The mobile nature of modern ICT systems allows users to manage their tasks effectively and obtain appropriate supervision input as and when required. Such interactions can be user initiated, prompted by the system, or may result from supervisor intervention. The developed system contributes to such management and supervision by implementing the requirements described further.

Characteristics of learning identified as important for task-shifting [34], such as being learner-centered, experiential, problem-orientated, and context-appropriate can be appropriately provided by ICT, especially as a refresher after initial, more formal training.

The relationship between CHWs within a team and between CHWs and their supervisors is crucial to a successful task-shifted intervention [35]. ICT can contribute to a good relationship by providing effective and transparent communication between CHWs among themselves, and between CHWs and their supervisors or general practitioners.

Campbell and Scott [8] defined community embeddedness as “...when community members ‘own’ the project by having substantial control over the selection, monitoring, activities and priority-setting of CHWs.” Whereas this definition goes far beyond the physical location of the intervention, an important aspect of community embeddedness is that interventions can be provided where this is deemed most suitable by the community, and often this may be the patient’s own home. Especially for the older adults targeted in this work, mobility may be a limiting factor in the delivery of care through health centers.

One of the main goals of the intervention is to strengthen the autonomy of the patients and highlight the role they can play in their own therapy. The intervention combines psychosocial techniques tailored to individual participants with embedded support mechanisms for nonspecialist health workers delivering the intervention. Behavioral activation was used as the main psychosocial approach in view of its demonstrated feasibility and efficacy for the treatment of depression [36]. It is a simple technique to apply, requires only a short period of professional training [37], and is suitable for delivery by nonspecialists [38,39]. The intervention further incorporates elements of psychoeducation (education about depression and simple coping strategies) and relapse prevention.

The CHW performs all interactions and sessions with the patient using the Android app. Upon logging in with a username password combination, the Android app’s starting screen gives the user an overview of the patients in their care (see Figure 1).

When selecting a user and choosing Start Session, the correct session content will be compiled for the chosen user and an intervention session started. The various screens in each session can be navigated using swipe actions to move to neighboring screens similar to browsing through a book page by page. In addition, chapters in the session can be selected using a horizontal menu bar at the top of the screen, and pages within these chapters can be selected using a vertical menu at the left-hand side of the screen. Wherever the interface presents or requests dynamic patient or user-specific data, a custom-built application programming interface (API) is used to provide a quick and easy means of defining user interfaces with automated storage and retrieval of information. The information thus obtained is stored in database tables.

The first screen of every intervention session (apart from the first session) displays the homework assignments agreed upon in the previous session (Figure 2) and allows the patient and CHW to start the new session discussing and recording progress and exploring potential barriers and solutions for homework assignments.

After reviewing their homework, the patients will respond to a number of questionnaires. On every visit, depression is assessed with the 9-item Patient Health Questionnaire (PHQ-9) [40], and mood is rated on a Likert rating scale (Figure 3). To assist CHWs in this task, a video presented during the first session explains to patients why and how the PHQ-9 is conducted, and the CHWs and patients have access to these videos throughout all future sessions. On the basis of the answers provided in the PHQ-9, the system will, if any of the answers indicate the presence of a depressive symptom, automatically add an extra question to assess how PHQ-9 symptoms have affected the patient’s life. Any positive response to the ninth question of the PHQ-9 on suicidal ideation automatically determines the immediate assessment of suicidal risk. For this, questions frequently used to assess immediate suicide risk (if there was any suicide attempt during the last 14 days and when was this attempt, and if the person has plans and means for another attempt) are shown on the screen. If the answer to any of these questions is positive, the CHW will immediately see instructions in the app on how to proceed in this situation, that is, to stay with the patient until a friend or family member has arrived. Other information, such as medication use (Figure 4), can be updated if required using a questionnaire.

Having gathered and updated the patient’s mental and physical health status, the intervention continues with multimedia resources that explain aspects of the patient’s symptoms, behaviors, and ways to improve these symptoms. In a series of videos specifically designed for this intervention, an actor explains the symptoms of depression, the vicious circle leading to a worsening of the depression and ways to counteract this vicious circle, the importance of doing pleasant activities to improve mood, how avoidant behaviors are linked to depression, how to plan pleasant (healthy) activities, and how to prevent relapse. As CHWs have limited training in the symptoms of depression and the principles underlying the intervention, these videos support the CHWs in providing patients with the required information on their depressive symptoms and the purpose of the intervention. One of the main goals of the intervention is to strengthen the autonomy of the patients and highlight the role they have in their own improvement. For this reason, in all sessions the patients do activities on the app, supported by the CHW, and plan activities to do between sessions. This strategy helps patients to develop the autonomy to identify and deal with symptoms of depression at present and in the future.

