This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
The integration of body-worn sensors with mobile devices presents a tremendous opportunity to improve just-in-time behavioral interventions by enhancing bidirectional communication between investigators and their participants. This approach can be used to deliver supportive feedback at critical moments to optimize the attainment of health behavior goals.
The goals of this systematic review were to summarize data on the content characteristics of feedback messaging used in diet and physical activity (PA) interventions and to develop a practical framework for designing just-in-time feedback for behavioral interventions.
Interventions that included just-in-time feedback on PA, sedentary behavior, or dietary intake were eligible for inclusion. Feedback content and efficacy data were synthesized descriptively.
The review included 31 studies (15/31, 48%, targeting PA or sedentary behavior only; 13/31, 42%, targeting diet and PA; and 3/31, 10%, targeting diet only). All studies used just-in-time feedback, 30 (97%, 30/31) used personalized feedback, and 24 (78%, 24/31) used goal-oriented feedback, but only 5 (16%, 5/31) used actionable feedback. Of the 9 studies that tested the efficacy of providing feedback to promote behavior change, 4 reported significant improvements in health behavior. In 3 of these 4 studies, feedback was continuously available, goal-oriented, or actionable.
Feedback that was continuously available, personalized, and actionable relative to a known behavioral objective was prominent in intervention studies with significant behavior change outcomes. Future research should determine whether all or some of these characteristics are needed to optimize the effect of feedback in just-in-time interventions.
Recent advancements in technology, particularly the advent of activity monitors and other wearable body sensors, have the potential to influence innovations in diet and physical activity (PA) assessment and interventions. According to 2013 Pew statistics [
Wearable sensors, particularly Internet-connected sensors, can dramatically enrich the temporality and frequency of health behavior data collection by facilitating self-monitoring and reducing self-report biases. Another important but less realized advantage of wearable sensor technology is its bidirectional communication capability. The latest health trackers are equipped with interactive software apps housing algorithms that allow data to be processed in real time to deliver actionable feedback at critical moments in a person’s daily life to facilitate the attainment of predetermined health behavior goals. These features are likely to enhance bidirectional communication between investigators and their study participants or between patients and their health care providers, thereby improving the users’ engagement with the technology and subsequently facilitating intervention adherence and improving health outcomes. Technology-enhanced interventions are likely the future of behavior change research; however, the use of fast-advancing technologies that enable just-in-time interventions is outpacing the adaptation of theory-based intervention design [
Performance feedback is a key, theory-based behavior change strategy [
Several studies have reviewed the efficacy of digital health technology to promote weight control, PA, and healthy diets. These reviews generally support the use of technology for self-monitoring and intervention delivery, but they also acknowledge that the content and design of future interventions will need more rigorous evaluation to optimize their effects [
The primary goal of this review was to (1) Provide a review of diet and physical activity (PA) interventions that use just-in-time feedback as a behavior change technique (BCT); (2) Characterize key aspects of the reviewed studies’ feedback content characteristics, prompting style, and delivery methods; and (3) describe how the implementation of these key aspects differed by studies that found significant effects of using feedback to motivate behavior change (intervention efficacy). Our secondary goal was to develop a practical framework for designing feedback that could be incorporated into technology-enhanced just-in-time interventions. We focused on feedback characteristics inferred from the theoretical and historical foundations of the use of performance feedback as a health behavior change strategy: timeliness, personalization, action orientation, and goal orientation.
Two authors (SMS and YL) with the assistance of a medical librarian (RSH) devised systematic strategies to search the Ovid MEDLINE, Ovid EMBASE, PubMed, Cochrane Library, Scopus, and PsychInfo databases for all relevant literature published through December 2016. Searches were limited articles written in the English language and conducted in humans. Database search strategies included the use of controlled vocabulary (eg, Medical Subject Headings and Emtree) and keywords to identify studies addressing PA or diet in conjunction with feedback. Keywords included physical activity, exercise, diet, eating, intervention, and feedback. Additionally, the bibliographies of topically relevant review papers and all included studies were examined to identify any additional studies.
Eligible studies included (1) Just-in-time feedback as an intervention component and (2) Targeted behavior changes that included PA, sedentary behavior, or dietary intake. Just-in-time feedback was defined on a case-by-case basis as any feedback that focused on participants’ daily PA, sedentary behavior, or dietary intake and that was provided within 1 min to 1 day of assessing current performance, as appropriate to each intervention or behavior change goal. Studies with multiple feedback components were included, but at least one type of delivered feedback must have met this definition of just-in-time feedback.
