Digital directly observed therapy to monitor adherence to medications: a scoping review

Abstract Background: Technology-based directly observed therapy (DOT) is more cost-effective and efficient compared with in-person monitoring visits for medication adherence. While some evidence shows these technologies are feasible and acceptable, there is limited evidence collating information across medical conditions or in the context of HIV prevention, care, and treatment. Objectives: We conducted a scoping review to understand the current evidence on the acceptability, feasibility, and efficacy of digital DOT to improve medication adherence and, specifically, to determine if digital DOT had been used to improve adherence for HIV prevention, care, and treatment Methods: We searched the electronic databases PubMed, Embase, and the Web of Science in January 2021 for any published studies with terms related to digital technologies and DOT. We included peer-reviewed studies in any population, from any country, for any outcome, and excluded conference abstracts. We included three types of digital DOT interventions: synchronous DOT, asynchronous DOT, and automated DOT. We provide an assessment of the current evidence, gaps in literature, and opportunities for intervention development regarding the use digital DOT to improve antiretroviral therapy (ART) adherence, specifically in the field of HIV. Results: We identified 28 studies that examined digital DOT. All studies found digital DOT to be acceptable and feasible. Patients using digital DOT had higher rates of treatment completion, observed doses, and adherence compared with in-person DOT, although data were limited on adherence. Only one study examined HIV prevention, and none examined ART adherence for HIV treatment. Conclusions: Digital DOT is acceptable and feasible but has not been used to remotely monitor and support ART adherence for people living with HIV.


Introduction
2][3][4][5] Technology-based DOT strategies have been shown to be more cost-effective and reduce provider and patient burden compared with regular monitoring visits for medication adherence. 6Although some evidence suggests that these technologies are feasible, accessible, and effective for DOT, there is limited evidence collating information across medical conditions which could inform the development of new interventions to improve medication adherence.Specifically, limited evidence exists describing the use of digital DOT interventions in the context of HIV prevention, care, and treatment.
In the United States, the largest drop-offs in the HIV care continuum are from poor retention in care and viral suppression, whereby 50% of those diagnosed are not retained in care and 43% of those prescribed antiretroviral therapy (ART) do not achieve viral suppression. 7][10][11][12] Nearly two-thirds of new HIV diagnoses are attributed to people living with HIV (PLWH) who are aware of their HIV infection and either not in care or are receiving care but not virally suppressed, making ART adherence a top priority for curbing incidence. 13][18] Among PLWH, young adults, Black Americans, and sexual and gender minorities have some of the lowest rates of viral suppression and engagement in care. 7,193][24][25][26][27] Thus, improved strategies are needed to increase adherence to medications to prevent new HIV infections among those uninfected and improve virologic suppression among PLWH.
We conducted a scoping review to understand the current evidence related to the acceptability, feasibility, and efficacy (i.e.impact on medication adherence) of digital health technologies for DOT to monitor adherence in any population for any outcome, and to determine if digital DOT had been used to improve medication adherence for HIV prevention, care, and treatment.We conducted a scoping review instead of a systematic review to broadly synthesize evidence on the use of digital DOT to improve medication adherence and, specifically, to assess the scope of evidence in the field of HIV.Here, we provide an assessment of the current evidence, gaps in literature, and opportunities for intervention development regarding the use of digital DOT to improve medication adherence in general and specifically among PLWH and those at high risk for HIV acquisition.

