This report describes the findings of a study that explored factors affecting adherence to ART in four public facilities in Botswana. The study used both quantitative and qualitative methods. The strength of this mixed-method approach is its usefulness in addressing the complexity associated with issues relating to adherence. The aim of the study was to help design interventions that can assist in maintaining high rates of adherence, a critical need if the danger of widespread resistance to treatment is to be avoided (Paterson et al., 2000; Orrell et al., 2003, Stevens et al., 2005). Paterson reported that ARV patients who achieved less than 80% adherence have a virological failure rate of 87%. As Botswana scales up access to ART, it is important to maintain high levels of adherence.

This section will first present the discussion on the quantitative results and then explore the integration of the quantitative and qualitative results. The quantitative data were collected using an adherence survey tool and exit questionnaires. The qualitative data were collected using exit interviews, semi-structured interviews, FGDs and observations. The respondents included ARV users, health workers, members of the local community and policy-makers.

A total of 514 respondents participated in the quantitative part of the study. Adherence in this study was defined as the number of times that patients actually take their drugs during a given period as a proportion of the number of times that they are recommended to do so. Optimal adherence was defined as an adherence rate of 95% or higher, since this is the level generally deemed necessary for treatment success and to avoid the development of resistance to treatment (Paterson et al., 2000). Levels of adherence below 95% were considered to be sub-optimal.

Three tools were used to measure adherence: pill counts over one month; one-month patient self-report with a visual analogue; and two-day recall using a 24-hour 'sun and moon chart'. The mean adherence rates were 95% for the pill count, 92% for the one-month self-report and 98% for the two-day recall. Using the pill count method it was found that 75% of the ART users achieved the optimal adherence levels of over 95%. Sixty per cent of participants reported optimal adherence using the visual line one-month recall and 96% reported optimal adherence using the two-day recall.

All three of these measures have both strengths and weaknesses. The pill count is in some ways the most 'objective', since it measures the number of pills left over from the previous refill, and can act as a very good proxy indicator for actual pill intake. It is, however, subject to manipulation by patients who may fear a bad response from health workers during dispensing if they have not achieved optimal adherence. In contrast, the one-month self-recall did not involve any health workers in this study. It was administered by researchers who were trained to act empathetically to the potential problems faced by respondents. As a result, ARV users might have been less inclined to overestimate medicine intake. However, a degree of desirability bias is always possible; as is the inability to recall medicine intake over such a relatively long period. The two-day recall has the advantage of a short time-span, which means that memory of medicine intake is likely to be good. However, it can overestimate medicine intake because patients may feel ashamed to admit to forgetting to take their medications so recently. While the two-day recall measure may be useful for on-the-spot individual patient adherence counselling, it may not be useful for long-term adherence monitoring. We recommend that both one-month recall (using visual analogue scale) and the pill counts be undertaken for routine adherence monitoring.

Recognition of the limitations of each of these three individual measures resulted in our calculating a composite adherence rate in the expectation that this would allow the strengths of one method to compensate for the limitations of the others. This composite adherence rate was calculated by taking the mean of the three measures and was found to be 77%. This is comparable to the adherence rates found in other countries (Bangsberg et al., 2000; Chesney, 2000; Liu et al., 2001; Nemes et al., 2004; Safren et al., 2005). Weiser et al. (2003) predicted that if treatment costs were removed as a barrier, adherence rates in Botswana would rise from 54% to 74%. We found that waiting times, costs and logistics are challenges to adherence.

The finding that 23% of patients had sub-optimal adherence rates is still unacceptably high. In view of the serious implications of non-adherence for public health, there is a critical need for targeted intervention strategies to increase the level of adherence. Monitoring and evaluation of the adherence strategies are an important component of any ART programme and can be useful in determining rates and identifying the factors that influence adherence. There is also a need for studies to investigate changes in patient adherence over time. As Horne (1998) argued, adherence should not be considered as a stable characteristic of any group of individuals but rather as a variable behaviour of an individual that can change over time.

The analysis of qualitative data was by thematic approach and 28 themes were initially identified. Further analysis identified the following key factors as facilitators or constraints of adherence (not in order of importance): the inter-related issues of disclosure: social support; acceptance of HIV status; treatment adherence partner; gender; remembering or forgetting to take the medicines; alcohol/substance abuse; transport costs and distance to health facility; side-effects; food; education; perceived effectiveness of treatment; self-efficacy; counselling; implementation of national guidelines; and a management system for addressing adherence and quality of care.

The study collected data at individual/household, community, health system and national levels. The following key issues relating to adherence operate primarily at the individual and household level, but they also incorporate aspects from the community and health facility.

