Correspondence: Dr Maria C. Rodriguez-Barradas, Infectious Diseases Section (MS 111G), Michael E. DeBakey VAMC, 2002 Holcombe Blvd, Houston, TX 77030, USA. Tel: 713 794 8856; fax: 713 794 7045; e-mail: email@example.com
Despite the effectiveness of highly active antiretroviral therapy (HAART), HIV remains a major cause of mortality in the USA, largely as a result of poor HIV treatment adherence. In this study we assessed the association between five patient-centred factors and adherence to HIV treatment.
We surveyed 244 adults at two HIV clinics in Houston, Texas between October 2009 and April 2010. Participants were given a questionnaire and their charts were reviewed for clinical data. Survey items assessed the following factors: self-assessed HIV knowledge, awareness of disease biomarkers, intention to adhere to HIV treatment, health literacy and decision-making style. The primary outcome measure was HAART adherence during the previous month. Logistic regressions were performed to calculate the effect of each factor on adherence.
All participants had HIV/AIDS and were on HAART at enrolment. Eight per cent of participants were female, 57% were African-American and 16% were Hispanic. Mean age was 58.1 years. Sixty-eight per cent were adherent to HAART during the last month. On univariate analysis, a preference for wanting choices, correct knowledge of recent HIV viral load level, and intention to adhere to HIV treatment were significantly associated with adherence. On multivariate analysis, only intention to adhere to HIV treatment remained statistically significant after adjusting for other factors (odds ratio 2.2; 95% confidence interval 1.1 to 4.3).
Intention to adhere to HIV treatment was significantly associated with self-reported adherence to HAART. Interventions that bolster patients' intentions to adhere to HIV treatment during clinical encounters may improve adherence to HAART and HIV control.
Advances in treatment for HIV infection have resulted in significant improvements in HIV control. The introduction of highly active antiretroviral therapy (HAART) in 1996 marked a transformation in patient outcomes . Nonetheless, HIV remains a leading cause of mortality in the USA for people 25 to 54 years old . Studies show that treatment adherence is one of the most important factors in preventing virological failure and emergence of resistant virus [3-5]. Improving patient adherence to HAART regimens allows the full benefits of HAART therapy to be realized, including maximal and lasting suppression of viral replication, immune reconstitution, and improved survival [5-7]. Past studies evaluating adherence to protease inhibitor therapy concluded that adherence of 95% or greater was required to achieve viral load control, while current guidelines recommend “excellent adherence.” [6-8] The literature on HIV adherence indicates that over 10% of patients miss one or more doses of medications per day; over 33% have missed doses within the past 4 weeks . In a 2006 meta-analysis, Mills et al. found that adherence rates were better in Africa than in North America . HAART adherence rates are comparable to those in other chronic illnesses; however, successful HIV treatment outcomes require higher adherence rates .
Many factors affect HAART adherence, including psychosocial problems, physician−patient communication, and HAART regimen characteristics [5, 10-15]. Studies of patient-centred factors have explored the role of health literacy and decision-making styles [11, 16-21]. In one study, HIV-related knowledge was found to be predictive of adherence . In the present study, we concurrently evaluated several patient-centred factors with potential relevance to HAART adherence, including behavioural intention, medical decision-making preferences, health literacy, awareness of HIV biomarkers, and self-assessed knowledge about HIV (Fig. 1) .
Participants were recruited from HIV clinics at the Michael E. DeBakey VA Medical Center (MEDVAMC) and the Thomas Street Health Center (TSHC) in Houston, Texas. The MEDVAMC provides out-patient care to approximately 800 HIV-positive veterans from southeastern Texas. The TSHC is operated by the Harris County Hospital District and provides comprehensive medical services to around 5000 adult residents of Harris County and surrounding counties with HIV/AIDS. The TSHC is among the largest HIV clinics in the USA.