Studies were excluded if (1) the intervention-targeted behaviors were off-topic (eg, studies of clinical education, personnel, management, medication adherence, blood glucose self-monitoring, and symptom management); (2) no intervention outcome results were reported (eg, protocol papers); (3) the time frame for providing feedback was greater than 1 day (eg, weekly performance summaries); (4) no or inadequate feedback was provided (eg, studies of performance tracking without evaluation); or (5) they reported the secondary outcomes of an included intervention.
Data were extracted into a structured coding form according to Preferred Reporting of Systematic Reviews and Meta-Analyses (PRISMA) guidelines [
Assessing each study for risk of bias was performed using the 2010 version of the Quality Assessment Tool for Quantitative Studies [
A meta-analysis was not possible owing to substantial heterogeneity in study design, study quality, intervention type, and outcome measures, as well as a lack of studies that explicitly tested the efficacy of using feedback as a BCT. Furthermore, the primary and secondary outcomes of the studies varied widely. We limited our summary of results to primary and secondary study outcomes that were specific to changes in PA, dietary intake, or body weight or body composition. Consistent with the design of a previous review [
The literature search yielded 4239 studies, of which 909 were duplicates, leaving 3330 articles to be screened for eligibility. A total of 3083 articles were excluded upon title or abstract screening because they were unrelated to diet or PA, had no reported outcomes, were nonintervention studies, or had ineligible feedback features. Thus, 246 full-text articles were assessed for eligibility. After 215 articles that did not meet the inclusion criteria were excluded, 31 studies with a total of 6623 participants were included in the review (see PRISMA diagram,
The studies’ characteristics are summarized in
The 31 studies’ risks of bias are summarized in
Preferred Reporting of Systematic Reviews and Meta-Analyses (PRISMA) diagram.
The 31 studies varied in the content, frequency, timing, and delivery of feedback, as well as in the methods used to monitor targeted behaviors and the theoretical foundations that guided the feedback content.
In 24 studies [
By its design, this review included only studies that delivered just-in-time feedback. In 10 studies [
A total of 17 studies [
The studies’ outcomes are described in
Regarding feedback content, 3 of the 4 studies with significant findings [
On the basis of the results of our review, we developed a practical framework (
Practical framework for designing just-in-time feedback.
Most behavioral theories posit that setting a goal is a key behavior change strategy. Behavioral objectives can be framed as goals or targets. Behavioral goals reflect a desired outcome that is achieved incrementally (eg, over a day); can be static (eg, 10,000 steps per day), or adaptive (adjusted incrementally based on performance achievements) and can be self-selected or assigned by an interventionist or health care professional. Comparatively, behavioral targets can be considered intermediate markers or behavioral mediators of achieving a goal that may be explicitly stated or implicitly understood. For example, a behavioral target could be to take a 5-min activity break after sitting for an hour (ie, reduce sedentary behavior) in an effort to achieve a 10,000-steps-per-day goal (ie, increase PA). Thus, behavioral targets are often set at shorter intervals (ie, within the day) than are goals to keep individuals on track for goal attainment. Behavioral targets can also refer to maintaining a continuously assessed marker within a certain predefined range from moment to moment (ie, blood glucose levels) and can be set at moment-to-moment level to achieve an often implicitly understood distal goal (ie, glycemic control). The chosen behavioral objective determines the context for which the feedback content is designed, as well as the appropriate time frame in which it is delivered.
The method used to measure the behavior on which feedback is being provided is an important factor for interventions that incorporate just-in-time feedback. The assessment method determines the cognitive load of self-monitoring, the quality of the data, and the frequency and timeliness with which feedback is delivered. Advancements in wearable sensor technology influence these aspects of data collection, particularly in the area of PA behavior. Newer activity monitors can not only objectively and continuously measure movement (eg, steps) and estimate energy expenditures but also transfer data to another device (eg, mobile phones) or servers wirelessly. This enables the generation and delivery of feedback without any user-initiated input.
Just-in-time feedback is defined as providing the right support at the right moment and in the right amount [
Rather than being generic or group-based, personalized feedback is based on an individual’s own performance and goal. The intent of personalizing feedback is to inform a person about his or her current performance relative to his or her behavioral goal or target (ie, the discrepancy). This message can be communicated visually as graphs or charts (eg, a progress bar) or quantified as text (eg, 3000 more steps to meet your goal) to meet the research needs or the user’s preference.
