Purpose: Smartphone-based behaviour change interventions have been developed predominantly using psychological theories, which have limitations when considering more dynamic digital behaviour change intervention such as just-in-time adaptive interventions (JITAIs). It is unclear how many dynamic model-based interventions have been undertaken in the domain of physical activity (PA) and sedentary behaviour (SB). Therefore, this review aims to quantify the number of studies that have used dynamic models to develop smartphone-based interventions to promote PA and reduce SB, describe their features, and where possible evaluate their effectiveness.

Methods: Databases including PubMed, PsychINFO, IEEE Xplore, Cochrane and SCOPUS were searched from inception to 15 May, 2019 using terms related to mobile health, dynamic models, and activity. Two researchers conducted the screening and data extraction, independently. Outcomes included general characteristics, dynamic model, theory/construct integrated, and outcomes measured (SB and PA behaviours). Data were synthesized narratively. There were limited scope for a meta-analysis because of the variability in the study results.

Results: A total of 1086 reports were screened. Eventually, 12 reports describing 8 studies were eligible and included. Only three studies targeted SB, two of which attempted to interrupt SB by light-intensity walking, and one focussed on both standing and movement. Social Cognitive Theory (SCT) was the major psychological theory on which the studies were based. JITAIs were described in three studies. Dynamic Decision Network, Control Systems engineering, Behavioural Analytic Algorithm and Exploit-Explore Strategy each were employed by one study. Android was the main smartphone operating system in all but one study that used iOS. Four studies used built-in smartphone sensors (i.e. accelerometers) to measure activity, two of which used phone’s GPS as well. Three studies used wearable activity trackers. Six studies reported on effectiveness of the intervention.

Discussion: To our knowledge, this is the first systematic review that reports on smartphone-based studies to interrupt SB and promote PA with a focus on integrated dynamic models. Current findings highlight the scarcity of dynamic model-based smartphone studies to interrupt SB or to promote PA. Future research is required to assess the effectiveness of dynamic models to promote PA and reduce SB.

Purpose: Evidence indicates that ecological momentary assessments (EMAs) are a valid, reliable, and feasible method of data collection. Mobile physical activity (PA) EMAs have demonstrated better correlation with accelerometer estimates than traditional self-report methods. However, existing evidence also shows discrepancies in mobile EMA methodology (e.g., triggering processes, time to follow-up), as well as limitations in terms of usage of identical mobile devices and inability to capture context. The purpose of this study was to develop a novel, replicable methodology of mobile EMAs to capture prospective PA within free-living social and physical contexts by leveraging citizen-owned smartphones running on both Android and iOS systems.

Methods: Data were obtained from the adult cohort of the SMART Platform, an innovative citizen science and mobile health initiative for active living surveillance. 538 citizen scientists (≥18 years) residing in Regina and Saskatoon, Canada, provided PA data during 8 consecutive days using a custom-built smartphone app. After rigorous pilot testing and feedback from citizen scientists, a time-triggered EMA was developed to capture daily prospective PA. The EMA enabled reporting of light, moderate, and vigorous PA, as well as physical and social contexts of PA via complex looped linking of intensity and context questions. Retrospective PA was reported using International Physical Activity Questionnaire (IPAQ). For both measures, PA intensities were categorized into mean light and moderate-to-vigorous PA/day. Wilcoxon signed ranks tests and Spearman correlation procedures were conducted to compare PA intensities reported via EMA and IPAQ.

Findings: Daily time-triggered EMAs were able to capture not only prospective light and moderate-to-vigorous PA, but also enabled PA reporting across varied physical and social contexts. Moreover, EMA and IPAQ intensity measures showed moderate correlation.

Conclusions: These findings suggest that time-triggered mobile EMAs are an effective method to record comprehensive prospective PA accumulated across multiple physical and social contexts. With approximately 6 billion smartphones estimated to be in circulation by the year 2020, these ubiquitous tools can be leveraged via citizen science to understand active living patterns of large populations in free-living conditions through EMAs. 

Purpose: The University of Vermont Wellness Environment (WE) program is a neuroscience-inspired behavioral change program to promote a healthy environment through classroom and residential halls. A customized application ‘WE App’ was developed to incentivize healthy behaviors such as exercise and meditation among college students. A 14-item survey was administered daily through the App to monitor wellness behaviors. The purpose of the study is to evaluate the association between self-reported exercise and other wellness behaviors from a large cohort of college students.  

