Correction: mHealth Interventions to Promote Anti-Retroviral Adherence in HIV: Narrative Review

[This corrects the article DOI: 10.2196/14739.].


Introduction
Background HIV remains to be a significant public health concern, causing an estimated 770,000 deaths worldwide, with significant mortality from poorly managed disease, even in developed nations [1,2]. Particularly in the developing world, there is significant morbidity with 79.1% of HIV prevalence centered in Africa (25,700,000 people) and Southeast Asia (3,500,000 people) [3,4].
The burden of this disease is not only health related but also impacts economics and development. A large proportion of deaths from HIV occurs in young and middle-aged adults; the same group of people who raise children, are caretakers, work, and teach future generations. Especially in lower-income countries, families who have had a death from HIV cope by pulling children out of school, reducing food intake, or using up agricultural resources prematurely [5]. In addition, although first-line antiretrovirals (ARVs) cost US $80 per year, the cost of second-and third-line regiments (used with drug resistance) may be as high as US $2200 per year [6].
Despite all these bleak consequences, studies have shown that with adherence to ARV therapy, a patient aged 20 years could expect an additional 43.3 years of life [7]. In contrast, those who are not adherent and develop AIDS have a median survival of 12.1 months [8]. Adherence to ARVs is key to prolonging life, improving immune function and quality of life, and preventing transmission [9][10][11][12][13][14].
Without proper adherence, there is a risk of disease progression or drug resistance. Despite increasing access to ARVs worldwide, adherence is far from perfect [15,16]. One meta-analysis reported adherence at 62%, and studies on subgroups, such as illicit drug users, are as low as 27% [17,18].
Factors such as psychiatric disorders, cognitive impairment, social stigma, substance abuse, and volatile housing can prevent optimal adherence [19,20]. Burch et al [21] found clear associations between markers of the social determinants of health and HIV outcomes. Rachlis et al [22] found that substance abuse and unemployment were associated with suboptimal care. The risk of transmission and resistance make it crucial to develop methods for improving adherence to medications. Indeed, international organizations have called for strategies to improve adherence to medication and its importance [23][24][25].
Consumer informatics, particularly mobile health (mHealth), has shown promise in improving ARV adherence. Indeed, research has shown that patients living with HIV are interested in apps to support HIV self-management, and much past research has been done to support disease management [26,27].
Given the widespread adoption of cellphones (95% of the population) and smartphones (77%) in different parts of the world, mHealth is becoming an important tool to improve care [28]. Furthermore, in North America, access to mobile data and the internet ranges close to 80%, and with expanding mobile computing capabilities, this represents further opportunities [29].

Objectives
We believe that the major benefits that mHealth could offer would be helping with medication and appointment adherence, improving HIV education, and increasing engagement in care. Some studies have shown promise in improving viral suppression and CD4 count. However, there are existing gaps in current reviews [30]. First, as technology is a rapidly growing field, constantly updated work is needed to summarize developments. Second, and most importantly, people who live with HIV are a diverse group of patients with a variety of challenges, including living in developing countries with poor health infrastructure and having comorbid substance use issues [31][32][33][34]. Each of these patient populations has unique characteristics and concerns that may influence care. For instance, access to the internet and smartphones, health care, income, and psychiatric comorbidities could be vastly different in various groups. The majority of existing reviews homogenize and summarize the existing literature as a whole and do not consider the differences in patient populations [35][36][37][38][39]. An intervention may have varying levels of success in each group and likely needs to be tailored to fit unique needs. For instance, a smartphone app for disease education may be successful in college-educated, smartphone-owning patients; however, it may have a vastly different effect on a patient in the developing world without a smartphone or an advanced education. The underlying differences in patient groups need to be considered, and our review groups studies into subgroups and discusses the implications of such differences.

Search Strategy
Two separate databases with peer-reviewed, reputable articles were used, PubMed and Academic Search Elite. All relevant articles reviewed were within the time period of 2010 to 2019.
For PubMed, we used the following search terms: "mHealth AND HIV AND medication adherence," "app AND HIV AND adherence," "smartphone AND app AND HIV AND anti-retroviral adherence," and "mobile phone AND anti-retroviral adherence AND HIV." On reviewing the literature using these search terms, we added the search terms "WelTel" and "CAMPS and retroviral" to capture relevant studies found in the references of the articles retrieved from the initial search.
We used the search terms "mHealth AND HIV AND medication adherence," "mobile phone AND anti-retroviral adherence AND HIV," and "app AND HIV AND adherence" for the Academic Search Elite database search.
Finally, we used the Journal of Medical Internet Research database, a group of journals that focuses on informatics implementations in health care, to identify articles listed under "mHealth for Treatment Adherence."

