Behavior Change Effectiveness Using Nutrition Apps in People With Chronic Diseases: Scoping Review

Introduction

Background

Cardiovascular disease (CVD), cancer, diabetes mellitus (DM), and obesity are common chronic diseases [1], and their prevalence is reaching a substantial epidemic level internationally [2]. Chronic diseases are defined by the Centers for Disease Control and Prevention broadly as “conditions that last one year or more and require ongoing medical attention or limit activities of daily living or both” [3]. Chronic diseases affect hospitalization, mortality rates, and people’s overall health and quality of life (QOL) [1]. For example, CVD remains the most prevalent cause of morbidity and mortality in high-income countries despite significant advances in treatment over the last 5 decades. Recent epidemiological data show that CVD mortality is no longer declining and is indeed rising in some communities [4], and hospitalization rates are universally increasing [5]. Furthermore, chronic conditions such as cancer, CVD, DM, and chronic respiratory diseases caused approximately 33.2 million deaths worldwide in 2019 [6]. The prevalence of obesity has increased in all World Health Organization regions since 2000, which affects other chronic conditions as it is a risk factor for the development of CVD, DM, and several cancers [6].

As the impact of chronic diseases continues to increase, finding strategies to improve care, access to care, and patient empowerment becomes increasingly essential. Therefore, mobile health (mHealth) technology is rapidly gaining popularity among health care providers and consumers [7]. mHealth technology is defined as mobile devices (ie, mobile phones or monitoring devices) intended to be worn, carried, or accessed by patients or health care providers to monitor health status or improve health outcomes [8]. Among mobile devices, the most used are smartphones, with more than three-quarters of Americans having one, and at least one-third of smartphone owners use a health app [7,9]. Health care providers use mHealth to access clinical information, collaborate with care teams, communicate over long distances with patients, and facilitate real-time monitoring and interventions. These apps provide an opportunity to increase health care access for vulnerable populations [10]. Patients use mHealth to track their health data, access their clinical records, and communicate with their providers [11].

Furthermore, a meta-analysis [12] reported promising results for mHealth interventions in the improvement of patient outcomes such as body measurements (ie, weight and waist circumference), metabolic and physiological measurements (ie, blood pressure and glucose levels), adherence to and safe use of medication, physical activity performance, meal management, and awareness of health conditions and treatment options. Mobile apps that provide tools intended to facilitate nutrition care via smartphone technologies provide patients with more autonomy, thus empowering them and offsetting patient disengagement [13]. Moreover, mobile app–based interventions effectively improve diet and diet-related health outcomes, and effect sizes are comparable with those of traditional nondigital interventions [14]. For example, mHealth interventions can help improve lifestyle behaviors related to CVD [15,16]. A recent meta-analysis [17] found that using mHealth interventions for CVD was associated with improved blood pressure and hospitalization rates. In addition, mHealth apps have emerged as supportive tools in managing cancer as they reduce the financial burden, provide access to information, and facilitate communication [18]. Different studies and meta-analyses of patients with cancer have shown the benefits of mHealth, which include reducing fatigue or pain [19,20], providing distance physical activity programs [21,22], the use of social networks to improve QOL [23], and monitoring of symptoms [24]. mHealth offers improved and cost-effective care to people with type 2 DM (T2DM) through improved diabetes self-management [25,26]. These apps seem to increase the perception of self-care by contributing to better knowledge among people with T2DM [10]. Individuals with DM also become more confident in dealing with their illness, primarily because of decreased fear resulting from a lack of information [27,28]. For obesity, a systematic review [29] has found that technology-based interventions can provide a moderately effective way of facilitating lifestyle changes and weight loss. New technologies could help reduce economic costs, improve accessibility and adherence to treatment, and increase patient motivation and weight control [30].

