**4. Discussion**

This pilot study evaluated the feasibility and acceptability of conventional F2F training compared with digital training programs to build the capacity of non-specialist health workers for delivering HAP, a brief evidence-based psychological treatment for depression. The primary goals of this study were to test the study procedures and, importantly, to inform modifications and refinements to the training programs in preparation for a large-scale fully powered effectiveness trial. While we previously demonstrated the interest in using digital technology for accessing training programs among non-specialist health workers, the current study substantially expanded on our prior work by testing these programs in the field, allowing the opportunity to gain insights about the use of the training programs in real world settings.

This study highlighted the need for several significant modifications to the digital training program. These included: the need to modify the timing and structure of the F2F training to accommodate participants' long commutes from distant villages, as well as to account for their family responsibilities such as childcare; the need to ensure that the digital program content could be accessed entirely offline given the low internet connectivity in rural areas in the Sehore district; the need for a more comprehensive orientation session for using the smartphone app to access the training program and navigating the Learning Management System, including use of more pictures and screenshots of 'how to' examples to account for low digital literacy among participants (this was also reflected by the large number of technical assistance calls received during this pilot study); and modifications to the provision of technical support, to allow early identification of participants who may be struggling to complete the digital training and enable a more timely response to technical challenges that could arise. The challenge of poor internet connectivity was similarly reported in a prior trial from Pakistan, where efforts to

address this concern also involved ensuring access to the training content in an offline format [30]. This prior study reported that the online training approach required a stable internet connection that may not be available in many remote, rural, resource-poor settings; hence, to increase the feasibility of the online training program, the researchers used an offline tablet-based application to deliver the training to frontline health workers [30]. Following the focus group discussions, we made substantial modifications to the remote support component of the DGT+ arm, given that participants expressed high interest in having a member of our research team contact them to provide encouragement and motivation on a regular basis.

Our study aligns with an emphasis in the digital mental health research literature that it is necessary to consider the perspectives of users in order to support adoption, engagement, and sustained use of digital interventions [65–67]. Despite the large number of technical challenges that participants mentioned throughout this pilot study, we were reassured by participants' continued interest in learning about the mental health treatment related content. This was consistent between participants in the digital training programs and F2F training, suggesting recognition among non-specialist health workers of the importance of depression care in their communities. This is an essential first step towards successfully scaling up mental health services in primary care settings in the Sehore district, as well as across the state and nation.

Another important finding in this study was about which cadre of non-specialist health workers would be most suitable for completing the training program based on their availability. We learned that the MPWs had too many other competing demands, and were frequently called away by their superiors for attending to urgent duties, which is reflected in their low completion rates across the three training programs. However, for non-specialist health workers who are frequently required to travel for other work related activities, use of a digital training program may offer the opportunity for these individuals to gain the necessary skills to deliver mental health services while accommodating their already busy workload. In addition, digital training holds the potential to train all types of health workers on depression care, as the program can be accessed on a smartphone, which could potentially expand access to mental health services at the community level, thereby advancing efforts to achieve the Mental Health Care Policy goal of providing universal mental health care services for all [13].

While this was a pilot study primarily focused on assessing the feasibility and acceptability of the different training programs, we found that scores on the competency assessment improved for the F2F and DGT+ participants. This is a promising finding, suggesting that digital training with added support may be equally effective compared to the regular classroom or in-person training in terms of gaining knowledge and skills. Also, digital training is potentially more convenient, feasible, and scalable for building the capacity of non-specialist health workers when compared to conventional in-person training, which is supported by similar studies and recent reviews from other low-income and middle-income countries [30,68,69]. Additionally, the findings also indicate that the training content is appropriate for gaining the knowledge and skills related to HAP delivery, as reflected by improved scores on the competency measure. However, the DGT participants did not show significant improvements, suggesting that the use of digital technology alone may not be sufficient for contributing to knowledge acquisition. Importantly, the addition of support initiated by our research team appeared to greatly improve program completion for DGT+ participants (79%) compared to DGT participants (57%). This is consistent with prior studies of online education programs that have demonstrated that the use of digital training is most effective when supplemented with access to remote or in-person support [30,36]. For example, the recent study of the Technology-Assisted Cascaded Training and Supervision system for Lady Health Workers, conducted in rural areas in Pakistan, found that use of digital technology in combination with in-person support and training contributed to comparable improvements in competence as conventional face-to-face training [30]. Additionally, a study in Zambia using technology to train community health workers highlighted a similar finding that in-person support is required to address technical challenges related to poor network coverage, mobile hardware, and software [70]. If these challenges are not addressed, it can negatively impact

the delivery process and training outcomes [70]. Further, a recent review of mobile technologies for education and the training of community health workers in low-income and middle-income countries indicated the value of digital training methods for augmenting periodic in-person training activities, while highlighting that digital training programs could be embedded within existing health care services to allow opportunities for continuing education among community health workers [69].

