Health workers’ perceptions and experiences of using mHealth technologies to deliver primary healthcare services: a qualitative evidence synthesis

Abstract Background Mobile health (mHealth), refers to healthcare practices supported by mobile devices, such as mobile phones and tablets. Within primary care, health workers often use mobile devices to register clients, track their health, and make decisions about care, as well as to communicate with clients and other health workers. An understanding of how health workers relate to, and experience mHealth, can help in its implementation. Objectives To synthesise qualitative research evidence on health workers' perceptions and experiences of using mHealth technologies to deliver primary healthcare services, and to develop hypotheses about why some technologies are more effective than others. Search methods We searched MEDLINE, Embase, CINAHL, Science Citation Index and Social Sciences Citation Index in January 2018. We searched Global Health in December 2015. We screened the reference lists of included studies and key references and searched seven sources for grey literature (16 February to 5 March 2018). We re‐ran the search strategies in February 2020. We screened these records and any studies that we identified as potentially relevant are awaiting classification. Selection criteria We included studies that used qualitative data collection and analysis methods. We included studies of mHealth programmes that were part of primary healthcare services. These services could be implemented in public or private primary healthcare facilities, community and workplace, or the homes of clients. We included all categories of health workers, as well as those persons who supported the delivery and management of the mHealth programmes. We excluded participants identified as technical staff who developed and maintained the mHealth technology, without otherwise being involved in the programme delivery. We included studies conducted in any country. Data collection and analysis We assessed abstracts, titles and full‐text papers according to the inclusion criteria. We found 53 studies that met the inclusion criteria and sampled 43 of these for our analysis. For the 43 sampled studies, we extracted information, such as country, health worker category, and the mHealth technology. We used a thematic analysis process. We used GRADE‐CERQual to assess our confidence in the findings. Main results Most of the 43 included sample studies were from low‐ or middle‐income countries. In many of the studies, the mobile devices had decision support software loaded onto them, which showed the steps the health workers had to follow when they provided health care. Other uses included in‐person and/or text message communication, and recording clients' health information. Almost half of the studies looked at health workers' use of mobile devices for mother, child, and newborn health. We have moderate or high confidence in the following findings. mHealth changed how health workers worked with each other: health workers appreciated being more connected to colleagues, and thought that this improved co‐ordination and quality of care. However, some described problems when senior colleagues did not respond or responded in anger. Some preferred face‐to‐face connection with colleagues. Some believed that mHealth improved their reporting, while others compared it to "big brother watching". mHealth changed how health workers delivered care: health workers appreciated how mHealth let them take on new tasks, work flexibly, and reach clients in difficult‐to‐reach areas. They appreciated mHealth when it improved feedback, speed and workflow, but not when it was slow or time consuming. Some health workers found decision support software useful; others thought it threatened their clinical skills. Most health workers saw mHealth as better than paper, but some preferred paper. Some health workers saw mHealth as creating more work. mHealth led to new forms of engagement and relationships with clients and communities: health workers felt that communicating with clients by mobile phone improved care and their relationships with clients, but felt that some clients needed face‐to‐face contact. Health workers were aware of the importance of protecting confidential client information when using mobile devices. Some health workers did not mind being contacted by clients outside working hours, while others wanted boundaries. Health workers described how some community members trusted health workers that used mHealth while others were sceptical. Health workers pointed to problems when clients needed to own their own phones. Health workers' use and perceptions of mHealth could be influenced by factors tied to costs, the health worker, the technology, the health system and society, poor network access, and poor access to electricity: some health workers did not mind covering extra costs. Others complained that phone credit was not delivered on time. Health workers who were accustomed to using mobile phones were sometimes more positive towards mHealth. Others with less experience, were sometimes embarrassed about making mistakes in front of clients or worried about job security. Health workers wanted training, technical support, user‐friendly devices, and systems that were integrated into existing electronic health systems. The main challenges health workers experienced were poor network connections, access to electricity, and the cost of recharging phones. Other problems included damaged phones. Factors outside the health system also influenced how health workers experienced mHealth, including language, gender, and poverty issues. Health workers felt that their commitment to clients helped them cope with these challenges. Authors' conclusions Our findings propose a nuanced view about mHealth programmes. The complexities of healthcare delivery and human interactions defy simplistic conclusions on how health workers will perceive and experience their use of mHealth. Perceptions reflect the interplay between the technology, contexts, and human attributes. Detailed descriptions of the programme, implementation processes and contexts, alongside effectiveness studies, will help to unravel this interplay to formulate hypotheses regarding the effectiveness of mHealth.

1. Through being connected to other health workers and across various healthcare services, health workers appreciated that mobile devices allowed them to better co-ordinate the delivery of care. Moderate confidence Due to no/very minor concerns regarding coherence, minor concerns regarding adequacy and methodological limitations, and moderate concerns regarding relevance 2. Lower-level health workers valued being able to reach higher-level health workers via mobile devices, and perceived the advice and support they received as improving their care and as satisfying to clients. When higher-level professionals responded in anger, it made lower-level health workers reluctant to call them. Moderate confidence Due to no/very minor concerns regarding coherence, relevance, and adequacy, but moderate concerns regarding methodological limitations 3. When higher-level health workers failed to respond and support lower-level workers through mobile devices, lower-level sta had negative perceptions of these devices. One study emphasised the importance of having health professionals' buy-in with mobile health to ensure that mobile devices were optimally used to support lay health workers. Moderate confidence Due to no/very minor concerns regarding coherence and relevance, minor concerns regarding methodological limitations, but moderate concerns regarding adequacy 4. The use of mobile devices allowed some health workers to feel connected to their peers within their own organisations. However, others preferred face-to-face communication with their peers. Moderate confidence Due to no/very minor concerns regarding coherence, minor concerns regarding methodological limitations and relevance, and moderate concerns regarding adequacy 5. Some health workers relayed that mobile devices improved their reporting to supervisors and encouraged them to report more truthfully. Others compared mobile device-facilitated supervision to "big brother watching". Some supervisors thought that mobile devices allowed them to better identify sta who needed support. methodological limitations 6. Health workers had positive experiences with using instant messaging through WhatsApp. This application was seen as cheap and suitable for a range of activities, such as communicating with peers and posting photos as evidence of work done.
Hampshire 2016; Henry 2016; Schoen 2017 Very low confidence Due to serious concerns regarding methodological limitations and adequacy, moderate concerns regarding relevance, and no/ very minor concerns regarding coherence 7. Even when health workers received messages that were automated, rather than sent directly from a manager or supervisor, this was still experienced and responded to, as a kind of supervision. Some lower-level health workers experienced it as supportive to their work, while others felt guilty for not providing correct care as per these messages. Moderate confidence Due to no/very minor concerns regarding coherence, relevance, and adequacy, and moderate concerns regarding methodological limitations 9. At times, health workers used their mobile devices to access the Internet for health information, and found it useful when they were with clients who needed the information. This interaction also included health workers providing clients with additional information beyond the healthcare intervention. But, if the only way that health workers could access online information, required them to use their own money to purchase data, then this could be prohibitive to them accessing such information. Low confidence Due to no/very minor concerns regarding coherence, minor concerns regarding methodological limitations and relevance, and serious concerns regarding adequacy 10. mHealth held the promise of increasing service efficiency for many health workers, but the experience of whether this promise was borne out in practice, varied in the accounts of health workers. It was experienced as efficient if it improved feedback, speed and workflow, but inefficient when the technology was slow and time consuming. Some were concerned that if mHealth was too efficient, making work faster, that this may justify sta cutbacks. High confidence Due to no/very minor concerns regarding relevance and adequacy, and minor concerns regarding methodological limitations and coherence 11. Health workers frequently reported mobile devices as overcoming the difficulties of rural and geographically challenging contexts when it made it possible for them to provide health care without having to travel. Some report-ed that reducing travel time allowed them more time with their clients. Moderate confidence Due to no/very minor concerns regarding coherence, relevance, and adequacy, but moderate concerns regarding methodological limitations 13. Through mHealth, health workers were able to use treatment and screening algorithms that were loaded onto mobile devices. Their perceptions of using these electronic algorithms ranged from finding it easy and useful, to threatening their clinical competency, and an information overload. There were also some concerns that erroneous data entry may lead to wrong treatment guidance. High confidence Due to no/very minor concerns regarding coherence, relevance, and adequacy, and minor concerns regarding methodological limitations 14. Using mobile devices to record routine client or surveillance data was mostly perceived by health workers and their managers as helpful for decision making, and increasing community and health worker appreciation of these data. Moderate confidence Due to no/very minor concerns regarding coherence, relevance, and adequacy, but moderate concerns regarding methodological limitations 15. In most cases health workers perceived mobile health as more advantageous than paper. However, some continued to prefer paper. High confidence Due to no/very minor concerns regarding coherence, relevance, and adequacy, and minor concerns regarding methodological limitations 16. mHealth interventions sometimes required health workers to perform tasks that were peripheral to regular service delivery, such as registering clients onto the system. These more menial tasks were sometimes regarded as undermining to professional sta . Very low confidence Due to serious concerns regarding methodological limitations and adequacy, and moderate concerns regarding coherence and relevance 17. Some health workers experienced the use of mHealth as generating an extra workload when, for instance, it resulted in reaching more clients needing care, or having to maintain both a mobile health and paper system. Some workers disliked this, particularly when their superiors did not perceive their mobile health work as part of their job description. Others did not object to the additional work, yet others wanted to be remunerated. High confidence Due to no/very minor concerns regarding coherence, relevance and adequacy, and minor concerns regarding methodological limitations mHealth led to new forms of engagement and relationships with clients and communities 18. Through mobile devices, health workers and clients could communicate directly with each other, which health workers reported as improving care and their relationship with clients. When clients initiated the contact, health ological limitations, and moderate concerns regarding adequacy 19. Health workers were aware of the importance of protecting confidential client information when using mobile devices, and the confidentiality risks in cases of stolen phones and using their SIM cards in colleagues' phones. Health workers were alert to clients' concerns when they shared personal information concerning stigmatised issues, such as HIV/AIDS and interpersonal violence, and suggested ways to keep the information confidential. They emphasised building a trusting relationship with clients prior to using the devices. High confidence Due to no/minor concerns regarding methodological limitations, coherence, relevance, and adequacy 20. Health workers were concerned that concentrating too much on the mobile technology during client consultations could be to the detriment of their service and interaction with clients. Low confidence Due to serious concerns about adequacy, moderate concerns regarding relevance, minor concerns regarding methodological limitations, and no/very minor concerns regarding coherence 21. Health workers had differing reactions to being contactable via mobile devices during and outside of working hours: some felt it was useful, some were ambivalent about it, and others objected to it. Workers suggested setting boundaries to protect themselves from this. Moderate confidence Due to no/very minor concerns regarding methodological limitations and coherence, minor concerns regarding relevance, and moderate concerns regarding adequacy 22. Health workers experienced the use of mobile technology to provide health care, as being met with both trust and skepticism from clients and the communities they served. They described how trust or skepticism in the device was translated into trust or skepticism of their service when using the device. Some found that using mobile devices raised their social status with clients, and even their families. Others were concerned that using expensive equipment would emphasise inequity between themselves and clients. High confidence Due to no/very minor concerns regarding coherence, relevance, and adequacy, and minor concerns regarding methodological limitations 23. Health workers experienced clients as having an opinion not only about their use of mobile devices, but as having an opinion on the devices themselves, which influenced how they responded to care delivered with the support of these devices. Health workers ascribed clients' enthusiasm for mobile devices as due to these clients' perception of the devices as prestigious, offering trustworthy information, and providing confidentiality. They perceived clients as more receptive when these clients were familiar with the devices used. There were concerns that clients who felt that the use of these devices during care was too time consuming, and would respond negatively to its use. Moderate confidence Due to no/very minor concerns regarding coherence, relevance and adequacy, but moderate concerns regarding methodological limitations 24. Some interventions required clients to have phones as well as health workers. Health workers described this as challenging for multiple reasons, including clients not Moderate confidence Due to no/very minor concerns regarding coherence, minor concerns re-having phones, changing their phone numbers regularly, not knowing how to use a phone, being a target of crime because of possession of the phone, and women being prohibited from accessing phones. Health workers suggested competitive pricing to increase clients' access to phones, and to issue clients with phones. Tewari 2017; van der Wal 2016; Wol -Piggott 2018 garding methodological limitations and adequacy, and moderate concerns regarding relevance 25. Health workers were ambivalent about interventions that required clients to use the health workers' mobile devices during consultations. Their optimism was tempered by concern that there was a loss of meaningful engagement with clients.

Bacchus 2016; Coetzee 2017
Low confidence Due to serious concerns regarding adequacy, moderate concerns regarding relevance, and no/very minor concerns regarding methodological limitations and coherence 26. Health workers reported that their access to mobile devices was beneficial to clients and communities who were too poor to own mobile phones.