Inclusion and Exclusion Criteria
A total of 386 search results were screened using the title or abstracts for relevance, and 61 articles were fully reviewed and are discussed in this study.
Articles were included if they were experimental in nature (encompassing qualitative or quantitative studies as well as randomized controlled trials [RCTs] or other studies). Study designs included RCTs, case studies, cohort studies, cross-sectional studies, and qualitative studies (focus groups, interviews, observations, and surveys). No age restrictions were set, although the majority of the literature focused on adults and studies were required to have all participants diagnosed with HIV. A significant number of articles were excluded (273/325, 84%). Reasons for exclusion were irrelevance to the topic discussed, inclusion of patients not diagnosed with HIV, articles written in a language other than English, reviews rather than original research on this topic, categorization as nonexperimental irrespective of whether qualitative or quantitative (for instance, a commentary of mHealth), or if they were protocols for experiments.
We defined mHealth in this study as any intervention that involved the use of mobile phones, smartphones, or other wireless devices (such as smart pill bottles). Although no formal definition of mHealth exists, this description is in agreement with the World Health Organization [40]. Pilot trials found on literature review were included in our review.

Overview
A total of 61 papers were identified from the period of 2010 to 2019. In accordance with the mHealth categories previously described by the United Nations, we categorized apps according to type and found that these apps were focused on either treatment support (reminders) or education [41]. In addition, we categorized apps as either push or pull and as one-way versus dual communication. Interventions that were primarily reminders tended to use push, whereas those that were education used pull ( Table 1).
As factors such as technology accessibility, socioeconomic background, and internet or cellular connectivity vary based on geographical location, the authors chose to separate the discussion into studies performed in lower-and higher-income countries. Studies were divided into categories (upper middle and upper vs lower middle and low) as per the definition by the United Nations [42].
The risk factor for HIV acquisition in upper middle-and upper-income countries was primarily men who have sex with men (MSM) or intravenous drug users (IVDU). The majority of these studies were performed in North America (Canada and the United States).
For lower middle-and low-income countries, the majority of these studies were based on the African continent, with some conducted in Asia.
On an evaluation, using the Oxford Centre for Evidence-Based Medicine's Levels of Evidence, there was a large amount of high-level evidence in developing countries, with many RCTs [59]. On the basis of this, we believe that conclusions regarding efficacy are possible. There is also a high level of evidence (with a large number of RCTs) in higher-income countries; however, there remain some limitations to those studies ( Table  1). The authors believe that further work may need to be performed in higher-income countries. 75 randomized to iTAB, which were daily personalized texts built from focus groups and focused on responsibility to others, self-esteem, nonadherence risks, harm reduction, reminders, spirituality, celebration of health, and disease control (50) vs control (25). Also assessed methamphetamine use with daily texts Pilot at UCSD e whose patients: Given the rising rates of cellular access in the developing world, mHealth represents a significant opportunity to improve care [64].
A variety of studies have been conducted on mHealth interventions in the developing world, with a wide range of study methodologies. The quantitative experimental designs (including cohort studies and RCTs) are summarized in Table  1 for ease of comparison. Seven studies with quantitative data were available, with the majority being text-or SMS-based modalities (5 studies) and a minority being voice call based (2 studies) [43][44][45][46][47][48][49].
Other studies investigated user acceptability and surveyed patient concerns regarding the use of mHealth. A pilot study in rural Uganda by Musiimenta et al [65] enrolled 63 patients and randomized them to a scheduled SMS, triggered SMS (by missed or delayed doses), or no SMS. The feedback received was positive, citing usefulness for motivation, reminders, and social support. However, confidentiality, shared phones, ability to use the phone, and availability of electricity were concerns [65]. Similarly, an explorative study by Lepere et al [66] demonstrated that mHealth was acceptable among people who live with HIV in Cote d'Ivoire, Togo, and Burkina Faso and advocated for its use. Mbuagbaw et al [67] also demonstrated, in a qualitative arm of the CAMPS trial, that these interventions were acceptable, although content, timing, technical challenges, and privacy need to be considered.

