Mobile messaging is often used in behavioral weight loss interventions, yet little is known as to the extent to which they contribute to weight loss when part of a multicomponent treatment package. The multiphase optimization strategy (MOST) is a framework that researchers can use to systematically investigate interventions that achieve desirable outcomes given specified constraints.
This study describes the use of MOST to develop a messaging intervention as a component to test as part of a weight loss treatment package in a subsequent optimization trial.
On the basis of our conceptual model, a text message intervention was created to support self-regulation of weight-related behaviors. We tested the messages in the ENLIGHTEN feasibility pilot study. Adults with overweight and obesity were recruited to participate in an 8-week weight loss program. Participants received a commercially available self-monitoring smartphone app, coaching calls, and text messages. The number and frequency of text messages sent were determined by individual preferences, and weight was assessed at 8 weeks.
Participants (n=9) in the feasibility pilot study lost 3.2% of their initial body weight over the 8-week intervention and preferred to receive 1.8 texts per day for 4.3 days per week. Researcher burden in manually sending messages was high, and the cost of receiving text messages was a concern. Therefore, a fully automated push notification system was developed to facilitate sending tailored daily messages to participants to support weight loss.
Following the completion of specifying the conceptual model and the feasibility pilot study, the message intervention went through a final iteration. Theory and feasibility pilot study results during the preparation phase informed critical decisions about automation, frequency, triggers, and content before inclusion as a treatment component in a factorial optimization trial.
ClinicalTrials.gov NCT01814072; https://clinicaltrials.gov/ct2/show/NCT01814072
Mobile device use is ubiquitous in the United States: in 2018, 95% of adults owned a cell phone and 77% owned a smartphone [
Messaging (also referred to as
Although the first randomized controlled trial examining messaging as a health intervention was only conducted in 2005 [
One common functionality of messaging is merely to deliver prompts or reminders about behaviors that need to be performed daily or multiple times per day. The complexity of reminders can vary from a once daily prompt (eg, to take medication [
Personalized and tailored messages are established to be more effective than generic messages [
MOST comprises several phases as depicted in
The multiphase optimization strategy (MOST).
The first objective in the preparation phase of this work was to define a conceptual model to guide the message component development and testing. We did so in alignment with social cognitive theory. The second objective was to determine the feasibility of constructing and delivering messages to participants and the acceptability and preferences regarding the messages to inform implementation of the message component in the future optimization trial. Finally, we structured the logic and fully programmed the component to be tailored and automatically delivered with fidelity.
Social cognitive theory is one model for behavioral change that can be helpful in describing how a messaging intervention could contribute to weight loss [
Behavioral facilitation, or the provision of instrumental and informational support, refers to supporting the individual with knowledge, skills, resources, or adjustments to the environment that can make unhealthy behaviors easier to change [
The proposed conceptual model (
Optimization of Remotely Delivered INtensive Lifestyle Treatment for Obesity Study (Opt-IN) messaging conceptual model.
To test feasibility and acceptability of messaging and preliminary efficacy for producing weight loss, an 8-week weight loss pre-post feasibility pilot study was conducted using a commercially available smartphone app (ie, Lose It!), coaching calls, and text messaging. The purpose of this feasibility pilot study was to determine whether the messaging strategy was acceptable to participants and to determine the feasibility of messages. As such, the feasibility pilot study was not powered. The ENLIGHTEN study was approved by Northwestern University’s Institutional Review Board (STU00078810).
Participants were recruited from a large Chicago area employer. Enrollees had to be between 18 and 60 years of age, have a BMI between 25 and 40 kg/m2, weight <300 lb, be weight stable, and own an Android (Google Mobile Services) or iPhone (Apple) smartphone. Participants were excluded if they had unstable medical conditions, high risk for cardiovascular symptoms with physical activity, diabetes requiring insulin, Crohn’s disease, obstructive sleep apnea, active binge eating, substance abuse or dependence, or active suicidal ideation. Participants were also excluded if they took medication known to cause significant weight gain or loss, used an assistive device for mobility (eg, wheelchair, walker, cane), had been hospitalized for a psychiatric disorder within the past 5 years or could not read the study questionnaires. Female participants could not be pregnant, trying to get pregnant, or lactating. Each participant completed an informed consent process and selected the timing and frequency of text messages (up to 3× per day, including weekends) that were manually sent to them by a coach throughout the week. Text messages addressed the topics of diet, physical activity, and weight change. Each message either targeted the construct of supportive accountability (eg, “Way to get that exercise in this week!”) or facilitation (eg, “It’s a beautiful day, get out, and enjoy an after dinner walk!”).