Sessions conclude with the scheduling of a new appointment. This scheduler recommends the date for the next appointment based on the patient’s current progress in the therapy and structures the process of making an appointment for a new visit to complete the current session or to move on to the next session in the therapy. As internet connectivity is not always available at the patients’ homes, the app contains all information for the patients under the care of the CHW who uses the device.

As mentioned previously, an important goal of the system is to empower CHWs with limited training in mental health service provision to provide a psychosocial intervention. To this end, the system provides automated support to the CHWs. This support consists of session content recommendations, as well as support in scheduling intervention visits and further recommendations personalized to that patient, such as warnings regarding adherence to the intervention, suicidal ideation, and progress in the intervention. The CHW has access to these recommendations and warnings in a dedicated screen in the app. Where necessary, the system forwards recommendations and warnings to supervisors by email.

To further support the supervision of CHWs, he or she can, in consultation with the patient, choose to make audio recordings of (parts of) the session for use in future intervention or supervision sessions. These audio recordings are stored on the device and are only accessible to the CHW.

The Web interfaces allow individuals with supervisory roles access to the system. The information shown to these individuals varies slightly based on the objective to show only the most pertinent information. Those that are responsible for the running of the intervention at a high level (Trial Managers or their equivalent in the health system) see aggregate data on a per-patient basis, whereas clinical supervisors see more detailed information for each patient related to the clinical care in which one of their CHWs is involved.

In the usability assessment, some of the users indicated that they found the system inconsistent and expected to require technical assistance during normal use of the system. Whereas these results may partly stem from bugs encountered and resolved during the pilot, they do indicate that further guidance is required. Solutions may be to hide part of the interface, such as the list of pages in a chapter displayed on the left-hand side of the screen by default and more extensive logging of use of the app, such that issues can be identified automatically. For example, longer than expected or usual activity on a particular screen of the app may indicate that users encountered issues with the content presented on the screen and such findings can be discussed in supervision meetings.

Decision support was visible to end users in the form of recommendations on the course of action after 3 initial sessions resulting in the patient being allocated to a high-intensity (8 more sessions) or low-intensity (5 more sessions) intervention, questionnaires on suicidality, and advice to discuss patients that did not show progress (based on trends in PHQ-9 responses) with supervisors. The algorithm for allocation to the low-intensity or high-intensity treatment was not disclosed to the health workers, so they could not influence the allocation of patients to the second phase of treatment.

In a full-scale trial or clinical roll-out of this intervention, CHWs will be supervised by the clinical supervisors who normally have supervisory responsibilities for these CHWs. During the pilot, supervision was performed by research psychologists who conducted the pilot study. Data captured by the system were used to discuss the progress of patients and to make suggestions for future sessions.

The functionality to make audio recordings of sessions was added to the app during the pilot upon request. For simplicity and ethical considerations, it was decided not to send the resulting audio files to the server. Hence, reviews can only be done face-to-face with the CHW, who is the only person with access to the files created on the device. It would be worth considering whether the benefits that could be derived from making the audio files available on the server would weigh up against potential issues around privacy. It may also be possible to implement a consent mechanism for audio uploads to the server on a per-user or per-instance basis. A further consideration is that the storage of audio data would require more server storage capacity, but this should not be a limiting factor with modern servers.

Accountability functionality was enabled by the patient data gathered during intervention sessions, but also by logs relating to session timing and appointments. The app logged when a session was started and finished and kept track of appointments made with the patient. It also allowed the CHW to make note of missed appointments and the reasons. These data proved invaluable during the pilot in several ways. For example, the patient responses gathered during sessions allowed automated content selection during interventions and the generation of patient-specific prompts to CHWs, thus providing robust support to the latter in following protocols and procedures. The data on scheduled, completed, and missed sessions provided valuable information on patient progress and potential barriers to their involvement in the intervention. For (trial) supervisors, the logged data were presented in such a way that these individuals obtained a high-level overview of the intervention delivered by CHWs and progress of their patients, thus allowing timely corrective action when required.