Method: A total of 668 WE and 596 non-WE participants who were college freshman and sophomores provided daily survey data. The average number of daily surveys completed per participant was 136 out of a possible 209 days from October 2017 to early May 2018. Generalized linear mixed models were used to estimate the association of exercise and other wellness and risk behaviors for both WE and non-WE students after controlling for gender, race, and academic year.

Results: Results revealed a significant association between higher engagement in exercise and better mood (β=0.1, p < .0001), shorter sleep duration (β=-0.05, p < .0001), higher consumption of fruit or vegetable (β=0.1, p < .0001), higher consumption of water (β=0.19, p < .0001), and less non-academic related screen time (β=-0.04, p < .0001). At baseline, compared to non-WE participants, WE participants had statistically significantly higher daily consumption of fruit and vegetables (p = .0006), more mindfulness practice (p < .0001); and lower prevalence of overall alcohol use (p < .0001), having a shot of liquor (p < .0001), using marijuana (p <. 0001), smoking cigarettes (p < .0001), using illicit drugs (p = .005), and taking unprescribed pills (p = .0034).

Conclusion:  These findings demonstrate favorable outcomes for using technology to track health and risk behaviors among college students. Exercise was positively associated with mood and a range of health behaviors and negatively associated with multiple types of substance use, suggesting that exercise may be an important target for health-promoting interventions among undergraduate students.

Purpose: Ecological momentary assessment (EMA) has become an increasingly popular survey methodology due to its strengths in capturing exposures and health-related behaviors (e.g., physical activity and dietary intake) that vary throughout the day. A concern is that multiple surveys administered many times a day could result in burden or panel conditioning that may deflate measure validity. The current study examines EMA survey burden and patterns of variation across days.

Methods: The Family Matters EMA study was administered to a diverse population of American, primarily low-income families in a Midwest urban city (N=150). Primary caregivers (91% female) responded to a minimum of four daily surveys administered over the course of a week to understand characteristics of the home food environment. Caregiver daily survey burden and overall mental health (depressive symptoms, coping, and overall stress) was measured to assess difficulty in completing surveys that day. Time series analyses with conditional fixed effects regression modeled within-participant variation in survey burden.

Results: Average burden was 1.2 ± 1.1 indicating overall low survey burden. Parents reporting higher burden were more likely to be born outside the United States (P=0.02) and to prefer speaking a language other than English inside the home (P=0.04). Across 1,392 survey days, participants reported no burden 25% of the time. Severe burden was rare (12% of days), affecting less than half of respondents (n=66). Burden did not increase as the study progressed, caregiver stress level and depressed mood were positively correlated with EMA burden (P=0.003 and P=0.009 respectively), and compliant days were less burdensome (P<0.001) and were predictive of lower next day burden (P=0.001). Parent survey burden was not different on weekends and weekdays (P=0.511).

Conclusions: EMA methodologies appear to be a reasonable design to assess how parent-level exposures relate to child dietary intake and physical activity throughout the day. Burden appeared transient in the current study which may indicate external factors, rather than the survey instruments, affected burden. Researchers should develop strategies to support foreign-born and non-English speaking participants to capture complete observation days. Multiple data collection methods (dietary recalls and accelerometry) may minimize potential missing data on non-compliant days.

Purpose: We aimed to study the association between the evolution of objectively-assessed Physical Activity (PA) patterns and weight change during a six-month period within a real-life setting. Physical activity was assessed objectively via wearable activity trackers and not through self-reported questionnaires which are known to be prone to social desirability and recall bias. The originality of our study comes from the fact that our results are based on one of the few large, real-life, prospective studies of a large population of 26,935 connected users of commercially available digital health tools.

Methods: We analyzed data from 26,935 connected device users (wearable activity trackers and connected digital scales), with 11,911,291 available measures of daily steps and 12,357,814 available weight measures. Users were categorized according to their six-month weight change as stable weight, weight gain (>5% of initial weight), and weight loss (<-5%). PA patterns were derived from information on both PA level and regularity of the level of PA practice, which were estimated using daily steps values. Multinomial logistic regression models were used to analyze initial PA patterns and their evolution over six months in association with the six-month weight change.

Results: Our results suggest in our population that evolutions of PA patterns characterized either by maintaining globally a high PA level (i.e. average PA level ≥ 5,800 steps/day) within a six-month period or an increase in the PA level resulting in an average daily steps ≥5,800 steps/day at the end of this six-month period were associated with weight loss [odds ratio (OR)=2.22 (95% CI: 1.97-2.49), OR=2.47 (95% CI: 2.06-2.95), respectively].