Feedback that is actionable aims to instruct a patient or participant to engage in behaviors that will improve the likelihood of goal attainment. Action plans are designed to promote small or large behavior changes with a high likelihood of success and should indicate
Our comprehensive literature search yielded 31 studies that met our eligibility criteria. Most of these studies provided feedback that was
To our knowledge, our review is unique in that its primary goal was to examine the use of just-in-time feedback in diet and PA interventions. Only one other similar systematic review has been conducted. In that review [
Another key finding of our review was that only a few of the reviewed studies provided goal-oriented feedback that was actionable. Of the 31 studies included in the review, 23 incorporated behavioral goals; however, only 5 [
Our review is strengthened by its focus on key theory-based characteristics of feedback delivered as behavior change interventions. We focused on only diet and PA interventions rather than looking more broadly across additional health behaviors. We did this in part because it is unclear how generalizable our findings might be to other behaviors or health-related outcomes. In addition, studies eligible for inclusion could have included multiple types of feedback, but at least one form of feedback had to meet our definition of just-in-time feedback. We believe this approach strengthens this review by enabling us to make conclusions that facilitate the progress of intervention science into a future in which feedback can be generated and delivered just in time, thereby preparing researchers for continued advancements in technology. Finally, our synthesis of the available data enabled us to develop a framework for designing just-in-time feedback for health behavior change interventions.
Despite these strengths, we were unable to conduct a meta-analysis primarily because of the variability in targeted behavior and study outcomes. Additionally, because it was not clear whether the included studies monitored the delivery, receipt or viewing, or comprehension of the provided feedback, we were not able to conclusively determine the efficacy of using feedback or which feedback feature(s) might be more effective than others. However, we found that feedback was continuously available, goal-oriented, or actionable in 3 of the 4 studies with significant intervention effects. In addition, the sample sizes, intervention durations, and interventions outcomes of studies with significant findings ranged widely. Interventions of longer duration could have been more likely to have significant findings. However, the duration of studies with significant findings was generally shorter than studies with nonsignificant findings (4 weeks to 12 months vs 2 weeks to 24 months). Excluding studies for providing feedback more than 24 hours after a person performed the target behavior limited the number of eligible studies; however, we believe that it was consistent with the advancement of body sensor technology and therefore important to the context of the review. Another important limitation to consider is that most of the included studies did not recruit participants from a representative, diverse population, thus limiting the generalizability of the findings. Finally, the cost of wearable body sensors and wireless devices could be a potential limitation for scaling up the technology-based interventions.
Before this study, few reviews had critically examined the use of feedback in diet and PA interventions. As advancements in technology continue to improve bidirectional communication between investigators and their participants, optimizing feedback messages will be key to future interventions. The systematic review and the framework we propose represent a foundation for designing feedback messages for future just-in-time diet- or PA-based interventions. Investigators may use the framework to ensure feedback developed for their interventions contain content that is theoretically and empirically supported to have a positive effect on behavior change. However, it is unclear from this review how many of the proposed components are needed to effectively motivate behavior change. Empirical research will be needed to determine the optimal combination of feedback components.
Data extraction form.
Characteristics of included intervention studies.
Risk of bias summaries for included intervention studies.
Descriptions and summaries of feedback features in included intervention studies.
Summaries of outcomes and feedback efficacy for included intervention studies.
behavior change technique
feedback intervention theory
intervention group
physical activity
personal digital assistant
Preferred Reporting of Systematic Reviews and Meta-Analyses
social cognitive theory
The authors acknowledge Catalina Malinowski for her assistance with article review and data extraction. The preparation of this manuscript was supported in part by NIH grants R21CA215415, R01HL119255, R01CA186700, R25ECA056452, and P30CA016672; a Janice Davis Gordon Memorial Postdoctoral Fellowship in Colorectal Cancer Prevention; an Institutional Research Grant from The University of Texas MD Anderson Cancer Center; the Chandler Cox Foundation; the Duncan Family Institute for Cancer Prevention and Risk Assessment; and the Center for Energy Balance in Cancer Prevention and Survivorship. The supporting sources had no involvement in the study design; collection, analysis, and interpretation of data; writing the report; or the decision to submit the report for publication.
None declared.