Efficacy of an mHealth self-management intervention for persons living with HIV: the WiseApp randomized clinical trial

Abstract

Importance

Progression of HIV disease, the transmission of the disease, and premature deaths among persons living with HIV (PLWH) have been attributed foremost to poor adherence to HIV medications. mHealth tools can be used to improve antiretroviral therapy (ART) adherence in PLWH and have the potential to improve therapeutic success.

Objective

To determine the efficacy of WiseApp, a user-centered design mHealth intervention to improve ART adherence and viral suppression in PLWH.

Design, Setting, and Participants

A randomized (1:1) controlled efficacy trial of the WiseApp intervention arm (n = 99) versus an attention control intervention arm (n = 101) among persons living with HIV who reported poor adherence to their treatment regimen and living in New York City.

Interventions

The WiseApp intervention includes the following components: testimonials of lived experiences, push-notification reminders, medication trackers, health surveys, chat rooms, and a “To-Do” list outlining tasks for the day. Both study arms also received the CleverCap pill bottle, with only the intervention group linking the pill bottle to WiseApp.

Results

We found a significant improvement in ART adherence in the intervention arm compared to the attention control arm from day 1 (69.7% vs 48.3%, OR = 2.5, 95% CI 1.4–3.5, P = .002) to day 59 (51.2% vs 37.2%, OR = 1.77, 95% CI 1.0–1.6, P = .05) of the study period. From day 60 to 120, the intervention arm had higher adherence rates, but the difference was not significant. In the secondary analyses, no difference in change from baseline to 3 or 6 months between the 2 arms was observed for all secondary outcomes.

Conclusions

The WiseApp intervention initially improved ART adherence but did not have a sustained effect on outcomes.

KEY POINTS.

Question. Does the WiseApp Mobile intervention improve antiretroviral therapy (ART) adherence in persons living with HIV in New York City?

Findings. In this randomized controlled efficacy trial of 200 persons living with HIV, individuals randomized to the WiseApp intervention arm had a significant increase in ART adherence compared to those in the control arm during the first 59 days of the trial, but the effect was not sustained over the entire trial period.

Meaning. The WiseApp intervention initially improved ART adherence but did not have a sustained effect on outcomes.

INTRODUCTION

Despite efforts to achieve UNAIDS 95–95–95 targets, deficits remain in HIV viral suppression and antiretroviral therapy (ART) adherence among some groups of persons living with HIV (PLWH).¹ Progression of HIV disease and premature deaths among PLWH has been attributed foremost to poor adherence to HIV treatment regimens.^2,3 Timely access to ART and subsequently sustained ART adherence is central to therapeutic success and is a critical determinant of long-term health outcomes (eg, viral suppression) in PLWH.⁴ For many chronic diseases, such as diabetes or hypertension, drug regimens remain effective even after treatment is resumed following a period of interruption. In the case of HIV, however, loss of virologic control because of ART nonadherence may lead to the emergence of drug resistance and loss of future treatment options.⁴ Effective interventions for PLWH with poor ART adherence is essential to improve health outcomes and continue making progress toward HIV target goals.

The abundant use of mobile health (mHealth) technologies creates opportunities for health behavior management tools that were not previously possible⁵ and has the potential to be used to support the healthcare needs of PLWH. The use of mHealth can reduce geographic and economic disparities and personalize healthcare,⁶ which is particularly relevant to PLWH since many are from underserved and minority groups.⁷

mHealth technology is especially relevant in the context of improving ART adherence in PLWH.^8,9 mHealth is focused on the use of mobile information and communication technologies to support care delivery through meeting information, communication, and documentation needs of clinicians, patients, and other healthcare workers and facilitating health resource monitoring and management.¹⁰ mHealth can provide mechanisms for improving the efficiency and effectiveness of care provided while reducing administrative burden.