Chang 2011; van der Wal 2016
Very low confidence Due to serious concerns regarding relevance and adequacy, moderate concerns regarding methodological limitations, and no/very minor concerns regarding coherence 27. Health workers felt that health promotion and educational messaging directed at clients using mobile health interventions, impacted positively on clients' health behaviours, but cautioned against repetitive showing of health promotion videos. In one instance, issuing clients with mobile phones led to increased use of healthcare services. Moderate confidence Due to no/very minor concerns regarding methodological limitations, coherence, and relevance, but moderate concerns regarding adequacy Health workers' use and perceptions of mHealth could be influenced by factors tied to costs, the health worker, the technology, the health system and society, poor network access, and poor access to electricity 28. Some health workers accepted bearing the costs of mHealth interventions themselves, but were dissatisfied when phone credit to use the phones was not delivered on time. Health workers felt that clients appreciated it when health workers called them, as it saved them costs. High confidence Due to no/very minor concerns regarding coherence, relevance and adequacy, and minor concerns regarding methodological limitations 29. Health workers' digital literacy impacted on their experience and perceptions of the use of mobile devices in health service delivery: being digitally literate resulted in positive experiences and perceptions, whilst low digital literacy caused concerns about job security and embarrassment when making mistakes in front of clients. For some workers, prior exposure to mobile devices did not affect their perceptions and use of mobile health. Some turned their lack of digital literacy into building a relationship with clients by asking clients to show them how to use the devices. Not using the devices often enough, resulted in loss in digital literacy. Moderate confidence Due to no/very minor concerns regarding coherence, relevance, and adequacy, but moderate concerns regarding methodological limitations 30. Health workers expressed a need for training and familiarity with mobile devices to overcome their initial anxiety in using the devices. Peer training from technologically proficient colleagues was experienced as valuable. In several cases, health workers wanted refresher training and pointed to the importance of training replacement sta . Not having mentors who used mobile devices, impacted negatively on lower-level workers' ability to learn how to use these devices. High confidence Due to no/very minor concerns regarding coherence, relevance, and adequacy, and minor concerns regarding methodological limitations 32. Health workers highlighted that mobile technology applications should be user-friendly, easy to learn, and improve the quality of their care. When the applications were not easy to use, health workers became frustrated and reluctant users of mobile devices. High confidence Due to no/very minor concerns regarding coherence, relevance, and adequacy, and minor concerns regarding methodological limitations 33. Health workers held mixed views on choosing between tablets and smartphones. Some felt that the type of content on the device was more important than the device itself. However, other health workers preferred tablets over smartphones, mainly because the bigger size of the screen was perceived as easier for client engagement. Due to no/very minor concerns regarding methodological limitations and coherence, but moderate 37. Health workers offered programmatic and implementation recommendations to improve mobile health interventions. The most cited of these was that the interventions be expanded to other settings and services, beyond what they were using it for as described in the studies. Other recommendations included raising community awareness about mHealth programmes, being involved in developing programmes, and appointing a 'mobile health champion'. Workers also suggested that those collecting surveillance data, must be informed of how the data are used. Due to no/very minor concerns regarding coherence, minor concerns regarding adequacy, and methodological limitations, and moderate concerns regarding relevance 39. The main challenges health workers experienced in using mobile devices, were poor network connectivity, access to electricity, and the costs to recharge devices. Solutions offered, included using solar panels, using the powered-up phone of a colleague, or reverting back to the paper-based system. Sometimes poor connectivity resulted in client dissatisfaction because it created delays in receiving health care. Health workers' commitment to their clients motivated them to cope with these and other challenges. Health workers discussed challenges, beyond network and electricity issues, that sometimes were just an annoyance or a concern, but at other times also impeded their mHealth activities, and their ability to provide a service assisted by the use of mobile devices. These included damaged devices, loss and the of devices, having to carry two devices, not being able to readily buy phone credit when needed, not being able to send long messages because of character limitations, and the limitations of the language capabilities of their devices. The growing interest in mHealth as a research topic is reflected in the 25 e ectiveness reviews published in the Cochrane Library (Appendix 1). Two overviews of reviews have also identified 29 systematic reviews (Marcolino 2018; Mbuagbaw 2015), of which 17 were non-Cochrane Reviews. These Cochrane and non-Cochrane reviews cover mobile health technologies that vary in their type and purpose, from the use of email for clinical communication between healthcare professionals (Pappas 2012), to the use of mobile phones for healthcare appointment reminders (Gurol-Urganci 2013). The evidence on the e ectiveness of mHealth cited in these reviews also varies. The overview of reviews from Marcolino 2018 shows mixed results and a lack of long-term studies, although some evidence suggests an e ect on some health outcomes. The growing importance and interest in mHealth is also reflected in the launch of two new journals, one of which is within the Lancet group of journals, namely The Lancet Digital Health (www.thelancet.com/ journals), and mHealth (mhealth.amegroups.com).

Why is it important to do this review?
The release in April 2019, of the World Health Organization (WHO) guideline on digital interventions for health system strengthening (WHO 2019), attests to recognition at the highest level of global health, that mHealth is now a significant component in the delivery and support of healthcare policy, guideline and decision-making processes. Processes, such as the development of this guideline, should be supported by "… social scientific studies explicating processes of technology adoption …" (Chib 2015). Identifying, appraising and synthesising the qualitative evidence of health workers' perceptions and experiences of mHealth programmes, complement the reviews of mHealth e ectiveness and help improve our understanding of the barriers to, and facilitators of, its successful implementation (Chang 2013;Grimsbø 2012;Medhanyie 2015), as well as helping us to understand the outcomes, implementation, and feasibility of these programmes. This is particularly important as decision makers move from assessing the options to implementing the intervention, and thus need to consider more than whether an intervention works or not, but also the extent to which it may be acceptable in di erent contexts (Langlois 2018). This review is one of two qualitative evidence syntheses, that have been used alongside a suite of reviews of e ectiveness, to inform the recently published WHO guidelines (WHO 2019); the other Cochrane Review focuses on clients' and peoples' perceptions and experiences of targeted digital communication, accessible via mobile devices for reproductive, maternal, newborn, child and adolescent health (Ames 2019).

How this review might inform or supplement what is already known in this area
The Cochrane and non-Cochrane e ectiveness reviews (Agarwal 2018a; Agarwal 2018b; Agarwal 2018c; Braun 2013; Gonçalves-Bradley 2018a; Gonçalves-Bradley 2018b; Vasudevan 2018; Vervloet 2012), showed mixed or inconclusive results. In order to understand this heterogeneity, we need to go beyond the numbers and explore the context in which the interventions are delivered, and the experiences of the people involved in the delivery (Langlois 2018). This may lead to a better understanding of possible reasons why mHealth interventions have worked di erently in di erent contexts. It is therefore, important to supplement the evidence of e ectiveness by exploring the barriers and facilitators to the successful implementation of mHealth interventions, through qualitative studies that take contextualised experience into account (Glenton 2013). This would support the call by some of the e ectiveness reviews that "… clients' and healthcare providers' evaluation and perceptions of the safety of the interventions, potential harms, and adverse e ects … should be assessed" (Gurol-Urganci 2013), and "… barriers to trial development and implementation should also be tackled [in future studies]" (Atherton 2012). This qualitative evidence synthesis intends to be both complementary to the e ectiveness reviews, as well as providing robust evidence in its own right.

O B J E C T I V E S
To synthesise qualitative research evidence on health workers' perceptions and experiences of using mHealth technologies to deliver primary healthcare services, and to develop hypotheses about why some technologies are more e ective than others.

M E T H O D S
Criteria for considering studies for this review

Types of studies
We included primary studies that used qualitative methods for data collection (e.g. interviews, focus group discussions, document analysis, and observations), and qualitative methods for data analysis (for instance, thematic analysis, and grounded theory).
We excluded primary studies that collected data using qualitative methods but did not perform a qualitative analysis (e.g. openended survey questions where the responses were analysed using descriptive statistics). We included mixed-methods studies when it was possible to extract data that resulted from the qualitative methods. We included studies regardless of whether they had or had not been carried out alongside studies of e ectiveness of mHealth.

Study participants
We included studies that focus on the perceptions and experiences of the following.
1. All categories of health workers (i.e. professionals, paraprofessionals and lay health workers) who were involved in providing primary healthcare services to clients. We defined a paraprofessional health worker as someone with some form of secondary education and subsequent informal and/or formal training, lasting a few months to more than a year (Olaniran 2017). We defined a lay health worker as any health worker who performs functions related to healthcare delivery, is trained in some way to provide these functions, but has received no formal professional or paraprofessional certificate or tertiary education degree (Lewin 2005). Where appropriate, we distinguished between di erent categories of health workers, for example, health professionals and lay health workers. 2. Any other individuals or groups involved in delivering and managing mHealth programmes which aimed to provide or support primary healthcare services to clients. These individuals or groups included administrative sta , information technology sta , managerial and supervisory sta , they may have been based in a primary healthcare facility or in the community, but could also have been employed at a district or national level. The criterion was that they were to be involved in supporting the delivery of primary healthcare services or the mHealth programmes or both, irrespective of their placement.
Given the review's focus, i.e. health workers' use of mHealth to deliver primary healthcare services, we excluded the perceptions and experience of clients in this review. We also excluded participants identified as technical sta who developed and maintained the mHealth architecture used, for example, those involved in writing the so ware programmes or who provided technical support to the end users.

Settings
We included studies of mHealth programmes that were part of primary healthcare services delivery. For the purposes of this review, we defined 'primary healthcare services' as one or any combination of the following. These services could be implemented in public or private primary healthcare facilities, in the community and workplace, or the homes of clients. We included studies conducted in any country.
While our review focuses on primary healthcare services as a microlevel health system, we understand and acknowledge that these services are embedded within broader, meso-level, i.e. district health systems, which deliver health care at secondary and tertiary levels (Gilson 2012; Langlois 2018). These district-level systems are in turn, shaped by the socioeconomic, political, and health system contexts at a macro level, i.e. national and global levels (Langlois 2018). It is therefore, to be anticipated that the barriers and facilitators to the successful implementation of mHealth programmes will be found across the three tiers, ranging from micro-level issues, such as protecting the confidentiality of primary healthcare client information (Labrique 2013), to meso-and macrolevel issues, such as reliable network coverage (Aranda-Jan 2014), and the integration of mHealth platforms into higher-level existing electronic health systems (Aranda-Jan 2014), and that this will be reflected in the experience of participants.

mHealth interventions
This review focused on health workers' perceptions and experiences of their use of mHealth devices to provide and support primary healthcare services.
In this review, mHealth devices were defined as mobile devices that are used to create, store, retrieve, and transmit data in real time between users (see Appendix 2 for more technical definitions related to these devices).
We included interventions in which health workers used mobile devices to provide and support any type of primary healthcare service, which revolved around uni-and bi-directional communication between health workers and clients, between health workers themselves, and between health workers and programme sta , other than health workers. In some instances, there was no direct interpersonal communication per se, but only health workers interacting with digital information available on the devices. Examples of these communications and interactions with data, included client registration and tracking, disease surveillance, various forms of decision support during consultations, for instance algorithms loaded on the devices, automated messaging to health workers, and stock notifications. We accessed the World Health Organization (WHO) taxonomy for digital health interventions (WHO 2018), and added their classification to our description of

Search methods for the identification of studies Electronic searches
In our search for synthesised evidence, we searched PDQ-Evidence (www.pdq-evidence.org) and the Cochrane Library (www.cochranelibrary.com) for related reviews on 21 February 2018. We scanned any identified reviews to assess if any of the studies included or cited in the reviews could potentially also be included in our review.
We searched the following databases for primary studies without any language, date, or geographic restrictions. We did not search Global Health in 2018 as we had no access to this database.

Searching other resources
We screened the reference lists of all the included studies and key references (i.e. relevant systematic reviews).

Grey literature
We conducted a grey literature search in the following sources to identify studies not indexed in the databases listed above. The search strategies for the main databases can be found in Appendix 3.
We re-ran the search strategies in February 2020. We screened these records and potentially relevant studies are awaiting classification; we will assess these studies at the next update.

Selection of studies
We collated all titles and abstracts identified through the search strategy into one reference management database Covidence. A er removing duplicate records, each record was independently assessed by the first review author and any one of the other review authors, for its potential inclusion eligibility. We excluded records that were not relevant to the topic of this review. Therea er, we retrieved the full text of all of the abstracts and titles that have been assessed as potentially eligible. Using the same process as for the abstracts, each full text was independently assessed by the first review author and any one of the other review authors, based on the review's inclusion criteria. To minimise bias, a review author was not permitted to assess a full text to which (s)he was an author. Given the high number of full texts we had to assess, we recruited an additional researcher and trained her to assist us with these assessments. We resolved disagreements between review authors through email correspondence and face-to-face discussions. When the two review authors could not reach consensus, we reverted to a team decision through email correspondence. In one instance, these email discussions resulted in a refinement of our inclusion criteria: though we included mobile health communication in our protocol, we did not specify the equipment used for emailing, and during a team discussion we agreed to exclude papers in which email was sent from stationery devices, such as a laptop used by a general practitioner in his/her consultation room. We contacted several study authors for more study information, when the information in the full text was insu icient to determine inclusion or exclusion of the study.

Translation of languages other than English
Abstracts of three studies required translation. Two of these were in Spanish and one in French. We translated the abstracts of these studies, using open source so ware (Google Translate: translate.google.com), and excluded the studies based on the translated version of the abstracts. No full-text studies required translation.