Patients were recruited between October 2009 and April 2010 from clinic waiting rooms. Patients were approached by study staff after checking in for appointments. Verbal consent was obtained and patients were given the survey. There were no exclusion criteria, provided patients could consent. The survey was self-completed, with research personnel available to answer questions and assist participants with reading difficulties. Medical records of all participants were reviewed to obtain laboratory data. The study was approved by the Institutional Review Board for Baylor College of Medicine and Affiliated Hospitals, and by the Harris County Hospital District and MEDVAMC research committees.
Study participants were asked to complete a survey containing 48 questions about their demographic background, decision-making preferences, health literacy, intention to adhere to HIV treatment, HIV laboratory test names, results and target values, and self-assessed HIV knowledge. Participants were also asked about their adherence during the preceding week and month. Demographic questions assessed race, ethnicity, years since HIV diagnosis, years of clinic attendance, and education level. Four infectious disease physicians with HIV medicine expertise reviewed the survey for thoroughness and appropriateness.
Medical ethics commonly stresses the importance of patient autonomy . Yet research has found that, while most patients wish to know about the risks and benefits of treatment options and to be involved in choosing among them, they do not necessarily want to be responsible for generating options through medical problem-solving . We selected questions designed to reflect these two concepts of autonomy and responsibility. Participants were classified as accepting or not accepting responsibility for medical problem-solving, and wanting more or less opportunity to choose from different treatment alternatives.
Health literacy was assessed by asking, “How confident are you filling out medical forms by yourself?” [26, 27] Respondents answered either: “extremely,” “quite a bit,” “somewhat,” “a little bit,” or “not at all.” Past studies have established the predictive validity between this single-item screener and more comprehensive measures of health literacy [26, 27]. Answers were scored from 1 to 5 along a Likert scale.
Intention to adhere to HIV treatment
Thirteen questions appraised participants' intention to adhere to HIV treatment. We have previously reported on the development and factor analysis of this measure, the Intention to Adhere to HIV Treatment Scale, and its validation with the criterion standard of HIV control . Questions were adapted from surveys about other chronic conditions or created specifically for this study . Questions reflected the domains of self-efficacy, risk perception, and outcome expectancy. These factors are components of the Health Action Process Approach, a health-related model of behavioural intention based in social cognitive theory .
General HIV knowledge
Four general HIV knowledge questions were included in the form of self-assessed ratings: “I know a lot about living with HIV infection,” “I know a lot about how HIV is spread from one person to another,” “I know a lot about medications to treat HIV infection” and “I know a lot about the side effects of medications used to treat HIV infection.” Answers were scored from 1 to 6 along a Likert scale. We have validated the HIV knowledge scale previously .
Knowledge of HIV biomarkers
To assess participants' awareness of the biomarkers of their HIV infection, they were asked to respond, in free text format, to the question, “What are the names and results of two important blood tests related to HIV infection that you and your doctor follow for your HIV management?” The correct laboratory answers were “CD4 count” and “HIV viral load.” Responses such as or similar to “T4,” “T cells” or “helper cells” and “viral load,” “HIV load” or “virus in the blood” were all considered correct. Answers were scored as correct or incorrect.
Participants were then asked to record their most recent test results and identify target values for these laboratory tests. Participants' self-reported test results were compared with CD4 count and HIV viral load results recorded in their chart during the last 6 months. CD4 count was considered correct if the reported value was within the same range as either of the two most recent values recorded in the chart, using the following ranges: < 50, 50–200, 201–500 and > 500 cells/μL. For viral load, answers within 1 log HIV-1 RNA copies/mL of the chart value were considered correct. The correct answer for CD4 count target was any number above 200 or any phrase indicating a goal of increased value (e.g. “as high as possible”). The correct answer for HIV viral load target was any number less than 400 or any phrase consistent with “undetectable” such as “no virus.” Following the reporting standards for HIV viral load during the study period, an undetectable viral load was defined as < 400 copies/mL. Answers were scored as correct or incorrect. Two authors reviewed and scored the responses. Discrepancies were resolved by consensus.