To that end, we developed an mHealth intervention, the WiseApp, using rigorous user-centered design research with underserved PLWH that draws on considerable formative work with PLWH, HIV clinicians, HIV case managers, and the Centers for Disease Control and Prevention (CDC) (U01PS003715, Principal Investigator [PI]: Schnall).¹¹ We then built the mHealth application (app) and integrated it with a smart pill box (CleverCap) to allow PLWH to self-manage their HIV and monitor their medication adherence in real time. The WiseApp is derived from formative work funded by and in collaboration with the CDC (U01PS003715) to design a self-management app for PLWH¹² with the goal of being more widely applicable across populations with chronic illness populations who require medications and adding self-management strategies. A comprehensive process for the design of the self-management app was guided by the Information Systems Research framework and incorporated end-user feedback throughout the design process.¹³ The resultant WiseApp is comprised of the following functional components: (1) testimonial videos of PLWH, (2) push-notification reminders, (3) medication trackers, (4) health surveys, and (5) a “To-Do” list outlining their tasks for the day, such as medications to take (see Figure 1). A key component of the WiseApp is a medication tracker linked to an electronic pill bottle (ie, smart pill box). The WiseApp sends tailored reminders based on the feedback from the linked device (ie, CleverCap), such as medication reminders if the pill bottle has not been opened. The WiseApp is unique because in most cases mHealth technology has been developed without incorporating patient-centered outcomes research. There are currently hundreds of applications (apps) for PLWH, yet they have not been conceptualized using evidence-based research and/or a patient-centered design.¹⁴ Consequently, research is needed to improve the understanding of how mHealth tools can be appropriately designed, functionally operated, and effectively used by PLWH to enable the dissemination of evidence-based information.¹⁴

Figure 1.

Screenshot of the WiseApp.

METHODS

Study design and participants

This study was a randomized controlled trial (RCT) of the WiseApp intervention group versus an attention control group on ART adherence in PLWH which was measured daily through the CleverCap. Recruitment was completed in New York City at HIV and dental clinics, and community-based organizations (NCT03205982). Online advertising through Craigslist and other social media tools were also used to recruit study participants. Study enrollment was from January 31, 2018 to April 13, 2021, and included a complete pause on enrollment due to COVID-19 from March to July 2020.

Eligibility criteria included: (1) 18 years of age or older, (2) have a diagnosis of HIV, (3) speak and understand English or Spanish, (4) live in New York City, (5) own a smartphone, (6) currently taking ART medications, and (7) report the past 30 days adherence of 80% or less as measured using the Visual Analogue Scale or have a viral load of over 400 copies/mL.

Columbia University served as the institutional review board for all study activities (the full trial protocol is available open access elsewhere).¹⁵ Written informed consent was obtained for each participant. Participants were incentivized for study visits ($40 for the initial visit, $50 at 3 months with up to $25 additional for completing app challenges, $60 at 6 months with up to $25 additional for completing health-related activities such as quizzes and activities within the app) and received the WiseApp, a FitBit, and CleverCap pill bottle at the initial visit. Both study arms also have a history tab to monitor whether or not they completed their assigned daily goals. Both study arms also received the CleverCap pill bottle, with only the intervention group linking the pill bottle to the WiseApp.

Intervention

The WiseApp comprised the following functional components: testimonials of lived experiences, push-notification reminders, medication trackers, health surveys, chat rooms, and a “To-Do” list outlining tasks for the day. Testimonials of lived experiences were drawn from publicly available videos comprised of content related to patients’ experiences disclosing their HIV status, communicating with providers about health problems, and overcoming ART adherence challenges. Both study arms received a fitness tracker (ie, FitBit) that connects to the WiseApp. The WiseApp intervention originated from formative work to design a self-management app for PLWH. The intervention group received videos and health surveys all centered on medication adherence and managing living with HIV and the intervention group also received daily app notification reminders for taking their medication.

Attention control

The attention control group was given access to a mHealth app developed by our study team which was comprised of health promotion videos and surveys focused on a healthy lifestyle including exercise, diet, and sleep. Participants were also given a step goal and reminders to reach 5000 steps per day to improve their cardiovascular health.

Randomization

Study participants were randomized (1:1) to the WiseApp intervention or the attention control arm with a variable-permuted randomized block design with the block size randomly selected between blocks.¹⁶ The treatment assignments in the block design were predetermined by the study statistician before beginning the RCT and remained static throughout the trial. Random assignments were concealed from the participants for the duration of their participation while study staff members were aware of the participant’s randomization. At the baseline visit, study staff opened a sealed study envelope and created the profile in the app that aligned with the study group.

