HIV-infected adolescents have low adherence to antiretroviral therapy: a cross-sectional study in Addis Ababa, Ethiopia

Introduction For antiretroviral therapy (ART) to work effectively, adherence is very crucial. However, most studies done on ART adherence are either on children or on adults. There is limited information on the level of adherence among adolescents. Methods Using a cross-sectional study design, we interviewed 273 HIV-infected adolescents receiving ART from three hospitals in Addis Ababa. We used a structured questionnaire to measure adherence levels using patient self-reports. Bivariate and multivariate methods were used for analysis. Results We interviewed 273 adolescents aged 13 to 19 years, and 144 (52.7%) of the participants were girls. Their mean age was 15.4 years (SD± 1.75). The self-reported adherence rate of the respondents was 79.1% (216/273). On bivariate analysis, variables like WHO clinical stage, being on Cotrimoxazole Prophylactic Therapy (CPT), marital and living status of the parent, whether parent was on ART or not and having special instructions for ART medications were associated with optimum adherence. However of those, only WHO stage IV (adjusted OR, 12.874 95% CI, 2.079-79.706), being on CPT (adjusted OR, 0.339 95% CI, 0.124-0.97) and adolescents with widowed parent (adjusted OR, 0.087 with 95% CI, 0.021-0.359) were found to be significantly associated with optimum ART adherence. Conclusion The level of self-reported ART adherence among HIV-infected adolescents at the three hospitals was below the recommended threshold. Though earlier presentation of adolescents to care should be encouraged, more targeted adherence support should be planned for those who present at an early stage of their illness.


Introduction
Adolescence is the period between 10 and 19 years of age with about one-fifth of the world's population belonging to this age group [1]. In the sub-Saharan Africa (SSA) including Ethiopia, one-third of the total population is aged between 10 to 24 years [2].
Adolescence is a period of dynamic change representing the transition from childhood to adulthood. During this stage, rapid changes in physical, emotional, cognitive and social characteristics take place [1]. In many resource-poor settings, adolescents are now the emerging group of HIV-infected population as survival of children with perinatally acquired HIV infection into adolescence is increasingly being documented [3,4]. In addition to the perinatally acquired infection, adolescents and young adults are vulnerable to acquiring HIV through sexual route [4]. Eighty-two percent of the estimated 2.1 million adolescents aged 10-19 years living with HIV by the end of 2012 were in sub-Saharan Africa [5]. Though, less than one percent of Ethiopian youth tested positive for HIV on the 2011's Ethiopian Demographic and Health Survey (EDHS), according to the regional estimates of HIV prevalence among youth, Addis Ababa (capital of Ethiopia) has the second highest prevalence next to Gambella region [6]. Similarly, HIV is significantly prevalent among youth in Addis Ababa, particularly among out-of school and female youth [7]. This suggests that there is a need to pay attention to factors affecting access to and outcomes of treatment among adolescents and young adults. The most important factor which determines the success of ART is sustainable and optimum adherence to therapy [8] as poor adherence is associated with treatment failure and the development of viral resistance [9].
Adherence to medication has been described as the proportion of prescribed medications that is actually taken. It is measured on a scale from zero to 100%. The recommended optimal adherence level for ART to be effective is above 95 percent [10]. However, there is limited information on the levels of adolescent ART adherence in resource-poor settings [11]. Studies done on ART adherence in Ethiopia are either on children or on adults and those showed high adherence rates [12,13]. Studies on the level and predictors of ART adherence among adolescents in Ethiopia are lacking. Therefore, in this study, we aimed to contribute to filling this information gap. software package by considering the following assumptions:

Methods
Proportion of non-adherence among unexposed, 23.1%; 80% power; 95% confidence interval; Odds ratio of 2 and 10% Nonresponse rate. Since this study focused on a unique population of limited size, the finite population correction is applied and the final sample size was n= 303. This was allocated proportionally to the three facilities after determining the population size in each facility.That is total number of adolescents on HAART at each institution. Using the ART registers and site electronic database, we prepared a list of all adolescents aged between 13-19 years and who were on ART as a sampling frame. We then selected study participants using systematic random sampling technique.

