Violence as an obstacle to HIV medication adherence and viral suppression: A daily diary investigation among adolescent boys

Abstract

For the 1.6 million adolescents currently living with HIV, adherence to antiretroviral therapy (ART) is vitally important for reducing HIV-associated morbidity and mortality, and for preventing onward HIV transmission. Unfortunately, ART adherence is particularly low among adolescents. One important barrier to adherence is exposure to violence; however, the impact of cumulative versus acute exposure to violence on adherence is unclear. We tested whether violence exposure was associated with ART adherence and viral load over a one-year period among adolescent boys living with HIV in Soweto, South Africa. Participants (N = 239) completed weekly ecological momentary assessment (EMA) surveys of their recent experiences with violence and ART adherence. They provided biological samples at baseline and at the one-year follow-up to measure viral load. Higher average exposure to violence over the course of the study year was significantly associated with lower ART adherence (i.e., a cumulative impact). When we restricted violence exposure to intimate partner violence only, we found that participants were significantly less likely to take their HIV medication on days when they experienced physical intimate partner violence (i.e., an acute impact). We also found a significant positive association between exposure to violence over the course of the study and participants’ viral load at follow-up, even after controlling for baseline viral load. Our findings suggest that violence impacts two important HIV care outcomes through multiple pathways and highlight the importance of differentiating cumulative versus acute victimization in violence research. To improve health outcomes among adolescents with HIV and prevent new infections, it will be necessary to include boys in violence prevention and support services and to address violence exposure in HIV care.

Violence as an obstacle to HIV medication adherence and viral suppression: A daily diary investigation among adolescent boys

There are approximately 1.65 million adolescents and young adults living with HIV (ALHIV) globally (UNICEF, 2023), with the largest population residing in South Africa (UNAIDS, 2022). Antiretroviral therapy (ART) can effectively manage HIV, reducing morbidity and mortality among those who use it consistently (Volberding & Deeks, 2010). Moreover, consistent ART adherence can suppress viral load to the point where the virus is undetectable and can no longer be transmitted sexually (Cohen et al., 2016). Unfortunately, ART adherence is particularly low among adolescents, even those who acquired HIV perinatally and have been engaged in treatment for most of their lives (Adejumo et al., 2015; Enane et al., 2018; Salou et al., 2016). Increasing ART adherence among adolescents is necessary to improve health outcomes and prevent onward transmission to sexual partners. Researchers have identified several factors that may affect ART adherence, including exposure to violence (Hudelson & Cluver, 2015; Kim et al., 2017).

Violence victimization is a critical barrier to ART adherence and viral suppression. Previous research has identified an association between ART adherence and several types of violence, including emotional, physical, and sexual intimate partner violence (IPV) (Cluver et al., 2023; Kidman & Violari, 2018). Similar work has demonstrated associations between violence victimization from family members or peers and adherence (Casale et al., 2021) or viral load (Merrill et al., 2021). Beyond a history of victimization, it may also be important to consider the frequency and timing of violence, as some studies have found a cumulative effect of violence on adherence and viral load (Cluver et al., 2018; Hatcher et al., 2021; Kacanek et al., 2016). However, the majority of this research is done with girls and women or, less commonly, non-disaggregated samples (Hatcher et al., 2015; Johnson et al., 2023; Kuchukhidze et al., 2023).

It is crucial to extend the existing work on violence and ART adherence to include boys and men. In urban areas of South Africa, witnessing and experiencing violence is very common for boys (Closson et al., 2016; Meinck et al., 2016). Further, there are important gender differences in the type and timing of violence that boys experience. South African boys report more frequent violence, more types of violence (i.e., polyvictimization), and violence in more contexts (e.g., home, school, and community) than girls (Kaminer et al., 2013; Stansfeld et al., 2017).

Various types of violence may have differential impacts on boys’ wellbeing. For example, although boys and young men are typically conceptualized as perpetrators of IPV, they also experience victimization from their partners (Merrill et al., 2020). IPV victimization is stigmatized (Taccini & Mannarini, 2023), and often intersects with gendered and cultural stereotypes. South African men who have experienced abuse from their female partners report feelings of shame, emasculation, and isolation (Kgatle & Mafa, 2021). Many also believe that traditional support options for victims (e.g., police, clinics) are only for women (Majola et al., 2023). Because of these additional layers of stigma and isolation, IPV victimization may pose unique risks to adolescent boys’ wellbeing, resulting in negative coping or avoidance behaviors, including ART nonadherence (Sweeney & Vanable, 2016).

