Clients’ perceptions and experiences of targeted digital communication accessible via mobile devices for reproductive, maternal, newborn, child, and adolescent health: a qualitative evidence synthesis

Abstract Background Governments and health systems are increasingly using mobile devices to communicate with patients and the public. Targeted digital client communication is when the health system transmits information to particular individuals or groups of people, based on their health or demographic status. Common types of targeted client communication are text messages that remind people to go to appointments or take their medicines. Other types include phone calls, interactive voice response, or multimedia messages that offer healthcare information, advice, monitoring, and support. Objectives To explore clients' perceptions and experiences of targeted digital communication via mobile devices on topics related to reproductive, maternal, newborn, child, or adolescent health (RMNCAH). Search methods We searched MEDLINE (OvidSP), MEDLINE In‐Process & Other Non‐Indexed Citations (OvidSP), Embase (Ovid), World Health Organization Global Health Library, and POPLINE databases for eligible studies from inception to 3‐6 July 2017 dependant on the database (See appendix 2). Selection criteria We included studies that used qualitative methods for data collection and analysis; that explored clients' perceptions and experiences of targeted digital communication via mobile device in the areas of RMNCAH; and were from any setting globally. Data collection and analysis We used maximum variation purposive sampling for data synthesis, employing a three‐step sampling frame. We conducted a framework thematic analysis using the Supporting the Use of Research Evidence (SURE) framework as our starting point. We assessed our confidence in the findings using the GRADE‐CERQual (Confidence in the Evidence from Reviews of Qualitative research) approach. We used a matrix approach to explore whether potential implementation barriers identified in our synthesis had been addressed in the trials included in the related Cochrane Reviews of effectiveness. Main results We included 35 studies, from a wide range of countries on six continents. Nineteen studies were conducted in low‐ and middle‐income settings and sixteen in high‐income settings. Some of the studies explored the views of people who had experienced the interventions, whereas others were hypothetical in nature, asking what people felt they would like from a digital health intervention. The studies covered a range of digital targeted client communication, for example medication or appointment reminders, prenatal health information, support for smoking cessation while pregnant, or general sexual health information. Our synthesis showed that clients' experiences of these types of programmes were mixed. Some felt that these programmes provided them with feelings of support and connectedness, as they felt that someone was taking the time to send them messages (moderate confidence in the evidence). They also described sharing the messages with their friends and family (moderate confidence). However, clients also pointed to problems when using these programmes. Some clients had poor access to cell networks and to the internet (high confidence). Others had no phone, had lost or broken their phone, could not afford airtime, or had changed their phone number (moderate confidence). Some clients, particularly women and teenagers, had their access to phones controlled by others (moderate confidence). The cost of messages could also be a problem, and many thought that messages should be free of charge (high confidence). Language issues as well as skills in reading, writing, and using mobile phones could also be a problem (moderate confidence). Clients dealing with stigmatised or personal health conditions such as HIV, family planning, or abortion care were also concerned about privacy and confidentiality (high confidence). Some clients suggested strategies to deal with these issues, such as using neutral language and tailoring the content, timing, and frequency of messages (high confidence). Clients wanted messages at a time and frequency that was convenient for them (moderate confidence). They had preferences for different delivery channels (e.g. short message service (SMS) or interactive voice response) (moderate confidence). They also had preferences about message content, including new knowledge, reminders, solutions, and suggestions about health issues (moderate confidence). Clients' views about who sent the digital health communication could influence their views of the programme (moderate confidence). For an overview of the findings and our confidence in the evidence, please see the 'Summary of qualitative findings' tables. Our matrix shows that many of the trials assessing these types of programmes did not try to address the problems we identified, although this may have been a reporting issue. Authors' conclusions Our synthesis identified several factors that can influence the successful implementation of targeted client communication programmes using mobile devices. These include barriers to use that have equity implications. Programme planners should take these factors into account when designing and implementing programmes. Future trial authors also need to actively address these factors and to report their efforts in their trial publications.


Description of the topic of interest
Within the field of digital health, there are a variety of ways digital technologies may be used for public health purposes. This review focused on digital targeted client communication (DTCC). Digital targeted client communication may be used to transmit health event alerts to specific population groups; deliver health information based on a known health status or demographic; alert and remind about a particular health behaviour; or transmit diagnostic results to clients (WHO 2018). Targeted communication can also be further customised according to an individual's specific needs, resulting in 'tailored client communication,' whereby message content, timing, and frequency are matched to the needs and preferences of an individual (Hawkins 2008). The communication can be unidirectional and bidirectional, but initial contact is from the health system, as opposed to on-demand information service and telemedicine, where the client initiates the first contact with the health system (WHO 2018). The purpose of the DTCC would be to improve health and well-being, healthcare services, and/or the functioning of the health system. Typical interventions include sending brief text messages as a reminder to adhere to health visits and medical treatment, to provide clients with health information, to monitor their progress, and/or to provide medical advice and support.

Why it is important to do this review
Through the World Health Assembly Resolution on Digital Health, Ministries of Health recognised that digital technologies can potentially bring value to the health system, but called for a better understanding of best practices and the promotion of evidence-based digital health interventions and standards (WHO 2018a). This resolution also highlighted the need to ensure that "digital health solutions complement and enhance existing health service delivery models, strengthen integrated, people-centred health services and contribute to improved population health, and health equity, including gender equality" and noted the lack of evidence on the impact of digital health in these respects (WHO May 2018).
To address this need, the World Health Organization (WHO) embarked on developing evidence-based guidelines to inform government-led investments in digital health interventions for health system strengthening, including mechanisms to bolster access to reproductive, maternal, newborn, child, or adolescent health (RMNC-AH) services. This qualitative evidence synthesis is among a series of systematic reviews informing the WHO guidelines on digital interventions for health system strengthening. The scope of this synthesis reflects the WHO's assessment of global intervention priorities in this area. In addition to contributing to the WHO guideline, the findings of this review will be of interest more generally to programme planners and policymakers when deciding if and how to implement DTCC via mobile device in their setting. This review will complement the two WHO-commissioned reviews that focus on the effectiveness of targeted digital communication via mobile device (Palmer Ongoing a, Palmer Ongoing).

O B J E C T I V E S
To explore clients' perceptions and experiences of targeted digital communication via mobile devices on topics related to reproductive, maternal, newborn, child, or adolescent health (RMNCAH).

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

Topic of interest
We focused on clients' perceptions and experiences of digital targeted client communication (DTCC) via mobile devices in the areas of reproductive, maternal, newborn, child, or adolescent health (RMNCAH).

Types of studies
We included primary studies that used qualitative study designs such as ethnography, phenomenology, case studies, and grounded theory as well as qualitative process evaluations. We included primary studies that used qualitative methods for data collection (e.g. individual interviews, focus group discussions, diaries, document analysis, open-ended survey questions, and observation) and that used qualitative methods for data analysis (e.g. thematic analysis, framework analysis or grounded theory). We excluded primary studies that collected data using qualitative methods but did not perform a qualitative analysis (e.g. open-ended survey questions where the responses are analysed using descriptive statistics only). We included mixed-methods studies when it was possible to extract data that were collected and analysed using qualitative methods. We included studies regardless of whether they had been carried out alongside studies of the effectiveness of digital health interventions.

Types of interventions
We included studies exploring clients' experiences and perceptions of targeted digital communication (e.g. text messages and interactive voice response) accessible via mobile devices (see Table 1). This could include perceptions and experiences of the content of the message, the delivery mechanism itself, the sender, or other aspects tied to this form of communication.
We defined 'digital targeted client communication' (DTCC) as the transmission of targeted health content to a specified population, or to individuals within a predefined health or demographic group. This transmitted information can fall along a continuum of tailored (personalised to an individual person's condition) to standard, general, untailored communication. It can include the transmission of individualised notifications according to a specific individual's clinical care plan as well as the transmission of predetermined content developed for the identified population group (Hawkins 2008). Eligible individuals need to be identified and subscribed into a system that allows the transmission of the health information via digital device to a number they have requested. Additionally, the timing and content of the transmitted information should be determined by the health system, and not by a client seeking information ondemand.

Examples of targeted client communication could include:
• providing targeted health education, promotion, or information to clients based on known health or demographic characteristics; • providing alerts, notifications, and reminders to a client based on a clinical care plan or protocol, such as in the case of medication adherence and appointments to see a healthcare provider.
In contrast, untargeted client communication is the transmission of health promotion content to the general population or an undefined target population.
By mobile devices, we mean mobile phones or handheld mobile devices of any kind (but not analogue landline telephones), as well as tablets, personal digital assistants, and smartphones that facilitate communication to a targeted group of clients via different channels including short message service (SMS), voice, interactive voice response, multimedia messages, and social media when used for instant messaging purposes. For a specific list of included and excluded types of delivery mechanisms see Table 1 below. We included communication that was one-way (e.g. triggered by a system to the defined population groups) or two-way (e.g. allows for discussion or question and answer between the targeted population and the health system). Two-way or bidrectional communication was included if the first communication was initiated by the health system or healthcare provider to a client's mobile device. Studies of bidirectional communication initiated by clients to contact the health system were included in another review related to telemedicine and client-to-provider consultations (Gonçalves-Bradley 2018, WHO 2019.
We included studies where the digital component of the intervention was delivered as part of a wider package, or if we judged it to be the major component of the intervention. The focus of the study needed to be on one of the intervention areas listed in Table 1.