Sampling from the included studies
We identified 23 studies from our 2015 search. We included all 23 studies in our analysis. In 2018, we repeated our search, and identified an additional 30 studies. While small sample sizes can lead us to have less confidence in a finding, large sample sizes can also threaten our ability to carry out a thorough qualitative analysis (Glenton 2018; Sandelowski 1995). We therefore decided to select a sample of these 30 studies. Several of the studies we had identified in our 2015 search had a number of methodological limitations. This had led us to downgrade our certainty in several of the findings we had developed during our first analysis. We therefore decided to sample studies from the 2018 search based on our assessment of their methodological limitations. While we had included all studies identified from our 2015 search regardless of their methodological limitations, we only included studies from the 2018 search that we assessed as having no to moderate concerns regarding their methodological limitations. This led us to sample 20 of the 30 studies from the 2018 search (see Table 1 for the exclusion reasons of the 10 studies we appraised as having serious methodological limitations). In the main, studies that were not For both the 2015 and 2018 search studies, we extracted study information, such as country, the health worker category, the healthcare issue addressed, and the specific mobile health technology used, into an Excel spreadsheet. This served as a tool to refer to the study details during the data extraction and coding.
The data coding, extraction and synthesis process was an iterative process, aligned with the thematic synthesis process outlined by Thomas 2008. For the 23 included studies from the first search, two review authors (WO, KD) independently read each study as a whole, including the background, methods, results, discussion, and conclusions sections, to get a sense of their meaning and their contribution to answering the review question. Each review author therea er conducted a line-by-line coding of the data of the first study. They then met and agreed on the codes and supporting data. They used this code list to code the second paper, thus beginning the process of translating the data from one study into the next. New codes that emerged from the second, and subsequent studies, were added to the list, and we returned to the already coded studies, to determine if these codes applied to that data also. As the code list was amended, the authors began the process of organising the codes into broad themes, which in some cases had subthemes attached to it. Using the thematically coded data, the same two review authors jointly wrote up discreet findings. Since many of the extracts did not neatly fit within any theme, we continued the iterative process of trying to make sense of the extracts, by regrouping them with other extracts from which similar underlying issues had emerged, and eventually synthesised all the extracted data into findings.
The same two review authors (WO and KD) that led the analysis for the first 23 studies, did so for the 20 new studies from the 2018 search. By the time we started coding these 20 new studies, we already had an existing list of themes and subthemes to use as a deductive coding framework. However, we were cognisant that the new set of studies might yield data not yet captured in our framework. We therefore approached the analysis both deductively and inductively, reading the data to determine if and where it fit within the existing framework, and for what new insights it yielded. Data extracts were therefore grouped by WO and KD, both into existing categories, as well as into new categories that emerged from the data. Upon completing this for all the new studies, one review author (WO) amended the texts of the existing findings to reflect the additional data. The rest of the author team verified that all the supporting data were reflected in the amended and new findings. Upon completing this for all the new studies, one review author (WO) amended the texts of the existing findings to reflect the additional data. We also constantly evaluated each extract against our inclusion criteria and review objectives, deciding up until the very end, whether or not it was an appropriate fit. The findings thus represent the final translation of the coded data across all of the 43 included sample studies.
The aim of the data synthesis was to develop a set of findings we believe represent a trustworthy, coherent, and detailed understanding of the perceptions and experiences of those who deliver and support primary healthcare services through using mobile devices. As detailed above, we synthesised the coded data into a set of 42 discreet findings. Therea er, the one review author (WO) involved in dra ing the findings, thematically analysed these findings and grouped them into four overarching themes. These themes provide a coherent overview of our findings.

Assessing the methodological limitations of included studies
At a minimum, all included studies had to have used qualitative data collection and analysis methods. Prior to the data coding, extraction, synthesising, and writing the findings from both searches, two review authors (WO, KD) independently assessed the methodological limitations of the included studies using an adapted Critical Appraisals Skills Programme (CASP) tool (Atkins 2008). We assessed each study on the following nine criteria.
1. Adequately described setting and context 2. A well described sampling strategy that is appropriate 3. A well described data collection strategy that is appropriate 4. An adequately described data analysis method that is appropriate 5. Su icient evidence to support the claims made/findings 6. Adequate evidence of researcher reflexivity 7. Demonstrated sensitivity to ethical concerns 8. Adequately described study limitations 9. Any other concerns raised by the review authors Based on their assessment, the two review authors (WO, KD) independently graded each study as having no, or very minor, minor, moderate, or serious methodological limitations. Therea er, they met and reached consensus on their respective assessments.

Assessing our confidence in the synthesis findings
Three review authors (WO, JAW, KD) used the GRADE-CERQual (Confidence in the Evidence from Reviews of Qualitative research) approach to summarise our confidence in each finding (Lewin 2018).
GRADE-CERQual assesses confidence in the evidence, based on the following four key components.

Methodological limitations of included studies: the extent to
which there are concerns about the design or conduct of the primary studies that contributed evidence to an individual review finding. 2. Coherence of the review finding: an assessment of how clear and cogent the fit is between the data from the primary studies and a review finding that synthesises those data. By cogent, we mean well supported or compelling. 3. Adequacy of the data contributing to a review finding: an overall determination of the degree of richness and quantity of data supporting a review finding.

Library
Trusted evidence. Informed decisions. Better health.
Cochrane Database of Systematic Reviews 4. Relevance of the included studies to the review question: the extent to which the body of evidence from the primary studies supporting a review finding is applicable to the context (perspective or population, phenomenon of interest, setting) specified in the review question.
A er assessing each of the four components, we made a judgement about the overall confidence in the evidence supporting the review finding. We judged confidence as high, moderate, low, or very low. A sample (40%) of the final assessment was peer reviewed by a fourth review author (NL), and we adjusted some of the assessments a er reaching consensus with the fourth review author. We started with high confidence in all findings, and then downgraded any findings where we had important concerns regarding any of the GRADE-CERQual components.

Summary of qualitative findings table and evidence profiles
We presented summaries of the findings and our assessments of our confidence in these findings in Summary of findings for the main comparison. We presented detailed descriptions of our confidence assessment in Appendix 4.

Linking the review findings to Cochrane intervention Reviews
We sought to understand how our findings were related to, and could help to inform, the findings of six of the Cochrane Reviews of e ectiveness that were used to inform the WHO guideline on digital interventions for health system strengthening (WHO 2019). These reviews assessed the e ectiveness of the following mHealth interventions.
1. Birth and death notification via mobile devices (Vasudevan 2018) 2. Stock notification and commodity management via mobile devices (Agarwal 2018a) 3. Client to provider telemedicine (Gonçalves-Bradley 2018a) 4. Tracking of client's health status and services received (Agarwal 2018b) 5. Health provider decision support via mobile devices (Agarwal 2018c) 6. Health provider to health provider telemedicine (Gonçalves-Bradley 2018b) Each of these interventions was also the topic of a recommendation in the WHO guideline (WHO 2019).
As part of the WHO's guideline process, our qualitative evidence was used as a source of information about intervention acceptability and feasibility. The WHO technical team prepared GRADE evidence-to-decision tables for each recommendation. Each table included evidence from the relevant Cochrane Review of e ectiveness. In addition, each table included evidence from this qualitative evidence synthesis regarding the acceptability and feasibility of each intervention. The WHO's technical team prepared these tables, with input from the review authors. The technical team and review authors of this synthesis also collaborated on a supplementary document presenting evidence about the acceptability and feasibility of all these interventions. The guideline panel used these tables and supplementary documents as the basis for their recommendations.
While our review was not directly linked to the e ectiveness reviews, the findings from our review may be used to shed light on the outcomes observed in the e ectiveness reviews, by o ering insight into contextual factors, including health worker preferences, that may have influenced outcomes, either positively or negatively. Furthermore, the findings from our review may be used to develop hypothesis for subsequent consideration and assessment in future e ectiveness reviews, seeking to understand why some mHealth technologies are more e ective than others.

Review author reflexivity
The review author team represents diverse professional backgrounds, with a range of research experiences and expertise that could have influenced their input in conducting this review. All of them are experienced qualitative researchers. Except for one review author (KD), everybody has had previous experience in conducting primary mHealth research in the context of primary healthcare services in low-income settings in South Africa, and have published on this (Coetzee 2017; Leon 2012; Neupane 2014; Watkins 2018). FG has also experience in conducting telemedicine research in high-income contexts (Gri iths 2017). Our experiences in conducting e ectiveness studies and process evaluations of mobile health programmes, included positive, negative, and mixed results. This provided us with a good platform for engaging and understanding the complexities and nuances of qualitative research of mobile health interventions.
The review authors reflected on the influence our perspectives might have on the conduct of the review, and in some cases tried to moderate this influence, in a number of ways. During the screening of abstracts/titles and full texts, the team constantly referred to each other to resolve conflicts, and in many instances a team decision was called upon. As is standard practice within qualitative research, the two review authors (KD, WO) who did the data coding, extraction, and synthesising, and wrote the findings, constantly discussed with each other how their own background and position, may have a ected their analysis and writing of the findings.
WO realised that at times his research experiences resonated strongly with some of the included studies, and was aware that this could lead him to give these data more importance than was due. Conversely, he was aware that he could be more dismissive towards data which contradicted his experiences. KD questioned the weight he attributed to certain data, ensuring that all data were equally represented in the final set of findings. WO and KD repeatedly questioned each other's interpretation of the data and how it fitted with the existing findings. They also called upon other members of the author team to verify that the findings were reasonable reflections of the supporting data. JAW, KD, and WO also used the same process of constant discussion and being aware of their personal perspectives when appraising their confidence in the findings. Finally, the contact editor of this review read each finding and its supporting data closely. She pointed to any mismatch between the supporting data and a finding, and critically engaged with our interpretation of the data, which led to a refinement of our analysis and writing of the findings.

Library
Trusted evidence. Informed decisions. Better health.

Results of the search
We screened 7225 records. Fi y-three studies met our inclusion criteria. We purposively sampled 43 of these studies for inclusion in our analysis ( Figure 1). All of the sampled studies were published between 2005 and 2018; see Methods section -'Sampling from the included studies' for a description of how we sampled these studies. In February 2020, we re-ran the search strategies. We screened those records and 85 studies that we identified as potentially relevant are listed under Studies awaiting classification; we will assess these studies at the next update.

Description of the included studies
In this section, we describe the studies that we sampled for analysis. For a more detailed description of each included and sampled study, see Characteristics of included studies. For an overview of the studies that we included but did not sample, see Table 1.

Study participants
In 17 (40%) of the sampled studies, the participants included both lay health workers and a range of professional health workers, such as nurses, paramedics, doctors, midwives, pharmacists, and laboratory sta . Non-medical professional participants, found in four studies, referred to programme managers, administrators, decision makers, and social workers. A further 10 (23%) and 12 (27%) studies only included lay health workers or health professionals, respectively. In four (9%) of the studies, the participants were only described as 'healthcare providers'.

Types of mobile devices
Mobile phones, also described in the studies as 'iPhones', or 'smartphones', were used in 31 (72%) of the sampled included studies, with either personal digital assistants (PDAs) and tablets in 11 (26%) studies. A combination of mobile phones and tablets was used in one study. Most of the mobile health interventions, in total 18 (42%) of the studies, comprised of so ware loaded onto the mobile device which guided the consultation process, for example, a screening algorithm that allowed health workers to diagnose clients. In the remainder of studies, the intervention comprised of in-person and/or text message communication, collecting of surveillance data, and a range of other interventions, such as health promotion materials in the form of videos on the devices, accessing the Internet, and in one instance, allowing low-level workers to send wound care images to higher-level workers for treatment advice. The healthcare issue addressed through the mobile health programmes was in the main, maternal, neonatal and child health, with 17 studies (40%) reporting on this. Other healthcare issues included communicable and non-communicable diseases, cardiovascular diseases, and intimate partner violence.

Methodological limitations of the studies
Details of our assessments of the methodological limitations of the individual sampled studies can be found in Table 2. We assessed 10 studies as having no or very minor methodological limitations, 13 as having minor methodological limitations, and 14 and six studies respectively, as having moderate and serious methodological limitations. Of the 43 studies, 41 were published in peer-reviewed journals, and two as full Masters theses (Barnabee 2014; van der Wal 2016). Most peer-reviewed journals set a word limitation, which is not suitable for the comprehensive reporting of qualitative research. Our concerns regarding the general lack of rich data and thick description, as well as o en poor descriptions of how participants were sampled and recruited, and of researcher reflexivity, may be attributable to these word limitations. Many studies provided poor descriptions of ethical considerations, apart from mentioning that ethical approval was obtained. All studies provided at a minimum a brief description of the study context, who the participants were, the mHealth programme, and to some extent, the study limitations.

Confidence in the review findings
Out of 42 review findings, we had high confidence in 13 findings, moderate confidence in 18, low confidence in six, very low confidence in five (Summary of findings for the main comparison).
Our explanation for each GRADE-CERQual assessment is shown in the evidence profile in Appendix 4.

Review findings
From the synthesised data, we dra ed 42 individual findings (Summary of findings for the main comparison), which we organised into four overarching themes. Theme 1 deals with how mHealth changed how health workers worked with each other, in particular through connecting lower-level health workers with higher-level health workers, and peers with each other. Theme 2 describes how mHealth changed how health workers delivered care, and includes health workers' perceptions and experiences about issues, such as accessing information from the Internet, providing care over distance, and using treatment algorithms. In Theme 3, we present how mHealth led to new forms of engagement and relationships with clients and communities, mainly because mHealth allows direct, and o en bi-directional communication between health worker and client. This theme also covers issues such as elevated health worker status that comes from health workers using modern technology and needing to protect client information on their devices. Theme 4 details how health workers' use and perceptions of mHealth can be influenced by factors tied to costs, the health worker, the technology, the health system and society, and how poor network access and poor access to electricity could make mHealth di icult.