The use of mobile apps for mHealth and of general health apps is increasing rapidly. However, these apps focus on improving general health care concerns, with limited apps focusing on specific chronic diseases and their nutritional intervention. The available evidence on the effectiveness of mHealth apps toward behavior change to improve chronic disease outcomes is also limited. Systematic reviews to date regarding nutrition apps have focused on healthy participants or examined the effects of dietary apps on diet improvement but not on chronic diseases [31]. Most reviews were inconclusive, with the authors recommending further research in this area to demonstrate possible benefits [32-34]. From a clinical point of view, it is essential to know if an app designed for chronic diseases produces behavior changes to improve an individual’s chronic disease outcomes. From an app creator’s point of view, it is crucial to see what needs to be done when developing an app directed toward people with chronic diseases to enhance the app’s effectiveness on behavior change and produce positive outcomes. The goal of this review was to define behavioral theories associated with mHealth use, evaluate behavior change techniques (BCTs), and describe the behavior change effectiveness using mHealth nutrition interventions in people with chronic diseases (ie, CVD, DM, cancer, and obesity).

Models and Theories of Health Behavior and Clinical Interventions

Health care and self-management of chronic conditions require effort and commitment on the part of the patient to follow treatment plans. These treatments involve many behaviors, such as dietary intake, physical activity, and prescription drug use. [35]. Theories and models are used in treatment planning to understand and explain the health behavior of individuals and can help guide clinicians to develop interventions that increase the effectiveness of health behavior change. The role of behavioral theories and models in informing digital health interventions is important to support sustainable health behavior change. A brief explanation of some of the theories and models of behavior used in these studies may help with understanding their constructs [36]: (1) the transtheoretical model shows how individuals move through 6 stages of behavior change (ie, precontemplation, contemplation, preparation, action, maintenance, and termination), which can be used to support behavior change and self-efficacy for the cessation of unhealthy behaviors and encourage a healthy lifestyle; (2) Social Cognitive Theory emphasizes self-efficacy and focuses on behavior change and outcome expectations by mastering steps to behavior change, observing others who are successful, improving mood, and ultimately increasing self-efficacy for health behavior change; (3) the Health Belief Model is based on the concept of expectancy-value and constitutes an individual’s belief that their health condition is serious, their actions will reduce their risk or illness, there is a benefit to taking action for their health condition, and they have the ability to take action for health behavior change (self-efficacy); and (4) the theory of planned behavior emphasizes that motivation that is directly influenced by ability (ie, an individual’s self-efficacy or perceived control over outside factors) is key to making a health behavior change [36]. There are other theories that can be used as well. For health behavior change, there is a general acceptance of the Health Belief Model along with a focus on self-efficacy found in many behavioral models such as the Social Cognitive Theory [35]. In addition, a commonly used theory to guide lifestyle interventions is the transtheoretical model [36]. There are several theories identified in this review that are important to chronic disease research on health behavior.

Methods

Information Sources and Search Strategies

In April 2022, a systematic literature search was conducted in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) strategy for each of the following diseases: obesity, DM, CVD, and cancer [37]. The PRISMA method allows for the transparent selection and analysis of literature for inclusion. The databases used in the search were MEDLINE, EBSCO, PubMed, ScienceDirect, and Scopus. These databases were chosen based on their extensive health- and technology-related literature [38]. The search strategy was similar for all diseases. It included the following keywords searched among titles, keywords, and abstracts: (1) mobile applications or apps or mobile apps or mHealth or eHealth AND nutritioneducation AND obesity, (2) mobile applications or apps or mobile apps or mHealth or eHealth AND nutritioneducation AND type 2 diabetes, (3) mobile applications or apps or mobile apps or mHealth or eHealth AND nutritioneducation AND cardiovascular disease, and (4) mobile applications or apps or mobile apps or mHealth or eHealth AND nutritioneducation AND cancer.

Eligibility Criteria

To summarize the evidence available on the topic, we included primary research studies involving an app and a nutrition intervention. Studies were excluded from the review if they did not involve an app or nutrition intervention, were written in a language other than English, were duplicates from other database searches, or were literature reviews.

