Gestational diabetes mellitus (GDM) has become a public health problem as the most common complication during pregnancy, with rising numbers globally [1]. Following GDM, women have a 10-fold higher risk for type 2 diabetes [2] and a high risk for metabolic syndrome and other cardiometabolic diseases [3,4]. Previous lifestyle interventions among women with a history of GDM who had already developed impaired glucose tolerance reduced the risk of developing type 2 diabetes by approximately 50% [5]. However, recent research highlighted the importance of starting interventions earlier, particularly in the first year postpartum [6], as the years following GDM offer a window of opportunity for preventive lifestyle interventions to prevent the progression of cardiometabolic disturbances early on [7].

Qualitative research showed that lifestyle interventions within the 10 years following GDM face specific challenges such as prioritization of the family and a mother’s perceived lack of resources [8]. These challenges are reflected in the modest effects of previous behavior interventions in the first few years post-GDM [6,7]. Hence, experts call for innovative, flexible, and personalized interventions specific to the needs of women during the first few years post-GDM and the socio-ecological contexts of health behaviors [9]. These needs include a practical and accessible tool for behavior change support that is compatible with daily family life [9-11]. Mobile health (mHealth) solutions offer easy access, use in daily family life, and integrated behavior change methods [12,13]. At present, the ideal mHealth platform for women post-GDM appears to be smartphone apps since they are frequently used for health purposes by women during the first few years post-GDM [14].

The aim of this project was to systematically plan and develop a theory- and evidence-based mHealth intervention for women during the first 5 years post-GDM that supports behavior change in several lifestyle areas. To ensure that the standards of both academia and industry were met, we decided to develop a smartphone app—the TRIANGLE app—with industrial partners using the intervention mapping protocol. Thereby, we aimed at both tackling some of the limitations from earlier prevention programs for women post-GDM [7,9] and testing the usability of the novel smartphone app with the intended users. This paper provides a detailed description of the systematic planning and development of the TRIANGLE intervention, including the user study.

We planned and developed the TRIANGLE intervention with the intervention mapping approach [15] as much as possible with the given resources. The process was guided by an understanding of the intervention context through the social-ecological model, systems thinking, stakeholder involvement, and ethical principles to identify and decide on the best intervention points. Further, we applied theory- and evidence-based practices using the 6 intervention mapping core processes as overarching principles. The core processes supported the use of different data sources for effective decision making by guiding us to pose a question related to the problem, brainstorm answers, review the literature, access and use theory, consider the need for new research, and define the working list of answers [15]. The 4 of the 6 intervention mapping steps that were applied in the scope of this project were a needs assessment with a logic model of the problem (step 1), a logic model of change (step 2), program design (step 3), and program production (step 4). These 4 steps provide the prerequisite for the 2 remaining steps—implementation planning (step 5) and evaluation planning (step 6)—not included in this publication.

As part of intervention mapping step 4, the TRIANGLE user study (DRKS trial registration: DRKS00012736) tested the novel TRIANGLE app in 2 test rounds with 5 or more participants of the intended target group (women in their first 5 years post-GDM) in a mixed methods design. The user study was conducted at the Medical Center of the Ludwig Maximilian University of Munich. The content of the tested TRIANGLE app was limited for the purpose of this study (Multimedia Appendix 1). Participants gave their written informed consent and signed the TRIANGLE app's privacy statement. The study was approved by the ethics committee of the Medical Faculty of the Ludwig Maximilian University of Munich.

Participants were primarily recruited by phone from the patient base of the Medical Center of the Ludwig Maximilian University of Munich between mid-June and mid-July 2017. The inclusion criteria were a medical diagnosis of GDM in a recent pregnancy, 3 months to 5 years postpartum, postnatal core muscle recovery (assessed by enquiring about the completion of a 10-week course in the German health care system and a question on postnatal sporting restrictions), ownership of an iPhone 5 to 7 Plus (Apple Inc), and fluency in German. The exclusion criteria comprised age <18 years, current pregnancy, cardiopulmonary disease or restrictions in the locomotor system contraindicating a sports intervention, gastrointestinal disease contraindicating a nutrition intervention, psychiatric disease requiring therapy, other serious illness contraindicating a lifestyle intervention according to the principal investigator, and alcohol or drug abuse.