The interrelated issues of disclosure, social support, acceptance of HIV status, treatment support partner and gender

The qualitative data showed that acceptance of one's HIV status and belief in the efficacy of treatment were important facilitators of adherence to treatment, while non-acceptance of HIV status was reported to be a major constraint. Respondents agreed that individual acceptance of HIV status translated into the kind of behavioural changes that are required for a good treatment outcome. For example, when ARV users are in a state of denial about their HIV status, ARVs are often hidden from friends and relatives and medication may be skipped to avoid discovery when other people are around.

The data also suggested that those who had accepted their HIV status were more likely to disclose to other people. Since acceptance is closely related to disclosure, most ARV users felt that non-disclosure of HIV status or of being on ART was a likely predictor of sub-optimal adherence to ART. Without disclosure it becomes difficult to use medication in the presence of other people. In view of the complexity of the medication schedule (in terms of time and consistency), it is difficult to have the privacy that would keep people from noticing regular medicine use. However, even though a significant number of respondents believed that disclosing to relatives and friends had the potential to improve adherence, some of them also perceived that disclosure carried a number of risks such as: emotional injury, loss of intimate relationships and job loss. Other studies have also found that disclosure is a complex process with varying consequences, such as greater intimacy or rejection, feeling of relief or remorse and enhanced status or 'spoiled image' (Ormazu, 2000; Klitzman et al., 2004; Zea et al., 2005). It was generally believed that without disclosure, the level of social support would not be adequate.

Adherence to the treatment regimen was also related to the availability of information material as well as emotional support from family members. The positive role of children in reminding parents when it was time to take their pills was found to be a common facilitating and supportive factor. In addition, the level of family social welfare (income status, collective efficacy, successful resolution of previous conflict, trust) was found to determine the kind of social support an ARV user receives. Men complained that where people had not invested in family social welfare, there was more negative energy/emotion in the family characterized by neglect, sarcasm, negative affective states and constant reminders of the recipient's dependency. Where a husband had previously deserted the family home and where lack of forgiveness prevailed, social support was often lacking. Investment in family social welfare was found to be essential for future social support.

Gender was perceived to have an influence on acceptance of status. Many respondents agreed that women were more likely to accept their status and to seek health care. The desire for women to stay alive in order to care for their children and other loved ones, despite being blamed by men for bringing the disease, was said to be a major facilitator of adherence. The better health-seeking behaviour of women was attributed to the fact that women have always been close to the health care system through the maternal and child health clinics, a system which pre-dates the HIV era. Prior involvement in prevention of mother-to-child transmission (pMTCT) programmes before starting on ART also meant that women were better sensitized and specifically targeted with HIV interventions. While women were thought to accept their HIV status more readily than men and were therefore more likely to seek help from the treatment centres, men were perceived to be difficult, refusing to accept their status or to disclose. This may partly explain why more women than men are on treatment. Since acceptance of HIV status, disclosure and gender were found to be the main emerging themes in the qualitative data, further studies are needed to explore these variables in greater depth.

Positive attitude and belief in the efficacy of ART

The FGDs revealed that, despite being often preoccupied with their own health concerns and fears, a majority of participants reported that the availability of ARVs had given them a "new lease of life." ARV users tended to have adequate information about the natural progression of HIV. The study also found that one of the strongest qualitative predictors of medication adherence was a personal belief in the efficacy of ART. This belief was also buttressed by the fact that ARVs are potent and as such, better than traditional herbs. There was an overriding belief among ARV users that traditional herbs do not increase CD4 count in the body or decrease the viral load, while ARVs do. This is a critical finding, in a country where patients tend to use both traditional and modern medicine. Those who held this belief also said that HIV was increasingly a manageable disease. In deciding whether or not to start taking ARVs and to continue taking the medication, a patient has to weigh up the costs and benefits of treatment.

Other studies have confirmed an association between adherence and a belief in the efficacy of the pills used in treatment (Eldred, 1997; Ferris, 1996; Smith, 1997). Several studies have shown that patients with a higher level of knowledge about the effectiveness of ART and belief that poor adherence could promote viral resistance and treatment failure have a greater ability to adhere to their medication.

Remembering/forgetting, alcohol and substance abuse

Forgetfulness was the most common reason cited by those who had problems with adhering to their medication, a finding that is consistent with other studies (Brigido et al., 1998; Chesney et al., 2000; Golin et al., 2002; Turner, 2002). However, the specific reasons for forgetfulness could not be quantified. Respondents cited work and home duties, travelling for work (e.g. cattle posts) or social events (e.g. funerals and weddings) as factors that led to forgetfulness. The qualitative data highlighted alcohol abuse as a major barrier to adherence to ART. ARV users, health workers and members of the community all identified a direct link between alcohol abuse and sub-optimal or non-adherence. The participants acknowledged the use of mechanical devices (radios, cell phones, and watches), children, parents and partners as helpful ways of being reminded to take their medicines.