Procedures

Following the PRISMA 2020 method, the systematic review process included four steps: (1) identification of records through the database search, (2) screening of duplicate and excluded records, (3) eligibility assessment of full-text records, and (4) final analysis of included records. This process can be seen in Figure 1. The results of the database searches were exported to Microsoft Excel (Microsoft Corp) for further analysis. Duplicates were removed. The remaining studies were moved to the eligibility phase, in which the authors assessed the eligibility of the articles based on the inclusion and exclusion criteria of the full text. The reason for exclusion was listed for each excluded article. The citation of the article, name of the mobile app used if available, type of intervention used, year of publication, population and sample details, study design details, purpose of the research, behavioral effectiveness, behavioral techniques, theory used, outcome measures, and results were recorded for all included studies. The authors independently reviewed the articles found in each search, and a second screening was performed by a different author. The process for screening was as follows: SG initially screened articles found in the obesity search, MA initially screened articles found in the DM search, ESG initially screened articles found in the cancer search, and AC initially screened articles found in the CVD search. SG performed a second screening of articles found in the DM search, MA performed a second screening of articles found in the obesity search, AC performed a second screening of articles found in the cancer search, and ESG performed a second screening of articles found in the CVD search.

‎View this figure

Results

CVD Results

Demographic, Participant, and Study Design Details

There were 6 studies and mHealth app nutrition interventions comprising 451 patients included for CVD. Of the 6 studies, 4 (67%) randomized controlled trials, 1 (17%) qualitative descriptive design study, and 1 (17%) intervention evaluation were analyzed. The populations analyzed had diverse ethnicities, education levels, ages, and diagnoses. Furthermore, there were different diets or treatments among the studies, with variations in targeted behavioral domains and outcomes measured in the mHealth app nutrition interventions. App features for these studies included food tracking [8,39-41], physical activity tracking [8,41,42], nutrition and exercise knowledge and recommendations [8,40-43], scheduled reminders [39,40,42,43], clinician portals [41-43], connectivity to digital health devices [42], demonstration of the desired behavior [40,43], and challenges [41]. The aim of all studies (6/6, 100%) was to evaluate the effectiveness of mHealth in the CVD population. The included studies on nutrition apps for people with CVD can be found in Table 1.

Targeted Behavior and Outcome Measures

The most commonly measured outcome in the CVD studies was usability or engagement in 83% (5/6) of the studies, followed by diet intake or quality and metabolic and physiological measurements (ie, blood pressure and urinary sodium) in 67% (4/6) of the studies. In addition, treatment adherence (ie, diet, cardiac rehabilitation, and medication) was measured in 50% (3/6) of the studies. Other measured outcomes included physical activity, weight loss, dietary knowledge, sustainability, tobacco or alcohol consumption, hospital readmissions, cost-effectiveness, food purchases, and self-management in 17% (1/6) of the studies.

BCTs Used

The most commonly used BCT was motivation or encouragement, which was used in all reviewed studies (6/6, 100%). Knowledge or education and self-monitoring were used in 83% (5/6) of the studies. Both prompts or cues and feedback were used in 50% (3/6) and 33% (2/6) of the studies, respectively. The least common BCTs included graded tasks or challenges, demonstration of behavior, self-efficacy, and goal setting in 17% (1/6) of the studies reviewed.

Behavioral Theory or Model

In total, 17% (1/6) of the analyzed studies were based on behavioral theories or models. The others (5/6, 83%) did not mention a theory. The transtheoretical model, Social Cognitive Theory, the theory of planned behavior, the Health Belief Model, the precaution adoption model, and goal-setting theories were used as the basis for 17% (1/6) of the interventions in this review section. Evidence for informing digital technology interventions reveals that the Health Belief Model has been widely used for goal setting and lifestyle changes for reducing cardiovascular risk as it focuses on confidence in one’s ability to take action [36].

Behavior Change Effectiveness

CVD-specific mHealth apps significantly improved the completion of cardiac rehabilitation, were significantly cost-effective, and resulted in substantial weight loss and less salt purchases in 17% (1/6) of the studies. Furthermore, significant engagement was observed in 67% (4/6) of the studies. In addition, some studies (1/6, 17%) showed improvement in blood pressure and self-management and rated the app quality as good. However, linking BCTs and theoretical frameworks to behavior change and CVD health measures is challenging. Therefore, this area of study should be investigated further.