Types of participants
The review focused on the following population groups as defined in relation to the WHO guideline for which this review was commissioned (WHO 2019).
We included studies that focused on the perceptions and experiences of clients. We define clients as "an individual who is a potential or current user of health services; may also be referred to as patient or non-patient who uses health information and services" (WHO 2019).
We included studies that focused on the perceptions and experiences of clients in one or more of the following groups.
• Adolescent and youth populations (ages 10 to 24 years) that were users/potential users of sexual and reproductive health (SRH) services. Studies that included other population groups were included if participants' age had been disaggregated or where it was explicitly mentioned that a minimum of 70% of participants were between the ages of 10 and 24 years. • Adult users/potential users of SRH (age 18+). Studies that explicitly stated that they also included population groups under 18 years of age were included where it was explicitly mentioned that a minimum of 70% of the participants were above the age of 18 years.

Search methods for identification of studies Electronic searches
Information Specialist John Eyers developed the search strategies in consultation with the review authors. We searched the following electronic databases for eligible studies between 3 and 6 July 2017, dependant on the data base (see Appendix 2).
• MEDLINE (OvidSP) • MEDLINE In-Process & Other Non-Indexed Citations (OvidSP) • Embase (Ovid) • World Health Organization Global Health Library • POPLINE Using guidelines developed by the Cochrane Qualitative Research Methods Group for searching for qualitative evidence (Booth 2011, Harris 2018, Noyes 2015, as well as modified versions of the search for the associated or 'sister' effectiveness reviews (Palmer Ongoing a, Palmer Ongoing), we developed search strategies for each database. There were no language or geographic restrictions on the search. We used 1993 as the cut-o date for the search, as the first commercial SMS message was sent in December 1992. A similar approach was taken in the related effectiveness reviews (Palmer Ongoing a, Palmer Ongoing).

Searching other resources
We asked the Guideline Development Group network for the WHO guideline on Digital Health Guidelines for Health System Strengthening to identify and send in any studies that fit the inclusion criteria on 28 June 2017 (WHO 2019).
We sent a public call for papers to global listservs, including Global Digital Health Network and Implementing Best Practices (IBP).We handsearched the database www.mHealthEvidence.org for any studies that met our inclusion criteria on 17 August 2017. This database is designed to bring together literature on digital health from a global perspective to help stakeholders quickly access up-to-date, relevant evidence.
We searched PubMed for all studies linked to the trials included in the related effectiveness reviews in September and October 2018.

Selection of studies
We collated records identified from different sources into Covidence, a systematic review screening tool (Covidence). We identified duplicates and removed them. Three review authors then independently assessed the titles and abstracts of the identified records to determine potential eligibility, discarding those that were clearly irrelevant to the topic. Review authors HA and EA screened all titles and abstracts, and TT resolved any conflicts.
We obtained the the full text of all the papers identified as potentially relevant, and two review authors (TT, EA or HA) independently assessed these for inclusion in the review. NL resolved disagreements. See Characteristics of excluded studies for a list of the excluded studies and the main reasons for exclusion.

Library
Trusted evidence. Informed decisions. Better health.

Translation of studies in languages other than English
Although we searched for languages spoken by at least one member of the review team (French, English, Scandinavian languages), all of the identified or included studies were in the English language.

Sampling of studies
As qualitative evidence synthesis aims for variation in concepts rather than an exhaustive sample, and because large numbers of studies can impair the quality of the analysis, we purposefully sampled from the 52 articles that met our inclusion criteria.
We developed a sampling frame that took into consideration the population group, data richness, and closeness of the study data to the review objective.
Firstly, we divided the studies that met our inclusion criteria by client group as listed above in the inclusion criteria. As there were a limited number of included studies for pregnant and postpartum women (up to six weeks) (seven studies) and for pregnant and postpartum women (up to six weeks) living with HIV (two studies), all studies were included.
Secondly, we assessed the included studies within each client group for data richness, using a scale of 1 to 5 (see Appendix 3), and also looked at how closely the data from the study matched the review objectives. Studies with 'thin data' or that did not provide a close match to our review objective were not sampled.
In total, we sampled 35 studies to be included in the analysis (see Table 2 below).

Data extraction
We performed data extraction using a data extraction form designed specifically for this synthesis. We used the form to extract key themes and categories relevant to the synthesis objective using the Supporting the Use of Research Evidence (SURE) framework (SURE Collaboration 2011). We used a second form to extract information about first author, date of publication, language, country of study, context (urban, rural), and participant group to which the intervention was directed. We also extracted information on research method and if theoretical or conceptual frameworks were used.
HA extracted data from all the sampled studies. EA double-checked the data extraction and verified that all relevant data were extracted.

Assessment of the methodological limitations of included studies
To assess the methodological quality of the included studies, we applied a quality appraisal framework to each study. We used an adaptation of the Critical Appraisal Skills Programme (CASP) quality assessment tool for qualitative studies (CASP 2018). Other reviews of qualitative evidence have also used this tool (Ames 2017; Glenton 2013; Gopinathan 2014; Lewin 2010). The adapted tool that we used included the following eight questions.
1. Are the setting/s and context described adequately? 2. Is the sampling strategy described, and is this appropriate? 3. Is the data collection strategy described and justified? 4. Is the data analysis described, and is this appropriate? 5. Are the claims made/findings supported by sufficient evidence? 6. Is there evidence of reflexivity? 7. Does the study demonstrate sensitivity to ethical concerns? 8. Any other concerns?
HA conducted the initial assessment, and NL and TT reviewed the assessments. We accept that there is no 'gold standard' approach for assessing the methodological quality of primary qualitative studies, but believe that this adapted CASP checklist fit our needs in the context of this synthesis.
We did not exclude any studies based on our assessment of methodological limitations, but used this information to assess our confidence in the synthesis findings, as part of the GRADE-CERQual (Confidence in the Evidence from Reviews of Qualitative research) approach (Lewin 2018).

Data management and synthesis
For our synthesis, we first grouped articles according to client group as defined above. However, since there were only two articles focusing on pregnant and postpartum women living with HIV, we combined these with the other studies focusing on pregnant and postpartum women.
We conducted an initial framework analysis using the SURE framework to identify themes in the data. We did this within each of the population groups and then looked across population groups. The SURE framework has been used as an analysis framework in other studies and reviews (Glenton 2013;Glenton 2016;Gopinathan 2014;Lewin 2010;Muloliwa Forthcoming;Oku 2017). We used the headings and subheadings from the SURE framework as a starting point for the analysis and then adapted them through an iterative process. Next, within each section of the framework we did a thematic analysis of the extracted data to identify our synthesis findings. For example, data included in the framework under 'health systems constraints -accessibility of care' were thematically synthesised, and findings around access to digital devices and interventions were identified (see Findings 5 to 9 in the Results). Another example is around the framework area of knowledge and skills. Here we adapted the category to look at knowledge and skills in relationship to using a mobile device. A final example of adaptation is under health systems constraints, relationships with norms and standards. We adapted this category to address issues related to privacy and confidentiality. Some areas of the framework were le empty and discarded. Once findings were identified, HA read through all of the sampled studies again to double-check data extraction. We also went through the findings and identified those where the contributing studies were only/predominantly from high-income (HIC) or LMIC settings. The same was done for the different client groups. We have indicated this in the detailed description of the relevant findings.
We then thematically analysed the 25 identified findings in order to group them into six related overarching categories to provide a narrative for the Findings section. Some categories reflect those within the SURE framework, whereas others have been reorganised to address different issues raised by clients. The six overarching categories related to the general acceptability of and preferences around DTCC; the varying degrees of access to network services, phones, and messages; communication delivery and format preferences; communication content preferences; privacy and confidentiality regarding personal health information; and the perceptions of intervention impact.
To create the summary of findings for the Abstract we took all of the findings with moderate or high confidence in the evidence and worked them together into a clear story line.

Appraisal of confidence in the review findings
Four review authors (HA, CG, SL, NL) used GRADE-CERQual to assess the confidence that can be placed in each review finding (Lewin 2018). Each finding was assessed by at least two review authors.
The GRADE-CERQual approach assesses confidence in the evidence based on the following four components (Lewin 2018a).
1. 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 that data. 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. 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 phenomenon of interest (perspective or population, context, setting) specified in the review question.
After assessing each of the four components, we made a judgement about the overall confidence in the evidence supporting each review finding. We judged confidence as high, moderate, low, or very low. The final assessment was based on consensus among the review authors. All findings started as high confidence and were then graded down if there were important concerns regarding any of the CERQual components (Lewin 2018).

'Summary of qualitative findings' tables
We concluded the appraisal of confidence in each review finding by drafting a 'Summary of qualitative findings' table that presents the findings and our assessment of confidence in these findings, as well as an explanation of this assessment, based on the GRADE-CERQual approach.