Moreover, we need new methods to create a more accurate and comprehensive picture of how violence affects ART adherence. With regard to the impact of cumulative violence victimization specifically, there are several hypothesized pathways linking it to later health behaviors. These include the neurobiological effects of trauma exposure (Laricchiuta et al., 2023), depression, PTSD, and substance misuse (Ndirangu et al., 2022; Quinn et al., 2024), and socioeconomic vulnerabilities, like reduced access to housing and healthcare (Glynn et al., 2021; Sprague et al., 2016).

In addition, violence may have an acute, proximal impact on behavior, though less work has been done in this area. One recent study found that adolescents in high-violence contexts who are exposed to community violence experience a same-day increase in stress, negative mood, and hostility – all of which have the potential to influence behavior (Deane et al., 2021). Exposure to violence is not static, but changes day to day. Similarly, ART adherence is not a static behavior, but should be viewed as an accumulation of daily choices to take (or not take) medication. Ecological momentary assessment (EMA) can effectively capture variability in adherence behavior on a day-to-day basis; it can also be used to estimate how much of this variability is due to having experienced violence that day (Shacham et al., 2019). Further, EMA methods capture behavior in real time and in a participant’s natural environment, reducing bias typically observed when reporting sensitive behaviors in study settings.

Delivering EMA with electronic data collection methods, like brief surveys via a mobile phone, has been shown to be acceptable and feasible with adolescent populations (Heron et al., 2017). Relevant to the current paper, researchers have used EMA methods to identify correlates of ART nonadherence, including mood and substance use, among adults in high-income countries (Cook et al., 2017; Shacham et al., 2019; Turner et al., 2020). To our knowledge, however, this approach has not been used to investigate whether violence influences ART adherence in adolescent boys from lower- or middle-income countries.

The purpose of this study was to test whether experiencing violence impacted ART adherence, and consequently viral load, in a sample of South African adolescent boys living with perinatally-acquired HIV. We used EMA surveys to sample a single day each week over the course of a year, asking participants about their experiences, behaviors, and mood at each time point. We hypothesized that recent violence victimization (i.e., in the past day or the past week) would be associated with ART nonadherence at that time point. Moreover, we hypothesized that participants who reported more frequent exposure to violence would be more likely to have a detectable viral load at the end of the study year.

Methods

Study setting and procedures

This study took place in Soweto, South Africa. Soweto, a township in Johannesburg, is a densely populated and majority-Black African urban area whose residents face high levels of unemployment and poverty (COGTA, 2020). Approximately 13% of the greater Johannesburg population is living with HIV (Johannesburg, 2020).

These data were collected as part of a larger study on the effects of violence on health behaviors among South African adolescent boys living with and without HIV. Given the research question, the present study analyzed data from participants living with HIV. Enrollment and data collection occurred between November 2020 and June 2023. Participants were eligible for the study if they were between 15 and 19 years old, male, living in or around Soweto, and in a current dating or sexual relationship. ALHIV were eligible for this study if they had a documented history of HIV before the age of ten, indicating likely perinatally-acquired HIV (PHIV). With advancements in ART effectiveness and accessibility, the number of adolescents with PHIV now outnumbers those who acquired the virus behaviorally (UNICEF, 2016). There are differences between these groups in their psychosocial experiences (e.g., stigma) and risk behaviors (e.g., risky sex), which may shape treatment engagement (Sherr et al., 2018). Therefore, recruitment was restricted to adolescents with PHIV.

To recruit ALHIV, study recruiters approached patients at public HIV care clinics. Those interested visited the research site at Perinatal HIV Research Unit, affiliated with the University of the Witwatersrand, for screening. Adolescents completed informed consent or assent procedures; caregivers also provided informed consent for those under age 18. Once enrolled, participants completed a baseline survey in the clinic and provided a blood sample to measure viral load. They then returned home with a smartphone to complete the EMA surveys.

Participants received a prompt to complete one EMA survey each week for one year. These surveys collected reports of participant mood, experiences, and behaviors both in the past 24 hours and over the past week. Additional information about the EMA survey structure has been published previously (Dietrich et al., 2024). Throughout the study, participants could communicate with the study team through a WhatsApp or 24-hour hotline. In addition, responses associated with potential psychological risks (e.g., suicidality) and reports of sexual or physical victimization triggered a notification to the South African study investigator. In accordance with documentation from the South African Department of Health, flagged participants were contacted by counselors who assessed the participant’s state, informed them of resources, and, if needed, provided referrals to care. One year after the baseline enrollment, participants returned to the clinic to complete a follow-up survey and provide a blood sample for viral load testing. The study was approved by the human research ethics committees of the University of the Witwatersrand in South Africa and Stony Brook University in the United States.