Cancer Results

Demographic, Participant, and Study Design Details

In total, 9 cancer-related studies comprising 645 patients were included. Of the 9 studies, 3 (33%) focused on survivors of breast cancer, 2 (22%) focused on breast cancer, 2 (22%) focused on esophageal cancer, 1 (11%) focused on patients with either gastric or colon cancer, and 1 (11%) focused on pancreatic cancer. Various study designs were used: 22% (2/9) prospective quasi-experimental studies, 22% (2/9) prospective pilot studies, 11% (1/9) interventional observation studies, 11% (1/9) nonrandomized 2-group controlled design studies, 11% (1/9) randomized controlled trials, 11% (1/9) randomized pretest-posttest design studies, and 11% (1/9) clinical trials were analyzed. These studies can be found in Table 2. Each app had different features, including digital diaries [7,44-49], coach feedback [7,44-47,49-51], personalized physical exercise programs and nutrition plans [49,50], general nutrition and physical exercise guidelines [45,51], psychological support courses [45,51], a secure message portal to their health care teams [7,48], health knowledge education [46,48,51], and daily tasks [48,51].

Target Behavior and Outcome Measures

The most targeted behaviors were physical activity, observed in 56% (5/9) of the included studies, and dietary behavior, observed in 78% (7/9) of the studies. The questionnaires measured both behaviors. Furthermore, QOL was measured in 67% (6/9) of the studies, followed by app use and satisfaction in 56% (5/9) of the studies. Less frequently measured outcomes included app adherence in 44% (4/9) of the studies, depression in 33% (3/9) of the studies, and fatigue in 22% (2/9) of the studies.

BCTs Used

Self-monitoring and feedback were the most frequently used in mHealth apps for cancer (8/9, 89% of the studies), followed by goal setting and motivation in 44% (4/9) of the studies each. However, only 22% (2/9) of the studies used knowledge.

Behavioral Theory or Model

Most of the analyzed studies (6/9, 67%) seemed not to be based on common theories. Only 33% (3/9) of the studies reported at least one theory. The only reported theories were Social Cognitive Theory, Goal-Setting Theory, and the comprehensive intervention model. This result is surprising given that most apps used feedback, goal setting, and motivation to conduct behavior change. Therefore, further research could use behavioral theory or models to enhance their apps for this chronic disease. However, in physical interventions (no digital technology involved), some evidence showed that the transtheoretical model combined with other models was successful in breast cancer screening behavior as it is based on the stages of behavior change and interventions can be tailored to each individual, which increases their empowerment to make change [52].

Behavior Change Effectiveness

Cancer-specific mHealth apps helped significantly improve QOL in 44% (4/9) of the studies, followed by changes in anthropometrics in 33% (3/9) of the studies. A total of 11% (1/9) of the studies reported an increase in QOL and a decrease in distress level. In contrast, physical activity and nutritional status were only significantly improved in 22% (2/9) of the studies. Similar to the CVD section, linking BCTs and theoretical frameworks to behavior change and cancer health measures is challenging. Therefore, this area of study should be investigated further.

DM Results

Demographic, Participant, and Study Design Details

In total, 13 studies comprising 1559 patients were included. Of the 13 studies, we analyzed 10 (77%) randomized controlled trials, 1 (8%) single-arm feasibility study, 1 (8%) case report, and 1 (8%) uncontrolled study. The studies included in this review involving nutrition apps for diabetes can be found in Table 3. Overall, there was a diverse population of patients diagnosed with prediabetes (glucose: 5.55-6.94 mmol/L or 100-125 mg/dL; HbA_1c: 39-46 mmol/mol or 5.7%-6.4%), type 1 DM, T2DM, and gestational DM (2-hour oral glucose tolerance test level of ≥9 mmol/L). The features of each app included food tracking [53-62], education, knowledge, or recommendations [54-56,59-65], physical activity tracking [54-56,59-62], weight monitoring [56,59-62], glucose monitoring [55,57,60,61,64], blood pressure monitoring [60], community support [54,59,60], feedback [54-56,59,61,62,64], health coaches [55], clinician portals [60-62], connectivity to digital health devices [56,57], scheduled reminders [56,58,59,62], gamification [59,63], and goal setting [59]. In addition, several targeted behavioral domains and outcomes were measured as a result of the mHealth app nutrition intervention. This factor revealed some interesting insights for future investigations into digital health interventions among people with DM.