Study assessments

Participants completed standardized quantitative assessments of demographic characteristics (ie, age, race/ethnicity, and employment status) and health literacy¹⁷ at baseline via Qualtrics. The primary outcome for this study was a change in ART adherence measured daily through the CleverCap^TM dispenser (Figure 2). The CleverCap automatically records each time a participant opens the dispenser. We collected adherence data each day from the start to the end of the trial (day 1–6 months), and it is a count response (number of times taking medication each day).

Figure 2.

CleverCap electronic monitoring pill bottle.

A number of secondary outcome measures were collected at baseline, 3 months, and 6 months. Blood draws were obtained through venipuncture at each study visit to measure CD4 counts and viral load. Self-reported adherence to ART was measured through the Center for Adherence Support Evaluation (CASE) Index.¹⁸ The CASE Adherence Index is a self-administered instrument with items scored such that higher values indicate better adherence, and the maximum total score is 16. Scores of 11 or higher on this index indicate good adherence (Cronbach's α = .79). Participants also completed the Healthcare Provider Engagement (HPE) Scale measuring their engagement with healthcare providers.¹⁹ Items are scored such that higher scores indicate a more negative relationship with their healthcare provider, where the maximum total score is 52. Participants also indicated their last primary care visit within the past year.

Statistical analysis

We enrolled 200 PLWH who reported suboptimal adherence to their ART regimen. We estimated that this would have greater than 80% power to detect at least a 10% difference in adherence to ART medication between the WiseApp intervention and the control arm. We made the following assumptions in our sample size and power calculation: a 75% retention rate by the end of the trial for both the control and intervention arms and that each person would be on a once-daily regimen, a conservative assumption of high intraclass correlation coefficient of .5 for the same participant at different times, and the adherence rate would be less than or equal to 80% at baseline. All calculations were based on a 2-sided test with alpha at .05 levels and power calculations being based on ART adherence.

We utilized a generalized linear mixed model (GLMM) to exam the WiseApp intervention efficacy on our primary outcome and secondary outcomes. The primary outcome, ART adherence, was measured daily from day 1 to the end of the RCT (or the last day before dropped out). A logistic GLMM with an individual-level random intercept was used to estimate the percent of ART adherence at each day. In this model, the dependent variable was ART adherence (coded as Yes/No) at each day, and the main dependent variables were (1) intervention arms; (2) days (from day 1 to end of the study); and (3) interaction term between arm and day. Since this is an RCT design, we report unadjusted results (https://www.ema.europa.eu/en/documents/scientific-guideline/guideline-adjustment-baseline-covariates-clinical-trials_en.pdf). The main analysis of all participants was followed by a subanalysis by limiting to participants with a detectable viral load at baseline.

Secondary outcomes, including CD4 count, viral load, primary care visits, HPE, and CASE adherence and App Use, were measured at baseline, 3 months, and 6 months. GLMMs were used with appropriate link functions based on types of outcomes: identity link for continuous outcomes, logit link for binary outcomes, and log link for count outcomes. In this model, outcomes were modeled at the 3 time points (baseline, 3, and 6 months). We calculated interaction terms between study arm (ie, intervention vs control) and each indicator for time following the baseline observation, indicating a difference in the rate of change from baseline to each timepoint across the 2 arms. GLMM estimates for longitudinal data were unbiased under the missing at random assumption. Analyses were conducted in SAS 9.4.

RESULTS

Sample characteristics

From July 2017 to April 2021, 1175 individuals were screened, 200 were enrolled, and 200 were randomly assigned to 2 arms: 99 individuals to the WiseApp intervention and 101 to the attention control arm (Figure 3).

Figure 3.

CONSORT diagram.

Demographic characteristics of our study participants are detailed in Table 1. In terms of race, 146 (74%) of participants were identified as Black/African American, 29 (15%) as unknown/missing, 13 (7%) as White, 1 (0.5%) as American Indian/Alaskan Native, and 2 (1%) as Native Hawaiian/other Pacific Islander. By ethnicity, 51 (26%) participants were identified as Hispanic/Latino (any race). Mean (SD) age was 49.3 (10.5) years. A total of 11 (6%) of participants were working full-time, 26 (13%) of participants were working part-time, and 102 (52%) of participants were unemployed or retired.