Higher-Income Countries Group
Countries classified as high or high middle income were analyzed. The primary risk factors in these studies were either IVDU or MSM, in agreement with the previous literature [68,69].
Different studies have investigated aspects of mHealth apps other than efficacy. WelTel was studied in Vancouver, British Columbia, on patients involved with substance abuse using both quantitative and qualitative methodologies [62,84]. One such study was done by Campbell et al [85], who studied WelTelBC1 from an economic perspective on 85 viral load-positive vulnerable patients (defined as substance use and other psychosocial factors), finding that the cost of caring for a highly vulnerable patient was US $347.74, whereas the overall median cost was US $36.72.
Multiple quantitative studies investigated the acceptability of mHealth interventions. In Seattle, 224 patients with HIV were randomized to receive two-way pagers with personalized adherence messages. Of these patients, 55% identified as gay or lesbian, and the majority were white males. Participants were surveyed for the usefulness of the pager and asked whether they adhered to their medications; 73% reported liking the pager, 51% began finding messages annoying, and 48% believed that the pager did not improve their adherence. The overall response rate was only 42.8% but declined over 3 months. The self-reported pager adherence was 90.8% (SD 33%) compared with the medication event monitoring system showing 53.6% (SD 37%) adherence [86].
Similarly, a Peruvian study by Krishnan et al [87] with 359 transgender women or MSM patients asked about mHealth preferences using a 5-point Likert scale. They found an overall positive uptake, with a mean of 3.21 for technological medication reminders and 3.56 for anonymous internet chats to discuss HIV issues.
Other studies, using focus groups and other quantitative study methods, were used to determine how best to deliver mHealth interventions. For instance, Krishnan et al [87] also found that daily messages were preferred in terms of frequency within their study population. In Bangkok, Anand et al [88] interviewed 16 MSM and 2 transgender women. They gathered information on preferences and how to address needs with electronic health. The group found that 39% (7/18) of patients preferred instant messaging and 11% (2/18) preferred phone calls. Of these, 39% (7/18) wanted a private website chat room and 11% (2/18) preferred Skype (Skype Communications) video chats. All patients desired personalized reminder messages; 50% (9/18) wanted reminders on an instant message app, 17% (3/18) preferred a stand-alone app, and 33% (6/18) wanted it as an SMS. The theme that emerged was web based, accessible, and reliable disease information [88].
Work performed in higher-income countries also had diversity in intervention type. For instance, Skrajner et al [89] enrolled a patient with cognitive impairment in a video-conferenced, cognitive-psychosocial program that improved the pill count adherence rate from 75% to 97.9% in 1 month (80% in 2-3 months). Similarly, the weCare team, which has done non-HIV work on social media, is currently developing a tailored Facebook, text, and social and sexual networking app-based intervention to help engage care [90]. Hwayoung et al [91] combined an electronic pill bottle, fitness tracker, and phone alerts and found that the pill bottle encouraged adherence and self-management of medications. However, they mentioned that the smart pill bottle was too small to fit all the pills, although they liked how it was easy to open and discrete [91].
Finally, Schnall et al [92] created the mVIP app, which attempted to provide self-care strategies for self-management of different HIV-related symptoms. They found improvements in a variety of symptoms, including anxiety (P=.001), depression (P=.001), neuropathy (P=.002), fevers or chills or sweats (P=.04), and weight loss (P=.02). Results for adherence were mixed, showing improvement only when measuring it via the center for adherence support evaluation adherence index [92].

Lower-Income Countries Group
Many studies have been performed using mHealth to improve adherence, with most in sub-Saharan Africa. Some of the studies were rigorous, large RCTs, such as Kenya WelTel [43]. There were some limitations in other studies where mHealth was not studied independently. For instance, in the study by Maduka et al [45], the effects of adherence counseling and SMS messages were not separated, and in the trial by Haberer et al [46], the role of social supporters was unclear. However, most of the studies indicate an improvement in adherence and VL (Table  1) [43][44][45][46]48,49]. Furthermore, it seems that mHealth represents a cost-effective intervention compared with the alternatives, with 1 study showing that a nurse could manage 1000 patients per week [43].
Although most studies found benefits, there is a gap in the literature regarding the design of a successful intervention. It was found that daily messages were not optimal, habituation likely being a significant factor [43][44][45][46][47]. Pop-Eleches et al [44] found that weekly interventions were more effective than daily interventions and long versus short messages made no difference, potentially reflecting that the messages were used more as reminders.
Most other successful interventions agreed with tapering schedules, twice weekly messaging in Nigeria, weekly WelTel Kenya1 messages, and the weekly South Indian study [43,45,46,49]. Another variable was message interactivity and content. On the basis of the existing literature, it is unclear if the interactivity of messages is important. In terms of content, Pop-Eleches et al [44] showed that a motivational message "This is your reminder. Be strong and courageous, we care about you" was not better than a generic impersonal message. These findings potentially indicate that an interactive or customized message does not negate the effects of habituation. Further work is necessary to determine the exact type of message that generates the maximal benefit.
Cultural contexts likely play a role in feasibility, and caution should be exercised when extrapolating results to different locations. Different views on health care, HIV stigma, baseline demographics, and cultural or religious views could be important when designing an intervention. The HIVIND study illustrates this, with high baseline adherence further boosted by the Hawthorne effect [48].
These studies collectively show a role for mHealth in improving HIV care in low-income or lower middle-income nations; however, we still need to delineate the best design. A less-than-daily message frequency seems to be optimal; however, interactivity and content need to be considered. Geographical and cultural differences may also affect efficacy.

