Using a mHealth system to recall and refer existing clients and refer community members with health concerns to primary healthcare facilities in South Africa: a feasibility study

Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=zgha20 Global Health Action ISSN: 1654-9716 (Print) 1654-9880 (Online) Journal homepage: https://www.tandfonline.com/loi/zgha20 Using a mHealth system to recall and refer existing clients and refer community members with health concerns to primary healthcare facilities in South Africa: a feasibility study Willem Odendaal, Simon Lewin, Brian McKinstry, Mark Tomlinson, Esme Jordaan, Mikateko Mazinu, Pam Haig, Anna Thorson & Salla Atkins To cite this article: Willem Odendaal, Simon Lewin, Brian McKinstry, Mark Tomlinson, Esme Jordaan, Mikateko Mazinu, Pam Haig, Anna Thorson & Salla Atkins (2020) Using a mHealth system to recall and refer existing clients and refer community members with health concerns to primary healthcare facilities in South Africa: a feasibility study, Global Health Action, 13:1, 1717410, DOI: 10.1080/16549716.2020.1717410 To link to this article: https://doi.org/10.1080/16549716.2020.1717410 © 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group. Published online: 10 Feb 2020. Submit your article to this journal Article views: 696 View related articles View Crossmark data ORIGINAL ARTICLE Using a mHealth system to recall and refer existing clients and refer community members with health concerns to primary healthcare facilities in South Africa: a feasibility study Willem Odendaal a,b , Simon Lewin a,c , Brian McKinstry d , Mark Tomlinson e,f , Esme Jordaan g,h , Mikateko Mazinu g , Pam Haig i , Anna Thorson j and Salla Atkins j,k a Health Systems Research Unit, South African Medical Research Council, Cape Town, South Africa; b Department of Psychiatry, Stellenbosch University, Stellenbosch, South Africa; c Division of Health Services, Norwegian Institute of Public Health, Oslo, Norway; d Usher Institute of Population Health Sciences and Informatics, The University of Edinburgh, Edinburgh, UK; e Department of Global Health, Institute for Life Course Health Research, Stellenbosch University, Stellenbosch, South Africa; f School of Nursing and Midwifery, Queen’sUniversity,Belfast,UK; g Biostatistics Unit, South African Medical Research Council, Cape Town, South Africa; h Statistics and Population Studies, University of the Western Cape, Cape Town, South Africa; i Family South Africa (FAMSA) Karoo, Oudsthoorn, South Africa; j Department of Global Public Health, Karolinska Institutet, Stockholm, Sweden; k New Social Research and Faculty of Social Sciences, Tampere University, Tampere, Finland ABSTRACT Background: Lay health workers (LHWs) are critical in linking communities and primary healthcare (PHC) facilities. Effective communication between facilities and LHWs is key to this role. We implemented a mobile health (mHealth) system to improve communication and continuity of care for chronically ill clients. The system focused on requests from facility staff to LHWs to follow up clients and LHW referrals of people who needed care at a facility. We implemented the system in two rural and semi-rural sub-districts in South Africa. Objective: To assess the feasibility of the mHealth system in improving continuity of care for clients in PHC in South Africa. Method: We implemented the intervention in 15 PHC facilities. The clerks issued recalls to LHWs using a tablet computer. LHWs used smartphones to receive these requests, communicate with clerks and refer people to a facility. We undertook a mixed-methods evaluation to assess the feasibility of the mHealth system. We analysed recall and referral data using descriptive statistics. We used thematic content analysis to analyse qualitative data from semi-structured interviews with facility staff and a researcher fieldwork journal. Results: Across the sub-districts, 2,204 clients were recalled and 628 (28%) of these recalls were successful. LHWs made 1,085 referrals of which 485 (45%) were successful. The main client group referred and recalled were children under 5 years. Qualitative data showed the impacts of facility conditions and interpersonal relationships on the mHealth system. Conclusion: Using mHealth for recalls and referrals is probably feasible and can improve communication between LHWs and facility staff. However, the low success rates highlight the need to assess facility capacity beforehand and to integrate mHealth with existing health information systems. mHealth may improve communication between LHWs and facility staff, but its success depends on the health system capacity to incorporate these interventions. ARTICLE HISTORY Received 19 September 2019 Accepted 24 December 2019 RESPONSIBLE EDITOR Peter Byass, Umeå University, Sweden KEYWORDS Client referral; community-based services; continuity of care; healthcare facility; lay health workers; mobile health; primary healthcare; recall