Measures

Independent variables:

At each EMA survey, participants reported experiences of verbal, physical, or sexual violence using three questions adapted from the World Health Organization Violence against Women Assessment (García-Moreno et al., 2005). Participants also indicated whether they witnessed violence in their home or community using an item adapted from the Adverse Childhood Experiences – International Questionnaire (WHO, 2012). Participants reported experiences with each type of violence across two time periods: within the past week and within the past 24 hours. Exposure was coded dichotomously (yes/no) by type of violence (i.e., verbal, physical, sexual, witnessed) and by time frame (i.e., past week and past 24 hours). We also created an overall dichotomous (yes/no) variable for any victimization, so that if a participant reported at least one type of victimization in the past week and past 24 hours, they were categorized as experiencing any victimization during that time frame. If participants reported any verbal, physical, or sexual violence, they were asked to indicate who had hurt them from a list of potential perpetrators including an intimate partner (i.e., a casual sexual partner, boyfriend/girlfriend, or spouse), a family member, a peer, and a stranger. The full list of violence measures is available in the Appendix.

Dependent variables:

To measure ART adherence, at each mobile survey participants were asked “In the past 24 hours, how did you do taking your HIV medicines?” Participants who reported taking all of their medicines were categorized as adherent. Participants who reported missing or skipping any pills were categorized as nonadherent.

Viral load was measured at the baseline survey and again at the follow-up survey one year later. Viral suppression at both time points was defined as < 200 copies/mL.

Covariates:

The baseline survey collected sociodemographic information from participants to be included as controls in the analyses. We asked participants to report their age in years (continuous), their race (categorical, condensed to Black African or other) and their enrollment in school (dichotomous, enrolled in school or not enrolled in school).

Analyses

Descriptive statistics were used to characterize the study sample, to calculate rates of violence victimization on a participant- and survey-level, and by type of perpetrator (intimate vs. nonintimate), and to calculate adherence and viral suppression.

To model the impact of daily (past 24-hour) violence victimization on ART adherence, we used multivariable generalized linear mixed effect models (GLMMs). The dichotomized variable representing victimization was analyzed as a between-subject (interindividual) component and within-subject (intraindividual) component. The between-subject component was the participant’s mean level of victimization across the surveys over the study year (unit = 0.1). The within-subject component was the deviation of the victimization report on a given survey from the participant’s mean, which indicated their daily variation from their average level (unit = 1.0). We created separate models for each type of victimization (i.e., verbal, physical, sexual violence, witnessing violence, or any violence), with the between-subject and within-subject components of victimization as the independent variables and ART adherence as the outcome. We created both a base model and a second model adjusting for participant demographics (i.e., age, race, school enrollment) and survey week. All GLMMs include a random effect at the participant level.

To distinguish the temporal ordering of violence and adherence, we ran additional sensitivity analyses using the same GLMMs as above, with past-week violence instead of past 24-hour violence.

To explore the impact of intimate partner violence specifically, we created a new subset of GLMMs for physical and verbal violence, using only reports of victimization by participants’ intimate partners. We created separate models for physical and verbal IPV victimization in the past 24 hours and in the past week, with the between-subject and within-subject components of victimization as the independent variables and ART adherence as the outcome. We ran both a base model and a second model adjusting for covariates (i.e., age, race, school enrollment, survey week) with a random effect at the participant level.

To analyze the impact of violence victimization on viral suppression at follow-up, we used multivariable logistic regression. For each type of victimization, the exposure variable was defined as the proportion of weeks that an adolescent experienced victimization. We created a base model which controlled for participant demographics (i.e., age, race, and school enrollment) as confounding effects, and a second model further controlled for participants’ viral load at enrollment. Statistical analysis was performed using SAS 9.4 (SAS Institute, Inc., Cary, NC) with a significance level set at 0.05.

Results

Impact of daily victimization on ART adherence

Daily adherence

There were 251 ALHIV enrolled in the study. Of these, 239 submitted at least one mobile survey with ART adherence data. On average, participants completed 28.5 (54.8%) of the 52 weekly surveys. The mean age was 17 years old, and most participants self-identified as Black African (91%) and were enrolled in school (76%). In the analysis of associations between victimization experiences and ART adherence, a total of 6,624 EMA survey observations (O) were used. At the survey level, participants reported taking all of their pills in 5,613 (85%) surveys. When aggregated at the participant level, the average participant reported medication adherence 81% of the time, with a standard deviation of 27%.

