Antiretroviral therapy (ART) has transformed an HIV infection from a likely death sentence to a manageable chronic condition [1], but the efficacy of ART hinges on maintaining high (at least 80%-85%) mean medication adherence [2-4]. Globally, approximately 53% of people living with HIV have access to ART, but only 44% of people living with HIV are virally suppressed [5]. In sub-Saharan Africa, only one-fourth of people living with HIV are virally suppressed [5], which results in avoidable cases of drug resistance [6] and death [7]. Structural (eg, drug availability), social (eg, stigma), and economic (eg, distance to clinic and clinic fees) ART adherence barriers have been documented in the literature [8-10], but patient behavior has been identified as a key factor determining the lack of viral suppression [11]. Recent studies have shown that mean ART adherence ranges from 60% to 80% in Uganda, and only 30% to 60% of patients achieve 85% adherence [12-14]. Thus, novel behavioral interventions are needed to help establish and maintain high ART adherence habits among people living with HIV in Uganda.

Daily habits (or routines) are a commonly reported strategy for maintaining high medication adherence among patients who successfully manage chronic diseases [15,16], but forming new routines is often difficult for patients to do on their own. According to recent psychology research, it takes approximately 3 months of repeatedly performing a daily behavior in response to the same contextual cue [17-20], as outlined in the Habit Formation Model in Figure 1A [21], before the behavior becomes routinized. Once routinized, the cognitive processes that govern the behavior move to neurological systems that operate nonconsciously [22-25]. In addition, behaviors that are routinized no longer require high intrinsic motivation to be performed, and thus can enable even the most vulnerable ART treatment initiators, such as those with limited motivation, to maintain high long-term adherence.

A common intervention method for establishing new routines is to anchor the targeted behavior to an existing routine that acts as the contextual cue. For example, taking ART medication after brushing one’s teeth in the morning or after completing evening prayers. This method is often called anchoring (Figure 1B) and has been shown to be an effective intervention for promoting physical activity routines [26], improving dietary routines [27], and maintaining smoking cessation [28]. However, these existing studies typically enrolled participants with high intrinsic motivation for the targeted behavior [29-33], and therefore have little potential for real-life clinical situations where there is no extra support for individuals with low intrinsic motivation (as is typical for many patients in HIV care [34]). Moreover, less than half of the participants in these existing studies successfully used their anchor and maintained the desired behavior in the long term [35].

Behavioral economic theory demonstrates the need for ongoing support during the time it takes to complete the routinization process and provides proven intervention methods for delivering such support to enhance existing anchoring interventions. The behavioral economic biases of lack of salience of ART adherence (eg, over time, more pressing needs of daily life dominate one’s attention and focus) and present bias (eg, excessively undervaluing the future health benefits of one’s actions) help to explain why many people have trouble adhering to their healthy intentions [36,37]. Fortunately, behavioral economics also suggests 2 methods for countering these biases: (1) SMS text messages that can be used to reinforce the information provided at recruitment and to increase the salience of the anchoring routinization strategy and (2) small behavioral economics–based incentives that have successfully changed a range of health behaviors by countering present bias [38-41]. Therefore, this study will test whether incentives for linking daily pill taking to the timing of an existing routine behavior can establish and maintain high ART adherence routines in a feasible and scalable manner.

In addition, this intervention is being targeted to treatment initiators to leverage the fresh start effect [42], a period of heightened motivation and attention. ART adherence instructions are initially salient for treatment initiators; however, over time, they fade from attention. Among adults with HIV in sub-Saharan Africa, forgetfulness is the most frequently reported barrier to maintaining high long-term adherence [43,44]. Therefore, treatment initiators may need subsequent support for adhering to the ART medication protocol until the behavior has been successfully routinized, which will be provided in the form of SMS text messages and incentives.