to care Background South Africa is a middle-income country with a high degree of economic inequality [1]. The disease burden of the country is characterised by high rates of communicable, non-communicable, maternal and perinatal and injury-related deaths, also unequally distributed within the population [2]. The leading causes of death are communicable diseases (33.6% of all deaths are caused by HIV/AIDS and TB), and noncommunicable diseases, such as cerebrovascular and ischaemic heart diseases, diabetes, and hypertension, account for 19.3% of deaths [2]. This epidemiological situation requires a comprehensive response at all levels of the health system, including strengthening districtbased primary healthcare (PHC) and developing innovative interventions [3]. Despite many innovations to improve the public health services in South Africa, such as the National Health Insurance scheme [4], and the PHC re-engineering programme [5], substantial challenges in service provision remain. An important challenge in South Africa is the shortage of skilled health professionals [6], which often leaves existing skilled staff overburdened and demotivated [7]. To address this challenge, a task-shifting approach is widely used in South Africa. One component of this is the use of lay health workers (LHWs) to support nurses and other healthcare professionals in implementing primary healthcare (PHC) at the community level [8]. Currently, LHWs play a critical role in extending the PHC system into communities [5]. South Africa has CONTACT Willem Odendaal [email protected] Health Systems Research Unit, South African Medical Research Council, Cape Town, South Africa GLOBAL HEALTH ACTION 2020, VOL. 13, 1717410 https://doi.org/10.1080/16549716.2020.1717410 © 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. approximately 72,000 LHWs organised in ward-based outreach teams [9]. They perform a variety of tasks, such as medication administration, child health surveys, referring people with illness symptoms to facilities, and recalling clients to for various reasons, for instance, to receive tests results or medication [10]. LHWs are also key to ensuring that clients are not lost to follow-up [10,11]. Despite these important functions being performed by LHWs, their turnover is high [12], resulting in a need for constant retraining of staff [13]. This turnover can partly be influenced by the characteristics of LHWs’work: long distances between communities and facilities to receive instructions [14–16], or to report on their work [17], that they usually have to travel on foot. This is added to by other issues, including inadequate reimbursement and poor communication between LHWs and facility staff [15]. Mobile health (mHealth) technologies, defined as medical and public healthcare practices supported by mobile devices such as mobile –and smartphones, client-monitoring devices, and tablets [18], are increasingly used in lowand middle-income countries (LMICs) to help solve health system challenges [17,19,20]. The technology has the potential to address some of the issues that LHWs experience, including traveling time, administrative tasks, and communication with facility staff [17,19–21]. There is also potential for these technologies to improve clients’access to care and potentially reduce LHWs’ workload [17,22]. mHealth interventions can also contribute to improving the continuity of care for chronic conditions [23–26]. Given this potential, we developed and implemented a mHealth system within a rural and semi –rural PHC programme in South Africa. The programme aimed to improve continuity of care for clients, while improving communication between community-based LHWs and health facilities. This paper reports on a formative evaluation of this programme, focusing on the feasibility of implementing a mHealth system to improve the continuity of care for PHC service users. The findings of the evaluation contribute to understanding the feasibility of these programmes and factors that may affect their sustainability. Methods Aim To assess the feasibility of a mHealth system to improve the continuity of care for clients in a PHC programme in South Africa. Study design Thiswasamixed-methodsevaluationstudy,implemented over 10 months, from June 2015 to March 2016. We used both quantitative and qualitative methods of data collection and analyses to assess the feasibility of the multi-component mHealth system [27,28]. This required a research team experienced in both approaches [29]. We used an embedded design in which the primary focus was on the quantitative dataset. We then used the qualitative data to understand and explain the quantitative findings [30]. Setting We planned to implement the mHealth system in arural setting, as we thought that these settings would benefit most from the improved communication offered through a mobile health system. Following consultations, the Western Cape Department of Health suggested