Violence victimization

Table 1 describes participants’ reported experiences with victimization. At the survey level, participants reported any violence in 901 surveys (14%). At the participant level, the majority (81%) reported experiencing at least one type of violence in any week over the survey year: witnessing violence was most common (59%), followed by experiencing verbal (55%), physical (46%) and sexual violence (25%). On average, participants reported any victimization on 9% of the days over the survey year.

Table 1.

Descriptive table of past 24-hour and past-week victimization, aggregated by surveys and by individuals

Any victimization reported, by survey (O=6,624)
	Past 24-hours	Past week
Any victimization	513 (8%)	901 (14%)
Witnessed violence	223 (3%)	384 (6%)
Physical victimization	124 (2%)	255 (4%)
Verbal victimization	254 (4%)	449 (7%)
Sexual victimization3	51 (1%)	110 (2%)
Any victimization reported, by participant (N=239)
	Past 24-hours	Past week
Any victimization	144 (60%)	194 (81%)
Witnessed violence	94 (39%)	140 (59%)
Physical victimization	69 (29%)	109 (46%)
Verbal victimization	93 (39%)	132 (55%)
Sexual victimization	29 (12%)	59 (25%)
Mean victimization rate, by participant (N=239)
	Past 24-hours	Past week
Any victimization	9%	16%
Witness violence	4%	7%
Physical victimization	2%	5%
Verbal victimization	5%	8%
Sexual victimization	2%	3%

Impact of violence victimization on daily adherence

Table 2 presents the regression results for the impact of victimization on daily ART adherence. Based on the within-person component, there was no significant association between daily fluctuations in victimization and medication adherence. However, examining the between-person component revealed that individuals who experienced higher average levels of victimization were less likely to adhere to treatment. To represent a meaningful change, we modelled the effect of a 10% increase in victimization. Boys who had a 10% greater mean level of any victimization, compared to other boys, had about a 30% reduction in the odds of taking their HIV medication (OR 0.74; 95% CI 0.61, 0.90). Sensitivity analyses using past-week victimization, rather than past 24-hour victimization, produced a very similar estimate (OR 0.75; 95% CI 0.65, 0.87). Likewise, the estimates changed very little between types of victimization (OR 0.64 to 0.75 across both time frames), though occasionally failed to reach significance.

Table 2.

Generalized linear mixed effect regression results of past-week and past 24-hour victimization on past 24-hour medication adherence

		Model I		Model II1
Victimization, by type and timing	Unit	OR (95% CI)	p 2	OR (95% CI)	p 2
Any victimization [past 24 hours]
within-person component	1	0.82 (0.60, 1.12)	.22	0.84 (0.61, 1.14)	.26
between-person component	0.1	0.73 (0.60, 0.89)	.001	0.75 (0.62, 0.90)	.003
Witnessed violence [past 24 hours]
within-person component	1	0.97 (0.61, 1.54)	.91	0.99 (0.62, 1.57)	.96
between-person component	0.1	0.70 (0.52, 0.93)	.01	0.71 (0.53, 0.96)	.02
Physical victimization [past 24 hours]
within-person component	1	0.67 (0.37, 1.22)	.19	0.67 (0.37, 1.23)	.20
between-person component	0.1	0.58 (0.33, 1.03)	.06	0.64 (0.36, 1.13)	.12
Verbal victimization [past 24 hours]
within-person component	1	0.98 (0.64, 1.50)	.93	1.00 (0.65, 1.53)	.99
between-person component	0.1	0.62 (0.46, 0.83)	.002	0.65 (0.48, 0.87)	.004
Sexual victimization [past 24 hours]
within-person component	1	0.42 (0.16, 1.09)	.07	0.43 (0.16, 1.11)	.08
between-person component	0.1	0.71 (0.50, 1.00)	.049	0.72 (0.52, 1.01)	.06
Any victimization [past week]
within-person component	1	0.82 (0.64, 1.05)	.12	0.84 (0.65, 1.07)	.16
between-person component	0.1	0.74 (0.64, 0.86)	<.001	0.75 (0.65, 0.87)	<.001
Witnessed violence [past week]
within-person component	1	0.81 (0.57, 1.15)	.24	0.82 (0.58, 1.17)	.28
between-person component	0.1	0.72 (0.57, 0.92)	.007	0.74 (0.59, 0.94)	.014
Physical victimization [past week]
within-person component	1	0.72 (0.48, 1.09)	.12	0.73 (0.48, 1.10)	.13
between-person component	0.1	0.63 (0.46, 0.86)	.004	0.65 (0.48, 0.89)	.007
Verbal victimization [past week]
within-person component	1	0.94 (0.67, 1.32)	.73	0.97 (0.69, 1.35)	.84
between-person component	0.1	0.72 (0.59, 0.88)	.001	0.74 (0.61, 0.90)	.003
Sexual victimization [past week]
within-person component	1	0.70 (0.36, 1.36)	.29	0.71 (0.37, 1.37)	.31
between-person component	0.1	0.73 (0.53, 1.01)	.054	0.75 (0.55, 1.03)	.07

¹Covariates: age, race, school enrollment, and survey week.

