Predictors of low antiretroviral adherence at an urban South African clinic: A mixed-methods study

Background Low adherence to antiretroviral treatment (ART) in people living with HIV (PLHIV) remains a critical issue, especially in vulnerable populations. Although ART is responsible for greatly reducing the mortality and morbidity associated with HIV, low treatment adherence continues to impact the effectiveness of ART. Considering that a high level of adherence to ART is required for the excellent clinical outcomes with which ART is often associated, understanding the complex contextual and personal factors that limit high levels of treatment adherence remains paramount. Poor adherence remains an issue in many South African communities many years after the introduction of ART. Objectives Our study sought to understand the specific factors and the interactions among them that contribute to non-adherence in this patient population in order to devise successful and contextually appropriate interventions to support ART adherence in PLHIV. Methods This mixed-methods study employed a study-specific questionnaire (N = 103) and semi-structured interviews (N = 8) to investigate the factors linked to non-adherence at the Heideveld Community Day Centre in Cape Town, South Africa. Results Over half (57.3%) of participants were ART non-adherent. Non-adherence was correlated with younger age, negative self-image and a low belief in the necessity of ART (P < 0.05). In patient interviews, alcohol use, treatment fatigue and stigmatisation emerged as contributors to suboptimal adherence. Conclusion The results suggest that there remains a need for context-sensitive interventions to support PLHIV in South African communities. Future research needs to ensure that these targeted interventions take these factors into consideration.


Introduction
Adherence to antiretroviral treatment (ART) in people living with HIV (PLHIV) promotes suppression of the HIV viral load and reduces patient morbidity and mortality. Yet despite ART's promotion of greatly improved health outcomes for PLHIV, low treatment adherence undermines its effectiveness.
Importantly, structural barriers have been highlighted by the literature as potential factors contributing to difficulties with adherence. Poverty, transportation costs faced by low-income individuals and high levels of alcohol abuse are some of the factors previously shown to be associated with disengagement from ART. 3,4,12 Therefore, a combination of individual-level factors, such as side effects, a lack of social support, local conditions, and structural and systemic barriers, likely interact to contribute to a patient's overall level of adherence. Considering that clinic-patient discussions about personal barriers to adherence have been previously shown to mediate adherence levels, 5,13 understanding the factors contributing to a patient's low adherence and the interaction among those factors is crucial if successful and contextually appropriate interventions to support adherence are to be devised.
This mixed-methods study sought to expand and update knowledge about the contributors to poor adherence to ART in sub-Saharan Africa by investigating the factors currently mediating differential levels of adherence in a diverse urban patient population. Considering that the factors contributing to non-adherence may manifest differently based on temporal context and locale, it is crucial to understand the current factors contributing to suboptimal ART adherence so that health systems may plan efficacious interventions for their patient populations.

Methods
This was a sequential, exploratory mixed-methods study that employed a combined quantitative and qualitative research design in distinct phases. The initial phase was a quantitative design and was questionnaire-based. The second phase was qualitative in nature and was conducted to expand on our quantitative data.

Study setting and population
The study was conducted at the Heideveld Community Day Centre (CDC), a public 8-h day clinic operating in the Cape Town Metro Health District in Cape Town, South Africa. Located in the community of Heideveld, the clinic acts as a first line of access for PLHIV in Heideveld and in surrounding townships. The clinic serves a mostly low-income population and offers various health services, including chronic care, emergency care, ART services and social work counselling.
Patients who are HIV positive attend clinic appointments with registered nurses or a medical doctor. Antiretroviral treatment is also distributed to patients on site.

Eligibility
A convenience sample was used for both phases. Initial identification of potential participants and recruitment procedures occurred outside the consultation room after patients' appointments at the antiretroviral (ARV) clinic. Participants in the quantitative phase met the following eligibility criteria: age ≥ 18 years old and previously diagnosed with HIV. All participants were registered as patients at the Heideveld CDC.
Qualitative data collection was limited to HIV-positive individuals who met the eligibility criteria above and indicated non-adherence to their ART based on their responses in the preceding quantitative phase. Nonadherence was assessed based on three adherence questions: • Sometimes if you feel worse, do you stop taking your HIV medication? (yes/no) • Over the past weekend, did you take all of your HIV medication? (yes/no) • In the last 30 days, on how many days did you miss at least one dose of any of your ARV medication? Write in number of days: ____ Non-adherents were classified as those who answered 'yes' to question 1 and/or 'no' to question 2 and/or reported missing two or more doses in the preceding 30 days.