Data collection comprised 2 visits (visit 1 [V1] and visit 2 [V2]) at the Medical Center of the Ludwig Maximilian University of Munich (details in Multimedia Appendix 2) with questionnaires, think-alouds with semistructured interviews (test setup in Multimedia Appendix 3), and app user logs collected during and between the 2 visits (Multimedia Appendix 4). Participants used the TRIANGLE app for a predefined period between the 2 study visits: 1 week (+ maximum 7 days) in group 1 and 4 weeks (± maximum 7 days) in group 2.

At V1, participants installed the TRIANGLE app on their iPhone and accessed their personalized account with an individual login code. They further received a Garmin vívosmart HR fitness tracker (Garmin Ltd), a step stool, and the TRIANGLE paper notepad. Participants were instructed to use the app on 5 or more days/week for the duration of the study, to conduct the fitness self-test, and to use the additional program materials. The client-server system of the TRIANGLE intervention tracked their activities on the app (Multimedia Appendix 4). The questionnaires completed during V1 included a social anamnesis, medical history, smartphone and app usage, the System Usability Scale (SUS) [16], and the user version of the Mobile Application Rating Scale (uMARS) [17]. The SUS and uMARS were also assessed during V2.

During a one-on-one think-aloud session followed by a semistructured interview, participants first received a standardized introduction video to get familiar with the test method (ie, to think aloud) before completing 30 small tasks from a standardized test protocol in the TRIANGLE app. An investigator (ALP) read aloud the tasks one by one and both observed and recorded (using audio and the iPhone screen, upon a user’s consent only) how a user installed and registered on the TRIANGLE app, attended the introduction video, and navigated through the features. Thereby, the investigator was seated outside the user’s field of vision (Multimedia Appendix 3) and noted the user’s actions in the TRIANGLE app, expressed thoughts, questions, problems, and body language in a respective checklist structured per task. If a user interrupted his or her thinking aloud, the investigator gave prompts such as, “What is on your mind?” The semistructured interviews (audio-recorded upon the user’s consent only) addressed the user’s overall impression and subjective experience with the TRIANGLE app, as well as some specifics (at V2 only) about the didactic design, content, expectations, and initial habit change. We aimed at timely insights and improvements via an explorative rapid thematic analysis before the start of the clinical pilot study. Hence, the investigator reviewed the screen and audio recordings and amended the written notes in case of a missed observation during the session instead of preparing a full transcript of each record.

Collected data were pseudonymized (4-digit code) and TRIANGLE app data were encrypted (2048-bit Secure Sockets Layer and end-to-end encryption) and stored on the server of the Medical Center of the Ludwig Maximilian University of Munich.

Each TRIANGLE user group contained 5 or more participants to detect approximately 80% of the issues during usability testing [18,19]. Because of the small group sizes, no statistical group comparisons were conducted. We analyzed all data in Microsoft Excel 2016 MSO (Microsoft Corporation) and Tableau Desktop 2019.3 (Tableau Software, LLC). Values are presented as counts (n) with percentages or as mean (SD). The analyses were based on all participants with the respective complete outcome measure. We excluded app activities collected during the think-aloud sessions for the app usage analyses between the 2 visits. Data from the observations during the think-alouds and the semistructured interview questions were analyzed in an explorative rapid thematic analysis with the respective numbers of participants [20,21].

The demographic characteristics of the user sample are shown in Table 3.

Overall, participants gave positive feedback on the TRIANGLE app during the think-alouds with semistructured interviews. Participants used the app intuitively, except for 1 participant who showed initial problems with the swiping gesture. Further, participants considered the 3 core features easy to use and helpful. The combined data from the think-alouds and semistructured interviews showed 7 themes in our exploratory analysis—onboarding process (Table 4), navigation bar, challenge system, coaching, library, settings, and other feedback—with positive and negative statements for each theme. All positive statements supported the current format of the intervention.

To improve the TRIANGLE app, we phrased crucial negative statements as problem descriptions and formulated the resulting changes (Table 5). We prioritized refinements and revisions of the TRIANGLE intervention based on the frequency and relevance of user feedback and the feasibility within this funding period. The main features of the TRIANGLE app remained the same before and after the pretest (Figure 2, post-testing). Additional changes to the TRIANGLE app besides those listed in Table 5 included a revised introductory video; a recipe section; more specific content for the late postpartum phase, such as pelvic floor training; an information section about the app, colored headings and subheadings per lifestyle area; and a revised paper-and-pencil questionnaire. Based on these adaptations, we revised and produced all of the TRIANGLE program materials.

The goal of this project was to plan and develop one of the first smartphone app–based interventions to change cardiometabolic risk behaviors in women during the first 5 years post-GDM [14,28,92,94]. The resulting TRIANGLE intervention confirmed the value of using the systematic intervention mapping framework to plan a complex health promotion program [95]. The iterative development process led to a detailed theory- and evidence-based intervention model that was translated into the TRIANGLE app and then tested by its intended users.