Table 1.

Characteristics of N = 198 WiseApp RCT participants, overall, and by study condition

Characteristics	Control group (n = 101)	Intervention group (n = 97)	Total (n = 198)
	Mean (SD)	Mean (SD)	Mean (SD)
Age (years)a	49.34 (10.2)	49.19 (10.9)	49.26 (10.5)
Sex at birthb	N (Col%)	N (Col%)	N (Col%)
Male	72 (72.0)	42 (43.8)	114 (58.2)
Female	28 (28.0)	54 (56.3)	82 (41.8)
Gender identity
Male	67 (66.3)	39 (40.2)	106 (53.5)
Female	28 (27.7)	55 (56.7)	83 (41.9)
Transgender male/transman/FTM	2 (2.0)	0 (0.0)	2 (1.0)
Transgender female/transwoman/MTF	4 (4.0)	2 (2.1)	6 (3.0)
Other	0 (0.0)	1 (1.0)	1 (0.5)
Sexual orientation
Homosexual/gay/lesbian	33 (32.7)	24 (24.7)	57 (28.8)
Heterosexual/straight	52 (51.5)	57 (58.8)	109 (55.1)
Bisexual	11 (10.9)	11 (11.3)	22 (11.1)
Queer	0 (0.0)	1 (1.0)	1 (0.5)
Asexual	0 (0.0)	1 (1.0)	1 (0.5)
Other	5 (5.0)	3 (3.1)	8 (4.0)
Race
African American/Black	71 (70.3)	75 (77.3)	146 (73.7)
American Indian/Alaskan Native	1 (1.0)	0 (0.0)	1 (0.5)
Native Hawaiian or Other Pacific Islander	2 (2.0)	0 (0.0)	2 (1.0)
White	8 (8.0)	5 (5.2)	13 (6.6)
Multiracial	2 (2.0)	5 (5.2)	7 (3.5)
Unknown/missing	17 (16.8)	12 (12.4)	29 (14.7)
Ethnicity
Hispanic/Latinx	27 (26.7)	24 (24.7)	51 (25.8)
Not Hispanic/Latinx	74 (73.3)	73 (75.3)	147 (74.2)
Relationship status
Single	62 (62.0)	52 (53.6)	114 (57.9)
In a relationship	24 (24.0)	33 (34.0)	57 (28.9)
Divorced/separated/widowed	12 (12.0)	12 (12.4)	24 (12.2)
Other	2 (2.0)	0 (0.0)	2 (1.0)
Employment status
Working full-time	6 (5.9)	5 (5.2)	11 (5.6)
Working part-time	14 (13.9)	12 (12.4)	26 (13.1)
Working off the books	1 (1.0)	3 (3.1)	4 (2.0)
Unemployed/retired	54 (53.5)	48 (49.5)	102 (51.5)
Student	2 (2.0)	5 (5.2)	7 (3.5)
Disabled	32 (31.7)	32 (33.0)	64 (32.3)
Highest education level
None	0 (0.0)	3 (3.1)	3 (1.5)
Elementary school	2 (2.0)	1 (1.0)	3 (1.5)
Some high school, no diploma	16 (15.8)	23 (23.7)	39 (19.7)
High school diploma or equivalent	25 (24.8)	34 (35.1)	59 (29.8)
Some college	41 (41.0)	15 (15.5)	56 (28.3)
Associate degree or technical degree	8 (7.9)	8 (8.3)	16 (8.1)
Bachelor/college degree	8 (7.9)	8 (8.3)	16 (8.1)