We propose to test the intervention, Incentives and Reminders to Improve Long-term Medication Adherence (INMIND), in a pilot, parallel group randomized controlled trial (RCT) at the Mildmay Uganda HIV clinic in Kampala, Uganda, with 2 intervention groups and 1 control group, using an even allocation ratio of 1:1:1 among the 3 study groups. All (150/150, 100%) participants (including those in the control group) will receive information about the importance of behavioral routines, as it is a part of the standard adherence counseling for treatment initiators at Mildmay Uganda, and will create personalized ART adherence anchoring strategies. The first intervention group will additionally receive SMS text message reminders of their anchoring strategy, and the second intervention group will receive both SMS text message reminders and small incentives conditional on taking ART pills within a time window that corresponds to participants’ personalized anchoring strategy. We hypothesize that the SMS text message reminders and incentives will increase participants’ use of their anchoring strategy, and thus, medication adherence will be better maintained after the intervention ends in both the intervention groups relative to the control group. We also expect to see stronger maintenance in the intervention group receiving both reminder messages and incentives than that in the intervention group receiving only reminder messages.

This pilot RCT has been funded by the National Institute of Mental Health in the United States (R34MH122331) and approved by the RAND Corporation’s Human Subjects Protection Committee (2020-N0632), Mildmay Uganda Research Ethics Committee (0701-2021), and Uganda National Council for Science and Technology (HS128ES).

This study will use a 3-armed RCT (2 intervention groups and 1 control group) with randomization at the individual level. Refer to the SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials) checklist for a guide to the key items reported in this protocol (Figure 2). The interventions will be administered for a period of 3 months from the baseline survey. During the intervention period, medication event monitoring system (MEMS) data readings will be collected monthly, wherein prize drawings for eligible participants in the incentives group will be conducted. As the population of interest is treatment initiators, the monthly study visits are expected to coincide with the monthly clinic visits as mandated for newly diagnosed clients who are becoming accustomed to ART treatment. During the postintervention period, the team will follow participants and continue MEMS data collection during the participants’ regular clinic visits (expected to be every month for the first 6 months after treatment initiation and every 1-3 months thereafter) for a period of 6 months, thus bringing the total study period to 9 months.

As the long-term goal of the study is viral suppression, viral load data will also be collected during the routine assessment. Viral load tests for treatment initiators are conducted after the first 6 months of ART initiation and then changed to 12-month periods as per the clinic and Ugandan Ministry of Health guidelines. Survey assessments will be administered at baseline, 3 months, and 9 months for all (150/150, 100%) participants.

It is important to note that the pilot study is being preceded by a formative phase and succeeded by an adaptation phase, both of which are designed to collect qualitative data pertaining to the overall design, feasibility, and acceptability of the interventions. The results of the formative phase will be used to adapt the pilot intervention materials and timeline before intervention administration. The adaptation phase is expected to provide feedback on several critical parameters of the interventions, which will be used to mold the intervention in preparation for a full-scale RCT. Figure 2 shows the timing of the study activities.

The study will be performed at Mildmay, Uganda, an HIV clinic in Kampala, Uganda. This clinic specializes in providing comprehensive HIV and AIDS prevention, care, and treatment services to >105,000 patients with HIV. It offers integrated health services to a diverse population base; of the 15,000 patients served at the main site in Lweeza, 11% are children younger than 18 years, 65% are women, and 100% of the patients are on ART. Mildmay also has well-established electronic medical records infrastructure, making it one of the growing number of facilities using electronic medical record systems in Uganda.

Apart from the patient-facing work, Mildmay also provides technical assistance to organizations and governments, hosting training and educational courses for >1500 professionals per year. Furthermore, the Mildmay Uganda laboratory is accredited by the South African National Accreditation System under International Organization for Standardization 15,189:2012 and specializes in virology and other tests. Mildmay also has numerous ongoing research projects involving international researchers. It has a standing community advisory board comprising church leaders, elected councilmen, health care providers (external to Mildmay), and advocates of clients with HIV.

The objective of this study is to establish preliminary efficacy, acceptability, and feasibility of the interventions. Consequently, 150 participants will be recruited from Mildmay’s main site for the pilot study. The sample will be representative of the patient population at Mildmay, with an approximately 70:30 ratio of female to male clients.

