 91, 92)	—
Strategies/plans/tips for helping children eat more vegetables	—	2 (68, 89)	—
Information about local dietitians, weight-loss programs	—	1 (86)	—
Warnings about high-level energy or fat of scanned foods	—	1 (83)	—
Context triggered advice/intervention (e.g., based on location)	—	—	1 (93)

¹Values are n (reference numbers).

Study type.

General healthy eating and diet-health links were the only content areas targeted across all study types. All nutrition-improvement trials provided content about food composition (13 of 13), but single-component trials were more likely to combine this with education about feeding practices (2 of 7), whereas multicomponent trials added diet-health link information (2 of 6). Education about diet-health links was more commonly provided in application-development projects (9 of 21) and qualitative studies (2 of 4) than with nutrition-improvement trials (2 of 13).

The provision of tailored feedback about food logging or barcode scanning in the form of healthiness ratings (traffic light or star ratings), tips or personalized strategies, and links to additional information were the only content delivery strategies to be used across all study types. Tailored feedback was the delivery strategy most used across all study types but was less commonly used in application-development projects (9 of 21) than with nutrition-improvement trials (9 of 13) and qualitative studies (4 of 4).

Target population.

The nutrition content areas addressed by applications varied little depending on the target population. However, applications targeting adolescents and young adults were more likely to incorporate diet-health link education (1 of 2 and 5 of 7, respectively) than those targeting overweight (2 of 10) or low-SES (1 of 6) populations.

Similarly, the education delivery strategies used did not substantially vary between different target populations. Feedback was used slightly more frequently in applications for overweight (8 of 10) than for general adult populations (6 of 13), whereas environmental support strategies were more common in applications targeting low-SES (3 of 6) and general adult (7 of 13) populations than toward overweight (0 of 10) and young-adult (1 of 7) populations.

Aim of application or program.

There was some variation in the nutrition content areas addressed between studies with different aims. In addition to food composition education, applications supporting general nutrition improvement, grocery shopping, and food access more commonly included diet-health link education (8 of 20, 2 of 5, and 2 of 4, respectively) compared with applications supporting weight loss (1 of 6) or parenting practices (0 of 3). Similarly, applications supporting general nutrition improvement and food access were the only types to provide education regarding feeding practices (4 of 20 and 1 of 4, respectively).

Because of their specific aims, applications supporting grocery shopping and food access most commonly delivered nutrition content through environmental supports (5 of 5 and 2 of 4, respectively) and less so through feedback (2 of 5 and 1 of 4, respectively). Conversely, general nutrition-improvement and weight-loss applications relied more heavily on feedback strategies (12 of 20 and 5 of 6, respectively) than on nutrition tools (4 of 20 and 1 of 6, respectively) or environmental supports (2 of 20 and 0 of 6, respectively).

Evaluation of content.

Few reports described any consumer evaluation of the nutrition content delivered by applications. The evaluation that was reported pertained only to content tailoring and presentation. The tailoring of educational information and environmental supports, such as food product recommendations, to match the nutritional needs of subpopulations (age, life-stage, SES, cultural heritage, and disease status) was considered important for general (51) and young-adult populations (95, 97). Regarding tailored feedback, both positive and negative reinforcement were considered motivating by overweight (90) and young-adult populations (99), although young adults were clear this should not feel like “telling off” (93). A user-friendly and age-appropriate presentation of content was important for adolescent and young-adult populations (57, 76), with star representations of healthiness reportedly appealing to adults and traffic-lights appealing to adolescents (77). Images of portion sizing and rating the relative healthiness of meal options helped with nutrition knowledge development for young adults (76) and low-SES populations (79). Finally, young adults also reported that information presented must be branded by a credible source (93, 95).