Supplementing the related Cochrane e ectiveness reviews with synthesised qualitative findings
We explored how the findings from our synthesis related to, and could help to inform, the findings of the two related Cochrane reviews of effectiveness of DTCC (Palmer Ongoing a; Palmer Ongoing).
To do this we utilised a matrix approach similar to the one used previously by Candy 2011, Ames 2017, and Munabi-Babigumira 2017. This approach has also been described by Harden 2018. Our matrix explored whether potential implementation barriers that we identified in our synthesis had been addressed in the programmes evaluated in the related reviews of effectiveness.
To create the matrix we undertook the following steps: firstly, we selected the synthesis findings that we had assessed as having high or moderate confidence and that presented potential barriers to the implementation of targeted client communication programmes. Secondly, we created 10 questions reflecting these potential barriers, and placed these in a table. Finally, we assessed whether any attempt had been made to address these implementation barriers in the trials that were included in the two related Cochrane Reviews of effectiveness.
To carry out this assessment, we examined the publications included in the two Cochrane Reviews of effectiveness (Palmer Ongoing a; Palmer Ongoing). We also performed a further search for additional publications that could be related to the trials. We did this by (1) examining the reference lists of the main trial publication; and (2) searching for each trial in PubMed, and doing an advanced search for 'Similar articles'. The advanced search for 'Similar articles' used the first author of the trial to identify possible related studies that had this author as a co-author, and selected any that appeared to be related to the trial.

Researchers' reflexivity
Within qualitative research, researchers are expected to reflect on their own background and position, and how it will affect the design, analysis, and reporting of their research. Throughout the data synthesis, the authors were aware of their own positions and reflected on how these could influence the data synthesis and study design. Several of the authors have both primary and evidence synthesis research experience in digital health (reporting positive, negative, and neutral findings), and they considered themselves to be agnostic as to the outcome of this evidence synthesis.

Included studies
We screened 9531 abstracts and assessed 142 full-text articles. Fi y-two studies met our inclusion criteria. From these 52 studies, we sampled 35 studies for analysis ( Figure 1). The 17 studies that met the inclusion criteria but were not sampled into the synthesis can be found in Table 1. All of the sampled studies were published between 2009 and 2017. All of the included studies were published in English. Sixteen of the sampled studies were from high-income countries: Australia (1), Canada (2), the UK (4), and the USA (9). Nineteen of the sampled studies were from low-or middle-income countries: Cambodia (1), Cameroon (2), Ghana (1), India (2), Kenya (2), Lesotho (1), Nigeria (2), Peru (3), Sierra Leone (1), South Africa (2), and Uganda (2).
Client populations were adolescents and youth (12 studies); adult users/potential users of reproductive health services (10); pregnant and postpartum women (including those living with HIV) (9); and parents and caregivers of children under five years of age (4).

Methodological limitations of the included qualitative studies
There was poor reporting of the participant voice in some of the included studies. For example, many studies included limited firstorder constructs or data extracts, and these were often not labelled with an identifier of the participant. We also found poor reporting of researcher reflexivity across many of the studies, which limited transparency regarding the role of the researcher. All studies gave some description, even if very brief, of the context, participants, sampling, methods, and analysis.

Confidence in the findings
Based on our CERQual assessments, we had high confidence in four findings and moderate confidence in nine findings, indicating that the studies were a good representation of the phenomenon of interest. We had several findings where we had low (nine) or very low confidence (three), indicating that the studies were a weaker fit with the representation of the phenomenon of interest. Our main concerns were connected to the methodological limitations of the studies and the relevance and adequacy of the data. Common methodological limitations included a lack of researcher reflexivity as well as poor reporting of ethical considerations, sampling, and representation of the participant's voice in the findings. The data were often assessed as being only partially relevant, mainly because the included studies represented few regions; had a focus on a certain target population (e.g. youth) or a specific topic (HIV/AIDS); or because many of the included studies explored participants' perceptions of hypothetical situations or digital health Cochrane Database of Systematic Reviews interventions, or both. Finally, our concerns about adequacy were mainly tied to the limited number of studies included in some findings and the thinness of the data contributing to some findings.
The GRADE-CERQual evidence profile tables supporting the assessment of confidence in each finding can be found in Appendix 4. We start each section of the findings with a link to the 'CERQual summary of qualitative findings' table where a summary assessment of the findings from that section is presented.

Findings and categories identified in the data
From our synthesis, we developed a set of individual findings, and then organised these findings into six overarching categories related to (1) the general acceptability of and preferences around digital health interventions; (2) the varying degrees of access to network services, phones, and messages; (3) communication delivery and format preferences; (4) communication content preferences; (5) privacy and confidentiality regarding personal health information; and (6) perceptions of intervention impact. Unless specifically addressed in the detailed finding, the data were not specific to HIC or LMIC setting or to a specific client group. For a description of the context and client group in each study contributing to a finding, please refer to the evidence profiles in Appendix 4.

General acceptability of and preferences around digital targeted client communication
('Summary of qualitative findings' table for Findings 1 to 4 is shown in Table 2.) Finding 1: Overall, participants had a range of views regarding acceptance of the idea of receiving health information through their mobile devices. This was due to factors such as familiarity with the technology, convenience, control, being able to save and re-read messages later, cost, seeing it as a simple way of providing a reminder for medication or appointments, and the sense that someone was thinking about them and cared enough to send a message (low confidence in the evidence).
Many studies from a variety of contexts and client groups presented data related to the range of participants' views regarding the acceptance of DTCC and the factors that influenced this acceptance ( . Some participants felt that it was more personal than other methods of delivering health information such as posters and billboards (Gold 2010). Others perceived it as a way of boosting already-existing interventions or curricula, such as school-based HIV curricula (Cornelius 2009).
However, some participants felt that DTCC delivered via mobile device were not acceptable for their setting and chose not to partici-pate in the interventions. An example of this was seen in a family planning intervention in Sierra Leone where some husbands had a problem with their partner being called by the healthcare provider or their partner participating in the family planning intervention, or both (Jalloh-Vos 2014). Others felt that in some situations it was important to still have the opportunity to speak to a person on the phone or in person (face-to-face), for example if they were having a strong craving for a cigarette and needed support in that moment (Naughton 2013). Some participants felt that DTCC would be useful for certain population groups such as younger audiences Willoughby 2017;Wright 2011), youth just starting on HIV medication, patients not adhering to medication, and those with less education .
Familiarity with technology, especially SMS, was one reason that participants put forward for accepting digital health interventions (Brown 2014;Cornelius 2009;Jennings 2013;Smillie 2014). Youth and pregnant women described using digital technology as something they already did frequently, that fit their learning styles, and was not a foreign approach (Brown 2014;Cornelius 2009;Munro 2017). Others mentioned that they were already using SMS to request healthcare providers to call them back or to set up health appointments (Jennings 2013).
Participants also noted that messages delivered through digital mechanisms were very convenient and in some cases cost-effective. Some felt that they would be very beneficial for families who lived a long distance from the health facility and would save them time and money (Calderón 2017). Messages were perceived as being quick and much easier than going to the health facility to get a pamphlet, going to a doctor's appointment ( Participants liked that they could save and re-read messages as well as have control over receiving, keeping, or deleting the information (Brown 2014;Evans 2016;Munro 2017

Cochrane Database of Systematic Reviews
Finding 2: In discussing the pros and cons of DTCC compared to in-person meetings with a healthcare provider, some participants perceived interacting with a healthcare provider as preferable, warmer, and something to which they were accustomed. Others also felt that people could receive a faster response using digital communication and that the messages were more convenient and less judgemental. However, some liked having direct access to both healthcare providers and DTCC (very low confidence).
A few studies, from both LMIC and HIC contexts (Calderón 2017; Nachega 2016; Naughton 2013; Sloan 2017; Smillie 2014), described a range of participants' preferences for digital health interventions compared to in-person visits to healthcare providers. The majority of studies in this finding looked at the perspectives of pregnant and postpartum women and parents. Some clients liked having direct access to both healthcare providers and to digital health interventions, as each played a different role (Naughton 2013; Sloan 2017; Smillie 2014). Some felt that the digital health interventions were more convenient, reliable, flexible, and faster and provided more frequent support (Nachega 2016; Naughton 2013; Sloan 2017; Smillie 2014). Clients who were pregnant and trying to quit smoking often preferred the SMS interventions, as they felt healthcare providers judged them and made them feel uncomfortable (Sloan 2017). Clients in some studies liked the digital health interventions but still felt it was important to have access to in-person visits with healthcare providers or speaking with someone when needed (Calderón 2017; Naughton 2013; Smillie 2014).
Finding 3: Participants said that they liked two-way digital communication as this allowed them to engage directly with a healthcare provider, which they trusted more; to receive answers to their questions and have opportunities for discussion; and to receive a more immediate response. However, some participants felt that for some topics they would feel uncomfortable talking to a healthcare provider through a digital channel, due to issues related to shyness and privacy, and would prefer to use SMS (very low confidence).
Some studies from both LMIC and HIC contexts found that participants wanted or liked to have the option of engaging directly with healthcare providers through DTCC in order to receive answers to their questions ( In general, participants felt that these types of two-way communication options would be useful and provide them with answers to their questions when they needed them, as well as allow them to maintain contact with their healthcare providers in between appointments if questions or concerns were to arise (Rana 2015). One participant in one study stated that two-way communication would not be acceptable, as the person on the other end would then be informed of his HIV status .
Some participants preferred voice calls for engaging with healthcare providers above communicating with them through SMS (Akinfaderin-Agarau 2012). There were different reasons for this. Some participants felt that they could ask detailed questions and receive detailed answers as well as discuss the various problems or challenges they were facing (Akinfaderin-Agarau 2012; Rodrigues 2015), or that the service could be more trusted because they spoke to someone directly (Akinfaderin-Agarau 2012). Others thought that it would provide more opportunity for discussion and ensure that the message was well received by the intended recipient, and that they could receive an immediate response (Jennings 2013).
However, some participants explained that they preferred SMS services, as they would allow the participants to be more open and to ask about issues they would be too shy to bring up when speaking directly with someone (Akinfaderin-Agarau 2012; Smillie 2014). Some participants felt that they would not feel comfortable talking to someone in person, especially if that person was new or unknown to them (Smillie 2014). SMS was also viewed as advantageous for brief and relatively confidential receipt of information (Jennings 2013).
Finding 4: Some participants expressed a concern that some people might view digital targeted communication from healthcare providers as a replacement to seeking appropriate medical assistance, which might have adverse impacts. While some saw digital health as a way to increase access to care, others noted that text messaging might be seen by poorer people as a cheaper or sufficient healthcare option, which might decrease appropriate health-seeking behaviour (very low confidence).
Participants in one study from the USA exploring college students' views on receiving SMS for sexual health promotion expressed concern that other people might become over-reliant on digital health interventions because they were seen as a cheaper option than going to the doctor (Willoughby 2017). There was a worry that people would use the digital communication intervention instead of seeking appropriate medical attention. The participants thought that this could especially be the case for people with few resources.