Electronic medical records and the hospital electronic database will be used to screen the client population for initial eligibility based on age and ART initiation period, as noted in the eligibility criteria in the following sections. For this step, the hospital staff (and not the study coordinator) will mine the electronic database for this information to create a master list of eligible clients. Given the focus on treatment initiators, this process will be conducted weekly to identify newly diagnosed clients who are eligible to participate in the study.

A daily list of expected clients will be generated from the master list for the study coordinators to identify those deemed eligible and who are due for a clinic visit. Once a client is identified, a prebaseline visit will be initiated, with the study coordinator approaching the client and inquiring about their interest in participating in an ongoing study. On confirming interest, clients will be taken to a separate study room to verify their eligibility and to obtain their consent to participate. Consented participants will be given a MEMS cap and instructed to store one of their HIV medications in a pill bottle with the MEMS cap attached. In addition, they will be given a study appointment approximately 1 month after this initial visit for adherence data retrieval and baseline survey administration. The first month of adherence data will be used as a baseline, and the intervention will not begin before this follow-up visit.

The study sample will consist of male and female clients aged ≥18 years who have started ART at Mildmay within the preceding 3 months. Treatment initiation is an important eligibility criterion given the conceptual framework of the pilot study, which suggests that habit formation during initiation is a key driver for sustained ART adherence over time. On the basis of electronic records data from Mildmay, each month, at least 100 to 150 clients with these characteristics begin treatment at the hospital (a total of 1758 clients in 2018); therefore, a large pool of potentially eligible clients will be available for recruitment. Given the nature of the intervention, inclusion will also require participants to own or have access to a phone for at least 5 days per week throughout the duration of the intervention and be willing to receive SMS text messages throughout the intervention period.

Children aged <18 years were excluded for 2 reasons: (1) the pediatric clinic is separate from the main recruitment site (which primarily caters to adult clients) and (2) the intervention may require alteration to account for the specific needs of children and adolescents. In addition, clients who are not mentally fit to understand the consenting or study procedures and clients who speak neither English nor Luganda (the local language spoken by most people in and around Kampala) were excluded. As both intervention groups rely on receiving SMS text messages, clients who do not own a mobile phone or have access to one will also be excluded.

Participation in another adherence study and inability or nondesire to use MEMS caps regularly throughout the course of the study will also be a basis for exclusion. Consequently, if the baseline MEMS reading suggests that participants opened their pill bottle <30% of the days and if that was not a consequence of low adherence, the participant will be given a transport refund and be unenrolled from the study. Finally, clients who come outside regular clinic working hours will be excluded.

Participants will be randomly assigned to either 1 of the treatment arms (treatment group 1 [T1]=messages; treatment group 2 [T2]=incentives) or the control arm after consenting but will only be informed of their assignment after the baseline survey is over to minimize any potential influence of the assignment on the baseline survey responses. The distribution ratios for the randomization will be 1:1:1, and the assignment will be conducted through a computer-generated randomization component built into the baseline survey administration software called Questionnaire Development System. The random assignment (to either the control arm or 1 of the 2 intervention arms) will be revealed at the end of the baseline survey to both the participant and study coordinator, who will not know the respondent’s treatment assignment during the survey. Given the nature of the intervention, neither the interviewers nor the participants can be blinded to the treatment status. However, the data analyst who will conduct the impact analysis will be blinded to treatment assignment.

The study will assess the following outcomes.

All data collected during the course of the study will have only the existing clinic identifiers as the unique participant identifiers. All other identifiable information will be stored separately. Data collection and storage hardware (ie, tablets and computers) will be password-protected, and physical storage spaces will have a locking mechanism for security. All physical storage spaces will be located at the Mildmay RAND office in Kampala, with access granted only to key personnel and the principal investigator. These physical storage spaces will be used to store the consent forms and other physical tracking documentation.