Targeted Behavior and Outcome Measures

The most frequently targeted behaviors were glycemic control measured using HbA_1c and DM self-efficacy or self-management in 46% (6/13) of the reviewed studies. In addition, dietary behavior, levels of engagement, user acceptability, or motivation, and weight or BMI were measured in 46% (6/13) of the studies, followed by waist circumference in 23% (3/13) of the studies. Less frequently measured outcomes included physical activity, stages of change, carbohydrate counting accuracy, QOL, health care use behavior, and lipids in 23% (3/13) of the studies.

BCTs Used

Numerous BCTs can be used to induce behavior change. All the DM studies reviewed (13/13, 100%) used knowledge and education, followed by self-monitoring in 77% (10/13) of the studies. Both social support or encouragement and autonomous personalized feedback were used in 54% (7/13) of the studies. Prompts or cues were used in 31% (4/13) of the studies, followed by graded tasks and gamification in 15% (2/13) of the studies. Real-time feedback and goal setting were used in 15% (2/13) of the studies reviewed. Unfortunately, no standard definition of techniques is included in the BCTs, making it challenging to identify the techniques used in the interventions [66].

Behavioral Theory or Model

Some of the analyzed studies (7/13, 54%) seemed not to be based on behavioral theories or models. In total, 46% (6/13) of the studies reported at least one theory. The transtheoretical model was most frequently used in 23% (3/13) of the studies, followed by Social Cognitive Theory, self-determination theory, models centering on cues and triggers, theory of planned behavior, behavioral economics, positive psychology, and motivational interviewing as other reported theories. Some research has shown that Social Cognitive Theory has been used in feasibility studies among populations with diabetes as it can increase confidence and promote greater sustained effort to change, making it a guide for digital technology interventions. This theory also includes skill training, which can benefit diabetes management and education programs [67,68].

Behavior Change Effectiveness

DM-specific mHealth apps improved glycemic control by significantly reducing HbA_1c values in 46% (6/13) of the studies. In addition, 15% (2/13) of the interventions [53,62] resulted in a decrease in the medications used for glycemic control. Some studies (6/13, 46%) showed significant engagement; however, 17% (1/6) of these studies showed a decline in engagement over time, and 17% (1/6) did not have follow-up data for engagement. Sustainability needs to be considered for the effectiveness of these types of interventions. There were also significant changes in anthropometrics in 31% (4/13) of the reviewed studies. DM self-efficacy and self-management, decrease in energy intake, and increase in physical activity were observed in 8% (1/13) of the studies. Not all studies analyzed the same outcomes for each intervention, making it difficult to link BCTs and theoretical frameworks to behavior change and health measures.

Obesity Results

Demographic, Participant, and Study Design Details

A total of 18 studies comprising 253,775 patients were included. Among these 18 studies, we analyzed 5 (28%) randomized controlled trials, 5 (28%) experimental studies, 3 (17%) feasibility studies, 2 (11%) observations, and 1 (6%) prospective cohort study. The studies involving nutrition apps among people with obesity can be found in Table 4. The studies focused on obesity prevention and treatment in many diverse populations that ranged in age, socioeconomic background, and physical status (ie, pregnancy and post partum). App features included physical activity tracking [69-78], food tracking [70,72-83], knowledge, education, or recommendations [70,72-76,78,79,82-86], push notifications or scheduled reminders [74-76,78,79,81-84,86], weight monitoring [75,76,78,81,85,86], behavior demonstration [69,75,83], motivational challenges [69,75-78,83], goal setting [69,73-75,78,81,84,85], feedback [69,70,73,75,76,81-83,86], community support [70,75,78,85], connectivity to digital health devices [70,72,81,86], clinician portals [72,73,82], and access to a health coach [76-78,81]. Furthermore, numerous behaviors were targeted, and the outcomes were analyzed to determine the effectiveness of a mobile nutrition intervention in promoting healthy weight. This information will help develop approaches and techniques for digital health behavior change interventions to prevent and treat obesity.

