Factors Affecting the Implementation of Electronic Antiretroviral Therapy Adherence Monitoring and Associated Interventions for Routine HIV Care in Uganda: Qualitative Study

Background High, sustained adherence to HIV antiretroviral therapy (ART) is critical for achieving viral suppression, which in turn leads to important individual health benefits and reduced secondary viral transmission. Electronic adherence monitors record a date-and-time stamp with each opening as a proxy for pill-taking behavior. These monitors can be combined with interventions (eg, data-informed adherence counseling, SMS-based adherence support, and/or alarms) and have been shown to improve adherence in multiple settings. Their use, however, has largely been limited to the research context. Objective The goal of the research was to use the Consolidated Framework for Implementation Research (CFIR) to understand factors relevant for implementing a low-cost electronic adherence monitor and associated interventions for routine HIV clinical care in Uganda. Methods We conducted in-depth qualitative interviews with health care administrators, clinicians, and ART clients about likes and dislikes of the features and functions of electronic adherence monitors and associated interventions, their potential to influence HIV care, suggestions on how to measure their value, and recommendations for their use in routine care. We used an inductive, content analysis approach to understand participant perspectives, identifying aspects of CFIR most relevant to technology implementation in this setting. Results We interviewed 34 health care administrators/clinicians and 15 ART clients. Participants largely saw the monitors and associated interventions as favorable and beneficial for supporting adherence and improving clinical outcomes through efficient, differentiated care. Relevant outside factors included structural determinants of health, international norms around supporting adherence, and limited funding that necessitates careful assessment of costs and benefits. Within the clinic, the adherence data were felt likely to improve the quality of counseling and thereby morale, as well as increase the efficiency of care delivery. Existing infrastructure and care expenditures and the need for proper training were other noted considerations. At the individual level, the desire for good health and a welcomed pressure to adhere favored uptake of the monitors, although some participants were concerned with clients not using the monitors as planned and the influence of poverty, stigma, and need for privacy. Finally, participants felt that decisions around the implementation process would have to come from the Ministry of Health and other funders and would be influenced by sustainability of the technology and the target population for its use. Coordination across the health care system would be important for implementation. Conclusions Low-cost electronic adherence monitoring combined with data-informed counseling, SMS-based support, and/or alarms have potential for use in routine HIV care in Uganda. Key metrics of successful implementation will include their impact on efficiency of care delivery and clinical outcomes with careful attention paid to factors such as stigma and cost. Further theory-driven implementation science efforts will be needed to move promising technology from research into clinical care. Trial Registration ClinicalTrials.gov NCT03825952; https://clinicaltrials.gov/ct2/show/NCT03825952


Introduction
High and sustained adherence to antiretroviral therapy (ART) is critical for achieving HIV viral suppression, which in turn leads to important individual health benefits and reduced secondary viral transmission [1,2]. However, the attention paid to adherence in clinical settings varies. Health care providers may ask about missed doses or perform pill counts, but these measures tend to overestimate adherence [3][4][5]. Pharmacy refill data is less biased and can improve detection of nonadherence [6]; yet all of these adherence measures are obtained with clinic visits and potentially after viremia has led to drug resistance [7]. Electronic adherence monitoring involves smart pill containers that record a date-and-time stamp with each opening as a proxy for pill-taking behavior. Real-time versions of electronic adherence monitors contain modems that transmit these data via cellular networks for internet-based review. Although limited by the need to use the monitor for each dose, these monitors provide daily adherence records that can trigger timely adherence interventions, potentially before the loss of viral suppression.
Several studies have suggested the effectiveness of electronic monitoring for adherence support. One study conducted in China found an increase in ART adherence when short message service (SMS) reminders triggered by real-time detection of missed doses were combined with data-informed counseling (ie, adherence records were used at clinic visits to develop solutions to adherence challenges) [8]. A similar study of triggered SMS reminders in South Africa (that did not include supported counseling) observed a decrease in sustained treatment interruptions [9]. Further, a randomized controlled trial in Uganda tested real-time adherence monitoring plus SMS reminders to patients and SMS notifications to social supporters (ie, friends or family who could help support adherence) and found improved average adherence and a reduction in sustained interruptions [10]. Other studies have used non-real-time electronic adherence monitoring data to inform counseling and also found increased ART adherence [11,12]. Importantly, these devices have been shown to be feasible and acceptable in these and other settings, although sometimes with technical challenges [13].
Despite this evidence of improved adherence and promise for HIV outcomes, electronic adherence monitors and associated interventions have largely remained in the research context-a fate common among mobile health (mHealth) interventions [14]. One clear initial barrier to implementation has been cost. Electronic monitors have traditionally cost more than US $100 each and require data transmission and hosting fees; SMS may necessitate additional development and other fees. Recently, a low-cost electronic monitor with integrated SMS messaging was developed with total costs of less than US $30 per patient per year. To contextualize these costs, a modeling analysis found that adherence monitoring-based interventions could be considered cost-effective in sub-Saharan Africa at up to $50 per person-year [15]. Intervention adoption, however, is influenced by many factors other than cost and can be holistically considered through the Consolidated Framework for Implementation Research (CFIR) [16]. The CFIR includes 5 domains: (1) intervention characteristics (eg, design, cost), (2) outer setting (eg, organizational knowledge of patient needs, external policies), (3) inner setting (eg, culture, relative prioritization), (4) individual characteristics (eg, beliefs about the intervention, self-efficacy), and (5) process (eg, planning for implementation, engaging leaders).
Here we present an exploratory analysis guided by CFIR and involving ART clients, clinicians, and health care administrators in which we sought to understand factors relevant for implementing electronic adherence monitoring and associated interventions for routine HIV clinical care in Uganda.