HAART adherence during the past week and month was measured using previously validated questions [30, 31]. Adherence during the past month was our primary outcome measure and was evaluated by asking participants to report their adherence during the past month using a visual scale ranging from zero to 100% . This scale is correlated with unannounced pill counts, and inversely correlated with HIV viral load [30, 32]. Participants were considered “adherent” if they reported adherence of 90% or greater. Participants reporting adherence of < 90% were deemed nonadherent . While many studies use a 95% cut-point, recent literature has challenged that level in persons on ritonavir-boosted protease inhibitors and contemporary nonnucleoside reverse transcriptase inhibitors [33-35]. We wanted to maximize sensitivity at detecting clinically adequate adherence, as well as maximize power, and chose a cut-point of 90%. Adherence during the past week was evaluated by asking participants, “During the last 7 days, how many pills did you take?”  Answers included: “all of my pills every day,” “most of my pills” or “none.” Respondents who answered “all of my pills every day” were classified as adherent. All others were classified as nonadherent. This item was adapted from a questionnaire that was developed for use in community-based HIV/AIDS research sponsored by the National Institutes of Health and conducted by the Terry Beirn Community Programs for Clinical Research on AIDS .
Univariate analyses compared the adherent and nonadherent groups in terms of demographic variables (age, race, ethnicity, gender, years since HIV diagnosis, years in clinic and level of education). The adherent and nonadherent groups were also compared in terms of factors we hypothesized might influence adherence: decision-making style; health literacy; awareness of HIV disease markers, results and goals; intention to adhere to treatment and HIV knowledge. The significance of differences between the adherent and nonadherent groups was determined using the nonparametric Wilcoxon two-sample test for continuous variables and the χ2 test for categorical variables.
A series of hierarchical multivariate logistic regression analyses were then conducted. Characteristics that significantly differed on univariate analysis between the adherent and nonadherent groups were included in the multivariable analyses. Some factors that did not differ on univariate analysis but were important in our theoretic model were also included. Multivariate model 1 evaluated the relationship of adherence with self-assessed knowledge scores. Model 2 evaluated the relationship of adherence with scores on the Intention to Adhere to HIV Treatment Scale . Model 3 evaluated the relationship of adherence with correct reporting of recent HIV viral load results. Model 4 evaluated the relationship of adherence with decision-making styles. Model 5 included each of the variables from models 1–4 in one combined model. P-values of < 0.05 were considered statistically significant. Data were analysed using sas statistical software (SAS Institute, Cary, NC).
The survey was distributed to 311 patients at two HIV clinics. Fifteen incomplete surveys were excluded from analysis. An additional 43 were excluded because the respondents were not on HAART. Nine were excluded because the respondents had started HAART too recently to measure adherence during the past month. Results from the remaining 244 questionnaires were included in this analysis (Fig. 2).
Sixty-eight per cent of the patients were adherent during the last month. Table 1 shows the baseline characteristics of the study population with comparisons between adherent and nonadherent participants based on the primary outcome of adherence during the last month. The mean age of the participants was 52 years; 92% were male and more than 50% were Black. Sixty per cent had some college education and 70% had adequate functional health literacy. Sixty-two per cent reported having an HIV diagnosis for over 10 years. Seventy-nine per cent came from the VA site. The adherent and nonadherent groups did not differ significantly in terms of age, gender, race, education, health literacy, time since diagnosis, or site.
Table 1. Demographic and clinical characteristics of patients on antiretroviral treatment categorized by 1-month adherence >90% or ≤ 90%
All patients (n = 244)
Adherent (n = 165)
Nonadherent (n = 79)
Adherence was assessed by visual analogue scale, with values > 90% classified as adherent.
SD, standard deviation.
