Abstract
We sought to determine the extent to which the Theory of Planned Behaviour (TPB) was applicable in predicting medication adherence among South Africans receiving antiretroviral therapy (ART). Regression analyses revealed that the linear combination of attitudes towards adherence, perceived behavioural control and perceived group norms explained 12 percent of the variance in intentions to adhere to ART. We also found a non-significant relationship between intentions to adhere to treatment and self-reported adherence. The results call into question the extent to which TPB is helpful in understanding a health-promoting behaviour such as medication adherence among South Africans receiving ART.
Adherence to antiretroviral therapy is necessary to ensure positive health outcomes for persons living with HIV and AIDS, specifically in terms of minimizing the progression of the disease and thus reducing mortality (Boulle et al., 2008). Considerable evidence has shown that sub-optimal adherence, for example, failing to attend clinic appointments and take pills at the required times can lead to a low CD4 count, a high viral load, poor quality of life and ultimately death (Mannheimer et al., 2005). Poor adherence among patients enrolled in an antiretroviral treatment programme may also increase patients’ risk of developing resistant strains of HIV, thus requiring the use of more expensive second-line drugs than first-line. As funding for ART programs is limited, ensuring optimal adherence is essential, not only to reduce patient mortality and improve quality of life, but also to limit the wastage of drugs, clinicians’ time and clinic resources.
Several barriers to ART adherence have been identified, namely structural barriers associated with the environment in which patients live (Coetzee et al., 2010; Kagee and Delport, 2010; Kagee et al., 2010); mental health problems (Nel and Kagee, in press; Simoni et al., 2011); low levels of perceived social support (Holstad et al., 2006); substance abuse (Tucker et al., 2004); and concerns about stigma (Klitzman et al., 2004). Key factors associated with adherence that were identified in a meta-analysis included fear of disclosure, substance abuse, forgetfulness, suspicions of treatment, complexity of treatment, decreased quality of life and competing demands on time (Mills et al., 2006). Further, in a study testing the relationship between autonomous regulation and ART adherence, this relationship was confirmed but mediated by self-efficacy (Lynam et al., 2009). Together, these findings bring into focus the need for autonomous motivation in ART users to be encouraged by health care personnel.
Adherence has been defined by the World Health Organisation (2003) as ‘the extent to which a person’s behaviour – taking medication, following a diet, and/or executing lifestyle changes corresponds with agreed recommendations from a health care provider’ (p. 3). A high rate of adherence, which is concerned with the correct dosage taken within the correct time period and in the correct manner, is necessary for treatment to be successful. Dose adherence refers to the number and proportion of doses taken, schedule adherence refers to adherence to doses taken on time and dietary adherence refers to doses taken correctly with food (Schonnesson et al., 2006). Under ideal circumstances all these dimensions of adherence are required. Non-adherence may thus take various forms, for example, not taking the medication at all, taking the medication at the wrong time, taking the wrong dose due to misunderstanding treatment directions, prematurely terminating the medication, or adjusting the regimen to reduce side effects and toxicity without consulting the health provider (Miller, 1997). Preparatory behaviours that impede proper pill-taking adherence include not filling prescriptions, self-adjusting the regimen to modulate side effects and toxicities and incorrectly understanding the doctor’s instructions (Chesney, 2003).
Theories of health behaviour seek to explain why individuals engage in health-related behaviours, which in the present study is ART adherence. The Theory of Planned Behaviour (TPB) postulates that volitional behaviour is associated with the intentions to engage in a specific behaviour (Rhodes and Courneya, 2004), in this case correct ART adherence. Intentions in turn are influenced by attitudes, perceived behavioural control and perceived group norms about the behaviour (Ajzen 1991; 2002). The TPB has been found to be a useful predictor of health behaviour, including HIV/AIDS health behaviour (Janepanish et al., 2011; White et al., 2010). Its utility in previous research has suggested that it may be similarly applicable, at least in part, in the context of understanding ART adherence. To our knowledge, only two previous studies have used the TPB in predicting adherence among South African samples. Fincham and colleagues (2008) used the TPB to predict dietary and fluid adherence among haemodialysis patients and showed that while attitudes and perceived behavioural control (PBC) explained 15.5 percent of variance in self-reported adherence, the full model was not optimal in explaining variance in self-reported adherence. Kagee and van der Merwe (2006) used the TPB to predict treatment adherence among patients seeking care for diabetes and hypertension at public health clinics in South Africa. These authors showed that the TPB explained 47 percent of the variance intentions to adhere to treatment and that perceived behavioural control was the strongest predictor of intentions to adhere to treatment.