Study Setting
This study was based at the Kabwohe Clinical Research Centre (KCRC) in rural southwestern Uganda. The KCRC ART Clinic provides PEPFAR-subsidized care for more than 6000 individuals living with HIV. It is a health center level IV facility, which also provides other comprehensive primary health care services and is governed by the Ministry of Health (MoH) and a district health officer. Specialized care is available through regional and national referral hospitals, and community-level care is provided through lower-level health facilities and community health workers. The research team met with a local community advisory board and KCRC leadership prior to initiating the study and incorporated their feedback in the study design.

Study Participants
We stratified ART clients by duration of ART use (less than vs more than 6 months) and residence type (rural vs periurban); within these categories, we identified clients randomly (ie, every 10th patient attending clinic) to understand the average experience in the clinic. Given the hierarchical nature of the Ugandan health care system, we identified up to 5 health care administrators/clinicians from each of the following cadres: MoH officials; regional referral hospital administrators; district health officers; and health center III/IV clinic administrators, physicians, nurses, and ART adherence counselors. Inclusion criteria for all participants were aged over 18 years and engagement in HIV care through one of the above-defined roles. Additionally, ART clients had to have HIV infection per clinic records and own a cellular phone (familiarity with cellular technology was felt important to inform their input on the intervention). Exclusion criteria for all participants were unwillingness or inability to provide informed consent.

Electronic Adherence Monitor and Associated Interventions
We studied the evriMED electronic adherence monitor (Wisepill Technologies, Figure 1), which can function with or without real-time data transmission. It can be paired with any combination of the interventions presented in Table 1.

Description Intervention
Records of monitor openings are reviewed on a smartphone, tablet, or computer and discussed at each clinic visit to identify specific challenges and develop effective solutions to overcome future adherence barriers.
Data-informed counseling SMS messages are sent to patients daily to encourage adherence (eg, through establishing the habit of daily pill taking and/or reminding patients that the clinic supports them). SMS are sent regardless of the recorded adherence. One-way scheduled SMS a to patients When real-time monitors are used, SMS are sent to patients when one or more doses are taken late or missed. The SMS are sent to the patient and/or a social supporter (ie, a person who knows the patient's HIV status and is willing to provide support).

One-way triggered SMS
Both scheduled and triggered SMS allow for a callback from study staff to provide support directly at that time. Two-way SMS Monitors are programmed to make audio-visual alerts when it is time to take medication. Alarms a SMS: short message service.

Qualitative Interviews
Interviews were conducted by authors JBT and RB, who are both bilingual in the local language (Runyankole) and English and highly experienced, well-trained male qualitative research assistants. One-time interviews were digitally recorded for later transcription and took place in private settings; most occurred in the study office or near the clinic, although all MoH interviews were conducted in the participants' offices in Kampala, and some participants were interviewed at home, per their preference. Interviews with health care administrators/clinicians were conducted in English, which is commonly used in professional settings; interviews with ART clients were conducted in Runyankole or English per participant preference. Interviews began with an introduction to the research assistants, followed by statements of no conflicts of interest, a desire for honest perceptions (favorable or unfavorable), and the overall goals of the study. Participants were then asked for basic demographic data. A description of the electronic adherence monitors, associated interventions, evidence for their use, logistical requirements, and costs was subsequently read to participants (Multimedia Appendix 1). Participants were also shown an electronic adherence monitor and the software interface for displaying adherence data ( Figure 2). Interview guides (Multimedia Appendix 2 and 3) were designed to obtain unbiased impressions of the technology and its potential for supporting ART in routine care, while also assessing each of the 5 domains in the CFIR. The guides were tailored for anticipated perspectives of health care administrators/clinicians versus ART clients. Initial questions in both guides asked about likes and dislikes of the features and functions of monitor and associated interventions and were followed by questions about their potential to influence HIV care, suggestions on how to measure their value, and recommendations for their use in routine care. Health care administrators/clinicians were also asked about the technology in relation to other health care priorities (ie, the outer setting). Questions were informally pretested with KCRC staff and clients and revised to ensure clarity and utility. Research assistants wrote debriefs after each interview to capture body language, participant mood, and any other nonverbal aspects of the interviews. Transcripts were reviewed for quality among authors LG, BFB, JBT, and RB and corrected as needed. Participants were interviewed until thematic saturation was achieved.