that the study be conducted in two rural sub-districts in the Eden Health district (Flowchart 1), one of the seven health districts in the Western Cape Province of South Africa. These two study sub-districts were selected because each already had a well-functioning team of LHWs. TheEdendistricthasatotalestimatedpopulation size of 613,124 [31]. The two study subdistricts represented approximately 21% of the total district population and were substantially different in size (the population of Sub-district 1 estimated at 26,064 and Sub-district 2 at 101,298) [32,33]. Most residents’first language is Afrikaans (73%), followed by isiXhosa (25%) and English (2%) [34]. In 2016, 40.5% of the population in the district lived below the poverty line (USD 320 per month) [31]. The Eden district performs slightly worse on key health indicators, compared to the Provincial average. For example in 2015, 16% of newborns were underweight in the district, compared to the 14.5% average in the Western Cape, and the maternal mortality rate was higher in Eden (69.9/100 000 live births) when compared to the Provincial rate (58.3/100 000) [31]. PHC facilities in the Eden sub-districts provide basic healthcare services, including treatment for TB, HIV/AIDS and non-communicable diseases, and maternal and child health services. The hospital in Sub-district 1 that participated in the study provides maternity services, basic surgery, and emergency services. Clients who needed to consult a doctor had to book an appointment in advance as doctors visited the respective facilities only on certain days of the month. Participants As noted earlier, each of the selected two study subdistricts had a well-functioning team of LHWs (Flowchart 1), and most of these LHWs lived in the 2W. ODENDAAL ET AL. communities in they worked. In each of these districts, the Provincial Department of Health had contracted a non-governmental organisation (NGO) to recruit, manage and pay LHWs to provide community-based PHC services on behalf of the Department of Health. This form of contracting out is common in the Western Cape Province [35]. The LHWs received non-professional training on various topics related to the services they provide, which include supporting clients whom the health facility assigns to them as well as health promotion activities. The LHWs worked 4.5 h a week and earned between 99 and 122 US$ per month, depending on their level of training. At the time of the study, the NGO in Subdistrict 1 had two LHW teams (29 LHWs in total), and the NGO in Sub-district 2 had four teams (64 LHWs in total). Each team was supported by a supervisor, who was a retired nurse. Demographic data for the LHWs and supervisors within each subdistrict is detailed in Table 1. The mHealth system The three key role players involved in the implementation of the mHealth system were the LHWs, their supervisors, and mHealth clerks, henceforth referred to as clerks. Each facility manager appointed one staff member, in most cases from the administrative staff, to act as clerk. Managing the mHealth system was in addition to the clerks’other duties. LHWs were given smartphones for the project, and clerks and nurses tablets, to manage the system. The main feature of the system was to enable LHWs to record their routine client visits, that is monitoring how they were doing on treatment, do pill counts, and conducting general health assessments in clients’ households, on project-funded smartphones. The system enabled real-time access for supervisors to these reports. The system also enabled two-way communication between LHWs and clerks. LHWs, supervisors and clerks received 2 days of training on how to use the system, offered by a for-profit mHealth service provider (Mobenzi), who developed the system. Thereafter, the implementation staff had 2 weeks to practice using the system before the system went live. Though there were a paper-based recall and referral system in use before the intervention, it was not standardised. The mHealth system could be considered a completely new system to the participating facilities. The mHealth recall and referral process were as follows (Figure 1): Firstly, the healthcare professionals at the facility instructed the clerk to ask an LHW to locate and advise a client to return to the facility (hereto referred to as recalls, Figure 2). The clerks issued these requests through the tablet, and these were received by the LHW on their smartphones while working in the community. Real-time communication ensued between clerks and LHWs when they discussed recall progress using the system. The supervisors’role in the recall process was added in month 5 of the implementation, after this was Table 1. Supervisor and LHW demographics. Sub-district 1 Sub-district 2 Supervisors LHWs Supervisors LHWs Female 2 29 4 62 Male - - - 2 Average age 58 35 58 34 Age range 55 –63 22 –59 47 –68 23 –60 Average number of years working as supervisor or LHW 8 6 3 5 