Distance to health facility, employment status and costs

The Government of Botswana is responsible for the overall management of the national ART programme and for making treatment available free of charge. However, the study data suggest that patients are burdened by the cost of transportation to and from facilities and by the cost of food while waiting to be attended to. Forty-four per cent of the patients in our quantitative survey reported an increase in expenditures as a result of being on ART, 37% reported a loss of income, and of those who had missed an appointment, 12% said it was for lack of funds to pay for transport to reach the health facility. One in four ARV users interviewed said they had to travel at least 50 km to visit the treatment centre and the maximum distance travelled was 200 km. These findings suggest that many patients have significant treatment-related financial problems, even though the Government is providing ARVs free of charge.

Quantitative data suggested that there was an association between employment and optimal adherence in that respondents who were employed were more likely to adhere than those who were unemployed. It makes conceptual sense that treatment-related costs (e.g. transport) could be a reason why the unemployed failed to achieve optimal adherence. However, the cost of transport can also be a serious problem for workers on a daily rate, the self-employed and casual workers, who not only have to pay for their trip to and from the clinic, but who are also likely to have to forego their daily wage in order to keep appointments at the health facility. Based on this finding, the provision of travel vouchers or reimbursement of travel expenses would be potent facilitators of adherence.

Side-effects

Quantitative data revealed that of the 58% of ARV users who reported having experienced side-effects, only 8% cited side-effects as a reason for failure to achieve optimal adherence. This is contrary to other studies carried out in developed countries where the most frequently cited reason for stopping medication was side-effects. (Chesney et al., 2000; Ammassari et al., 2001). There are several possible reasons for these differences in results. First, in the case of Botswana, the entry level for ART is CD4 count < 200. At this level, patients may be so sick that the side-effects experienced could be perceived as a symptom of the disease itself. Second, a sociocultural belief exists among Batswana that it is easy and painless for a disease to get inside the body, but difficult and painful for the disease to get out. According to this belief, the experience of side-effects may be interpreted as the disease exiting from the body and therefore may be tolerated. Third, pre-treatment counselling, where patients are educated about the ARV treatment plan, including possible side-effects, may result in the perception of side-effects as a reasonable risk to be tolerated in view of the magnitude of benefits expected and the alternatives available. These reasons could explain the under-reporting of side-effects/adverse events as stated by one of the health workers in the qualitative interview. This could also suggest that, despite lack of continuity in counselling, this service seems to have an impact on adherence.

The treatment-related increase in food demand did not appear to have a significant impact on adherence. Although 2% of respondents in the quantitative study cited lack of food as a reason for sub-optimal adherence to ART, subsequent interviews revealed that, while lack of food was an inconvenience, it did not stop them taking their ARVs. Education level is usually used as a measure of socioeconomic status but most studies so far did not find education as a predictor of adherence when controlled for other socioeconomic factors (Carrieri et al., 2002; Nemes et al., 2003; Murphy et al., 2004). Our study also confirms this finding and may also be an indicator that information, education and communication is working well.

Health facility level

Health care workers performed different activities such as: pre-treatment counselling; ongoing counselling; referral to social workers of patients with adherence problems; pill counts; and collation of adherence data and submission of monthly statistics to the national programme. The health workers all maintained that there is no strategic management system for addressing ART adherence problems.

All four of the facilities surveyed use the Botswana Guidelines on Antiretroviral Treatment. The current version (2005) recognizes sub-optimal adherence as a common reason for treatment failure and the most important cause of the emergence of drug-resistant strains of HIV. The guidelines recommend seven strategies to improve adherence, including: establish trust with the patient and family; serve as an educator and source of information; provide ongoing support and monitoring; intensify management in periods of low adherence by more frequent visits, recruitment of friends and family and deployment of other team members; utilize a health team approach; provide training to support the ART staff; use the family care model approach and provide care to the family as a unit.

While these recommendations are excellent, there is no clarity as to how the health care worker should achieve the objectives. The guidelines acknowledge the need to involve family and friends in coping with the demands of adherence, but fail to address the ethical issues involved in areas such as privacy, autonomy and confidentiality, which are genuine concerns for patients and health workers.

Study limitations

It would have been interesting to examine the adherence rates in relation to the timing of drug and food intake. However, due to the fact that most of our study participants were on first-line medication that did not require food restrictions and the fact that a previous question included the times medication were taken, these measures could not be used. This study had several limitations. For example, some of the data collection methods used in this study relied on self-reports of adherence behaviour, which is prone to response biases. In addition, the study used a client sample that was currently on ART and did not include individuals who had discontinued treatment. It is also possible that people who achieved sub-optimal adherence may no longer be visiting the treatment facilities.