Recruitment and data collection Quantitative
Quantitative data, including demographic information, stigma assessments, and patients' beliefs and concerns about their ART, were collected between September and December 2019 at the Heideveld CDC. Participants were recruited by research staff operating inside the clinic within discreet and private rooms. Research staff utilised information sheets in the patient's language of choice (English, Afrikaans or isiXhosa) to describe the study's goals and the option for participation in the study. Interested patients were directed to appropriate members of the team, who provided additional information about the exact nature of the study.
Informed consent procedures as well as questionnaires were administered to participants by trained research staff members conversant in English, Afrikaans and/or isiXhosa. All processes took place in private rooms within the clinic, and participants were advised that they could withdraw at any point. After completing a questionnaire, participants were offered R80.00 (about $6.00) as compensation for their time.

Qualitative
Qualitative data collection was conducted between January 2020 and March 2020 in the Heideveld CDC. Individuals who self-reported as ART non-adherent based on the study criteria and who previously agreed to be contacted telephonically were eligible for participation in the qualitative phase. Semistructured interviews (n = 8) lasted approximately 45 min and occurred in private office spaces or other locations within the clinic. Interviewees received R100.00 (around $7.00) for participation.

Quantitative measures
All questionnaires were translated from English to Afrikaans and isiXhosa by professional translators. The surveys were then back-translated by other translators to confirm the accuracy of the translations.

Demographics
The questionnaire collected demographic data, including age, sex, education level, the number of months on ARVs and estimated income levels.

Adherence levels
Adherence to ARVs was assessed using three different questions ( Table 1). The first two questions, which were answered dichotomously (yes/no), are based on assessments developed and used in other adherence studies. 14 The final question, which asks patients to estimate how many days they have missed their ARV medication out of the past 30 days, has been validated against electronic drug monitoring as part of a three-item self-report measure. 15 Adherence levels (adherent or non-adherent) were characterised based on answers to these three questions (Table 1).

HIV-related stigma
The HIV stigma questionnaire consists of 12 items, with four groups of three items each comprising different subscales: personalised stigma, disclosure concerns, concerns with public attitudes and negative self-image. The 4-point Likert item set was previously adapted from the 40-item HIV Stigma Scale (HSS) and has been shown to have acceptable internal consistency (Cronbach's α > 0.7) and comparable psychometric properties to the full-length scale. Differential item functioning for this abridged measure is negligible in different cultural contexts and in different languages. 16 Lower subscale and sum scores reflect higher levels of stigma. Individuals with sum scores of ≤ 23 were classified as having high experienced stigma, whereas individuals with sum scores of ≥ 36 were classified as having low levels of stigma.

Accessibility factors
The accessibility questionnaire consists of six questions related to patients' access to the clinic and their medications, as well as their ability to understand clinic staff. The accessibility questionnaire was created by the research team based on a literature review of barriers to treatment adherence in the South African context.

Behaviour and knowledge
The behaviour component of the questionnaire consists of 10 questions from the 'Specific Concerns' and 'Specific Necessity' components of the Beliefs about Medicine Questionnaire (BMQ-Concerns and BMQ-Necessity, respectively). 17 This 5-point Likert item set has been shown to have high internal consistency in many contexts and high test-retest reliability.
The knowledge section was developed by the research team and consists of four true/false items testing patient knowledge about HIV and ARVs.

Data analysis Quantitative
Data were processed and analysed using IBM Corporation's Statistical Package for the Social Sciences (SPSS) 24 for Mac. Descriptive frequencies were calculated for demographic characteristics, answers to dichotomous adherence questions and responses to 4-point and 5-point Likert items in the stigma, accessibility and behaviour portions of the questionnaire. The distribution of the summed scores on the HIV stigma survey portion and the BMQ-Necessity and BMQ-Concerns sections was determined to be roughly normal. Pearson's coefficients and Spearman's coefficients were calculated, and significance levels were set at p < 0.05.

