Details of the Partner2Lose protocol have been published [19]. In brief, Partner2Lose was a parallel, 2-arm randomized controlled trial that carried out a comparative evaluation of a partner-assisted intervention and a participant-only intervention for weight management. The primary outcome was participant weight at 24 months. To be eligible to enroll, participants must have (1) been cohabitating with a partner, (2) had a BMI of 27-29.9 kg/m² and one obesity-related comorbidity or a BMI of ≥30 kg/m², and (3) had a desire to lose weight. Partners had to have a BMI ≥18.5 kg/m² to participate. Additional inclusion and exclusion criteria can be found in the previously published protocol.

Enrollment for Partner2Lose started in January 2019. Five cohorts of 45-50 couples were sequentially recruited and randomized to the participant-only arm or partner-assisted arm. The intervention included 6 months of weight loss initiation and 12 months of weight maintenance, followed by a 6-month period of no intervention. In the partner-assisted arm, partners participated in the intervention alongside the index participant and received communication skills training. Data collection will be completed in March 2023.

Index participants in both study arms received the nutrition, physical activity, and weight management intervention, a standard reduced-calorie weight management approach established in previous trials [20,21]. The weight loss initiation phase consisted of group classes every 2 weeks co-led by a registered dietitian and exercise physiologist and focused on unique dietary and physical activity education topics. At each class, participants were asked to select a goal topic from a menu of 3-4 of them (see examples in Table 1) related to the class education topics. Participants provided their goal selection to a research staff member after the group meeting. These goal selections were used to inform tailored text messages to participants.

Partners in the partner-assisted arm also received communication and support skills training [22,23]. Partners attended half of the group meetings, where they received the same nutrition and activity education as index participants. The participants and their partners were asked to select a partner support plan to support the participants’ goal from a prespecified list of options (see Table 1). Partners provided their support plan selection to a research staff member after the group meeting. These support plans were used to inform tailored text messages to partners.

During month 7, participants transitioned to the weight loss maintenance phase, where participants started to receive telephone support from the registered dietitian. During each of the 8 calls that were delivered, participants reflected on satisfaction with outcomes of weight loss and formed relapse prevention, self-monitoring, and social support plans. Partners in the partner-assisted joined 5 telephone calls during this period.

To maximize the opportunity for scalability, it is beneficial to automate components of the intervention where appropriate. To that end, we used REDCap to automate the delivery of text message communications to participants and their partners for the duration of the intervention. Specifically, 2 databases were developed, which ran in parallel: a weight loss initiation database and a weight maintenance and reminder database. The choice to develop 2 databases was to simplify the database building processes, as the databases were used for different purposes and thus had different programming requirements (described below).

The weight loss initiation database was used to reinforce class contents and support goal achievement during the weight loss initiation phase (months 1-6). The weight maintenance and reminder database was used to deliver a battery of automated reminders throughout the 24-month study period. It was also used to deliver automated text messages that concentrated on behavioral maintenance principles (eg, relapse prevention) during the weight maintenance phase (months 7-18). Both REDCap databases stored text message contents and participant-level information, including role (ie, participant or partner), record ID, and cell phone number.

To specify text message content, conditions, and schedule, we used a feature in REDCap, known as “Automated Survey Invitations” (Figure 1). This feature allows messages to be sent, via email or text message, at a designated time (eg, immediately, in 2 hours, or in 14 days) after a set of prespecified conditions are met. Conditions are specified from data fields in the respective REDCap database, such as participant role. An example condition for the goal database would be “[participant role=partner].” Under this condition, the messages would only be sent to someone if their documented role in REDCap was “partner.” Conditions used for the databases are described in greater detail below.

The default for the Automated Survey Invitation feature in REDCap is for the messages to include a hyperlink to a survey for the recipient to complete. For this study, however, text messages were unidirectional and were not intended to solicit a response; hence, we did not want to send a hyperlink to a survey in our text message. Therefore, we used an open HTML comment to eliminate the hyperlinked survey feature in the body of the Automated Survey Invitation to make the system deliver a plain text message.