Years in post –range 4 months –7 years 1 –10 years 6 –10 years 3 –10 years No demographic data were collected about the clerks. Eden district with 7 sub-districts Sub-district 1 Sub-district 2 6 Primary healthcare (PHC) facilities including 1 hospital 9 PHC facilities NGO 1 2 Supervisors 29 LHWs NGO 2 4 Supervisors 64 LHWs Supervisor 3 17 LHWs Supervisor 2 18 LHWs Supervisor 1 17 LHWs Supervisor 1 12 LHWs Supervisor 2 17 LHWs Supervisor 4 12 LHWs Flowchart 1. Settings in which the study was implemented. GLOBAL HEALTH ACTION 3 requested by them. The late addition was due to preproject consultations suggesting that this functionality was not necessary. Secondly, when LHWs identified a person, who could have been an existing client or someone else in the community with a health problem, for example, headaches or wounds requiring care, they would advise that person to seek care at the facility (hereto referred to as referrals,Figure 3). LHWs sent a notification of these instances to the clerk’s tablet, using their smartphones. The clerks closed recalls and referrals, respectively, as successful, when the person arrived at the facility, or unsuccessful when the person failed to attend at the facility. The digital health interventions included in the mHealth system evaluated in this study targeted healthcare providers and can be classified as follows, using the World Health Organization’s classification of digital health interventions [36]: ●Interventions focused on client health records: Longitudinal tracking of clients’health status and services (classification number 2.2.1) ●Interventions focused on healthcare provider decision support: Provide checklist according to protocol (classification number 2.3.2) ●Interventions focused on healthcare provider communication: Communication from healthcare provider(s) to supervisor (classification number 2.5.1) ●Interventions focused on referral coordination: Manage referrals between points of service within health sector (classification number 2.6.2) Figure 1. mHealth recall and referral system. 4W. ODENDAAL ET AL. ●Interventions focused on health worker activity planning and scheduling: Schedule healthcare provider’s activities (classification number 2.7.2) Data collection The main question we wanted to answer using quantitative data was whether the mHealth system allowed facility staff and LHWs to, respectively, recall existing clients and refer community members with health concerns to healthcare at the health facilities. All recall and referral data via the mHealth system in the two study sub-districts were stored on a Mobenzi server and exported to Excel by the research team. The data included a date and time stamp, sender and recipient, geographic location, and content of the messages between LHW and clerk. As reasons for recalls and referrals were recorded without predetermined categories, the research team manually coded the recall and referral reasons, categorising them according to most frequent reasons –the codes developed are shown in Table A1. As we collected service indicators that were not in use in the study sites prior to this study, baseline data were not available. The qualitative component of the study aimed to provide an understanding of the implementation processes and how the participants perceived the mHealth system. These data were intended to help us contextualise the quantitative findings through incorporating participant perspectives. We used two methods of qualitative data collection: semistructured individual and group interviews, conducted at the end of the project with all of the participating LHWs, supervisors, clerks, and facility managers; and a fieldwork journal kept during the implementation of the study. WO collected the data. The interview questions included how LHWs, supervisors, and clerks felt about using the mHealth system; whether this system changed their practices; their views regarding barriers and facilitators to implementation; and how the mHealth system compared to the paper-based system. We invited the facility managers to join the interviews with the clerks, as it was important for us to ascertain their perceptions, experiences, and recommendations, too. We include the findings from seven clerk/manager interviews, as these cadres were key to the implementation of the mHealth system within the facility, and had a good overall view of the implementation processes. In total, 12 of the 15 clerks, and three of the eight facility managers participated in the interviews. In some instances, facility managers were responsible for two facilities, and the four participating mobile facilities were managed by some of the ‘fixed facility’ Figure 2. Example of the recall format of correspondence between the LHW and clerkfacility. The figure shows pseudonyms for client, LHW and facility, and LHWs in the study were referred to as community care workers (CCWs). Figure 3. Example of the format for a referral sent by