Yet, social cognitive theoretical models such as the TPB have received criticism for being overly focused on individual decision-making and therefore of limited use in non-western cultural contexts (Campbell, 2003). To this extent, it was considered useful to determine the extent to which the TPB was applicable in the South African cultural and political context.
There is a considerable body of research from many parts of the world to suggest that perceived stigma is associated with poor ART adherence (Dlamini et al., 2009; Klitzman et al., 2004). Also, it has been noted that patients fearing that others would discover their positive status would often incur considerable expense to travel to clinics far from their homes to avoid being seen seeking care for their condition and would have to hide their medication to avoid discovery by family members (Coetzee and Kagee, 2010; Hardon et al., 2007). While it has been suggested that the availability of ART may play a role in changing in people’s perceptions of HIV/AIDS from that of a death sentence to a manageable chronic illness (Kagee, 2008), it is simultaneously assumed that perceived stigma is an important barrier to adherence.
In the context of many millions of persons living with HIV gaining increasing access to ART, understanding adherence among this population is an important public health and psychological concern. The aim of the present study was to determine the extent to which the TPB, together with perceived stigma, could explain adherence behavior among patients receiving antiretroviral care.
Method
Participants
One-hundred-and-one patients receiving antiretroviral treatment were recruited by means of convenience sampling at a peri-urban public hospital in South Africa. The clinics provided treatment for HIV including anti-retroviral treatment (ARVs) to patients. The following inclusion criteria were applied: (a) only participants aged 18 years of age and older were included in the study; (b) participants were able to understand either spoken English or Afrikaans; (c) participants were physically and psychologically capable of engaging in an interview and completing a battery of self-report instruments. Patients who indicated they had been diagnosed with bipolar disorder, schizophrenia or related psychotic disorders were excluded from the study.
Procedure
Patients seeking ART services were informed by clinic nurses about the research study. They were told that if they were interested in learning more about the study, they could approach the researcher in a private room where the study would be explained to them. Patients who agreed to participate in the study were asked to sign an informed consent form, following which they were asked to complete a questionnaire battery administered in English, Afrikaans or Xhosa.
Upon completion of the questionnaire, each participant received a supermarket voucher as a token of thanks for their participation. On the recommendation of the clinic director we also offered snacks to all patients, regardless of whether they participated in the study and a supermarket voucher to the clinic nurses for handing out flyers about the study to patients. Ethical clearance was obtained from the Stellenbosch University Health Research Ethics Committee and permission to conduct the research was obtained from the Western Cape Department of Health, the hospital superintendent and the clinic director.
Instruments
Attitudes toward treatment adherence
Two subscales of the Adherence Attitude Inventory were used to test attitudes toward treatment adherence. The Adherence Attitude Inventory is a 28-item, Likert-type instrument (Lewis and Abell, 2002), that assesses cognitive functioning, patients/health worker communication, self-efficacy and commitment to treatment adherence. The reliability of the instrument as indicated by Cronbach alpha was 0.75 (Lewis and Abell, 2002). The subscales Commitment to Adherence and Patient-Provider Communication subscales were used in the study.
Perceived subjective norms
An eight-item questionnaire was developed to calculate perceived subjective norms (with four Likert scale responses ranging from ‘strongly disagree’ to ‘strongly agree’). A similar questionnaire was used in a study conducted by Kagee et al. (2008) and the Cronbach alpha was 0.61. Their questionnaire was modified to apply ART users.
Perceived behavioural control (PBC)
PBC was measured using an eight-item questionnaire assessing self-efficacy to engage in adherence-related activities. A similar questionnaire used in a study by Kagee and van der Merwe (2006) was adapted for ART users.
Intentions to adhere to treatment
An eight-item, three-response-option Likert-type scale was constructed to measure intentions to engage in various adherence-related activities. A similar questionnaire used by a study conducted by Kagee and van der Merwe (2006) had a Cronbach alpha of 0.72.
Perceived HIV stigma
The HIV stigma scale was used in the present study to assess perceived stigma (Berger et al., 2001). This 40-item scale includes four subscales: Personalized stigma, disclosure concern, negative self-image and concern with public attitudes. The overall Cronbach alpha for the HIV Stigma Scale is 0.96 and the alphas for the subscales ranged from 0.90 to 0.93 (Berger et al., 2001).