Finding 2: Lower-level health workers valued being able to reach higher-level health workers via mobile devices, and
perceived the advice and support they received as improving their care and as satisfying to clients. When higher-level professionals responded in anger, it made lower-level health workers reluctant to call them (moderate confidence in the evidence) Apart from facilitating contact between same-level health workers (Finding 1), lower-level health workers particularly valued being able to reach higher-level health workers through mobile devices ( health workers perceived direct contact as improving relationships between lower-and higher-level workers (Chang 2011). In contrast, when higher-level professionals responded in anger, it made lowerlevel health workers reluctant to call them (Ayiasi 2015).

Finding 3: When higher-level health workers failed to respond
and support lower-level workers through mobile devices, lower-level sta had negative perceptions of these devices. One study emphasised the importance of having health professionals' buy-in with mobile health to ensure that mobile devices were optimally used to support lay health workers (moderate confidence in the evidence) The data suggests that those participants who expressed an opinion desired good communication and co-ordination with their seniors and others in their health systems context. Lack of anticipated co-ordination, support and responsiveness, in particular in emergency cases (Mwendwa 2016), through mobile technology's connectedness with higher-level sta or emergency services, led to a negative experience of the intervention amongst lower-level health workers (Cherrington 2015; Huq 2014; Mwendwa 2016;Quinn 2013; Toda 2017): [Lower-level health worker]: "The technology is good but let the higher levels take it seriously otherwise there is no need of sending the instant reports. The last time is [sic] sent a suspected tetanus baby but he died without anyone coming to see the child at the facility I had referred to" (Toda 2017). This lack of responsiveness was described as limiting the e ectiveness of the intervention (Mwendwa 2016). In contrast, direct communication was expressed as having a positive impact on the facility and in turn on client experience. In one instance, a lack of managerial interest in mHealth-facilitated disease surveillance demotivated the lower-level workers from consistently using mHealth for their work (Toda 2017). In one study (van der Wal 2016), matters were complicated when the supervising doctor was not given a smartphone, as he could not supervise what the lower-level health workers were doing with the phones. His lack of a phone led to his lack of supporting these workers and encouraging them not to use the intervention application either. In turn, these lower-level workers wanted him to be given a phone, not only for his buy-in, but also because they believed that this would enhance how they worked together. Cochrane Database of Systematic Reviews connection]" ..."I've always come in once a week, because I'm dying to see everybody" (Valaitis 2005).
Finding 5: Some health workers relayed that mobile devices improved their reporting to supervisors and encouraged them to report more truthfully. Others compared mobile device-facilitated supervision to "big brother watching". Some supervisors thought that mobile devices allowed them to better identify sta who needed support (moderate confidence in the evidence) Some health workers felt that mobile devices improved reporting to their supervisors (Barnabee 2014; Medhanyie 2015), their relationship with their supervisors (van der Wal 2016), and encouraged truthful reporting (Chang 2011). On the other hand, there were supervisors who did not think that mobile devices were a safeguard against false reporting (Jennings 2013). Mobile health-facilitated supervision le some supervised workers with a sense of "big brother [is] watching" (Valaitis 2005). Mobile health resulted in work being more visible to the supervisors (Henry 2016). In instances where clients complained about health workers who did not visit them, workers reverted to having clients signing a paper record for such proof, and suggested electronic signatures be made available on their devices (Schoen 2017). In the context of conducting disease surveillance, health workers expressed a need for face-to-face interactions with those overseeing the surveillance (Toda 2017). Some supervisors expressed that mobile technologyfacilitated supervision allowed them to be more aware of their sta 's work, in particular when the latter experienced problems. These supervisors perceived this increased awareness of sta 's performance as positive because they could address the problems that came to their attention as a consequence (Henry 2016; Madon 2014; Mwendwa 2016): "[WhatsApp] has made me learn a thing or two, it has made me get to know characters as far as community health volunteers are concerned … I can even gauge performance when it comes to community health volunteers" (Henry 2016). Finding 7: Even when health workers received messages that were automated, rather than sent directly from a manager or supervisor, this was still experienced and responded to, as a kind of supervision. Some lower-level health workers experienced it as supportive to their work, while others felt guilty for not providing correct care as per these messages (low confidence in the evidence) Automated text messages about illness and client management, and in one study, motivational messages (Jones 2012), sent to lower-level health workers' mobile phones, were perceived by some of them as supervision (Jones 2012), and they felt it improved their care and knowledge (Cherrington 2015; Ilozumba 2018; Jones 2012; Mwendwa 2016). Workers valued the messages' conciseness, and saw it as providing up-to-date information and as being a useful reminder to provide correct treatment (Jones 2012): "It kept me on task...if I forgot, I would turn it on and it would pop up, "You're late"...it was wonderful" (Cherrington 2015). Some felt motivated by receiving automated treatment messages, but others felt guilty for not providing correct care as recommended by these messages (Jones 2012). There were concerns that the text messages were too repetitive in the information they carried, and that motivational messages on their own, without treatment guidance, were less meaningful than those with treatment guidance (Jones 2012). Finding 9: At times, health workers used their mobile devices to access the Internet for health information, and found it useful when they were with clients who needed the information. This interaction also included health workers providing clients with additional information beyond the healthcare intervention. But, if the only way that health workers could access online information, required them to use their own money to purchase data, then this could be prohibitive to them accessing such information (low confidence in the evidence) Some health workers used their mobiles devices to gain access to the Internet where they accessed health information, clinical guidelines, health promotion material, and other information thought to be needed by clients (Bacchus 2016; Hampshire 2016; Schoen 2017; Watkins 2018): "Doctors and nurses who used the Internet for work reported using search engines on their phones such as Google, to access clinical information on diseases or prescription drugs. A few nurses reported accessing the digital versions of government clinical guidelines" (Watkins 2018). They found the quick access to such information useful, particularly when they were with clients who needed more information about a certain condition and its treatment (Hampshire 2016). The cost of data, when borne personally, sometimes prevented health workers from searching for information (Watkins 2018). In one study, health workers used their tablets to give non-health information to clients who wanted to further their education, and thus needed access to information to enable reaching this goal: "We don't have tablets usually so I used the tablet to do some personality tests of my clients who wanted to be in school" (Bacchus 2016).
Finding 10: mHealth held the promise of increasing service e iciency for many health workers, but the experience of whether this promise was borne out in practice, varied in the accounts of health workers. It was experienced as e icient if it improved feedback, speed and workflow, but ine icient when the technology was slow and time consuming. Some were concerned that if mHealth was too e icient, making work faster, that this may justify sta cutbacks (high confidence in the evidence) Health workers' experiences of the e iciency of mobile devices varied across and within studies. In general, e iciency was related to work being done more quickly, whilst ine iciency was related to the extra time and work it took when using these devices. . In one study, health workers were concerned that the ability to work faster because of mobile health was used by management to support sta cutbacks (Valaitis 2005). There were also health workers who experienced mobile health overall, or in part, as ine icient. This included perceptions of the technology as slow, time consuming, and increasing workload: "It complicates and increases our work. One has to stop everything she is doing and concentrate when sending reports" (Mwendwa 2016).
The perceived increase in workload was in part because it was experienced as more cumbersome and taking longer to complete work (van der Wal 2016), compared to standard practise (Ginsburg 2016; Hao 2015; Kolltveit 2017; Mwendwa 2016), and sometimes because the application was slow (Schoen 2017). Mobile health also increased workloads when better screening procedures resulted in detecting more clients who needed treatment (Lodhia 2016). ..those who are in the most remote areas who have the highest prevalence for blindness will now be linked to the health system and so people will be able to find them and treat them" (Lodhia 2016). It was also used to schedule visits in advance which avoided wasteful travelling (Hampshire 2016). Some workers pointed out that finding clients in these contexts still required being provided with detailed client information on mobile devices (Rothstein 2016). Using these devices saved travelling time in an urban setting too, which allowed health workers to spend more time with their clients (Valaitis 2005).   In one study, the data sharing between di erent services was regarded as achieving a policy goal (Murray 2011). Workers reported specific advantages to data generated through mobile devices compared to paper-based reports, which included that it was easy to format the reports to their needs, such as viewing individual or aggregated data ( Finding 15: In most cases, health workers perceived mobile health as more advantageous than paper. However, some continued to prefer paper (high confidence in the evidence) The majority of health workers mentioned advantages to their use of mobile devices, compared to using paper-based systems , and a more durable platform compared to paperbased systems, for example with papers that can be damaged by rain (Nguyen 2015; van der Wal 2016). For some workers, these benefits of mobile devices over paper-based systems, as well as the perceived improvement of care, led to a commitment to use these devices (van der Wal 2016). In one study, being issued with mobile devices made health workers feel more professional compared to when they used a paper-based system (Valaitis 2005). Yet, in one study, it was found that despite the mentioned advantages, most of the health workers did not use these devices during their visits to clients (Schoen 2017). Some workers preferred paper-based systems, as they perceived these as safer to store information, more flexible, and had concerns about malfunctioning technology Cochrane Database of Systematic Reviews were slower than their paper-based systems (Schoen 2017), and that it was easier to correct errors on paper forms because this did not require technical knowledge (Nguyen 2015).
Finding 16: mHealth interventions sometimes required health workers to perform tasks that were peripheral to regular service delivery, such as registering clients onto the system. These more menial tasks were sometimes regarded as undermining to professional sta (very low confidence in the evidence) In some instances, the use of mobile devices to deliver or support healthcare services resulted in tasks that were additional to health workers' routine care practices. These tasks included registering clients onto the Finding 19: Health workers were aware of the importance of protecting confidential client information when using mobile devices, and the confidentiality risks in cases of stolen phones and using their SIM cards in colleagues' phones. Health workers were alert to clients' concerns when they shared personal information concerning stigmatised issues, such as HIV/AIDS and interpersonal violence, and suggested ways to keep the information confidential. They emphasised building a trusting relationship with clients prior to using the devices (high confidence in the evidence) Health workers were conscious of protecting clients' confidential information ( Workers were also aware of clients' anxieties as to how their information will be used during mHealth-facilitated consultations (Coetzee 2017), particularly when the information they provided , and an explanation to clients on how their information will be protected (Lodhia 2016). Some health workers preferred sending automated text messaging to clients from a non-traceable number, rather than using their personal phones to send these messages (Murray 2015). Other health workers pointed to the importance of building a trusting relationship with clients prior to using these devices, perceiving this as a way to mitigate clients' concerns about confidentiality (Coetzee 2017). Health workers were also concerned about risks to client information in cases of stolen phones, and when using their SIM card, with client information, in colleagues' phones (Lodhia 2016; Mwendwa 2016). Finding 22: Health workers experienced the use of mobile technology to provide health care, as being met with both trust and skepticism from clients and the communities they served. They described how trust or skepticism in the device was translated into trust or skepticism of their service when using the device. Some found that using mobile devices raised their social status with clients, and even their families. Others were concerned that using expensive equipment would emphasise inequity between themselves and clients (high confidence in the evidence) Health workers reported that mHealth devices raised their social status ( Being seen as a trustworthy health worker made clients more receptive to workers' use of mobile devices (Ilozumba 2018): "I am not telling. The mobile is telling. They will hear its words. Whoever is nearby they also become silent and will here all things" (Ilozumba 2018). Some health workers believed that using mobile devices helped them build a relationship with clients, and that satisfied clients told each other about the quality of care they received through mobile devices (Jones 2012;Khan 2015). This encouraged more clients to come and see them (Khan 2015 Cochrane Database of Systematic Reviews support (Cherrington 2015). Health workers said that some clients were attracted from outside their catchment areas because of the mobile devices they used. Some clients not having the illness condition for which the health workers used the devices for, would question why they were not treated with these mobile devices (Vedanthan 2015). In contrast, there were health workers who experienced clients and communities as less responsive if they were sceptical as to the devices and how it can be used (Lodhia 2016; Mwendwa 2016): "Some mothers still do not know Rapid SMS and do not understand how useful the system is so they are uncooperative and do not give us information" (Mwendwa 2016). Workers suggested that this skepticism be addressed through community education (Lodhia 2016). There were also concerns amongst workers that using expensive devices working in resourceconstrained communities would emphasise the social inequity between clients and health workers, and would impact negatively on connecting with underprivileged clients (Valaitis 2005).

Finding 23: Health workers experienced clients as having an
opinion not only about their use of mobile devices, but as having an opinion on the devices themselves, which influenced how they responded to care delivered with the support of these devices. Health workers ascribed clients' enthusiasm for mobile devices as due to these clients' perception of the devices as prestigious, o ering trustworthy information, and providing confidentiality. They perceived clients as more receptive when these clients were familiar with the devices used. There were concerns that clients who felt that the use of these devices during care was too time consuming, and would respond negatively to its use (moderate confidence in the evidence) Health workers felt that some clients were enthusiastic about the use of mobile devices in support of their care, as they perceived it as prestigious and modern ( . Discontinuation of the use of mobile devices caused problems with new and returning clients, who expected it to be used (Mitchell 2012). Some workers perceived clients to be more willing to report sensitive information through mobile devices, than doing so verbally or on paper (Bacchus 2016; Westergaard 2017): "So if they feel safe enough to do it on the tablet, feeling like it's a little anonymous, it starts to break down those walls and maybe next time they'll want to talk about it" (Bacchus 2016). Workers also reported clients' preference for mobile devices when these o er non-invasive medical diagnosis, as compared to invasive diagnosis, such as drawing blood (Ginsburg 2016). They also perceived clients to be more receptive to mobile health, if they were already familiar with the type of device used by the health workers (Garg 2016; Lodhia 2016). In contrast to the positive reactions, health workers were concerned that clients who perceived using mobile devices during care, as too time consuming, would respond negatively to such devices (Ginsburg 2016; van der Wal 2016): "The few negative reactions from clients related to them being impatient with the lengthy consultations. [Health worker]: Before, we mainly asked about danger signs but now we have many questions… some patients were impatient to answer all the questions" (van der Wal 2016). Other health workers reported that clients, particularly the elderly, disliked the use of mobile devices during consultations (Schoen 2017).