Professional or graduate degree	1 (1.0)	5 (5.2)	6 (3.0)
Annual income
Less than $10 000	52 (51.5)	49 (50.5)	101 (51.0)
$10 000–$19 999	25 (24.8)	22 (22.7)	47 (23.7)
$20 000–$39 999	13 (12.9)	8 (8.3)	21 (10.6)
$40 000 or more	2 (2.0)	5 (5.2)	7 (3.5)
Do not know	9 (8.9)	13 (13.4)	22 (11.1)
Health insurance source
Through job	3 (3.0)	3 (3.1)	6 (3.0)
Through someone else's job	1 (1.0)	1 (1.0)	2 (1.0)
Through affordable care act	5 (5.0)	3 (3.1)	8 (4.0)
Paid by respondent	1 (1.0)	3 (3.1)	4 (2.0)
Medicaid/Medicare	90 (89.1)	90 (92.8)	180 (90.9)
AIDS drug assistance program	6 (5.9)	4 (4.1)	10 (5.1)
Other source	1 (1.0)	2 (2.1)	3 (1.5)
Uninsured	2 (2.0)	1 (1.0)	3 (1.5)
Insurance status unknown	1 (1.0)	1 (1.0)	2 (1.0)
Confidence with medical forms
Extremely	61 (60.4)	48 (49.5)	109 (55.1)
Quite a bit	23 (22.8)	21 (21.7)	44 (22.2)
Somewhat	13 (12.9)	25 (25.8)	38 (19.2)
A little bit	3 (3.0)	2 (2.1)	5 (2.5)
Not at all	1 (1.0)	1 (1.0)	2 (1.0)
Last primary care visit
Last 3 months	87 (86.1)	88 (90.7)	175 (88.4)
3–6 months ago	8 (7.9)	6 (6.2)	14 (7.1)
6–9 months ago	2 (2.0)	1 (1.0)	3 (1.5)
9–12 months ago	1 (1.0)	0 (0.0)	1 (0.5)
I do not know	3 (3.0)	2 (2.1)	5 (2.5)
Indication of past-year alcohol problemb	47 (47.0)	38 (39.6)	85 (43.4)
Past 3-month substance use	74 (73.3)	60 (61.9)	134 (67.7)
Unstably housedc	6 (9.7)	13 (22.0)	19 (15.7)
Detectable viral load (≥20 copies/mL) via blood drawb	64 (63.4)	55 (56.7)	119 (60.1)
	Mean (SD)	Mean (SD)	Mean (SD)
ART adherence visual analog scalea	71.8 (26.1)	66.8 (26.6)	69.4 (26.4)
CASE adherence score	9.1 (3.1)	8.6 (2.9)	8.8 (3.0)
CD4 count (blood draw)d	570.8 (367.9)	564.1 (403.4)	567.5 (384.5)
Healthcare provider engagement score	17.4 (8.2)	18.2 (8.1)	17.8 (8.1)
Newest vital sign health literacy score	1.5 (1.7)	1.6 (1.4)	1.6 (1.5)
Post-study system usability questionnaire score	2.1 (0.9)	2.0 (1.0)	2.1 (0.9)
Mental health	Mean T-Score (SE)	Mean SE (T-Score)	Mean T-Score (SE)
PROMIS-29 depression subscalea	49.8 (4.3)	49.9 (4.3)	49.9 (4.3)
PROMIS-29 anxiety subscalea	51.2 (4.2)	52.5 (4.0)	51.9 (4.1)