Several limitations with this study warrant consideration. Firstly, this was a pilot study looking at acceptability and feasibility of the training program, and therefore the sample size was small and not adequately powered to detect differences in competency outcomes between groups. Additionally, participants' satisfaction and acceptability ratings were generally very positive, suggesting a potential desirability bias. On the other hand, appropriateness was ranked lower, suggesting the need for improvements to the training programs to promote engagement and sustained interest, which was further reflected during the focus group discussions. The self-report measure used to assess competency outcomes was translated into Hindi and adapted to the local context, though the psychometric properties of this measure have not yet been established for use in rural India. It will be important that further efforts seek to validate this self-report competency measure to support its widespread use in diverse contexts in India. While more scalable and efficient to administer, the use of a self-report measure for competency presents other disadvantages compared to conventional competency assessment methods such as role plays or direct observation because it may not capture the application of skills during direct interactions between the health worker and patient. Furthermore, we made conscious efforts to limit potential bias during the quantitative and qualitative data collection. For instance, the quantitative surveys about satisfaction and acceptability with the training programs may have been subject to social desirability bias, where participants may have reported highly positive responses. To minimize this potential risk, members of our research team overseeing data collection were not involved in the intervention development, and they also reassured participants that there are no right or wrong answers to the questions about program satisfaction because honest feedback is most important for finding ways to improve the training program and content for the future. To minimize a similar risk of social desirability bias in the qualitative data collection, we ensured that the facilitators of the focus group discussions and note taking were also not members of our team involved in the training program development, as they may have influenced participants' responses. Members of our research team who were not involved in the development of the training program conducted the focus group discussions and collected field notes. Given that our study was primarily aimed at informing a subsequent large-scale trial, we did not conduct an in-depth thematic analysis of our qualitative findings. Therefore, in future research developing digital applications, we can build on our approach presented here to strengthen the qualitative methods for analysis and interpretation of participants' feedback and recommendations about program design.

We made an effort to recruit only participants who had not previously participated in our formative research as a method to minimize prior exposure. However, there is still a possible risk of contamination [71], which we did not assess, though we believe that this risk was low. Furthermore, the non-specialist health workers were recruited from real world settings; therefore, it is not possible to fully minimize contamination in such settings, as health workers may look up information about the training materials on the Internet or may talk to each other about the program content in routine encounters in the workplace. Participants were recruited from three community health centers in a single district in Madhya Pradesh, indicating that these findings may not generalize to other settings in India in terms of context and culture, or other settings globally. However, many of the findings reported here relate to the use of digital training in a low-resource setting and overcoming challenges such as low digital literacy and poor bandwidth likely apply to many other settings. Our finding that some cadres of health workers, such as the MPWs, were not able to complete the training due to their prior engagement with work commitments highlights potential challenges for scaling up this type of training program due to competing priorities. Therefore, our findings may only generalize to health workers who have the time available and who are interested and willing to learn about treatment for depression. To achieve the goal of universal access to mental health services, it will be necessary to consider what cadres of health workers are available and ideally positioned to successfully complete the training program and provide care for depression and other mental disorders as part of their routine service delivery. Furthermore, we restricted our sample to non-specialist health workers with minimum 8th standard education to ensure that they were literate and able to follow the written training materials, to access and navigate the training program on the smartphone app, and to answer the questions on the competency assessment measure. This type of training would likely not be suitable for health workers who may be illiterate, or who may not be able to operate a smartphone. Even though we found that roughly half of the sample had ever used a smartphone, all participants randomized to either of the digital training programs were able to learn to use the smartphone and access the training program content. This further attests to the interest among non-specialist health workers to use digital technology to support their work, which has been consistently reported in prior studies [68,69].