Targeted Behavior and Outcome Measures

Anthropometric measurements (ie, weight, BMI, and waist-to-hip ratio) were the most targeted health outcomes, with behavior change in 78% (14/18) of the reviewed studies, followed by diet quality and physical activity in 61% (11/18) of the studies. In addition, engagement, user acceptability, or motivation levels and body composition were measured in 39% (7/18) and 28% (5/18) of the studies, respectively. QOL, behaviors (ie, goal setting and self-efficacy), and CVD measures (ie, blood pressure, heart rate, glucose, and lipids) were included as outcomes in 22% (4/18) of the studies. Less frequent measures included stress parameters (ie, chronic stress and cortisol levels), diet knowledge, reduced screen time, and behavior and health maintenance, observed in 6% (1/18) of the studies.

BCTs Used

BCTs were incorporated into each intervention. Self-monitoring was the most commonly used BCT (14/18, 78% of the reviewed studies), followed by knowledge or education (12/18, 67% of the studies). Goal setting was used in 56% (10/18) of the studies, feedback and encouragement were used in 56% (10/18) of the studies, and prompts or cues and intention formation were used in 50% (9/18) of the studies. Feedback through private messages was used in 39% (7/18) of the studies, followed by community or social support and demonstration of behavior in 33% (6/18) of the studies each. Graded tasks and gamification or incentives were used in 17% (3/18) of the studies. Finally, information about health benefits and consequences was used in 6% (1/18) of the studies in this section. Clinician portals were used in 17% (3/18) of the interventions to improve patient care. Although it is not a BCT, it is worth mentioning as a vital intervention component.

Behavioral Theory or Model

Some interventions (10/18, 56%) did not mention the basis of behavioral theories or models. A total of 6% (1/18) of the studies mentioned that they were rooted in theoretical models but did not specify which ones. Social Cognitive Theory was the most frequently used behavior model in 33% (6/18) of the studies. Self-determination theory was used in 11% (2/18) of the studies, followed by the socioecological model, control theory, and social support theory in 6% (1/18) of the studies. Furthermore, Social Cognitive Theory has been used in feasibility studies on mHealth technology lifestyle interventions for obesity, for example, by using the principle of verbal persuasion through methods such as personalized encouragement that help individuals realize that they have the capability to make the necessary healthy lifestyle changes to lose weight [87].

Behavior Change Effectiveness

Obesity-specific mHealth apps improved weight through significant reductions in 39% (7/18) of the studies. Some studies (4/18, 22%) showed a substantial decrease in body fat percentage and waist circumference. There were also significant changes in cardiovascular measurements (ie, blood pressure and lipids) and QOL in 11% (2/18) of the reviewed studies. Changes in self-efficacy, decreases in energy intake and screen time, and increases in knowledge of nutrients and physical capabilities were observed in 6% (1/18) of the studies. In total, 11% (2/18) of the studies analyzed and reported significance in the maintenance of weight loss, whereas 11% (2/18) of the studies reported a decline in changes over time. The sustainability of these interventions is an area for future research to determine the effectiveness of these types of interventions.

Discussion

Principal Findings

Overall, mHealth apps used for various chronic disease populations did improve health. These studies showed significant improvements in QOL, cardiac rehabilitation completion, glycemic control (ie, HbA_1c), weight reduction, and reduction in physiological measures (ie, blood pressure and lipids). In addition, some studies (3/46, 7%) showed improved self-efficacy and self-management of chronic diseases. Although many studies (35/46, 76%) did not measure long-term effectiveness, some (3/46, 7%) showed significance in maintaining weight loss, whereas others (3/46, 7%) showed a decline in changes or engagement in app use over time. A total of 24% (11/46) of the studies measured maintenance of health behavior change. Of these 11 studies, 7 (64%) sustained behavior change for approximately 6 to 12 months, and 4 (36%) showed a decline in behavior change or discontinued app use.