Strategic approach and behavioral change theory

An explicit discussion of the strategic approach underpinning application development, in addition to a description of findings from existing literature, was provided in 2 of the 7 dietary self-monitoring trials (42, 44), 9 of the 13 nutrition-improvement trials (36, 48, 50–52, 56, 57, 60, 62), 15 of the 21 application-development projects (65, 68, 70, 72–74, 77, 79, 80, 85, 86, 88–90, 92) and was mentioned in 4 of the 6 qualitative application-development studies (93, 95, 96, 98). Multiple strategies were used in one self-monitoring trial (42), 5 nutrition-improvement trials (48, 50–52, 57), and 6 development projects (65, 74, 77, 86, 88, 89).

Study type.

The most common strategic approach involved the use of behavior change theory in application development, generally through the use of classic behavioral change models. The most commonly used models were used across both nutrition-improvement trials and application-description articles. These were Social Cognitive Theory, which informed the development of 7 applications (50–52, 57, 74, 77, 89); the Transtheoretical model (36, 52, 79) and Self-determination theory (57, 74, 89), both informing 3 applications; and Transportation theory (50, 77) and the Health Action Process Approach (56, 60), both informing the development of 2 applications. Additional models reported in nutrition-improvement trials and application-development projects were Control Systems Theory (44), Fogg’s behavioral model (51, 62), the Behavior Change Wheel (72, 86), the Health Belief Model (74, 77), the Theory of Planned Behavior (48), the Elaboration Likelihood Model (77), and the Precaution Adoption Process Model (77).

Additionally, some application-development projects and qualitative studies were not informed by a specific model but were guided by the general principles of broader theoretical content areas. These included principles from goal-setting theory (48), ecological momentary intervention (48), behavioral self-management (52), learning theory (51), motivational-enhancement approaches (52), behavioral economic theory (65), Atkin and Michie’s principles of individual behavior change (65), microlearning principles (70), cognitive behavioral therapy (88), and health communication and education theory (85). Mindful eating, social engagement, low burden, low-judgement approaches (80), self-regulation strategies of goal setting, self-monitoring, feedback (42, 93, 95), highlighting affective consequences, providing information, and rewards (98) were also suggested or used. Two qualitative studies recommended using both physical appearance and health-related motivations to facilitate application engagement for young adults (98) and social support for older adults (96).

Finally, rather than or in addition to using behavior change theory, 10 applications were developed through user involvement in application development, an iterative participatory design, or after conducting a needs assessment (42, 50, 52, 65, 68, 73, 86, 88, 90, 92). This approach was seen across all study types.

Target population.

Any strategic approach was more likely to be used in applications targeting a specific population (adolescents, 2 of 3; low-SES groups, 6 of 6; overweight populations, 10 of 15) than those for unspecified groups (young adults, 4 of 8; general adult, 8 of 15). This same pattern was also evident for the use of classic behavioral change models (adolescents, 2 of 2; low SES groups, 5 of 6; young adults, 0 of 4; general adults, 3 of 8). Applications targeting low-SES and overweight populations were also more likely to report involving users in their development process (3 of 6 and 4 of 10, respectively).

Aim of application or program.

The frequency with which applications with different aims were developed by using a strategic approach was similar (general-nutrition improvement, 15 of 24; weight loss, 7 of 11; grocery shopping support, 2 of 5; parenting practices support, 3 of 3; food access support, 3 of 4). Those aiming to support parenting practices were more likely to be informed by classic behavior change models than those for grocery shopping or food access support (3 of 3, 0 of 2, and 1 of 3, respectively), with the reverse pattern seen for the use of a needs-assessment approach (1 of 3, 2 of 2, and 2 of 3, respectively).

Consumer testing

Consumer testing of the application before intervention delivery was carried out in 7 of the 13 nutrition-improvement trials (36, 48, 50, 52, 53, 62, 63) and 17 of the 21 discreet application-development projects (65, 68, 70, 72–77, 79, 80, 85, 86, 88–90, 92). No description of consumer testing before trial delivery was provided in any of the reports of dietary self-monitoring trials or qualitative application-development articles. One study extensively elicited feedback about application experience and released software updates during the trial (52), and evaluation of the application to a lesser extent after trial completion was a common approach. Application-development project reports generally provided a more comprehensive description of consumer testing methods.