Varying degrees of access to network services, phones, and messages
('Summary of qualitative findings' table for Findings 5 to 9 is shown in Table 3.) Finding 5: Participants reported varying degrees of access to network services, including cell networks (for calls and SMS) and the internet. In addition, some participants had poor access to electricity to charge their phones. These factors were reported to be barriers to using the DTCC (high confidence).
Studies from a range of income settings found that issues related to network services and electricity acted as a barrier to people's use of DTCC ( For instance, when network coverage was poor, some participants in a Nigerian study recommended that SMS was the best option as they were more likely to be transmitted when the network was unstable, whereas voice calls would not connect or would be dropped (Akinfaderin-Agarau 2012). Participants in a study from Canada described living in mountainous areas with no network coverage (Smillie 2014). Participants in low-income settings also mentioned that not being able to charge their phone due to power outages or lack of access to electricity was a barrier to participating in digital For some participants, their access to a phone was controlled by others. This could be because they could not afford to purchase a phone themselves (Jalloh-Vos 2014), or because physical access to the mobile phone was controlled by another person (see Finding 18) (Akinfaderin-Agarau 2012; Jalloh-Vos 2014; Rana 2015). This group was mainly comprised of women and adolescents, and this is discussed further in Findings 7 and 18.
Finding 7: Some participants, particularly women and adolescents, had their access to phones controlled or restricted by others, especially if they had to share or borrow a phone. They noted that they would often have to explain why they wanted to use the phone, and who they wanted to call, to allay suspicions about this communication. They mentioned that this was a barrier to accessing DTCC and made it difficult to keep their messages private (moderate confidence).
A few studies from LMIC settings in Africa found that some participants, particularly women and adolescents, had their access to phones controlled or restricted by others, especially if they had to share or borrow a phone (Akinfaderin-Agarau 2012; Flax 2017; Jalloh-Vos 2014; Rana 2015). They mentioned that this was a barrier to accessing DTCC and made it difficult to keep their messages private. In some contexts, women had their mobile phone use controlled by their husbands or other family members. The women would often have to explain why they wanted to use the phone and who they wanted to call (Akinfaderin-Agarau 2012; Jalloh-Vos 2014; Rana 2015), for example to allay suspicion that they were talking to their boyfriends or having an affair. For some women, this would mean having to find an alternative phone to use if they did not want their husband to know they were using a digital health service. For example, in one study, some women did not want their husbands knowing they were receiving information on family planning (Jalloh-Vos 2014). In some settings, women and girls were also viewed by their society as not having time to use phones due to greater domestic obligations than their male counterparts (Akinfaderin-Agarau 2012). Youth in one study also reported facing restrictions related to using phones at school (Rana 2015).
In one study, women in a women's group all shared a single phone. They elected one group member to control the phone and share the messages. This group member was then responsible for distributing the messages from the DTCC. In this context, the majority of the participants accepted this form of phone sharing and believed it was functional (Flax 2017).
In all studies contributing to this finding, participants felt that sharing a phone or having access to their phone controlled by someone else delayed the delivery of the message (Flax 2017; Rana 2015), and decreased the privacy and confidentiality around their personal information (Akinfaderin-Agarau 2012; Jalloh-Vos 2014; Rana 2015).

Finding 8: Participants believed that the cost of participating in DTCC should be free or very low, as cost could present a barrier to participation, particularly for young people and those on lower incomes. Participants felt that there should be little or no charge for costs such as joining the digital health intervention, downloading applications (apps), or for sending and receiving mobile messages/phone calls (high confidence).
Participants in several studies felt it was important for digital health interventions to have little or no cost, as these costs could present a barrier to participation ( If the intervention could not be offered at no cost to the client, then participants felt that the interventions should be very low cost and that cheaper options should be used, for example SMS instead of voice calls (Akinfaderin-Agarau 2012). In some cases, messages sent to participants were free, but if a participant wanted to reply they had to use their own airtime. Some participants thought that this would prevent people from using the bi-directional functions within digital health interventions (Rana 2015).

Finding 9: Participants' ability to access digital communication was sometimes limited by their language skills and their personal level of literacy and/or techno-literacy (moderate confidence).
Some studies, the majority from LMIC settings ( One study from the USA found that Latina women needing cancer screening receiving interactive voice recordings as reminders understood how to access messages but that the language used in the messages was not familiar to them (Greaney 2014), as illustrated in the following quote.

Library
Trusted evidence. Informed decisions. Better health.
Cochrane Database of Systematic Reviews "'If your response is 'yes', then press the star button.' 'Star' is what Americans say, but on the telephone there is no star; it is an asterisk. It all depends on who you are speaking with. If the person you are speaking with understands that that is a star then let's press the star, but if the person understands it's an asterisk then he/she will begin to look for a star" (Greaney 2014).
Participants with literacy issues often preferred voice calls to SMS, as they could talk with the caller and ask for clarifying information (Akinfaderin-Agarau 2012; Jalloh-Vos 2014). Participants in two studies said they had learned or were willing to learn how to text in order to participate in digital health interventions (Calderón 2017; Smillie 2014).

Communication delivery and format preferences
('Summary of qualitative findings' table for Findings 10 to 13 is shown in Table 4.) Finding 10: Participants often had preferences for how often health messages were sent, the time of day they were sent, and the duration of the DTCC. However, there was variation in what most participants felt was appropriate timing and frequency, and these preferences were often linked to the health issue on which the messaging was focused; whether people had their own phone or had to share a phone; and the participant's particular circumstances. Participants were particularly concerned about being bombarded with too many messages; whether the timing of the messages was convenient for them; and/or whether messages arrived in connection with the behaviour the message was trying to target (moderate confidence).
Many studies discussed and presented participants' preferences related to timing, frequency of messages, and duration of digital health projects ( . However, within and across studies and client groups, there was no consensus as to the ideal timing, frequency, or duration, as this was linked to personal preferences, contextual factors (such as attending school), and the behaviour or information the text message was trying to target.
With regard to frequency, participants did not want in general to feel pestered or bombarded by too many messages (Evans 2016;Gold 2010; Willoughby 2017), but described a fine balance between feeling bombarded and not receiving enough information to reinforce the messages. For example, participants receiving medication reminders were open to receiving multiple texts a day, whereas those receiving more general/less tailored information wanted messages much less frequently. However, in a number of studies no clear consensus emerged on the optimal frequency of the messages (Evans 2016;Gold 2010;Rana 2015).
Participants' preferences for message frequency could also be linked to owning a phone or having to share (Jennings 2013). Those who had to share a phone wanted messages only a few times a week with the ability to stop the messages if the phone owner was away, whereas those who owned their own phone were open to daily messages (Jennings 2013).
With regard to the timing of delivery of the messages, preferences for timing varied among population groups. It was important to participants that messages arrive when they could be seen and accessed, for example not late at night, when they would not be seen until the next day (Rana 2015; Ware 2016). Some adolescents and young adults thought during school hours would be fine (Cornelius 2009;, whereas others believed that this could cause problems based on restrictions around phone use (Cornelius 2009;Rana 2015;Wright 2011). Some participants felt that it was important that the message arrive in connection with the behaviour it was targeting Gold 2010;Menacho 2013; Naughton 2013; Rana 2015; Ware 2016), or in good time before an appointment (Odeny 2014), for example on a Friday night before going out to remind them of condom use . Participants also liked the option of tailoring the timing of messages to fit their lives .
With regard to duration, no consensus emerged on how long the intervention should last (Evans 2016).
Finding 11: Participants had different preferences for various delivery channels available for sharing information through DTCC, including mobile messaging, interactive voice response, or speaking with a healthcare provider. These preferences were influenced by a number of factors including cost, convenience, the ability to store messages and re-read them, familiarity with the channel, personal preferences, the nature of the content being delivered, the nature of the topic, language and literacy considerations, and the ability to have a discussion with a real-life person (moderate confidence).
Some studies presented data related to participants' preferences for the delivery channel used to share information for digital health interventions ( However, participants in a few studies raised concerns with text messaging or preferred other delivery channels. They thought, for example, that text messages were more appropriate for younger audiences (Willoughby 2017), or that the highly convenient nature of the SMS was a negative that would only have a short-term impact, as people would delete them or ignore them (Naughton 2013; Rana 2015). Finally, others who were unfamiliar with SMS technology felt that they were too passive and unfamiliar and so they ignored them .