The main difference between the sustainability of health behavior change was the inclusion of a clinician portal or access to a health coach. Griauzde et al [56] measured postintervention qualitative data that provided reasons for app satisfaction, dissatisfaction, and ways to improve. Reasons for satisfaction included “encouraged self-reflection,” reasons for app dissatisfaction included “did not consider personal circumstances,” and strategies to improve the intervention included “increased interpersonal contact.” Of the 46 studies, 9 (20%) included a clinician portal as a feature to enhance the app intervention, allowing for additional communication between clinicians and patients. Of the 9 studies, 7 (78%) had an effective health behavior change. Silk et al [88] reported acceptability and easy-to-use features regarding the integration of the clinician portal into their mHealth intervention. This component should be considered for health behavior change interventions and investigated further.

Of the 46 nutrition apps in this review, 37 (80%) included some type of self-monitoring feature. Of those 37 apps, 35 (95%) had at least one significant improvement or good usability rating to promote engagement with mobile apps for chronic disease self-management. This is in alignment with other research that shows that the most common self-management application for mHealth is a tracking feature [89]. In addition, when surveying clinicians working in diabetes and weight management patient care settings, the most adoptable apps included self-monitoring features [90]. However, the apps that did not include self-monitoring or tracking features showed significant improvements as well. In addition, there was variability in other features combined with tracking features (ie, knowledge, goal setting, coach feedback, and clinician portals), making it difficult to attribute behavior change to self-monitoring features alone. More research is needed to correlate specific features with health behavior change.

Except for cancer populations, a key finding in this review was that 41% (19/46) of the nutrition app interventions targeted weight management, and 58% (11/19) of those studies were effective in health behavior change. This finding is similar to that of Fakih El Khoury et al [91], who reported that dietary mobile apps positively affected measured nutritional outcomes in chronic diseases, especially weight loss. The key finding for cancer populations was that mHealth nutrition apps can significantly improve QOL. A similar result was found in a review analyzing the use of mHealth involving nutrition in a chronic kidney disease population [92]. A total of 28% (13/46) of the studies examined changes in glycemic control, HbA_1c, and other biometric measurements (ie, blood pressure and lipids). In total, 69% (9/13) of those studies showed a significant reduction in their measurements when using a nutrition app in their intervention. In addition, 15% (2/13) of those studies showed a decrease in medication treatment for glycemic control [53,62]. Eberle et al [93] concluded that mHealth apps with a nutrition intervention effectively enhanced diabetes management, which is comparable with our review results.

There was diversity in the chronic disease state; study design; number of participants; and variety of in-app features, BCTs, and behavior models used in the studies. However, linking these factors to a particular behavior change and health outcome posed to be difficult, which is not different from Taj et al [94] in their review of digital health behavior change technology. This review provides insight into the theories and theoretical frameworks or models used in nutrition-focused mHealth interventions to increase understanding and translate them into practice, which is essential for developing behavior change for the sustainability of health improvement. For example, this scoping review found that most apps used BCTs, such as self-monitoring in 80% (37/46) of the interventions, knowledge in 72% (33/46) of the interventions, feedback in 54% (25/46) of the interventions, and goal setting in 20% (9/46) of the interventions, for effective health behavior change, as another review [31] demonstrated.

In addition, our review found that less than half (19/46, 41%) of the studies based their nutrition apps on a behavioral theory or its constructs. Some studies (14/46, 30%) that based their interventions on theories had improved engagement, satisfaction, or app use among participants and suggested that incorporating a behavioral theory in mHealth interventions is an effective strategy, which is a comparable finding with that of another review [91]. However, other factors play a role in usability, acceptability, and overall user satisfaction. The app features, BCTs, and theory-based interventions will all affect the effectiveness of mHealth nutrition apps in health behavior change toward chronic diseases.