To deliver the text messages to participants, we used a third-party communication application integrated within REDCap, known as Twilio (Twilio Inc). Twilio is a cloud communication platform that uses web services application programming interfaces (APIs) to make and receive phone calls and text messages. To do so, a Twilio account, 34-digit string identifier (SID), authentication token (Auth Token), and purchased phone number are required. One Twilio phone number can be used for the duration of the study and for all participants. The Account SID and Auth Token act as the Twilio account’s username and password and are used to inform Twilio from which account the API requests are derived. As such, we created a Twilio account to retrieve an Account SID and Auth Token, which we entered into our REDCap databases. We disabled the “Request Inspector” feature on our Twilio account to ensure that the server did not store participant information. Disabling this feature is a requirement by our institution to remain HIPAA (Health Insurance Portability and Accountability Act)-compliant.

Twilio has a pay-as-you-go payment structure and does not require an upfront fee to make an account. A monthly fee of US $1.00 for a Twilio phone number is required. Phone numbers can be selected on the basis of a local area code. At the time of this intervention, text messaging to domestic numbers costs US $0.0075 per text message.

REDCap automatically maintains a log display or comprehensive list of messages that are delivered or scheduled to be delivered (Figure 3). The log displays the date and time when the message was sent or is scheduled to be sent, as well as the participant’s cell phone number and record ID. The log also displays if there were any errors. The log can be viewed and downloaded at any time to monitor intervention fidelity. Intervention fidelity is further enhanced over long periods of time through automation and limiting the need for a person to send a message. Our study team looked at the log monthly to check for errors and confirm message delivery. If an error was seen, we investigated and resolved it.

Approval for Partner2Lose was obtained by the University of Wisconsin Health Sciences Institutional Review Board in December 2018 (protocol 2018-1400).

To effectively implement interventions in large populations, scalable intervention delivery approaches are needed. This study demonstrates a feasible approach to automate a significant component of participant communication and engagement over a 24-month behavioral weight management trial using a widely available software, REDCap.

The widespread use of mobile technology has expanded our opportunity to automate the delivery of intervention content without requiring face-to-face interactions and with minimal upkeep and maintenance by study staff [1]. Inasmuch, it reduces study burden and limits the need to travel to a study site. This is particularly useful for populations in rural areas or those who may work multiple jobs or have caregiving responsibilities. Text messaging is also cost effective, as indicated in our report and previous research [24], and unlike other digital interventions and strategies, it does not require a smartphone (with access to applications) or other devices, such as wearables, facilitating reach to sociodemographically diverse groups [1,25]. In fact, 97% of adults own a cellphone in the United States, with high rates across all race and ethnicity, household income, and educational attainment groups [1].

There are other examples of automated, tailored interventions being delivered in behavioral health trials. These interventions have used other software tools that require individual development and may involve smartphone apps. One example is the Nourish study [26] that leverages a smartphone application and evidence-based behavior change principles to improve adherence to the DASH eating pattern among adults with hypertension. The Nourish intervention delivers tailored and personalized feedback that is based on dietary tracking and goal attainment. This feedback also takes into consideration personal characteristics of the participant, such as working full time, to inform the feedback provided. Given the increased number of variables being considered in the feedback algorithm, in addition to it relying on an external application, REDCap likely would not be a feasible option for this study or others alike. These are important factors to consider when selecting a software platform. We chose REDCap for this study owing to institutional investment in which it supports a wide variety of research needs, and REDCap’s wide adoption and collaborative, international network. We also only sought to leverage text messaging for intervention delivery owing to its simplicity and equity in use across diverse populations. Thus, results from this study can be more easily replicated.