Testing design.

Field deployment (real-time use of the application by consumers in the intended setting), with or without participant observation, was the most common testing design used. This approach was taken in 3 nutrition-improvement trials (36, 48, 50) and 13 application-development projects (65, 68, 70, 72, 74, 76, 79, 80, 85, 88–90, 92). Other common approaches were testing of application components (36, 52, 75, 85) and focus groups (73, 85, 86). Less common approaches were online application demonstrations (92); weekly design, development, and testing sprints (68); cognitive walkthrough iterations (77); modified think-aloud protocols (90); and acceptability testing (53), and 2 papers reported “testing” with no elaboration on method (62, 63). Thirteen studies reported 1 round of testing (48, 50, 62, 63, 65, 70, 72–74, 79, 80, 88, 89), 7 reported 2 rounds (36, 52, 53, 75, 76, 90, 92), 1 reported 3 rounds (86), and 3 reported multiple rounds (68, 77, 85).

Testing duration.

Of the trials using a field deployment design and reporting the duration of user testing, 1 tested the application for less than 1 wk (88), 5 for 2–4 wk (50, 72, 74, 79, 80), 3 for 6–10 wk (48, 65, 76), and 2 for between 6 and 10 mo (70, 90).

Population sampled.

The most common population sampled for consumer testing was the target user group of the application, with 5 nutrition-improvement trials (36, 48, 52, 53, 62) and all 17 of the application-development projects (65, 68, 70, 72–77, 79, 80, 85, 86, 88–90, 92) using this approach. Initial application testing was also conducted by people within or close to the development team (50, 52) and project stakeholders (68). One nutrition-improvement trial research team tested their application with nutritionists and nurses rather than target users (63).

Sample size.

The size of the testing sample varied greatly. The most common size was between 11 and 20 testers (9 studies) (48, 72, 79, 85, 86, 88–90, 92), followed by ≤10 testers (4 studies) (73, 76, 85, 90), 21–50 testers (3 studies) (36, 74, 77), 51–100 testers (2 studies) (70, 80), 101–200 testers (2 studies) (75, 92), and least commonly >200 testers (1 study) (65).

Instruments used.

The instruments used for evaluating applications also varied greatly, with many studies using multiple instruments. The most common evaluative tools used were online surveys, interviews, and/or focus groups about accessibility, usability, satisfaction, integration into daily routine, and user experience (36, 48, 65, 70, 72–77, 79, 80, 85, 86, 88–90, 92) (18 studies). During field-deployment testing, 4 studies incorporated a pre- and posttest of nutrition improvement (48, 65, 79, 88), 3 logging of application use (links followed, button pressing, messages responded to) and (48, 70, 72), 2 analytics for data uptake and usage metrics (65, 85), and in 1 study testers kept a diary of when and where the application was used, progress, enjoyment, and thoughts related to nutrition content (79). When testing was not completed through a field-deployment design, paper or interactive-application mock-ups on a device were always used (52, 73, 77, 85, 86) (5 studies). Other instruments used were the System Usability Scale (48, 90), the Questionnaire for User Interaction Satisfaction (92), the Paper Prototyping Method (90), and a ranking questionnaire with tester rating of the importance of application features (77, 90).

Table 4 synthesizes the results of this review into a summary checklist outlining the key development, feature, content, and evaluation strategies for development teams to consider when designing an application-development project.

TABLE 4.