We attempted to tackle some of the limitations of previous interventions during the first few years post-GDM. One of these limitations was the use of no or a single cognitive theory and few behavior change methods [96]. The more behavior change methods a health promotion program applies, the more likely it will affect behavior [97]. Hence, we used a multitheory approach at the habit-goal interface, applying 39 behavior change methods. The applied behavior change methods addressed 11 behavioral determinants and complied with the respective parameters of each method. The TRIANGLE intervention is the first intervention to prioritize habit change for women during the first 5 years post-GDM based on evidence that most health behaviors are habitual [71] and women are susceptible to unhealthy habits during the first few years post-GDM [98].

The TRIANGLE app contains a comprehensive multimedia database of newly developed lifestyle content with evidence-based information and exercises at the habit level to meet the needs of women during the first 5 years post-GDM. The content includes psychosocial well-being along with the more common DPP goals for nutrition and physical activity [7]. This addition was suggested in recent reviews [58,99] and is neededeg to prevent postnatal depression in women post-GDM [99]. Compared with previous programs during the first 5 years post-GDM, our extensive list of 81 performance objectives distinguishes between preparatory and habitual actions. Further, our performance objectives include many subbehaviors and supportive behaviors to achieve a greater behavioral outcome. In addition, we target major health decisions and communication goals as necessary actions. Few other programs during the first few years post-GDM address postpartum topics such as pelvic floor health and sleep disruptions [98], which are included in the TRIANGLE intervention.

In addition, we considered the real-life intervention context and created a practical tool for both health care practitioners and women during the first few years post-GDM [100]. This tool has the potential for widespread implementation, low-cost intervention delivery in daily life, and human coaching by multiple health care professionals. The resulting tool also shows the importance of academia-industry cooperation in order to fulfil both standards in mHealth, as recently requested by women in the first few years post-GDM [14]. Our academia-industry cooperation led to a highly usable and accepted mHealth product with unique themes, components, scope, and sequence, while most other programs during the first 5 years post-GDM were based on the DPP model. We set up strong technical assistance to support the program engagement goal of at least 5 out of 7 days per week versus the more traditional weekly or monthly interactions [101]. Further, we translated the components of skill training, behavior change support, and knowledge transfer into a unique challenge-coaching-library feature system with flexible pacing. In contrast to the flexible pacing desired by women in the first 2 years post-GDM [11], most other programs followed fixed sequences such as the DPP curriculum [89,102]. Current programs are testing different features for smartphone-based interventions in the first year post-GDM, varying degrees of automation, and cross-links to other platforms and thus will soon provide further insights [14,28,94].

Lastly, the systematic planning and development led to a transparent description of the underlying theory of the TRIANGLE intervention to inform future intervention models, related studies, or projects. To date, such a detailed description of a behavior intervention post-GDM is missing, making it difficult to compare and identify the most successful intervention components for this target group. Hence, our model may be of value for other work groups to replicate or learn from certain decisions at each intervention mapping step.

The TRIANGLE user study demonstrated that all of the main components of the TRIANGLE intervention were accepted by the intended users (women during the first 5 years post-GDM), including a first beta version of the TRIANGLE app for the iPhone. The study confirmed the value of mixed methods usability testing [103]. The SUS ratings of participants indicated “excellent” system usability [104] of the TRIANGLE app. The uMARS quality ratings of the TRIANGLE app (ie, 4.2 out of 5.0 points at V2) lay just below the top range of uMARS quality ratings for health apps (ie, 4.3 points [105])—mainly because of the lower engagement mean score compared with the other subscores. This indicated a need to enhance engagement, yet may also be related to both the limited content and the limited coach interaction in the user study. The increase in the app perceived impact mean score after several days of app usage demonstrated that users perceived a higher influence of the app on their behavior over time, as they may have started to change their health behaviors.

Overall, a smartphone app seemed to match the needs of women during the first 5 years post-GDM because participants used the TRIANGLE app regularly at almost any time of the day throughout the different lifestyle modules. The regular overall usage of the TRIANGLE app confirmed the feasibility of our usage goal of 5 active days per week. However, just like for a similar app [93], usage varied widely between participants. Each participant used each lifestyle area and thus confirmed the value of the added psychosocial module, as suggested by a recent review [58]. The high usage of the interactive challenge system stressed the importance for women to check off completed tasks for self-monitoring [93] and to view their own progress [14] in the first years post-GDM. In contrast, coaching requirements were much lower and indicated a realistic scope for implementation in routine care, especially when considering that some coaching activities may be automated in the future.