The REDCap Consortium is an international community of software platform partners [10]. While each institution maintains its own version of the software, developed features are shared across the community. This allows for interoperability of data transfer across sites and best practices related to security and regulatory requirements. The initial costs of using software such as REDCap requires institutional investment; however, there is less of a need to rebuild custom software tools across the enterprise once the infrastructure exists, resulting in reduced long-term spending [27]. Using REDCap, or a similar platform, is also cost savings for individual study teams, as the scalability of automated processes is exponential and more affordable than manual processes that incur labor costs over time. For example, increasing a sample size from 100 to 10,000 would traditionally require a significant increase in labor costs and data tracking; however, software and automation can greatly reduce the per-participant cost.

Automating the participant communication component of our intervention delivery increased our ability to monitor and manage intervention fidelity. Specifically, using REDCap’s survey invitation log feature, we were able to identify messages delivered or scheduled to be delivered that had an error. This feature allowed us to make informed and responsive modifications such as reinitiating messages or stopping double messaging. It should be noted that our intervention did not ask or prompt participants to respond to text messages, so there could be delivery errors not identified by REDCap or our team. Nonetheless, the databases used in this study demonstrated a high rate of fidelity over time in message delivery as demonstrated by the very low error rate.

The tools and features in REDCap are regularly modified and updated to improve the platform’s functionality and effectiveness. As such, the workflow described in this paper would benefit from being adapted to leverage the tools and features available today. For example, if we were to implement automated text messaging in our next trial, we would use REDCap’s “Alerts and Notifications” feature to send the text messages via Twilio. This relatively newer feature provides an option to specify delivery conditions to tailor the text message content by each participant, however, does not require the text message to be connected to a survey in REDCap. The feature described to send text messages in this report, Automated Survey Invitations, would be more appropriate to send text messages or communications that solicit a survey response, such as an outcome survey or adverse event questionnaire. Nonetheless, both of these features have tracking logs, in which intervention delivery and survey completion can be tracked and maintained by the study team.

We note several limitations to automating intervention content and communication using REDCap and Twilio, which should be considered before design and implementation. First, it requires participants to have a consistent mobile phone, which may limit the generalizability of the automated communication approach. Although mobile phone ownership is high and continues to grow in the United States [1], there is still a small percentage of individuals who do not own a mobile phone and thus cannot be reached using this approach. For example, 8% of adults aged 65 years or older do not own mobile phone and would be inaccessible via this modality [1]. Second, the phone number that is used to deliver the text messages via Twilio is a robot. Thus, if someone were to reply to the text message, there would not be a study team member monitoring the responses or be available to reply. This can be a safety concern if a participant is seeking medical advice from a study team member; however, to minimize this concern, we notified participants at the start of the trial that responses would not be monitored. Additionally, it is worth noting that participants received messages from the same phone number over the course of the study and were able to save the number so they would not think it was spam. Twilio has the capability for 2-way SMS text messaging; however, at the time of implementation, REDCap only had the capability of receiving text messages that met predefined criteria. For safety or technical concerns, we provided alternative contact modalities for the participants to reach the trial staff, including an email ID and a phone number. Third, although our data indicate that the error rate was small in this study, it is possible that errors went unidentified as our intervention did not prompt a response from participants. Moreover, if there is a programming error, it is likely to impact several study participants rather than a single participant. Fourth, given the use of text messaging, it was not possible to record if a participant read the intervention communication. Lastly, accounting for holidays and weather can be challenging, and automated messaging may occur when not desired by participants. In the future, this may be mitigated by asking participants communication preferences or tailoring text messages to times of the day that are best suited to them.

The proliferation of mobile phones coupled with research management software offers study teams the opportunity to integrate automation at scale into their intervention implementation and delivery. Our trial automated the delivery of intervention content and communications using REDCap for 5 cohorts of participants and their partners in a behavioral health trial with longitudinal outcomes. Our approach provides a framework that can be used in future behavioral health interventions to create an affordable, reliable and accessible method for intervention engagement.