All data collection and on-ground study activities will be conducted by the study team in Uganda. This team will include a team leader, 2 lead interviewers, and 3 supporting team members. The design of the data collection instruments and protocols, quality monitoring of the qualitative and quantitative data, and data analysis will be conducted by the study team based in the United States. Data collected on the ground will be transferred to the US team on a weekly basis through a secure web portal (Kiteworks).

Published material will not contain any personally identifying information. There is no data monitoring committee because the trial was deemed to have minimal risk. The study team in the United States will still perform data monitoring and quality assurance exercises weekly during the 9-month study period.

Missing data for some variables will be imputed if a participant remained enrolled in the study. When participants drop out, multiple logistic regression models will be fit to assess whether this dropout is random. If it is not, nonresponse weights using logistic regression will be developed to adjust for dropout. All analyses will reflect these design effects in the calculation of SEs and statistical tests of significance.

We hypothesize that our 2 intervention groups will display high mean medication adherence and high routinized medication adherence (ie, pill taking within –1 to +1 hour of participants’ chosen anchor) during the 3-month intervention relative to the control group. After using their anchoring strategy more frequently, we hypothesize that our 2 intervention groups will better maintain their mean medication adherence and routinized medication adherence over the 6 months after the intervention is withdrawn. Finally, we hypothesize that our second treatment group, which receives both SMS text message reminders and small incentives for using their anchor during the 3-month intervention, will more strongly maintain their mean medication adherence and routinized medication adherence during the 6-month postintervention period.

INMIND has a strong scientific premise that addresses a critical knowledge gap in the literature around the design of interventions for establishing and maintaining long-term behavior change. A growing body of literature in the field of psychology targets long-term behavior change [18,52]; however, most intervention methods are one-off interventions that do not support participants during the approximately 3-month routinization process [17,20,53,54]. Behavioral economics–based interventions have also had limited efficacy in maintaining long-term behavior change. For example, incentives have successfully changed a range of health behaviors by countering present bias [38-41], including improved ART adherence [55], but these interventions typically show only short-term impact that dissipates after the incentives are withdrawn [38,52]. In our own previous studies, we also found that incentives did not have persistent effects, and only participants who showed timely adherence (an indicator that they potentially anchored pill taking to an existing routine) maintained high adherence after the incentives were withdrawn [51].

Our combined intervention approach attempts to leverage the successful components of these existing psychology-based and behavioral economics–based interventions to better maintain ART adherence. If successful, this intervention will help to expand the understanding of the habit formation process and common psychological barriers to successfully adhering to new health behaviors. The study also addresses biological variables such as age and sex appropriately and incorporates them in both the impact analysis and when testing for age and sex differences in behavioral biases and intrinsic motivation. Such analyses will be especially useful in designing a large-scale intervention that can assist all treatment initiators in establishing and maintaining long-term ART adherence.

The study has some limitations. First, the study will recruit the sample from 1 clinic in Uganda, limiting the generalizability of our results. Second, the small sample size limits the team’s ability to detect clinically meaningful effect sizes. Although this is not the intended aim of this feasibility and acceptability study, the small sample size is still a limitation to the statistical analyses. Third, ART adherence will only be measured over 9 months; therefore, future studies will be needed to assess ART adherence over long durations. Fourth, ART adherence will be measured using MEMS caps, which is currently one of the most accurate ways to measure medication adherence, but conscious manipulation of the pill bottle openings by the participants is still a possibility that may lead to an overestimation of the study’s adherence outcomes for participants in the incentives group, who may use deception to increase their chances of a prize drawing.

The team will use peer-reviewed publications and conference presentations as the primary means of disseminating results. The findings will be relevant to those interested in the behavioral mechanisms that underlie successful long-term ART adherence and, more broadly, the mechanisms underlying long-term behavior change. In addition, the findings will be used in the design of a large-scale RCT that we will use to rigorously assess the effectiveness of the INMIND intervention for establishing long-term ART adherence habits. In addition to future RCTs, we plan to use detailed MEMS data on the timing of daily pill taking to inform statistical models of the habit formation process, which will guide new intervention designs for promoting ART adherence habits.