Finding 24: Some interventions required clients to have phones as well as health workers. Health workers described this
as challenging for multiple reasons, including clients not having phones, changing their phone numbers regularly, not knowing how to use a phone, being a target for crime because of possession of the phone, and women being prohibited from accessing phones. Health workers suggested competitive pricing to increase clients' accesses to phones, and to issue clients with phones (moderate confidence in the evidence) Health workers identified that interventions that required communication between health workers and clients, might pose several challenges to the clients. These included clients who regularly changed their phone numbers without informing the health worker (Hirsch-Moverman 2017): "For others you find that the patient has given you a certain number, in a blink of an eye he has changed it without telling you that he doesn't use that number anymore" (Hirsch-Moverman 2017); clients who did not have phones (Chang 2011; Tewari 2017; Wol -Piggott 2018), or did not always have their phones with them (Tewari 2017); clients who did not have money to buy phone credit or access to electricity (Huq 2014; Wol -Piggott 2018); and clients who were afraid of being robbed of their phones: "some of them …. they say they can't have these telephones because they come very early." It was explained that women queuing in the dark might be targets for criminals (Wol -Piggott 2018). There were also clients who did not know how to use mobile phones (Tewari 2017; Wol -Piggott 2018). In one case, women's access to mobile phones was prohibited as a consequence of gender discrimination (Hirsch-Moverman 2017). Competitive pricing of mobile phones and phone credit costs increased clients' accesses to phones, and thus eased health worker -client communication (van der Wal 2016). In one study, health workers reported that providing clients with mobile phones, promoted these clients' general social connectedness (Murray 2015).

Finding 25: Health workers were ambivalent about
interventions that required clients to use the health workers' mobile devices during consultations. Their optimism was tempered by concern that there was a loss of meaningful engagement with clients (low confidence in the evidence) Health workers had mixed reactions to clients using their (health workers') mobile devices on their own during consultations. In one instance, workers expressed that giving clients the device to access surveys and health promotion material on their own during the consultation allowed clients to deal with sensitive topics in private (Bacchus 2016). Yet, these same health workers were concerned about the loss of meaningful interactions during this process, and raised concerns that they were unable to engage with the clients about the topic at hand, unless the client was willing to share the activity with them: "The challenge is how to keep it personal. If [women] answer positive on the tablet and then you just close the tablet and "oh thank you" and put it away then you've just told her, all I needed was for you to answer the questions. I'm not really here to help you. You have to say okay so this is how you answered and this is how you scored, let's talk more about that. The computer can't do that part, all it can do is take down the information and it's up to the nurse or home visitor to expand upon it and actually Cochrane Database of Systematic Reviews get her the assistance that she needs" (Bacchus 2016). In another instance, health workers felt that having material on the device which they could share with clients was useful on days when they (the health workers) were tired, and not up to interacting with the client (Coetzee 2017). However, it has to be noted that in this study, the authors pointed out that they did not agree with the health workers' interpretation that using the device to show health promotion material when they were tired, was a good thing.
Finding 26: Health workers reported that their access to mobile devices was beneficial to clients and communities who were too poor to own mobile phones (very low confidence in the evidence) Health workers reported that their access to mobile devices benefited clients and communities who were too poor to own mobile phones, or to a ord paying local merchants for the occasional use of a mobile phone (Chang 2011). Health workers' access to a mobile phone enabled them to access higher-level care, on behalf of these clients: "I was saying that you can find a whole village without a phone. So, giving these PHWs phones, is helping a lot… By giving these PHWs phones I think it helped them a lot because they just go and contact their PHW, and the PHW calls us" (van der Wal 2016). "…people have been motivated to take the drugs because they know that once you get to their home and you press the phone to send a message it means that you are reporting the patient that he has not taken the pills properly….Patients are motivated now to take their pills" (Chang 2011). In one study, health workers thought that clients were motivated to improve their adherence because they were aware that information about their adherence was being relayed to clinic sta by field sta in real time via the mobile devices (Chang 2011). Some workers reported that clients were more responsive to visual material than verbal messages (Coetzee 2017; Lodhia 2016). Some attributed this to a cultural preference for visual information over verbal information (Coetzee 2017). Other health workers cautioned that repetitive showing of health promotion videos would not improve uptake of the health messages (Bacchus 2016; Coetzee 2017): "My client looked at me one time, and she said "how many more times do we have to do this?" (Coetzee 2017). Workers reported that issuing phones to clients for health-related usage led to increased use of healthcare services (Murray 2015). Health workers perceived that graphic displays on a device helped clients to better understand their condition: "It's wonderful. I got better results than I expected...If patients see the risk bar, they understand very well that they have a high risk of CVD...We gained knowledge from this percentage display too...This is 100% beneficial to the doctor" (Praveen 2014).

Theme 4: Health workers' use and perceptions of mHealth could be influenced by factors tied to costs, the health worker, the technology, the health system and society, poor network access, and poor access to electricity
Finding 28: Some health workers accepted bearing the costs of mHealth interventions themselves, but were dissatisfied when phone credit to use the phones was not delivered on time.
Health workers felt that clients appreciated it when health workers called them, as it saved them costs (high confidence in the evidence) The cost implications for health workers, of mHealth was discussed across the studies, with di ering opinions as to the appropriateness and a ordability of bearing costs personally (Hampshire 2016; Messinger 2017; Watkins 2018; Wol -Piggott 2018). Bearing the costs personally was accepted by health workers either as part of their altruism (Hampshire 2016; Messinger 2017; Watkins 2018): "I know some nurses who will never use their own phones because they have no passion for the job. But, for some of us, it is the passion for the patients and the work that makes us continue" (Hampshire 2016); or from a sense that the investment would generate a greater demand for their services and thus better income (Khan 2015;Messinger 2017). There was less satisfaction with interventions that failed to deliver promised phone credit, on time (van der Wal 2016).
Health workers also felt that clients appreciated it when the health workers saved them call costs by phoning them, rather than the other way round: "Even if any client calls me then I cut the line and call back from my phone. If I do so, client will say I am kind to them" (Messinger 2017). . Some health workers expressed initial hesitancy, but over time and with training, became more comfortable, to the point of expressing concern that they were becoming dependent on the intervention devices ( Cochrane Database of Systematic Reviews the technology could lead to dual systems, paper and electronic, and poor integration into normal routines (Murray 2011). Health workers, unfamiliar with the technology, were regarded by their trainers as needing emotional reassurance (Murray 2011). This emotional response was also seen in another study where health workers claimed to feel encouraged if sending a message worked, but feeling hopelessness if it did not (Mwendwa 2016). In one study, the authors expressed that those health workers that understood the mHealth application, expressed no concerns, but for other health workers who lacked this understanding, there was concern over errors in reading laboratory results (Hao 2015). A lack of ongoing training was said to lead to insecurity, and this insecurity in turn reduced enthusiasm for the intervention and hindered In one study, where nurses already used the Internet or web applications on their phones for personal use, they were also more interested in gaining computer skills for their work, than those with no prior experience of using the Internet (Watkins 2018). It was not always clear where the di iculty stemmed from, faulty technology or unfamiliarity with the technology, but what was clear in one case, was that those who struggled, were the lowest educated (Ilozumba 2018). Poor digital literacy was not limited to health workers, but also to those with whom they had to engage, such as village leaders who expressed that they needed training because they were not able to understand reports and information given on the mobile phone (Madon 2014). One study suggested that irrespective of computer and smartphone literacy levels, that poor uptake was related to di iculty with typing, with health workers suggesting di iculty in use as a result (Shao 2015). Some health workers turned their digital illiteracy into an opportunity to build relationships with clients by asking clients to show them how to use the devices: "One of the older home visitors revealed that she used her own lack of experience with technology as a way of encouraging young women to open up to her with the computer tablet (it's like I'm saying you're really tech savvy with this and it's sort of like a prop you know. Like we're going to talk about this, but you get to use this tablet)" (Bacchus 2016). Others asked their children to assist them, which they viewed as appropriate given that there was no confidential information on the device (Coetzee 2017). It was reported that health workers with low client caseloads used the devices infrequently, and thus may forget how to use them (van der Wal 2016). In some instances, workers felt that poor competency with mobile devices threatened their job security (Madon 2014; Murray 2011): "At the same time, some VHWs were discriminated against for being unable to adapt to new ways of working as a result of the technology. For example, in one of the villages with low prior phone possession, some VHWs with visual impairments who had become used to entering health information onto paper registers were declared as being unfit to work because they could not easily operate the mobiles" (Madon 2014), and felt embarrassed when making mistakes whilst being with clients (Coetzee 2017). In one study while all doctors reported using the Internet at least once a week, the same was only true for a quarter of the nurses (Watkins 2018).