Higher-Income Countries Group
The major risk factor for HIV in higher-income countries was either IVDU or MSM [93][94][95][96][97][98]. Although in higher-income countries different factors were identified in HIV management compared with lower-income countries, optimizing adherence is still the key.
In those with IVDU as a risk factor, compliance is important as it has implications for disease transmission. Finding meaningful interventions to improve adherence is crucial as there are higher rates of drug use in patients with HIV+, and these patients often face barriers to adherence [20,99].
However, there are a series of unique challenges for these patients that have implications in implementation and interpretation of studies. An inherent difficulty in studies performed on patients with a history of substance misuse is their heterogeneity, with a variety of substances used, including alcohol, marijuana, cocaine, amphetamines, and opioids. Prolonged use of drugs such as methamphetamines can cause psychotic episodes, resulting in more chaotic lives than other addictive substances [100,101]. Opioids have substitution agents that represent an opportunity for frequent, recurrent health care interaction in dispensing [102]. An intervention efficacious in one group may not be efficacious in another.
In addition, a variety of barriers exist that confound the ability of mHealth interventions to help with adherence. These include unstable financial and housing situations as well as psychiatric comorbidities [103]. Some of these barriers may limit the ability for any intervention to be beneficial without first addressing them.
The majority of the interventions studied in patients with substance misuse were text or SMS based, whereas those focusing on populations with MSM as the major risk factor had more advanced smartphone apps (Table 1) [50][51][52][53][54][55][56][57][58][60][61][62][63]. A potential explanation is the financial resources required for a smartphone and an accompanying plan, which supports the need for stable financial situations before a successful intervention.
Overall, across risk groups in higher-income countries, there seemed to be a theme of advanced mHealth interventions (using apps, social media, and game-like interventions). Examples in our review included EPIC Allies, the talking avatar, and the weCare interventions that contrasted mainly to call and text-related interventions in lower-income countries [72][73][74]90]. This could potentially be explained by the availability of smartphones and internet connectivity in higher-income countries [28,29].
Regardless of risk factors, a limitation of many of these individual studies was the small sample size. For instance, the iTAB study only showed an effect on methamphetamine use, but as this factor could contribute to nonadherence, a larger sample size may have had an effect on adherence [61].
In addition, studies outside North America should be interpreted with caution if implemented in North America, as many different considerations exist. For instance, different regions of the world preferentially use different messaging apps [104]. These may have different functions and capabilities. Certainly, further work is required [105,106].
A reasonable conclusion from the totality of the data is that there is a role in improving adherence in higher-income countries, but specific nuances need to be considered on a case-by-case basis. There are limitations affecting the ability to make conclusions, including small sample sizes, heterogeneous populations, and heterogeneous interventions [53,54,60,62,[80][81][82]88].

Conclusions
This study attempted to review the existing literature on mHealth in HIV adherence, divided the literature to the largest subgroups for which evidence was found, and discussed existing studies with each of these groups as specific contexts.
On the basis of an evaluation of both the results and the quality of evidence available, we believe a clear role for mHealth interventions in developing countries exists; however, further work is needed before a final conclusion can be made in the higher-income countries, where the 2 main subgroup risk factors were MSM and IVDU. In addition, a gap in the literature in all 3 groups is the exact nature of interventions (optimal message frequency, content, and intervention themes).
In addition, as demonstrated by the variety and scope of interventions in our review, there is a difference between digital and mobile access, with some groups having the hardware and connectivity to access advanced apps such as smartphone apps, whereas other groups simply have mobile access and can only access text or SMS or voice call interventions.
Our review of the literature provides an optimistic outlook on the role of mHealth in improving HIV care; however, there are limitations to our work. Our narrative, compared with a systematic methodology, provides a broad and comprehensive overview of the subject area but does not have the stringent inclusion and exclusion criteria of a systematic review. However, our review provides a broad view of the subject and identifies specific focus areas for future systematic reviews. Specific areas of future work include better delineation of the efficacy of different types of interventions in higher-income countries and in specific risk groups within higher-income countries.