^a n = 197 due to missing data.

^b n = 196 due to missing data.

^c n = 121 due to missing data.

^d n = 183 due to missing data.

SD: standard deviation; Col%: column percent; SE: standard error; percent may not add to 100% due to rounding error.

Primary outcome

The full sample for analysis of intervention efficacy was n = 198. Participants randomized to the WiseApp arm had significantly higher daily ART adherence rates compared to those randomized to the control arm from day 1 (69.7% vs 48.3%, OR = 2.5, 95% CI = 1.4–3.5, P = .002) to day 59 (51.2% vs 37.2%, OR = 1.77, 95% CI 1.0–1.6, P = .05) of the study period. Adherence rates then began to decline over the study period for both arms (P < .0001 for both arms) and the rate of decline was faster for the intervention arm as compared to the control arm (P < .0001). At day 60, there was no significant difference in ART adherence between the 2 study arms (Figure 4A), but the intervention arm had adherence rates than the control arm participants.

Figure 4.

ART adherence between study arms. (A) Full sample analysis; (B) subgroup analysis.

The sample of cases with detectable baseline viral load for subanalysis was n = 119. In this subsample, participants randomized to the WiseApp arm had a nonsignificant higher ART adherence compared to those randomized to the control arm right at the start of the intervention (50.6% vs 44.2%, OR = 1.3, P = .48). For both arms, adherence declined over time throughout the entire study period (P < .0001 for both arms). The decline was significantly faster for the intervention arm than that for the control arm (P < .0001), and by around 2 months, there was no difference in ART adherence between the 2 arms (Figure 4B).

Secondary outcomes

In all study participants, even those who were detectable at baseline, there was not a significant difference in detectable viral load change between the 2 arms (P = .89). For each arm, there was not a significant decrease in detectable viral load from baseline (P = .09–.49). For those randomized to the intervention arm, CD4 count did not significantly change from baseline to 3 months (P = .81) and 6 months (P = .11). For those randomized to the control arm, CD4 count declined significantly from baseline to 3 months (declined 16.2%, P = .04) and 6 months (declined 25.4%, P = .01). However, the difference change between the 2 arms was not statistically significant (P = .25). For those randomized to the intervention arm, CASE adherence index increased significantly from baseline to 3 months (increased 0.38 points, P = .0002) and 6 months (increased 0.39 points, P = .0002). For those randomized to the control arm, CASE adherence index increased significantly from baseline to 3 months (increased 0.38 points, P = .0001) and 6 months (increased 0.39 points, P = .0002). There was no significant difference in change in CASE adherence between the 2 arms (P = .99).

There was no observed difference in self-reported attendance at primary care visits between the 2 arms (P = .51). There was also no significant difference in healthcare provider engagement between the 2 study arms (P = .19). Retention rates for each of our study arms remained at or above 75% throughout the study. Our retention rates were based on both attendance at the study visits (see Figure 4) and on reports of daily adherence. At 3 months, we retained 84% of study participants in the intervention arm and 89% in the delayed intervention arm. At 6 months, we retained 80% of study participants in the intervention arm and 85% in the delayed intervention arm.

App Use was measured as the mean number of days that each participant used the app. At 3 months, intervention group participants used the app a mean of 57.8 days as compared to the control group which used the app a mean of 46.4 days with a statistically significant difference between the groups (P < .05). At 6 months, mean app use decreased in both groups to 42.0 days in the intervention group and 40.5 days in the control group with no significant difference between groups (P = .77).

DISCUSSION

The WiseApp for PLWH demonstrated significant improvements in ART adherence from baseline to 2 months. The effect was most pronounced at the start of the intervention with these effects waning over time. In the first 59 days after the start of the intervention, adherence was more than 20% higher for the intervention group as compared to the control group. In our subgroup analysis, we limited our analysis to those who were virally unsuppressed and saw a similar effect although it was less pronounced. These findings suggest the potency of the intervention especially during early use and suggest the need for a booster or a combination intervention approach after the initial effects of mHealth technology intervention wanes.

A seeming limitation of this study was the lack of concordance between the daily pill count data and biologic outcomes. This discrepancy is not substantive given that the pre-specified primary outcome of the study is pill bottle opening and is not viral suppression. Nonetheless, we acknowldge that validation of the behavioral results with biologic outcomes would strengthen the findings.

Despite this limitation, these findings are still significant given that we were able to significantly improve medication adherence during the first 2 months of the study. This has implications across health conditions given the need to ensure medication adherence and may be more aligned with the adherence threshold often applied for other chronic diseases such as hypertension.²⁰ Importantly, earlier iterations of antiretroviral medications required very high levels of adherence (95%),²¹ whereas today’s antiretroviral therapies are more potent, efficacious, and ultimately more forgiving. As a result, lower adherence levels are sufficient for maintaining viral suppression. Given this new reality, there is still limited data but studies suggest that 64–75% adherence rates may be adequate for viral suppression.^22,23

Importantly, our study was powered based on daily pill counts (180 data points vs 3 data points per person) which has exponentially more power than the 3 time points of viral suppression and self-reported adherence, collected at baseline, 3 and 6 months. As a result, we would expect there to be a discrepancy because these different outcomes, and it would be unlikely for us to find a significant result in viral suppression and self-reported adherence since these 2 measures were hugely underpowered as compared to daily pill counts.