Materials and methods
The scoping review was conducted according to the standard protocol for Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews checklist (PRISMA-ScR). 28We included quantitative or qualitative studies of digital DOT technologies on adherence from any prior date until January 8, 2021.Digital DOT interventions were defined as interventions that allow a human or technology (e.g. a smartphone app) to digitally observe or view a participant consume a medication at the time of dosing.We included three types of digital DOT interventions: synchronous DOT, asynchronous DOT, and automated DOT.3][4] Automated DOT interventions use smartphone apps with artificial intelligence and facial recognition to confirm medication ingestion without the use of a provider. 1,29We did not include studies with ingestible sensors or reminders via text message as these constitute a different type of intervention that do not include observation by a provider or researcher.Therefore, the studies that met inclusion criteria were those that 1) included the use of a digital DOT technology, 2) focused on improving medication adherence for any medical condition, 3) were peer-reviewed publications, and 4) were published in English.We included interventions in any population and did not restrict by comparator group or study design.The outcomes of interest were medication adherence and the acceptability and feasibility of the intervention.Cost-effectiveness outcomes were not included.We did not restrict studies by type of reported statistics (e.g.odds ratio).
We reviewed published, peer-reviewed literature across a range of studies, from pilot and descriptive studies to randomized trials.The first search was conducted on November 23, 2020, with a second search on January 8, 2021, to confirm that we were able to identify key publications of interest.We searched the electonic databases PubMed, Embase, and the Web of Science, including published peer-reviewed literature and excluded conference abstracts.We used the following keywords: (digital Ã OR technolog Ã OR application Ã OR "app" OR "apps" OR smartphone Ã OR smart phone Ã OR mobile OR mhealth Ã OR telehealth OR tele-health OR telemedicine OR tele-medicine OR video Ã OR videoconferenc Ã OR telerehabilitation OR tele-rehabilitation OR remot Ã ) AND ("directly observed therap Ã " OR "directly observed treatment Ã " OR "directly observed intervention Ã " OR "directly observed" OR "direct observation").We also conducted a manual search of references of citations identified through the electronic searches, to identify any additional relevant citations.
Abstracts were imported from each electronic database and combined within Covidence online software, which automatically removes duplicates. 30One coauthor screened the abstracts and examined the full text of articles in relation to inclusion criteria.A second co-author then reviewed a random sample of 10% of the articles and flagged discrepancies which were reviewed again by the first co-author.If there were any discrepancies, an additional 10% of articles were reviewed and so on.A third co-author was consulted for any unresolved discrepancies.Additional studies that were abstracts, scoping reviews, or commentaries on the topic were only included in the introduction and discussion to provide information on newly conducted research or opinions in the field.Data charting was performed using a standard form that was created by the team through Covidence software and included variables of interest to extract.One reviewer extracted information from all studies including the citation, year, population, inclusion criteria, sample size, study design, medical condition (e.g.HIV, tuberculosis [TB]), outcome, intervention description, if incentives were provided, follow-up period, and findings.The same process for review of extracted data was used as was done for the inclusion process whereby a second co-author reviewed a random sample of 10% of the articles and flagged discrepancies.Extracted information was based on agreement between the two co-authors.We conducted a critical appraisal of study quality and potential biases for each study by examining the strength of the study design (e.g.pilot, programmatic data, randomized trial) and sample size.Given that most of the studies were not randomized trials and used different research designs and outcomes, we opted not to use a standardized critical appraisal tool.Study data were summarized overall and by medical condition (e.g.HIV) being examined to assess the overall evidence for medication adherence according to the type of medical condition and type of intervention (e.g.automated versus asynchronous DOT).Data are presented in table format by medical condition and type of intervention.A protocol was developed and distributed internally prior to conducting the review .

Results
After exclusion of duplicate citations, a total of 93 studies were identified from the three databases (Figure 1).The titles and abstracts were screened for relevance by one reviewer, and 64 were retained for full text review.Reasons for exclusion are reported in Figure 1.A total of N ¼ 28 studies were eligible for inclusion.These studies are summarized in Table 1, grouped by medical condition and type of intervention.
Most studies examined medication adherence related to TB (n ¼ 22, 79%; Table 1).Of the remaining six studies each examined adherence to medications for different medical conditions or needs: schizophrenia, stroke, HIV prevention, asthma (inhaler use), dietary supplement use, and opioid use.Most publications were cohort studies (n ¼ 19; 68%), five were single arm pilot studies (18%), three were randomized trials (11%), one was a non-randomized experimental study (4%), and one was a pilot study where patients were randomized to either immediate or delayed DOT with an mHealth app intervention (4%

Asynchronous DOT
2][33][34][35][36][37][38][39][40][41][42] Participants were satisfied with the intervention, would recommend it to a friend, would use it in the future and thought it was convenient, private, and easy to use.Both patients and providers found the app acceptable and thought it saved time and money compared to in-person DOT. 38,39,43A study of inhaler use in children found that all but one child mentioned that daily recoding of their use of the inhaler was fun and enjoyable. 31In five studies, patients stated that they preferred or would choose asynchronous DOT over inperson DOT. 32,35,38,40,41Concerns with asynchronous DOT included technical difficulties, privacy concerns, and discomfort taking a video of oneself.
Treatment completion and adherence were high across studies (82-93%).8][39] A randomized trial of 226 patients reported that 70% of asynchronous DOT patients achieved 80% or more scheduled observations compared with 31% of in-person DOT patients. 37Only one study explicitly compared adherence between asynchronous DOT and a control group that measured adherence through in-person DOT visits.The study found that medication adherence on DOT was comparable to that of in-person DOT (94% vs 98%, P ¼ 0.17). 39