Qualitative
Qualitative narrative data were generated using transcripts from semi-structured interviews. Two researchers doublecoded eight transcripts independently and compared them side by side to identify inconsistencies. Codes were grouped thematically based on their viability in responding to the research question, and the relationships between codes were identified. Illustrative participant quotes are included below.

Ethical considerations
The study received ethics approval from the Dartmouth College (#00031086) and University of Cape Town Institutional Review Boards (HREC 464/2018) and provincial approval from the Western Cape Department of Health.

Quantitative findings
The 103 subjects in this study had a median age of 42 years and were mostly female; a plurality of participants identified as isiXhosa speakers. Most subjects had not completed any secondary schooling. Post-secondary education was rare. Income levels were low among the study participants, with the median income bracket under R5000/year. A substantial majority of participants had been initiated onto ART at least two years prior to the study date (Table 1).

Adherence levels
More than half of all participants in the quantitative phase (57.3%; n = 59/103) indicated some degree of non-adherence to their ARV protocols as determined by their answers to the three adherence questions listed in

Levels of stigma among study population
The average total score on the HSS was 30 (median, 30; range, 12-46; s.d., 7.3), reflecting intermediate levels of overall stigma. Almost a fifth of participants (17.5%; n = 18/103) reported numerically low stigma scores (≤ 24), indicating high levels of overall stigma. Similar scores were reported across the four subscales (disclosure concerns, negative self-image, concern about public attitudes, personalised stigma). The lowest subscale score (mean, 6.61; s.d., 2.49) was reported for the disclosure concerns category, indicating a higher concern about disclosure of HIV status among study participants as compared to the other types of stigma.
As indicated in Figure 1

Predictors of non-adherence
There was a significant inverse correlation between older subject age and the number of days an individual did not take their medication during the previous 30 days (r = -0.246, P < 0.05). There were no significant correlations between adherence status (adherent vs non-adherent) and other demographic factors, including gender, income level, time on ARVs and education level.
There was a significant relationship between an individual's negative self-image subscore on the HSS and the number of days that individual did not take their medication in the previous 30 days (Pearson's r = -0.338, P < 0.001), indicating that a more negative self-image predicted greater non-adherence. All other associations between stigma scale subscores and adherence were statistically insignificant (P > 0.05).
An individual's score on the BMQ-Necessity was significantly correlated with the number of days that individual skipped their medication during the previous month (r = 0.272, P < 0.001), reflecting that a lower belief in the necessity of one's HIV medications predicted a higher number of days being non-adherent to ART. There was a significant correlation between an individual's score on the BMQ-Concerns and the number of days that individual skipped their medication during the previous month (r = -0.266, P < 0.001), indicating that greater concern about one's medication (a lower BMQ-Concerns subscore) was correlated with the degree of non-adherence.
There was a significant difference in the mean knowledge scores between individuals who were classified as adherent (mean, 3.23) versus non-adherent (mean, 2.81) (P < 0.05), with non-adherent individuals reporting lower average knowledge about HIV and their ARV protocols.

Qualitative findings
Several themes emerged from careful review of the transcripts. We describe three major themes selected for their importance in elucidating the reasons behind the differential levels of adherence to HIV medications.

'They isolate you' -Fear, stigma and discrimination
In several of the interviews, patients describe a pervasive awareness of the potential for stigmatisation and discrimination because of their HIV status. Several patients opined that an individual's seropositivity was an important part of one's social identity in the eyes of community members, and they expanded upon the potential consequences of making one's positive HIV status public:  However, several patients cited alcohol as the most common social deterrent to adherence to HIV medication. The use of alcohol was particularly prevalent in male patients. Some patients mentioned that they feared the adverse effects of mixing alcohol with their medication, so they opted to either drink alcohol or take their medication but not both. Others noted that alcohol use caused them to forget their medications: 'Yeah, you sometimes, maybe, are drinking too much. And then you go and like just … fell asleep. Maybe wake up at night, which is you can't take both of your medications. 'Others, they don't care to take their medication. Sometimes they used to say to me, "If you take and take, you'll see.'' They come now better. They stop. If they feel like they need to start again, they run back to the hospital. … Others, they say they're only tired. Tired to take it every day.' (Female patient, #80, 41 years old)