Finding 30: Health workers expressed a need for training and familiarity with mobile devices to overcome their initial anxiety in using the devices. Peer training from technologically proficient colleagues was experienced as valuable. In several cases, health workers wanted refresher training and pointed
to the importance of training replacement sta . Not having mentors who used mobile devices, impacted negatively on lower-level workers' ability to learn how to use these devices (high confidence in the evidence) Some health workers experienced anxiety in understanding and using mobile devices, and felt that training and familiarity with these devices were needed to overcome this anxiety ( Finding 33: Health workers held mixed views on choosing between tablets and smartphones. Some felt that the type of content on the device was more important than the device itself. However, other health workers preferred tablets over smartphones, mainly because the bigger size of the screen was perceived as easier for client engagement (very low confidence in the evidence) In one study, half of the health workers were given smartphones and the other half were given tablets (Shao 2015). Their opinion about the two di erent tools were reported as being similar, showing more concern for the algorithm loaded on the devices, than the devices themselves (Shao 2015). In contrast, workers in another study who were issued with mobile phones, thought that tablets would have been better, because it would have made it easier to show clients what they were doing on the devices, and that it looked more professional (Schoen 2017).
Finding 34: Some health workers felt that sustainable, at scale mHealth programmes required approval and stewardship from political leaders, such as ministries of health. Leadership interest in mHealth interventions was described as motivating to health workers. Health workers suggested that such leaders should be engaged early and continuously throughout the programme, and be provided with evidence of e ectiveness, so as to secure their support. The lack of high-level stewardship impacted negatively on the mHealth programme (low confidence in the evidence) Some health workers felt motivated when political interest was shown in mobile health programmes: "Even where healthcare professionals had to organize and manage the telemedicine use on their own, they could still feel some support from leaders when they experienced their attitudes toward telemedicine to be positive. [Health worker]: "There is an interest for this intervention by the leaders, but they are not so interested in knowing more about the intervention. My leader is pleased with the fact that I am handling it all" (Kolltveit 2017). They perceived that sustainable, at scale use of mobile devices required approval and stewardship from higher-level leaders, including decision makers at national ministries of health (Ginsburg 2016; Kolltveit 2017; Lodhia 2016; van der Wal 2016). They felt that early and continued engagement with these leaders facilitated their support (Ginsburg 2016). Workers also reported that not including higher-level professionals in programmes, a ected intervention uptake (Kolltveit 2017).
Receiving mobile devices and the means to use it, was perceived by some health workers as an acknowledgement of their work (van der Wal 2016). They also mentioned that evidence of e ectiveness and cost-e ectiveness were important to ensure higher-level support for mobile health programmes (Lodhia 2016). When there was lack of higher-level stewardship, health workers felt it impacted negatively on the mobile health programme (Kolltveit 2017).
Finding 35: Health worker accounts pointed to the strong influence of the health systems and social context in which the intervention was embedded. Contextual and systems issues, such as di erence in language use between clients and health workers, gender discrimination, discomfort with professional hierarchies, poverty, resource constraints, sta attrition, and more, all of which were external to the technology and the physical device, influenced how health workers experienced mHealth and the use of mobile devices for service delivery, in their di erent contexts (moderate confidence in the evidence) Health workers' accounts showed that the systems in which mobile health programmes were implemented, and contextual issues external to the devices itself, shaped their experiences and perceptions regarding their use of the devices (Wol -Piggott 2018). These included language di erences between workers and clients, which made communication di icult irrespective of using mobile devices or not (Khan 2015); cultural practices, such as gender discrimination against female use of mobile phones (Huq 2014); and educational and professional di erences which caused strained relationships between lower-and higher-level workers: "Two village doctors reported that the lack of comfort with the call centre doctors resulted in their reluctance to use the intervention. As one village doctor said, "When I talk with my patient I feel like we are brothers during our conversation and the patient feels comfortable (about sharing problems). But when I consult with call centre sometimes I didn't feel that warmth probably because they are from another place and we never met [sic]"" (Khan 2015. Furthermore, health workers felt that client poverty made the uptake of mobile health services challenging for clients if it required that they access a mobile device themselves, (Huq 2014; Tewari 2017). Sta attrition and shortages were o en reported as main barriers to optimal implementation of the intervention, and uptake of the use of the devices as an additional activity, on top of existing sta ing problems ( Cochrane Database of Systematic Reviews They have to be trained" (Rothstein 2016). The same applied to workers who reported that unsupportive higher-level professionals made them less enthusiastic about mobile health (Praveen 2014; van der Wal 2016). Workers also found it challenging to learn how to use the new devices amidst the rush and unpredictable workday routines in primary healthcare facilities (Rothstein 2016). They also pointed out that other health systems' problems, such as when drugs were out of stock (Shao 2015), overstretched laboratory services (Shao 2015), and health workers not knowing pre-mHealth reporting guidelines (Toda 2017), made it di icult to use mobile devices. In one disease surveillance programme which used mobile devices to record and report the surveillance data, sta felt that face-to-face supervision was an important enabler for high-quality data, but resource constraints and poor road conditions made such supervision impossible, and that this lack of face-to-face supervision, not the technology, impacted negatively on the quality of the fieldwork sta (Toda 2017). Across several studies, health workers suggested improving the social marketing of these programmes, including advertising and sharing information with users, clients, and communities, as this may enhance the acceptability and uptake of these programmes (Ginsburg 2016;Khan 2015;Lodhia 2016;Mwendwa 2016). They also recommended that those using the devices be consulted as end users during the planning and implementation of the intervention: "I [supervisor] think there needs to be more involvement of other people (users), to know, what do they think, how can this be done better, what are their inputs rather than pushing it down and ask them just to use the system" (Hao 2015); be included in decisions about system changes to be introduced because of the mobile health programme (Hao 2015); be given money for phone credit (Barnabee 2014); and that all paper-based stationery, not just certain documents, made available on the mobile devices (Schoen 2017). Some recommended an automated response service for when health workers are in emergency situations (Mwendwa 2016). Workers advocated for the appointment of a 'champion' who could provide technical and intervention assistance (Kolltveit 2017; Murray 2015): "The importance of having a colleague who could champion this intervention was described as a prominent success condition: "Well, we have some among us who facilitate it all. They have encouraged us, and when some of us think this is too much work, they have been there with their enthusiasm. This enthusiasm has been valuable to us"" (Kolltveit 2017). There were recommendations that those collecting routine data using mobile devices, should be informed as to how the data are put to use, and that not getting this feedback made them feel like students who sat an exam without getting their results (Madon 2014).
Finding 38: Health workers had several technical recommendations to improve mobile health devices, for instance solar panels to counter poor electricity access and using photos to track clients' recovery from illness. Other recommendations included using sturdier devices, bigger screens, and having common applications, such as work scheduling on the devices (moderate confidence in the evidence) Health workers suggested ways to improve the technical aspects of mobile health interventions. Password protection was proposed to keep client information confidential, installing tracking so ware to mitigate loss or the of a device, and having the alert protocols put in place (Coetzee 2017 Cochrane Database of Systematic Reviews when typing on a tablet (Schoen 2017). In one instance workers thought that tablets, with their bigger screens, would make it easier for clients to see what the workers were doing on their devices, compared to small screen mobile phones (Schoen 2017).
Finding 39: The main challenges health workers experienced in using mobile devices, were poor network connectivity, access to electricity, and the costs to recharge devices. Solutions o ered, included using solar panels, using the powered-up phone of a colleague, or reverting back to the paper-based system. Sometimes poor connectivity resulted in client dissatisfaction because it created delays in receiving health care. Health workers' commitment to their clients motivated them to cope with these and other challenges (high confidence in the evidence) The most cited challenges for health workers using mobile devices were poor network connectivity ( In one study it served as a deterrent to downloading the intervention so ware application (Ginsburg 2016). Poor connectivity impacted on clients, who became impatient in waiting to connect to a doctor or waiting to get a text message prescription (Khan 2015). To circumvent these challenges, workers used private solar panels (van der Wal 2016), though some complained that it took a long time to charge their devices (van der Wal 2016). In other instances, health workers solved poor connectivity by uploading their work at times when connectivity was good, but that may have been inconvenient to them (van der Wal 2016), or by looking for places with good connectivity: "We also have problems with network connectivity. For example, so the uploading may be a challenge. And sometimes they're typed, but it doesn't go through. Sometimes it doesn't go through at all. You have to go and climb a tree" (Rothstein 2016). Some had to walk long distances in search of good reception and/or electricity (Ginsburg 2016; Madon 2014; Nguyen 2015; Rothstein 2016). Some workers used their own resources to ensure connectivity (Quinn 2013; Watkins 2018), whilst others reverted back to the paper-based systems when the mobile devices did not work (Nguyen 2015; Schoen 2017). Given problematic access to electricity, workers at times had to use the powered-up phone of a colleague or otherwise put their SIM card into a colleague's powered-up phone (Mwendwa 2016). In one case they used informal business networks to have their phones charged at a low cost (Chang 2011). Poor connectivity did not just challenge health workers, it challenged the success of the intervention being conducted as intended. It resulted in delays in providing health care (Quinn 2013), which upset some clients (Khan 2015), it caused complaints from workers that work was lost (Schoen 2017), devices 'froze' whilst they were working (Schoen 2017), and resulted in slow transmission and receiving of information (Schoen 2017; van der Wal 2016). However, health workers' commitment to their clients motivated them to cope with these and other challenges: " … [despite] feeling discouraged because of lack of Internet connectivity, they [health workers] kept stressing that the mHealth tools were in their community's best interest, good for their health system, and therefore felt a professional duty to accept and use the application. [Health worker]: "Our priority is mother and child care, so it is important that we fully succeed, isn't it?"" (van der Wal 2016).
Finding 40: Health workers expressed dissatisfaction with mobile devices when technology changes were too rapid, showed a dislike for typing, and were concerned that mHealth impersonalised their interaction with clients. Since these dissatisfactions were only infrequently raised within the data set, it is unclear if these perceptions reflect wider experience (low confidence in the evidence) Some studies o ered challenges raised by health workers that were not commonly discussed across the included studies. Health workers expressed dissatisfaction with aspects of mobile devices, which included when technology changes were too rapidly introduced (Valaitis 2005), or when their expectations of the devices were not met, for example when they anticipated it would make manual data capturing unnecessary, yet they still had to do it (Hao 2015), or not being able to record two visits simultaneously, particularly because it took too long to load the application for each new visit (Schoen 2017). There were also health workers who did not like typing on a mobile device (Schoen 2017). Other health workers felt that the devices at times impersonalised their interactions with clients when they required emotional support: [Home visitor]: "It's [tablet] cold…it's just her interacting with a machine. So there's no sympathy, there's no condolences. There's no, I want to say loving interaction. No, um, it's like no comfort, no support you know" (Schoen 2017). The infrequency with which these were raised, makes it unclear if these are widely-held perceptions, reflecting a more general or broader experience.
Finding 41: Health workers discussed challenges, beyond network and electricity issues, that sometimes were just an annoyance or a concern, but at other times also impeded their mHealth activities, and their ability to provide a service assisted by the use of mobile devices. These included damaged devices, loss and the of devices, having to carry two devices, not being able to readily buy phone credit when needed, not being able to send long messages because of character limitations, and the limitations of the language capabilities of their devices (moderate confidence in the evidence) Apart from challenges with network connectivity and electricity issues, health workers listed a number of other challenges they faced when using mobile devices. These challenges included lost and damaged devices (Murray 2015; Schoen 2017), working in highcrime areas and subsequently facing personal safety risks and risks of stolen devices (Chang 2011; Coetzee 2017; Schoen 2017; Toda 2017; Valaitis 2005). CHWs agreed that they were concerned about the safety and use of tablets in some of the communities and households they visited: "I was also afraid because of the places that I go to. The places that I go to criminals will be looking at me while they did not mind me before" (Coetzee 2017); workers' low proficiency with English and the unavailability of local language characters ( Cochrane Database of Systematic Reviews and work phone: "Many of the health workers complained about carrying two phones, the smartphone we gave them and their private phone. As a result, many of them stopped carrying their private phone and started using the smartphone as their primary phone" (Medhanyie 2015). Some workers complained when they lost work because it was not backed-up (Cherrington 2015). In one instance, not having a vendor to purchase phone credit from, was perceived as a barrier to use mobile devices (Hampshire 2016), and in another, not having someone to repair their devices (Murray 2015; Schoen 2017). When health workers in one intervention uploaded photos, videos and music (sometimes private), it slowed the technology and disabled the mHealth application that they had to use (van der Wal 2016). These challenges were not simply an annoyance or irritation, they also impeded health workers in their service delivery. For example, one health worker explained that she/he could not call an ambulance for a violent psychiatric patient because she/he did not have credit on their phone and all shops were closed, so no credit could be purchased (Hampshire 2016). Character limitations in text messages prevented some health workers from sending required lab results (Hao 2015). Device shortage in combination with network and Internet coverage made for di iculties in sending images, client information, and for seeking advice (Lodhia 2016).
Finding 42: Health workers complained when the tasks asked of them in mHealth interventions were felt to be beyond their clinical capacity, and when support from higher-level workers was absent (very low confidence in the evidence) Health workers found it di icult to communicate or explain information to the client provided to them via their mobile devices when this information was beyond their clinical capacity. Examples of this included receptionists who had to screen clients for cardiovascular illnesses and relay the results to the clients: "Receptionists were unsure how to respond to patients' questions and generally felt this duty was not part of their role. They did not see the relevance of screening for stroke prevention. Patients would say, "Is this my heart rate?" and I would say, "I don't really know"" (Orchard 2014). Another example was lay health workers who preferred to refer screened clients to a doctor to receive their screening results (Praveen 2014). This di iculty was also expressed when there was an absence of higher-level support in following up on the screening results (Praveen 2014).

Results of linking the review findings to the Cochrane intervention reviews
As described in the Methods section, our review was used alongside several Cochrane intervention Reviews on the e ectiveness of mHealth interventions, commissioned by the WHO to inform their guidelines on digital interventions for health system strengthening (WHO 2019). These reviews assessed the e ectiveness of diverse types of mobile devices used by health workers to improve their delivery of care. These included mobile devices for birth and death notification (Vasudevan 2018), stock notification (Agarwal 2018a), client to provider and provider to provider telemedicine (Gonçalves-Bradley 2018a; Gonçalves-Bradley 2018b, tracking of clients' health status (Agarwal 2018b), and decision support (Agarwal 2018c). The findings from our qualitative evidence synthesis were used as a source of information about intervention acceptability and feasibility, which the WHO's guideline panel used as a basis for their recommendations. The GRADE evidence-to-decision tables where this evidence is presented alongside evidence from the relevant Cochrane Reviews of e ectiveness are available in the Guidelines appendices (www.who.int/reproductivehealth/ publications/digital-interventions-health-system-strengthening).
In addition, we also planned to develop hypotheses about why some mHealth technologies are more e ective than others. A er an assessment of the results from the reviews of e ectiveness, as well as of our own review findings, we chose, however, not to develop specific hypotheses about individual mHealth technologies. One reason for this was that the preliminary findings of the reviews of e ectiveness (apps.who.int/iris/bitstream/handle/10665/311980/ WHO-RHR-19.10), showed large evidence gaps, while the evidence that was found was o en of low or very low certainty. While there was some evidence of impacts on health outcomes, the available evidence suggested that these types of interventions may make little or no di erence to the outcomes that were measured. While we do not currently have a clear picture of whether some mHealth technologies are more e ective than others, the findings from our own review suggest that it is unrealistic to expect consistent positive outcomes in mobile health programmes. This includes outcomes tied to the health of clients, to service delivery and to the organisation of care. The qualitative evidence in our own review illustrates how these programmes are comprised of many interlinking and at times complex, components, for instance, health system arrangements whereby the programme is being implemented. Each of these components o er di erent issues that may contribute to positive outcomes, but also pose definite challenges that render mHealth susceptible to poor outcomes. These facilitators and barriers relate to, amongst others, (i) the devices and technology itself, for example, short-life batteries and user-friendly so ware (Findings 32, 41); (ii) health workers' aptitude for mobile devices and their digital literacy (Findings 15,24,29); and (iii) upstream issues such as health system arrangements and high-level stewardship (Findings 34,35). Therefore, we consider the following two key hypotheses based on our findings, may help explain and understand the e ectiveness of mHealth programmes in delivering and supporting primary healthcare services.
There are self-evident benefits to mHealth, for example linking people living in rural areas to forms of health care that they would never otherwise have had access to (Findings 1,2,22), it eliminates inhibiting transport challenges (Finding 11), access to real-time data (Findings 5,10,14), and its portability allows health workers to access healthcare information at the point-ofcare (Findings 9,12). However, seeking and delivering health care, regardless of the mode, remains a relational transaction between a client and a health worker. Our findings show that all of the usual relationship issues that make health worker-client relationships successful or challenging, issues such as being trustworthy, being seen to o er high-quality care, being seen to be knowledgeable about the condition being treated, remain. However, with the addition of mHealth, this becomes complicated when for example, a health worker's trustworthiness may be determined by the sense of trustworthiness of the information they relay from the mobile device, or a health worker's status is elevated because they are seen as being worthy enough of being entrusted by their employer with an expensive device. The existing relationship complexities, therefore interact with the new relationship complexities brought by mHealth. The use of a mobile device as part of the transactional relationship between health workers and clients is therefore, but one element of a complex relationship, and understanding the Cochrane Database of Systematic Reviews eventual health outcome achieved, requires acknowledgement of this complexity, and acceptance that there may be factors which cannot be controlled for or easily explained. However, a health worker who is already regarded as trustworthy, and is able to show extra competence through use of a mobile device, is likely to be even more highly regarded, and the message they deliver through the use of the device may be regarded as more credible. The opposite however, is equally likely, a previously well-regarded health worker, using a faulty device, may lose face with clients, and so may begin to lose credibility, and struggle to convince clients to adopt the behaviour that they are suggesting. The outcome therefore is sensitive to the interaction between these multilayered, complex factors and interactions.
No matter the sophistication of the mobile health technology or programme, if it is implemented in a health system that has challenges, then the mHealth programme is also likely to be challenged. In those instances, it is the dysfunctional system, rather than the device, technology, or programme, which becomes the barrier to positive e ectiveness outcomes. Mobile health programmes are embedded in larger systems, and therefore impacted by contextual issues external to the technology itself, and how health workers use mobile devices. This was evident in health workers' reporting how health system and contextual factors impacted on their optimal use of the devices (Findings 3,16,34,35,39). Examples of these factors include network connectivity, access to electricity, sta shortages, unresponsive emergency services, inadequate supervision, and strained relationships between lowerand higher-level health workers. It is our contention that the reverse also holds true: mHealth can have a positive e ect when used within an already functional system and can sometimes close a specific gap in a system that is not fully functional.
In summary, the complexities of healthcare delivery and of human interactions defy dra ing simplistic hypotheses that can predict the e ectiveness of mHealth programmes to provide and support primary healthcare services. The e ectiveness of these programmes results from the interplay between technology, context, and the human attributes of clients and health workers.
Detailed programmatic and process description information and realist evaluations (Pawson 2001), alongside e ectiveness studies will be a starting point to unravel this interplay and formulate hypotheses regarding the e ectiveness of mobile health. Below, we o er a few recommendations regarding implementation practices that may improve the likelihood of positive outcomes when using mobile devices to provide and support the delivery of primary healthcare services.