Automated DOT
Of the two studies that assessed acceptability of automated DOT (aDOT), all found the intervention acceptable and feasible. 1,29Patients reported that the artificial intelligence (AI) platform improved the doctor-patient relationship, 29 high satisfaction, that they would recommend the app to a friend, that they would use it in the future, and that it helped with medication adherence. 1Patients with little experience using a smartphone were able to successfully use aDOT. 29wo studies assessed adherence measured using the AI platform versus adherence through mobile DOT or plasma drug concentration levels and found increases in adherence with aDOT.One study in patients who had experienced a stroke found that adherence based on plasma drug concentration levels was 100% (15 of 15) in patients using aDOT compared with 50% (6 of 12) in the control group.The second study found that mean adherence over 24 weeks was 90% (standard deviation [SD] ¼ 24.9) for participants who were monitored using aDOT, compared with 72% (SD ¼ 39.8) in  those who were monitored in person by study staff (or a third party) for a difference of 18% (95% confidence interval [CI] ¼ À2.0 to 37.7; P ¼ 0.08).Among all 28 studies, only one used digital DOT with an HIV prevention outcome. 1 Among young men who have sex with men in San Francisco, California, automated DOT was highly acceptable, 84% reported the app helped with taking pre-exposure prophylaxis (PrEP), and median PrEP adherence was 91%. 1

Discussion
We identified 28 studies that examined digital DOT interventions to improve adherence to medication for any health indication.40][41][42]44,46,47,49,50,54,55 Patients using digital DOT had higher rates of treatment completion, observed doses, and adherence compared with in-person DOT. 29,56More research is needed from large randomized trials examining the impacts of digital DOT on adherence compared with other interventions and comparing different types of digital DOT.Evidence largely shows that digital DOT is acceptable, feasible, and has the potential to improve medication adherence.
Given the challenges with adherence to PrEP and ART, digital DOT has a high potential to improve HIV prevention, care, and treatment.However, we identified only one study that had an HIV prevention outcome. 1 The automated DOT approach used AI in a smartphone app and was found to be highly acceptable with high PrEP adherence. 1This study highlights the potential benefit of using digital DOT technologies to increase adherence in the context of HIV prevention, care, and treatment.We did not identify any studies examining ART adherence for HIV; however, two excluded studies used mobile phone-based reminders to increase ART adherence. 57,58Our findings highlight the potential benefit of using digital DOT technologies to increase adherence in the context of HIV prevention and the gap in research focused on digital DOT to improve adherence to HIV treatment.
3][64][65][66] A systematic review examining the HIV care continuum in YLWH found that only 54% of youth who initiate ART achieve viral suppression and an additional 43% are retained in care. 7 There are some limitations to our review.We reviewed only three databases and included studies in English.Therefore, there may have been studies that were not captured in our review that would have been using a more systematic review approach.While there were 28 studies that used digital DOT, there were few studies that were randomized trials and most had a small number of participants.Most studies focused on TB outcomes and were in the United States.More data are needed from regions outside of the United States and from large trials and those that examine adherence outcomes using digital DOT compared to other methods as well as studies that compare different types of digital DOT to each other.Additionally, while over 96% of individuals age 18-29 in the United States have a smartphone, 67 we acknowledge that there will be more marginalized groups who may not have smartphone access and this intervention may not be accessible to them.Further studies are also needed to review implementation outcomes with the scale up of digital DOT in healthcare settings and should also investigate digital DOT and other tools to assess adherence for non-clinic-based provision of long-acting injectables for HIV treatment and prevention.
Given the recent COVID-19 pandemic, remotebased research and the use of digital tools to support medication adherence are becoming increasingly important. 68Digital DOT may offer a solution by which to remotely monitor and support medication adherence.However, there is a critical gap in literature on the use of these strategies to increase ART adherence among PLWH.It may be a particularly useful tool to promote ART adherence among groups experiencing inequities in HIV care such as youth, and people of color.Digital DOT may increase convenience and reduce patient and provider burden compared to in-person DOT.Compared to in-person DOT, several studies found improved treatment completion and adherence with digital DOT.However, randomized studies assessing adherence outcomes are limited.More data are needed from larger, experimental randomized trials, in settings with limited access to healthcare such as rural settings and comparing the different methods of digital DOT that now exist.

Figure 1 .
Figure 1.Flow diagram for study inclusion.

Table 1 .
Characteristics of 28 studies included in the scoping review.

Table 2 .
Suboptimal ART adherence and retention in care can increase risk of treatment failure, HIV transmission, and poor health outcomes.Thus, YLWH are a key population with whom to examine digital DOT interventions aimed at improving HIV clinical outcomes.Acceptability, feasibility, and adherence outcomes among 28 studies included in the scoping review.