Age (years) [mean (SD)]
Gender [n (%)]
Race/ethnicity [n (%)]
Education [n (%)]
≤ High school
Some college or more
Health literate [n (%)]
Time since HIV diagnosis [n (%)]
≤ 5 years
> 10 years
Site [n (%)]
Thomas Street Health Center
Michael E. DeBakey Veterans Administration Medical Center
CD4 count (cells/μL) [mean (SD)]
HIV viral load (log10 copies/mL) [mean (SD)]
HIV viral load < 400 copies/mL [n (%)]
Differences in HIV control between adherent and nonadherent subjects
As expected, adherent participants had significantly higher CD4 count than nonadherent participants (mean 528 vs. 431 cells/μL, respectively; P < 0.01), had significantly lower viral load (mean 1.8 vs. 2.3 log10 copies/mL, respectively; P < 0.001) and were significantly more likely to have undetectable viral load (97% vs. 81%, respectively; P < 0.0001). Overall, 217 patients (92%) had undetectable viral load.
None of the demographic parameters was significantly associated with adherence in univariate analyses. Nor was any association found between adherence and health literacy (Table 1). No significant association was found between adherence and having a preference for accepting responsibility for medical problem-solving, self-assessed HIV knowledge, awareness of CD4 count test name, result or target, or awareness of the HIV viral load test name or target (Table 2). In contrast, a preference for making choices, correct knowledge of HIV viral load, and a high score on the Intention to Adhere to HIV Treatment Scale were significantly associated with adherence in univariate analyses (Table 2). Similar results were obtained whether the measure of adherence was during the past month or during the past week (data not shown); hence, all further analyses used HAART adherence during the last month as the primary outcome.
Table 2. Univariate analysis of patient-centred factors associated with highly active antiretroviral therapy (HAART) adherence during the last month
Adherent (n = 165)
Nonadherent (n = 79)
*An intention score of < 5.5 was the referent category.
Separate multivariate analyses were performed using each factor that was significantly associated with adherence on univariate analysis, including a preference for making choices, correct knowledge of HIV viral load, and high score on the Intention to Adhere to HIV Treatment Scale (Table 3), as well as factors that did not differ on univariate analysis but were important in our theoretic model (i.e. self-assessed knowledge scores and decision-making style). These analyses showed no association between adherence and self-assessed knowledge (model 1). Positive associations were found between adherence and a high intention score (model 2), correct reporting of HIV viral load (model 3), and a preference for having choices. No association was found between adherence and accepting responsibility (model 4). In model 5, which included all the factors of models 1–4, the only variable significantly associated with adherence was a high intention score. Respondents with a high intention score had 2.2 times higher odds of being adherent during the last month (95% confidence interval 1.1 to 4.3). The associations between wanting choices and correct knowledge of viral load results were no longer significant.
Table 3. Multivariable logistic regression analyses of patient-centred factors associated with highly active antiretroviral therapy (HAART) adherence during the last montha
Odds ratio (95% confidence interval)
aAll models were adjusted for clinical site, age, race, ethnicity, gender, years since HIV diagnosis, education and health literacy. The dependent variable in the logistic regression models was whether or not the respondent was adherent during the last month.
bThe moderate score of 4.5 to < 5.5 for self-assessed knowledge was the referent category.
cAn intention score of < 5.5 was the referent category.
dNot knowing viral load results within the last 6 months was the referent category.
eNot wanting choices was the referent category.
fNot accepting responsibility was the referent category.
This study assessed several patient-centred factors with potential relevance to HAART adherence. Previous studies have evaluated the role of psychosocial problems, psychiatric comorbidities, HAART regimen characteristics, and physician communication style [5, 7, 14, 36-38]. Patient-centred factors such as decision-making style, health literacy and behavioural intention have also been examined previously [11, 16-21]. Osborn et al. found an association between health literacy and self-reported HAART adherence . Kalichman et al. confirmed this with a pill-count study of adherence . Beach et al. found that patients with a preference for sharing decisions with their provider were more adherent to HAART than patients who preferred that their provider make decisions for them . Only a few studies have examined the association between HIV-related knowledge and HAART adherence [17, 22]. Weiss et al. compared participants' answers to eight questions about HIV with self-reported adherence levels, and found that participants with six to eight correct answers were significantly less likely to miss doses of HAART than participants with fewer correct answers . Nelsen et al. validated the Intention to Adhere to HIV Treatment Scale as a measure of behavioural intention associated with HIV control . A consensus has yet to emerge as to which of these factors is most important in determining adherence. Our study is the first to evaluate concurrently the role of decision-making style, health literacy, disease marker awareness, intention to adhere to HIV treatment and HIV knowledge. In particular, our study advances understanding about the importance of intention to adhere to HIV treatment. Our hypothesis that self-assessed HIV knowledge would predict adherence was not supported.