²p-values were from Type III Tests based on generalized linear mixed-effects model.

Reported intimate partner violence victimization

When respondents reported victimization, they were asked to indicate whether the perpetrator was an intimate or non-intimate partner (Table 3). We used this information to understand whether participants’ experiences of physical and verbal victimization varied by perpetrator type. In some cases, the respondent had experienced violence from more than one perpetrator in the reporting period. Participants reported more physical and verbal violence from non-intimate compared to intimate partners.

Table 3.

Reports of victimization by partner type

Victimization reported, by survey (O=6,624)
	Past 24-hour	Past week
Physical victimization
Intimate partner	36 (1%)	71 (1%)
Non-intimate partner	94 (1%)	186 (3%)
Verbal victimization
Intimate partner	44 (1%)	72 (1%)
Non-intimate partner	215 (3%)	371 (6%)
Victimization reported, by participant (N=239)
	Past 24-hour	Past week
Physical victimization
Intimate partner	17 (7%)	33 (14%)
Non-intimate partner	57 (24%)	91 (38%)
Verbal victimization
Intimate partner	26 (11%)	45 (19%)
Non-intimate partner	81 (34%)	113 (47%)

Reported intimate partner violence victimization and daily adherence

We examined the association between physical and verbal IPV and ART adherence (Table 4). We did not find evidence that the between-person component of IPV was associated with adherence. Rather, there was a significant association between the daily fluctuation of physical IPV victimization (the within-person component) and ART adherence. Participants’ odds of taking their medication were about 70% lower on days when they reported experiencing physical IPV in the past 24-hours (OR 0.28; 95%; CI 0.10, 0.76). This association was slightly attenuated for past-week physical IPV (OR 0.45; 95% CI 0.20, 0.97).

Table 4.

Generalized linear mixed effect regression results of past week and past 24-hour intimate partner physical and verbal victimization on past 24-hour adherence

		Model I		Model II1
Intimate Partner Victimization, by type and timing	Unit	OR (95% CI)	p 2	OR (95% CI)	p 2
Physical victimization [past 24 hours]
within-person component	1	0.28 (0.10, 0.77)	.013	0.28 (0.10, 0.76)	.013
between-person component	0.1	0.66 (0.29, 1.53)	.33	0.81 (0.34, 1.89)	.62
Verbal victimization [past 24 hours]
within-person component	1	0.67 (0.25, 1.76)	.41	0.67 (0.26, 1.77)	.42
between-person component	0.1	0.59 (0.31, 1.11)	.10	0.67 (0.35, 1.27)	.22
Physical victimization [past week]
within-person component	1	0.45 (0.21, 0.98)	.044	0.45 (0.20, 0.97)	.041
between-person component	0.1	0.81 (0.52, 1.27)	.36	0.89 (0.56, 1.40)	.60
Verbal victimization [past week]
within-person component	1	0.92 (0.40, 2.11)	.84	0.92 (0.40, 2.11)	.84
between-person component	0.1	0.76 (0.50, 1.17)	.22	0.82 (0.53, 1.26)	.36

¹Covariates: age, race, school enrollment, and survey week.

²p-values were from Type III Tests based on generalized linear mixed-effects model.

Reported violence victimization and viral load at follow-up

Viral load

We captured HIV viral load at baseline and at the one-year follow-up visit for 227 participants who had EMA data on violence victimization. About two-thirds of participants were virally suppressed at follow-up (Table 5).

Table 5.

Distribution of viral load change from baseline to follow-up

Viral Suppression Trajectory	Participants (%)
Stayed Suppressed (<200 copies/mL)	141 (58%)
Suppressed → Detectable	23 (10%)
Detectable → Suppressed	19 (8%)
Stayed Detectable	44 (18%)

Reported violence victimization and viral suppression at follow-up

We examined the association between exposure to violence and viral load at follow-up. Average exposure was calculated using participants’ past-week reports of violence in the mobile surveys, as these were more inclusive of all violent episodes than the past 24-hour reports. Multiple types of victimization were associated with increased odds of having a detectible viral load (e.g., OR 1.35 for a 10% increase in experiences of any violence), after adjusting for participant demographics (Table 6, Model I). Even after adjusting for baseline viral load (Model II), a 10% increase in exposure to any violence was still associated with a 27% increase in the odds of having a detectible viral load at follow-up.