Conclusions: Our findings indicate that increasing PA levels, irrespectively from the baseline level, may be beneficial in the short term. Our results also suggest that health benefits can already be observed below the 10,000 steps per day and emphasize the idea that “some PA is better than none”. Our work may have important public health implications when encouraging adults to engage in PA that is monitored as steps/day, especially at-risk low-active adults for which adherence to the 10,000 steps/day may be too ambitious or unrealistic to achieve.

Purpose

Diet monitoring and assessment is becoming increasingly crucial for individuals living with a diet-related disease or wanting to follow a healthy lifestyle, as well as healthcare professionals aiming to monitor or assess their patients' diet or the eating habits of populations. goFOOD^TM uses artificial intelligence algorithms, smartphones and embedded sensors for time- and cost-efficient dietary monitoring and assessment, with accuracy being the cornerstone of the entire research effort.  

Methods

The goFOOD^TMLite app is designed and developed for visual data recording of eating habits. The app allows the recording of food/beverage images or videos and provides a diet log to the individual or healthcare professional. The acquired data are concurrently used to enhance an algorithmic pipeline implementing the automatic detection, recognition, segmentation and 3D reconstruction of food. The information about the type, segment and volume of food is used along with food composition databases, in order to estimate its calories and macronutrient content (carbohydrate, protein, fat). The goFOOD^TM app provides different versions that address the needs of both dietitians and the general population for real-time, cost-efficient, automatic dietary assessment. With the use of a video or two images goFOOD^TM outputs the meal’s calories and macronutrient content in kcal and grams, while with the use of one image it outputs the nutrient content as a traffic-light system.

Android users need to capture two separate images or a video and to place a designated reference card beside their meal for proper size estimation. iPhone X users are able to simultaneously capture both images with one shutter click, since the app exploits the iPhone’s two integrated cameras. No reference card is required in this case, as well as in the case of the single image input.

Results

goFOOD^TM supports 24 broad and 324 fine food categories. It is evaluated on the MADiMa2017 database, which contains 80 central-European dishes with known food categories, weight, volume and nutrient content. The average error percentage in volume estimation is in the order of 20%. 

Conclusions

The goFOOD^TM versions address a variety of needs and exploit different technologies adjusting their functionalities for accuracy, simplicity or speed.

Purpose: Fitspiration is a social media phenomenon purported to inspire viewers to lead healthier lifestyles, however psychological impacts aren't well understood. Some fitspiration images have negative effects on body image, and research is yet to confirm whether credible fitspiration accounts can positively influence physical activity without producing negative psychological effects. This study aimed to (1) develop a screening tool to identify credible Instagram fitspiration accounts and (2) test the reliability of the tool.

Methods: Aim 1: Social media literature was examined to guide development of the screening tool. A video recording is taken of the Instagrammers' bio and most recent 15 posts and screened in three stages. Stage 1 examines the images. Accounts are excluded if any images contain nudity or inappropriate clothing, sexualisation or objectification, portrayal of extreme bodies (extreme thinness or muscularity), or fewer than 4 fitness-related posts. Stage 2 examines captions and hashtags. Accounts are excluded if any content contains thinspiration or other negative messages (e.g. encouraging unhealthy attitudes towards the body, diet or exercise). Stage 3 collects information about the account holder (e.g. demographics) and content (types of posts) from posts and account bios. Aim 2: The tool was trialled on 100 Instagram fitspiration accounts identified through Google searching (search phrase "top Instagram fitspiration accounts"). The first 50 Google results were examined. Accounts were included if they were mentioned in >1 Google result (n=60), or based on highest number of followers (n=40). Inter-rater reliability of two independent raters was determined using kappa.

Results: From the 100 accounts audited, 65 were excluded at stage 1. Two were excluded at stage 2. Agreement between raters was excellent (96%, κ=0.91, p<.001). Stage 3 (n=33 included accounts) indicated that account holders were predominantly female (63%), from the US (67%) and held a relevant qualification (58%). Posts included varied content, such as details of workouts, exercise videos, inspirational quotes, and other lifestyle posts.

Conclusions: This project developed a feasible and reliable tool to identify credible Instagram fitspiration accounts that may have a variety of research applications in future.