Self-reported adherence
The self-reported adherence measure was used to assess adherence (Morisky et al., 1986). The scale questions are: Do you ever forget to take your medicine? Are you careless at times about taking your medicine? When you feel better do you sometimes stop taking your medicine? Sometimes if you feel worse when you take the medicine, do you stop taking it? In the original study by Morisky et al. the alpha reliability of the scale was shown to be 0.61.
Data analysis
All statistical procedures were performed using the Statistical Package for the Social Sciences (SPSS). Frequencies, percentages, means, standard deviations and ranges were calculated for the various independent variables. Base 10 and natural base logarithmic transformations were performed in order to normalise skewed distributions. Multicollinearity between predictor variables was assessed using the Variance Inflation Factor (VIF) statistic. Hierarchical regression analysis was used to test the linear combination of the TPB variables on intentions to adhere to antiretroviral medication.
Results
Description of the sample
A total of 121 patients were invited to enrol in the study of which 101 agreed to participate. The mean age was 35 years (SD = 7.05 years). Most of the sample, (82.2%) were female; 66 percent were Black, and 42.6 percent stated they were unemployed. Of the total sample, 45 percent stated they were single, 11 percent stated they were widowed, six percent were separated, nine percent were divorced, and 29 percent were married or living together. In terms of annual family income, 42 percent reported less than ZAR10,000; 17 percent reported between ZAR10,000 and ZAR40,000: 6 percent reported more than ZAR40,000; and 35 percent stated that they did not know their annual family income.
Data screening
Tests of collinearity revealed that this was not a concern as correlations among predictors were less than 0.80 and no Variance Inflation Factor was greater than 10. The sample sizes, means, standard deviations, ranges of the key variables are presented in Table 1. The internal consistencies of various instruments as measured by Cronbach’s alpha and the inter-correlations between the variables are reported in Tables 2 and 3, respectively. The measure of self-reported adherence originally had modest internal consistency (α = 0.56), which improved to 0.59 when two were removed from the analysis, namely, forgot to take medication during the last two weeks and forgot to take medication over the weekend.
Descriptive statistics of key variables.
Cronbach’s alpha of the measures.
Correlation Matrix of TPB variables and self-reported adherence.
Correlation is significant at the 0.05 level (two-tailed)
Correlation is significant at the 0.01 level (two-tailed)
Predicting intentions to adhere to ART
Table 4 presents the summary of the hierarchical regression analysis for the variables that predicted intentions to adhere. Table 5 presents the regression statistics for the two models that were tested, namely, the TPB variables in predicting intentions at Step 1, and the TPB variables as well as perceived stigma at Step 2. The linear combination of the TPB variables in Step 1 significantly explained 12 percent (R2 = 0.12; F (3, 97) = 4.52, p = 0.01) of the variance in intentions to adhere to treatment. Although the addition of perceived stigma explained an additional three percent of the variance in intentions, this addition was non-significant: F (1, 96) = 3.09, p = 0.08. We also tested the relationship between intentions to adhere and self-reported adherence but this correlation coefficient was also non-significant. A further test of the linear combination of the predictor variables in explaining self-reported adherence indicated a non-significant result.
Summary of Hierarchical multiple regression analysis for variables predicting intentions to adhere to treatment (Regression model 1 and 2).
A. Predictors: (Constant), Attitudes towards treatment adherence, Subjective norms, Perceived behavioural control
B. Predictors: (Constant), Attitudes towards treatment adherence, Perceived subjective norms, Perceived behavioural control, Perceived stigma
C. Dependent variable: Intentions to adhere to treatment
Parameters for variables predicting intentions to adhere to treatment (N = 101).
Note: CI = confidence interval; LL = lower limit, UL = upper limit.
Dependent variable: Intentions to adhere to treatment
Attitudes towards treatment adherence
Perceived subjective norms
Perceived behavioural control
Perceived stigma
Discussion
The TPB was able to explain 12 percent of the variance in intentions to adhere to antiretroviral treatment, a medium effect size for multivariate models in the social sciences (Cohen, 1988). The correlation between intentions and reported adherence was non-significant, however. These results suggests that the linear combination of attitudes towards antiretroviral treatment, perceived behavioral control and perceived subjective norms was able to account for a significant, albeit modest, proportion of the variance in behavioral intentions.