Table 6.

Logistic regression model results of victimization rate on detectible viral load (>=200) at follow-up, with and without adjustment for baseline viral load

		Model I1		Model II2
Victimization	Unit	OR (95% CI)	p 3	OR (95% CI)	p 3
Any victimization	0.1	1.35 (1.15, 1.60)	< .001	1.27 (1.05, 1.54)	.014
Witnessed violence	0.1	1.39 (1.06, 1.82)	.017	1.25 (0.89, 1.77)	.20
Physical victimization	0.1	1.78 (1.24, 2.55)	.002	1.49 (0.96, 2.33)	.08
Verbal victimization	0.1	1.25 (1.02, 1.53)	.034	1.11 (0.86, 1.43)	.42
Sexual victimization	0.1	1.20 (0.91, 1.58)	.21	1.14 (0.80, 1.61)	.47

^1:Covariates: age, race, school enrollment

^2:Covariates: age, race, school enrollment and baseline viral load

^3:p-values were from Type III tests based on logistic regression models.

Discussion

ART adherence and viral suppression are vitally important for maintaining the health of ALHIV and for preventing new infections. In this study of South African boys living with HIV, we found negative associations between experiencing violence and two key HIV care outcomes (i.e., ART adherence and viral suppression). These findings highlight exposure to violence as an important structural determinant of health, especially in relation to HIV. However, the study also uncovered conflicting evidence on the relative role of cumulative and same-day exposure to violence, raising new questions about how violence exerts influence over ART adherence. These results have several important implications for how we might disrupt the pathways from violence to nonadherence, and ultimately improve viral suppression.

Violence exposure was common among study participants. Over three-quarters of the sample reported any violence and about half reported witnessing violence, physical victimization, or verbal victimization over the study year. Physical and verbal IPV victimization were less common, though still reported by 14% and 19% of the sample, respectively. The rates of violence over the study year are higher than past-year estimates reported in other studies of South African adolescents (Cluver et al., 2023; Meinck et al., 2016). It is possible that these differences are due, in part, to differences in data collection methods (i.e., asking adolescents to recall the past week vs. the past year). Researchers might consider using shorter, repeated measures surveys to collect more accurate data on adolescent experiences.

We found that boys who had experienced more violence were substantially less likely to take their medicine. Specifically, the between-person component of violence victimization, or the average exposure level over time, was an important determinant of ART adherence. This finding held across different types of victimization. However, we did not find similar evidence that exposure to violence exerted its influence on the same day (or even within the same week). The within-person component, or daily fluctuations from the average, trended in the expected direction but did not reach statistical significance. These findings align with previous, cross-sectional work, which also concludes that cumulative violence exposure is associated with ART adherence (Cluver et al., 2018). The EMA structure of this study allows us to move beyond traditional designs and clarify the role of timing by comparing cumulative violence to daily changes in exposure. Specifically, these findings indicate that violence is having a lasting impact on adolescents’ adherence behaviors, but individual incidents of violence are not triggering an immediate behavioral response. This difference in timing might suggest potential mediators in the pathways between violence exposure and adherence behavior.

Interestingly, our findings and interpretation differ when we focus on exposure to intimate partner violence (IPV) only, as opposed to violence from any perpetrator. We found a significant association between the within-person component of physical IPV and ART adherence. That is, participants were far less likely to take their medication on days when they also experienced physical IPV. This is in line with a study among U.S. adults, which found that it was participants’ momentary affective state that affected ART adherence, and not their average mood (Cook et al., 2017). That study differed from ours in that it focused on changes in mood rather than changes in violence exposure.

Understanding how violence operates at the trait-level (i.e., the between-person component) and state-level (i.e., the within-person component) can provide insights into the mechanism. For example, the trait-level influence of violence on adherence may point to psychosocial pathways, whereby repeated exposure to violence negatively affects an individual’s cognitions and temperament. This may lead to the development of mental illness and maladaptive behaviors which then reduce adherence over time. More specifically, potential mediators of the association between cumulative violence exposure and health behaviors in adolescents include depression (Connolly et al., 2022), post-traumatic stress disorder (Brief et al., 2004), suicidal ideation (Farrell & Zimmerman, 2019), maladaptive coping behaviors (e.g., substance use (Gonzalez et al., 2011)), the stigma of victimization (Arnocky & Vaillancourt, 2014), and chronic stress, including the physiological and neurocognitive impacts of allostatic overload (Guidi et al., 2020).