an LHW to the health facility. The figure shows pseudonyms for client, LHW and facility, and LHWs in the study were referred to as community care workers (CCWs). GLOBAL HEALTH ACTION 5 managers. The remaining three facilities and clerks were not available at the time that WO conducted these interviews. The staff of some facilities were interviewed together as this was the most convenient approach for gathering data from staff in remote, neighbouring facilities. The fieldwork journal detailed the researcher’s reflections and observations during the fieldwork visits. For instance, visiting the LHWs in the most remote areas highlighted the challenges of regular contact with, and reporting to, supervisors based in the main towns in the respective sub-districts. Analysis Recalls were categorised as successful if the clerk recorded that the client attended the facility as requested. Failed recalls included the following categories: (i) clients who failed to attend the facility as requested; (ii) unclosed recalls, i.e. where either the clerk or LHW did not respond to the other’slatest correspondence; (iii) LHWs who did not view the recall; and (iv) clerk errors, for example, when the clerk sent the request to the wrong LHW. There were no time limits on keeping recalls open. Referrals were categorised as successful if a person attended the facility for the reason he/she was referred by the LHW. Referrals expired within 14 days of being issued and were then categorised as failed referrals. We collected data from June 2015 until February 2016, the second last month of the study, to ensure that recalls and referrals issued in February could be acted upon by the end of the study in March 2016. We used Excel and R statistical software (https:// www.r-project.org/) for the descriptive statistics, calculating facilityand sub-district level averages of recall and referral numbers and success rates, and to present participant demographics. In order to understand and contextualise our results, we applied a qualitative content analysis approach to the interview data from clerks and facility managers [37]. We used Atlas.ti version 8.1 (https://atlasti.com/product/ v8-windows/) to conduct this content analysis. As we were primarily interested in explaining the recall and referral outcomes, we analysed the interview data deductively at the manifest level [37]. We used both the apriori themes from the interview guides, including, e.g. barriers and facilitators to implementation, while allowing additional themes to emerge from the data to explain our quantitative results. SA read and reread the transcripts to familiarise herself with the data and generated condensed meaning units from the data. These meaning units were then further reduced to codes, from which categories were generated that related directly to the quantitative data, see Table 2 for an example of the analysis. The analysis was checked by WO, and differences resolved by discussion. WO referred to the fieldwork journal as the interviews were being analysed, looking for content that could illuminate and further explain the Table 2. Example of qualitative analysis. Categories Sub-categories Example codes Examples of extracts from the data Lack of support and time ●Need technical support ●Maintaining dual systems ●Additional workload ●After hour recalls LHWs had difficulties in maintaining two systems ‘Some of them struggled to use paper and the phone. So I either carry on with the paper or I carry on with the phone …’ Communication and interpersonal challenges ●Users struggled adapting to mHealth ●LHW-facility staff relationships LHW: difficulties in the beginning because not familiar with smart phone ‘like I’ve said, in the beginning when I first went for this training I was nervous. Because I have never used something like this in my life. The only time I might have used a phone, then it was a phone with buttons. But it’s not these modern phones.’ Closing recalls and referrals ●Facility staff communication delay ●System description ●Immediacy NC: mH offered immediacy to LHW-facility communication ‘And when I walk in there the sister would quickly tell me, listen, I’ve passed on this or that to the tablet, or I’ve given this and that to that person. And almost miraculously, when I get to those carers and I’ve already got it from the sister early in the morning, then I get to the CCWs: you know what, I got this and that referral.’ Effects of mHealth ●Remote settings ●Improved communication and reporting ●Improved recalls ●Paper vs mHealth NC: very difficult to get data from remote living LHWs “…I got those every second week. I had to arrange it with them because we have to stick to our petrol budget; in other words, I can’t drive to [Place A] and [Place B] every week. I was able to go about once a month, and then I had it delivered every second week at the [Place C], that’s the satellite, so I had an arrangement with that man that I would collect it on a Friday.” Categories, sub-categories and example codes and quotes (sub-category in bold where codes and quotes are related). 