Analysis
We used an inductive, content analysis approach [17] to explore factors that could influence the implementation of electronic adherence monitors plus associated interventions. We identified the aspects of CFIR [16] that participants indicated were most relevant to their context and potential implementation of the technology. In an iterative process, authors LG and JEH read the first 20% of transcripts, formulated codes, and assembled and pilot-tested a codebook. LG subsequently used the codebook to code the qualitative data, which was entered into qualitative analysis software (Dedoose, SocioCultural Research Consultants LLC). JEH and LG then developed categories by characterizing core concepts, developing labels, writing operational definitions, and selecting illustrative quotes from the interviews. Themes were reviewed with the qualitative research assistants but not participants to ensure accurate reflection of the participants' stated perspectives.

Ethics
All participants provided written informed consent. This study was reviewed and approved by the institutional review boards at the Mbarara University of Science and Technology, Ugandan National Council for Science and Technology, and Partners Healthcare. The study was registered with ClinicalTrials.gov [NCT03825952].

Participant Characteristics
We interviewed 34 health care administrators/clinicians with a mean age of 37 (SD 10) years; 56% (19/34) were female. Four administrators worked in the MoH, while 5 worked in regional referral hospitals and 5 at the district level; 5 clinicians each were doctors, nurses, and adherence counselors. We also interviewed 15 ART clients with a mean age of 40 (SD 13) years; 60% (9/15) were female, and 67% (10/15) had taken ART for less than 6 months. A total of 60% (9/15) lived in periurban settings, and 40% (6/15) lived in rural settings. None of the individuals approached for participation declined. Interviews lasted an average of 51 minutes. Table 2 presents the main factors participants felt would influence implementation of the adherence monitors and associated interventions. The factors are organized within the 5 domains of CFIR. Thematic saturation was achieved with the 49 study participants. Note that all perspectives reflect hypothetical use of the monitors and associated interventions.

Intervention
Both health care administrators/clinicians and ART clients uniformly stated that the value of the adherence monitor and associated interventions would be seen in their ability to improve knowledge of and support for adherence. They were seen as better than the current, limited approach to adherence monitoring. To enable uptake of the monitoring and associated interventions, participants also highlighted the need to provide in-service training. Opinions differed as to whether additional staff and/or infrastructure would be needed. Within the clinic setting, participants stated that the adherence monitors and associated interventions could have a significant impact on clinical care delivery, which they felt should be considered in any implementation plans.

Individual
Both categories of participants also identified several individual-level characteristics that could impact implementation. For instance, ART clients who prioritized improved health expressed enthusiasm to use the intervention. They described a welcomed pressure to adhere.
What I have liked is that it may report you to the clinic that you are not taking your drugs well, and as a patient, this will force you to take your drugs well and live longer. [ART client, male, 37 years (ITA-1-005)] Additionally, clients indicated that the monitors would enable them to demonstrate good adherence to the health care workers. This motivation reflected their appreciation of the care they receive in clinic.
I will know that the counselor is trying to help me on the basis of that data, to make sure that I live a healthy life and I cannot feel bad about it because I will know that she/he cares about my life... This counseling will help me to change my behavior because I will know that they will keep posted with my adherence. [ART client, female, 48 years (ITA-1-012)] That said, both categories of participants felt inaccuracies may arise in the data if clients do not use the monitors as planned, which would limit the value of the adherence data.
He/she might remove the pills, take them somewhere, and forgets or some time passes before he/she swallows the medicine. I would feel there would be a camera or a way of being sure that the client has taken the drugs after opening and removing them.
[Clinic administrator, female, 34 years (HCA-1-020)] Poverty was seen as a principal influencing factor. Many participants felt clients could not pay for the monitors or associated SMS or pay for electricity to charge the monitors or their cell phones to receive or send SMS. Participants were also concerned that both traditional illiteracy and technical illiteracy (ie, ability to use technology)-two indirect effects of poverty-would limit some clients' ability to use the SMS.
[SMS] reminders in Uganda have not really yield the positive, because we have had reminders to mothers to attend antenatal... but the issue is that the phone must be charged, and the person must be with the phone and able to read the message. Nearly all participants commented on the potential for stigma in the event of lost privacy (eg, others seeing the monitors or SMS); however, views differed on the importance of these factors in using the intervention. Opinions seemed to stem from the extent to which clients accepted their HIV diagnosis and disclosed their status to others.
So you have to educate them on the advantages and functions of these monitors. If they feel out place carrying them, they may feel stigmatized. They may feel segregated and end up not using them. [ Participants felt that these individual-level factors were critical for uptake of the monitors and associated interventions. Challenges would be important to address through counseling and education.