Discussion
More than half of individuals surveyed reported some degree of non-adherence to their ART protocols, with over a quarter having missed four or more doses in the prior 30 days. This figure is significantly higher than several other recent studies investigating suboptimal ART adherence in the greater sub-Saharan African context. 18,19,20 Younger age was significantly correlated with a lower degree of adherence. The major correlates of low adherence were negative self-image, a lower belief in the necessity of one's ART regimen and a greater degree of concern about ART. Lower scores on the knowledge section of the questionnaire was also significantly correlated with a lower degree of adherence to ART. The qualitative data in our study suggested that internalised stigma, alcohol use and treatment fatigue also mediated patients' lack of optimal adherence to ARV medications.
Although the proportion of patients reporting high overall stigma levels was relatively low compared to other studies evaluating HIV-related stigma in similar contexts, 20,21 several http://www.sajhivmed.org.za Open Access patients indicated that the fear of discrimination related to one's seropositivity made it difficult to maintain proper adherence in social settings. In particular, all of the men interviewed in the qualitative arm of the study mentioned that taking their medications was especially difficult in social settings, where alcohol use and fear of stigmatisation are prevalent. Although both high degrees of stigma and alcohol use have been previously correlated with low ART adherence, 7,20,22,23,24 our study suggests that patients are more likely to fear stigmatisation in settings where alcohol is being used. This fear of stigma, combined with concerns about ART-alcohol interactions, resulted in several patients' decision to not take their ART in social settings, even though patients understood the importance of adherence to their overall health. Moreover, patients who had experienced social isolation or derision because of their status in the past were particularly reluctant to take ART outside their homes, fearing stigmatisation and discrimination by friends, colleagues and community members.
Several patients noted that treatment fatigue made it difficult to take their medications regularly, a finding in line with other recent studies. 25,26,27,28 In addition, despite an overall high belief in the necessity of their ART and their proficient knowledge about how ART works, many patients felt inclined to skip medications should they experience side effects or feel better. In patients with underlying treatment fatigue, beliefs in the necessity of ART and the inconsequential nature of sometimes skipping medications were often held simultaneously. Thus, an individual's knowledge of and belief in the positive effects of ART was not sufficient to prevent both fatigue and non-adherence to ART.
The findings of this study suggest ways in which clinicians may improve the identification of patients who may benefit from intensified counselling and/or longitudinal efforts to support ART adherence. An understanding of the major correlates of non-adherence may allow clinicians to better target interventions toward those especially vulnerable to non-optimal adherence levels and address concerns about taking ART. Considering the impact of one's social environment on adherence, the efficacy of interventions may be enhanced by efforts to help patients navigate such settings.
This study was conducted in a community clinic that serves a low-income patient population in South Africa, with a strong background diversity of study subjects allowing for robust analyses of the correlation between the social drivers of low ART adherence. Two primary limitations constrain the power of inferences one might make based on the findings of this study. Due to the use of convenience sampling and the relatively small sample size, the findings of this study are limited in their generalisability, especially to patient populations with significantly different demographic characteristics. Moreover, social desirability bias might have influenced the manner in which participants answered the surveys or interview questions. Therefore, as several other studies have found, the self-reported non-adherence levels may be an underestimate. 29,30 Conclusion This sequential, exploratory mixed-methods study provides research evidence that suggests that low ART adherence remains highly prevalent in an urban ART clinic in South Africa. We confirmed prior studies showing that negative selfimage, lower beliefs in the necessity of one's medications and higher degrees of concern about one's treatment regimen were major predictors of low ART adherence. We build upon these data by showing that despite overall high levels of patient belief in the necessity of their medications, lower knowledge with regard to HIV and the requirements of ART also predicted low ART adherence. In addition, our study shows that treatment fatigue and alcohol use pose major challenges for ART adherence, especially among young people who are aware of the necessity of their ART regimens. These data provide improved ways that ART clinics may identify patients at risk for low adherence and suggest interventions that may improve adherence, particularly in at-risk individuals such as younger people, those with low degrees of social support and those engaging in high levels of alcohol use.