During the pilot, the geolocation of a session was not recorded and CHWs were also not prevented from accessing a session for a patient when there was no future appointment for that patient. Although this may be seen as overly rigid, there may be merit in locking the app for a user until shortly before the next scheduled appointment. This would prevent inadvertent data logging for a user (for example, because of briefly accessing the next session in the app when merely intending to review the session the next patient will receive) and would also encourage CHWs to maintain complete diaries of patient visits.

Treatment recommendations and warnings regarding a lack of progression or suicidality were displayed both in the app and in the supervisors’ Web interface. For full accountability, a protocol should be implemented that requires CHWs and their supervisors to indicate that the recommendations and warnings have been reviewed and followed up.

The patient data logged in the sessions were used to personalize subsequent sessions, thus facilitating CHWs to provide a contextualized intervention. This, in turn, enabled sessions to be conducted efficiently (for example, the medical questionnaire can be reviewed and updated very quickly) and allowed an objective review of progress (through reviewing homework agreed on in a previous session and by allowing read-only access to tabular and graphical representations of PHQ-9 response, mood ratings, and homework completion on a session basis). In addition, aggregate information derived from the data gathered during sessions provided supervisors a good insight in patient progress and allowed for timely intervention where required.

Although, where possible, data input was structured and made quicker using drop-down lists and other multiple-choice widgets, some elements of the intervention required considerable free-text input. We provided keyboards for this purpose and though this may suit some users, others may find such input difficult on a tablet device.

Although generally positive, the responses to items 3 and 7 on the SUS scale regarding EoU and learning rate, respectively, indicate that users were not always confident in using the system. Requests for an interface with dummy patients for each of the sessions (which was implemented during the pilot) support our belief that these shortcomings can be remedied with improved in-app training facilities.

During the pilot, automated communication was limited to patient-specific warnings being shown in the Web interfaces for supervisors and researchers. Such automated communications should be extended to automated emails to supervisors and, where necessary, to routine care health providers involved in the care of patients that make use of the system to fully comply with accountability requirements.

Due to time constraints, user-initiated communications were not implemented during the pilot study, but CHWs were able to contact supervisors using the device’s normal phone app and Google Hangouts. A WhatsApp group was used frequently for communication among CHWs and researchers involved in the trial. The limited implementation of this user-initiated communication in the pilot study version of the system does not seem to have had a negative effect on the good relationship between health care workers in a team, but this is likely because of the fact that these health care teams were already well established.

Mobile delivery of the intervention in patients’ homes was facilitated using the mobile intervention interface on Android tablets that contained data for all the patients under the care of a given CHW using the device. Most tablets were shared by, normally, 2 CHWs, and the protocols required users to synchronize their devices before leaving the community health center. In a few instances, this sharing resulted in CHWs using a different tablet that did not contain the data for their patients. A login in the app with subsequent synchronization would have resulted in all data being downloaded, but this was not always performed, on occasion resulting in CHWs not being able to perform the session as planned. Although the tablets were equipped with subscriber identification module cards, a connection to the system’s servers may not always be available at a patient’s home. For this reason, it is desirable to implement a procedure within the app that unobtrusively ensures the data for the CHW using the tablet are indeed available on the device.

Task-shifting has been the preferred strategy to address the lack of specialized resources in LMICs. However, this problem is of such magnitude that even transferring responsibilities to low-cadre health workers is not enough to overcome the problem. More innovative solutions are needed. One such solution is for technology to provide assistance to these low-cadre health workers. This is already happening in many parts of the world, and the uptake is increasing rapidly [48]. Our pilot suggests that technology indeed may have a positive effect on the ability of low-cadre health workers to effectively deliver a psychosocial mental health intervention. Although our results lack statistical power, they suggest the intervention improved symptoms of depression and that the support provided to the CHWs by the system allowed them to deliver the intervention with very little training and support. Moreover, the pilot addresses the needs of a significantly underserved population, which does not have access to routine mental health services delivered by highly skilled practitioners. The combination of task-shifting and technology is a key enabler for provision of mental health services to this underserved population.