6W. ODENDAAL ET AL. views that participants shared during the interviews. The qualitative and quantitative data are reported in parallel in the results below. Ethics, consent and permissions Prior to conducting this study, ethical approval was obtained from the South African Medical Research Council (EC016-11/2014). The approval included the participant information sheet and informed consent form signed by all participants. All interviewees provided signed informed consent, and the interviews took place in a private space in their respective facilities, at a time that was convenient for them. The interviews were conducted in the language preferred by participants, which was predominantly Afrikaans. Interviews were audio-recorded, transcribed and translated into English. Before the study commenced, we provided the LHWs with an information flyer in plain language and asked them to use this when describing the study to their clients. LHWs were instructed during the training to only use the mHealth system after clients were briefed and had given consent to participate. Results The quantitative results for recalls of PHC clients and referrals of clients and community members are presented below, according to success rates. The categories emerging from the qualitative data –lack of support and time, communication and interpersonal challenges, closing recalls and referrals, and the effects of mHealth –are reported together with the quantitative data. We first present the recall results and then discuss the referral results. Recalls In total, 2,204 client recalls were issued across the two sub-districts, of which 28% (628/2,204) resulted in clients attending at the facility as requested (Table 3). Most recalls were initiated in Sub-district 2 (1642, 74% of total recalls) as could be expected due to the larger sub-district population. However, the recall success rate in Sub-district 1 was 53% (301/ 562), compared to 20% (372/1,642) in Sub-district 2. The two most common recall categories across the two sub-districts were recalling children under five (23%, n = 514), for example, for growth or nutrition monitoring or vaccinations, and facility appointment reminders (10%, n = 230), that is reminding clients of upcoming appointments, and in many cases, of rescheduled appointments (see Table A1 for further detail). Table 3 shows the recall success rates according to the recall reasons. There was a high number of medication collection recalls in Sub-district 2 (21% of the total recalls) because of fewer community-based medication dispensing outlets available for the population than what was available in Sub-district 1. Many Subdistrict 2 clients, therefore, had to collect their medication at the nearest health facility and thus became part of the mHealth recall system. The high number of facility appointment reminder recalls in Subdistrict 1 (27% of total recalls in that sub-district) were due to frequent doctor appointment rescheduling (personal communication, facility manager, 8 April 2016). Qualitative data suggested that clerks Table 3. Recall success rates according to the recall reason. Reason for recall Subdistrict 1 N (%) Sub-district 2 N (%) Total across subdistricts N (%) a Children < 5 years (e.g. deworming, Vitamin A, immunisation) Successful recalls 101 (64%) 75 (21%) 176 (34%) Total recalls 158 356 514 (23%) Diagnostic tests (being tested or receiving results for all conditions excluding TB/HIV and AIDS) Successful recalls 22 (44%) 11 (9%) 33 (20%) Total recalls 50 117 167 (8%) Finding defaulting clients (including TB/HIV clients) Successful recalls 4 (44%) 11 (15%) 15 (19%) Total recalls 9 72 81 (4%) Medication collection Successful recalls 16 (55%) 51 (15%) 67 (17%) Total recalls 29 348 377 (17%) Non-communicable disease care Successful recalls 5 (83%) 7 (39%) 12 (50%) Total recalls 6 18 24 (1%) Obstetrics/Gynaecology (including family planning) Successful recalls 4 (27%) 3 (5%) 7 (9%) Total recalls 15 62 77 (3%) Reminding clients about facility appointments Successful recalls 86 (57%) 9 (11%) 95 (41%) Total recalls 150 80 230 (10%) TB/HIV/AIDS care Successful recalls 8 (44%) 10 (16%) 18 (21%) Total recalls 18 64 82 (4%) Other b Successful recalls 55 (43%) 150 (29%) 205 (31%) Total recalls 127 525 652 (30%) Total Successful recalls 301 (53%) 327 (20%) 628 (28%) Total recalls 562 (26%) 1,642 (74%) 2,204 (100%) a The % reported for the total of each recall reason is the proportion of the total recalls across sub-districts. b Included a range of health issues, such as wound care, eye care, having to see the occupational therapist or social worker, and mental healthcare. It also included unspecified reasons, when the recall/referral simply stated that the client needed to seek care at the facility. GLOBAL HEALTH ACTION 7 thought that the mHealth system supported