During the think-aloud sessions, participants used the app intuitively and considered the features useful. Further, the positive user statements during the semistructured interviews confirmed the chosen features and lifestyle modules. We specified and clustered emerging usability issues and negative statements as problem statements to guide smaller adaptations before the full program production, similar to related apps [14,93]. The desire of single participants for new features will have to be quantified before changing the TRIANGLE app system. Some of these desired features for apps in the first year post-GDM are currently being tested by other work groups, such as a built-in pedometer to translate other types of physical activity into step counts [28].

The development of the TRIANGLE intervention using the intervention mapping approach had several limitations. The comprehensive intervention mapping approach is resource consuming, and this project offered limited human and financial resources and a tight timeline due to the fast-paced technological environment. Hence, shortcuts as suggested by the authors of intervention mapping [15] had to be taken, similar to other projects applying this approach [106,107]. These shortcuts led to limited stakeholder involvement [107], especially in the planning group [15]; the focus on individual behavior change in women post-GDM without targeting the behavior of environmental agents such as their partners [108]; BMI-based tailoring only; and a respective prioritization of performance objectives, change objectives, behavior change methods, and practical applications. Lastly, we conducted intervention mapping steps 1 to 4 only and thus lack the planning and data on large-scale implementation (step 5), evaluation (step 6), and cost-effectiveness of the TRIANGLE intervention.

The user study sample was rather homogenous and may exhibit a socioeconomic bias due to its inclusion of iPhone users only and women with a high education level who were mostly native speakers. Women during the first 5 years post-GDM participated voluntarily, which may have preselected those with a higher motivation to test the TRIANGLE app. Therefore, the sample’s usage of the app, respective feedback, and suggested adaptations may not be representative of the priority population. In addition, the limited content available for the user study may have biased the distribution of activities per lifestyle area. Future studies on the TRIANGLE app need to address these factors and test its usability on the Android system.

In the scope of larger future projects, we suggest to additionally support behavior change by key environmental agents such as partners [108], health care professionals [94], or families [109] and to develop specific content for cultural subgroups that could not be addressed in this project [110]. Further, other cardiometabolic risk behaviors of women during the first 5 years post-GDM that we excluded need to be addressed, such as breastfeeding [111] and smoking [112]. In addition, new strategies associated with health benefits for women during the first 5 years post-GDM that contribute to the existing behavioral goals should be incorporated, such as intuitive eating [113] or postpartum emotional health [102]. Next, we identified other subgroups that need to receive tailored content in the future, such as those with low socioeconomic status [26]. Untapped behavior change methods may be added, such as motivational interviewing. Motivational interviewing was recently implemented in a digital lifestyle modification program [114] and is promising for future mHealth interventions. In general, the craving for more content needs to be addressed (eg, by weekly new content cycles [93]). Lastly, new features that further increase engagement with the app will have to be added and some of the coaching needs to be automated.

The intervention mapping steps 1 to 4 were used to plan, develop, and pretest the digital TRIANGLE intervention to prevent cardiometabolic disease in women during the first 5 years post-GDM. This project description with a detailed intervention model extends previous programs post-GDM and may guide the development of similar complex smartphone solutions to support behavior change. The TRIANGLE intervention uses one of the first interactive smartphone apps to address individual habit changes in several lifestyle areas with diverse behavior change methods and flexible pacing. The TRIANGLE app meets academic and industrial standards as a result of a respective cooperation. A user sample from the target group accepted the novel TRIANGLE intervention, rated the TRIANGLE iOS app to be of high quality, and perceived it to have a large impact on behavior. The TRIANGLE app's core features of “challenge system,” “human coaching,” and “library” seem suitable for large-scale implementation and the overall feedback was positive. However, some additional user needs and subgroups will have to be addressed to further refine the TRIANGLE intervention. The Test TRIANGLE study (German clinical trials register DRKS00012996), a small multicenter randomized controlled clinical trial, explores whether the TRIANGLE intervention is followed and accepted by women during the first years post-GDM and whether it changes their cardiometabolic risk behaviors. After the programming of the app for the Android system, a larger clinical trial needs to assess the program’s efficacy and cost-effectiveness in a representative sample. Intervention mapping steps 5 and 6 may guide program implementation and evaluation in a routine care setting, with increased stakeholder involvement during the respective planning.