Study variables:
The dependent variable in this study was 'adherence to ART' and independent variables included sociodemographic, behavioural and clinical factors. A participant was said to have optimal ART adherence if he or she took ≥ 95% of the prescribed pills correctly for the four days prior to the study. . We recruited nurses working in the ART clinics of the selected facilities as interviewers. We then provided them with two days training on the study and on the standard operating procedure. We pre-tested the tool on non-sampled adolescents and modified it accordingly. Nurses collected data by face to face interview with the adolescents in a private room. For adolescents whose HIV status was not disclosed, the nurses took data regarding ART medications adherence from parents or caregivers. We used patient medical charts to retrieve clinical information.

Data entry and analysis:
We used Epi Info version 3.5.4 for data entry and SPSS version 21 for analysis. First, we did descriptive statistics to explore the socio-demographic characteristics of the respondents, the adherence rate and clinical characteristics of the adolescents. Then, we explored for the association between the various independent variables with optimum ART adherence on bivariate analysis. We then identified those variables which were associated with adherence at a level of P-value of less than 0.05.
We included those variables in multivariable analysis.

Ethical considerations: The School of Public Health at Addis
Ababa University and the Institutional Review Boards of the hospitals reviewed and approved the research protocol. We sought permission to conduct the study from the Medical Directors of the three hospitals. Adolescents and caregivers received information about the study. We obtained written informed consent from adolescents aged 18 and above. For adolescents under 18 years, we obtained consent from parents or legal guardians in addition to verbal assent by the young. To ensure confidentiality of all study participants, we used no direct identifiers in the data collection, storage or report writing. All electronic documents were password protected and all paper documents were stored in a locked cabinet.
We prevented accidental disclosure of HIV status to those nondisclosed adolescents. This was dealt with by training the interviewer on the study protocol and by collecting data regarding adherence from parent or caregiver.

Socio-demographic Characteristics of Participants:
We interviewed 273 adolescents aged 13 to 19 years with a response rate of 91%, and 52.7% of them were girls. The mean age of the participants at study enrolment was 15.4 years (SD ± 1.7). Nearly all (97.8%) were living in urban settings. One-third of the adolescents had both parents dead and 39 (14.3%) were cared for by their grandparents. The majority 165 (60.5%) of the adolescents were cared for by either of their parents. Only 45 (16.5%) of the caregiver's had external monetary support for the adolescent. Table   1 provides the socio-demographic characteristics of the study participants.

Behavioural characteristics of adolescents on ART:
Out of the 273 adolescents interviewed, 9(3.3%) have tried to smoke a cigarette. Among those, the earliest age when a whole cigarette was smoked was 14years, 3 out of the 9 did at this age. None of the nine did smoke cigarette daily. Similarly  Factors associated with optimal ART adherence: Table   4 Describes factors associated with optimal ART adherence on bivariate and multivariate analyses. On bivariate analysis, we examined whether various explanatory vvariables are associated with optimal adherence, considering a P-value of less than 0.05 as a cut-off for statistical significance. We found six variables namely; baseline WHO clinical stage, being on CPT, caregiver's marital and living status, whether parent was on ART and special instruction with ART medications to have a significant association with optimal ART adherence. We then conducted a multivariate analysis controlling for those six covariates to adjust for possible confounder