this rescheduling well, as it speeded up messages getting to clients via LHWs, adding to the system’s feasibility. To facilitate ownership of the mHealth system, we decided in advance to allow clerks and LHW teams to adapt the system in any way that made it easier for them to use. The interview data suggested that this happened in relation to the process of issuing recalls, which differed across facilities. In some facilities, recall requests were made through a meeting among the healthcare professionals, while in others, clerks received a stack of folders or were told verbally or through stickers or book notes to recall clients. Facility staff also noted that when the period for facility audits, i.e. reporting on how well the facility performed against service targets, was approaching, certain client categories would be prioritised for recalls, to meet these targets. One clerk noted the following when asked about the improvement in recall numbers: Audits! …When an audit is coming, then the patients are called in because we are worried that we won’t get it [the targets] right. (Sub-district 1, Facility 1 clerk) The qualitative data from over half of the participating facilities indicated that interpersonal relationships and communication patterns in the facility and with LHWs could impact substantively on closing recalls, and thus impacted on the feasibility of the mHealth system. As reported below, interpersonal conflicts meant that information about clients arriving was not communicated, hampering system implementation: The communication between us [clerks and healthcare professionals] was just not right to say that the patient did come back….(Sub-district 2, Facility 1 clerk) Referrals A total of 1,085 referrals were recorded across the two sub-districts, of which 45% (485/1,085) were successful (Table 4). Sub-district 1 had 84 referrals (8% of the total referrals), compared to the 1,001 referrals (92% of the total referrals) in Sub-district 2. The success rate in Subdistrict 1 was 33% (28/84), compared to the 46% (457/ 1,001) in Sub-district 2. Table 4 shows the referral success rates according to the referral reasons. Data from the implementation journal suggested that one facility in Sub-district 2 had a higher referral success rate than the others because the nursing staff had prioritised LHW referrals. Clients arriving at this facility were, therefore, more closely followed by clerks. This was corroborated by the interview data, where some clerks indicated that certain referrals and recalls were made and closed based on priorities and targets set by the facility –and sub-district management: Every month they [management] check …if there are reports that have to be signed off …So she [the clerk] knew that these had to go out. (Sub-district 1, Facility 2 clerk) The interviews suggested several reasons why success rates for recalls and referrals were low. Closing the Table 4. Referral success rates according to the reason for referral. Subdistrict 1 N (%) Sub-district 2 N (%) Total across subdistricts N (%) Reason for referral Total (%) a Children < 5 years (e.g. deworming, Vitamin A, immunisation) Successful referrals 8 (22%) 172 (57%) 180 (53%) Total referrals 36 303 339 (31%) Diagnostic tests (being tested or receiving results for all conditions excluding TB/HIV and AIDS) Successful referrals 2 (100%) 8 (29%) 10 (33%) Total referrals 2 28 30 (3%) Finding defaulting clients (including TB/HIV clients) Successful referrals 1 (50%) 19 (34%) 20 (34%) Total referrals 2 56 58 (5%) Male medical circumcision Successful referrals 0 5 (29%) 5 (29%) Total referrals 0 17 17 (2%) Medication collection Successful referrals 6 (32%) 30 (38%) 36 (37%) Total referrals 19 79 98 (9%) Non-communicable disease care Successful referrals 2 (50%) 5 (28%) 7 (32%) Total referrals 4 18 22 (2%) Obstetrics/Gynaecology (including family planning) Successful referrals 0 (0%) 50 (56%) 50 (56%) Total referrals 3 90 93 (9%) Physical symptoms Successful referrals 2 (33%) 38 (45%) 40 (44%) Total referrals 6 85 91 (8%) Reminding clients about facility appointments Successful referrals 2 (100%) 5 (31%) 7 (39%) Total referrals 2 16 18 (2%) TB/HIV/AIDS care Successful referrals 0 30 (65%) 30 (65%) Total referrals 0 46 46 (4%) TB/HIV/AIDS testing Successful referrals 2 (67%) 60 (43%) 62 (33%) Total referrals 3 140 143 (13%) Other b Successful referrals 3 (43%) 35 (28%) 38 (29%) Total referrals 7 123 130 (12%) Total Successful referrals 28 (33%) 457 (46%) 485 (45%) Total referrals 84 (8%) 1,001 (92%) 1,085 (100% a The % reported for the total of each referral reason is the proportion of the total referrals across sub-districts. b Included a range of health issues, such as wound care, eye care, having to see the occupational therapist or social worker, and mental healthcare. It also included unspecified reasons, when the recall/referral simply stated that the client needed to seek care at the facility. 8W. ODENDAAL ET AL.