Only behavioural intention was predictive of HAART adherence in multivariate models that also contained self-assessed knowledge, knowledge of viral load, and decision-making style. Adherence, as expected, was highly associated with controlled viral load. These results suggest that intention to adhere to HIV treatment is an important factor in achieving virological control. Behavioural intention is well described in the social science literature, where it has been found to predict other health-seeking behaviours [29, 39-41]. Behaviours as diverse as smoking cessation, diet, cancer screening and sunscreen use have been examined with such models [42-50]. Intention and treatment adherence have been associated in other chronic conditions, such as hypertension . Intention may therefore serve as the best composite measure of internal motivation, and may be a domain for which interventions could be designed. In addition, studies have shown that treatment adherence can wane over time . By targeting intention, adherence may be better maintained. Surveying patients for their intention to adhere to HIV treatment could provide initial and ongoing assessments of patients' readiness to comply with antiretroviral treatment.
We assessed HIV-related knowledge using two methods—by asking participants to rate their own level of HIV knowledge, and by objectively assessing their awareness of the names and target values of HIV laboratory results and asking them to report their most recent results for these tests. Our hypothesis that both self-assessed and objectively assessed HIV knowledge would be associated with adherence was not confirmed. Self-assessed HIV knowledge showed no association with adherence. Among the items evaluating participants' knowledge about specific markers, only correct knowledge of recent HIV viral load was associated with adherence in the univariate analysis; this association was lost in the most inclusive multivariable model that included intention. Previous work has demonstrated the importance of providing patients with specific information about chronic disease control and contrasting patients' current status with clinically important (goal) levels . The model of expectations vs. fantasies described by Oettingen et al. suggests that comparing one's current status to an explicit goal is an important determinant of behavioural intention . Although most of our patients correctly identified their treatment goals (high CD4 count and low viral load), most did not know their own values.
Limitations of our study include the fact that adherence was measured using self-report, rather than objective measures such as pill count. The free text format used to evaluate knowledge of laboratory markers may have impaired some patients' ability to demonstrate their knowledge; a multiple-choice survey might have yielded different results. Scores on our self-assessed knowledge survey tended to be high, indicating either that our study population was relatively knowledgeable about HIV or that participants tended to overestimate their knowledge. It is possible that, if measured objectively, participants' HIV-related knowledge scores may be associated with adherence. Similarly, our respondents tended to indicate a high level of intention to adhere to HIV treatment. The lack of a sizeable group of participants with low levels of self-assessed knowledge, intention or nonadherence impacts our ability to generalize these finding to other groups. A large percentage of our survey respondents came from the VA, which may reflect a selection bias or response bias. Additionally, our study was cross-sectional, we did not follow participants over time, and the majority were enrolled in the HIV continuity clinics for > 5 years.
In conclusion, this study evaluated several patient-centred factors related to adherence. The only factor that was found to be significantly associated with adherence was intention to adhere to treatment. Evaluation of intention may be a useful tool to inform providers of the likelihood that patients are and will remain adherent to antiretroviral treatment.
Funding: This work was supported by the use of resources and facilities at the Houston VA Health Services Research and Development Center of Excellence (HFP90-020) at the Michael E. DeBakey VA Medical Center, and the resources and facilities of the Harris County Hospital District. ADN is also supported by a Doris Duke Charitable Foundation Clinical Scientist Development Award. BWT was supported by a VA Career Development Award from RR&D (B4623).