Review author reflexivity
We described our initial positioning earlier (see 'Review author reflexivity' in the 'Methods' section above). Our views remained the same during the review, though our continued team discussions led to more nuanced definitions regarding mobile health, whether it is used to deliver or support healthcare services, and primary health care. Whilst writing the 'Discussion' and the 'Conclusions', we were particularly aware of the risk of overlooking data that refuted our own experiences that mobile health intervention outcomes are usually a mix of having positive and no e ects.

Summary of main findings
For a summary of the main findings, please see the 'Plain Language Summary'.

Overall completeness and applicability of the evidence
The majority of studies (74%) on which our findings are based, were conducted in low-or middle-income countries. Most of these were from Africa (71%) while only one study was from Latin America (Brazil). Of the 12 studies conducted in high-income countries, five were conducted in the USA, two in Canada, and one each in Australia, Ireland, Norway and Scotland. We downgraded our confidence in several of our findings because of the limited range of settings that the data were from.
Regarding the health issues addressed in the mobile health programmes, our data are a fair representation of the main services o ered in primary health care in the study countries. Seventeen studies focused on maternal, neonatal and child health, four addressed HIV/AIDS, and two, malaria and tropical diseases. Other health issues included, amongst others, cardiovascular diseases, intimate partner violence, eye care, hypertension, wound care, and mental health. Similarly, there was an even spread of studies reporting on lay and professional health workers' experiences and perceptions: 10 studies reported on lay health workers, 12 on health professionals, and in 17, both categories were reported on. The health worker category was unclear in the remaining four studies.
We would also argue that the included studies mirror by and large, current device and application use, with tablets and personal digital assistants (PDAs) less used compared to mobile phones, most likely due to costs and convenience. Tablets and PDAs were used in only 11 studies (26%), and mobile phones, identified as iPhones or smartphones, in 31 studies (72%). In one study, both tablets and mobile phones were used. Device applications ranged from text messaging, screening and diagnostic algorithms, and preloaded health promotion materials, to the recording of surveillance data and work scheduling.
Studies exploring healthcare workers' use of digital health strategies are now published increasingly o en. We have listed 85 studies under "Studies awaiting classification" that appear relevant for our review and that have been published since we finalised our analysis. In our next update of this review, we will sample from these studies, focusing in particular on those findings that we have downgraded using the GRADE-CERQual approach.

Comparison with other studies or reviews and implications for the field
A mixed-methods review by Konttila and colleagues (Konttila 2019), on key competencies required from healthcare professionals to use digital technology, and the organisational factors that shape their use of it, resonates with several of our findings. They reported that using digital devices impacted both positively and negatively on the health worker-client relationship (Findings 18,22); led to concerns amongst health workers that they might lose their clinical competencies (Finding 13); were influenced by the users' level of digital literacy (Finding 29); and required refresher training and technical support (Findings 30,31). Importantly, these authors echoed our views that digital technology is embedded within larger Cochrane Database of Systematic Reviews systems that can either advance or undermine the e ectiveness of mobile health programmes (Konttila 2019).
We concur with Dunn and colleagues, that primary research is needed to understand the pathways of how mobile health e ects change (Dunn 2018). It is clear from our findings that the e ectiveness of mobile devices cannot be separated from those who use it to deliver health care. Mobile health interventions are a combination of device and health worker, each with unique attributes, but also jointly, that impact clients' health behaviours, and how primary health care is supported and delivered. Though our review reported on health workers perceptions of how clients perceived them and the devices they use, asking clients themselves may further our understanding of the pathways of e ect. We are in agreement with Lee et al (Lee 2016), that clearer descriptions of the mobile health interventions may help to better understand its impact. Contrary to Marcolino's review of e ectiveness reviews (Marcolino 2018), where it was found that the majority of studies were conducted in high-income countries, 74% of the studies in our review were conducted in low-or middle-income countries. No obvious reason for the di erence was found, but it does suggest a need for a more even spread of studies conducted in low-, middleand high-income countries.
It appears from this review and others, that mobile health programmes are increasingly maturing from their 'pilotitis' status (Huang 2017), in so far as pilot studies that usually have small client participant numbers. Across the included studies, four had programmes serving between 101 to 1000 clients, 22 were implemented across multiple facilities, communities, and subdistricts, and seven of these were implemented in 20 or more sites. One study reported on a national mobile health programme. However, programmes remain relatively small in terms of health worker participants: only three studies had more than 100 worker participants, and 26 studies had between 11 and 100 workers. Mobile health programmes su er from 'pilotitis' with respect to sustaining it over time. Only two have been running for six years or longer, three between two and three years, and two studies were implemented between one and two years. The implementation period of the remaining studies was either unclear or shorter than one year. From our full-text screening, it is also clear that the number of qualitative studies reporting on mobile health programmes is increasing.
Our findings resonate with results from primary studies that did not meet our inclusion criteria, but which reported benefits and challenges similar to what we found, when using mobile devices to deliver or support primary healthcare services. This includes benefits, such as reducing travel time, and its related expenses, for health workers and clients alike ( Our synthesis identified issues that have been less reported in peerreviewed literature. These include: health workers advocating for political stewardship from ministries of health; workers reporting that it is more likely to turn intervention care into routine care when the mobile platform is integrated with other routine electronic systems; an acknowledgement that there is still a need for faceto-face interaction with colleagues and clients, and workers who o en use their personal devices without remuneration for its associated costs. Though not surprising, there were health workers who perceived some clients to be more responsive and taking ownership for their health, because their consultations included the use of mobile devices. The reasons o ered for this ranged from clients being concerned about the immediacy with which workers could report when they defaulted on treatment, to clients perceiving the devices as more trustworthy than standard care.

Limitations of the review
As we no longer had access to the Global Health Ovid database for the 2018 search, this could be considered a limitation of the review. The 2015 search yielded only 870 records from that database, and it could be expected to have yielded less in 2018, given the shorter time period. However, it remains that some eligible studies may have been missed.
Through our sampling, we excluded 10 studies because of serious methodological limitations, and it is possible that some of these may have contained data that could have added nuances to our findings and/or resulted in new findings. However, this concern is to an extent balanced out by the fact these additional studies would likely not have increased the confidence in existing findings because of their methodological limitations.

Implications for practice
Below are a set of questions that are drawn from the highand moderate-confidence findings in this review, and that may help implementation agencies, ministries of health, programme managers, and other stakeholders to plan, implement, or manage mobile health programmes.

Health systems questions
1. Will health workers be part of the planning, implementation, and evaluation processes of mobile health programmes? Will their views be sought, and their perspectives taken at each stage of the programme? 2. To what extent is political buy-in from health ministries required, and achieved, for the successful implementation of the mobile health programme? 3. Has a proper assessment been made on whether health workers' use of mobile devices is adding to or alleviating their workload? How will the extra workload that may occur, be accommodated for? 4. If your intervention is intended to improve e iciency and coordination, is the health system in which it is set prepared for the extra demands that this may imply? For example, if a health worker calls for an ambulance or for professional backup, will such support be available; if an mHealth screening intervention results in increased clients at facilities, will the existing capacity of facilities be able to handle the increased workload? If no preparation is in place for extra demands on the health system, have you engaged with those who may be required to provide additional services, so they can make preparations? 5. Do higher-level health workers have the time and means to respond when lower-level workers send them requests via mobile devices, and have lower-level workers' use of mobile Cochrane Database of Systematic Reviews devices been properly explained to all higher-level workers with whom they interact in delivering health care? 6. Does your intervention require health workers at the same or di erent levels of hierarchy, to interact with each other? If so, are these health workers prepared for, and willing for the changes that may arise as a result of this interaction, such as new forms of supervision and accountability, immediacy of contact, and telephonic request for advice? What needs to be done to better prepare these relationships for the anticipated changes in expectations on all parties as a result of mHealth?

Technical and infrastructural questions
1. Does your setting have the necessary infrastructural and technological capacity to support the level of sophistication intended by the intervention, for example: is there su icient electricity supply and electricity coverage, network capacity, technical support, and vendors to purchase phone credit or data for the level of intervention that you intend to implement? Have you considered how these might vary by region? 2. Are the devices being used in the intervention su iciently sophisticated for the level of intervention being planned, and are these devices replaceable or repairable within your setting? Have you considered who will repair them, and who will cover the costs? 3. When planning mHealth programmes, has the number of sta and clients who have access to mobile devices been taken into account? Are there strategies in place when clients change their mobile phone numbers? 4. Has adequate provision been made for health workers to have su icient phone credit and data, without having to use their own resources? 5. Is there a strategy to integrate the mobile health platform within existing electronic health information systems? Have you considered the requirements to ensure interoperability?
Questions about health worker training and skills 1. Has the programme management budgeted for adequate training of initial sta , refresher training and in-service training for new sta members? 2. What is the level of digital literacy amongst those health workers who will implement the intervention, as well as managers and supervisors who will support them? What further interventions are needed to ensure adequate skill levels are present at the beginning of the intervention and maintained over the course of the intervention? 3. Has the programme management identified 'champions' amongst the workers whom they can call upon to assist those struggling with the devices? 4. When the device allows the health worker to screen and diagnose clients, are they clinically equipped to respond appropriately to the results of the screening and diagnosing? Are they able to explain the results to the patient? 5. Is there a system in place to allow sta who dislike, or who are not su iciently digitally literate to use mobile devices, to continue with standard practice, such as a paper-based system for recording work?
Questions about sociocultural acceptance and equity 1. Has enough been done to raise community-and client-level awareness of the mobile health programme, and its implications for the services delivered by it? 2. What is the level of cultural acceptability of mHealth in the proposed setting? What is the existing level of trust between healthcare workers and the community? Have you considered that low levels of trust may be exacerbated by mHealth, for example fears about personal data? 3. What other interventions are needed to increase trust, enhance acceptability of mHealth, and reduce skepticism amongst recipient communities? 4. Are there specific social or geographical barriers which may interact with the intervention, such as women not being allowed access to phones? How might these be addressed in advance? 5. Have you considered how barriers to mHealth use may further increase inequity, and what other interventions are required to reduce these inequities?

Implications for future research
1. More studies are needed from high-income countries and lowand middle-income countries outside of Africa. 2. In general, researchers should aim for better reporting of their studies. This includes providing detailed information on: a. the contexts in which the mobile devices are used, as this is likely to shape the acceptability, feasibility, and e ectiveness of using mobile devices; b. their methods of sampling, data collection and analysis; and c. reflections on how the researchers' views and positions may have influenced the results. 3. Suggestions regarding how to report mHealth interventions can be found in the 'mHealth evidence reporting and assessment (mERA) checklist' (Agarwal 2016). 4. Researchers should give prominence to participants' voices in their studies, and present rich data, where important for a proper understanding of the phenomenon. 5. More qualitative research should be conducted alongside e ectiveness studies to explore the results of e ectiveness studies. We suggest that detailed programmatic and realist evaluations (Pawson 2001), become part of e ectiveness studies. 6. Though individual mobile health programmes may be implemented at scale regarding client participants, and to a lesser extent, health worker participants, more longitudinal research of these programmes are needed to assess the sustainable integration of mHealth into standard care.

A C K N O W L E D G E M E N T S
We gratefully acknowledge the following individuals and institutions who have played an important part in this review.
We are grateful to the following sta from the Norwegian Satellite of EPOC.
1. Claire Glenton supported the review from start to finish. Without this support to the novice review authors, the review would not have been possible. She contributed significantly to reviewing the findings and ensuring they were supported by the data.

Cochrane Database of Systematic Reviews
Modi 2015 Development and formative evaluation of an innovative mHealth intervention for improving coverage of community-based maternal, newborn and child health services in rural areas of India Serious concerns due to insufficient information on data analysis, author reflexivity, and poor data to support the study findings.
Jalloh-Vos 2013 Mobile health: connecting managers, service providers and clients in Bombali district, Sierra Leone Serious concerns due to insufficient information on sampling, data collection and analysis, and no reference to author reflexivity. It is also a serious concern not knowing if the cited data refer to mid-or end-intervention time points.