Process
Consistent with the above-noted emphasis on cost-effectiveness, health care administrators/clinicians described funders as playing a key role in the implementation process. Support and resources were expected from the MoH and/or other organizations contracted for HIV services delivery in Uganda (eg, USAID-supported Elizabeth Glazer Pediatric AIDS Foundation). Most ART clients and clinicians reported having insufficient funding themselves to implement the monitors or interventions.
The clinic has no funding specifically to buy this [evriMED]  Health care administrators/clinicians also indicated that sustainability of the monitors and associated interventions would depend on the target population. Opinions on the target population, however, varied widely in both categories of participants. Some suggested including all ART clients, while others wanted to select those with high risk or documented adherence challenges. Participants felt that large numbers of clients would limit implementation, although excluding clients could also cause challenges. Participants took a holistic view of the implementation process, including all of the above-noted factors related to the intervention, individual, inner setting, and outer setting.

Principal Findings
This qualitative study of ART clients, health care administrators, and clinicians explored factors that may influence the implementation of low-cost electronic ART monitors and associated interventions for routine HIV care in Uganda. To our knowledge, this study is the first to use an implementation science framework to explore the means to move a technology-based adherence intervention from research into clinical care. The intervention was largely seen as favorable and beneficial for improved clinical outcomes with efficient, differentiated care. Concerns centered primarily around potential for stigma, device misuse, possible need for additional resources, and cost in the setting of competing demands for limited resources.
Participants identified improvements for the intervention that could address some of their concerns. For instance, stigma could be reduced by simple alterations, such as variable monitor sizes and making alarms optional. Counseling specifically around disclosure could also alleviate concerns for stigma, and education could support fidelity of use, even with low levels of literacy among clients. Within the clinic, leveraged use of existing infrastructure and staff would increase efficiency and reduce concerns about cost. And, perhaps most importantly, demonstrated value in clinical outcomes for the lowest possible cost could position implementation well against competing demands for limited resources in HIV care. These insights are particularly valuable, as most research on electronic adherence monitoring has focused only on measurement, even when considering the context of routine clinical care [18].
The potential value of an improved approach to adherence monitoring and support was endorsed by all participants. Steady progress is being made toward the UNAIDS 90-90-90 goals, and many are already achieving the high adherence necessary for viral suppression. Yet sustained adherence remains a challenge over time. Up to one-third of clients in sub-Saharan Africa are viremic at 2 years of therapy [19] and similar numbers have stopped ART and been lost from care at 5 years [20]. Increased adherence monitoring and support could play a critical role in reaching the 10-10-10 currently eluding the current care models.
Several tuberculosis treatment programs globally have begun to implement electronic adherence monitoring and support as part of routine care [21], although little evidence has been published on the implementation of these approaches. Studies driven by implementation science frameworks could facilitate uptake, assess for fidelity of implementation, and understand the impact on clinical outcomes. These data will be critical to determine how well potential benefits translate into real-world settings. Indeed, some preliminary reports with other digital adherence monitoring approaches suggest challenges with patient engagement and accuracy [22] that will need to be systematically addressed.

Limitations
This study has limitations. First, although we interviewed health care workers and administrators from all levels of the health care system, the ART clients came from a single site. That said, KCRC and the client population characteristics are largely reflective of ART delivery in rural East Africa. Second, study findings reflect perceived views that were not influenced by actual use of the monitors or associated interventions. Future work will involve deployment of the technology with subsequent reflections on the implementation process. Strengths of the paper include use of a comprehensive implementation science framework and in-depth exploration of factors relevant for the implementation process.

Conclusions
In conclusion, we found that low-cost electronic adherence monitoring combined with data-informed counseling, SMS-based support, and/or alarms has potential for use in routine HIV care in Uganda. Key metrics of successful implementation will include their impact on efficiency of care delivery and clinical outcomes with careful attention paid to factors such as stigma and cost. Given that most interventions fail to progress from research to practice, further theory-driven implementation science efforts will be needed to realize the benefits of this promising technology.