Library
Trusted evidence. Informed decisions. Better health.

Cochrane Database of Systematic Reviews
A few studies presented data that participants preferred interactive voice response (IVR) to receive digital health interventions (Greaney 2014; Missal 2016; . Often this was related to low literacy levels in a community or that participants found them easier, more interactive, and that they attracted more attention than an SMS, meaning that they would not be easily missed . However, some participants felt that they should be able to repeat or re-read the message at a time more convenient to them, for example through linked delivery on platforms such as voice, video, or SMS (Missal 2016).
A number of studies presented data where participants discussed their preferences for speaking directly with a healthcare provider on the phone to other delivery platforms ( Participants also discussed other delivery channels that they liked. These included an app (Mitchell 2016), written letters (Greaney 2014), voicemails, Facebook, and reminder alarms (Rana 2015).

Finding 12: Participants appreciated personalised health information and discussed their preferences for options to make interventions more relevant to individuals. This could include sender-based personalisation or receiver-based options.
Reasons for these preferences included engaging the user, enhancing credibility, increasing feelings of ownership, control over their personal information, and feelings of privacy. Preferences for tailoring included making digital health messages personalised by using an individual's name; allowing participants to choose the content, topic, and language of their messages; providing information relevant to the participant's setting (local information); allowing them to select the timing and frequency of the message; providing personalised reminders (e.g. for vaccination or medication); and allowing participants to have control over privacy settings (low confidence).
Some studies discussed participants' thoughts about personalised or customised messaging and their preferences for options to make interventions more relevant to individuals ( . For example, participants in one study on HIV messaging in Kenya to prevent mother-to-child transmission suggested that two different sets of messages be developed for those who had and those who had not disclosed their HIV status to their partner. This would help women avoid risking disclosure of their status (Jennings 2013). Participants in another study discussed the importance of using personal pronouns to make messages more relevant to the user. They felt that naming functions within the digital health intervention as, for example, " MY test plan", would help participants to take ownership of these functions (Goldenberg 2015).
Participants in several studies mentioned ways in which the people receiving communication could personalise or customise digital health interventions that they believed would be important. These included the following.
• Being able to request the time of day the message(s) would be sent ( Participants in a number of studies mentioned ways in which the people designing and sending communication could personalise or customise digital health interventions that they believed would be important. These included the following. • Using unique access codes in order to prevent children and other adults from unintentionally gaining entry into discussed personal health information (Calderón 2017). • Using the individual's name (Evans 2016).
• Delivering the message in different languages (Evans 2016).
• Including community-specific information (Evans 2016;Munro 2017). • Being able to, for example, text "STOP" if they wanted to stop receiving messages ). • Explaining the reason for the appointment as well as the date (e.g. the specific vaccinations to be received on that day) (Odeny 2014).
Finding 13: Participants mentioned various message formats that they preferred. These included a preference for short, concise, personalised, clear, and direct messages in a language they could understand and in full text rather than "text speak" (low confidence).
A number of studies presented data related to participants' preferences for message format ( In one study (Missal 2016), participants commented on their experiences with interactive voice response (IVR). Most found the content and language of the message useful. However, some found the message too fast, short, and sometimes the voice was not clear Cochrane Database of Systematic Reviews enough. They were unable to request for the message to be played again if they had missed or misunderstood the content.

Communication content preferences
('Summary of qualitative findings' table for Fndings 14 to 17 is shown in Table 5.) Finding 14: Participants' perceptions of who sent the digital health communication could influence their trust in and perception of the credibility and value of the DTCC and the information it provides. Participants said they wanted a known, identified phone number; messages sent from a reliable, trusted, credible source such as health professionals or official sources; and in some cases to feel like the messages were sent by a person (even if sent from an automated service). However, some participants, such as those with stigmatised health conditions, preferred an unmarked sender to protect their privacy (moderate confidence).
Several studies found that participants' perceptions of who the sender of the message was could influence their trust in and perception of the digital health intervention's credibility and value . Participants in many studies identified that the sender should be known and identifiable. If this was not the case many stated that they were more likely to ignore or delete the message (Akinfaderin-Agarau 2012; Greaney 2014; Smillie 2014), as their phones were already receiving many messages linked with telemarketing (Menacho 2013;Missal 2016). If a sender was unknown, it was also felt that this would decrease the credibility of the message. However, some participants preferred an unmarked sender in order to protect their privacy. This was in the case, for example, of a stigmatised health condition such as HIV, where if the sender's phone number were identifiable, a person's HIV status may be inadvertently revealed (Rana 2015). These participants felt that a solution would be to allow participants to choose if they wanted the number or name of the sender to be attached to the message (Rana 2015).
Participants also wanted the messages to come from a reliable, trusted, and credible source (Brown 2014;Cates 2015;Evans 2016;Lau 2014;Willoughby 2017). They were interested in the credentials or education of the person creating or responding to the message. Many felt that it was important that the information be written and sent by health professionals and from official sources (Cates 2015; Greaney 2014; Lau 2014; Mbuagbaw 2012). In some cases, if this was unclear, this led them to worry about the accuracy of message content and the intentions of the proposed intervention (Calderón 2017). If the participant knew and trusted the source, they also felt that their data would be protected (Evans 2016).
Many participants felt that it was important that the messages felt like they were coming from an actual person rather than from an automated system ( Lau 2014; Naughton 2013; Rodrigues 2015; Willoughby 2017). Some participants felt that since the messages were written by a health professional, the computer's role in the automated sending of the messages was irrelevant (Naughton 2013). These participants suggested that messages which were not repeated and were delivered at various times of day using the par-ticipant's name would feel more like they were sent by a person (Naughton 2013).
Finding 15: Participants said that the tone of digital health communication mattered to them. Their preferences varied but included a tone that was: motivational, friendly, encouraging, polite, respectful, congratulatory, personalised, upbeat, positive, humorous, and relatable. Some participants highlighted that they did not like feeling pressured, lectured, shamed, or frightened by digital health messages (low confidence).
A number of studies, the majority from HIC contexts ( Finding 16: Participants had preferences regarding the content they receive through DTCC. They wanted varied content that provided new knowledge and reminders, as well as explanations, solutions, and suggestions about health issues. They were interested in content related to health, illness, and treatments and practical topics such as health facility location and transportation. They wanted this information to be relevant and acceptable to their personal circumstances and local setting (moderate confidence).
A number of studies presented data related to participants' preferences regarding the content they received through digital health interventions (Brown 2014 ). An example of such practical advice is the suggestion that mothers who struggled with breastfeeding (or the idea of breastfeeding) could pump their milk and give it to their infants in a bottle instead of switching to formula milk (Brown 2014), as expressed in the quote below from a teenage mother in the USA.
"I was kind of grossed out by actually breast feeding so I had decided to just use formula. I don't know why I was so grossed out, I just was. But then you said in one of the messages that you can pump and feed the breast milk through a bottle, and the baby still gets all those benefits. It makes sense, but I just never thought of it. Because of that message, I started to do that and my baby still gets breast milk, but otherwise I would have given up" (Brown 2014).
Finally, participants were interested in content related to health, illness, and treatments and practical topics such as health facility location and transportation (Brown 2014 Finding 17: Some participants felt that including elements in the mobile-based platform in which participants are asked for a response (e.g. via knowledge quizzes or multiple-choice questions or a practical tool allowing access to additional information, such as a nutrition calculator) could increase the engagement of users with the intervention, its content, and provide additional information to them. In one study, participants suggested that it would be helpful if the response was quick, simple, and convenient (low confidence).
Participants in a few studies, all from HIC contexts (Cornelius 2009; Munro 2017; Naughton 2013; Wright 2011), felt that bi-directional communication or content that somehow engaged clients (e.g. using quizzes or replies) would be more useful than one-way communication. Participants mentioned that for this to be the case, the interaction should be made convenient and allow for quick and simple responses. Such interactive options would help keep participants' attention and prolong engagement with the intervention. Participants in one study from the UK, women who had smoked during a previous pregnancy and receive SMS support for smoking cessation, also felt that when the communication asked for a reply they would think more about the content of the messages they were receiving (Naughton 2013), as noted by one participant below.
"An interactive text inviting a reply would make 'you think more about the text message' otherwise 'you don't have to do anything with it so you read it and then forget about it'" (Naughton 2013).