Alternatively, the state-level influence of IPV on adherence suggests that adolescent boys are exhibiting an immediate behavioral response to physical violence from a partner. It is possible that boys are experiencing stress and negative emotions in the immediate aftermath of IPV (Deane et al., 2021), which interrupts adherence (Cook et al., 2017; Dalton & Hammen, 2018). IPV may also affect medication behaviors because it poses a unique threat to the boy’s self-identity. Violence in the community and from authority figures, like parents, is common for South African adolescent boys (Closson et al., 2016). Violence from an intimate partner is distinct: it may contradict sociocultural norms that support men’s superiority over women (Russell et al., 2014) and define only girls and women as victims (Majola et al., 2023). This may create an immediate threat to boys’ self-concept and have a more acute link to medication adherence.

Finally, it is possible violence exhibits both trait- and state-level influences on adherence. In all of our analyses, both the between-person and within-person effects of violence trended in the expected direction, even when they did not reach significance. There may be a relative balance between these two effects, depending on the event, the perpetrator, and violence exposure considered. This balance may explain why some studies have found mixed or null effects of cumulative IPV victimization on medication adherence (Miller et al., 2023; Rotheram-Borus et al., 2019), further underscoring the need to consider differences between cumulative and acute influences of violence in research designs.

Importantly, we found an association between violence victimization and later viral suppression, with the association remaining significant even after adjusting for baseline viral suppression. These findings are consistent with previous research from the U.S. and Zambia (Kacanek et al., 2016; Merrill et al., 2021). We build on the prior research by collecting weekly assessments of victimization that are less prone to recall bias and establishing a temporal order. Viral suppression is critical for the health of people living with HIV and reducing the risk of onward transmission; therefore, these findings highlight the need to address violence victimization in order to manage the HIV epidemic.

Strengths and limitations

Our findings underscore the importance of carefully considering both the frequency and the timing of violence victimization on HIV care outcomes. Using EMA methods, we collected data at multiple timepoints, in real time, and in a real-world setting. Our approach and analytic methods enabled us to test the influence of both average levels of and daily fluctuations in violence exposure, as well as establish a temporal order for key analyses. For instance, violence exposure always preceded the measurement of viral suppression. In the primary analyses of violence and adherence in the past 24 hours, we cannot establish the temporal order. In sensitivity analyses, however, we better distinguished the temporal order by modelling past-week reports of violence and adherence in the past 24 hours, with little change in the association. Future research should seek to further clarify how the timing, frequency, and duration of violence victimization impacts adherence behaviors. Similarly, we asked about the type and perpetrator of violence at each survey, revealing important nuances in the pathways between victimization and adherence across violence experiences. Finally, the study captured both the proximal risk behavior (i.e., adherence) and the cumulative impact on health (i.e., viral load). The association between violence and viral suppression establishes that violence is having a demonstrable impact on an objective and critical health outcome.

There are also limitations to the study. Although participants completed their surveys on private devices, the questions asked about sensitive behaviors, and participant answers may have been subject to social desirability bias. Another challenge is that the EMA survey captured IPV with only a few brief questions. We did not collect information on who initiated violence between intimate partners, and it is possible that boys reported victimization during the same episode in which they were also perpetrators. Finally, this sample focused on boys with perinatally-acquired HIV who we recruited from HIV clinics. The results may not be generalizable to ALHIV who acquired HIV behaviorally, who are not clinically engaged, or who reside in very different contexts.

Implications

This study has several important implications. Guidance from UNAIDS (2021) recommends integrating psychosocial support in HIV care, including support for gender-based violence experienced by girls and young women. Although these services are absolutely needed, our results show that adolescent boys are also frequent victims of violence – not only from partners, but from family members, peers, and others in the community – with resulting detrimental effects on ART adherence. Thus, when integrating violence services in HIV care, it will be important to include outreach to and engagement of boys.