The TPB can be used to understand intentions to engage in adherence to ART. These data add to the growing number of studies showing the utility of TBP variables in predicting adherence to various prescribed health behaviours, for example, dietary and fluid intake (Fincham et al., 2008); medication for diabetes and hypertension (Kagee and van der Merwe, 2006); exercise-based therapy in rehabilitation (Yardley and Donovan-Hall, 2007); and exercise – rehabilitation therapy among cardiac patients (Blanchard et al., 2002). The study results indicate that when considering intervention options to enhance adherence, TPB variables such as attitudes, perceived behavioural control and group norms may need to be taken into account.
We were surprised by the non-significant relationship between perceived stigma and intentions. It is possible that this finding may be peculiar to the sample who had received psychosocial services, which included how to deal with stigma, from a non-governmental organization allied with the treatment clinic. It is also possible that a rational construct such as intentions was non-orthogonally related with perceived stigma, thus accounting for this result. In other words, the TPB and related constructs may be inappropriately applied in the context where the study was conducted. Such a view is echoed in Marks’ (2008) criticism of the TPB, observing that theories such as the TPB may not reflect the complexity and interplay among social, cultural, economic and political factors that influence health behavior. Indeed, social cognitive theoretical models such as the TPB and the Health Belief Model have been criticized for being individualistic in focus and of limited applicability in countries that make up the global South (Campbell, 2003). It has been argued that individual and intra-psychic factors insufficiently consider the socio-cultural complexity needed to foster a health-enabling community within which persons living with HIV can engage in health behaviours such as adherence. The unexplained variance in adherence intentions may thus in part be due to the context within which ART-users live. Structural conditions specific to resource-constrained environments such as poor access to affordable transport, long waiting times at clinics, poor interpersonal interaction with health care personnel and stigma (Coetzee et al., 2011; Kagee and Delport, 2010) may exist as additional extra-individual barriers to adherence intentions that lie beyond the TPB’s ability to explain. To this extent, contextual factors, such as the nature of the public health care system in South Africa and other resource-constrained countries, may also play an important role in determining adherence intentions. It is at this nexus, between cognitive processes and the contextual factors that create a health enabling environment, that further theoretical, conceptual and empirical research is needed to understand adherence intentions and behaviour. The findings of the study underscore Marks’ (1996) call for placing health psychology theories, of which the TPB is an example, into its sociopolitical and community context and it is likely that the utility of the theory will have limitations if applied in a de-contextualized manner. The data of this study, along with the substantial body of literature critical of the TPB (e.g. Marks, 2008; 1996) call into question the presumed utility of the TPB in resource-constrained contexts.
We also found a non-significant relationship between intentions and self-reported adherence. These findings are similar to those of Fincham et al (2008) and Kagee and van der Merwe (2006). Among possible explanations for this apparent lack of association are recall bias and social desirability bias. Recall bias occurs when a respondent’s answer is influenced by his or her memory of the event rather than actual observation of the event. Social desirability bias occurs when respondents tailor their answers to questions in a manner that is in keeping with what they think the questioner wishes the answer to be. In the present sample, the self-reported adherence questionnaire had an internal consistency of 0.59 which did not meet the criterion of 0.70, recommended by Field (2000). After the necessary items were deleted from the questionnaire, the internal consistency could still not be improved. It is possible that a construct such as self-reported adherence may be vulnerable to the biases mentioned previously and that a construct as complex as adherence may need to be assessed in a more comprehensive manner, such as electronic monitoring of pill-container activity (e.g. Lyimo et al., 2011).
In addition to these biases may be the possibility that intentions to adhere simply do not correlate with actual self-reported behaviour, signaling the possibility that the model may not be applicable beyond the point of intentions in the context of ART adherence. Additional research is necessary to evaluate this possibility by, for example, conducting assessment of actual adherence in real time through the use of electronic diaries, cellular technology, or daily assessment so as to minimize the likelihood of recall bias.
The study was part of a longer cross-sectional investigation into the correlates of poor adherence to antiretroviral treatment. The study questionnaire took on average almost one hour to complete, introducing the possibility of respondent fatigue. As a consequence, participants may have concentrated less well on questions occurring at the tail end of the questionnaire. Also, as is the case in many public health clinics in sub-Saharan Africa, the sample consisted mainly of women and may have limited generalizability to male ART users.
Adherence is assumed by some theorists to result from rational social-cognitive decision-making processes (Ogden, 2000). Rational decision-making models, of which the TPB is an example, often do not take into consideration the role of personality, unconscious processes, mental status and socio-political context. These factors could potentially play an important role in furthering understanding of adherence. To this extent the role of these factors in explaining ART non-adherence awaits further investigation.