We are a German research team with expertise in integrative medicine and cognitive behavioral therapy consisting of two medical doctors, one public health researcher, and one behavioral cognitive psychologist. Moreover, the team is experienced in nondigital as well as digital clinical research using RCTs, mixed-methods approaches, and stakeholder engagement. Since 2012, we have developed and evaluated four app-based mHealth interventions for the pain conditions dysmenorrhea [12-14], and back pain and neck pain [15].

This paper aims to support developers of mHealth interventions in the early stages of their career. Toward this end, we applied a continuous consensus procedure. Content units were identified by discussions in multiple rounds to cluster the selected topics using inductive and deductive coding strategies, as well as mind mapping. Topics were categorized by two reviewers (AR and DP) and any discrepancies were resolved in face-to-face discussions.

The lessons described in this paper are based on our experience conducting the four app-based mHealth RCTs described in the following section.

Research is often conducted from the perspectives of health professionals or researchers, and other perspectives such as those of the patients, customers, and policymakers are neglected. As described in previous research, the sustainability of a developed mHealth intervention often meets obstacles when not in line with the interests of the ministries of health (eg, national strategic goals and laws), especially when it comes to data security; hence, the app developed might not be appropriate for the present infrastructure [16]. Moreover, most mHealth interventions in the form of RCTs are developed as single-stage interventions; however, popular apps usually combine multiple-stage interventions. Hence, the sustainable use of a well-developed app might not only be dependent on its effectiveness but also on its multipurpose applicability to the target group [17]. Accordingly, neglecting the opinions of stakeholders can lead to trials that do not meet the needs of their target group, evaluate irrelevant outcomes, and are therefore not sustainable. Moreover, the user’s app experience is greatly influenced by factors such as design, language, and adaptivity; in turn, the effects of these factors might broadly be influenced by user characteristics such as age, gender, and education. Therefore, we recommend involving stakeholders to understand the target group as early as possible for ensuring a sustainable app and trial.

To show efficacy, effectiveness, and cost-effectiveness [18,19] of an mHealth intervention, a clinical trial is required. An app and trial each come with their own specific characteristics and needs. An app may be determined by the platform it is available for, its design language, its respective features, or its possible dependencies on valid sensor data or a server. These aspects of the app directly affect the complexity (eg, study sample and adherence) and privacy aspects (eg, data sharing) of the corresponding trial, thereby affecting the necessary resources and development time. Moreover, the choice of hardware and software may influence the structure of the trial, since not every feature is available on a given operating system or device, which may significantly shape the scale of the intervention [11]. On the other side, a trial may be determined by the target group, the clearly defined intervention and control settings, and its research outcomes. These aspects of the trial directly affect the choice of app platform (eg, platform preferences of target populations have to be considered), design of the app (eg, use of age-specific graphical elements), included features (eg, questionnaires), and their technical realization (eg, use of the smartphone’s camera). Previous research has shown that approximately half of study results in mHealth interventions are either unclear or negative; however, the number of mHealth interventions and their popularity are consistently increasing [17,20]. This clearly shows a gap necessary to explore in this field of research, and emphasizes the urgent need for well-planned and performed RCTs that are flexible to the field of mHealth interventions.

Therefore, we recommend that an mHealth app always be developed simultaneously with its respective trial to avoid a waste of time and resources.

Currently, mHealth apps need to be developed by multiprofessional teams with expertise in areas such as iOS and Android development, design, backend development, regulatory affairs, psychology, and business development; hence, complexity is already high. Considering the amount of time, expertise, and resources needed for a clinical trial, the appropriate clinical evaluation of mHealth apps substantially adds a magnitude of complexity. Many factors of app and trial development cannot be neglected because of increased user expectations as well as technological and regulatory standards. Moreover, it is difficult to ensure a focused analysis of the interventional effect because mHealth apps are complex interventions including features such as BCTs, combinations of different therapeutic approaches (eg, exercise and diet), and connected devices and services, which are prone to regulation within the health care sector. Knowing that three-quarters of mobile phones are being used in low- or middle-income countries, minimizing complexity might also be beneficial for the sustainability of the mHealth interventions developed, since the latest versions of devices or repair services might not always be within the user’s reach, especially in rural areas [22]. The market of mHealth interventions is often trend-based and dependent on current preferences of the target group. An evidence-based mHealth intervention must therefore not only focus on the quality of the interventions provided but also be able to react timely to changes in the market (eg, software or hardware update) so as to maintain interest of the target group or avoid user frustration. In addition, with a less complex app, the development team might be able to react to bugs or market changes more quickly. To increase the speed of development and the availability of the mHealth intervention for users, we recommend minimizing complexity.

The multidisciplinary team necessary for mHealth app development is usually not available in a research setting. Resources such as designers, frontend or backend developers, and access to technical infrastructure need to be taken into account in addition to the research resources. A lack or insufficient management of these resources leads to a longer time in development, potentially higher financial burden, and less substantial results. We therefore recommend performing prior analyses of existing and necessary resources for app and trial development to adequately manage the mHealth study necessities, and to increase the overall speed of app development to keep up with the rapidly progressing market.

Besides scalability and efficiency, behavior change is a key component of mHealth [21]. Michie et al [23] defined the smallest, observable, replicable intervention component with the potential to bring about change in behavior as a BCT [14,24]. BCTs could also be defined as “a systematic procedure included as an active component of an intervention designed to change behavior” [23]. In digital interventions, app features and functions could be designed based on different BCTs to improve user engagement [25]. For example, “prompts/cues” could be implemented as an app notification to remind users to fill in questionnaires. Feedback on behavior could potentially maintain users’ motivation by providing instant feedback. Further, “goal-setting” and “self-monitoring” are also commonly implemented BCTs in smartphone apps. In a systematic review, 344 BCT apps were reviewed and rated [26]; however, on average, these apps only showed low to moderate functionality, meaning that only a slight amount of BCT was used, and therefore the full spectrum of potential behavior change due to BCTs was not unfolded.

We recommend involving behavior change specialists and to perform early user testing. As a fundament, the BCT taxonomy [21] and the behavior change wheel framework [27] might be helpful.