It is also necessary to develop interventions to increase ART adherence among adolescents. ALHIV have lower adherence levels than adults (Adejumo et al., 2015; Enane et al., 2018), but recent systematic reviews and meta-analyses have concluded that there are few interventions designed for adolescents in low- and middle- income countries that significantly improve ART adherence (Alcon et al., 2020; Griffee et al., 2022; Ridgeway et al., 2018). However, a recent meta-analysis of interventions from both high-income and low- and middle-income contexts found that incorporating psychosocial components (e.g., motivational interviewing, cognitive-behavioral therapy, peer support) in interventions improves both ART adherence and viral load (Laurenzi et al., 2021) Therefore, it may be beneficial to prioritize adolescents’ physical and psychological wellbeing, with ART adherence as a secondary goal. For example, a family-based intervention program, designed to strengthen support and communication between ALHIV and their caregivers, improved HIV knowledge and medication adherence in South Africa and Thailand (Bhana et al., 2010; Nestadt et al., 2019). Similarly, community-based interventions that connect adolescents to peer mentors (i.e., ALHIV trained to deliver counseling and support) improve adolescents’ mental health and viral suppression over time (Denison et al., 2020; Mavhu et al., 2020). With these examples in mind, future interventions might aim to reduce adolescents’ exposure to violence, provide peer support for violence-affected youth, and discuss positive means of coping with violence in order to improve HIV outcomes.

Further, it may be necessary to develop multiple, intersecting interventions that address the various mechanisms linking violence and adherence. This study found that cumulative exposure to violence was associated with ART nonadherence. Longstanding patterns of violence in an adolescents’ community, family, and social groups likely impact adherence through psychosocial pathways, by increasing their vulnerability to psychological conditions like depression (Connolly et al., 2022) and substance misuse (Gonzalez et al., 2011) and reducing their ability to access positive coping behaviors. Clinicians could consider building partnerships with community programs that address mental and behavioral health and developing a streamlined protocol to screen adolescents for violence exposure and refer them to these community partners (Grimes et al., 2024). This study also found that exposure to IPV was associated with same day ART nonadherence, suggesting that boys have an acute reaction to violence from intimate partners. School-based interventions can teach adolescents skills to navigate and cope with relationship conflict, like emotional regulation, de-escalation, and conflict management (Cheong et al., 2015; Eadeh et al., 2021).

Moreover, community-level interventions that reduce stigma against people, including men, who have experienced IPV may indirectly improve adherence. Adolescents may also experience increased risk of violence when accessing treatment; many patients report walking or taking public transportation to clinics and having to wait in long lines to access care, leaving them vulnerable to harassment or community violence (Kagee et al., 2011). Infrastructure changes that improve access to public transportation and increase staffing and security in clinics could reduce these risks. Similarly, targeted interventions within the healthcare system are necessary to engage men in preventive care, including screening for violence victimization (Beia et al., 2021). Finally, amending South African legislative and policy documents on violence prevention to explicitly include men as potential victims of violence could ameliorate male victims’ feelings of isolation and helplessness (Van Niekerk et al., 2015).

Conclusion

We found that violence negatively impacts ART adherence and viral suppression among boys living with HIV in South Africa. To our knowledge, this is the first study to test both between-person and within-person effects of violence on ART adherence, and the first to isolate the relationship among boys specifically. Our findings highlight the importance of differentiating accumulated versus acute victimization in violence research. In order to improve the health outcomes of ALHIV and prevent secondary transmission, it is necessary to include boys in violence prevention and support services and to integrate violence in HIV care programs.

Appendix

List of Violence Measures

Survey question	Answer options	Adapted from
Witnessed violence
In the past week, did you see someone being stabbed, shot or threatened with a knife/gun in your community?	Yes No Refused	ACE-IQ
You said that you saw someone being stabbed, shot or threatened. Has this happened in the past 24 hours?	Yes No Refused
Physical violence
In the past week, did someone hit, push, slap, choke or otherwise physically hurt you?	Yes No Refused	WHO Assessment
Who physically hurt you?	Girlfriend/wife Boyfriend/husband Female casual sexual or romantic partner Male casual sexual or romantic partner Family member Peer/someone my own age Female stranger Male stranger Refused
Has this happened in the past 24 hours?	Yes No Refused
Verbal violence
In the past week, did someone threaten to hurt you, or insult you or make you feel bad about yourself?	Yes No Refused	WHO Assessment
Who threatened or insulted you?	Girlfriend/wife Boyfriend/husband Female casual sexual or romantic partner Male casual sexual or romantic partner Family member Peer/someone my own age Female stranger Male stranger Refused
Has this happened in the past 24 hours?	Yes No Refused
Sexual violence
In the past week, did someone have forced sex with you or touch you in a sexual way when you did not want them to?	Yes No Refused	WHO Assessment
Who had forced sex/touched you?	Girlfriend/wife Boyfriend/husband Female casual sexual or romantic partner Male casual sexual or romantic partner Family member Peer/someone my own age Female stranger Male stranger Refused
Has this happened in the past 24 hours?	Yes No Refused