Discussion
In this study, we assessed the magnitude of ART adherence among HIV-infected adolescents in Addis Ababa and we also looked into the factors associated with optimal ART adherence. We found the ART adherence level to be lower than the recommended level and less advanced disease stage, taking CPT along with ART and those cared for by widowed parents were significantly associated with poorer adherence levels. The level of non-adherence (20.9%) is quite high and puts the adolescents at higher risk of drug resistance and treatment failure. Also, of great concern is that significant proportion of the respondents reported missing full day's medication of one or more days in the four days prior to the study.
Furthermore, six of the adolescents had zero percent adherences; did not take any ART medication in the four days. This suggests the need for more targeted adherence interventions for adolescents in early disease stage, those on concomitant medications and adolescents living with widowed parents. The low self-reported adherence in our study is in line with other study findings. A study was done among American adolescents aged 12-18 years revealed low ART adherence level [15]. This study, unlike ours, longitudinally followed a cohort of 231 HIV + adolescents. Though it used a similar self-report method to assess adherence it was further validated by various additional ways. Another study done in Botswana came up with adolescent ART adherence rate of 76.9%.
Despite a small sample size, this study was done in a similar group of patients, 13-20 years and used patient self-reports method [16].
Similarly, other studies as well showed low ART adherence among adolescents [17][18][19]. In contrast to these a study done in Gaborone, Botswana reported that high proportion of the studied adolescents had excellent ART adherence. This study considered excellent adherence using pill count method when greater than or equal to 95% of the prescribed doses for one month were taken by the end of the month. However, in this study the investigators used a smaller sample size and the study was done in a single health Page number not for citation purposes 5 facility [20]. A systematic review of 50 articles involving 10725 adolescents on ART reported overall adherence level to be 62.3%.
In this study, ART adherence level for Africa was found to be 84% which is slightly higher than our finding of 79.1% [21]. Most of the studies on ART adherence in Ethiopia are either on children or on adults. A study in Addis Ababa assessed the ART adherence among children and found a higher adherence level, 86.9% [12]. Similarly, a prospective study of adult HIV patients in Ethiopia found a higher ART adherence rate, 94.3% [13]. Contrary to these findings a study which measured adherence using unannounced home-based pill count revealed a very low adherence level, 34.8%. This study was done among HIV-infected children below 15 years who were attending paediatric ART clinic of Tikur Anbessa Hospital [22].
The poorer adherence level among adolescents in earlier disease stage is consistent with finding from other studies. A study in Ethiopia revealed that children and adolescents in WHO stage III/IV were more likely to adhere [22]. This could be because those who are relatively healthy will be reluctant about taking their medications. Gibbs study revealed that symptomatic HIV disease was associated with better adherence [23]. Catz et al also found that healthy HIV-infected out patients had lower rates of adherence to medical appointments than the symptomatic ones [24]. A study in Uganda also found that those who had been hospitalized two or more times had better adherence [25]. On the contrary, the previously mentioned American cohort study showed that those with late HIV disease stage were less likely to be adherent compared with those in the early stage of the disease [15]. The poorer adherence among adolescents living with widowed parents could be because married parents tend to be emotionally and economically better and might also get support from their partner in giving care and support to the adolescent. Moreover, this may also be because the adolescent is living with a single parent who may be dealing with his or her own HIV status or could be too sick to take care of the adolescent. On the other hand, poorer adherence among those who were on CPT could be due to high bill burden but a false sense of security with CPT or even misunderstanding CPT as a replacement for ART could be another factor. However a study in Addis Ababa described that those who took CPT besides ART were more than three times likely to adhere than those who didn't (OR = 3.65 with 95% CI, 1.24-10.74) [12]. Accurate measurement of adherence to therapy is oftentimes difficult. There are different ways of adherence assessment including; patient self-reports, pill count method, biochemical assays of drug levels and electronic monitoring system. All of these techniques have their own limitations. A major limitation of self-reports is that they assess only short-term adherence and may often overestimate it. Moreover, this method assumes that patients can correctly recall their behaviour and are providing honest answers. Pill count method measures adherence by counting the returned excess pills which should have been taken. Here patients are expected to return the excess pills on their refill visit date. Similar to the former this method tends to overestimate adherence as patients tend to discard the package inadvertently. In addition, some patients may also discard packages purposively to appear adherent. The latter two techniques, assays of drug levels and electronic monitoring system, tend to be sophisticated and costly [26]. Our study has some limitations. First would be the use of ART nurses to collect data on medication adherence. This might introduce social desirability bias and lead to under-reporting of non-adherence by adolescents. However, we used ART nurses because we wanted to keep sensitive HIV-related adolescent information confidential. Secondly, for a small group of adolescents whose HIV status was not disclosed, 27 (9.9%), nurses took data regarding ART medications adherence from parents or caregivers. There is a chance that caregivers may not accurately recall adolescents' adherence information. However, we used them with the intention of preventing accidental disclosure of HIV status to those non-disclosed adolescents. Thirdly, with regard to selecting study facilities, we purposefully chose three public hospitals in Addis providing ART and HIV care services for a large number of HIVpositive adolescents in the city. These hospitals, however, are not the only sites providing ART for adolescents in Addis Ababa.

Conclusion
The findings of this study indicated that the ART adherence rate among adolescents in Addis Ababa is low. Advanced WHO stage and having a married parent were associated with better ART adherence. On the other hand being on CPT was associated negatively with ART adherence. Health care providers should strengthen adolescent ART adherence counselling services in the

Competing interests
The authors declare no competing interests.

Authors' contributions
Naod Firdu prepared the study proposal, collected and analysed the data, interpreted the findings and wrote the manuscript. Fikre