Privacy and confidentiality regarding personal health information
('Summary of qualitative findings' table for Findings 18 to 19 is shown in Table 6.) Below we present findings specifically related to privacy and confidentiality. However, this theme is also touched upon in a number of other findings including Finding 7.
Finding 18: Some participants with health issues that are often seen as stigmatised or very personal (e.g. HIV, family planning, and abortion care) worried that their confidential health information would be disclosed or their identity traced due to their participation in DTCC. In general, people's perceptions of information delivery channels (SMS, interactive voice response, voice call) were influenced by how confidential they felt the delivery channels to be (high confidence). . These participants worried that the SMS or phone conversation would reveal their status to people who picked up their phones or who overheard their conversations. However, in one study from Kenya, participants felt that receiving an SMS or phone call would protect their privacy more than a face-to-face appointment, as conversations in a health facility were easily overheard by others. Some participants in a family planning digital communication study in Sierra Leone feared that participating in the intervention would compromise their privacy because their husbands or family members would find out they were using family planning methods (Jalloh-Vos 2014). Similarly, a participant in a postabortion care intervention in Cambodia was worried that her medical history of abortion would be discovered by her family if she was to receive a phone call in their presence or if someone else were to answer her phone  Cochrane Database of Systematic Reviews tial, and more readily available (Curioso 2009). Others felt that they had more control over text messaging as they could prevent the messages from appearing on their phone screens, could lock their phones, and could delete messages .
Finding 19: Some participants proposed strategies to address their concerns regarding confidentiality and privacy. These strategies for communication included neutral, coded, or discreet language; access codes; communication that does not disclose the sender; coming from a trusted sender; and the ability to tailor and control content, timing, and frequency of their messages (high confidence).
In some studies, many from LMIC settings (

Perceptions of intervention impact
('Summary of qualitative findings' table for Findings 20 to 25 is shown in Table 7.) Finding 20: Some participants thought that participating in DTCC had influenced their behaviour, while others did not. Reasons given for the changes in behaviour included receiving new knowledge; receiving strategies on how to initiate discussion with a partner or healthcare provider; being motivated or reassured by the intervention; and being reminded, for example, to take medication or make an appointment. Some participants who believed that the intervention did not have any influence on their behaviour found that the digital health interventions were not relevant to them (low confidence).
A number of studies found that participants thought that taking part in DTCC had influenced their behaviour ( Digital targeted client communication also provided some participants with specific strategies, for example how to discuss a health topic with a significant other. Some studies found that this led to participants talking to their partners about testing for sexually transmitted diseases Jennings 2013); preparing for labour and delivery (Missal 2016); or discussing questions with a healthcare provider (Munro 2017). Participants felt that without these strategies they might not have taken the steps to have these conversations about their health.
Messages also motivated or reassured participants about their health decisions. Some participants felt a sense of confidence in their parenting decisions, as the messages they received validated their parenting choices (Brown 2014); or made them feel like the difficult conversations about disease testing were the right thing to do   Cochrane Database of Systematic Reviews undertaking some health tasks. They were concerned that, in the absence of these reminders, they would adhere poorly to care plans (low confidence).
A few studies, from LMIC settings in Africa (Jalloh-Vos 2014; Mbuagbaw 2012; Rana 2015), found that some participants were concerned about over-reliance on digital reminders, and that in their absence adherence might become problematic. Two studies explored perceptions related to adherence to HIV medication, and another family planning and antenatal/postpartum care. These concerns were important to participants that mentioned the problem of reliance, as they believed that the digital health interventions would eventually end and people needed to remember themselves (Rana 2015). For example, in Sierra Leone, both women and men mentioned that if the nurse did not call to remind them about family planning they would forget, and that it was the nurse's job to remind them to come (Jalloh-Vos 2014). Two studies from LMIC settings found that participants believed that digital health interventions could potentially save them time and money by giving them access to health care through their mobile phones instead of having to go to the health facility (Calderón 2017; Smith 2017). In Peru (Calderón 2017), women over 18 who had at least one child believed that this would be useful to families living in resource-poor communities, where families often cannot afford the cost or time related to transport to the health facility or the consultation. They felt that the intervention would be used more by families that live further away from a health facility, described as follows.
"People like me who work all day could use it, we have to use it, because they then already have someone to ask, you will not be with that doubt 'do I give or not give the medication to the baby?' or 'how much medication do I have to give him/her?' I don't know, so when we have someone to call or to send text messages to, you will not hesitate to do it as they say" (Calderón 2017).
Women in Cambodia also felt that the digital health intervention saved them time and money, as they could receive their family planning counselling over the phone instead of having to travel to the health facility. This saved them both the fees associated with transport and the consultation with the healthcare provider .
Finding 24: Some participants felt that digital health interventions provided them with feelings of support and connectedness, as they felt that someone was taking the time to send them messages. A few participants felt that in some cases the sense of caring and support that they received from healthcare providers through digital health interventions had a positive influence on their relationship with their healthcare provider (moderate confidence).
A number of studies, the majority from LMIC settings ( , presented data related to participants' reflections around the caring and supportive nature of DTCC. In some cases, the participants in these studies felt that someone was interested in their situation, was invested in their well-being, and cared about them. This led some participants to feel encouraged and to have increased self-confidence and feelings of self-worth. For others, the messages provided support, guidance, and information, often giving a sense of direction, reassurance, and motivation to participants. Support was presented in various ways, for example to take medications or providing counselling. In some cases, this increased dialogue between healthcare workers and participants had a positive influence on their relationship, for example by making each group more aware of the other's expectations (Jalloh-Vos 2014; Smillie 2014; Smith 2017).
Finding 25: Participants described how they shared digital communication content more broadly with friends, family, and community members. Many participants felt that the information would be useful to others (moderate confidence).
Some studies found that participants shared messages they had received or thought they would share messages with friends, family, colleagues, or neighbours ( Participants in one study felt that sharing messages and experiences could help to create a sense of community (Wright 2011).
In another study, participants were excited about the possibility of creating message content to share with friends and family (Cornelius 2009).
Finally, one study sent SMSes to pre-existing community-based women's groups about maternal and child health topics. Not all of the women in the group were pregnant or had children. However, these participants mentioned that they still found the content interesting and discussed it with other group members, family, and neighbours (Flax 2017).

Supplementing the Cochrane Reviews of e ectiveness with synthesised qualitative findings -matrix results
As described in the Methods section, we used a matrix approach to explore whether potential implementation barriers that we had identified in our synthesis (see Table 8 below) had been addressed in the programmes evaluated in the related Cochrane Reviews of effectiveness (Palmer Ongoing a; Palmer Ongoing).
The relevance of our synthesis findings to the reviews of effectiveness was strengthened by the fact that the studies included in the qualitative synthesis and those included in the effectiveness reviews came from similar settings. Around half of the studies in both the synthesis and the reviews were from low-income countries (primarily African countries), whilst the remaining studies where from high-income countries (primarily the USA). In addition, around 10 of the qualitative studies included in our synthesis appear to have been carried out to inform the development of some of the trials. Cochrane Database of Systematic Reviews Table 8 presents an overview of our matrix assessment (a more detailed version can be found in Appendix 5, and a list of all of the studies included in the analysis is shown in Appendix 6). In summary, we found that some of the included trials did describe efforts to address some of the potential barriers to implementation that we had identified in the qualitative synthesis. However, most of the trials referred to only a small number of barriers, and in some trials potential participants were actively excluded if they were dealing with these barriers.
One barrier identified in our synthesis was related to situations where the target group did not own a functioning mobile phone. In more than half of the trials, trialists dealt with this barrier by making mobile phone ownership a condition for trial participation. Between 0.3% and 63% of eligible participants were excluded from trial participation because they did not own a phone or could not receive text messages. Some trialists did attempt to find solutions to this problem, however. For instance, some trialists only required that participants had access to a phone, for instance through family, friends, or neighbours Gibson 2017;Kassaye 2016;Odeny 2014); in other trials participants were provided with phones (e.g. Cook 2015; Ingersoll 2015); and in one trial, local healthcare providers were given phones to share with women who had no access to phones (Joshi 2015).
Other barriers identified in our synthesis were tied to poor access to electricity to charge phones and poor access to network services. However, trialists rarely referred to these issues. If mentioned, this was usually because access to specific networks or to stable electricity was a condition for participation in the trial. Trials that did attempt to deal with these barriers included one trial based in the USA (Ingersoll 2015). Here, assessments carried out prior to the trial showed that many people had inconsistent cellular or internet service, but that they could usually receive text messages. The trialists therefore built a texting system rather than one that would require a consistent cellular signal or internet access (Ingersoll 2015). Another exception was a Kenya-based trial (Pop-Eleches 2011). As many participants in this trial needed to charge their phones at feebased charging stations in nearby markets, the trialists provided financial support to cover these fees (Pop-Eleches 2011).
Another barrier was tied to the expenses associated with the trial interventions. This was also confirmed as a problem by some trialists, who described poor participation because of cost barriers (Ahlers-Schmidt 2012; Lester 2010; Norton 2014). Around one-third of the trials described efforts to address this problem in some way, for instance by offering vouchers to participants (e.g. Lund 2012; Omole 2018), or by offering participants free access to various internet sites in return for receiving mobile advertising (Gold 2011).
Our synthesis also pointed to problems in receiving messages when people changed their phone numbers or sim cards. Again, this issue was rarely referred to in the trials. When mentioned, this was usually as an explanation for participant loss to follow-up or as a condition for trial participation. Examples of trials that did try to address this issue include a trial from India, where participants with new phones were able to update their contact details with their local health worker (Joshi 2015); a USA-based trial , where participants were asked to update information about their phone number every month (Devine 2014); and a trial from Ecuador, where trialists gathered several different contact avenues to maximise their ability to contact participants (Maslowsky 2016).
Two related barriers were tied to situations where people had to share phones or had their phones controlled by others, and situations where these people, as well as people with their own phones, were concerned about privacy issues when receiving messages. Very few trialists discussed situations whereby people's phones might be controlled by others, and at least two-thirds did not refer to privacy issues. However, several trialists did describe efforts to address issues of privacy and confidentiality. These included trials targeting people about oral contraception (Castano 2012), or HIV treatment and prevention (Garofalo 2016; where trialists ensured that messages did not refer to the recipient's name, their health status, or the name of the medication; encouraged recipients to delete messages; and gave them information about privacy settings on their phones. Other barriers identified in our synthesis were created by language problems, low literacy, or limited techno-literacy. At least a third of the trials made no reference to these issues. Another third actively excluded people who did not speak one or two mainstream languages, and several trials also excluded people who could not read these languages or who were illiterate. Some trialists described efforts to overcome these problems, including designing text messages so that they scored low on readability scales Stockwell 2014); offering text messages in the local language (Kassaye 2016); or sending voice messages or pictorial messages instead of text messages (de Costa 2012; Smith 2015).
Our synthesis also found that people's trust in and perceptions of the information they received was influenced by their perceptions of the sender. Attempts to address this issue were rarely described by trialists. Exceptions included a UK-based trial, Cooper 2015, that highlighted in advertising materials that the intervention was supported by the National Health Service (NHS) (Emery 2018); and a USA-based trial , where messages were linked to celebrities that the target audience said were most interesting to them (Devine 2014).
The final issue we assessed in our matrix was the extent to which the target group had been given an opportunity to offer feedback about their needs, preferences, and experiences regarding the intervention. More than half of the trialists described collecting user feedback to develop or improve the intervention, although it was not always clear how information gathered before, during, or after the trial had influenced the intervention or would influence future versions of the intervention.