The ENLIGHTEN feasibility pilot study enrolled 9 participants (6/9, 67% female; 5/9, 56% black; mean age 42.4 years, mean weight 197.2 lb, mean BMI 31.8 kg/m2). Participants who completed the intervention (n=8) lost an average of 3% of their initial body weight (range: +0.75 lb to −14.75 lb) and preferred to be sent an average of 1.8 texts per day, on 4.3 days of the week, with a range of 2 to 7 texts per week.
The final stage of our preparation phase was to fully develop the message treatment component such that it could be implemented in the next phase, the optimization trial. To reduce the burden on staff and ensure fidelity of the treatment, we automated message sending by specifying programming logic for frequency and timing, content sent, and tailoring in response to user status.
One limitation of the feasibility trial to note was the significant staff burden required to manually send text messages to participants. As such, automation of sending messages was of paramount importance during further message program development. Automating, via a push notification protocol, not only reduces staff time but also enables the interventionist to maintain treatment fidelity due to the lack of human error that can produce untimely, missed, or poorly constructed text messages. Text messages are typically sent via SMS, which utilizes a 160-character restricted protocol. In contrast, we use Apple’s push notification service for iOS and Google Cloud Messaging service for Android. In style, length, and prominence, push notification messages are intended to be similar to SMS text messages. The advantages of the technology structure or architecture are 2-fold. First, neither participants nor study staff pays text message charges through their phone plan. Second, messages are integrated within the smartphone app to be used for self-monitoring in the optimization trial. Participants are alerted when a message arrives through content that pops up on either the iPhone’s lock screen or the Android’s notification bar. Clicking on the message opens the smartphone app and guides the participant to the message, providing an opportunity for the participant to continue engaging with the app after opening the message.
Specific timing of the messages was structured to balance the message value with the burden that notifications place on the participant [
Message preference schedules for participant selection.
Preference optiona | Monday | Tuesday | Wednesday | Thursday | Friday | Saturday | Sunday |
A | T1b | T2 | T3 | T4 | T5 | T6 | T7 |
B | T1/T2 | N/Ac | T3/T4 | N/A | T5/T6 | N/A | T7 |
C | N/A | T1/T2 | N/A | T3/T4 | N/A | T5/T6 | T7 |
aSchedule of text messages to be sent across a week.
bT: Text.
cN/A: not applicable.
The message trigger acted as the beginning of a logic structure that we hypothesize to increase likelihood of seeing the message (
Logic structure for Optimization of Remotely Delivered INtensive Lifestyle Treatment for Obesity Study (Opt-IN) text message triggers.
The content of all messages relates to behavioral facilitation or supportive accountability from the conceptual model in
Sample weekly text schedule with example content.
Weekly | Monday | Tuesday | Wednesday | Thursday | Friday | Saturday | Sunday |
Topic | Weight | MVPAa goal | Adherence | Behavioral | MVPA | Fat goal | Calorie goal |
Message | You have lost X pounds in the past week! Way to go—remember to maintain these behavior changes as you continue with your weight loss! | Try to schedule in your activity! Make it a priority so you can meet those goals! | Focus on getting back on track with food entry on your phone in order to help you know your caloric and fat intake! | Adding in a small breakfast like a yogurt or an apple can make a big impact on your weight loss success! | Nice work exercising this week! Keep it up for the rest of the week. | If you're having trouble adding the right kinds of fat into your diet or recording your foods, don't be afraid to ask for suggestions on your next call. | It looks like you went over your calorie goal yesterday. Don't worry, slips are normal! Get back on track today. |
aMVPA: moderate-vigorous physical activity.