Shieshia 2014
Strengthening community health supply chain performance through an integrated approach: using mHealth technology and multilevel teams in Malawi Serious concerns due to insufficient information on participant demographics, sampling, data collection and analysis, and no reference to author reflexivity.
van Heerden 2017 App-supported promotion of child growth and development by community health workers in Kenya: feasibility and acceptability study Serious concerns due to insufficient information on data collection, and no reference to author reflexivity. There is insufficient data to support the study findings.  S45 S1 OR S2 OR S3 OR S4 OR S5 OR S6 OR S7 OR S8 OR S9 OR S10 OR S11 OR S12 OR S13 OR S14 OR S15 OR S16 OR S17 OR S18 OR S19 OR S20 (S1 OR S2 OR S3 OR S4 OR S5 OR S6 OR S7 OR S8 OR S9 OR S10 OR S11 OR S12 OR S13 OR S14 OR S15 OR S16 OR S17 OR S18 OR S19 OR S20 TI ( "portable media player" or "portable media players" ) OR AB ( "portable media player" or "portable media players" ) 7 S37 TI ( ipod or ipods or "i pod" or "i pods" ) OR AB ( ipod or ipods or "i pod" or "i pods" ) 151 S36 TI ( mp3player* or mp3 W0 player* or mp4player* or mp4 W0 player* ) OR AB ( mp3player* or mp3 W0 player* or mp4player* or mp4 W0 player* ) 63 S35 ( (handheld or "hand held") N3 console* ) OR ( (handheld or "hand held") N3 console* ) 3 S34 TI ( "computer tablet" or "computer tablets" or "pc tablet" or "pc tablets" or "palmtop computer" or "palmtop computers" or "palm top computer" or "palm top computers" or "pda computer" or "pda computers" or "pocket pc" or "pocket pcs" or "pda phone" or "pda phones" or blackberry or "palm pilot" or "palm pilots" or "pilot palm" or "pilot palms" ) OR AB ( "computer tablet" or "computer tablets" or "pc tablet" or "pc tablets" or Cochrane Database of Systematic Reviews "palmtop computer" or "palmtop computers" or "palm top computer" or "palm top computers" or "pda computer" or "pda computers" or "pocket pc" or "pocket pcs" or "pda phone" or "pda phones" or blackberry or "palm pilot" or "palm pilots" or "pilot palm" or "pilot palms" ) S33 ( (handheld or "hand held") W0 computer* ) OR ( (handheld or "hand held") W0 computer* ) 278 S32 TI ( "instant messaging" or "instant messenger" ) OR AB ( "instant messaging" or "instant messenger" ) 98 S31 TI ( whatsapp or "whats app" ) OR AB ( whatsapp or "whats app" ) 33 S30 TI "web messag*" OR AB "web messag*" 10 S29 TI ( mms W0 (messag* or service*) ) OR AB ( mms W0 (messag* or service*) ) 0 S28 TI ( ("multi media" or multimedia) W0 (messag* W0 service*) ) OR AB ( ("multi media" or multimedia) W0 (messag* W0 service*) ) 11 S27 TI ( sms W0 (messag* or service*) ) OR AB ( sms W0 (messag* or service*) ) 34 152 S19 TI ( smartphone* or smart W0 phone* ) OR AB ( smartphone* or smart W0 phone* ) 1,620 S18 TI ( "portable electronic" W0 (app or apps or application*) ) OR AB ( "portable electronic" W0 (app or apps or application*) ) 615 S17 TI ( (mobile or phone or telephone) W0 (app or apps or application*) ) OR AB ( (mobile or phone or telephone) W0 (app or apps or application*) ) 758 S16 TI ( (mobile or cellular) W0 (technology or technologies) ) OR AB ( (mobile or cellular) W0 (technology or technologies) ) 424 S15 TI (cell* W0 phone* or cellphone*) OR AB (cell* W0 phone* or cellphone*) 938 S14 TI ( mobile W0 phone* or mobile W0 telephone* or wireless W0 phone* or wireless W0 telephone* ) OR AB ( mobile W0 phone* or mobile W0 telephone* or wireless W0 phone* or wireless W0 telephone* ) inconsistency regarding methodological reporting in a few studies garding coherence equacy because the richness of the data are inconsistent across the studies evance because data from four studies are pilot studies with its associated focused support to participants which is not true to real life nor concerns regarding methodological limitations and relevance, and moderate concerns regarding adequacy

Finding 5
Some health workers relayed that mobile devices improved their reporting to supervisors and encouraged them to report more truthfully. Others compared mobile device-facilitated supervision to "big brother watching". Some supervisors thought that mobile devices allowed them to better identify sta who needed support. Moderate concerns regarding the methodological limitations because of inconsistent support of the included data for this finding No/very minor concerns regarding coherence Minor concerns regarding adequacy because some of the data are supported by one study only Minor concerns regarding relevance because the perceptions are mostly from lay health workers

Moderate confidence
Due to no/very minor concerns regarding coherence, minor concerns regarding relevance and adequacy, and moderate concerns regarding methodological limitations

Finding 6
Health workers had positive experiences with using instant messaging through WhatsApp. This application was seen as cheap and suitable for a range of activities, such as communicating with peers and posting photos as evidence of work done. Serious concerns regarding the methodological limitations because the study contributing most to the finding had a poor description of the context, sampling of participants, data collection and analysis.
No/very minor concerns regarding coherence Serious concerns regarding adequacy because the finding is supported by only three studies Moderate concerns regarding relevance because the finding is based on only three studies, with two of them being from Sub-Saharan Africa Very low confidence Due to serious concerns regarding methodological limitations and adequacy, moderate concerns regarding relevance, and no/very minor concerns regarding coherence

Finding 7
Even when health workers received messages that were automated, rather than sent direct- Health workers' perceptions and experiences of using mHealth technologies to deliver primary healthcare services: a qualitative evidence synthesis (Review) Copyright © 2020 The Authors. Cochrane Database of Systematic Reviews published by John Wiley & Sons, Ltd. on behalf of The Cochrane Collaboration.

106
ly from a manager or supervisor, this was still experienced and responded to, as a kind of supervision. Some lower-level health workers experienced it as supportive to their work, while others felt guilty for not providing correct care as per these messages.
search components are presented it is not described in sufficient detail garding coherence cause it is thin data cause most of the data relate to lay health workers only, mostly from one study concerns regarding methodological limitations, and no/very minor concerns regarding coherence

Finding 8
The task optimisation enabled through mHealth interventions was widely valued by health workers. Moderate concerns regarding the methodological limitations because in two supporting studies it was not clear how participants were selected, poor data collection description, and limited researcher reflexivity. In addition, we have only positive perceptions, which raises a concern about possible bias No/very minor concerns regarding coherence No/very minor concerns regarding adequacy No/very minor concerns regarding relevance

Moderate confidence
Due to no/very minor concerns regarding coherence, relevance, and adequacy, but moderate concerns regarding methodological limitations

Finding 9
At times, health workers used their mobile devices to access the Internet for health information, and found it useful when they were with clients who needed the information. This interaction also included health workers providing clients with additional information beyond the healthcare intervention. But, if the only way that health workers could access online information, required them to use their own money to purchase data, then this could be prohibitive to them accessing such information. Minor concerns regarding methodological limitations as the majority of studies had no to minor methodological limitations No/very minor concerns regarding coherence Serious concerns regarding adequacy because of very thin data Minor concerns regarding relevance due to a limited number of countries in which the studies were conducted Low confidence Due to no/very minor concerns regarding coherence, minor concerns regarding methodological limitations and relevance, and serious concerns regarding adequacy

107
mHealth held the promise of increasing service efficiency for many health workers, but the experience of whether this promise was borne out in practice, varied in the accounts of health workers. It was experienced as efficient if it improved feedback, speed and workflow, but inefficient when the technology was slow and time consuming. Some were concerned that if mHealth was too efficient, making work faster, that this may justify sta cutbacks.
10 of the studies were pilot studies, which could bias the perceptions given the intensified support that is standard in pilot studies herence given there were only two studies reporting negative perceptions garding adequacy garding relevance minor concerns regarding methodological limitations and coherence

Finding 11
Health workers frequently reported mobile devices as overcoming the difficulties of rural and geographically challenging contexts when it made it possible for them to provide health care without having to travel. Some reported that reducing travel time allowed them more time with their clients. Minor concerns regarding methodological limitations because more than half of the studies had poorly described data collection and analysis methods No/very minor concerns regarding coherence No/very minor concerns regarding adequacy Minor concerns regarding relevance as the finding is primarily applicable to only rural and geographically challenging contexts High confidence Due to no/very minor concerns regarding coherence and adequacy, and minor concerns regarding methodological limitations and relevance

Finding 12
Health workers appreciated the portability and work schedule flexibility of mobile devices. Moderate concerns regarding the methodological limitations because the majority of studies had insufficient, poorly described methods and data collection; the data in one study was hand recorded, and no No or very minor concerns regarding adequacy No or very minor concerns regarding adequacy No or very minor concerns regarding relevance

Moderate confidence
Due to no/very minor concerns regarding coherence, relevance, adequacy, but moderate concerns regarding methodological limitations Through mHealth, health workers were able to use treatment and screening algorithms that were loaded onto mobile devices. Their perceptions of using these electronic algorithms ranged from finding it easy and useful, to threatening their clinical competency, and an information overload. There were also some concerns that erroneous data entry may lead to wrong treatment guidance.
Minor concerns regarding methodological limitations given that the majority of studies had no or minor limitations No/very minor concerns regarding coherence No/very minor concerns regarding adequacy No/very minor concerns regarding relevance High confidence Due to no/very minor concerns regarding coherence, relevance, and adequacy, and minor concerns regarding methodological limitations

109
mHealth interventions sometimes required health workers to perform tasks that were peripheral to regular service delivery, such as registering clients onto the system. These more menial tasks were sometimes regarded as undermining to professional sta .

2015; Murray 2015;
Wol -Piggott 2018 ological limitations because two of the included studies had poor descriptions of the context, data collection and analysis methods garding coherence because part of the finding is not coherent across the supporting studies garding adequacy because of a limited number of studies with very thin data regarding relevance because of the limited number of settings in which the studies were conducted methodological limitations and adequacy, and moderate concerns regarding coherence and relevance

Finding 17
Some health workers experienced the use of mHealth as generating an extra workload when, for instance, it resulted in reaching more clients needing care, or having to maintain both a mobile health and paper system. Some workers disliked this, particularly when their superiors did not perceive their mobile health work as part of their job description. Others did not object to the additional work, yet others wanted to be remunerated.

112
health workers. Health workers described this as challenging for multiple reasons, including clients not having phones, changing their phone numbers regularly, not knowing how to use a phone, being a target of crime because of possession of the phone, and women being prohibited from accessing phones. Health workers suggested competitive pricing to increase clients' access to phones, and to issue clients with phones. 2014; Murray 2015; Tewari 2017; van der Wal 2016; Wol -Piggott 2018 one study had serious methodological limitations with inadequate descriptions of context, data collection and analysis methods garding coherence equacy because of the limited number of studies and thin data relevance because of the limited range of settings and health worker categories nor concerns regarding methodological limitations and adequacy, and moderate concerns regarding relevance

Finding 25
Health workers were ambivalent about interventions that required clients to use the health workers' mobile devices during consultations. Their optimism was tempered by concern that there was a loss of meaningful engagement with clients.

Bacchus 2016; Coetzee 2017
No/very minor concerns regarding methodological limitations No/very minor concerns regarding coherence Serious concerns regarding adequacy because the finding is based on only two studies Moderate concerns regarding relevance as none of the two studies were conducted in low-and lower-middle-income countries, and only reported lay health workers' perceptions Low confidence Due to serious concerns regarding adequacy, moderate concerns regarding relevance, and no/ very minor concerns regarding methodological limitations and coherence

Finding 26
Health workers reported that their access to mobile devices was beneficial to clients and communities who were too poor to own mobile phones.

Chang 2011; van der Wal 2016
Moderate concerns regarding the methodological limitations because of poorly described data collection and analysis methods in the one study contributing most of the data No or very minor concerns regarding coherence Serious concerns regarding adequacy because the finding is based on very thin data Serious concerns regarding relevance because the finding is based on only two studies, which also limits the study contexts Very low confidence Due to serious concerns regarding relevance and adequacy, moderate concerns regarding methodological limitations, and no/very minor concerns regarding coherence Moderate concerns regarding methodological limitations because half of the studies had moderate concerns of which three had serious methodological limitations. The limitations included poorly described data collection and analysis methods, and in one study there were concerns that the data collection method could have biased participant responses No/very minor concerns regarding coherence No/very minor concerns regarding adequacy No/very minor concerns regarding relevance

Moderate confidence
Due to no/very minor concerns regarding coherence, relevance, and adequacy, but moderate concerns regarding methodological limitations

116
ers experienced mHealth and the use of mobile devices for service delivery, in their different contexts.

Finding 36
It was important for health workers that mobile health interventions be integrated with other existing electronic health information systems. This interoperability made it more likely that mobile devices would be integrated into standard care practices, while the absence of integration frustrated health workers. Due to no/very minor concerns regarding methodological limitations and coherence, but moderate concerns regarding relevance and adequacy

Finding 37
Health workers offered programmatic and implementation recommendations to improve mobile health interventions. The most cited of these was that the interventions be expanded to other settings and services, beyond what they were using it for as described in the studies. Other recommendations included raising community awareness about mHealth programmes, being involved in developing programmes, and appointing a 'mobile health champion'. Workers also suggested that those collecting surveillance data, must be informed of how the data are used.