Another significant development in the sector of health technology is that smartphones can also be embedded with sensors or coupled with wearable sensors for health monitoring, which could enhance a nutrition app’s effectiveness in health behavior change. Examples of these devices that could be coupled with nutrition apps include motion sensors such as accelerometers, gyroscopes, and magnetometers that measure motion and physical activity [95]; wearable devices such as glasses (analyzing the intake pattern with smart glasses) and rings (ring-type tactile sensors to detect food mass) for food intake measurement [96]; portable and handheld single-lead electrocardiogram devices in addition to fitness trackers to measure heart rate and heart rate variability [95]; wristwatches to measure glucose levels extracted from skin interstitial fluid through reverse ionophoresis or through saliva-, sweat-, or tear-based wearable biosensors [97]; a wrist-wearable watch with the function of a pulsimeter without a cuff to measure blood pressure [97]; and ingestible capsules that can be used for medical health monitoring [98]. These enhance personal health care and performance monitoring with the potential to complement nutrition apps and have a broad impact on our society.

Limitations

Although the results of this study showed health improvements achieved when using mHealth nutrition apps for behavior change in chronic diseases, some limitations need to be addressed. An important limitation is the lack of research on mobile apps’ long-term effects (>1 year) on disease state populations. Therefore, there is no conclusive result on their long-term behavior change to determine whether the behavior continues to improve or reverses compared with the baseline. Consequently, there need to be more long-term studies conducted. Second, many studies (24/46, 52%) did not include a control group for comparison, and their sample size was small, limiting the interpretation of the results of the studies. Third, many apps are at risk of becoming rapidly obsolete owing to the fast pace at which technologies progress and, therefore, new technological innovations must be considered [8]. For example, the latest mobile technologies can connect and interact with each other, update and track personal health data in real time, and send alerts to users [8]. Similarly, most health apps have encountered serious usability problems or have not undergone usability assessment [99]. Usability affects the efficiency and efficacy of the app (ie, time to complete tasks and errors) [8]. Usability must be considered to increase the chance of the app being successfully adopted by patients [8]. Barriers to using mHealth apps include the patient’s lack of integration of technology into everyday life [12] and difficulties using mobile apps [11]. In the older adult population, health problems such as cognitive changes related to aging, disability, and lack of confidence are reasons for not using digital technology [18,19]. Further research is needed to evaluate patients’ experiences with apps and the benefits gained as a result. There could be a slight bias from the user perspective as there were 2% (1/46) of apps that were Android-only or web-based applications. The others (45/46, 98%) were web-based applications or all-smartphone apps. There were a few apps (4/46, 9%) that were provided through the smartphone that participants were given as part of the study, and the smartphone was not specified. A few studies (4/46, 9%) did not report the data privacy rules or the effects of users and funders on the app interventions. In addition, there are important confidentiality and funding issues that must be considered when designing interventions [100]. Finally, proficient health care providers should be involved in the app development stage to address safety during self-management and health education [101]. There is a need for comprehensive, efficient, and flexible mobile apps for the self-management of disease states with more features to increase the number of long-term users and induce better self-management and patient empowerment [101,102]. We did not conduct a systematic review or meta-analysis and, thus, did not weigh the quality of evidence or study design against the reported results. Some studies (6/46, 13%) included few participants, and the diversity of study objectives, designs, and outcomes made it difficult to compare them. We reviewed the current evidence to expand the knowledge base regarding the impact of nutrition apps on chronic disease management and assess the effectiveness of health behavior change.

Conclusions

In this scoping review, the use of mHealth nutrition apps and their effects on health behavior change were analyzed for 4 diseases (ie, cancer, CVD, DM, and obesity). The results suggest that mHealth apps involving nutrition can significantly improve health outcomes for people with chronic diseases. The study design, demographics, targeted behavior, health outcomes, BCTs, behavioral theories, and behavior change effectiveness were profoundly diverse among these studies, indicating that a one-size-fits-all approach for designing and implementing nutrition apps as part of chronic disease treatment is not possible. Tailoring nutrition apps to specific populations is recommended for effective behavior change and improvement of health outcomes. In addition, some studies (7/46, 15%) showed sustained health behavior change, and some (4/46, 9%) showed a decline in the use of the nutrition apps. These results indicate a need for further investigation on the sustainability of the health behavior change effectiveness of disease-specific nutrition apps.