Summary of main results
The acceptability of digital targeted client communication (DTCC) is mixed. Some clients described DTCC in positive terms, as providing them with support and connectedness; giving them a feeling that someone cares about them; and, in some cases, having a positive influence on their relationship with their healthcare provider. However, clients who are dealing with health conditions that are often stigmatised or very personal (e.g. HIV, family planning, and abortion care) worry that their confidential health information will be disclosed or their identity traced.
Clients' perceptions and experiences of DTCC can be influenced by a number of factors. Participants believed that there should be little or no charge to participate in digital health interventions. They

Library
Cochrane Database of Systematic Reviews wanted messages that were from a trusted sender and that were polite and encouraging. They did not like feeling pressured, lectured to, or frightened. They wanted varied information that arrived at a time and frequency that was convenient for them. Content preferences included new knowledge, reminders, solutions, and suggestions about health issues presented in a clear, short, and personalised way.
In general, barriers to participating in digital health interventions included problems with network connectivity, access to electricity, device usability, and issues tied to data confidentiality and privacy.
Some of our findings addressed issues related to gender, equity, and human rights. For example, access to healthcare services via digital devices may be particularly helpful to clients with caring or work responsibilities; clients who live far from health facilities; and poorer people. However, access to digital health interventions may be particularly difficult for others who speak minority languages or who have low literacy skills or low digital literacy skills, who for example do not know how to open, read, or send a text message. Participation may also be difficult for clients with poor access to network services, electricity, or ownership of mobile phones. For clients, particularly women and adolescents, who have to share or borrow a phone or who have phone access controlled by others, it can be difficult to receive messages or keep them private. Clients with stigmatised health conditions such as HIV were also concerned about what would happen if their privacy was not protected, for example if the fact that they were participating in the digital health intervention led to the disclosure of their HIV status.

Overall completeness and applicability of evidence
The sampling approach we used in this review (see above) aimed to achieve a maximum variation of target populations, settings, delivery mechanisms, and content focus of the targeted digital health intervention.
We found studies that represented clients from all of the client groups included in the scope of this synthesis. However, studies from parents and pregnant and postpartum women and their partners were not as common. For example, we found very few studies researching DTCC for HIV-positive pregnant and postpartum women. We are therefore less certain whether these client groups have the same perspectives and expectations of targeted digital health interventions.
A majority of the studies covered topics related to sexual and reproductive health for young people and adults, with a focus on sexually transmitted infections and specifically HIV/AIDS. We are therefore less certain about whether participants receiving digital health interventions related to other topics have the same preferences for information and intervention delivery, such as vaccination and general child health.
The studies included in this review come from a wide variety of contexts and settings. There are a number of included studies from poorly resourced healthcare systems as well as high-income settings. The range of settings included in this review highlighted access issues as well as issues related to gender and equity across all settings.
The majority of the studies included in this synthesis looked at digital health interventions where the content was or would be delivered by SMS. We have only a few studies that explore percep-tions related to other delivery methods such as interactive voice response (IVR).
The data collection methods and study designs of the included studies may in some cases limit the applicability and completeness of the data reported. One example of this is the large number of studies that presented hypothetical digital health interventions (18 studies), both with and without examples of content, and asked participants to reflect on their preferences if they would participate. It is unclear whether these preferences based on hypothetical thinking would remain the same once the participant had actually experienced participating in the intervention and receiving the messages.
All of the included studies made use of individual or group interviews and focus group discussions as their main method of data collection. None used long-term ethnographic methods or field observation. While interviews and focus groups allow researchers to collect data on what people say, observational methods also allow researchers to collect data on what people do and why. This would have been appropriate for understanding how clients engaged with and used the various digital health interventions and allowed researchers to compare actions with interview and focus group data. Interviews and focus group discussions seem to be the most commonly used research methods amongst qualitative researchers exploring issues related to health. This could be because they are less time-consuming than longer-term ethnographic methods.

Agreements and disagreements with other studies or reviews
The review team identified one published qualitative evidence synthesis that had a similar scope to this review (O'Connor 2016). The O'Connor synthesis had a much broader scope and explored factors affecting client engagement and recruitment to digital health interventions. Also, these interventions did not have to be targeted at a specific audience and were open to any health intervention delivered by a digital technology. We are aware of an ongoing related qualitative evidence synthesis considering healthcare workers' experiences with and perceptions of targeted digital communication via mobile device ( . Two overviews of systematic reviews looking at text messaging (Hall 2015), and the impact of mHealth interventions (Marcolino 2018), have also been done. However, many of these reviews and overviews address health issues beyond the scope of our synthesis, such as obesity and chronic illness (Bacigalupo 2013; Hamine 2015; Peiris 2014).
In 2015, a review was published that explored the adoption of mHealth in low resource environments (Chib 2015). This review found that the majority of studies in these environments concentrated principally on pilot projects focused mostly on the introduction and implementation of new interventions. This finding is similar to the types of studies we found in this synthesis, where approximately half of the included studies were projects using hypothetical examples to develop targeted digital health interventions. A further six studies were pilot projects for new targeted digital health interventions. We only identified one study that explored client experiences related to an ongoing project at scale, Mobile Midwife in Ghana (Entsieh 2015). Other systematic reviews have identified a Our qualitative evidence synthesis highlighted issues of access to mobile phone technology, networks, and electricity. Other studies have highlighted this issue as well (Aranda-Jan 2014; Bukachi 2007). As discussed in the findings of our matrix analysis, future trials should take these access issues into consideration when defining participant inclusion criteria. At this time, little is known about how targeted digital health communication interventions are used or perceived by those who do not have access to phones, networks, or electricity, as the large majority of existing trials and qualitative studies include participants who have phones and access to networks and electricity. Our findings highlight a potential lack of digital literacy among clients in many settings, which could also affect the feasibility and acceptance of targeted digital health communication interventions. Similarly, the O'Connor 2016 review concluded that "more investment is also needed to improve computer literacy and ensure technologies are accessible and affordable for those who wish to sign up to them".
Finally, issues related to tailoring and personalisation of timing, format, content, and privacy were clearly described in the findings of this review. One other review, Gurman 2012, mentions that less than half of the interventions included described targeting or tailoring the content. A meta-analysis of tailored print health behaviour change interventions found that tailored interventions were more effective than non-tailored interventions for health promotion (Noar 2007). The O'Connor 2016 review also found that interventions that are personalised when possible should be a focus when creating digital health interventions. To support this thinking, research has found that successful intervention design demands a user-centred and iterative approach to developing new digital behaviour change interventions (Yardley 2016). Tailoring was also identified as a core interactive design feature in effective e-health interventions (Morrison 2012).

Summary of integrating the findings from this synthesis with the findings of relevant Cochrane e ectiveness reviews
We used a matrix approach to explore how the findings from our synthesis related to, or could help to inform, the findings of the two related Cochrane Reviews of effectiveness of DTCC (Palmer Ongoing a; Palmer Ongoing).
Our synthesis and the two intervention reviews were designed to complement each other, and used similar inclusion criteria where possible. The included studies in the synthesis and the two reviews were therefore broadly similar in terms of population groups, delivery mechanism, setting, and publication date. All three syntheses/reviews included studies with the same population groups (i.e. adolescents and adults that are users or potential users of reproductive health services; pregnant and postpartum women, including women living with HIV; and parents of children under five). In all three syntheses/reviews, the majority of programmes used text messages to communicate with their target audiences. Slightly more than half of the 35 studies in the synthesis were from LMICs. Slightly more than half of the 68 studies in the intervention reviews were also from LMICs. The studies sampled in our synthesis were published between 2009 and 2017, whilst the trials included in the intervention reviews were published between 2006 and 2018.
Our matrix shows that most of the potential barriers to implementation raised by participants in the qualitative research were not referred to in most of the trials. It is possible that trialists did attempt to find solutions but did not report these in their publications. It is also possible that our search strategy failed to identify all relevant publications, and that these solutions were reported elsewhere. For instance, many of the trialists did not describe how they dealt with privacy issues when participants received messages in the papers we examined. However, it is possible that some privacy issues were addressed to some extent during scientific ethics review and approval processes.
Some of the included trials did describe efforts to address some of these issues. Where this occurred, it would be useful to explore the impact, transferability, and potential sustainability of these efforts. Some of the solutions described by trialists are likely to be sustainable, including efforts to increase accessibility of messages through audio, pictures, and local languages; or protecting people's privacy though anonymising messages. However, other solutions may be less sustainable outside of a trial context or may be unacceptable for other reasons. These include the distribution of free phones or covering the costs of recharging phones. Other solutions, such as sharing of phones, may also lead to concerns about privacy.
In several trials, trialists simply excluded participants who were dealing with the barriers we identified in our synthesis. The exclusion of participants who do not own their own phones; who are likely to change their phone number or sim card; or because of language, literacy, or techno-literacy issues is problematic. Whilst the size of these challenges varies from setting to setting, people that experience these challenges may be the same people who need health services the most, but who access them the least. Interventions that specifically aim to increase people's access to and use of healthcare services should therefore make a particular effort to address these challenges.