Checklist for application-development projects targeting nutrition improvement in community settings

	Checklist item
Development	Work in development teams including researchers, practitioners, target users, and application developers
	Utilize branding from a credible/familiar source within the application (for example, academic branding)
	Conduct predevelopment qualitative needs assessment of the target user population
	Use a behavior change theory as a broad base for application design
	Include target users through all key stages of application development/design
	Allocate resources to and conduct predeployment field testing with the target user group
Features	Incorporate interactive features, an attractive interface, and nonrepetitive images and colors
	Include a feature from each feature domain identified in this review
	Tailor features broadly to those most suited to the target user group
	Enable individual-user customization of application features
	Allow for changes in individual-user customization over time
Nutrition content	Tailor information and environmental supports to subpopulations
	Consider including >1 of the content areas identified in this review
	Consider including >1 of the feedback strategies identified in this review
	Include both positive and negative reinforcement of eating behaviors
	Use graphical presentations, such as colors, traffic light, and star representations
Evaluation and reporting	Allocate resources to, conduct, and report on application usability and consumer satisfaction evaluation
	Utilize existing tools, such as those identified in this review, in usability and satisfaction evaluation
	Clearly define and evaluate health outcomes to be influenced by the application
	Design studies to identify the contribution of the application to outcomes in multicomponent-program trials
	Describe the application and its development in detail in publications, including pictures of the interface and application flow
	Utilize the technical terminology of the area (that used in this review) when writing reports

Conclusions

The number of studies identified in this review, and specifically the large number completed in the last 5 y, suggests researcher and practitioner interest in the use of smartphone applications to support nutrition improvement in community settings. Although the results themselves provide a useful map of application characteristics for application-development teams and practitioners to consider, herein we shall highlight some broader implications of the findings from this systematic scoping review for these groups and the research area in general. A discussion of the importance of application tailoring and consumer testing is followed by suggestions for how to incorporate behavior change theory.

The strongest evaluation theme to come through in this review is the desire for and the success of tailoring in application design. This finding is consistent with findings from others (3, 9, 100). In their review, Hermawati and Lawson (9) suggest that when applications do not target a very specific user group this may result in a mismatch of potential user needs and application characteristics and therefore poor application engagement. They suggest application developers more specifically define their target groups and include these target users in all stages of application development to ensure full tailoring of application features and content. Certainly this approach would achieve the specified aim; however, it may also result in a proliferation of highly specific applications targeting very small population groups. Additionally, this may make promotion and dissemination of applications more challenging. Another approach may be to keep target user groups relatively broad but provide opportunity for within-application customization [of the type described by Helf and Hlavacs (3)], such that the specific needs of different users can be served. Characteristics described here that can be incorporated into application design to achieve this include a choice between simple and detailed food logging, customizable goals and challenges, time- and location-customizable reminders and prompts, multiple options for display of progress and food healthiness, and customizable avatars. This approach is likely to meet a diversity of consumer needs within user groups and meet individual users’ needs as they change over time and their interaction with the application matures (49, 60).

To be successful in tailoring applications in the manner described above, consumer testing with target users before application release is essential. However, our findings show that this was carried out in just over half of the trials and development projects. This finding was also identified in a systematic review of applications for obesity prevention that focuses on user-centered design (9). Bugs and application functionality problems (such as requiring an internet connection) that would undoubtedly have been illuminated in consumer testing reportedly needlessly influenced application engagement in a number of trials (39, 42, 51, 57, 60). Reasons provided for not conducting user testing included practical timing and resource and funding constraints of studies (57); however, most reports provided no reasoning. Studies, especially costly randomized controlled trials incorporating made-to-order applications, must consider testing with target users before application deployment an essential intervention design step and allocate resources accordingly (9).

Similarly, the use of a strategic approach for application design, whether it be an interactive user-centered approach or mapping features to behavioral change theory, is known to be a critical step toward ensuring application effectiveness, engagement (21, 27), and quality (22). Behavior change theories are an integral component of successful interventions to improve dietary intake (101). However, providing a detailed summary of the behavior change theories that were utilized in the cited studies was beyond the scope of this review. A number of articles (25, 27, 102) have assessed the prevalence and/or role of health behavior theory in diet applications. The opportunity to theoretically tailor application characteristics is a major advantage of making an application to order instead of using one that is commercially available. Yet, we found just over half of the studies reviewed here described any type of strategic approach. This may partially be a feature of study reporting, but for the purposes of this review our classification of strategic approach and theory use was broad and inclusive, as can be seen in the results. Commercial application content reviews focusing on incorporation of behavior change techniques offer many useful suggestions for mapping features to behavior change theories for future application developers (21, 22, 25, 27–29). Additionally, our map of the approaches that have been used may serve as a starting point, and the feature domains reported in the findings generally correspond to different behavior change support strategies common to many classic behavioral change models (21, 28), thus inclusion of a feature from each domain could be a useful initial approach.