To increase relevance and engagement, the text content is tailored to the individual’s personal progression toward study goal attainment (ie, weight loss, daily calorie intake, daily fat gram intake, and physical activity). As such, most messages delivered are based on what the participant has (or has not) self-monitored within the smartphone app. The texts are designed to encourage and reinforce positive participant behaviors, not to be negatively critical or discouraging. Hence, content was written to reinforce not only achieving goals, but also coming close to goal attainment, much like a human health coach might. For example, if a participant with a 1200 daily calorie goal enters 1220 calories, a calorie goal attainment text message would still reflect that they did well staying in the calorie range, rather than telling them they exceeded their daily calorie goal. Similarly, if a participant had not self-monitored food in the app on the day an adherence to self-monitoring text was to be sent, but had self-monitored food the prior day, the adherence to self-monitoring text would reinforce their recording on some days, while encouraging them to continue setting patterns (
Optimization of Remotely Delivered INtensive Lifestyle Treatment for Obesity Study (Opt-IN) message tailoring: adherence to self-monitoring example.
Tailoring messages to make them feel both personalized and as if they are coming directly from the health coach was also considered an important factor in maintaining a strong therapeutic alliance between participant and their coach. Based on the assumption of this kind of social presence, we posit that the simulated human connection through messages created with this structure will be more efficacious than a
As part of the MOST framework, critical preparation work was completed to develop a messaging protocol that was theoretically grounded, responsive to user feedback, easy to scale in a remotely delivered intervention, tailored to the individual’s current state, and feasible to deliver with fidelity in a factorial trial. Despite widespread use of messaging as a component of health behavior change interventions, information about the derivation of message content is often absent, leaving it unclear how and why the messages were developed and implemented in a particular way [
The development of our messaging component included designing messages based on theory, testing the feasibility and acceptability of the messages, and identifying participant preferences regarding message receipt. The resulting messages are personalized, timely, and relevant to the participant, which may increase the likelihood that the participant will respond in a desirable manner by maintaining or improving a targeted behavior. The preparation work also allowed us to create a component with potential for high treatment fidelity, a critical need in intervention trials and especially for factorial trials with a high number of randomized conditions. The resulting technology has the unique ability to deploy messages either in response to interaction with the smartphone app or based on the participant’s preferred schedule. This capitalizes on participant receptivity: when the individual is actively using the app, a time they are most likely thinking about their health behaviors. If an individual is not interacting with the study app, sending messages based on their preferred schedule may provide a real time-time helpful reminder to re-engage with target health behaviors to prevent a lapse.
The features of our message component may very well be perceived as an artificially intelligent system. Messages are sent at times of high receptivity, with content that uses current participant status, and that responds in a way to support the self-efficacy of an individual much in the way a live health coach might. Human staff have availability constraints, cost a significant amount, and are prone to make errors in intervention delivery. Therefore, using human staff to coach individuals in a weight loss program has many barriers to scalability. The message program design described, if effective, may well be a first foray into artificial intelligent coaching that can provide just-in-time adaptive interventions. In sum, these essential preparatory activities supported the development of a theoretically sound, scalable, and low-cost treatment component that was feasible to deliver in our optimization trial design.
One possible critique of our work is that it had a significant upfront cost to design and program sufficient to meet our requirements and confer a realistic and human feel. The overall cost might be worth time and effort downstream if it relieves staff time across enough participants over enough time. One benefit is that once programmed, as we have done, it can be scaled up and distributed widely with a relatively low maintenance cost, an important requirement set during our preparation phase. Comparatively, many intervention components are developed as part of multicomponent treatment packages, sometimes without critical preparatory work, and tested as such in a randomized controlled trial, the results of which are unable to reveal whether the component has any important or significant effect on the outcome. We believe the upfront cost of preparation in the context of MOST is warranted to develop rigorous, robust, and testable components.
Although weight loss programs that include a text messaging component have been effective [
Although our preparation phase work and the subsequent Opt-IN trial will make progress in optimizing a cost-contained treatment package for obesity, its limitation relevant to mobile messaging is that it will inform the utility of including messaging in a one-size fits all treatment package. One could also optimize weight loss by testing the use of message systems in stepped sequences as in a sequential multiple assignment randomized trial [
All authors acknowledge prior support from R01DK097364 through the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK). SM is currently supported by F31DK120151 through NIDDK. BS and AP are also supported by UL1TR001422 through the National Center for Advancing Translational Sciences.
Conflicts of Interest: BS serves on the scientific advisory boards of Arrivale and Actigraph. The remaining authors declare no conflicts of interest.
multiphase optimization strategy
National Institute of Diabetes and Digestive and Kidney Diseases
Optimization of Remotely Delivered INtensive Lifestyle Treatment for Obesity Study