Review author reflexivity
As part of the synthesis process, we reflected on how our backgrounds and positions might have influenced our choice of review topic, study selection, data extraction, analysis, and interpretation of data. Our backgrounds are in health systems research, social sciences, epidemiology, and nursing, and while working on the synthesis we were employed by government research institutions in Norway, South Africa and the USA (HA, CG, SL, NL, EA) and by the WHO (TT). The synthesis was commissioned to inform a WHO guideline, specifically to address guideline questions regarding the acceptability and feasibility of digital health interventions. Three of us were key members of the WHO guideline technical team (TT, SL, CG). Some of us had been involved in primary research related to digital health interventions, whilst others (CG, HA) had no previous work experience on this topic. Before working on the synthesis, our viewpoints regarding digital health interventions ranged from being neutral to these types of interventions to a slight skepticism of digital interventions as a magic bullet in solving health issues. All of us started the process believing that the implementation of digital health interventions should be informed by robust evidence of effectiveness, acceptability, and feasibility. Whilst working on the synthesis, we became more convinced of the importance of supporting evidence-based decision-making in digital health -especially having seen from the studies included in this review, as well as from several other reviews commissioned for the WHO guideline, the range of challenges and constraints in implementing these interventions at scale, and in ways that protect the privacy of participants.

Implications for practice
The following questions, derived from our findings, may help health system or programme managers when implementing or planning for digital targeted client communication strategies to address issues of importance to their target population. It is important to consider local contextual factors including gender, age, cultural group, and education when implementing new digital targeted client communication strategies. 8. Have solutions been considered for tailoring or changing intervention content to ensure the privacy and confidentiality of clients and to avoid any harms that a break in this privacy may cause? 9. Has an attempt been made to explore how clients perceive different sources of digital health interventions as more or less reliable, trusted, and credible? Has an attempt been made to use those sources that are perceived as trusted, reliable, and credible to send digital health messages? 10.Have members of the client target group been given an opportunity to offer feedback about their needs, preferences, and experiences regarding the intervention during intervention development, implementation, and evaluation?

Implications for research
These implications have been derived from the CERQual assessment and the overview of the studies included in this review.
There is a need for better reporting of context, sampling, methods, and researcher reflexivity in qualitative studies. Future qualitative studies should report their methods clearly and include reflection on the researchers' roles in the study and how this may have impacted on the process and results of the study. More detail concerning setting and participants is also needed to identify underlying cultural or social phenomena (shared values or beliefs) that mediate the influence of communications, as these need to be addressed when designing targeted digital health interventions. A better representation of the participant's voice in the data in the studies included in this synthesis could have improved our confidence in some of the findings. For example, in some studies quotes were not labelled with a participant identifier, so we were unable to determine if the quotes came from multiple participants or the same participant.
Research about digital targeted client communication should aim to include a broader spectrum of participants in relation to phone ownership, literacy, and ability to use a smartphone. Researchers could also focus on exploring why some digital health interventions do or do not influence participants' actions and behaviour.
More research is needed on the public's preferences around the details of timing, amount, and content of digital health interventions from people who have actually participated in digital health interventions. There is a large body of hypothetical studies, and there are some studies that evaluate or discuss participants' experiences after participating in a pilot project or a research trial. However, we found only one study that interviewed participants involved in a digital health intervention that was being delivered on a national scale, in Ghana (Entsieh 2015).
Trials assessing the effectiveness of digital targeted client communication interventions should consider the issues identified in this qualitative evidence synthesis (Table 8 above) and should ensure that the design and assessment of the intervention are properly reported, for instance following existing reporting guidelines for digital health interventions (Agarwal 2016).

A C K N O W L E D G E M E N T S
We would like to acknowledge the contributions of Garrett Mehl in developing the protocol and Janan Dietrich for the searching and screening of other relevant studies linked to the matrix analysis.  (7)

Munabi-Babigumira 2017
Munabi-Babigumira S, Glenton C, Lewin S, Fretheim A, Nabudere H. Factors that influence the provision of intrapartum and postnatal care by skilled birth attendants in low-and middle-income countries: a qualitative evidence synthesis.   Overall, participants had a range of views regarding acceptance of the idea of receiving health information through their mobile devices. This was due to factors such as familiarity with the technology, convenience, control, being able to save and re-read messages later, cost, seeing it as a simple way of providing a reminder for medication or appointments, and the sense that someone was thinking about them and cared enough to send a message  Participants said that they liked two-way digital communication, as this allowed them to engage directly with a healthcare provider, which they trusted more; to receive answers to their questions and have opportunities for discussion; and to receive a more immediate response. However, some participants felt that for some topics they would feel uncomfortable talking to a healthcare provider through a digital channel due to issues related to shyness and privacy, and would prefer to use SMS  Participants reported varying degrees of access to network services, including cell networks (for calls and SMS) and internet. In addition, some had poor access to electricity to charge their phones. These factors were reported to be barriers to using the digital targeted client communication.

A D D I T I O N A L T A B L E S
High confidence   Participants often had preferences for how often health messages were sent, the time of day they were sent, and the duration of the digital targeted client communication. However, there was variation in what most participants felt was appropriate timing and frequency, and these preferences were often linked to the health issue on which the messaging was focused; whether people had their own phone or had to share a phone; and the participant's particular circumstances. Participants were particularly concerned about being bombarded with too many messages; whether the timing of the messages was convenient for them; and/or whether messages arrived in connection with the behaviour the message was trying to target.

Moderate confidence
Due to minor concerns regarding methodological limitations and moderate concerns regarding relevance     Some participants with health issues that are often seen as stigmatised or very personal (e.g. HIV, family planning, and abortion care) worried that their confidential health information would be disclosed or their identity traced due to their participation in digital targeted client communication. In general, people's perceptions of information delivery channels (SMS, interactive voice response, voice call) were influenced by how confidential they felt the delivery channels to be.

High confidence
Due to minor concerns regarding methodological limita-     Anti retroviral therapy (ARV); Human Papillomavirus (HPV); Prevention of mother-to-child transmission (PMTCT); Randomized control trial (RCT); Tuberculosis (TB) (Continued) Finding 2: In discussing the pros and cons of digital targeted client communication compared to in-person meetings with a healthcare provider, some participants perceived interacting with a healthcare provider as preferable, warmer, and something to which they were accustomed. Others also felt that people could receive a faster response using digital communication and that the messages were more convenient and less judgemental. However, some liked having direct access to both healthcare providers and digital targeted client communication.

Methodological limitations
Minor concerns about methodological limitations due to poor reporting of researcher reflexivity

Coherence No or very minor concerns about coherence
Relevance Serious concerns about relevance due to a fair number of studies where participants did not experience an mHealth intervention but were asked to comment about their preferences regarding a hypothetical intervention and partial relevance of the target group Adequacy Serious concerns about adequacy due to thin data from a small number of studies  Some participants, particularly women and adolescents, had their access to phones controlled or restricted by others, especially if they had to share or borrow a phone. They noted that they would often have to explain why they wanted to use the phone, and who they wanted to call, to allay suspicions about this communication. They mentioned that this was a barrier to accessing digital targeted client communication and made it difficult to keep their messages private.

Assessment for each CERQual component
Methodological limitations Minor concerns about methodological limitations due to poor reporting of sampling (unclear how participants were recruited in several studies) and researcher reflexivity

Coherence
Minor concerns about coherence, as the majority of participants in 1 study did not see phone sharing as a problem

Relevance
Minor concerns about relevance due to a focus on study populations that may have limited access to mobile phone ownership, e.g. due to age, gender, SES, or health condition (partial relevance) Have the trialists described efforts to address situations where members of the target group: 1. do not own a functioning mobile device; 2. have poor access to network services; 3. have poor access to electricity to charge mobile devices; 4. want to avoid expenses associated with the intervention, such as paying for airtime; 5. change their phone numbers or sim cards; 6. have access to the phone controlled by someone else; 7. have low literacy, differing language skills, or limited techno-literacy; 8. have concerns about privacy and confidentiality; 9. perceive different sources as more or less reliable, trusted, and credible; 10. have members of the target group been given an opportunity to offer feedback about their needs, preferences, and experiences regarding the intervention.