The population demographic was a marginalized, predominantly male cohort with markers of poor HIV engagement. Participants came from a diversity of backgrounds and exhibited a wide range of vulnerabilities and barriers to care that could be encountered in urban settings in the United States.^24–30 Vulnerabilities common in our population, including active substance use (68%) and unstable housing (16%), have been previously associated with poor ART adherence.^{24–27,31,32} More than half of our participants had a high school education as the highest level of education; 75% of our participants had an annual income of less than $20 000 confirming the low socioeconomic status of this vulnerable study group (https://www.populationpyramid.net/canada/2015/).³³ We also purposefully included women in our study, with 42% of our participants having been assigned female sex at birth, and 8 transgender participants. This was important because women living with HIV and transgender persons have been identified as having poor ART adherence.³⁴

As a part of our study, we were able to assess the biologic effects of the behavioral intervention at both 3 and 6 months. There was no significant difference in change in viral suppression which may be partially explained by 40% of participants being virally suppressed at baseline. We also did not see an improvement in CD4 count for the intervention group participants. However, we did see a decline in CD4 count, a negative health outcome, in our control group participants, suggesting a small biologic effect of our intervention in this small study. Nonetheless, ART is very efficacious now as compared to at the beginning of the epidemic and so many patients showing poor adherence may not yet have shown nonsuppression of the virus due to the potency of the current drug regimens.

Notably, the study retention rates for our study remained above our goal of 75% retention. This is especially notable since most of our follow-up period for our study occurred during the COVID-19 pandemic. Our study was completely paused for over 3 months with no ability to conduct study visits. Once we resumed our study, challenges persisted with participants being unwilling and unable to attend the study visits because of COVID-19 infections, transportation barriers, and fear of exposure to COVID-19 in the absence of a vaccine, which was not yet available.

There are several limitations to our study design which may have attenuated the effects of the intervention. First, both study groups received the CleverCap pillbox. Although only the intervention group received medication reminders, the use of the CleverCap may have served as a placebo for attention control group participants because the CleverCap was used to store their medication and did light up after each use. Additionally, there are limitations to the use of the CleverCap since this was used to store only the primary ART medication for our study participants and some of the study participants had more than 1 medication which they took each day. Finally, this study was conducted during the COVID-19 pandemic. The challenges to healthcare engagement, filling of prescription medication, and accessing our research study team were notable and may have attenuated the effects of our study intervention findings.

In summary, this study of a cohort of racially and ethnically diverse and socio-economically depressed PLWH in NYC showed evidence of improved measures of HIV care when exposed to an mHealth self-management intervention. The study is unique in that participants were provided with pillboxes linked to a mobile phone app allowing for increased engagement of marginalized populations that would otherwise be difficult to access. This intervention could potentially bring substantial health benefits to vulnerable populations in an era in which HIV care is largely limited by medication adherence. There are implications for this intervention beyond HIV care. Medication adherence is critical for health maintenance in many chronic conditions such as diabetes and cardiovascular diseases which are both leading causes of death and major contributors to health disparities in the United States. Given the success in uptake and use of the WiseApp intervention during a 6-month trial, adaptation for other chronic conditions and evaluation of health outcomes is warranted.

FUNDING

The Agency for Healthcare Research and Quality grant number R01HS025071 (to RS). The content is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Healthcare Research and Quality. The funding source had no role in the design of the study; the collection, analysis, and interpretation of data; the writing of the report; or the decision to submit the paper for publication.

AUTHOR CONTRIBUTIONS

RS, HC, HJ, DM, NRR, and SO made substantial contributions to the conception or design of the work. RS, GS HC, HJ, DM, and SG made substantial contributions to the acquisition, analysis, and interpretation of data for the work. GS, HJ, and RS drafted the work. RS, GS, HJ, HC, BB, NRR, MG, DCM, SG, and SO critically revised the manuscript for important intellectual content. All authors provided final approval of the version to be published. All authors agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

CONFLICT OF INTEREST STATEMENT

None declared.

DATA AVAILABILITY

The data underlying this article will be shared on reasonable request to the corresponding author.