In the same way, incorporating multiple domain features and nutrition content delivery strategies can assist in affecting multiple factors influencing food choices and therefore provide a more supportive environment for behavior change. For example, following the levels of food choice influence found in the ecological framework from Story et al. (2), providing general information and feedback on food logging may influence cognitions and therefore have an impact at the individual level (2). Adding social and team-related features may contribute to social support, changes in social norms, and role modeling at the level of the social environment (2). Similarly, although applications themselves cannot change the physical and macrolevel environments as such [although there is a case for reformulation and data collection that can be used for lobbying stimulated by some types of applications (81)], they can provide information about access to local foods or grocery shopping and menu selection assistance that can provide support for consumers to better manipulate these environments for their own health benefit. Our review shows that the majority of trials conducted have focused on generalized nutrition improvement and weight loss and as such typically do not incorporate supports for these higher levels of food choice influence. However, it is encouraging to see a variety of well-designed, theoretically based applications that have been developed to incorporate these features within the application-description studies that could be used in community nutrition improvement programs.

Poor-quality, vague, and inconsistent reporting hinders this evidence base in general. Likely because of strict journal word limits and competing priorities, reports of trials were often lacking comprehensive descriptions of applications and their development, testing, and evaluation. As such, our review is limited by the detail about applications provided in the reports, and this has previously been identified as an issue with this body of literature (5). In this study we did not specifically abstract information about the resources that developers used to inform the nutrition information disseminated in their applications; hence, whether the information they provide is accurate requires further evaluation. Some applications may have had additional characteristics that are not reported here, but because of practical constraints we could not individually download and try all the applications available commercially nor request access to made-to-order applications from authors. Additionally, because of inconsistencies in terminology used between studies there was a level of unavoidable subjective interpretation within data extraction. However, all outcomes of interest to the review were extracted from available reports, so it is as thorough as the quality of reporting in the literature allows. Similarly, critical definitions were discussed at regular author meetings, and all data extraction was completed by one author; therefore, consistency in data extraction and analysis was high. In moving the reporting of this area of research forward, we suggest authors take advantage of the option to include online-only supplementary material with their manuscripts to provide detailed information about made-to-order applications. Screenshots of the design of application interfaces and flow would be especially useful.

It is possible that heterogeneity within target group demographics can affect the success of the applications to enhance nutritional impact and limit the generalizability of results. Hence, careful consideration of how target population demographics may modify or confound the effectiveness of nutrition applications is an important component of the design process. A further limitation of the research area is the scarcity of information regarding length of use, impact on health outcomes, and sustained change achieved specifically through the use of applications. As this evidence base matures, this will be an important area for future research to systematically review and consequently provide concrete data on efficacy for future application projects. We originally intended to examine strategies for promoting and disseminating applications, but only one study reported on this (95), and it reported using media coverage, mail-outs, clinic posters, patient leaflets in service settings, service provider support, and promotion through target user networks and newsletters. Again, this issue with the evidence base has been reported previously as an issue with using traditional research designs, such as randomized controlled trials, to test the efficacy of applications (5). Finally, children and adolescents are at increased risk of developing unhealthy weight-control behaviors, and as such it is important that caution is taken to ensure that no harm is caused when promoting changes in eating behavior for this population group (103).

Despite these limitations, to our knowledge this is the first research to comprehensively and inclusively map the characteristics, development, and consumer testing of applications for nutrition improvement in community settings.

Collaboration between academics and application developers promotes an appropriate balance of evidence-based content and functionality (3–5). This review can be used as a starting point and foundation for these collaborations in designing applications for nutrition-improvement projects.